Cascaded Photonic Crystal Fiber Interferometers for Refractive Index Sensing

Cascaded Photonic Crystal Fiber Interferometers for Refractive Index Sensing Volume 4, Number 4, August 2012 Jun Long Lim, Member, IEEE Dora Juan Juan Hu Perry Ping Shum, Senior Member, IEEE Yixin Wang, Member, IEEE DOI: 10.1109/JPHOT.2012.2205911 1943-0655/$31.00 2012 IEEE

Cascaded Photonic Crystal Fiber Interferometers for Refractive Index Sensing Jun Long Lim, 1;2 Member, IEEE, Dora Juan Juan Hu, 1 Perry Ping Shum, 2;3 Senior Member, IEEE, and Yixin Wang, 1 Member, IEEE 1 RF and Optical Department, Institute for Infocomm Research, Agency for Science, Technology, and Research (A STAR), Singapore 138632 2 OPTIMUS, Photonics Centre of Excellence, School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798 3 CINTRA CNRS/NTU/THALES, UMI 3288, Singapore 637553 DOI: 10.1109/JPHOT.2012.2205911 1943-0655/$31.00 Ó2012 IEEE Manuscript received May 24, 2012; revised June 20, 2012; accepted June 20, 2012. Date of publication June 25, 2012; date of current version July 6, 2012. This work was supported in part by the A*STAR-Science and Engineering Research Council (SERC) Thematic Strategic Research Programme (TSRP), Singapore, under Grant 102-152-0012. Corresponding author: D. J. J. Hu (e-mail: jjhu@i2r.a-star.edu.sg). Abstract: We present a cascaded fiber device based on photonic crystal fiber (PCF) interferometers for refractive index (RI) sensing. The PCF modal interferometers have microbubbles at both sides of the splice regions. The microbubbles act as thick diverging lens that scatter light for efficient excitation of higher order cladding modes and attenuation of the core modes in the transmitted spectrum. Three resonance wavelengths are monitored, and their corresponding RI sensitivities are found to be 252, 187, and 207 nm/refractive index unit (RIU). The results, to our best knowledge, demonstrate that PCF interferometers with microbubbles are repeatable for high RI sensitivity, and the best crosstalk achievable is 0.295 nm/riu. Index Terms: Photonic crystal fibers (PCFs), interferometer, sensors, fiber optics systems. 1. Introduction Sensor network allows multiple sensing using the same set of source detector. Without sensor network, the cost will be overwhelming, and resources are wasteful to operate. Cost effective and intelligence optical sensor network has less expensive components, which come mainly from their source and detector component. In a multiple sensing application, if each sensor requires individual set of source and detection component, the cost of the sensor application would be prohibitively large. Hence, multiplexing is one way to reduce cost, time, and maintenance in sensor network. There are numerous types of multiplexing mechanisms where each has their own set of attributes. Photonic crystal fiber (PCF) gratings [1], [2] and interferometer [3] [8] sensors have been showing promising potential for temperature, strain, and/or refractive index (RI) sensing. It is worth mentioning that most research efforts, utilizing PCF interferometer with microbubble, have concentrated on strain and temperature sensing; RI sensing is challenging and is only possible with certain PCF structures for repeatability. In this paper, we cascade two PCF-based fiber sensors for RI sensing. Each sensor is constructed by a PCF modal interferometer that exploits the relative phase shift between two strongly coupled modes. Here in our experiment, we use an index-guiding multimode PCF, as shown in the Vol. 4, No. 4, August 2012 Page 1163

Fig. 1. (a) Schematic of the individual sensor fabricated with microbubble at the PCF SMF splice points. (b) Experimental setup. Fiber optics sensor arrangement with two cascaded sensors identified as A and B. Sensor A is immersed in a calibrated liquid (one droplet of 50 l) on a fixture for RI sensing. Inset shows the Scanning Electron Microscopy image of the PCF. Scale bar in the inset represents 10 m. inset of Fig. 1. Light can be guided in the cladding, and a substantial fraction of the optical power propagates as cladding modes inducing evanescent wave outside the fiber cladding. The all-silica PCF is fabricated by Yangtze Optical Fibre and Cable Company, China. The first-two rings of air holes are arranged in the triangular lattice with air-filling fraction 81.8%. The core size is 2.85 m. The first-two rings and third-to-fifth rings of air hole diameters are 5.8 m and 2.8 m, respectively. 2. Individual sensor The individual sensor, as shown in Fig. 1(a), is fabricated by splicing PCF to SMF, in which microbubbles form in the splice region after the fully collapsed zone. The fabrication process involves microhole collapse technique (MCT) using a fusion splicer machine [9] (Fujikura FSM- 40PM). The principle of the method is to substantially collapse the air holes at the splice region. There are a few parameters in the splicer setting that influence the splice result. First of all, the arc time and arc power are optimized in the fabrication, i.e., at 15.09 ma and 1 s, respectively. In addition, we re-arc the splice region for three times. The arc discharges here are weaker than the conventional SMF SMF splice setting (14.10 ma for 2 s). This adjustment will trap tiny air inside the splice interface and eventually aggregate all the air bubbles to a single microbubble when all the silica bridge collapsed onto the cladding. The weak arc discharges make it easier to control and optimize the fiber structure [10], in our case microbubble formation. Apart from the arc power and time, there is another important parameter known as offset to facilitate the control of air hole collapse. It shifts the arc position 80 m away from the SMF PCF interface on the SMF side. We use the same sets of arc parameters and perform MCT to test for the microbubble formation repeatability. Our results show that L MI median 19.8 m and standard deviation ¼ 2:71 m for sample size of 8. The success rate of microbubble formation is a hundred percent. We attribute the microbubble size variation to the angle of cleaved fibers ðþ before the splice. When G 2, we are able to obtain microbubble of repeatable geometry. For 2, microbubbles of irregular shape and unpredictable size are formed. We believe that this is due to imperfect cleaving of PCF facet. Ideally for fibers with a perfect cleave ¼ 0, the air gap between the SMFjPCF is uniform. Using the same splicing setting, the amount of air that is trapped to form the microbubble in the splice region is the same. As a result, the microbubble formation in the well-cleaved fibers would be highly repeatable in both shape and size. On the other hand, for an imperfect case 9 0, the air gap between the SMF PCF varies, causing some uncertainty during the microbubble formation during the splicing process. Therefore, well-cleaved fibers are desired to form consistent microbubble. Vol. 4, No. 4, August 2012 Page 1164

Dimensions of the fabricated sensors TABLE 1 After the first splice is done, the fiber is cleaved on the PCF (3 mm) side. Subsequently, the cleaved PCF end is spliced to SMF using the above procedures again. The fused silica at the splices enables the sensors to have very good mechanical strength. The cladding remains relatively unchanged at 125 m for SMF and 124 m for PCF. We place a sensor in a vertical position, fix the top end, and apply a steady force downwards by gravity at the bottom end with a load. Force of 1.96 N is applied by using Newton s second law of motion F ¼ ma, where m is mass (0.2 kg), and a is acceleration (9.80 m/s 2 ). The fiber did not suffer any distinguishable physical damage. While the force applied may not appear large, the force applied here would easily break conventional tapered fiber sensing devices. The dimensions of the fabricated sensors are tabulated in Table 1. The working principle is summarized as follows. When light enters the first microbubble in the splice region, it scatters two times by the two parabolic surfaces that excite higher order modes launching into the PCF. This is because the microbubble has an RI lower than the fiber; hence, it functions like a thick diverging lens. The first scattering occurs at the SMF air, whereas the second scattering occurs at the air-collapsed PCF. The SMF fundamental mode, centered at the middle of the parabolic surface, is attenuated significantly and scattered due to the parabolic profile. In the first fully collapsed region, fundamental mode and scattered modes begin to diffract. The diffraction excites multiple core and cladding modes into the PCF. Because the sensor reacts to environment RI changes around the cladding, its evanescent waves of the PCF cladding modes must have reached the external surface. In other words, the effective index of the cladding mode depends on the surrounding RI. The PCF core modes are isolated from the external RI changes. The core modes and cladding modes propagate at different speed along the PCF, hence will accumulate phase differences that are also dependent on wavelength. The modes propagate along the PCF until they reach the second fully collapsed region and begin to recombine. At the second microbubble, modes are scattered again before launching into the SMF. Inside the SMF, only the SMF fundamental mode will be guided and continue propagate to the OSA, which displays the interference pattern in the transmitted spectrum. We monitor the transmission spectrum while the second splice is fabricated. To control and optimize microbubble formation, four weak arc discharges are applied. This process not only creates microbubble, it also slows down the alteration of the microbubble. The repeated discharges during the microbubble formation modify the microbubble geometry slowly to remove fiber modes; hence, when sufficient light is scattered away by the microbubble, a very strong two-wave interference pattern is observed by the PCF modal interferometer. Two sensors with distinctive fringe minimal (dips) are selected and cascaded for measurement. We utilize the simple source sensors detector configuration, as depicted in Fig. 1(b). The setup consists of a broadband amplified spontaneous emission (ASE) light source and optical spectrum analyzer (OSA) as the source and the detector, respectively. The measured spectrum of the individual sensor in de-ionized (DI) water (RI ¼ 1:3323 at room temperature) is plotted in Fig. 2(a). Sensor A exhibits two transmission dips at 1546 nm and 1590 nm denoted by the solid line with star marker. Sensor B shows only one dip at 1557 nm in the spectral window from 1530 nm to 1600 nm denoted by the solid line with triangle marker. Assuming two Vol. 4, No. 4, August 2012 Page 1165

Fig. 2. (Color online) Experimental data (solid-lines) and theoretical data (dashed-lines) of the sensors in water RI ¼ 1:3323 at 24 C. Panels (a) and (b) show the spectrum of the individual sensors and the cascaded sensors, respectively. parallel waves are participating in the inference [11] for both sensors, the spectra of the sensor can be expressed as p I i ¼ I 1 þ I 2 þ 2 ffiffiffiffiffiffiffi I 1 I 2cosði Þ (1) where I 1 and I 2 are the irradiance of the two interfering waves. is the phase difference between the two coupling modes traveling at different optical path length. i is the sensor designation that refers to sensor A or sensor B. i (in radian) is expressed as i ¼ 2 n il i þ i (2) where L i is the microbubble-to-microbubble distance, is the wavelength, i is the initial phase difference of the interfering waves, and n is the effective index difference. We use (1) to fit the measured spectrum of the individual sensor and plot the simulated spectrum as dashed line in Fig. 2(a). The fit gives I 1 ¼ 1:0, I 2 ¼ 0:6, n 1 ¼ 0:0195, n 2 ¼ 0:0165, A ¼ 0:3456, and B ¼ 0:4398. When multiple fiber modes are excited in the fully collapsed zone, the light also experiences multiple Fresnel reflection and refraction at the cladding surrounding interface. To account for these and the scattering losses, we include another 20-dB loss mechanisms to (1). The simulated (dashed-lines) and measured (solid-lines) spectrum for sensor B superimpose well over each other. On the other hand, we observe a larger deviation for sensor A, which may be resulted from the simplified assumption of only two waves participating in the interference. Nevertheless, the simplified simulation is helpful to provide insight for the cascaded device of sensor A and sensor B, which we demonstrate in the experiment. The PCF length L ¼ L MM þ L MI þ L MO is noticeably shorter than reported PCF interferometer previously [3] [5], [7]. We fabricated longer sensors (i.e., 5 mm, 10 mm, 20 mm) using the same procedures and did not see a dependence on the free spectral range. We notice an elevated increase in fringes (seen as random fluctuation) in the transmission spectrum as L increases. When we subject the longer sensors to RI testing, we did not observe a dependable shift in spectrum. We elucidate the behaviors to the current fabrication settings where the microbubble geometry is not optimized to scatter away fiber modes at different PCF length. Vol. 4, No. 4, August 2012 Page 1166

3. Cascaded sensor Next, we cascade both sensors and measure the resultant spectrum in DI water denoted as solid line with circle in Fig. 2(b). The dips are matching those in the individual measurement in Fig. 2(a), i.e., at 1546, 1557, and 1590 nm. However, the dip of sensor B at 1557 nm plunges 3 db due to the broadening of sensor A left minima. Likewise, the dip of sensor A at 1546 nm is flattened due to the presence of sensor B. The simulated spectrum of the cascaded device is obtained by summing up (1) for individual sensors and plotted in Fig. 2(b) by dashed line. When cascading two PCF sensors together by splicing the SMF sides, the distance between two sensors is around 50 cm. There is no observable difference in the sensor spectrum for the distance ranging from 5 cm to longer length. The reason for selecting 50 cm is due to the ease of fabrication. In practical application, the distance can be extended to much longer values because SMF is a very low loss transmission media giving rise to propagation loss of 0.2 db/km. We examine the RI sensing by immersing only one sensor in the varying RI liquid, whereas the other sensor is immersed in DI water. The RI liquid is a mixture of glycerin (Fisher Scientific Certified ACS) and DI water measured by a commercial refractometer (Reichert r2 mini) with an RI measurement accuracy of 2 10 4. By mixing different concentrations of the mixture, we vary the RI of the solution from 1.3323 to 1.3751. In between consecutive measurement, we clean the device with 2-propanol and dry it in air. The refractometer has automatic temperature compensation, and we check the RI prior to each RI measurement. PCF provides excellent temperature stability since it is made of single material (pure silica). This advantage gives PCF relatively low thermo-optics coefficient and typical insensitive to temperature [1], [4] [7]. Thermal effect is minimized by carrying out the experiment in a temperature-stable room. Thermometer in the laboratory measures air temperature stability with fluctuation smaller than 0.1 C. Furthermore, the experiment was conducted on a thermal insulated fixture, and the setup was covered to reduce heat transfer from air. In view of ambient temperature variation less than 0.1 C, the temperature drift appears negligible [12]. For the clarity of the discussion, we number the dips (1, 2, and 3) according to their spectral position shown in Fig. 3. Fig. 3(a) demonstrates the spectrum change when sensor A is immersed in varying RI liquid and sensor B is immersed in DI water all the time. Fig. 3(b) is obtained by plotting the wavelength of the dips as a function of the varying RI from Fig. 3(a). The reason for carrying out this measurement is to demonstrate the simultaneous measurement using the cascaded sensor device, e.g., both sensors (A & B) can be used to measure different RI liquid. In addition, the measurement is also used to analyze the crosstalk between two sensors. It can be observed in Fig. 3(b) that dip 3 reserves its shape well while shifting to the longer wavelength as RI increases. In comparison, dip 1 gradually becomes indistinguishable due to the presence of sensor B. In addition, dip 2 shows differences in power, and its spectral position change is not visible. The resonance wavelengths of the cascaded device are monitored as a function of the varying RI and plotted in Fig. 3(b). The dip 1 (square) and dip 3 (circle) from sensor A have RI sensitivity of 252 nm/refractive index unit (RIU) and 207 nm/riu, respectively. They exhibit fair linearity when approximated by the linear fitting lines. Considering the OSA specification wavelength accuracy of 20 pm, the device theoretical sensitivity can be estimated to be 7:94 10 5 RIU. This estimated RIU affirms our device measurement accuracy to the commercial refractometer 2 10 4 used to calibrate the RI liquid. Dip 2 from sensor B shows a flat RI response; the gradient of the linear fitting line is 8.95 nm/riu. Similarly, Fig. 3(c) demonstrates the spectrum change when sensor A is immersed in DI water all the time and sensor B is immersed in varying RI liquid. Dip 2 resulted from sensor B shows clear red shift as RI increases, whereas the other dips resulted from sensor A show unnoticeable shifts with RI. Fig. 3(d) is obtained by plotting the wavelength of the dips as a function of the varying RI from Fig. 3(c). Dip 2 (diamond) shows good linearity around 187 nm/riu, whereas dip 1(square) shows RI response of 13.5 nm/riu, and dip 3 shows RI sensitivity of 0.295 nm/riu. Clearly, because of the proximity of dip 1 and dip 2, the measurements of their spectral positions are challenging, especially in the case shown in Fig. 3(a). From Fig. 3(a), because of the proximity of the resonance Vol. 4, No. 4, August 2012 Page 1167

Fig. 3. (Color online) (a) Spectra of the cascaded device when sensor A is immersed in varying RI liquid and sensor B is immersed in DI water throughout the measurement. (c) Spectra of the cascaded device when sensor B is immersed in varying RI liquid and sensor A is immersed in DI water throughout the measurement. (b) and (d) RI response of the resonant wavelength of the dips, i.e., dip 1 (square), dip 2 (diamond), and dip 3 (circle). Linear fitting lines are used to fit the RI sensitivity. The arrows indicated the direction the dips shifted when the RI increases. Error bars in (b) and (d) indicate the uncertainty in the reported measurement. dip of sensor A at 1546 nm and that of sensor B at 1557 nm, when sensor A is subject to varying RI liquid, the two dips become Boverlapped,[ and it is difficult to distinguish the dips individually. As a comparison, Fig. 3(c) shows a good example when sensor B is immersed in varying RI liquid, and the spectrum shows distinctive dips. In addition, there is crosstalk between dip 1 and dip 2 despite the measurement is carried out by immersing one sensor in the plain DI water. Comparatively, because dip 3 is spectrally distant from dip 2, its crosstalk is much less significant. The crosstalk cannot be removed by regular filter and accumulated over the number of sensors. In order to reduce crosstalk between two sensors, the spectral separation of the resonance dips of the two sensors has to be enlarged. In other words, the individual fabricated sensor should be carefully chosen with spectrally apart resonance dip from the other sensor. Therefore, for multiplexed sensing purpose, it is suggested to monitor dip 2 and dip 3 if sensor A and sensor B are exposed to different RI liquids for detection. The maximal I max and minimal I min of one sensor are at 20 and 40 db, respectively. This implies that a maximum of two sensors is permissible on our interrogator (OSA), which has a detection limit d l ¼ 70 db; when I min of one sensor is exactly at the I max of the other sensor, I max þ I min 9 di db. For three sensors, I max þ I max þ I min is less than G d l. This will prevent us from using the OSA for measurement. In practice, we may have to employ further active devices to detect or amplify the weak signals when cascading multiple sensors for multiplex sensing. Two examples of such active devices, commonly used in optical fiber communication, are erbium-doped fiber amplifier (EDFA) or semiconductor optical amplifier (SOA) placed in between sensors or as a preamplifier. These active devices are relatively cheaper than purchasing entire source and interrogator units. Other ways are to supply a stronger light source to raise the power budget above the minimum received power or use smaller ER sensors by tweaking the microbubble formation settings in terms of arc power and time and number of arc discharges. The distance between sensors is another limiting factor if the sensors are placed extremely close to one another. The modes from one sensor can travel to the next sensor and distort the whole measurement. To rapidly Vol. 4, No. 4, August 2012 Page 1168

strip away nonfundamental mode, one can easily apply a high RI glue or epoxy at the SMF in between the sensors. Lastly, we would like to mention that the sensors cannot be cascaded indefinitely. The accumulated noise figure in active devices and the insertion loss would eventually make the interrogation impractical. Although we can achieve better insertion loss ( 14 db) sensors, they are generally less sensitive. This is because more power is confined within the PCF, and less optical power is excited to the cladding modes for evanescent waves interaction with surrounding liquid. 4. Conclusion In summary, we report and experimentally demonstrate a cascaded fiber device for RI sensing. The sensors are fabricated by MCT that involves simple cleaving and splicing processes only. The resonance wavelengths of the device are located at different spectral positions, and they exhibit different RI response, making multiplexed RI sensing possible. The RI sensitivities of the cascaded device are 252, 187, and 207 nm/riu by measuring the wavelength shifts of the resonance dips at 1546, 1557, and 1590 nm. In addition, the cascaded device exhibits good linearity of RI response. The minimum crosstalk in the fabricated device is 0.295 nm/riu. The reported device can find promising applications such as biochemical sensing, which desires multiplexing capabilities. References [1] Y. Zhu, P. Shum, H.-W. Bay, M. Yan, X. Yu, J. Hu, J. Hao, and C. Lu, BStrain-insensitive and high-temperature longperiod gratings inscribed in photonic crystal fiber,[ Opt. Lett., vol. 30, no. 4, pp. 367 369, Feb. 2005. [2] Y. Zhu, P. Shum, J.-H. Chong, M. K. Rao, and C. Lu, BDeep-notch, ultracompact long-period grating in a large-modearea photonic crystal fiber,[ Opt. Lett., vol. 28, no. 24, pp. 2467 2469, Dec. 2003. [3] J. Villatoro, V. P. Minkovich, V. Pruneri, and G. Badenes, BSimple all-microstructured-optical-fiber interferometer built via fusion splicing,[ Opt. Exp., vol. 15, no. 4, pp. 1491 1496, Feb. 2007. [4] G. A. Cárdenas-Sevilla, V. Finazzi, J. Villatoro, and V. Pruneri, BPhotonic crystal fiber sensor array based on modes overlapping,[ Opt. Exp., vol. 19, no. 8, pp. 7596 7602, Apr. 2011. [5] R. Jha, J. Villatoro, and G. Badenes, BUltrastable in reflection photonic crystal fiber modal interferometer for accurate refractive index sensing,[ Appl. Phys. Lett., vol. 93, no. 19, pp. 191106-1 191106-3, Nov. 2008. [6] E. Li, G.-D. Peng, and X. Ding, BHigh spatial resolution fiber-optic Fizeau interferometric strain sensor based on an infiber spherical microcavity,[ Appl. Phys. Lett., vol. 92, no. 10, pp. 101117-1 101 117-3, Mar. 2008. [7] S. Silva, J. L. Santos, F. X. Malcata, J. Kobelke, K. Schuster, and O. Frazão, BOptical refractometer based on large-core air-clad photonic crystal fibers,[ Opt. Lett., vol. 36, no. 6, pp. 852 854, Mar. 2011. [8] D. J. J. Hu, J. L. Lim, M. K. Park, L. T.-H. Kao, Y. Wang, H. Wei, and W. Tong, BPhotonic crystal fiber based interferometric biosensor for streptavidin and biotin detection,[ IEEE J. Sel. Topics Quantum Electron., vol. 18, no. 4, pp. 1293 1297, Jul. 2012. [9] L. Xiao, M. S. Demokan, W. Jin, Y. Wang, and C.-L. Zhao, BFusion splicing photonic crystal fibers and conventional single-mode fibers: Microhole collapse effect,[ J. Lightw. Technol., vol. 25, no. 11, pp. 3563 3574, Nov. 2007. [10] L. Xiao, W. Jin, and M. S. Demokan, BFusion splicing small-core photonic crystal fibers and single-mode fibers by repeated arc discharges,[ Opt. Lett., vol. 32, no. 2, pp. 115 117, Jan. 2007. [11] E. Hecht, BInterference,[ in Optics, 4th ed. San Francisco, CA: Addison Wesley, 2002, pp. 385 389. [12] D. J. J. Hu, J. L. Lim, M. Jiang, Y. Wang, F. Luan, P. P. Shum, H. Wei, and W. Tong, BLong period grating cascaded to photonic crystal fiber modal interferometer for simultaneous measurement of temperature and refractive index,[ Opt. Lett., vol. 37, no. 12, pp. 2283 2285, Jun. 2012. Vol. 4, No. 4, August 2012 Page 1169