Highly sensitive temperature sensor based on apolymer-infiltrated Mach–Zehnder interferometercreated in graded index fiberFENGCHAN ZHANG,1,2 XIZHEN XU,1,2 JUN HE,1,2,* BIN DU,1,2 AND YIPING WANG1,2

1Key Laboratory of Optoelectronic Devices and Systems of Ministry of Education and Guangdong Province,College of Physics and Optoelectronic Engineering, Shenzhen University, Shenzhen 518060, China2Guangdong and Hong Kong Joint Research Centre for Optical Fibre Sensors, Shenzhen University, Shenzhen 518060, China*Corresponding author: hejun07@szu.edu.cn

Received 4 February 2019; revised 17 April 2019; accepted 20 April 2019; posted 22 April 2019 (Doc. ID 359227); published 9 May 2019

A highly sensitive temperature sensor is proposed anddemonstrated based on a UV-curable polymer-infiltratedMach–Zehnder interferometer (MZI) created in a graded in-dex fiber (GIF). The device was constructed by splicing ahalf-pitch GIF between two single-mode fibers and creatingan inner air cavity in one lateral side of the GIF coreby means of femtosecond laser micromachining. The air cav-ity and the residual GIF core functioned as two interferencearms of the MZI. Moreover, the GIF was used as a miniaturein-fiber collimator to reduce insertion loss of the air cavity.Experimental results show such an MZI device has a highrefractive index (RI) sensitivity of 24611.54 nm/RIU(RI � 1.545–1.565). Subsequently, thermo-sensitive poly-mer liquid was infiltrated into the air cavity, then cured withUV illumination, and annealed at 50°C for 12 h. The infil-trated MZI exhibits a high temperature sensitivity of−13.27 nm∕°C. In addition, this MZI also has excellent ther-mal stability and repeatability, compact structure, low inser-tion loss, and high fringe visibility. As such, the proposedMZI could be developed for high-accuracy temperaturemeasurements in many areas such as biomedical or oceano-graphic applications. © 2019 Optical Society of America

https://doi.org/10.1364/OL.44.002466

High-accuracy temperature measurements are required in vari-ous applications. For example, biomedical scientists need torecord the hypothalamic temperature with an accuracy of0.1°C for diagnosis [1]. Marine geologists need to monitorthe ocean temperature with an accuracy of 0.01°C to studyglobal climate change [2]. High-sensitivity temperature sensorsare beneficial for achieving such high-accuracy temperaturemeasurements, and hence are widely studied at present.Compared with other techniques, fiber-optic temperaturesensors are attractive due to their unique advantages of highsensitivity, compact size, and immunity to electromagnetic in-terference. The most widely used fiber temperature sensors arebased on fiber Bragg gratings (FBGs) [3]. However, FBGshave a typical low temperature sensitivity of ∼10 pm∕°C,

which is not sufficient for high-accuracy temperature measure-ments. Moreover, in-fiber Mach–Zehnder interferometers(MZIs) are also developed for temperature sensing. Variousin-fiber microstructures, such as tapers [4,5], internal aircavities [6,7], and core-offset splicing joints [8,9], have alreadybeen created to excite interference between the core mode andcladding modes in MZIs. However, these MZIs still have a lowtemperature sensitivity of ∼100 pm∕°C due to the small differ-ence in thermal-optic coefficients (TOCs) of the silica fibercore and cladding materials. Hence, researchers sought to in-crease the temperature sensitivity by use of large-TOCmaterial.For example, Yang et al. reported an MZI by selectivelyinfiltrating two air holes in photonic crystal fiber (PCF) andachieved a high sensitivity of 7.3 nm/°C [10]. In 2013,Liang et al. developed an MZI by selectively filling refractiveindex (RI)-matching liquid into one air hole of the PCFand achieved a high sensitivity of 16.49 nm/°C [11].Recently, Zhang et al. reported an MZI temperature sensorbased on a liquid-filled D-shaped cavity and achieved an ultra-high sensitivity of −84.72 nm∕°C [12]. Additionally, we havefabricated a Fabry–Perot interferometer (FPI) by filling mer-cury in a silica tube and exhibited an ultrahigh temperaturesensitivity of −41.9 nm∕°C [13]. However, it had poor thermalrepeatability and stability due to the unstable reflection of themercury surface. Therefore, most of the highly sensitive MZItemperature sensors at present are based on the infiltrationof thermal-sensitive liquid material, which leads to a smalloperation range and unstable performance.

In this Letter, we propose and demonstrate a highly sensitivetemperature sensor based on an MZI created in a half-pitchgraded index fiber (GIF) and infiltrated with UV-curable poly-mer. An air cavity was created in one lateral side of the GIF coreat the quarter-pitch position of GIF by use of femtosecondlaser micromachining. Subsequently, UV-curable polymerliquid was infiltrated into the air cavity, then cured with UVillumination and annealed at 50°C for 12 h. As a result, thepolymer-infiltrated air cavity and the residual GIF core func-tioned as two arms of the MZI. Moreover, the cured polymerhad a stable and large TOC of ∼−2 � 10−4 RIU∕°C [14],

2466 Vol. 44, No. 10 / 15 May 2019 / Optics Letters Letter

0146-9592/19/102466-04 Journal © 2019 Optical Society of America

resulting in a high sensitivity and stable performance of theMZI temperature sensor. In addition, the quarter-pitch GIFfunctioned as a miniature in-fiber collimator [15], which couldenlarge the optical mode area and reduce the divergence angle.Hence, the use of GIF can reduce the insertion loss and increasethe fringe visibility of the in-fiber cavity-based interferometer[16,17]. Experimental results show the polymer-infiltratedMZI has high temperature sensitivity, excellent thermal stabil-ity and repeatability, compact structure, low insertion loss, andhigh fringe visibility.

The structure and working principle of the proposed in-fiber MZI temperature sensor are shown in Fig. 1. An incidentbeam is directed into a half-pitch GIF through a lead-in single-mode fiber (SMF) (i.e., SMF1). Light travels along a quasi-sinusoidal trajectory in the GIF. Hence, at the middle ofthe half-pitch GIF (i.e., the quarter-pitch position), the modearea of the incident beam reaches the maximum, and thedivergence angle reaches the minimum. In contrast, the modearea reaches the minimum at the end of the half-pitch GIF. Anair cavity is created in one lateral side of the GIF core at thequarter-pitch position and infiltrated with UV-curable poly-mer. As a result, the polymer-infiltrated cavity and the residualGIF core can function as two interference arms of the MZI, inwhich the intensities of two beams are denoted as Ip and Ic ,respectively. At the end of the half-pitch GIF, the two beamsrecombine and interfere with each other. The light intensity inthe lead-out SMF (i.e., SMF2) is given by [18]

I out � I p � I c � 2ffiffiffiffiffiffiffiffiffiffiffiI p · I c

qcos�2πLΔn∕λ� φ0�, (1)

where L is the air cavity length, Δn � npolymer − ncore is the RIdifference of two interference arms, where ncore and npolymer arethe effective RIs in the GIF core and UV-curable polymer,respectively, λ is the light wavelength, and φ0 is the initial phasedifference. At each interference minimum, the phase shift inMZI should satisfy

2πLΔn∕λm� φ0 � �2m� 1�π, (2)

where m is an integer defining the fringe order, and λm is thewavelength of the mth order interference dip, defined as

λm � 2LΔn∕�2m� 1�: (3)

Note that the initial phase difference φ0 could be ignored inEq. (3) for simplicity if the absolute values of λm are not nec-essary in the following calculations. Moreover, the free spectralrange (FSR) can be determined by

FSR � λ2∕�L · Δn�: (4)

The fabrication process of the MZI temperature sensor includessix steps. In step 1, as shown in Fig. 2(a), a section of GIF(Yangtze Optical Fiber, 62.5/125GI 0.275) and a commonSMF (i.e., SMF1) were spliced together by means of electricalarc discharge. In step 2, as shown in Fig. 2(b), the GIF with ahalf-pitch length of 490 μm was obtained by using a fiber pre-cision cutting system with a precision of �5 μm. In step 3, asshown in Fig. 2(c), the SMF-half-pitch GIF structure wasplaced in the left motor and spliced with another SMF (i.e.,SMF2) placed in the right motor. In step 4, as shown inFig. 2(d), the SMF–GIF–SMF structure was fixed on a com-puter-controlled three-axis (xyz) translation stage with a stepresolution of 10 nm. An air cavity was then created by meansof femtosecond laser micromachining with a precision of�1 μm. The femtosecond laser (Spectra-Physics) with a wave-length of 800 nm, pulse duration of 120 fs, and repetition rateof 1 kHz was used. The average on-target laser power was set tobe 15 mW. The initial position of the laser focal spot is indi-cated by a red dot, which was initially focused on the surface ofGIF by an objective lens with an NA of 0.25. The air cavity wascreated by direct femtosecond laser ablation and cleaned by us-ing alcohol after micromachining. The top-view and side-viewmicroscope images of a fabricated MZI with an air cavity lengthof 80 μm are shown in Figs. 3(a) and 3(b). In step 5, as shownin Fig. 2(e), the UV-curable polymer was infiltrated into the aircavity and exposed to UV irradiation with a wavelength of365 nm and intensity of approximately 27 W∕cm2 for40 min. In step 6, as shown in Fig. 2(f ), the polymer-infiltratedMZI was kept at 50°C for 12 h. The annealing process canimprove the stability of the MZI temperature sensor.Moreover, the UV-curable polymer material can withstand awide temperature range from −80 to �90°C after being

Fig. 1. Schematic diagram of the proposed MZI temperature sensorcreated in a GIF and infiltrated with UV-curable polymer. Insets P1,P2a, P2b, and P3 show the near mode fields measured at the SMF1output, before the air cavity, after the air cavity, and at the half-pitchGIF output, respectively. Note that P3-max and P3-min were obtained atdifferent input laser wavelengths.

Fig. 2. Flow chart of device fabrication: (a) splicing SMF1 withGIF; (b) cutting the GIF into a half-pitch length; (c) splicing half-pitchGIF with SMF2; (d) fabricating air cavity by femtosecond laser micro-machining; (e) polymer infiltrating and curing with UV illumination;(f ) annealing.

Letter Vol. 44, No. 10 / 15 May 2019 / Optics Letters 2467

completely cured. The microscope image of the polymer-infiltrated MZI is shown in Fig. 3(c).

At first, three MZIs with different air cavity lengths of40 μm, 60 μm, and 80 μm were fabricated before infiltratingwith polymer. The fringe visibility and insertion loss of theseMZIs were investigated. The transmission spectra were mea-sured by using a broadband light source (BBS, Fiber Lake)and an optical spectrum analyzer (OSA, Yokogama, AQ6370C,wavelength resolution: 0.02 nm). As shown in Fig. 4(a), fringevisibilities of ∼21 dB, ∼20 dB, and ∼23 dB, and insertionlosses of ∼10 dB, ∼10 dB, and ∼11 dB were obtained inMZIs with air cavity lengths of 40 μm, 60 μm, and 80 μm,respectively. For comparison, the transmission spectrum ofthe SMF–GIF–SMF structure without the air cavity is alsoshown in Fig. 4(a), which exhibits a low insertion loss of∼0.25 dB. Then these MZIs were immersed in RI-matchingliquid (Cargille Labs, RI � 1.566). As shown in Fig. 4(b),the insertion loss is reduced to ∼2 dB, ∼4 dB, and ∼5 dBfor three MZIs with air cavity lengths of 40 μm, 60 μm,and 80 μm, respectively. Note that the insertion loss of theMZIs in this work is much lower than that reported in the pre-vious work by Li et al. [19], in which a similar air cavity wascreated in single-mode twin-core fiber (TCF), and their MZIhad an insertion loss of ∼10 dB. The results reveal the GIF canreduce insertion loss efficiently, i.e., the first quarter-pitch GIF(before the air cavity) functions as an in-fiber collimator, andthe second quarter-pitch GIF (after the air cavity) can shrinkthe mode field to match with the lead-out SMF. Hence, thedivergence angle in the air cavity is reduced, leading to reducedinsertion loss. Additionally, it was observed that the fringe FSRdecreased with an increasing cavity length. The result agreeswell with the expectations from Eq. (4). A smaller FSR isbeneficial for detecting the dip wavelength and improving mea-surement accuracy, while a larger FSR is out of the OSA mea-surement range. It is obvious the MZI with a cavity length of80 μm has a smaller FSR and an acceptable insertion loss of

∼5 dB. As a result, we chose the air cavity length of 80 μmfor fabricating the MZI temperature sensors in the followingexperiments. Furthermore, it should be noted that the widthof the air cavity was chosen to be 35 μm to optimize the lightintensity ratio of two interference arms and hence obtain themaximum fringe visibility.

Subsequently, we studied the RI response of MZIs byimmersing the air cavity into a series of RI-matching liquidswith RI � 1.300–1.650. The experimental results are shownin Figs. 5 and 6. It can be deduced from Eq. (3) that thedip wavelength of MZI shifts with a changing RI differenceΔn of the two arms. The RI sensitivity can be derived asdλm∕d �Δn� � λm∕Δn. It is obvious that the RI sensitivityis determined only by the RI difference Δn of the two armsand is independent of the air cavity length. The RI sensitivityreaches its maximum when the RI of the material infiltrated inthe air cavity approaches the RI of the GIF core. If the RI isincreased from 1.400 to 1.416, as shown in Figs. 5(a1) and5(a2), the dip wavelength exhibits a blue shift with an RI sen-sitivity of −14184.78 nm∕RIU. If the RI is increased from1.548 to 1.562, as shown in Figs. 5(b1) and 5(b2), the dipwavelength exhibits a red shift with a very high RI sensitivityof 24611.54 nm/RIU. If the RI is increased from 1.612 to1.628, as shown in Figs. 5(c1) and 5(c2), the dip wavelengthexhibits a red shift with an RI sensitivity of 11048.84 nm/RIU. In addition, the changes in RI have a negligible effecton the MZI dip intensities. Moreover, we measured the RI sen-sitivities of the MZI infiltrated with different RI-matching

Fig. 3. Microscope images of (a) top view and (b) side view of theMZI with an air cavity length of 80 μm before polymer infiltration,and (c) polymer-infiltrated Mach–Zehnder interferometer.

Fig. 4. (a) Transmission spectra of three MZIs with different aircavity lengths of 40 μm, 60 μm, and 80 μm before infiltrating withpolymer and (b) transmission spectra of three MZIs with different aircavity lengths of 40 μm, 60 μm, and 80 μm after infiltrating withRI-matching liquid (RI � 1.566).

Fig. 5. Transmission spectra evolutions of an MZI immersed in dif-ferent RI-matching liquids with RI range from (a1) 1.400 to 1.416,(b1) 1.548 to 1.562, and (c1) 1.612 to 1.628; dip wavelength anddip intensity of the MZI as functions of the RI of the infiltrated liquidsfrom (a2) 1.546 to 1.562, (b2) 1.400 to 1.416, and (c2) 1.612 to 1.628.

Fig. 6. Measured and calculated RI sensitivities of the MZI infil-trated with different RI-matching liquids (RI � 1.300–1.650).

2468 Vol. 44, No. 10 / 15 May 2019 / Optics Letters Letter

liquids (RI � 1.300–1.650). As shown in Fig. 6, the MZI has anegative RI sensitivity when the RI is lower than 1.491,whereas it has a positive RI sensitivity when the RI is higherthan 1.491. The absolute value of RI sensitivity reaches themaximum when the RI approaches a turning point of1.491. It can also be seen in Fig. 6 that the measured resultsagree well with the calculated results from Eq. (3).

Furthermore, we fabricated an MZI with an air cavity lengthof 80 μm and infiltrated the air cavity with UV-curable poly-mer (NOA 68, RI � 1.540 at room temperature). After infil-tration, the MZI was exposed to UV illumination for 40 minuntil the transmission spectrum was stabilized and the polymerliquid was fully cured. Finally, the polymer-infiltrated MZI wasannealed at 50°C for 12 h. The temperature responses of thepolymer-infiltrated MZI were studied before and afterannealing by use of a high-precision column oven (LCO102) with an accuracy of 0.1°C. As shown in Fig. 7, theMZI dip wavelength exhibits a blue shift with an increasingtemperature and a red shift with a decreasing temperature.Before annealing, the temperature sensitivities in the heatingand cooling processes are −12.91 nm∕°C and −13.26 nm∕°C,respectively. After annealing, the thermal stability and repeat-ability of the infiltrated MZI are improved significantly. Thetemperature sensitivities in the heating and cooling processesare −13.27 nm∕°C and −13.20 nm∕°C, respectively. Thevariation in dip intensity is smaller than 1.16 dB during thetemperature cycling process.

The high temperature sensitivity of the polymer-infiltratedMZI results from the high RI sensitivity of the MZI at RI �1.540 and the negative and large TOC of the UV-curablepolymer material (∼−2 � 10−4 RIU∕°C) [14]. Additionally,the annealed UV-curable polymer-infiltrated MZI temperaturesensor proposed in this work should have the capability of op-erating in a wider temperature range from −80°C to �90°C.The relatively small temperature range (from 24°C to 42°C)

demonstrated in Fig. 7 is partially limited by the oven usedin this experiment, and partially limited by the operation rangeof the OSA. This shortcoming could be solved by using eitheran OSA with a wider operation range or a high-resolution in-tensity demodulation scheme.

In summary, we demonstrated a highly sensitive temperaturesensor based on a UV-curable polymer-infiltratedMZI created ina GIF. Experimental results show such an MZI temperature sen-sor has a high temperature sensitivity of −13.27 nm∕°C, excel-lent thermal stability and repeatability, compact structure (lessthan 0.5 mm), low insertion loss of ∼5 dB, and high fringe vis-ibility of∼25 dB. Note that the insertion loss and fringe visibilityreported in this work are much better than previous reports[7,10–12,18,19]. As a result, the proposed MZI temperaturesensor could be further developed for high-accuracy temperaturemeasurements in many areas such as biomedical or oceano-graphic applications.

Funding. National Natural Science Foundation ofChina (NSFC) (61875128, 91860138, 61635007); Scienceand Technology Innovation Commission of Shenzhen(JCYJ20180507182058432, JCYJ20170302143105991,JCYJ20160427104925452); Development and Reform Com-mission of Shenzhen Municipality Foundation.

REFERENCES

1. K. Song, H. Wang, G. B. Kamm, and J. Pohle, Science 353, 1393(2016).

2. B. Bereiter, S. Shackleton, D. Baggenstos, K. Kawamura, and J.Severinghaus, Nature 553, 39 (2018).

3. A. D. Kersey, M. A. Davis, H. J. Patrick, K. P. Koo, and C. G. Askins, J.Lightwave Technol. 15, 1442 (1997).

4. P. Lu, L. Q. Men, S. Kevin, and Q. Y. Chen, Appl. Phys. Lett. 94,131110 (2009).

5. T. Chen, R. Chen, P. Lu, Q. Chen, and K. P. Chen, Electron. Lett. 47,1093 (2011).

6. L. Jiang, J. Yang, S. Wang, B. Li, and M. Wang, Opt. Lett. 36, 3753(2011).

7. T. Y. Hu, Y. Wang, C. R. Liao, and D. N. Wang, Opt. Lett. 37, 5082(2012).

8. J. T. Zhou, C. R. Liao, Y. P. Wang, G. L. Yin, K. M. Yang, and Z. Y. Li,Opt. Express 22, 1680 (2014).

9. X. Hao, Z. R. Tong, W. H. Zhang, and Y. Cao, Opt. Commun. 335, 78(2015).

10. M. Yang, D. N. Wang, Y. Wang, and C. R. Liao, Opt. Lett. 36, 636(2011).

11. H. Liang, W. G. Zhang, H. Y. Wang, P. C. Geng, S. S. Zhang, C. X.Yang, and J. L. Li, Opt. Lett. 38, 4019 (2013).

12. H. Zhang, S. C. Gao, Y. H. Luo, Z. S. Chen, S. S. Xiong, L. Wan, andX. C. Huang, Sensors 18, 1239 (2018).

13. K. M. Yang, J. He, Y. Wang, S. Liu, C. R. Liao, and Y. P. Wang, IEEEPhoton. J. 7, 6803509 (2015).

14. A. Priyadarshi, L. Shimin, S. G. Mhaisalkar, R. Raioo, E. H. Wong, andV. Kripesh, J. Appl. Polym. Sci. 98, 950 (2005).

15. X. Z. Xu, J. He, M. X. Hou, S. Liu, Z. Y. Bai, and Y. Wang, J. LightwaveTechnol. 36, 2827 (2018).

16. Y. N. Zhang, Y. J. Li, T. Wei, X. W. Lan, Y. Huang, and G. D. Chen,IEEE Photon. J. 2, 469 (2010).

17. D. Q. Ouyang, S. C. Ruan, C. Y. Guo, and H. F. Wei, J. LightwaveTechnol. 32, 2524 (2014).

18. Y. Wang, Y. H. Li, C. R. Liao, and D. N. Wang, IEEE Photon. Technol.Lett. 22, 39 (2010).

19. Z. Y. Li, C. R. Liao, Y. P. Wang, X. P. Dong, S. Liu, and K. M. Yang,Opt. Lett. 39, 4982 (2014).

Fig. 7. Transmission spectra evolutions of the MZI infiltrated withUV-curable polymer during the heating and cooling processes: (a1)before annealing and (a2) after annealing; dip wavelength and dip in-tensity as functions of the temperature during the heating and coolingprocesses: (b1) before annealing and (b2) after annealing.

Letter Vol. 44, No. 10 / 15 May 2019 / Optics Letters 2469