Highly sensitive refractive index sensor based on SPR with silver and titanium dioxide coating

A surface plasmon resonance (SPR) sensor based on dual-layered air hole shaped photonic crystal fiber (PCF) is proposed to realize the simultaneous measurement of refractive index (RI). The plasma materials silver (Ag) and titanium dioxide (TiO2) were sequentially coated on the outer surface of PCF to obtain enhanced sensing properties. By carefully adjusting the geometrical parameters, the simulation results show a maximum wavelength sensitivity of 72,000 nm/RIU for analyte refractive indices ranging from 1.26 to 1.365, which realizes the high-sensitivity sensing in the visible to near-infrared optical band. Moreover, the sensor attains a maximum figure of merit (FOM) of 229 and RI resolution of 1.29 × 10− 6. This work shows great potential for real-time, affordable, and accurate measurement in biomedical, biological and organic chemical domains.


Introduction
Over the past few years, surface plasmon resonance (SPR) sensing technology based on photonic crystal fibers (PCFs) has achieved notable applications in environmental monitoring, medical diagnostics, biochemical analysis, and food safety testing due to its outstanding performance and wide range of utilization (Cao et al. 2018;Liu et al. 2017b;Tong et al. 2018). The SPR can be characterized as the collective oscillation of free electrons at the metal-dielectric interface (Islam et al. 2018). Resonance occurs when the propagation constant of the incident electromagnetic photons matches the propagation constant of the surface electron oscillations, in which part of the energy of the incident light is transferred to the surface plasmons (SPs). The resonance conditions depend on the dielectric constant of the metal and the dielectric constant of the medium contacting it. When the dielectric 2 Structure and theoretical modeling In this paper, we employ finite element method (FEM) based COMSOL v5.4a software to conduct modeling and performance evaluation. The schematic of proposed Ag/TiO 2 coated PCF-SPR sensor is illustrated in Fig. 1(a). Upon a silica substrate, a symmetrical structure with a circular core and two layers of circular air holes is introduced. Differing diameters of air holes (d 1 and d 2 ) have a significant influence on optical mode coupling and propagation of light, which will be discussed in later sections. The air holes of diameter d 1 are designed to confine the energy of the core mode, makes the energy divided into four channels coupled to the SPP mode, thus the performance of sensor improved with the intense interaction between fundamental mode and plasmonic mode. The air holes of diameter d 2 are utilized to minimize confinement loss. A layer of silver film is assembled on the cladding surface. Then a TiO 2 film is attached to the silver film, and the additional TiO 2 film enhances the performance of the sensor compared to a mono-layer metal film. In our proposed sensor, the analyte is directly contacted with the external surface of the sensor, which avoids the inadequacies of internal sensing such as difficulty in cleaning and difficulty in ensuring uniform distribution of the analyte.
The refractive index of pure silica can be described by Sellmeier equation (Liu et al. 2017a): where n presents the refractive index of silica and λ stands for the wavelength of the incident light.
The RI of TiO 2 can be calculated as follows , where n presents the effective refractive index of TiO 2 and λ stands for the wavelength of the incident light. The confinement loss can be calculated using the imaginary part of the core mode complex RI by the following equation , where L indicates the confinement loss, λ means the incident wavelength, unit of micron. Im(n eff ) denotes the imaginary part of the complex refractive index.
The wavelength interrogation (WI) method is used to evaluate the performance of the proposed sensor, and thus the wavelength sensitivity can be calculated by the following equation , where Δλ peak stands for the shift in resonance peaks and Δn a represents for the variation of analyte RI respectively.
Resolution is also an important parameter that reflects the ability of the sensor to detect a small RI change of proposed sensor. The resolution of a sensor can be described by the following equation (Momota and Hasan 2018), where R indicates the sensor resolution, Δn a represents the variation of analyte RI, Δλ min represents the minimum wavelength resolution, assumed to be 0.1 nm, and Δλ peak represents the difference in resonance peak wavelength.
Note that, the overall performance of a sensor is defined in terms of figure of merit (FOM), which is the ratio of sensitivity to full width at half minima (FWHM), the FOM can be calculated by S/FWHM. Generally, to realize a high performance sensor the FOM should be as high as possible, which can be obtained with increasing sensitivity and decreasing FWHM.
The fabrication of the proposed sensor is not complicated. The different sized circular shaped air holes based in silica substrate can be accomplished by the stack-and-draw method (Mahdiraji et al. 2014;Knight 2003), and the stacked arrangement is illustrated Fig. 2(a). The 3D structure of the microstructured fiber obtained by the stack-and-draw method is shown in Fig. 2(b). Moreover, femtosecond laser etching ) and 3D printing (Ebendorff-Heidepriem et al. 2014) have also been applied to the fabrication of PCF most recently. Several convenient methods can be used of obtaining thin silver and TiO 2 layers on the outer surface of the sensor, such as chemical vapor deposition (CVD) and high-pressure microfluidic chemical deposition, and magnetron sputtering coating (Bayindir et al. 2004;Sazio et al. 2006;Zhang et al. 2007;Xie et al. 2017). Therefore, the manufacture of proposed sensor can be realized by using available fabrication technologies.

Investigation of sensor performance by optimizing different geometric parameters
Firstly, we analyzed the distribution of evanescent field, which indicated that there was a strong coupling between the fundamental mode and the SPP mode as shown in Fig. 1. At the resonance wavelength, Fig. 1(b-e) shows the field distributions at n a = 1.26 RIU, and Fig. 1(f-i) shows the field distributions n a = 1.36 RIU. Figure 1(f-g) shows the fundamental mode distribution, and Fig. 1(h-i) shows the SPP mode distribution. It can be seen that the SPP mode energy is higher in this case compared to n a = 1.26 RIU, indicating that the mode field is more strongly coupled at n a =1.36 RIU than at n a =1.26 RIU. Thus, stronger detection performance is achieved when the RI of the analyte is higher than 1.26 RIU. Since the proposed PCF-SPR sensor adopts symmetrical structure, the x polarization equals the y polarization to some degree which can be confirmed by the field distribution in Fig. 1, so the following discussion focuses on single one of the polarizations. The dispersion profile for fundamental core mode, plasmonic mode and confinement loss as a function of wavelength is shown in Fig. 3(a). The resonance loss curve obtains a sharp peak when the effective RI curves of core fundamental mode and SPP mode cross over, and the loss peak can be applied to detection of analytes effectively and easily. Figure 3(b) shows the confinement loss characteristics of the analyte RI as it changes from 1.26 RIU to 1.365 RIU. Notably, the refractive indices of a wide range of common inorganic and organic substances (e.g. water, methanol, ethanol, propanol, Fig. 2 a Stacked structure of the sensor. b 3D illustration of the sensor ether.) are in the range of 1.33 RIU to 1.365 RIU. Notice that, the characteristics of the loss curves for different RI analytes mainly determine the performance of the PCF-SPR sensor. Figure 3(b) demonstrates that with the increase of the analyte RI, the resonance loss peak shows red-shift and the loss peak height also increases obviously. Note that as the analyte RI changes slightly, the RI of the SPP mode changes as well while the core mode is unchanged, resulting in the phase matching point move to a longer wavelength . It was apparent that with increasing RI of the analyte, the shift of the resonance loss peak increases, meaning that higher sensing sensitivity can be accomplished. In addition, the maximum wavelength shift of 360 nm is obtained between the analyte RI of 1.36 RIU and 1.365 RIU. Thereby, the maximum wavelength sensitivity up to 72,000 nm/RIU can be obtained at 1.36. In the RI sensing range between 1.26 RIU and 1.365 RIU, the maximum resolution of the SPR sensor is 1.39 × 10 − 6 RIU with a 0.1 nm resolution of the spectrometer. Figure 3(c-d) shows the fitted plots of the resonance wavelength and FOM versus analyte refractive index, with an adjusted R-square of 0.99984 and 0.99606, respectively. The FOM obtained with different analytes is presented in Table 1, and it can be seen that the maximum FOM of 229 was reached at n a = 1.36 RIU.

Effect of core diameter on sensing performance
The influence of different core diameters at the fiber center on the sensing performance is investigated. Note that, in order to directly characterize the performance of the proposed sensor, we choose the RI of 1.36 RIU and 1.365 RIU as a reference. From Fig. 4(a) it can be noticed that the resonance peak with an RI of 1.365 RIU changes significantly when d changes. Overall, for d = 0.18 µm, the wavelength sensitivity achieves its maximum and the loss peak appears to be sharper. Thus, we choose d = 0.18 µm as optimum diameter of circular shaped air hole in the core.

Effect of cladding air hole diameter d 1 and d 2 on sensing performance
To obtain the optimal diameter of the cladding air holes, we carefully adjusted d 1 and d 2 to obtain the respective spectra as shown in Fig. 4(b and c). It is observed that there is only a slight effect of altering d 1 on the wavelength sensitivity. It is due to the introduction of the large bubbles with d 1 diameter in order to split the core mode into four channels, thus having a little effect on the coupling between the Core mode and SPP mode. As the d 2 varies, we observed from the Fig. 4(c) that the shift of the resonance loss peak is quite large and Fig. 4 a Confinement loss spectra at different diameter of core. b and c Confinement loss spectra for different air hole diameter d 1 , d 2 the change in wavelength sensitivity is also significant. The reason is that the air holes with d 2 diameter is introduced to reduce the confinement loss, which directly affects the energy coupling from the core mode to the SPP mode, and thus the size of d 2 has a greater influence on the loss than d 1 . Note that, although a larger wavelength shift is achieved when d 2 = 0.98 µm but the resonance loss is excessively high, so the performance is more acceptable for d 2 = 1.0 µm. Therefore, we choose d 1 = 1.5 µm, d 2 = 1.0 µm as the optimal air hole diameters.

Effect of pitch distance Λ 1 and Λ 2 on sensing performance
We also investigated the impact of the pitch distances between air holes. As shown in Fig. 5, with the increase of Λ 1 and Λ 2 , the loss peak with refractive index at 1.36 is regularly blue-shifted, while the loss peak with refractive index at 1.365 is more variable. After careful tuning, the wavelength sensitivity obtains a maximum when Λ 1 = 1.25 µm and Λ 2 = 1.80 µm, and the loss peak is sharper at the same time.

Effect of silver and TiO 2 thickness on sensing performance
The thickness of the silver and TiO 2 layers plays a critical role in the sensing performance. Some reports indicate that when using a double-layer composite as plasmonic material, a better result can be obtained with a total thickness of approximately 50 nm. Therefore, in order to thoroughly study the influence of the thickness ratio of silver and TiO 2 on the sensing characteristics, the thickness of the silver film and the TiO 2 film are set at 10 nm/40 nm, 30 nm/20 nm, 20 nm/30 nm, 10 nm/40 nm at first. The results are shown in Fig. 6(a). The resonance loss peak differed significantly for different ratios of silver to TiO 2 layers, which should be due to the reason that either an overly thick silver layer or an overly thick TiO 2 would hinder the energy transfer from the fiber core mode to the SPP mode. The relatively optimal wavelength sensitivity was obtained when Ag/TiO2 = 20 nm/30 nm. Therefore, the loss curves obtained by adjusting Ag when TiO 2 is 30 nm and by adjusting TiO 2 when Ag is 20 nm are investigated respectively, the results are shown in Fig. 6(b and c). When the thickness of Ag stays at 20 nm, the resonance peak is red-shifted, and the Fig. 5 a and b Confinement loss spectra with varying pitch distance Λ 1 , Λ 2 peak loss also decreased as the thickness of TiO 2 increases. Similarly, when the thickness of TiO 2 stays at 30 nm, the resonance loss peaks also appeared red-shifted variously with the increase of the Ag film thickness. From the above, we finally chosen A g = 20 nm and TiO 2 = 28 nm as the optimal bilayer structure parameters.

Effect of analyte thickness on sensing performance
We also investigated the effect of the analyte thickness on the sensing performance. The analyte thickness was set to 1 µm, 1.5 µm, 2 µm, 2.5 µm, and 3 µm, respectively, and the loss spectrum was shown in Fig. 6(d). It can be noticed that when T a is greater than or equal to 2 µm, the resonance peak changes slightly, whereas when it is less than 2 µm, the resonance peak changes significantly due to the insufficient energy coupling between core mode and SPP mode. Therefore, we choose T a = 2.0 µm as the optimal parameter.
A performance comparison of the proposed set of PCF-SPR sensors with prior sensors is presented in Table 1. Fig. 6 a Confinement loss spectra at different ratio of Ag and TiO 2 film thickness. b Confinement loss spectra for different Ag film thickness at TiO 2 film thickness = 30 nm. c Confinement loss spectra for different TiO 2 film thickness at Ag film thickness = 20 nm. d Confinement loss spectra at different analyte channel thickness variation

Conclusions
In this paper, we propose a symmetrical dual-layer air-hole arranged photonic crystal fiber SPR sensor coated with silver and TiO 2 externally. The thickness of the silver and titanium dioxide coatings has a significant impact on the sensor performance. Detailed numerical simulation of the sensor shows a maximum wavelength sensitivity of 72,000 nm/RIU and a maximum FOM of 229 RIU − 1 within the RI detection range between 1.26 RIU and 1.365 RIU. The sensor can be easily fabricated using existing technology. It is also convenient to conduct relevant practical applications due to the external sensing and symmetrical structure adopted. Thus, the proposed sensor demonstrates huge potential in biomedical and chemical fields and can be utilized for the accurate and precise detection in biomedicine and chemistry.

Conflict of interest
The authors declare no conflicts of interest.