Optofluidic refractive index sensor based on partial reflection

We demonstrate a novel optofluidic refractive index (RI) sensor with high sensitivity and wide dynamic range based on partial reflection. Benefited from the divergent incident light and the output fibers with different tilting angles, we have achieved highly sensitive RI sensing in a wide range from 1.33 to 1.37. To investigate the effectiveness of this sensor, we perform a measurement of RI with a resolution of ca. 5.0×10–5 refractive index unit (RIU) for ethylene glycol solutions. Also, we have measured a series of liquid solutions by using different output fibers, achieving a resolution of ca. 0.52 mg/mL for cane surge. The optofluidic RI sensor takes advantage of the high sensitivity, wide dynamic range, small footprint, and low sample consumption, as well as the efficient fluidic sample delivery, making it useful for applications in the food industry.

To date, a variety of optofluidic RI sensors have been reported by using integrated interferometers [27,28], Fabry-Pérot cavities [29,30], microring resonators [31,32] gratings [33,34], microstructured optical fibers [35,36], and surface plasmon resonance [37,38]. Although the sensitivity of the aforementioned sensors can be as high as 10 6 RIU by measuring the wavelength shift, these sensors require expensive instruments, e.g., optical spectrum analyzers. Alternatively, one can design a highly sensitive RI sensor based on partial reflection or refraction by measuring the intensity of the reflected or transmitted light or both [39,40]. It is well known that the reflectivity of liquid-solid interface can be approximated by combining Snell's law with the Fresnel equations of reflection. When the RI of the liquid increases to a value where total internal reflection (TIR) at the liquid-solid interface is no longer satisfied, the reflectivity will decrease sharply, leading to a strong increase of the transmitted light. This phenomenon provides an attractive feature for assembling tunable optofluidic device and RI sensors. For example, Lapsley and co-authors [39] reported a variable optical attenuator where the light attenuation was achieved by adjusting the RI of the liquid in the microfluidic channel and thus altering the reflectivity of the light at the sidewall of the microchannel. Recently, Weber and Vellekoop [40] reported an optofluidic sensor, which was consisted of one input fiber and two out fibers with fixed angles for guiding the incident light, reflected light, and refracted light, respectively. Also, integrated microlenses were used for collimating the divergent light. Note that an inherent disadvantage of this sensor is the narrow dynamic range, typically, 0.01 refractive index unit (RIU) for a fixed incident angle. Thus, one has to prepare many chips with different incident angles to determine samples within a wide range of RI.
In this work, we report an optofluidic RI sensor with one straight detection channel, two input fibers, and ten output fibers for RI sensing with high sensitivity and wide dynamic range. Compared with the existing counterparts that require microlenses for light collimation, our design takes advantage of the divergent incident light and the optimized position of the output fibers to realize high sensitive sensing in a dynamic range from 1.33 to 1.37. In this case, we can measure the RI of most aqueous samples with high sensitivity by choosing an output fiber with a specific tilting angle. Since microlenses are removed from our design, the fabrication of the sensor becomes much simpler and cost-effective. To investigate the effectiveness of the RI sensor, we perform a measurement of ethylene glycol (EG) solutions with a RI range from 1.33 to 1.37, achieving a resolution of ca. 5.0×10 -5 RIU. Furthermore, we present measurements of liquid concentrations with a resolution of ca. 0.52 mg/mL for cane surge. The optofluidic RI sensor shown here may provide a compact and versatile sensing platform for sensitive and fast detection of low-volume samples.

Reagents and instruments
All reagents and standards were of analytical grade and purchased from Sinopharm Chemical Reagent (Shanghai, China) unless otherwise stated. SU-8 photoresist and SU-8 developer were purchased from MicroChem Corp. (Newton, MA, USA). Polydimethylsiloxane (PDMS) (Sylgard 184) was purchased from Dow Corning (Midland, MI, USA). Ultrapure water (Siemens Labostar2-UV) was used throughout. Sample solutions were prepared before use. Optical micrographs were obtained using a charge coupled device (CCD) camera (DS-Fi1, Nikon, Japan) mounted on a microscope (Eclipse 90i, Nikon, Japan). Figure 1 shows an optical image of a PDMS microfluidic chip for multi-fiber RI sensing, in which it contains one detection channel, two fiber channels with a tilting angle of 71°, and ten output fibers channels with tilting angles range from 67° to 76°, respectively. The tilting angle of 71° was calculated by using Snell's law [40], and the tilting angles of the output fibers were defined according to the divergent incident light. The microfluidic chip was fabricated by using standard soft lithography process [41]. Briefly, uncured PDMS was poured onto a SU-8 master, followed by curing at 80 ℃ for 20 min. The cured PDMS slab was then peeled from the SU-8 master, and punched holes at the ends of the detection channel and the PDMS channels to introduce sample and uncured PDMS, respectively. The PDMS slab was bonded with a clean glass slide by using a plasma cleaner (PDC-32G-2, Harrick, USA). Typically, the detection channel was 2 cm in length, 125 μm in width, and 150 μm in depth with a rectangular cross section. In this case, optical fibers could be inserted into the fiber channels and well aligned by fiber channels. To remove the scattering light, the gap between the fiber and fiber channel was sealed by introducing uncured PDMS from the PDMS inlets, followed by curing at 80 ℃ for 20 min. Fig. 1 Optical image of a microfluidic chip for multi-fiber RI sensing, which contains one detection channel, two input fibers channels, four PDMS channels, and ten output fiber channels. The dot and dash lines indicate the divergent incident light and reflected light, respectively.

Procedures
In the experiment, 5 μL of sample solution was added into the sample inlet and was introduced into the detection channel by negative pressure generated by a syringe. After each measurement, the channel was flushed with ultrapure water. A 473 nm laser and a broadband light from a tungsten halide lamp (PHILIPS 7748XHP) were coupled into input fibers and served as incident light for visualization and RI measurements, respectively. The reflected light was collected by one of the ten output fibers and recorded by a spectrometer (Maya2000 Pro, Ocean optics, Dunedin, FL, USA).  As shown in Fig. 2(c), the vertical sidewalls of the detection channel serve as reflective interfaces. When the liquid sample is introduced into the detection channel, a RI mismatch between PDMS and the liquid sample in the detection channel will cause the incident optical beam (I) to be reflected and refracted at the PDMS-liquid interface, resulting in a reflected beam (R) and a transmitted beam (T). The angle of R(θ) is the same as the angle of I, allowing the reflected beam to be captured by the output optical fiber. The intensity of the reflected light is a function of RI of the liquid sample, and thus, we can determine the concentration of solutes by measuring the intensity of the reflected light. Note that we did not fabricate microlenses to collimate incident light, it offers a possibility to arrange several output fibers with different tilting angles to record the divergent reflected beam. In this case, each output fiber corresponds to an angle of reflection, providing an effective method to achieve a wide dynamic range. On the other hand, the insertion loss of the device is relatively greater than its counterparts with microlenses, and approximately 5% incident light could be collected by the output fiber for the microfluidic chip as shown in Fig. 2

Visualization of the reflected and refracted beam
To visualize the reflected and refracted beams, a 473-nm-wavelength laser was used as incident light and 0.01 mM fluorescein solutions with different RIs were used to visualize the refracted light. Figure 3 shows four typical images of an optofluidic RI sensor (==71°) with different fluids in the detection channel. When air (n=1) was in the detection channel, TIR occurred at the PDMS-air interface, and the divergent reflected light could be clearly seen in Fig. 3(a). When the tilting angle of the input fiber was 71°, the estimated angle of reflection was from 67° to 76°. When fluorescein solutions were injected into the detection channel, the fluorescence intensity increased obviously with an increase in RI from 1.33, 1.36, to 1.

[Figs. 3(b)-(d)]
, indicating an increase in the refracted light intensity and a decrease in the reflected one. When the fluorescein solution with a RI of 1.42 was in the detection channel, the sidewall of the detection channel could hardly be identified, and the incident beam transmitted through the detection channel without obvious refraction and reflection [ Fig. 3(d)].

Theoretical calculation
To achieve a fundamental understanding of the sensor's RI response, theoretical analysis was conducted using Fresnel equations. We calculated the reflection coefficients r s and r p for s-and ppolarizations by (1) and (2), respectively, with the total reflectivity calculated by (3) where n 1 =1.41 is the RI of PDMS, and n 2 is the RI of the liquid in the detection channel. Thus, the reflectivity is a function of n 2 . Figure 4(a) shows a theoretical analysis of the reflected light intensity of an RI sensor with different angles of reflection. For a fixed angle θ, we found that the intensity kept as unity for small n 2 because of the TIR at the PDMS-liquid interface. When n 2 increased to a value where TIR was no longer satisfied, a very sharp decrease in reflectivity was observed with very small change in n 2 . Based on theoretical prediction, the slope can be infinity right at the critical point, thus a very sensitive RI sensor could be realized. For example, when n 2 changed 0.001 RIU, the maximum change in the intensity could be 30%, however, the dynamic range of the sensor was ca. 0.01 RIU, which was relatively narrow for real application. As shown in Fig. 4(a), when θ increases from 69° to 74°, the critical RI of n 2 shifts from 1.33 to 1.37. Thus, we could achieve a wide dynamic range by choosing an output fiber with different θ.

Effect of d on the measurement of RI
To optimize d of the RI sensor, we fabricated a series of microfluidic chips with one input fiber and one output fiber for characterization. Owing to the large aperture and core diameter, a multi-mode fiber could collect light from a wide angle, and thus smearing out the sharp decrease of the light coming from a single collection angle as indicated by the theoretical calculations [ Fig. 4(a)]. In this work, we adopted SMFs with a core diameter around 10 m as input and output optical fibers, thus, the reflection light within a narrow range of angle could be collected by the output fiber. We found that d had a great impact on the performance of the sensor. Figures 4(b) and 4(c) show normalized intensity of the light collected by the output fiber with different tilting angles as a function of RI for d of 0.6 mm, and 1.5 mm, respectively. The results did not match with the theoretical analysis [ Fig. 4(a)], especially, the response curves for =69° and =71° can be hardly resolved in the RI range of 1.33-1.34. This mismatch could be attributed to the output fiber collecting the light with different angles of reflection. Increasing d is an effective method to narrow the range of the angle of reflection. When we increased d to 5 mm, the flare angle of the reflected light that could be collected by a SMF with a core diameter of 10 m decreased dramatically. To obtain the estimated flare angle of the reflected light, we considered the reflection area as a point, thus, the sides of the isosceles triangle were 5 mm and 10 m, respectively, resulting in a flare angle of approximately 0.1°. As shown in Fig. 4(d), all of the four response curves for the different tilting angles are well separated, and the slopes after the critical points become steeper, which matched well with the theoretical calculations [ Fig. 4(a)]. It is worth mentioning that we can arrange more output fiber on one side the detection to collect reflected light with different angles when d is 5 mm.   Figure 5 shows four typical curves recorded by the output fibers with a tilting angle of 69°, 71°, 72°, and 74°, respectively, which covers a RI range from 1.33 to 1.37. The result confirmed that the dynamic range could be extended by using different output fibers for sensing. When RI increased from 1.35 to 1.36, the normalized intensity for the output fiber with a tilting angle of 72° decreased from 103 to 23. Because the stability of the light source was measured with an outstanding reproducibility of 0.13% RSD, the RI resolution could be estimated as 5.0×10 -5 RIU based on 3 times the standard deviation of the light source. Note that the signals collected by the output fibers with a tilting angle less than 68° or greater than 74° were too weak for sensing owing to the Gaussian intensity distribution of the divergent incident beam.   To further evaluate the performance of the RI sensor, aqueous cane surge solutions with different concentrations (0.1 g/mL-0.3 g/mL) were prepared and analyzed. Because the RI of the cane surge solution is in a range from 1.34 to 1.37, we adopted three output fibers with tilting angle of 71°, 72°, and 74° for collecting the signals, respectively. Table 1 gives obtained values and the relative errors (RE) for the sensor, showing a wide dynamic range and high accuracy with an estimated resolution of 0.52 mg/mL.

Conclusions
In summary, we have developed a sensitive, robust, and inexpensive optofluidic RI sensor. We have realized highly sensitive RI sensing in a wide dynamic range of 1.33-1.37 with a RI resolution of 5.0×10 5 RIU by choosing an output fiber with different tilting angles for signal recording. Also, we have measured a series of liquid solutions by using different output fibers, achieving a resolution of ca. 0.52 mg/mL for cane surge. To further enhance the sensing performance, one might use a higher resolution lithography instrument to fabricate a SU-8 master with a smoother sidewall. We believe this optofluidic RI sensor can be used for real-time, low-cost, and multifunctional measurements in a wide range of applications.