Hg 2 + Optical Fiber Sensor Based on LSPR with PDDA-Templated AuNPs and CS / PAA Bilayers

: An optical ﬁber localized surface plasmon resonance (LSPR) sensor was proposed and experimentally demonstrated to detect Hg 2 + ions by functionalizing the optical ﬁber surface with gold nanoparticles (AuNPs) and chitosan (CS) / poly acrylic acid (PAA) bilayers. A flame-brushing technology was proposed to post-process the polydimethyl diallyl ammonium chloride(PDDA)-templated nanoparticles, avoiding the aggregation of AuNPs and achieving well-dispersed AuNPs arrays. LSPR stimulated by the AuNPs is sensitive to changes in the refractive index induced by Hg 2 + ions absorption on the CS / PAA bilayers. Experimental results demonstrated that the LSPR peak wavelength linearly shifts with the concentrations of Hg 2 + ions from 1 to 30 µ M with a sensitivity of around 0.51 nm / ppm. The sensor also exhibits good speciﬁcity and longtime stability.


Introduction
Mercury ions (Hg 2+ ) are one of the main sources of heavy metal pollution in water. Contamination by Hg 2+ , not only causes serious environmental problems, but also central nervous system disorders, kidney or liver disorders due to the toxicity and carcinogenicity [1]. Various sensing technologies have been developed for Hg 2+ detection in an aqueous solution. Spectroscopic-based techniques, such as atomic absorption spectroscopy [2] and inductively coupled plasma spectroscopy [3], are amongst the most commonly used methods and have accurate Hg 2+ detection capability. However, these spectroscopic techniques have the same drawback of high cost due to the maintenance required for this sophisticated equipment. Additionally, the sample preparation process makes these methods time consuming. All of these undoubtedly urge the generation of new sensing technologies with simplicity, rapidity, low cost, and high sensitivity.
Optical fiber sensing technology has attracted extensive attention and great research enthusiasm in the field of metal ion monitoring due to its unique advances such as low cost, light weight, ease of use, small in size, multiplexing, remote sensing, and high sensitivity [4][5][6][7]. Up to now, optical fiber heavy metal ion sensors have been reported by utilizing fluorescence [8], surface-enhanced Raman scattering [9], surface plasmon resonance [10][11][12], and localized surface plasmon resonance (LSPR) [13][14][15][16][17]. LSPR is usually excited near the surface of the metal nanoparticles to enhance the local electromagnetic fields, and the basic principle of heavy metal ion detection based on LPSR is highly responsive to changes in the refractive index around metal nanoparticles [18,19]. In view of enhancing the plasmon resonance and the sensitivity of the refractive index sensing, the deposition of well-dispersed metallic nanoparticles is extremely important to realize LSPR sensor with high sensitivity and good reproducibility. Colloidal gold nanoparticles (AuNPs) are commonly deposited on an optical fiber surface through a substrate via the electrostatic self-assembly method. The first reported deposition method was based on positively charged silane to assemble negatively charged nanoparticles [20]. This method usually requires a complex operation and lengthy deposition time. For instance, a total of 13 h was required when depositing AuNPs on the hydroxylated optical fibers through (3-mercaptopropyl) trimethoxy silane [21]. In addition to a lengthy deposition time, a large number of aggregates and low nanoparticle coverage were also observed in the silence method [22,23]. In order to reduce the deposition time, a polyelectrolyte assembly method was then proposed to simplify the preparation of optical fiber LSPR sensors [24][25][26]. At high nanoparticle coverage, this method was also reported to have aggregation [27]. A clear aggregation was also observed as directly depositing the AuNPs onto the surface of poly (methyl methacrylate) chips and optical fibers [28]. The aggregation of nanoparticles broadens the LSPR band and hence strongly affects the refractive index resolution, sensitivity, and experimental reproducibility [29].
In this paper, we proposed and experimentally demonstrated a heavy metal Hg 2+ sensor based on the optical fiber LSPR. We developed an alternative route to deposit well-dispersed AuNPs on bare optical fiber by fast flame-brushing the PDDA-templated nanoparticles. Chitosan (CS)/poly acrylic acid (PAA) multilayer film was employed to functionalize the optical fiber surface and then to facilitate Hg 2+ ions adsorption on the sensor for detection. LSPR spectrum shifts were monitored to characterize the sensor performance with various Hg 2+ concentrations.

Synthesis Method of the AuNPs
The AuNPs used in the experiments were prepared by the reduction of chloroauric acid using sodium citrate [30]. One-hundred milliliters of chloroauric acid solution (0.3 mM) stored in a flask was put into an oil bath at 105 • C. After the solution boiled, 2 mL of sodium citrate solution (1 wt%) was rapidly injected into the flask under vigorous stirring. The solution turned from transparent to purple and then turned wine red in about five minutes. After stirring for 30 min, the flask was taken out, and the mixture was stirred continuously until its temperature returned to room temperature. Finally, the formed wine-red AuNPs solution was preserved in a refrigerator at 4 • C for further use.

Preprocessing of the Optical Fiber
The hard polymer cladding multimode optical fiber (HP2140-A from YOFC) has a step-index profile and a numerical aperture of 0.37. The diameters of the core, cladding, and coating are 200, 230, and 500 µm, respectively. The materials of the three layers were pure silica glass, fluoroacrylate, and ethylene tetrafluoroethylene plastic, respectively. A section of the optical fiber was first intercepted at 20 cm long, and then the coating of the middle section was mechanically removed for approximately 2 cm. The removal of the hard polymer cladding was realized by immersing the middle section in acetone for around 90 min. Subsequently, this optical fiber segment was immersed for 30 min in piranha solution to hydroxylate the optical fiber surface that could remain at 80 • C.

Functionalization of the Optical Fiber
To fabricate the LSPR sensor, eight steps are required to functionalize the optical fiber, as shown in Figure 1. The bare middle section of the multimode fiber was hydroxylated by immersion in piranha solution (sulfuric acid and hydrogen peroxide) in a 7:3 volume ratio at 80 • C for half an hour, followed by thorough rinsing with deionized water and drying in a nitrogen flow. The fiber was placed in a vacuum drying oven at 110 • C for one hour, making the hydroxyl group more active. The prepared fiber was first immersed in PDDA solution (30 wt%) for eight hours and then in aqueous AuNP solution for 90 min. The AuNPs were successfully adsorbed onto the PDDA nanosheet with electrostatic interaction. After that, the surface of the fiber was rinsed with ultrapure water to remove the excess and unbounded AuNPs and was then baked in an oven at 60 • C for 2 h. The flame brushing technique was used to implement the high-temperature annealing process. Only the sensing section of the fiber was moved quickly over the flame of the alcohol lamp; thus, the AuNPs upon optical fiber substrate was clean. Then, we used 3-MPA as a crosslinker, which included both the terminated thiol and -COOH groups. The removal of the citrate in the flame brushing provides enough free space for the sulfhydryl group, and hence the sulfhydryl linking between AuNPs and 3-MPA becomes easier and stronger. We then used EDC/NHS to activate the other terminated -COOH groups of the 3-MPA. Finally, we loaded the chitosan (CS)/PAA bilayers on optical fiber by using the layer-by-layer self-assembly method. First, the treated fiber was immersed in CS solution for 7 min at PH = 2.0. Second, poly acrylic acid (PH = 5) was linked onto the chitosan through electrostatic adsorption between them. After each step, the fiber was rinsed with ultrapure water and dried in the vacuum drying oven to remove the CS or PAA that failed to absorb on the substrate. The first and second processes were repeated 4 times to form four CS/PAA bilayers, and to finish the final sensor device.
Appl. Sci. 2020, 10, x FOR PEER REVIEW 3 of 10 middle section in acetone for around 90 min. Subsequently, this optical fiber segment was immersed for 30 min in piranha solution to hydroxylate the optical fiber surface that could remain at 80 °C.

Functionalization of the Optical Fiber
To fabricate the LSPR sensor, eight steps are required to functionalize the optical fiber, as shown in Figure 1. The bare middle section of the multimode fiber was hydroxylated by immersion in piranha solution (sulfuric acid and hydrogen peroxide) in a 7:3 volume ratio at 80 °C for half an hour, followed by thorough rinsing with deionized water and drying in a nitrogen flow. The fiber was placed in a vacuum drying oven at 110 °C for one hour, making the hydroxyl group more active. The prepared fiber was first immersed in PDDA solution (30 wt%) for eight hours and then in aqueous AuNP solution for 90 min. The AuNPs were successfully adsorbed onto the PDDA nanosheet with electrostatic interaction. After that, the surface of the fiber was rinsed with ultrapure water to remove the excess and unbounded AuNPs and was then baked in an oven at 60 °C for 2 h. The flame brushing technique was used to implement the high-temperature annealing process. Only the sensing section of the fiber was moved quickly over the flame of the alcohol lamp; thus, the AuNPs upon optical fiber substrate was clean. Then, we used 3-MPA as a crosslinker, which included both the terminated thiol and -COOH groups. The removal of the citrate in the flame brushing provides enough free space for the sulfhydryl group, and hence the sulfhydryl linking between AuNPs and 3-MPA becomes easier and stronger. We then used EDC/NHS to activate the other terminated -COOH groups of the 3-MPA. Finally, we loaded the chitosan (CS)/PAA bilayers on optical fiber by using the layer-by-layer selfassembly method. First, the treated fiber was immersed in CS solution for 7 min at PH = 2.0. Second, poly acrylic acid (PH = 5) was linked onto the chitosan through electrostatic adsorption between them. After each step, the fiber was rinsed with ultrapure water and dried in the vacuum drying oven to remove the CS or PAA that failed to absorb on the substrate. The first and second processes were repeated 4 times to form four CS/PAA bilayers, and to finish the final sensor device.  Figure 2 illustrates the experimental setup for the detection of the refractive index and the heavy metal. To connect the sensing fiber into the optical test system, both ends of the sensing fiber were first ground with 600 mesh and 2000 mesh abrasive papers successively and then were inserted into the fiber adapters with an inner diameter of 250 µm, and finally were terminated with the fiber jumpers using the temporary FC-SMA connectors. This method is simple and easy to use, but the temporary connection is not stable, and the insertion loss is large, and hence a high-power light source is usually required. Alternately, another method was also investigated to obtain a strong connection with low insertion loss by fusion splicing the sensor fiber with two fiber jumpers. Both the jumper fibers and sensing fibers were first cut using a fiber cleaver (CT50, Fujikura, Japan) to obtain smooth and flat ends, and then both ends were spliced using a splicer (Fujikura 80s) with a larger arc current and longer arc duration compared with the case of splicing the standard telecom single-mode fiber. The lead-in jumper was connected with an unpolarized white light source (HL1000 with halogen lamp, Wyoptics, Shanghai, China), and the lead-out jumper was connected with the charge coupled device (CCD) spectrometer (USB 4000, Ocean Optics, US) with a wavelength range from 350 to 1000 nm. The LSPR absorption spectrum was monitored an average of 10 times. Before the measurement, the background signal was recorded as the basis by connecting a bare (not Au NP-coated) fiber instead of the LSPR sensor.

Experimental Setup for the Detection of the Refractive Index and the Heavy Metal
Appl. Sci. 2020, 10, x FOR PEER REVIEW 4 of 10 Figure 2 illustrates the experimental setup for the detection of the refractive index and the heavy metal. To connect the sensing fiber into the optical test system, both ends of the sensing fiber were first ground with 600 mesh and 2000 mesh abrasive papers successively and then were inserted into the fiber adapters with an inner diameter of 250 m, and finally were terminated with the fiber jumpers using the temporary FC-SMA connectors. This method is simple and easy to use, but the temporary connection is not stable, and the insertion loss is large, and hence a high-power light source is usually required. Alternately, another method was also investigated to obtain a strong connection with low insertion loss by fusion splicing the sensor fiber with two fiber jumpers. Both the jumper fibers and sensing fibers were first cut using a fiber cleaver (CT50, Fujikura, Japan) to obtain smooth and flat ends, and then both ends were spliced using a splicer (Fujikura 80s) with a larger arc current and longer arc duration compared with the case of splicing the standard telecom single-mode fiber. The lead-in jumper was connected with an unpolarized white light source (HL1000 with halogen lamp, Wyoptics, Shanghai, China), and the lead-out jumper was connected with the charge coupled device (CCD) spectrometer (USB 4000, Ocean Optics, US) with a wavelength range from 350 to 1000 nm. The LSPR absorption spectrum was monitored an average of 10 times. Before the measurement, the background signal was recorded as the basis by connecting a bare (not Au NPcoated) fiber instead of the LSPR sensor.  Figure 3 illustrated the evolution of the absorption spectrum when PDDA-coated optical fiber was immersed in the Au colloid solution for 90 min. In the beginning, the PDDA nanosheet had abundant positive charges, and hence easily absorbed negatively charged AuNPs on the surface. Therefore, a weak LSPR peak is immediately observed as long as the optical fiber is inserted into the Au colloid solution, and further, the absorption peak grows rapidly in the first few minutes. At 5 min, the absorption peak arrived at 0.24 at the peak wavelength of 528 nm. The location of the absorption peak is determined by the particle size and shape. At this stage, the AuNPs are isolated and devoid of interactions with neighboring particles. As time passed, the AuNPs were packed closer and began to form multilayers and aggregation of the surface. When the spacing between particles is small compared to the wavelength of light, the coupling of surface plasmons between closely spaced particles became increasingly significant. The aggregation effect generates a new red-shifted feature in the optical spectrum centered between 600 and 800 nm [31]. As shown in Figure 3, the first LPSR peak continued to grow, but a new LPSR band over 600 nm appeared at 10 minutes, and gradually became prominent. After 50 min, the intensities of both absorption peaks increased slowly and finally reached stability around 90 min.

Characterization of the Sensor Device
The aggregation of the AuNPs broadened the absorption spectrum, and furthermore, the existence of organic stabilizers and linkers made the surface unclean and limited its application. The flame brushing technique with the alcohol lamp provided a small heat range to heat only the 3-cm sensor area without destroying the coating protection of the adjacent area, Existence of the fiber LSPR sensor  Figure 3 illustrated the evolution of the absorption spectrum when PDDA-coated optical fiber was immersed in the Au colloid solution for 90 min. In the beginning, the PDDA nanosheet had abundant positive charges, and hence easily absorbed negatively charged AuNPs on the surface. Therefore, a weak LSPR peak is immediately observed as long as the optical fiber is inserted into the Au colloid solution, and further, the absorption peak grows rapidly in the first few minutes. At 5 min, the absorption peak arrived at 0.24 at the peak wavelength of 528 nm. The location of the absorption peak is determined by the particle size and shape. At this stage, the AuNPs are isolated and devoid of interactions with neighboring particles. As time passed, the AuNPs were packed closer and began to form multilayers and aggregation of the surface. When the spacing between particles is small compared to the wavelength of light, the coupling of surface plasmons between closely spaced particles became increasingly significant. The aggregation effect generates a new red-shifted feature in the optical spectrum centered between 600 and 800 nm [31]. As shown in Figure 3, the first LPSR peak continued to grow, but a new LPSR band over 600 nm appeared at 10 minutes, and gradually became prominent. After 50 min, the intensities of both absorption peaks increased slowly and finally reached stability around 90 min.

Characterization of the Sensor Device
The aggregation of the AuNPs broadened the absorption spectrum, and furthermore, the existence of organic stabilizers and linkers made the surface unclean and limited its application. The flame brushing technique with the alcohol lamp provided a small heat range to heat only the 3-cm sensor area without destroying the coating protection of the adjacent area, Existence of the fiber coating made the sensor structure robust. The organic stabilizers and linkers were burnt out in the high-temperature for cleaning the surface of AuNPs. Due to the high temperature, two or more AuNPs began to coalesce and form large particles [32]. Figure 4 illustrates the SEM images of the AuNPs before and after flame brushing, which clearly show that the AuNPs with smooth surfaces were well dispersed on the optical fiber substrate after flame brushing. Moreover, the LSPR absorption spectrum has only one peak located around 542.7 nm (Figure 3b).
Membranes with a layer-by-layer structure were formed by sequential depositions of CS and PAA alternately. After each layer, the LSPR spectral response of the sensor was monitored, as shown in Figure 3b. It is clearly observed that there is a red shift of the LSPR absorption peak after each layer functionalization process, indicating the CS/PAA bilayers was successfully immobilized on the sensor surface, and hence changes the effective refractive index.
Appl. Sci. 2020, 10, x FOR PEER REVIEW 5 of 10 coating made the sensor structure robust. The organic stabilizers and linkers were burnt out in the high-temperature for cleaning the surface of AuNPs. Due to the high temperature, two or more AuNPs began to coalesce and form large particles [32]. Figure 4 illustrates the SEM images of the AuNPs before and after flame brushing, which clearly show that the AuNPs with smooth surfaces were well dispersed on the optical fiber substrate after flame brushing. Moreover, the LSPR absorption spectrum has only one peak located around 542.7 nm (Figure3b).
Membranes with a layer-by-layer structure were formed by sequential depositions of CS and PAA alternately. After each layer, the LSPR spectral response of the sensor was monitored, as shown in Figure 3b. It is clearly observed that there is a red shift of the LSPR absorption peak after each layer functionalization process, indicating the CS/PAA bilayers was successfully immobilized on the sensor surface, and hence changes the effective refractive index.

Refractive Index Response
The refractive index sensing performance was tested by immersing the sensor probe in a series of sucrose solutions with different concentrations of 4.8 wt%, 11.4 wt%, 18.00 wt%, 24.00 wt%, 29.00 wt%, and 35.00 wt%. The external refractive index of the sucrose solutions can be estimated to be 1. 3400, 1.3500, 1.3605, 1.3706, 1.3794, and 1.3903, according to the empirical formula at room temperature [33,34]. Figure 5a,b illustrates the spectra evolution and the wavelength shift of the LSPR absorption peak. The results exhibit a nonlinear response, and the sensitivity is around 257.8 and Appl. Sci. 2020, 10, x FOR PEER REVIEW 5 of 10 coating made the sensor structure robust. The organic stabilizers and linkers were burnt out in the high-temperature for cleaning the surface of AuNPs. Due to the high temperature, two or more AuNPs began to coalesce and form large particles [32]. Figure 4 illustrates the SEM images of the AuNPs before and after flame brushing, which clearly show that the AuNPs with smooth surfaces were well dispersed on the optical fiber substrate after flame brushing. Moreover, the LSPR absorption spectrum has only one peak located around 542.7 nm (Figure3b).
Membranes with a layer-by-layer structure were formed by sequential depositions of CS and PAA alternately. After each layer, the LSPR spectral response of the sensor was monitored, as shown in Figure 3b. It is clearly observed that there is a red shift of the LSPR absorption peak after each layer functionalization process, indicating the CS/PAA bilayers was successfully immobilized on the sensor surface, and hence changes the effective refractive index.

Refractive Index Response
The refractive index sensing performance was tested by immersing the sensor probe in a series of sucrose solutions with different concentrations of 4.8 wt%, 11.4 wt%, 18.00 wt%, 24.00 wt%, 29.00 wt%, and 35.00 wt%. The external refractive index of the sucrose solutions can be estimated to be 1. 3400, 1.3500, 1.3605, 1.3706, 1.3794, and 1.3903, according to the empirical formula at room temperature [33,34]. Figure 5a,b illustrates the spectra evolution and the wavelength shift of the LSPR absorption peak. The results exhibit a nonlinear response, and the sensitivity is around 257.8 and

Refractive Index Response
The refractive index sensing performance was tested by immersing the sensor probe in a series of sucrose solutions with different concentrations of 4.8 wt%, 11.4 wt%, 18.00 wt%, 24.00 wt%, 29.00 wt%, and 35.00 wt%. The external refractive index of the sucrose solutions can be estimated to be 1.3400, 1.3500, 1.3605, 1.3706, 1.3794, and 1.3903, according to the empirical formula at room temperature [33,34]. Figure 5a,b illustrates the spectra evolution and the wavelength shift of the LSPR absorption peak. The results exhibit a nonlinear response, and the sensitivity is around 257.8 and 821.3 nm/RIU and at the RIU of 1.34 and 1.39, respectively. The full width at half maximum (FWHM) of the sensor is around 93 nm, and hence the Q-factor is calculated to be the ratio between the resonance wavelength and the FWHM, i.e., 6.1. The figure of merit (FOM) can be calculated as the quotient between the FWHM and the sensitivity, and hence the FOM of the sensor is expected to be 2.7 to 8.8 RIU −1 .
Appl. Sci. 2020, 10, x FOR PEER REVIEW 6 of 10 821.3 nm/RIU and at the RIU of 1.34 and 1.39, respectively. The full width at half maximum (FWHM) of the sensor is around 93 nm, and hence the Q-factor is calculated to be the ratio between the resonance wavelength and the FWHM, i.e., 6.1. The figure of merit (FOM) can be calculated as the quotient between the FWHM and the sensitivity, and hence the FOM of the sensor is expected to be 2.7 to 8.8 RIU −1 .

Hg 2+ Detection
Before Hg 2+ detection, the LSPR spectrum of the blank sample was obtained by immersing the prepared sensor probe in the ultra-pure water for 10 min and then drying it in a vacuum drying oven at 60 °C . In the Hg 2+ measurement, Hg 2+ can be absorbed by the amino (-NH2) and hydroxyl (-OH) groups in the CS to form stable chelates, and also be chelated with a large number of carboxyl (-COOH) groups in the PAA. This series of reactions will increase the degree of cross-linking between the two-layer sensing materials, and accordingly, increase the refractive index of the sensing materials. According to refractive index sensing in Figure 5b, a redshift of the LSPR peak is expected when increasing the concentrations of Hg 2+ solution. The LSPR spectrum with different Hg 2+ concentration was monitored when the sensor probe was immersed into the Hg 2+ solution for 10 min and then removed from the Hg 2+ solution and afterward dried completely in the oven. When the sensor probe was removed from the Hg 2+ solution, some Hg 2+ ions were immobilized by the use of the CS/PAA film, and hence the refractive index changed due to the Hg 2+ contact. The spectral measurement in the air instead of the solution has been proven to be effective [14]. Moreover, this method can maximally exclude the influence of solution salinity and PH because the sensor probe has been dried completely before measuring the LSPR spectrum. After each measurement, the sensor probe soaked in HCl solution at PH = 3 for 10 min until the resonant wavelength recovered to the original wavelength. In the acid solution, the -NH2 and -COOH groups in the sensor layer material were protonated, resulting in the desorption of the adsorbed Hg 2+ , thus ensuring the sensor can be reused. Figure 6a illustrates the normalized LSPR spectra of the blank sample and the Hg 2+ samples with different concentrations. The change in position of LSPR peak is quite evident. Figure 6b illustrates the wavelength shift of the absorption peak relative to the blank reference. The absorption peak wavelength shifted by 17.6 nm when the Hg 2+ concentration increased from blank to 30 M. The linear fitting of the peak response indicates the proposed sensor has a sensitivity of around 0.51 nm/M. When the concentrate increased to 50 M, the sensitivity began to decrease, and the wavelength shift seems to be saturated. A limit of detection (LOD) was calculated to be 0.52 M (3S) based on three times the standard deviation (SD) of measured signals of blank samples.

Hg 2+ Detection
Before Hg 2+ detection, the LSPR spectrum of the blank sample was obtained by immersing the prepared sensor probe in the ultra-pure water for 10 min and then drying it in a vacuum drying oven at 60 • C. In the Hg 2+ measurement, Hg 2+ can be absorbed by the amino (-NH2) and hydroxyl (-OH) groups in the CS to form stable chelates, and also be chelated with a large number of carboxyl (-COOH) groups in the PAA. This series of reactions will increase the degree of cross-linking between the two-layer sensing materials, and accordingly, increase the refractive index of the sensing materials. According to refractive index sensing in Figure 5b, a redshift of the LSPR peak is expected when increasing the concentrations of Hg 2+ solution. The LSPR spectrum with different Hg 2+ concentration was monitored when the sensor probe was immersed into the Hg 2+ solution for 10 min and then removed from the Hg 2+ solution and afterward dried completely in the oven. When the sensor probe was removed from the Hg 2+ solution, some Hg 2+ ions were immobilized by the use of the CS/PAA film, and hence the refractive index changed due to the Hg 2+ contact. The spectral measurement in the air instead of the solution has been proven to be effective [14]. Moreover, this method can maximally exclude the influence of solution salinity and PH because the sensor probe has been dried completely before measuring the LSPR spectrum. After each measurement, the sensor probe soaked in HCl solution at PH = 3 for 10 min until the resonant wavelength recovered to the original wavelength. In the acid solution, the -NH2 and -COOH groups in the sensor layer material were protonated, resulting in the desorption of the adsorbed Hg 2+ , thus ensuring the sensor can be reused. Figure 6a illustrates the normalized LSPR spectra of the blank sample and the Hg 2+ samples with different concentrations. The change in position of LSPR peak is quite evident. Figure 6b illustrates the wavelength shift of the absorption peak relative to the blank reference. The absorption peak wavelength shifted by 17.6 nm when the Hg 2+ concentration increased from blank to 30 µM. The linear fitting of the peak response indicates the proposed sensor has a sensitivity of around 0.51 nm/µM. When the concentrate increased to 50 µM, the sensitivity began to decrease, and the wavelength shift seems to be saturated. A limit of detection (LOD) was calculated to be 0.52 µM (3S) based on three times the standard deviation (SD) of measured signals of blank samples.

Specificity
The specificity of heavy metal detection is important for the practical water quality monitoring in the real aqueous solution, which contains various heavy metal ions. In order to study the specificity of the heavy metal sensor, the proposed LSPR sensor was experimentally employed to test the spectral response to K + , Na + , Ca 2+ , and Cu 2+ solutions besides the Hg 2+ solution. Figure 7a illustrates the wavelength shift of the above metal solution with a concentration of 30 M. It is clear that the wavelength shift of Hg 2+ is much larger than those of other ions.

Stability
To investigate the stability of the optical fiber sensor, the proposed LSPR sensor was directly exposed to air for two weeks at room temperature. Figure 7b compares the LSPR absorption spectra of the fresh sensor and the one exposed to air for two weeks. It appears that both the central wavelength and peak intensity of the resonance barely changed except for a tiny drift of the baseline. This result indicates the proposed LSPR sensor has relative long-term durability under normal storage conditions. Due to good stability in the air, it is suggested to employ the sensor when a measurement is required instead of constantly using the sensor in water for a long time.

Specificity
The specificity of heavy metal detection is important for the practical water quality monitoring in the real aqueous solution, which contains various heavy metal ions. In order to study the specificity of the heavy metal sensor, the proposed LSPR sensor was experimentally employed to test the spectral response to K + , Na + , Ca 2+ , and Cu 2+ solutions besides the Hg 2+ solution. Figure 7a illustrates the wavelength shift of the above metal solution with a concentration of 30 µM. It is clear that the wavelength shift of Hg 2+ is much larger than those of other ions.

Specificity
The specificity of heavy metal detection is important for the practical water quality monitoring in the real aqueous solution, which contains various heavy metal ions. In order to study the specificity of the heavy metal sensor, the proposed LSPR sensor was experimentally employed to test the spectral response to K + , Na + , Ca 2+ , and Cu 2+ solutions besides the Hg 2+ solution. Figure 7a illustrates the wavelength shift of the above metal solution with a concentration of 30 M. It is clear that the wavelength shift of Hg 2+ is much larger than those of other ions.

Stability
To investigate the stability of the optical fiber sensor, the proposed LSPR sensor was directly exposed to air for two weeks at room temperature. Figure 7b compares the LSPR absorption spectra of the fresh sensor and the one exposed to air for two weeks. It appears that both the central wavelength and peak intensity of the resonance barely changed except for a tiny drift of the baseline. This result indicates the proposed LSPR sensor has relative long-term durability under normal storage conditions. Due to good stability in the air, it is suggested to employ the sensor when a measurement is required instead of constantly using the sensor in water for a long time. (b) absorption spectra of the LSPR sensor tested immediately versus the one exposed to air for two weeks.

Stability
To investigate the stability of the optical fiber sensor, the proposed LSPR sensor was directly exposed to air for two weeks at room temperature. Figure 7b compares the LSPR absorption spectra of the fresh sensor and the one exposed to air for two weeks. It appears that both the central wavelength and peak intensity of the resonance barely changed except for a tiny drift of the baseline. This result indicates the proposed LSPR sensor has relative long-term durability under normal storage conditions. Due to good stability in the air, it is suggested to employ the sensor when a measurement is required instead of constantly using the sensor in water for a long time.

Conclusions
A Hg 2+ detection method was proposed and experimentally demonstrated by the LSPR with PDDA-templated AuNPs and multiple CS/PAA bilayers films. A new flame brushing technique was proposed to avoid the aggregation of the AuNPs and realize a well-dispersed AuNPs array on the optical fiber surface. The AuNPs were coated with a CS/PAA film using the LBL self-assembly method, which acted as the sensing materials. The results showed that the sensor had a detection sensitivity of 0.51 nm/µM with certain reusability and stability. The sensing film did not change significantly over two weeks.

Conflicts of Interest:
The authors declare no conflict of interest.