- Email: [email protected]
A photonic crystal fiber glucose sensor filled with silver nanowires
Optics Communications 359 (2016) 279–284
Contents lists available at ScienceDirect
Optics Communications journal homepage: www.elsevier.com/locate/optcom
A photonic crystal fiber glucose sensor filled with silver nanowires X.C. Yang, Y. Lu n, M.T. Wang, J.Q. Yao College of Precision Instrument and Opto-Electronics Engineering, Key Laboratory of Opto-Electronics Information Technology (Ministry of Education), Tianjin University, Tianjin 300072, PR China
art ic l e i nf o
a b s t r a c t
Article history: Received 19 May 2015 Received in revised form 28 September 2015 Accepted 29 September 2015
We report a photonic crystal fiber glucose sensor filled with silver nanowires in this paper. The proposed sensor is both analyzed by COMSOL multiphysics software and demonstrated by experiments. The extremely high average spectral sensitivity 19009.17 nm/RIU for experimental measurement is obtained, equivalent to 44.25 mg/dL of glucose in water, which is lower than 70 mg/dL for efficient detection of hypoglycemia episodes. The silver nanowires diameter which may affect the sensor's spectral sensitivity is also discussed and an optimal range of silver nanowires diameter 90–120 nm is obtained. We expect that the sensor can provide an effective platform for glucose sensing and potentially leading to a further development towards minimal-invasive glucose measurement. & 2015 Elsevier B.V. All rights reserved.
Keywords: Photonic crystal fiber Glucose sensor Silver nanowires
1. Introduction In recent years, glucose sensors have become attractive due to the increase of diabetes. To get a good treatment, patients have to measure the blood glucose concentration frequently. Then a minimal-invasive, compact, and straightforward with high measurement accuracy glucose sensor is necessary. Several methods for measuring the concentration of a glucose solution have been proposed, such as sensors based on surface plasmon resonances (SPRs) [1], interferometers [2], [3], resonant cavities [4], and midinfrared photoacoustics [5], [6]. However, these traditional methods present a number of disadvantages such as complicated structures, costly integration, lack of accuracy, low reliability, and difficulties in mass production. In [1], Lam et al. proposed a SPR system based on the gold coated prism for the measurement of glucose in aqueous solution with an extremely high detection resolution of 8.67 10 6 RIU, equivalent to 6.23 mg/dL of glucose in water. But the bulky structure makes it limited of mechanical reliability and impossible for highly integrate. In [6], Liakat et al. implemented a non-invasive setup to realize glucose sensing in biological fluids using mid-infrared light. Clinically accurate measurements as low as 30 mg/dL were obtained. However, water has strong capability to absorb mid-infrared light, then it can only penetrates up to 100 mm into human skin where blood capillaries are not reached. This would limit its widespread application. Optic fiber based SPR sensors have attracted much attention recently due to their high sensitivity, compact structure, easy to operate, n
Corresponding author. E-mail address: [email protected] (Y. Lu).
http://dx.doi.org/10.1016/j.optcom.2015.09.102 0030-4018/& 2015 Elsevier B.V. All rights reserved.
and low cost [7]. Moreover, only a small quantity of test solution is required, potentially leading to its further development towards a minimal-invasive glucose measurement of interstitial fluid [1]. Compared with traditional optical fibers, photonic crystal fibers (PCFs) have many other advantages. They are made of single material and have several geometric parameters that can be manipulated for larger flexibility of design [8]. The cladding air holes can be used as channels either optical transmission or analyte, which can realize the interaction of light and matter [9]. The sensing mechanism of PCF–SPR sensors is through coupling the leaky core mode to the plasmon to achieve resonance sensing. The PCFs' flexible design makes it easy to equate the effective index of the core mode to that of the material under test. Thus phase matching condition between the core mode and the plasmon can be easily achieved at the required wavelength and then resonance occurs [10]. As analyte is inside the cladding air holes of the PCF, the sensor can have high reliability and the package can be very compact. Moreover, PCF based sensors can realize high sensitivity. In [11], Wu et al. presented a microfluidic refractive index (RI) sensor based on a directional coupler architecture using solid-core PCFs. It works for analytes having refractive indices slightly higher than the PCF background RI and the sensitivity of 30,100 nm/RIU has been derived from measurements. In this paper, a PCF–SPR glucose sensor filled with silver nanowires has been analyzed through numerical simulations and demonstrated by experiments. We take a 8 cm LMA-10 PCF filled with analyte and silver nanowires as the active region. The proposed sensor is analyzed through the finite element method (FEM) using COMSOL multiphysics software. The simulated results show that a red-shift is obtained with the increasing of filling analyte RI
280
X.C. Yang et al. / Optics Communications 359 (2016) 279–284
and that 90–120 nm should be an optimal range of silver nanowires diameter for the proposed sensor. To demonstrate its feasibility, a setup has been built and the average spectral sensitivity 19009.17 nm/RIU for experimental measurement is obtained, equivalent to 44.25 mg/dL of glucose in water. The detection resolution of such a sensor is lower than 70 mg/dL for efficient detection of hypoglycemia episodes, which can provide a reference for the implementation and application of glucose sensors or other biochemical sensing.
2. Theoretical analysis The numerically simulated PCF is commercially available LMA10 produced by NKT Photonics. The diameter of the core is dc ¼10 μm. The diameters of the cladding air hole and the pitch are approximately d ¼3 μm and Λ ¼ 6.5 μm, respectively. The outer cladding diameter is D ¼125 μm. The whole cladding consists of seven layer air holes of hexagonal lattices. The RI of fused silica fiber is determined by Sellmeier equation and the RI of silver nanowires is referred to the Handbook of Optics. The simulated cross section of the LMA-10 PCF is shown in Fig. 1(a). The electromagnetic mode of the glucose sensor is solved by the FEM using COMSOL multiphysics software. In the numerical simulation, some silver nanowires are embedded in the air holes of the first layer as the sensing region. In [12], Lu et al. demonstrated that when the number of silver nanowires embedded in each air hole of the first layer is three, the sensor will get saturated spectral and intensity sensitivities. The sensitivities will remain relatively stable with the continuously increasing of the silver nanowires numbers. Moreover, the irregularity of the filled nanowires has no effect on sensitivity. In [13], Luan et al. reported that the silver nanowires are unlikely to suspended in the liquid (leave from the holes surface) because of the gravity effect, then they testified that the resonance wavelength will not change as long as the nanowires are still on the surfaces of the holes and that sensitivity of the sensor is relatively stable with the randomly filled nanowires. The cross section of the channels filled with analyte and silver nanowires is shown in Fig. 1(b). The purpose of the design is to enhance the coupling between a core-guided mode and a plasmon mode, and simultaneously to reduce the plasmon to plasmon mode coupling [14,15]. The electric field distributions of the fundamental mode of LMA-10 is shown in Fig. 1(c). Obviously, we can see that the fundamental mode is confined well within the core, and the arrows indicate the electric field direction. As a PCF–SPR sensor, the most crucial requirement is phase matching of a core-guided mode and a plasmon mode. Fig. 2 shows the changes of effective refractive indices of the two modes in the vicinity of the phase matching point of the proposed sensor when glucose solution is 10 g/L. Theoretically, phase matching requires equating the propagation constants of the two modes, implying that the effective refractive indices of the two modes have to be close. The effective RI of a core-guided mode is close to that of a core material. The effective RI of a plasmon mode is determined by the background silica, the adjacent analyte, and the metallic coating, usually at a value not significantly larger than the filling analyte [16]. When the phase matching is satisfied at a certain wavelength, the energy of a core-guided mode is transferred to the plasmon mode and a resonant loss peak will be observed at this wavelength. To investigate the relationship between the loss spectra and the glucose concentration, samples ranging from 10 g/L to 60 g/L are used in the simulation. The RI of glucose solution can be calculated by [2]:
Fig.1. (a) The simulated cross section of LMA-10. (b) Cross-section of the channels filled with analyte and silver nanowires. (c) The electric field distributions of the fundamental mode.
n = 0.00011889c + 1.33230545
(1)
where c is the glucose concentration (g/L) and n is the glucose solution RI. Confinement loss is the light confinement ability within the core region, and the corresponding expression is defined as:
X.C. Yang et al. / Optics Communications 359 (2016) 279–284
Fig.2. Dispersion relations of various modes for the proposed sensor. The black line and blue dash line represent the real parts of the effective refractive indices of the core-guided mode and the plasmon mode, respectively. The red curve shows the imaginary part of the effective RI of the core-guided mode. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
αloss(dB/m) = 8.86⋅k 0Im[neff ]
(2)
Where k0 ¼2π/λ is the wavenumber with λ in meters, Im[neff] is the imaginary part of the mode effective RI. The simulated results of the relationship between confinement loss ɑloss and different samples is shown in Fig. 3. We can see that the resonance wavelength red-shifts with the glucose concentration increase and the peak loss will also increase, which can be explained by the increase of glucose solution RI will vary the phase matching point and enhance the phase matching coupling. Then the spectral sensitivity of the corresponding sensor can be obtained as:
S (nm/RIU) = Δλ /Δn
(3)
where Δλ is the offset of the resonant wavelength and Δn is the change of glucose solution RI. When the RI of the glucose solution is changed, the resonance spectrum will be varied. Therefore, we can obtain information about the analyte according to the change of the resonance wavelength. Surface plasmon waves (SPWs) are very sensitive to the
Fig.4. Simulated results of the relationship between confinement loss and different diameters of silver nanowires.
diameter of the silver nanowires. Then different diameters of silver nanowires are embedded in the air holes in simulation. As illustrated in Fig. 4, the loss spectra of the sensor vary considerably with the change of the silver nanowires diameters. Generally, the phase-matching point shifts to a longer wavelength with the silver nanowires diameter increasing. When the diameter is too small (60 nm), the offset of the resonant wavelength Δλ is not apparent and the peak loss is low as glucose concentration changes from 10 g/L to 60 g/L. With the silver nanowires diameter increasing (90–120 nm), the resonance spectrum becomes sharper and the peak loss increases, also Δλ becomes larger, then the spectral sensitivity will increase. However, when the silver nanowires diameter continues to increase (150–210 nm), the higher order plasmon modes will be excited. As we see, a secondary peak at a shorter wavelength is magnified, which may introduce noises and make the detection more difficult. When the silver nanowires diameter is larger than 240 nm, the peak loss and Δλ begins to decrease and the resonance spectral becomes wider. A wider resonance peak will result in a increased standard deviation of the spectral variation [17]. Another detection method is known as the power detection. Assume that the wavelength of the light is λ, and the transmission length is L, then the power detection sensitivity can be defined as:
SA(RIU−1) =
Fig.3. Simulated results of the relationship between confinement loss and different samples of glucose solution.
281
∂P (L, λ , na) ∂α(λ , na) 1 1 = ∂na P (L, λ , na) α(λ , na) ∂na
(4)
Fig. 5 shows the simulated results of power detection sensitivity curves corresponding to the glucose sensor filled with different diameters of silver nanowires. We can see that the power detection sensitivity goes through a circle of rise and fall, which increases with the increase of silver nanowires diameter first (60– 210 nm). When the silver nanowires diameter is 210 nm, the glucose sensor has the maximum power detection sensitivity 513.68 RIU 1 and then it tends to decrease with the silver nanowires diameter continues increasing. Considering the width of the resonance peak, the spectral sensitivity and the power detection sensitivity, 90–120 nm should be an optimal range of silver nanowires diameter for the proposed sensor. So when performing experiments, we can tune the resonance wavelength to a desired value by adjusting the silver nanowires diameter.
282
X.C. Yang et al. / Optics Communications 359 (2016) 279–284
Fig.5. Simulated results of power detection sensitivity curves corresponding to the proposed sensor filled with different diameters of silver nanowires.
Fig.7. The cross section of LMA-10.
Fig.6. The scheme of the experimental setup for PCF-based SPR sensor.
3. Experimental analysis To confirm the implementing corresponding to the theory modeling [9], we design a setup for glucose sensing, which is shown in Fig. 6. Two ends of the LMA-10 are spliced with the single mode fibers by a commercially available splicer for measurement convenience. The total splicing loss for the two splicing points is about 6 dB. We use a supercontinuum broadband source (SBS) (EXB-1 from NKT photonics) as the light source and an optical spectrum analyzer (OSA) (USB 4000 from Ocean Optics) to monitor the transmission spectrum passing though the SMFs and the PCF. The beam is coupled into the single mode fiber (SMF) by lens, and then spreads into the LMA-10 filled with analyte. When the RI of the filled analyte is changed, the resonance spectrum will be varied. Therefore, we can obtain information about the analyte according to the change of the resonance wavelength. The cross section of LMA-10 is shown in Fig. 7. The background material of the proposed PCF is fused silica. To realize the SPR effect, silver nanowires are filled in the air holes of PCF with the analyte. The silver nanowires we used is a stable translucent colloidal suspension of silver nanowires in distilled water carrier. The diameter of the nanowires is about 90 nm and the average length is about 30 μm. The concentration of the silver nanowires is 1000 ppm. Moderate compatible surfactant is used to prevent the nanowires from agglomerating. As the silver nanowires colloids is mainly composed of distilled water and what we are concerned of is the variation of the solution RI (Δn), then we can get the samples by dissolving the glucose into the colloids. We use an electronic balance with a resolution of 0.1 mg (ME 104 from Mettler Toledo) to weigh the glucose. Then the glucose concentration is close to the requirement of Eq. (1). To demonstrate the glucose sensor's feasibility, we do a series of experiments. Samples of glucose solution are defined as 10 g/L, 20 g/L, 30 g/L, 40 g/L, 50 g/L and 60 g/L, respectively. The experimental results of the relationship between confinement loss and different samples are shown in Fig. 8. We can see that the
resonance wavelength shifts to the longer wavelength as the glucose concentration increases, which corresponds to the simulation. The reason is that a lower glucose concentration will lead to a lower solution RI, thus a lower effective RI of the core-guided mode, then the real parts of the core-guided mode will decrease, but the plasmon mode is unchanged, so the intersection point of the dispersion curves (see Fig. 2) for the plasmon mode and the core-guided mode, which corresponding to the resonance wavelength shifts to the longer wavelength. We can also see that some subpeaks will appear, which are some noises or other coupling effects. Ordinarily, the coupling can support several waveguide modes which result in several peaks, while only the most obvious resonance peak is the coupling between the plasmon mode and the core-guided mode, and is most suitable for detecting [9,18]. So in the experiment, we only consider the main resonance peak shift for sensing. When the glucose concentration changes from 10 g/L to 60 g/L, the offset of the resonance wavelength Δλ is 113 nm. According to Eqs. (1) and (3), the increment of solution RI is 5.9445 10 3 RIU, and the average spectral sensitivity 19009.17 nm/RIU can be obtained by experiments. The resolution of the spectrograph we used is 1 nm, then the detection resolution of the sensor is 44.25 mg/dL of glucose in water, which is lower than 70 mg/dL for efficient detection of hypoglycemia episodes. We use two different diameters of silver nanowires in the experiment to verify the simulated results. The diameters of the silver nanowires are 60 nm and 90 nm, respectively. The relationship between confinement loss and different diameters of silver nanowires is shown in Fig. 9. We can see that with the silver nanowires diameter increasing, the resonance wavelength shifts to the longer wavelength. As the glucose concentration changes from 10 g/L to 60 g/L, the offset of the resonance wavelength Δλ is 113 nm when the silver nanowires diameter is 90 nm, which is larger than 101 nm when the silver nanowires diameter is 60 nm. It means that we can get higher spectral sensitivity when using 90 nm silver nanowires. Also, 90 nm silver nanowires have a sharper resonance spectrum and higher peak loss than 60 nm. This is because 90 nm silver nanowires can support a stronger SPWs, and the plasmon mode will increase, but the core-guided mode is not changed, then the intersection point corresponding to the resonance wavelength shifts to the longer wavelength (see Fig. 2).
X.C. Yang et al. / Optics Communications 359 (2016) 279–284
70000
50000 40000 30000 20000 10000 0 -10000 300
50000 40000 30000 20000 10000 0
400
500
600
700
800
-10000 300
900
400
Wavelength(nm)
70000
Confinement Loss(dB/m)
Confinement Loss(dB/m)
700
800
900
40g/L 586nm
60000
40000 30000 20000 10000 0
50000 40000 30000 20000 10000 0
400
500 600 700 800 Wavelength(nm)
70000
-10000 300
900
400
500 600 700 800 Wavelength(nm)
70000
50g/L 605nm
50000 40000 30000 20000 10000 0
900
60g/L 621nm
60000 Confinement Loss(dB/m)
60000 Confinement Loss(dB/m)
600
70000
30g/L
50000
-10000 300
500
Wavelength(nm)
555nm
60000
-10000 300
20g/L
524nm
60000 Confinement Loss(dB/m)
Confinement Loss(dB/m)
60000
70000
10g/L
508nm
283
50000 40000 30000 20000 10000 0
400
500 600 700 800 Wavelength(nm)
900
-10000 300
400
500
600
700
800
900
Wavelength(nm)
Fig.8. The relationship between confinement loss and different samples of glucose solution. (a) 10 g/L, (b) 20 g/L, (c) 30 g/L, (d) 40 g/L, (e) 50 g/L, and (f) 60 g/L.
Moreover, a stronger SPWs will enhance the coupling efficiency between the core-guided mode and the plasmon mode, thus a sharper resonance spectrum and higher peak loss. The experimental results have a same trend with the simulation. The similarity of simulated and experimental results have been shown in Fig. 10, which all show linearly increasing relationships between the resonance peaks and the glucose concentration. We can see that the experiments and the simulations exhibit a good qualitative agreement in general with little discrepancies as the glucose concentration increasing. When the glucose concentration changes from 10 g/L to 60 g/L, the resonance wavelengths all shift to the longer wavelength, but the experimental resonance peak shifts (2.397 nm/g/L) much faster than simulated (0.146 nm/g/L). This is mainly caused by the difference of the glucose solution RI between the simulations and the experiments. In the simulations, the refractive indices of the samples are all calculated by Eq. (1). According to [2], the refractive indices calculated by Eq. (1) and presented in CRC Handbook of Chemistry and Physics are very close with a minor error of 4.1 10 4 (at 20 °C). However, in the experiments, the estimate of the glucose solution refractive
indices via Eq. (1) are not so accurate due to manual operation errors, the instabilities of liquid fluctuation or other factors of external environment. For instance, the weighing of glucose and the measurement of silver nanowires colloids in the process of making samples may generate errors. Besides, the environmental temperature can’t be absolutely stable, which may also change the glucose solution RI. Moreover, the glucose sensor has a high precision and is very sensitive to the solution RI, for example, a 1% RIU change of solution RI may lead to the resonance peak shifting nearly 200 nm. Then the results in experiments may have a discrepancy with the simulations. But they have a same trend when the glucose concentration changes from 10 g/L to 60 g/L. Generally, the simulations are only used to estimate the positions of the resonance peaks and the intensities of the peak losses approximately. 4. Conclusion In this paper, we analyze a PCF–SPR glucose sensor filled with silver nanowires and different concentrations of analyte both in
284
X.C. Yang et al. / Optics Communications 359 (2016) 279–284
obtained with the increasing of filling analyte RI and the extremely high average spectral sensitivity 19009.17 nm/RIU for experimental measurement is obtained. The silver nanowires diameters which may affect the sensor's spectral sensitivity is also discussed, and the optimal range of silver nanowires diameter for the proposed sensor 90–120 nm is obtained. The theoretical analysis and experimental results in this paper can provide a reference for the implementation and application of PCF-based SPR glucose sensors or other PCF-based SPR sensing.
Acknowledgments This work was supported by the National Basic Research Program of China (973 Program) (Grant number: 2010CB327801)
Fig. 9. The relationship between confinement loss and two different diameters of silver nanowires.
Fig.10. The similarity of simulated and experimental results as functions of glucose concentration.
theory and experiment. The results show that a red-shift is
References [1] W.W. Lam, L.H. Chu, C.L. Wong, Y.T. Zhang, Sens. Actuators B 105 (2005) 138–143. [2] Y.L. Yeh, Opt. Lasers Eng. 46 (2008) 667–670. [3] K.H. Chen, C.C. Hsu, D.C. Su, Appl. Opt. 42 (2003) 5774–5776. [4] Y. Chen, Z.Y. Li, H.X. Yi, Z.P. Zhou, J. Yu, Front. Optoelectron. China 2 (2009) 304–307. [5] J. Kottmann, J.M. Rey, J. Luginbühl, E. Reichmann, M.W. Sigrist, Biomed. Opt. Express 3 (2012) 667–680. [6] S. Liakat, K.A. Bors, et al., Biomed. Opt. Express 4 (2013) 1083–1090. [7] S. Binu, et al., Mater. Sci. Eng. C 29 (2009) 183–186. [8] R. Wang, J.Q. Yao, Y.P. Miao, Y. Lu, Sensors 13 (2013) 1–10. [9] Y. Lu, M.T. Wang, C.J. Hao, Z.Q. Zhao, J.Q. Yao, Photonics J. 6 (2014) 6801307. [10] X.Y. Fu, Y. Lu, X.H. Huang, et al., Opt. Appl. 4 (2011) 941–951. [11] D.K.C. Wu, B.T. Kuhlmey, B.J. Eggleton, Opt. Lett. 34 (2009) 322–324. [12] Y. Lu, C.J. Hao, B.Q. Wu, X.H. Huang, et al., Sensors 12 (2012) 12016–12025. [13] N.N. Luan, R. Wang, W.H. Lv, Y. Lu, J.Q. Yao, Sensors 14 (2014) 16035–16045. [14] Y. Peng, J. Hou, Z.H. Huang, Q.S. Lu, Appl. Opt. 51 (2012) 6361–6367. [15] Y. Peng, J. Hou, Y. Zhang, Z.H. Huang, R. Xiao, Q.S. Lu, Opt. Lett. 38 (2013) 263–265. [16] X. Yu, Y. Zhang, et al., J. Opt. 12 (2010) 015005. [17] I.M. White, X.D. Fan, Opt. Express 16 (2008) 1020–1028. [18] T. Biswas, R. Chattopadhyay, S.K. Bhadra, J. Opt. 16 (2014) 045001.
Contents lists available at ScienceDirect
Optics Communications journal homepage: www.elsevier.com/locate/optcom
A photonic crystal fiber glucose sensor filled with silver nanowires X.C. Yang, Y. Lu n, M.T. Wang, J.Q. Yao College of Precision Instrument and Opto-Electronics Engineering, Key Laboratory of Opto-Electronics Information Technology (Ministry of Education), Tianjin University, Tianjin 300072, PR China
art ic l e i nf o
a b s t r a c t
Article history: Received 19 May 2015 Received in revised form 28 September 2015 Accepted 29 September 2015
We report a photonic crystal fiber glucose sensor filled with silver nanowires in this paper. The proposed sensor is both analyzed by COMSOL multiphysics software and demonstrated by experiments. The extremely high average spectral sensitivity 19009.17 nm/RIU for experimental measurement is obtained, equivalent to 44.25 mg/dL of glucose in water, which is lower than 70 mg/dL for efficient detection of hypoglycemia episodes. The silver nanowires diameter which may affect the sensor's spectral sensitivity is also discussed and an optimal range of silver nanowires diameter 90–120 nm is obtained. We expect that the sensor can provide an effective platform for glucose sensing and potentially leading to a further development towards minimal-invasive glucose measurement. & 2015 Elsevier B.V. All rights reserved.
Keywords: Photonic crystal fiber Glucose sensor Silver nanowires
1. Introduction In recent years, glucose sensors have become attractive due to the increase of diabetes. To get a good treatment, patients have to measure the blood glucose concentration frequently. Then a minimal-invasive, compact, and straightforward with high measurement accuracy glucose sensor is necessary. Several methods for measuring the concentration of a glucose solution have been proposed, such as sensors based on surface plasmon resonances (SPRs) [1], interferometers [2], [3], resonant cavities [4], and midinfrared photoacoustics [5], [6]. However, these traditional methods present a number of disadvantages such as complicated structures, costly integration, lack of accuracy, low reliability, and difficulties in mass production. In [1], Lam et al. proposed a SPR system based on the gold coated prism for the measurement of glucose in aqueous solution with an extremely high detection resolution of 8.67 10 6 RIU, equivalent to 6.23 mg/dL of glucose in water. But the bulky structure makes it limited of mechanical reliability and impossible for highly integrate. In [6], Liakat et al. implemented a non-invasive setup to realize glucose sensing in biological fluids using mid-infrared light. Clinically accurate measurements as low as 30 mg/dL were obtained. However, water has strong capability to absorb mid-infrared light, then it can only penetrates up to 100 mm into human skin where blood capillaries are not reached. This would limit its widespread application. Optic fiber based SPR sensors have attracted much attention recently due to their high sensitivity, compact structure, easy to operate, n
Corresponding author. E-mail address: [email protected] (Y. Lu).
http://dx.doi.org/10.1016/j.optcom.2015.09.102 0030-4018/& 2015 Elsevier B.V. All rights reserved.
and low cost [7]. Moreover, only a small quantity of test solution is required, potentially leading to its further development towards a minimal-invasive glucose measurement of interstitial fluid [1]. Compared with traditional optical fibers, photonic crystal fibers (PCFs) have many other advantages. They are made of single material and have several geometric parameters that can be manipulated for larger flexibility of design [8]. The cladding air holes can be used as channels either optical transmission or analyte, which can realize the interaction of light and matter [9]. The sensing mechanism of PCF–SPR sensors is through coupling the leaky core mode to the plasmon to achieve resonance sensing. The PCFs' flexible design makes it easy to equate the effective index of the core mode to that of the material under test. Thus phase matching condition between the core mode and the plasmon can be easily achieved at the required wavelength and then resonance occurs [10]. As analyte is inside the cladding air holes of the PCF, the sensor can have high reliability and the package can be very compact. Moreover, PCF based sensors can realize high sensitivity. In [11], Wu et al. presented a microfluidic refractive index (RI) sensor based on a directional coupler architecture using solid-core PCFs. It works for analytes having refractive indices slightly higher than the PCF background RI and the sensitivity of 30,100 nm/RIU has been derived from measurements. In this paper, a PCF–SPR glucose sensor filled with silver nanowires has been analyzed through numerical simulations and demonstrated by experiments. We take a 8 cm LMA-10 PCF filled with analyte and silver nanowires as the active region. The proposed sensor is analyzed through the finite element method (FEM) using COMSOL multiphysics software. The simulated results show that a red-shift is obtained with the increasing of filling analyte RI
280
X.C. Yang et al. / Optics Communications 359 (2016) 279–284
and that 90–120 nm should be an optimal range of silver nanowires diameter for the proposed sensor. To demonstrate its feasibility, a setup has been built and the average spectral sensitivity 19009.17 nm/RIU for experimental measurement is obtained, equivalent to 44.25 mg/dL of glucose in water. The detection resolution of such a sensor is lower than 70 mg/dL for efficient detection of hypoglycemia episodes, which can provide a reference for the implementation and application of glucose sensors or other biochemical sensing.
2. Theoretical analysis The numerically simulated PCF is commercially available LMA10 produced by NKT Photonics. The diameter of the core is dc ¼10 μm. The diameters of the cladding air hole and the pitch are approximately d ¼3 μm and Λ ¼ 6.5 μm, respectively. The outer cladding diameter is D ¼125 μm. The whole cladding consists of seven layer air holes of hexagonal lattices. The RI of fused silica fiber is determined by Sellmeier equation and the RI of silver nanowires is referred to the Handbook of Optics. The simulated cross section of the LMA-10 PCF is shown in Fig. 1(a). The electromagnetic mode of the glucose sensor is solved by the FEM using COMSOL multiphysics software. In the numerical simulation, some silver nanowires are embedded in the air holes of the first layer as the sensing region. In [12], Lu et al. demonstrated that when the number of silver nanowires embedded in each air hole of the first layer is three, the sensor will get saturated spectral and intensity sensitivities. The sensitivities will remain relatively stable with the continuously increasing of the silver nanowires numbers. Moreover, the irregularity of the filled nanowires has no effect on sensitivity. In [13], Luan et al. reported that the silver nanowires are unlikely to suspended in the liquid (leave from the holes surface) because of the gravity effect, then they testified that the resonance wavelength will not change as long as the nanowires are still on the surfaces of the holes and that sensitivity of the sensor is relatively stable with the randomly filled nanowires. The cross section of the channels filled with analyte and silver nanowires is shown in Fig. 1(b). The purpose of the design is to enhance the coupling between a core-guided mode and a plasmon mode, and simultaneously to reduce the plasmon to plasmon mode coupling [14,15]. The electric field distributions of the fundamental mode of LMA-10 is shown in Fig. 1(c). Obviously, we can see that the fundamental mode is confined well within the core, and the arrows indicate the electric field direction. As a PCF–SPR sensor, the most crucial requirement is phase matching of a core-guided mode and a plasmon mode. Fig. 2 shows the changes of effective refractive indices of the two modes in the vicinity of the phase matching point of the proposed sensor when glucose solution is 10 g/L. Theoretically, phase matching requires equating the propagation constants of the two modes, implying that the effective refractive indices of the two modes have to be close. The effective RI of a core-guided mode is close to that of a core material. The effective RI of a plasmon mode is determined by the background silica, the adjacent analyte, and the metallic coating, usually at a value not significantly larger than the filling analyte [16]. When the phase matching is satisfied at a certain wavelength, the energy of a core-guided mode is transferred to the plasmon mode and a resonant loss peak will be observed at this wavelength. To investigate the relationship between the loss spectra and the glucose concentration, samples ranging from 10 g/L to 60 g/L are used in the simulation. The RI of glucose solution can be calculated by [2]:
Fig.1. (a) The simulated cross section of LMA-10. (b) Cross-section of the channels filled with analyte and silver nanowires. (c) The electric field distributions of the fundamental mode.
n = 0.00011889c + 1.33230545
(1)
where c is the glucose concentration (g/L) and n is the glucose solution RI. Confinement loss is the light confinement ability within the core region, and the corresponding expression is defined as:
X.C. Yang et al. / Optics Communications 359 (2016) 279–284
Fig.2. Dispersion relations of various modes for the proposed sensor. The black line and blue dash line represent the real parts of the effective refractive indices of the core-guided mode and the plasmon mode, respectively. The red curve shows the imaginary part of the effective RI of the core-guided mode. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
αloss(dB/m) = 8.86⋅k 0Im[neff ]
(2)
Where k0 ¼2π/λ is the wavenumber with λ in meters, Im[neff] is the imaginary part of the mode effective RI. The simulated results of the relationship between confinement loss ɑloss and different samples is shown in Fig. 3. We can see that the resonance wavelength red-shifts with the glucose concentration increase and the peak loss will also increase, which can be explained by the increase of glucose solution RI will vary the phase matching point and enhance the phase matching coupling. Then the spectral sensitivity of the corresponding sensor can be obtained as:
S (nm/RIU) = Δλ /Δn
(3)
where Δλ is the offset of the resonant wavelength and Δn is the change of glucose solution RI. When the RI of the glucose solution is changed, the resonance spectrum will be varied. Therefore, we can obtain information about the analyte according to the change of the resonance wavelength. Surface plasmon waves (SPWs) are very sensitive to the
Fig.4. Simulated results of the relationship between confinement loss and different diameters of silver nanowires.
diameter of the silver nanowires. Then different diameters of silver nanowires are embedded in the air holes in simulation. As illustrated in Fig. 4, the loss spectra of the sensor vary considerably with the change of the silver nanowires diameters. Generally, the phase-matching point shifts to a longer wavelength with the silver nanowires diameter increasing. When the diameter is too small (60 nm), the offset of the resonant wavelength Δλ is not apparent and the peak loss is low as glucose concentration changes from 10 g/L to 60 g/L. With the silver nanowires diameter increasing (90–120 nm), the resonance spectrum becomes sharper and the peak loss increases, also Δλ becomes larger, then the spectral sensitivity will increase. However, when the silver nanowires diameter continues to increase (150–210 nm), the higher order plasmon modes will be excited. As we see, a secondary peak at a shorter wavelength is magnified, which may introduce noises and make the detection more difficult. When the silver nanowires diameter is larger than 240 nm, the peak loss and Δλ begins to decrease and the resonance spectral becomes wider. A wider resonance peak will result in a increased standard deviation of the spectral variation [17]. Another detection method is known as the power detection. Assume that the wavelength of the light is λ, and the transmission length is L, then the power detection sensitivity can be defined as:
SA(RIU−1) =
Fig.3. Simulated results of the relationship between confinement loss and different samples of glucose solution.
281
∂P (L, λ , na) ∂α(λ , na) 1 1 = ∂na P (L, λ , na) α(λ , na) ∂na
(4)
Fig. 5 shows the simulated results of power detection sensitivity curves corresponding to the glucose sensor filled with different diameters of silver nanowires. We can see that the power detection sensitivity goes through a circle of rise and fall, which increases with the increase of silver nanowires diameter first (60– 210 nm). When the silver nanowires diameter is 210 nm, the glucose sensor has the maximum power detection sensitivity 513.68 RIU 1 and then it tends to decrease with the silver nanowires diameter continues increasing. Considering the width of the resonance peak, the spectral sensitivity and the power detection sensitivity, 90–120 nm should be an optimal range of silver nanowires diameter for the proposed sensor. So when performing experiments, we can tune the resonance wavelength to a desired value by adjusting the silver nanowires diameter.
282
X.C. Yang et al. / Optics Communications 359 (2016) 279–284
Fig.5. Simulated results of power detection sensitivity curves corresponding to the proposed sensor filled with different diameters of silver nanowires.
Fig.7. The cross section of LMA-10.
Fig.6. The scheme of the experimental setup for PCF-based SPR sensor.
3. Experimental analysis To confirm the implementing corresponding to the theory modeling [9], we design a setup for glucose sensing, which is shown in Fig. 6. Two ends of the LMA-10 are spliced with the single mode fibers by a commercially available splicer for measurement convenience. The total splicing loss for the two splicing points is about 6 dB. We use a supercontinuum broadband source (SBS) (EXB-1 from NKT photonics) as the light source and an optical spectrum analyzer (OSA) (USB 4000 from Ocean Optics) to monitor the transmission spectrum passing though the SMFs and the PCF. The beam is coupled into the single mode fiber (SMF) by lens, and then spreads into the LMA-10 filled with analyte. When the RI of the filled analyte is changed, the resonance spectrum will be varied. Therefore, we can obtain information about the analyte according to the change of the resonance wavelength. The cross section of LMA-10 is shown in Fig. 7. The background material of the proposed PCF is fused silica. To realize the SPR effect, silver nanowires are filled in the air holes of PCF with the analyte. The silver nanowires we used is a stable translucent colloidal suspension of silver nanowires in distilled water carrier. The diameter of the nanowires is about 90 nm and the average length is about 30 μm. The concentration of the silver nanowires is 1000 ppm. Moderate compatible surfactant is used to prevent the nanowires from agglomerating. As the silver nanowires colloids is mainly composed of distilled water and what we are concerned of is the variation of the solution RI (Δn), then we can get the samples by dissolving the glucose into the colloids. We use an electronic balance with a resolution of 0.1 mg (ME 104 from Mettler Toledo) to weigh the glucose. Then the glucose concentration is close to the requirement of Eq. (1). To demonstrate the glucose sensor's feasibility, we do a series of experiments. Samples of glucose solution are defined as 10 g/L, 20 g/L, 30 g/L, 40 g/L, 50 g/L and 60 g/L, respectively. The experimental results of the relationship between confinement loss and different samples are shown in Fig. 8. We can see that the
resonance wavelength shifts to the longer wavelength as the glucose concentration increases, which corresponds to the simulation. The reason is that a lower glucose concentration will lead to a lower solution RI, thus a lower effective RI of the core-guided mode, then the real parts of the core-guided mode will decrease, but the plasmon mode is unchanged, so the intersection point of the dispersion curves (see Fig. 2) for the plasmon mode and the core-guided mode, which corresponding to the resonance wavelength shifts to the longer wavelength. We can also see that some subpeaks will appear, which are some noises or other coupling effects. Ordinarily, the coupling can support several waveguide modes which result in several peaks, while only the most obvious resonance peak is the coupling between the plasmon mode and the core-guided mode, and is most suitable for detecting [9,18]. So in the experiment, we only consider the main resonance peak shift for sensing. When the glucose concentration changes from 10 g/L to 60 g/L, the offset of the resonance wavelength Δλ is 113 nm. According to Eqs. (1) and (3), the increment of solution RI is 5.9445 10 3 RIU, and the average spectral sensitivity 19009.17 nm/RIU can be obtained by experiments. The resolution of the spectrograph we used is 1 nm, then the detection resolution of the sensor is 44.25 mg/dL of glucose in water, which is lower than 70 mg/dL for efficient detection of hypoglycemia episodes. We use two different diameters of silver nanowires in the experiment to verify the simulated results. The diameters of the silver nanowires are 60 nm and 90 nm, respectively. The relationship between confinement loss and different diameters of silver nanowires is shown in Fig. 9. We can see that with the silver nanowires diameter increasing, the resonance wavelength shifts to the longer wavelength. As the glucose concentration changes from 10 g/L to 60 g/L, the offset of the resonance wavelength Δλ is 113 nm when the silver nanowires diameter is 90 nm, which is larger than 101 nm when the silver nanowires diameter is 60 nm. It means that we can get higher spectral sensitivity when using 90 nm silver nanowires. Also, 90 nm silver nanowires have a sharper resonance spectrum and higher peak loss than 60 nm. This is because 90 nm silver nanowires can support a stronger SPWs, and the plasmon mode will increase, but the core-guided mode is not changed, then the intersection point corresponding to the resonance wavelength shifts to the longer wavelength (see Fig. 2).
X.C. Yang et al. / Optics Communications 359 (2016) 279–284
70000
50000 40000 30000 20000 10000 0 -10000 300
50000 40000 30000 20000 10000 0
400
500
600
700
800
-10000 300
900
400
Wavelength(nm)
70000
Confinement Loss(dB/m)
Confinement Loss(dB/m)
700
800
900
40g/L 586nm
60000
40000 30000 20000 10000 0
50000 40000 30000 20000 10000 0
400
500 600 700 800 Wavelength(nm)
70000
-10000 300
900
400
500 600 700 800 Wavelength(nm)
70000
50g/L 605nm
50000 40000 30000 20000 10000 0
900
60g/L 621nm
60000 Confinement Loss(dB/m)
60000 Confinement Loss(dB/m)
600
70000
30g/L
50000
-10000 300
500
Wavelength(nm)
555nm
60000
-10000 300
20g/L
524nm
60000 Confinement Loss(dB/m)
Confinement Loss(dB/m)
60000
70000
10g/L
508nm
283
50000 40000 30000 20000 10000 0
400
500 600 700 800 Wavelength(nm)
900
-10000 300
400
500
600
700
800
900
Wavelength(nm)
Fig.8. The relationship between confinement loss and different samples of glucose solution. (a) 10 g/L, (b) 20 g/L, (c) 30 g/L, (d) 40 g/L, (e) 50 g/L, and (f) 60 g/L.
Moreover, a stronger SPWs will enhance the coupling efficiency between the core-guided mode and the plasmon mode, thus a sharper resonance spectrum and higher peak loss. The experimental results have a same trend with the simulation. The similarity of simulated and experimental results have been shown in Fig. 10, which all show linearly increasing relationships between the resonance peaks and the glucose concentration. We can see that the experiments and the simulations exhibit a good qualitative agreement in general with little discrepancies as the glucose concentration increasing. When the glucose concentration changes from 10 g/L to 60 g/L, the resonance wavelengths all shift to the longer wavelength, but the experimental resonance peak shifts (2.397 nm/g/L) much faster than simulated (0.146 nm/g/L). This is mainly caused by the difference of the glucose solution RI between the simulations and the experiments. In the simulations, the refractive indices of the samples are all calculated by Eq. (1). According to [2], the refractive indices calculated by Eq. (1) and presented in CRC Handbook of Chemistry and Physics are very close with a minor error of 4.1 10 4 (at 20 °C). However, in the experiments, the estimate of the glucose solution refractive
indices via Eq. (1) are not so accurate due to manual operation errors, the instabilities of liquid fluctuation or other factors of external environment. For instance, the weighing of glucose and the measurement of silver nanowires colloids in the process of making samples may generate errors. Besides, the environmental temperature can’t be absolutely stable, which may also change the glucose solution RI. Moreover, the glucose sensor has a high precision and is very sensitive to the solution RI, for example, a 1% RIU change of solution RI may lead to the resonance peak shifting nearly 200 nm. Then the results in experiments may have a discrepancy with the simulations. But they have a same trend when the glucose concentration changes from 10 g/L to 60 g/L. Generally, the simulations are only used to estimate the positions of the resonance peaks and the intensities of the peak losses approximately. 4. Conclusion In this paper, we analyze a PCF–SPR glucose sensor filled with silver nanowires and different concentrations of analyte both in
284
X.C. Yang et al. / Optics Communications 359 (2016) 279–284
obtained with the increasing of filling analyte RI and the extremely high average spectral sensitivity 19009.17 nm/RIU for experimental measurement is obtained. The silver nanowires diameters which may affect the sensor's spectral sensitivity is also discussed, and the optimal range of silver nanowires diameter for the proposed sensor 90–120 nm is obtained. The theoretical analysis and experimental results in this paper can provide a reference for the implementation and application of PCF-based SPR glucose sensors or other PCF-based SPR sensing.
Acknowledgments This work was supported by the National Basic Research Program of China (973 Program) (Grant number: 2010CB327801)
Fig. 9. The relationship between confinement loss and two different diameters of silver nanowires.
Fig.10. The similarity of simulated and experimental results as functions of glucose concentration.
theory and experiment. The results show that a red-shift is
References [1] W.W. Lam, L.H. Chu, C.L. Wong, Y.T. Zhang, Sens. Actuators B 105 (2005) 138–143. [2] Y.L. Yeh, Opt. Lasers Eng. 46 (2008) 667–670. [3] K.H. Chen, C.C. Hsu, D.C. Su, Appl. Opt. 42 (2003) 5774–5776. [4] Y. Chen, Z.Y. Li, H.X. Yi, Z.P. Zhou, J. Yu, Front. Optoelectron. China 2 (2009) 304–307. [5] J. Kottmann, J.M. Rey, J. Luginbühl, E. Reichmann, M.W. Sigrist, Biomed. Opt. Express 3 (2012) 667–680. [6] S. Liakat, K.A. Bors, et al., Biomed. Opt. Express 4 (2013) 1083–1090. [7] S. Binu, et al., Mater. Sci. Eng. C 29 (2009) 183–186. [8] R. Wang, J.Q. Yao, Y.P. Miao, Y. Lu, Sensors 13 (2013) 1–10. [9] Y. Lu, M.T. Wang, C.J. Hao, Z.Q. Zhao, J.Q. Yao, Photonics J. 6 (2014) 6801307. [10] X.Y. Fu, Y. Lu, X.H. Huang, et al., Opt. Appl. 4 (2011) 941–951. [11] D.K.C. Wu, B.T. Kuhlmey, B.J. Eggleton, Opt. Lett. 34 (2009) 322–324. [12] Y. Lu, C.J. Hao, B.Q. Wu, X.H. Huang, et al., Sensors 12 (2012) 12016–12025. [13] N.N. Luan, R. Wang, W.H. Lv, Y. Lu, J.Q. Yao, Sensors 14 (2014) 16035–16045. [14] Y. Peng, J. Hou, Z.H. Huang, Q.S. Lu, Appl. Opt. 51 (2012) 6361–6367. [15] Y. Peng, J. Hou, Y. Zhang, Z.H. Huang, R. Xiao, Q.S. Lu, Opt. Lett. 38 (2013) 263–265. [16] X. Yu, Y. Zhang, et al., J. Opt. 12 (2010) 015005. [17] I.M. White, X.D. Fan, Opt. Express 16 (2008) 1020–1028. [18] T. Biswas, R. Chattopadhyay, S.K. Bhadra, J. Opt. 16 (2014) 045001.