Open Access
Add to BibSonomy
Abstract
This study evaluates the interrogation techniques of a hybrid fiber optic sensor based on surface plasmon resonance (SPR) and multimode interference (MMI). The sensor is based on a single mode, fiber-no core, fiber-single mode fiber (SMF-NCF-SMF) structure with a deposited gold film layer. Both SPR and MMI effects are excited in a single sensor structure without enlarging the device size. However, at the same time, the interference fringe patterns are also mixed with the SPR transmission spectra, and the traditional SPR interrogation technique becomes unavailable since the resonant wavelength is hard to be located. In this study, the fast Fourier transform and different filtering algorithms are applied, both SPR signal and interference signal with different orders are separated effectively due to their different spatial frequency distributions, and they are processed individually for refractive index (RI) sensing. The experimental results verify that the overall RI sensitivity of the hybrid sensor is significantly enhanced. This study provides an important supplement to the traditional SPR and MMI functions.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
In recent years, a variety of cascaded fiber optic sensors have been reported for dual-parameter sensing [1–3]. Generally, the sensors are cascaded with two elements that possess different sensing performances in the same ambient conditions, such as highly integrated Fabry-Perot /FBG sensor [4,5], PDMS/ tapered fiber sensor [6], few-mode fiber/FBG [7], etc. However, these kinds of sensors are based on the combination of two different elements, which enlarge inevitably the sensor size. Besides, as for the cascaded fiber optic SPR sensors, they are fabricated by splicing a special fiber (photonic crystal fiber, etc.) between two multimode fibers [8,9]. Generally, these sensors can measure only one parameter at a time. In the authors’ previous work [10], we proposed a cascaded hybrid fiber optic sensor based on SPR and MMI for dual-parameter measurement. Based on the special design, both SPR effect and MMI effects were excited in a fiber structure without enlarging the sensor size. However, in this work, the SPR effect was excited in the visible region, and the MMI effect was excited in the near-infrared region of spectrum. As a result, two spectrometers were needed for detection since the transmission spectrum covered a wide wavelength range. Hence, the volume and the cost of the whole sensing system are still increased.
Besides, since both SPR effect and MMI effect are excited, the interference fringe patterns are mixed into the SPR spectrum and the final transmission spectrum curve becomes irregular. Hence the specific interrogation technique should be evaluated. The traditional interrogation techniques of SPR sensors include wavelength demodulation [11–13], intensity demodulation [13,14], phase demodulation [15,16], and angle demodulation [17]. Recently, there are also the studies relating to the interrogation technique optimization based on the curve shaping method [18] and the spatial frequency spectrum analyses [19]. However, all the above interrogation techniques are based on the ‘smooth’ SPR transmission spectra curves. When the hybrid fiber optic sensor based on SPR and MMI is considered, the SPR transmission spectrum curve is no longer ‘smooth’ since it also contains the interference fringe patterns. Under this condition, the resonant wavelength is difficult to be located, and the traditional interrogation technique is no longer available. In order to solve this problem, we previously conducted a curve smoothing operation based on the least square method [10]. This method was able to obtain a clear resonant wavelength shift trend at the expense of loss of interference signal. Hence, the novel optimized interrogation technique to take advantage of both SPR signal and MMI signal should be evaluated systematically, which leads to the objective of this study.
This study mainly discusses the interrogation process of the hybrid fiber optic sensor based on SPR and MMI. The fast Fourier transform (FFT) algorithm is applied to the originally acquired transmission spectra, hence the SPR signal and the interference signal with different orders are separated effectively in the spatial frequency domain. Then, by using different filtering algorithms, both SPR signal and MMI signal can be processed individually for refractive index (RI) sensing, and therefore the overall RI sensitivity of the hybrid sensor is significantly improved. Moreover, when compared with authors’ previous work, only the visible region of spectrum is used for detection in this study. Hence the volume and the cost of sensing system are largely reduced since only one spectrometer is needed. Finally, this study of interrogation technique analysis also provides an effective method for noise filtering in the SPR data processing research fields.
2. Principles of interrogation technique
Before discussing the interrogation technique, we firstly present briefly the sensor structure, as indicated in Fig. 1. It consists of SMF-NCF-SMF structure with deposited gold film. The NCF is made of the high-precision pure silica with a diameter of 125 $\mathrm{\mu}\textrm{m}$. During the fabrication process, the polymer coating of the NCF fiber is firstly stripped and the fiber length of 12 mm is retained. Then, it is spliced between the two single mode fibers (SMF, 8 $\mathrm{\mu}\textrm{m}$/125 $\mathrm{\mu}\textrm{m}$) to form the cascaded MMI structure. Finally, a gold film layer with a thickness of 30 nm is deposited uniformly onto the surface of the NCF by the magnetron sputtering fiber coating device (SKY Technology Development Inc. JGP450A) to excite the SPR effect. In this study, we mainly evaluate the interrogation technique of the hybrid fiber optic sensor, the analyses of theoretical details such as length selection of NCF, self-image effect of MMI, etc. are not discussed.
When the light transmits through the interface between lead-in SMF and NCF, the higher-order Eigen-modes LP0n (n=1,2,3…) are excited. Part of the evanescent wave energy is absorbed by the metal layer and converted as the surface plasmon energy. Correspondingly, the SPR effect is excited. Then, when the light transmits through the interface between the NCF and the lead-out SMF, the MMI effect occurs between the core mode and the higher-order cladding modes. It should be explained that the MMI effect may be partially weakened due to the evanescent wave absorption by the deposited metal layer. However, considering that the thickness of the deposited gold film layer is only 30 nm which can be ignored, both SPR and MMI effects are integrated into a single NCF structure without enlarging the device size. Finally, the output transmission spectrum contains both SPR absorption dip and interference fringe patterns, this is why the optimized interrogation technique is discussed in this study.
Next, the interrogation technique principle is illustrated. In this study, the interrogation of SPR signal in the visible light band is evaluated, and the key problem is to separate the effective SPR signal from the interference signal. In order to solve this problem, the fast Fourier transform (FFT) algorithm is used and its principle is indicated in Fig. 2. Firstly, the experimentally acquired transmission spectrum curve contains both the SPR absorption dip and the interference fringe patterns with different orders, and it can be considered as a superposition of all these series of components, as indicated in the decomposition map of Fig. 2. Among them, the SPR resonance absorption dip can be assumed as the fundamental component, and the interference fringe patterns can be assumed as the higher-order components. Then, the FFT performs as a converter, it transforms the wavelength distribution into the spatial frequency distribution. After the FFT, the SPR signal is transformed as the lower-order spatial-frequency component, while the interference fringe patterns are transformed as the higher-order spatial frequency components. The higher the interference order is, the larger the spatial frequency is. The final spatial frequency distribution is the combinations of all these spatial frequency components. Thus, the SPR components and the MMI components are separated effectively in the spatial frequency domain. Finally, based on the low-pass filtering and the inverse FFT algorithm, the spectra of SPR components can be retained and processed individually for sensing. Correspondingly, based on the band-pass filtering and the inverse FFT algorithm, the spectra of interference signal with specific interference order can be also obtained for individual sensing.
3. Experimental setup
Figure 3 represents the diagram of the experimental setup. Firstly, the white light source (DH-2000-BAL, Ocean Optics Inc.) with bandwidth ranging from 400 nm to 1000 nm is selected. Then, the wide-spectrum light launches into the sensor and the transmission signals are collected by the spectrometer (USB 4000, Ocean Optics Inc.). Finally, the transmission spectra are processed on computer. In this study, a series of NaCl solutions with concentration ranging from 0% to 24% are prepared for the tests. The solution concentrations correspond to the RIs varying from 1.3342 to 1.3762, which are calibrated by WYA-2S digital Abbe refractometer (calibrated at 550 nm). During the experiments, the prepared sample is fixed into an analyte container, and the container is used to protect the sensor and also load the analyte solution. The solution is injected into the container from one of the inlets to immerse the sensor and then flowed out from the other. When the data acquisition process finishes, the container is then cleaned by the pure water until the spectrum returns to the initial state. For each experiment, three tests are conducted and the average values are saved for the following analyses. Concerning the pretreatment of the transmission spectra, the background noise spectrum is firstly filtered out from the original spectrum. Then, the processed spectrum is subtracted by the initial spectrum to obtain the normalized spectrum.
4. Results and discussions
Firstly, the experimentally acquired transmission spectra are indicated in Fig. 4. The obvious SPR resonance absorption dips are observed between 500 nm and 800 nm. However, we also observe that the curves are not ‘smooth’ and there exist numerous ‘burrs’ on the curves. When compared with the light source spectrum (inset), we confirm that these ‘burrs’ do not result from the light source and the connected fiber optic, but from the MMI of the SMF-NCF-SMF structured sensor. As explained in part 2, the interference fringe patterns in Fig. 4 are superposed by a large number of interference modes with different orders, these interference fringe patterns mix into the SPR absorption dip and make the transmission spectrum curves irregular. Hence, the traditional interrogation technique based on the resonant wavelength shift becomes unavailable since the resonant wavelengths are difficult to be located. In the following study, the interrogation technique will be discussed thoroughly for the signal demodulation of this hybrid sensor.
In the first step, FFT is applied to the raw transmission spectra, and the result of spatial frequency distribution with RI=1.3342 is shown in Fig. 5(a). From the analyses in part 2, we know that the low spatial frequency component results from the SPR absorption dip, and the rest higher-order spatial frequency components on the right side result from the interferences. Next, we perform the low-pass filtering on the spatial frequency domain, hence only SPR spatial frequency components are retained, as shown in Fig. 5(b). Finally, the inverse FFT is conducted to the processed spatial frequency components, and they turn back to the wavelength-distributed spectra, as shown in Fig. 6(a). It is seen that the interference fringe patterns are removed successfully from the raw spectra, only the SPR absorption dips are retained. Under this condition, the clear resonant wavelength shift trend is observed, and the RI sensitivity is estimated as 1037 nm/RIU with linearity of 0.9958.
Fig. 5. (a) Spatial frequency distribution of the spectrum after FFT. (b) Spatial frequency distribution of the SPR component after low-pass filtering (RI=1.3342).
Fig. 6. (a) SPR spectra after inverse FFT; (b) Relationship between RIs and wavelength shifts for the SPR components.
However, the interrogation of SPR signal is not enough to illustrate the advantages of this hybrid sensor. In the next step, the interrogation of MMI signal based on the same spectra is discussed. In Fig. 5(a), we observe four dominant higher-order spatial frequency components on the right side, which are defined as peak A, B, C and D. It is predictable that these components result from the interferences of the core mode LP01 and higher-order cladding modes LP0n. Besides, the four dominant amplitude peaks are located at the spatial frequencies of 0.044, 0.069, 0.088, and 0.115 nm−1 respectively, which correspond to the average wavelength spacing of 22.7, 14.5, 11.4 and 8.7 nm. Hence, the effective RI difference $\Delta {n_{eff}}$ between the core mode and the cladding mode can be estimated by the following equation [20]:
In the second step, we apply the band-pass filtering algorithm to retain only the interference components with the specific order, and then we perform the inverse FFT to obtain their corresponding wavelength-distributed spectra. In this step, the dominant peaks A, B, C in Fig. 5(a) are selected, and their interference spectra after inverse FFT are shown in Fig. 7. We observe that the interference fringe patterns become denser as the interference order increases, and their spatial frequencies are estimated as 0.047, 0.070, 0.090 nm−1 according to the fringe pattern numbers between 400 nm and 1000 nm, which are consistent with Fig. 5(a). Besides, the maximum interference amplitude is found for peak B (LP01-LP03 interferences) in the wavelength-distributed spectra, which also agrees well with the above analyses in Fig. 5(a). Finally, for each interference fringe pattern, it shows the maximum interference amplitude in the middle of the spectrum (500 nm-800 nm) and then decreases toward both sides, which is consistent with the SPR absorption dip. Actually, due to the novel design of the hybrid sensor, both SPR and MMI effects are excited in a single sensor structure. The MMI occurs due to the excitation of evanescent wave. Meanwhile, the evanescent wave is also absorbed by the deposited gold film layer. Hence, the final interference fringe amplitude is modulated inevitably by the SPR absorption dip.
In the third step, we select the LP01-LP02 interferences as an example, and the spectrum evolution of the interference fringe pattern with different RIs is indicated in Fig. 8(a). We observe that the interference fringe patterns keep a good profile along with RI variations. When compared with the SPR transmission spectra in Fig. 6(a), we can observe that the depth of SPR absorption dip becomes larger as RI varies from 1.3342 to 1.3762. Correspondingly, the interference fringe amplitude in Fig. 8(a) also increases within the same RI range. This phenomenon confirms again the previous discussion that the interference fringe amplitude is modulated by the SPR absorption dip. Besides, as discussed in Fig. 7, the middle of the spectrum shows the most significant interference signal, hence the two interference dips in the middle of the spectrum are selected and defined as dip a and dip b, and their relationships between the wavelength shifts and different RIs are shown in Fig. 8(b). The resonant wavelengths of the two interference dips show obvious redshift along with RI variations, and their corresponding RI sensitivities are 142 nm/RIU for dip a and 100 nm/RIU for dip b. As for the MMI, the RI sensitivity is determined by the effective RI difference $\Delta {n_{eff}}$ between the interference modes. On the other hand, $\Delta {n_{eff}}$ is affected by the external medium RI, and its variation trend is dependent of the operating wavelength. Hence, it shows different RI sensitivities between the different interference dips.
Fig. 8. (a)Spectra evolution of the LP01-LP02 interferences with different RIs. (b) Relationship between RIs and wavelength shifts.
From Figs. 6(a) and 8(a), it is obvious that both SPR effect and MMI effect contribute to the resonant wavelength shift of the transmission spectra, and the details are discussed as follows. We select the spectra of pure SPR and LP01-LP02 interferences (dominant peak A) as an example, and their wavelength shifts as RI varying from 1.3342 to 1.3762 are shown in Fig. 9. It is seen that both SPR and MMI show the significant redshift when the RI increases, two different effects are excited in a single NCF structure and then superpose in the output transmission spectrum. Although the RI sensitivities of SPR and MMI cannot be simply superimposed, the overall RI sensitivity of the proposed hybrid sensor is still significantly improved when compared with the single SPR or MMI sensor with the same device size. Moreover, only the visible region of spectrum is used for detection, and therefore only one spectrometer is needed. By using the interrogation technique in this study, the wavelength shift caused by these two different mechanisms can be separated for individual sensing.
Finally, the RI sensitivities of different interference orders are investigated quantitatively, and the results are indicated in Fig. 10. In Fig. 10(a), we observe that the peak A of dip a show the highest RI sensitivity of 142 nm/RIU, while the peak D of dip a show the lowest sensitivity of 74 nm/RIU. A similar phenomenon is also observed for dip b. Besides, the linearity coefficients of the fitting curve varies from 0.9584 to 0.9976. The linearity error mainly comes from the band-pass filtering algorithm that we use to demodulate the interference spectra with specific interference order. The selection of the band width, the starting and the cutoff spatial-frequency can influence slightly the shapes of the interference fringe patterns, which finally result in the error of linearity coefficient. A metrological study is necessary, and this leads to the further researches in the future.
Fig. 10. Relationship between RIs and wavelength shifts for different interference orders. (a) dip a. (b) dip b
In this study, the most significant advantage of the proposed interrogation method is that the SPR signal and the interference fringe patterns are separated effectively after FFT due to their different spatial frequency components, hence they can be processed individually for sensing. Both SPR and MMI effects are excited in a single NCF structure without enlarging the device size, and the overall RI sensitivity is significantly improved. Besides, the MMI signal can be used for temperature compensation of the general RI sensors. Moreover, by using the proposed interrogation method, the hybrid sensor can also realize the dual-parameter measurement other than RI sensing. Finally, in the SPR signal process research fields, the interrogation technique by using the FFT algorithm can be also used for noise filtering since the effective SPR signal and the noises usually represent the different spatial frequency components.
5. Conclusion
In summary, the interrogation technique of the hybrid fiber optic sensor based on SPR and MMI is mainly discussed. The SPR signal and MMI signal are separately effectively by using the FFT algorithm due to their different spatial frequency components, and therefore both SPR and MMI effects can be processed individually for RI sensing, which improves the overall RI sensitivity of the hybrid sensor. Besides, both SPR signal and MMI signal are observed in the visible region of spectrum, hence only one spectrometer is needed, which largely reduces the volume and the cost of multiplexed sensing system. Finally, this study of interrogation technique analyses also provides a useful method for noise filtering in the SPR signal processing field. This study provides an important supplement to the traditional SPR and MMI function.
Funding
National Natural Science Foundation of China (51808347); Natural Science Foundation of Shenzhen University (2019109, 860-000002110218).
Acknowledgments
We gratefully acknowledge Mr. Fan Lin (Shenzhen University) for the preparation of the experimental solutions.
Disclosures
The authors declare no conflicts of interest.
References
1. W. Zhang, W. Zhuang, M. Dong, L. Zhu, and F. Meng, “Dual-Parameter Optical Fiber Sensor for Temperature and Pressure Discrimination Featuring Cascaded Tapered-FBG and Ball-EFPI,” IEEE Sens. J. 19(14), 5645–5652 (2019). [CrossRef]
2. Y. Liu, D. Yang, Y. Wang, T. Zhang, M. Shao, D. Yu, H. Fu, and Z. Jia, “Fabrication of dual-parameter fiber-optic sensor by cascading FBG with FPI for simultaneous measurement of temperature and gas pressure,” Opt. Commun. 443, 166–171 (2019). [CrossRef]
3. W. Zhang, L. Huang, F. Gao, F. Bo, G. Zhang, and J. Xu, “All-fiber tunable Mach-Zehnder interferometer based on an acousto-optic tunable filter cascaded with a tapered fiber,” Opt. Commun. 292, 46–48 (2013). [CrossRef]
4. Q. Liu, Z. L. Ran, Y. J. Rao, S. C. Luo, H. Q. Yang, and Y. Huang, “Highly Integrated FP/FBG Sensor for Simultaneous Measurement of High Temperature and Strain,” IEEE Photonics Technol. Lett. 26(17), 1715–1717 (2014). [CrossRef]
5. Y. Jiang, D. Yang, Y. Yuan, J. Xu, D. Li, and J. Zhao, “Strain and high-temperature discrimination using a Type II fiber Bragg grating and a miniature fiber Fabry–Perot interferometer,” Appl. Opt. 55(23), 6341–6345 (2016). [CrossRef]
6. R. Yang, Y.-S. Yu, C.-C. Zhu, Y. Xue, C. Chen, X.-Y. Zhang, B.-L. Zhang, and H.-B. Sun, “PDMS-Coated S-Tapered Fiber for Highly Sensitive Measurements of Transverse Load and Temperature,” IEEE Sens. J. 15(6), 3429–3435 (2015). [CrossRef]
7. X. Gao, T. Ning, C. Zhang, J. Xu, J. Zheng, H. Lin, J. Li, L. Pei, and H. You, “A dual-parameter fiber sensor based on few-mode fiber and fiber Bragg grating for strain and temperature sensing,” Opt. Commun. 454, 124441 (2020). [CrossRef]
8. Y. Wang, Q. Huang, W. Zhu, M. Yang, and E. Lewis, “Novel optical fiber SPR temperature sensor based on MMF-PCF-MMF structure and gold-PDMS film,” Opt. Express 26(2), 1910 (2018). [CrossRef]
9. Y. Zhao, Q. Wu, and Y. Zhang, “Simultaneous measurement of salinity, temperature and pressure in seawater using optical fiber SPR sensor,” Measurement 148, 106792 (2019). [CrossRef]
10. Y. Duo, C. Yuzhi, G. Youfu, T. Fei, L. Yong, L. Xuejin, and H. Xueming, “Low crosstalk hybrid fiber optic sensor based on surface plasmon resonance and MMI,” Opt. Lett. 45(1), 117 (2020). [CrossRef]
11. Q. Wu, Y. Zhao, Y. Zhang, and Y. Yang, “High sensitive applied load measurement using optical fiber tapered-loop probe with SPR effect,” Opt. Laser Technol. 114, 95–102 (2019). [CrossRef]
12. Y. Saad, M. Selmi, M. H. Gazzah, A. Bajahzar, and H. Belmabrouk, “Performance enhancement of a copper-based optical fiber SPR sensor by the addition of an oxide layer,” Optik 190, 1–9 (2019). [CrossRef]
13. Y. Chen, Y. Yu, X. Li, H. Zhou, X. Hong, and Y. Geng, “Fiber-optic urine specific gravity sensor based on surface plasmon resonance,” Sens. Actuators, B 226, 412–418 (2016). [CrossRef]
14. Y. Chen, Q. Xie, X. Li, H. Zhou, X. Hong, and Y. Geng, “Experimental realization of D-shaped photonic crystal fiber SPR sensor,” J. Phys. D: Appl. Phys. 50(2), 025101 (2017). [CrossRef]
15. Z. Tan, X. Hao, Y. Shao, Y. Chen, X. Li, and P. Fan, “Phase modulation and structural effects in a D-shaped all-solid photonic crystal fiber surface plasmon resonance sensor,” Opt. Express 22(12), 15049 (2014). [CrossRef]
16. H. Moayyed, I. T. Leite, L. Coelho, J. L. Santos, and D. Viegas, “Analysis of Phase Interrogated SPR Fiber Optic Sensors With Bimetallic Layers,” IEEE Sens. J. 14(10), 3662–3668 (2014). [CrossRef]
17. R. Naraoka and K. Kajikawa, “Phase detection of surface plasmon resonance using rotating analyzer method,” Sens. Actuators, B 107(2), 952–956 (2005). [CrossRef]
18. E. Rodríguez-Schwendtner, A. González-Cano, N. Díaz-Herrera, M. C. Navarrete, and Ó. Esteban, “Signal processing in SPR fiber sensors: Some remarks and a new method,” Sens. Actuators, B 268, 150–156 (2018). [CrossRef]
19. W. Udos, K.-S. Lim, C.-L. Tan, M. N. S. M. Ismail, C.-W. Ooi, R. Zakaria, and H. Ahmad, “Spatial frequency spectrum of SPR-TFBG: A simple spectral analysis for in-situ refractometry,” Optik 219, 164970 (2020). [CrossRef]
20. Y. Geng, X. Li, X. Tan, Y. Deng, and Y. Yu, “High-Sensitivity Mach–Zehnder Interferometric Temperature Fiber Sensor Based on a Waist-Enlarged Fusion Bitaper,” IEEE Sens. J. 11(11), 2891–2894 (2011). [CrossRef]
21. W. Xu, J. Shi, X. Yang, D. Xu, F. Rong, J. Zhao, and J. Yao, “Improved Numerical Calculation of the Single-Mode-No-Core-Single-Mode Fiber Structure Using the Fields Far from Cutoff Approximation,” Sensors 17(10), 2240 (2017). [CrossRef]
