1. Introduction

Fiber-optic refractometers have been receiving extensive interest in the context of label-free chemical and biological sensors because they are immune to electromagnetic interference, compact in size, and easy for signal delivery. There are a number of ways to implement refractive index (RI) sensing, such as etched fiber Bragg gratings (FBGs) [1], long period gratings (LPGs) [2], a single mode-multimode-single mode fiber structure [3], micro-bent fiber [4], open-cavity Fabry-Perot (FP) interferometers (FPI) [5], surface plasmon resonance [6] and multi-D-shaped fiber [7]. Recently, microstructured fiber (MSF) or photonic crystal fiber (PCF) based microfluidic refractometers have been developed due to their high sensitivity and small sample volume (down to nL). The micrometer-sized air holes of PCFs are inherent microfluidic channels allowing gas/liquid to pass through. Accordingly, the MSF- or PCF-based microfluidic refractometers have more flexibility in sensor design, e.g. absorption-based evanescent-wave sensors [8–10], FBGs and LPGs in different types of PCFs [11, 12], and open-cavity FPI [13]. Generally, to accomplish RI sensing with such PCF-based microfluidic refractometers, the capillarity effect is utilized to fill the liquids into the micro channels of PCFs. However, due to the small size of air holes and relatively long length of PCFs, the liquids filling process is time consuming. The typical measuring time for aqueous sample is in the range of 6 ~16 minutes [8, 9]. Although applying pressure pump can improve the liquid filling speed [10, 11, 13], the actual measuring time is still no less than 1 minute [10]. Therefore, there is a clear need for a PCF-based microfluidic refractometer technique that is able to provide a fast and convenient measurement.

In this work, we propose and demonstrate a novel open-cavity FPI microfluidic RI sensor with short response time based on a concave-core photonic crystal fiber (CPCF). Compared with the fiber-optic open-cavity FPI RI sensors for the-state-of-the-art, the proposed sensor provides outstanding advantages in sensor fabrication and performances. Generally, the open-cavity fiber-optic FPIs can be developed by varied methods, such as by femtosecond (fs) laser micromachining a micro-notch in the fiber core [14], lateral offset-splicing fibers [15], splicing a piece of homemade C-shaped fiber between two single mode fibers (SMFs) [16], and splicing a section of fiber tube (FT) between a section of SMF and a section of PCF to make typical SMF-FT-PCF sandwich structure [13, 17]. However, the sensors made by fs lasing micromachining and offset-splicing fibers suffer weak mechanical strength. This disadvantage brings no negligible limitations on the stability, convenience and efficiency of the measurement. The C-shaped fiber is not a commercial product, the fabrication process of which is complex and costly. As for the SMF-FT-PCF sandwich structure sensor, in order to make the liquid easily flow into or out of the sensor, the PCF needs to be cleaved as short as possible, usually tens micrometers. Otherwise, auxiliary pressure equipment is required to facilitate the liquids get into and out of the sensor [13]. However, after the shortening process of the PCF, a reflection beam is generated by the last end face of the PCF, which makes the interference pattern complex and increases the difficulty in signal processing. Although this reflection can be removed by roughening the last end face of PCF by wet chemical etching [18], fs laser micromachining [19], and terminating the PCF with an angle [20], such techniques increase the complication and cost of the fabrication and may destroy the SMF-FT-PCF sandwich structure. Meanwhile, it is well understand that a shorter FP cavity length can make a larger FSR, leading to a broader measurement range [16]. In the SMF-FT-PCF sensor, the length of FP cavity is determined by the FT length. However, in the fabrication process of SMF-FT-PCF sensor, multiple operations of cleaving and splicing are required to the FT. Due to the electron arc discharge, slight deformation and residual stress will be generated in the range around the splicing point of FT, after the first splicing operation. It brings no negligible difficulties in subsequent operations of cleaving and splicing if such operations locate very close to the original splicing point of FT, which makes difficulty to obtain FPI with very short FP cavity. Few works have been reported for the SMF-FT-PCF FPI with the cavity length less than 10 um [13, 17].

The problems of the aforementioned different types of open-cavity FPIs can be overcome by the proposed open-cavity FPI based on the CPCF. The CPCF is made by cleaving a high numerical aperture solid-core PCF with an axial tension, which can provide a very shallow concave core. The proposed sensor is fabricated by splicing the CPCF with a piece of SMF. The cladding diameter of the CPCF is 125 μm, which makes it compatible to splice with the conventional SMF, leading to a robust structure. The concave-core part of the CPCF naturally forms an FP cavity with a very short cavity length, can be less than 10 μm, which is shorter than that for the open-cavity FPI RI sensors reported before [5, 13, 17], leading to a broadened measurement range. Meanwhile, multiple large air holes in the CPCF cladding naturally function microfluidic channels that can be used to efficiently deliver the liquid sample into and out of the sensor. To shorten the channel length and eliminate the harmful reflection from the last end face of the CPCF, the CPCF is cleaved with an angled tension by a cleaver. With the less volume of the cavity, the shorter length, and the larger sectional area of the channels, the proposed sensor provides an easy-in and easy-out structure for liquids, leading to a great improvement of the measurement convenience and efficiency. The experimental results show that there is no need of pressure equipment for the liquid RI measurement by the proposed sensor, and the measuring times of proposed sensor for distilled water and ethanol is less than 359 and 23 ms, respectively, which clearly demonstrates that proposed sensor can provide on-line RI measurement. Moreover, the fabrication of the proposed sensor is very simple and low cost, only common fiber-optic splicer and fiber cleaver are required. The sensor reported here also provides other promising performances, including small size, high RI sensitivity, linear RI response, low temperature cross-sensitivity, and high repeatability. Such advantages make the proposed sensor attractive to the microfluidic RI measurement.

2. Fabrication and working principle

The schematic of the proposed FPI microfluidic refractometers is shown in Fig. 1(a). It consists of a section of SMF as lead-in fiber and a section of CPCF as reflection fiber. The end face of SMF and the inner face of the concave-core of CPCF function as the partial mirrors that form the FP cavity, while the concave-core part of CPCF functions as the sample chamber. The air holes in the CPCF cladding are used as open channels to let the liquid sample flow in and out the FP cavity freely. The CPCF is made by cleaving a section of high numerical aperture PCF (NKT MM-37-01) with an axial tensile force. As shown in Fig. 1(b), the multimode PCF has a solid core with a diameter of 38 μm, which is surrounded by 20 petal shaped air holes with an azimuthal diameter of 12 μm and a radial length of 24 μm. These air holes are close arranged, resulting in that the fiber core and the cladding are connected by 20 thin bridges of 3 μm thickness. The outer diameter of the porous zone in the PCF cladding is about 86 μm. The cladding diameter of the PCF is 125 μm, which makes it compatible to splice with the conventional SMF, leading to a robust structure of proposed sensor.

 

Fig. 1 (a) Schematic of proposed sensor structure; (b) Microscope picture of the cross section of the CPCF.

Download Full Size | PDF

The specific preparation of CPCF and fabrication of the proposed sensor are as follows: Firstly, a section of PCF with axial tension was cleaved by a fiber cleaver (Fujikura CT-30) to make a CPCF. As shown in Fig. 2(a), the left end of the PCF mounted on the cleaver was fixed on a fiber holder, while the right end was pulled axially by a weight through a pulley [not shown in Fig. 2(a)]. The weight’s gravitational force provides an axial force for the PCF. Due to the 20 relatively large air holes, the applied axial tension causes non-uniform axial strain between the fiber core and cladding, which further leads to a radial uneven fracture between the core and the cladding when the PCF was cleaved. Thus, after cleaving there was a raised section of the core on the left PCF producing a CPCF on the right, as shown in Fig. 2(b). Secondly, the CPCF was spliced to a section of SMF (Corning SMF-28) by the fusion splicer (Fujikura FSM-80s) with optimized discharge time and power. In order to avoid collapsing of the concave-core of CPCF, the CPCF was set 20 μm off the standard splicing point as shown in Fig. 2(c). The corresponding fusion result is shown in Fig. 2(d). It is clearly shown that a very short FP cavity was formed by the concave-core CPCF near the splicing point. The end face of the lead-in SMF and the inner face of the concave-core of the CPCF were the two reflective mirrors of the FPI, respectively. Lastly, in order to make liquid sample easy in and out of the FP cavity, the reflection fiber, CPCF, was shortened using a cleaver facilitated by an optical microscope and a linear translational stage. In addition, the harmful reflection from the outside end face of reflection fiber was also eliminated in this step. Specifically, the lead-in SMF was fixed on a fiber holder, while a tilted tensile force was applied on the CPCF by a slant pulling with a translation stage before cleaving as shown in Fig. 2(d). It is worth noting that, due to this tilted tensile force, a concave-core with angled end face was generated at the outside end of CPCF after that cleaving process [see Fig. 2(e)], which eliminates the reflection from the outside end face of CPCF and makes a clear interference spectrum as shown in Fig. 2(f). By this method the CPCF with very shallow concave-core, around 10 μm, was obtained. The typical microscope photograph of the proposed sensor is shown in Fig. 2(e). It can be seen that compared with the typical SMF-FT-PCF based FPI, the proposed sensor head is very compact, in which the CPCF is used both as the reflection fiber and the FP cavity; the total length of reflection fiber is 33.4 μm, which is also the whole sensor head length, while the FP cavity length is only 10.3 μm, which is completely determined by the depth of the concave-core of CPCF. With an 84 μm outer diameter porous zone of the CPCF, the volume of the cavity, which can be used to approximate the minimum amount of liquid sample required by the sensor, is calculated to be 0.06 μL. This is about 5 times less than reported in [13]. Therefore, with the less volume of the cavity, the shorter length and the larger sectional area of the channels, the sensor has an easy-in and easy-out structure for liquids, leading to great decrement of the measuring time.

 

Fig. 2 (a)~(d) Schematic diagrams and microscope photographs of fabrication processes of CPCF and sensor head; (e) Micro photograph of proposed sensor; (f) Reflection spectrum of a fabricated sensor in air.

Download Full Size | PDF

As shown in Fig. 2 (f), the free spectral range (FSR) of CPCF based FPI is 111.9 nm. The fringe visibility of the reflection spectrum is more than 15 dB. Due to the small reflectivity of the two FP reflection mirrors ($R\approx$3.5% and 0.2% when cavity is filled with air and water, respectively). Thus the low fineness FPI can be modeled using two-beam optical interference equation [15, 16]:

According to Eq. (1), the interference reaches its minimum when the phase of the cosine term becomes an odd number of $\pi$. The two adjacent interference minimums have a phase difference of $2\pi$. Therefore, the relationship between wavelength spacing of two adjacent minimums, which is known as FSR can be derived as [15]

It is noted that in the preparation of the CPCF, the structure of the concave-core is relative to the axial tension applied on the fiber. In order to investigate the influences of the axial tension on the depth of the concave core, several sensors with different applied axial tensions have been fabricated. As describe above, the axial tension was applied by pulling the PCF axially with a weight through a pulley. The values of the axial tension can be adjusted by changing different weight. The microscope photographs and the reflection spectra of the sensors fabricated with different axial tensions are shown in Fig. 3. It can be seen that when the weights used were 8, 10, 20 and 30 g, respectively (the applied axial forces were 0.0784, 0.098, 0.196 and 0.294 N, respectively), the FSRs of the resulted CPCF based FPIs were 135. 8, 111.9, 61.2, and 41.8 nm, respectively. By using the Eq. (2) again, the corresponding lengths of the FPIs, which are determined by the depths of the concave-core of CPCFs, are estimated as 7.6, 10.3, 20, and 28.3 μm, respectively. The experimental results suggest clearly that a weaker axial tension applied makes a shallower depth of the concave-core. Therefore, the depth of the concave core can be controlled by tuning the axial tension with different weight in the fabrication.

 

Fig. 3 (a)~(d) Microscope photographs of fabricated sensor head with different weights as 8, 10, 20 and 30g, respectively; (e)~(h) Reflection spectra for fabricated sensor1~sensor4 in air, respectively.

Download Full Size | PDF

3. Experiments and discussion

The response of the proposed sensor to the change of the RI in liquid was characterized with the experimental setup illustrated in Fig. 4. It consists of a broadband source (BBS, FiberLake-BBS) with output wavelength ranging from 1250 to 1650 nm, an optical circulator (OC), and an optical spectrum analyzer (OSA, YOKOGAWA AQ6370C) with a resolution of 0.02 nm. All the experiments were conducted by the proposed sensor with 10.3 um cavity length. The lead-in SMF of the proposed sensor was connected to port 2 of the OC. The BBS was connected to port 1 of OC to supply the light to the sensor. The reflection spectrum was taken via port 3 of OC and measured by OSA. The liquid sample used was the mixture of ethanol and distilled water. Its RI was controlled by stepwise adding 40 uL pure ethanol into 2 mL distilled water for 15 times. Thus the weight concentration of ethanol solution was increased from 0 to 19.11% by a step of approximate 1.5%. The RI values of the liquid sample were measured by a handheld Abbe refractometer (Reichert AR200) with measurement accuracy of ${10}^{-4}$RIU.

 

Fig. 4 The experimental setup for liquid sample RI measurement with the proposed sensor.

Download Full Size | PDF

The sensor reflection spectrum in air was first recorded by OSA. Then the sensor head was immersed into the liquid sample. Due to the easy-in and easy-out structure of the sensor, the liquid sample passed through the microfluidic channels of the sensor and filled the FP cavity, leading to a change of the reflection spectrum. After spectrum recording, the sensor head was lifted from the liquid sample. Through the microfluidic channels again, the liquid sample flowed out of the FP cavity and evaporated in air. We found that the reflection spectrum can always recover to that in air, suggesting a complete removal of the sample in the cavity so that the next liquid RI measurement would not be affected. Therefore, the proposed sensor can achieve a reliable and repeatable liquid RI measurement. The spectrum evolutions of the sensor with respect to the RI change of liquid sample are shown in Fig. 5. It is shown that the spectra red shift linearly with the increase of the RI. For example, the wavelength position of the fringe valley at 1495.34 nm in the distilled water shifted to 1517.03 nm when the ethanol-water solution concentration increased from 0 to 19.11%. Thus the sensitivity, defined as a change of the measured variable due to a unity change in the samples, is 1635.62 nm/RIU at 1500 nm as shown in Fig. 5(b). Note that the FSR of the proposed sensor in the distilled water is 86.45 nm, for an interferometric sensor, the measurement range is usually defined as the FSR of the interference spectrum divided by its sensitivity. Thus, the measurement range of proposed sensor is estimated as 0.0529 RIU, which is larger than that for the open-cavity FPI RI sensors reported before [5, 13, 17]. Although, there is a tradeoff between the measurement range and the detection limit for the FPI refractometer [16], the proposed FPI with very short and controllable cavity length provides a promising optional method for practical applications. It also shows that the response curve of the spectral valley has good linearity with $R^2$ better than 0.99. In addition, the peak power of the spectra degrades during the increasing of the liquid RI due to the reduction of the Fresnel reflectivity of the mirrors.

 

Fig. 5 Reflection spectra of sensor with different concentration of ethanol-water solution; (b) The relationship between RI change and valley wavelength shift. DW: distilled water

Download Full Size | PDF

In the experiments, we found that the spectrum changed very quickly when the sensor being immersing in or lifting from the liquid sample. To investigate the sensor response speed, the experimental setup shown in Fig. 6(a) is utilized, which is similar with Fig. 4, except that a narrow linewidth tunable laser (Alnair Labs TLG-200) with a linewidth less than 100 kHz replaces the BBS and a fast speed photoelectric detector (PD, THORLABS PDB430C) followed by an oscilloscope (Tektronix MDO3104) replacing the OSA. When the laser wavelength was set at a fixed value, for example 1550 nm, the sensor reflectivity at this wavelength was varied with different sample due to changed interference spectrum, as shown in Fig. 6(b). Then the corresponding reflective intensity change can be detected by the PD and displayed on the oscilloscope. Thus, the respond speed of the sensor can be measured by timing the intensity change of signal. In the experiments, the laser wavelength was set at 1552 nm. The response times for both the pure ethanol and distilled water were investigated.

 

Fig. 6 (a) The experimental setup for testing the measuring speed of proposed sensor, (b) The sensor reflection spectra for air, distilled water and ethanol, respectively. DW: distilled water

Download Full Size | PDF

Benefiting from the easy-in and easy-out structure of the sensor, the liquids filling in and flowing out speeds are very fast. The results are shown in Fig. 7(a) to (d). It is seen that at the moment of inserting the sensor head from air into pure ethanol, the initial intensity kept at 3.20 V until the time −842.27 ms, as shown in Fig. 7(a). Then as the ethanol filled in the FP cavity, the signal intensity quickly decreased to 0.1 V until the time −820.16 ms, which is because the sensor reflectivity for ethanol is much lower than that in air at 1552 nm. After −820.16 ms, the signal intensity maintained at 0.1V, indicating that the sensor cavity had been completely filled by ethanol. Therefore, the sensor actual response time, which is mainly dependent upon the ethanol filling time, can be presented by the transition time in Fig. 7(a), which is 22.11 ms. The sensor recovery time from ethanol to air was also investigated. As shown in Fig. 7(b), when the sensor was quickly lifted from ethanol to air, the signal intensity kept at 0.1V until 961.92 ms. Then as the ethanol began to flow out of FP cavity and evaporate, the signal intensity jumped to and stabilized at 3.20 V after 1417.96 ms, corresponding to a recovery time of 456.04 ms. Similarly, we tested the sensor measuring time and recovery time for distilled water. It is noted that due to higher viscosity, the distilled water took relatively longer time for filling in and flowing out. Even so the measuring time and recovery time are only 358.33 and 1269.05 ms, respectively, as shown in Fig. 7(c) and (d). It should be pointed that in our experimental system, the liquids fast filling in and flowing out speeds are dependent upon the sensor inherent easy-in and easy-out structure. Unlike the other typical PCF based microfluidic RI sensor [8, 10, 13], there is no pressure system required to facilitate the liquids to get in and out of sensor, which reduces the complexity and cost of the system. Moreover, the proposed sensor measuring speed is much faster than that in reported works, e.g. the distilled water measuring time of proposed sensor is 358.33 ms, while the water or aqueous solution measuring time of is 6 min in [9], and with pressure pump is 4 min in [8] and 1 min in [10], respectively. Accordingly, the proposed sensor measuring speed is 1004 times faster than that in [9], 700 times than that in [8], and 167 times than that in [10]. Obviously, such very fast measuring speed and the convenient measurement method make the proposed sensor very useful and competitive for an on-line microfluidic RI measurement. The RI sensitivity of the proposed sensor is 1635.62 nm/RIU at 1500 nm, which is comparable to that reported in open-cavity FPI RI sensor [13, 15]. Although its sensitivity is lower than absorption-based evanescent-wave sensor [9, 10], the shorter response time of proposed sensor provides a promising optional method for practical applications.

 

Fig. 7 (a) and (b) The response time of the sensor for being immersed into and lifted out of ethanol, respectively; (c) and (d) The response time of the sensor for being immersed into and lifted out of distilled water, respectively.

Download Full Size | PDF

Another advantage of the proposed sensor is the low temperature cross-sensitivity due to the small thermal-expansion coefficient of the silica material of the fiber ($\text{5}\text{.5}\times {\text{10}}^{\text{-7}}\text{/}°C$). Before measurement of the temperature response, the sensor was firstly annealed in a furnace by gradually increasing the temperature from room temperature to ~200 °C for two times. Then we recorded the reflection spectra though the OSA when the temperature decreased from 200 to 25 °C in the air as is shown in Fig. 8(a). It is found that the reflection patterns nearly keep the same with the variation of temperature, which vividly indicates the good temperature stability of our sensor. In order to study the temperature cross-sensitivity, the accurate wavelength values of the fringe valleys were obtained using the two-order polynomial fitting method and the results are shown in Fig. 8(b). The black scatters are the values of the valleys at different temperature, while the red line is the linear fitting result. It is seen that the measured temperature sensitivity (the slope) is 0.29 pm/°C and the corresponding temperature cross sensitivity is as low as $\text{1}\text{.77}\times {\text{10}}^{\text{-7}}$ RIU/°C. For a temperature variation of 100 °C, it only induces an error of $\text{1}\text{.77}\times {\text{10}}^{\text{-5}}$RIU for RI measurement, which means that, therefore, temperature compensation is not required for most applications.

 

Fig. 8 (a) Spectra of the sensor according to temperature variations. (b) Wavelength dependence with temperature.

Download Full Size | PDF

As discussed earlier, the proposed RI sensor with its inherent easy-in and easy-out structure for liquids can provide convenient, repeatable, and reliable RI measurement. Meanwhile, by this structure, the inner cavity of the proposed sensor is easy to be rinsed clean with pure ethanol to eliminate the influences of the dust and the impurity in the liquids on the sensor performance. To further verify the repeatability of the sensor, we repeated the RI measurement of distilled water several times in a prolonged time of more than 3 days. The same procedure described above was followed and the same distilled water sample was used for each measurement. The measured spectra are shown in Fig. 9(a). By using the two-order polynomial fitting method again, the accurate wavelength shift values of one fringe valley at 1583.32 nm were obtained as shown in Fig. 9(b). The peak-to-peak variation of wavelength shift is 0.04 nm. Using the sensitivity of 1635.62 nm/RIU of the proposed sensor, the variation of the equivalent RI is only$\text{2}\text{.5}\times {\text{10}}^{\text{-5}}$. Noting that water has a thermo-optic coefficient (dn/dT) of $\text{-8}\times {\text{10}}^{\text{-5}}/°\text{C}$ [21], a significant contribution to the observed RI variations may be from the temperature changes of the ambient environment. The experimental results vividly indicate a relative good repeatability of the proposed sensor for a long time.

 

Fig. 9 Sensor repeatability test. (a) Reflection spectra of the sensor for several measurements of distilled water; (b) Wavelength shifts for several measurements of distilled water.

Download Full Size | PDF

4. Conclusion

In summary, we have proposed and demonstrated a novel FPI microfluidic RI sensor based on concave-core photonic crystal fiber (CPCF) with fast response speed. The proposed sensor is fabricated by splicing CPCF to a section of well-cleaved single mode fiber. The concave-core part of CPCF naturally forms the FP cavity, while the air holes in the cladding of CPCF are used as the open channels for the sample flows in and out of the sensor. With the small volume of the cavity, the short length and the large sectional area of the channels, the open cavity FPI provides an inherent easy-in and easy-out structure for liquids measuring. Thus the response time is greatly decreased. The measuring times of the proposed sensor for distilled water and pure ethanol are less than 359 and 23 ms, respectively. The RI sensitivity of the sensor is characterized to be 1635.62 nm/RIU at the wavelength 1500 nm. We also have experimentally demonstrated that the sensor has additional advantages of good linearity and low temperature cross-sensitivity. These excellent performances render the proposed sensor a promising candidate for many applications including chemical and biological sensing.