- Email: [email protected]
Comparison of detection limit in fiber-based conventional, amplified, and gain-clamped cavity ring-down techniques
Optics Communications 407 (2018) 186–192
Contents lists available at ScienceDirect
Optics Communications journal homepage: www.elsevier.com/locate/optcom
Comparison of detection limit in fiber-based conventional, amplified, and gain-clamped cavity ring-down techniques K. Sharma a,b,1 , M.I.M. Abdul Khudus a,c , S.U. Alam a , S. Bhattacharya b , D. Venkitesh b, *, G. Brambilla a a b c
Optoelectronics Research Centre, University of Southampton, Southampton SO17 1BJ, UK Department of Electrical Engineering, Indian Institute of Technology Madras, Chennai - 600036, India Department of Physics, Faculty of Science, University of Malaya, 50603 Kuala Lumpur, Malaysia
a r t i c l e
i n f o
Keywords: Cavity ring-down technique Detection limit Evanescent wave sensing Optical fiber sensors Refractive index Tapered fiber
a b s t r a c t Relative performance and detection limit of conventional, amplified, and gain-clamped cavity ring-down techniques (CRDT) in all-fiber configurations are compared experimentally for the first time. Refractive index measurement using evanescent field in tapered fibers is used as a benchmark for the comparison. The systematic optimization of a nested-loop configuration in gain-clamped CRDT is also discussed, which is crucial for achieving a constant gain in a CRDT experiment. It is found that even though conventional CRDT has the lowest standard error in ring-down time (𝛥𝜏), the value of ring-down time (𝜏) is very small, thus leading to poor detection limit. Amplified CRDT provides an improvement in 𝜏, albeit with two orders of magnitude higher 𝛥𝜏 due to amplifier noise. The nested-loop configuration in gain-clamped CRDT helps in reducing 𝛥𝜏 by an order of magnitude as compared to amplified CRDT whilst retaining the improvement in 𝜏. A detection limit of 1.03 × 10−4 RIU at refractive index of 1.322 with a 3 mm long and 4.5 μm diameter tapered fiber is demonstrated with the gain-clamped CRDT. © 2017 Elsevier B.V. All rights reserved.
1. Introduction Cavity ring-down technique (CRDT) is a popular time-domain technique for measurements that require high sensitivity [1]. It has many significant industrial and bio-medical applications [2]. This technique is known for its advantages such as long effective path lengths, insensitivity to intensity fluctuations of the light source and above all, immunity to electromagnetic interference of the sensing set up [3]. CRDT was first used in 1980 to measure the reflectivity of highlyreflective mirrors [4] and later, for measurements of small cavity loss in free-space cavities [5]. Compared to free-space cavities, fiber-based cavities — consisting of a fiber loop (known as conventional CRDT) make the arrangement compact and portable, in addition to being alignment-free [6,7]. However, these cavities are affected by large inherent loss due to the insertion loss of components such as couplers and sample holders [8]. A suitable fiber amplifier was included in the fiber cavity to compensate this inherent loss and achieve long ring-down times and hence, improve detection limit [9]. An erbium
doped fiber amplifier (EDFA) was used in this particular demonstration of amplified CRDT since the wavelength of operation was in 1530– 1560 nm range. However, the pulse-to-pulse fluctuations in the gain and amplified spontaneous emission (ASE) noise of EDFA resulted in a reduction of the detection limit [10]. Post-processing techniques such as digital and adaptive filtering were suggested to remove the impact of ASE noise [11–13]. A nested-loop configuration (known as gainclamped CRDT) was also suggested to provide a constant (clamped) gain for all pulses, using a laser loop instead of an amplifier in the fiber cavity [9,14]. Conventional [2,6,7,15,16], amplified [9–13,17], and gain-clamped CRDTs [8,9,14,18] have been implemented in the past for various applications, albeit independently. However, to the best of our knowledge, there has been no systematic comparison of parameters such as minimum detectable change in loss and detection limit, between the three techniques through experiments. In this paper, refractive index (RI) measurements of sugar solution are performed at a wavelength
* Corresponding author.
E-mail addresses: [email protected] (K. Sharma), [email protected] (M.I.M. Abdul Khudus), [email protected] (S.U. Alam), [email protected] (S. Bhattacharya), [email protected] (D. Venkitesh), [email protected] (G. Brambilla). 1 Permanent address: Department of Electrical Engineering, Indian Institute of Technology Madras, Chennai - 600036, India. http://dx.doi.org/10.1016/j.optcom.2017.09.017 Received 9 June 2017; Received in revised form 3 September 2017; Accepted 5 September 2017 Available online 22 September 2017 0030-4018/© 2017 Elsevier B.V. All rights reserved.
K. Sharma et al.
Optics Communications 407 (2018) 186–192
Fig. 1. Schematic of the experimental set-up of fiber-based C-CRDT. Fig. 2. Schematic of the experimental set-up of fiber-based A-CRDT. An erbium doped fiber amplifier (EDFA) compensate the intrinsic loss of the cavity.
of 1550 nm using uncoated tapered fiber as an intra-cavity element in the three techniques. These measurements are used as a benchmark to quantify and compare the relative performance and detection limits of the three techniques. Conventional and amplified CRDTs have been used in the past by others at a wavelength of 1550 nm [19,20] and by our group at a wavelength of 1953 nm [21], for refractometric sensing using tapered fibers in independent demonstrations. The evanescent field in tapered fibers helps in achieving good sensitivity [22], which is further accentuated when the tapered fiber is included as an intracavity element. Time-domain refractometric measurements using the evanescent field in tapered fibers thus combine the advantages of fiberbased CRDT and that of tapered fibers. A nested-loop configuration is typically used in a gain-clamped CRDT for achieving a constant gain. Balancing the two loops for an optimal operation — especially for performing a CRDT sensing experiment is extremely challenging [14]. Some of the issues of the nested-loop are discussed in the literature [8,14,18] but a detailed procedure for its optimization is not available. In this paper, the systematic optimization of the nested-loop in gain-clamped CRDT is also discussed in detail.
by the amplifier. In this configuration, the equation of ring-down time is modified as [9] 𝜏=
In pulsed CRDT, light from a pulsed laser source is coupled into a high-finesse cavity. The time-dependent output of the cavity is monitored using a photo-detector and an oscilloscope to determine the cavity loss [1]. In a conventional fiber-based CRDT (referred to as C-CDRT henceforth), the cavity is constructed using two 99/01 couplers and a sample head (tapered fiber in this case), as shown in Fig. 1. A pulse is launched into the cavity using the first coupler. The power of this cavity ring-down (CRD) pulse decreases exponentially in each roundtrip, based on the total loss of the cavity. The second 99/01 coupler is used to extract output of the CRD cavity. The time taken for the output of the cavity to decay to 1∕𝑒 of its initial value is referred to as the ringdown time (𝜏), which is related to the total cavity loss (𝛼) through [6] 𝑡𝑟 𝑡𝑟 = , 𝛼 𝛼𝑐 + 𝛼𝑠
(2)
where G is the single pass gain of EDFA. The total loss in the cavity (𝛼) is controlled by adjusting the gain, G (by varying pump power of EDFA). In order to maximize the ring-down time, the gain has to be adjusted such that almost all the loss except the sample loss is compensated. A variable optical attenuator (VOA) can be used to finely balance the gain and loss of the cavity. This is a sensitive adjustment, because an exact match of gain and the total cavity loss would lead to lasing in the cavity, which has to be avoided. A careful adjustment of G can significantly increase the ring-down time and the effective number of round-trips, which in turn improves the detection limit [24]. However, the EDFA induces pulse-to-pulse gain fluctuations since the power of the CRD pulse progressively decreases with each round-trip. Consequently, the noise due to ASE also fluctuates, thus degrading the performance of the system. CRDT in the gain-clamped configuration is used to maintain the constant gain experienced by the CRD pulse in each round-trip by replacing the amplifier shown in Fig. 2 by a laser loop. The schematic of experimental setup of the gain-clamped CRDT (referred to as G-CRDT henceforth) is shown in Fig. 3. It consists of two nested-loops - an outer loop (with solid lines) that constitutes the CRD cavity and an inner loop (with dotted lines) which is a laser loop that provides a stabilized gain to the CRD pulse propagating in the outer loop. The outer and inner loops are inter-connected using two 50/50 couplers, with an EDFA in common. The lasing wavelength (λlas ) propagating through the inner loop and the CRD pulse propagating at signal wavelength (λsig ) in the outer loop share the same EDFA, thus enabling the gain experienced by the CRD pulse to be clamped by λlas . An optical band pass filter (BPF 1 – with center wavelength at λlas ) is used in the inner loop to ensure that λsig ≠ λlas , to avoid cross-talk between the two loops [18]. A variable optical attenuator (VOA 1) is used to carefully adjust the gain in the EDFA at λsig such that it can compensate the inherent loss of the outer loop. The output of a pulsed laser source is coupled into the outer loop using a 99/01 coupler. The other components in this loop are tapered fiber, a second 99/01 coupler to extract the output of the CRD cavity, a band pass filter (BPF 2) and a variable optical attenuator (VOA 2). BPF 2 (with center wavelength at λsig ) ensures that the wavelength propagating in the outer loop correspond to only λsig and VOA 2 is used to adjust the cavity loss of the outer loop, such that the outer loop does not lase. The outer loop has to be maintained just below the lasing threshold at λsig and the inner loop above the lasing threshold at λlas [14]. When the inner loop is lasing, the gain (or inversion) in the EDFA remains constant i.e. gain clamping occurs at all wavelengths except for λlas . This provides a constant gain to the CRD pulse in the
2. Theory
𝜏=
𝑡𝑟 𝑡𝑟 = , 𝛼 𝛼𝑐 + 𝛼𝑠 − 𝐺
(1)
where 𝑡𝑟 is the round-trip time of the cavity. 𝛼𝑐 refers to the inherent loss of the cavity, which includes the loss of couplers, splices, tapered fiber, and the absorption loss due to solvent in case of a liquid sample. 𝛼𝑠 is the loss due to the sample absorption and is referred to as the sample loss. The envelope of the output pulses of the cavity is fitted with an exponential function to extract 𝜏. For a given cavity, 𝑡𝑟 and 𝛼𝑐 are known; 𝛼𝑠 is then calculated using (1). The effective number of round-trips — defined as the ratio (𝜏/𝑡𝑟 ) [23], is limited by the inherent loss of the cavity in the case of C-CRDT. A schematic of the experimental setup for amplified CRDT (referred to as A-CRDT henceforth) is shown in Fig. 2. An EDFA is additionally included in the cavity to compensate for the inherent cavity loss. A band pass filter (BPF) is used to filter the out-of-band ASE generated 187
K. Sharma et al.
Optics Communications 407 (2018) 186–192
Fig. 4. Schematic of a tapered fiber showing waist, transition and un-tapered fiber regions.
3. Experimental set-up In the set-up used for C-CRDT (Fig. 1), the 99% ports of the two couplers are spliced in the CRD cavity along with a tapered fiber. The 1% ports of the two couplers are used to inject pulses from a pulsed laser source (1550 nm, average power: ∼10 dBm, repetition rate: 200 kHz, and pulse width: ∼5 ns) and to extract the output. The round-trip time (length) of the cavity is 41.6 ns (8.32 m). The output pulses are detected with a photo-detector (Thorlabs DET01CFC, bandwidth: 1.2 GHz) and recorded with an oscilloscope (LeCroy WaveSurfer 44Xs, bandwidth: 400 MHz). It is to be noted that a slower detector can also be used for detecting the nano-second pulses used in the experiments. In the A-CRDT set-up (Fig. 2), a home-made EDFA, a tunable BPF (TECOS WTFM-1550-S-FA; bandwidth: 0.4 nm), and VOA are included along with two couplers with split ratios of 99:01 and a tapered fiber. The round-trip time (length) of the cavity is 324 ns (64.8 m). The repetition rate of the pulsed source is reduced to 10 kHz (average power: ∼0 dBm) to account for increase in length of the cavity due to additional components. The BPF is tuned to the signal wavelength (1550 nm). In the G-CRDT set-up (Fig. 3), two 50/50 couplers are used to couple light between the inner and outer loops. BPF 1 (TECOS TFM-1550-S-FA; bandwidth: 2 nm) and BPF 2 (TECOS WTFM-1550-S-FA; bandwidth: 0.4 nm) are used to choose the inner loop lasing wavelength (λlas = ∼1554.1 nm) and outer loop signal wavelength (λsig = 1550 nm) respectively. The output spectra of the inner and outer loops are monitored using an OSA (Yokogawa AQ6370). The repetition rate of the pulsed source is 10 kHz with average output power of −10 dBm. The round-trip time (length) of the outer loop is 360 ns (72 m). The wavelength of the CRD pulse from the pulsed laser source is maintained at 1550 nm for all the experiments. Schematic of the tapered fiber used in the experiments is shown in Fig. 4. A standard single mode fiber (Corning-SMF28e) is adiabatically tapered using the modified flame brushing technique to result in a reduced diameter fiber, with uniform waist in between two transition regions [32]. The transition regions convert the core mode of the standard single mode fiber to the cladding mode of the taper waist, with the confinement provided by the cladding-ambient interface [32]. During fabrication, fiber is fixed on two computer-controlled translation stages using magnetic clamps and a micro-heater is used to heat the fiber being stretched until the desired diameter is obtained. An image of a tapered fiber (5 μm diameter), taken using a microscope (Nikon Eclipse LV100) with a 50x objective is shown in Fig. 5(a). Image of an un-tapered fiber with 125 μm diameter taken using the same objective lens is shown in Fig. 5(b) for comparison. The loss of fibers is monitored in real-time during taper fabrication. The fabricated tapered fibers have an excess loss smaller than 0.4 dB. Sample solutions with different RI (1.316–1.344) are prepared by dissolving sugar in water with concentrations between 0% and 20%. RI of these solutions is measured using a standard refractometer (Metricon Corporation Model 2010 Prism Coupler) at 1550 nm.
Fig. 3. Schematic of the experimental set-up of fiber-based G-CRDT. The inner loop (with dotted lines) is designed to stabilize the gain whereas the outer loop (with solid lines) functions as the CRD cavity.
outer loop, irrespective of the input power [25]. If a small input signal is given to the inner loop, it gets amplified using the available gain at that wavelength in the EDFA, without disturbing the lasing action. To achieve a clamped gain in a G-CRD experiment, the nested-loop must be operated under the following conditions: (1) the input power of CRD pulse to the inner loop should be low enough to avoid any gain competition with λlas ; (2) the pump current in the inner loop must be maintained above the lasing threshold; (3) the loss in the outer loop must be larger than the gain available at λsig in the EDFA; (4) the wavelengths of the two loops need to be optimized. Hence, the optimization process relies on tuning of the parameters such as the input power of CRD pulse, the EDFA gain, the pump current of the EDFA, the inner and outer loop loss, and operating wavelengths of the inner and the outer loops. The possible wavelengths of operation of the two loops depend on the gain saturation characteristics of the EDFA and its optimization is discussed in [18,26]. In all three configurations of CRDTs, a tapered fiber is spliced in the CRD cavity, and is immersed in a sample solution. Increase in the concentration of the sample solution increases the RI of the sample surrounding the tapered fiber. This enhances the fraction of power in the evanescent field (𝜂𝑐𝑙𝑎𝑑 )–defined as ratio of power propagating outside the fiber and the total power—thus increasing the overlap of sample with the evanescent field [21,27]. When a sample (sugar solution in this case) absorbs at the wavelength of propagation, it induces loss in the light propagation path [28]. Increase in the overlap between the sample and the evanescent field, therefore, augments the loss induced by the sample [29], which in turn reduces the ring-down time [27]. This reduction in ring-down time helps in measuring the RI of the sample. The performance metric of a refractometric sensor is its detection limit (DL), defined as the minimum RI that can be accurately measured by the system. It is calculated using a method similar to the one used in [30,21]. For a CRDT experiment, the minimum detectable change in loss (Δ𝛼) that can be accurately measured by the system can be calculated as [31] Δ𝜏 , (3) 𝜏2 where Δ𝜏 is standard error in 𝜏; quantified as standard deviation of the ring-down time [30,21]. DL can be calculated as
Δ𝛼 = 𝑡𝑟
4. Results
Δ𝛼 , (4) 𝑆 where S is the slope of sample loss versus RI curve. DL can be improved either by reducing Δ𝜏 or by increasing 𝜏. DL is calculated for all the three techniques separately. DL =
4.1. Optimization of the three cavities In C-CRDT (Fig. 1), the inherent loss of the cavity should be small enough such that sufficient number of round-trips can be achieved to 188
K. Sharma et al.
Optics Communications 407 (2018) 186–192
Fig. 5. Optical microscope images of (a) tapered and (b) un-tapered fibers.
Fig. 6. Change in ring-down time and cavity loss with pump current of EDFA in A-CRDT.
Fig. 7. Variation of gain at the signal wavelength with change in the input signal power for different VOA 1 loss values in G-CRDT. Gain clamping is observed for different magnitudes of gain, albeit for different ranges of input powers.
increase the effective path length and ring-down time. The inherent loss for our set up is measured to be 0.5 dB and hence, the effective number of round-trips is ∼7. In A-CRDT (Fig. 2), the pump current of EDFA can be adjusted to achieve a desired gain such that it can almost compensate for the inherent loss in the cavity. Fig. 6 shows the change in ring-down time and cavity loss with pump current of EDFA. Large pump current increases gain in the cavity, reduces the total cavity loss and thus increases the ring-down time [10]. If the pump current is too high, the gain in the cavity overcomes the loss and the cavity starts to lase. Hence, the pump current (limited to 120 mA in this experiment) should be maintained below the lasing threshold of the cavity. The VOA is be used to fine-tune the cavity loss. The effective number of round-trips in this case is expected to be ∼40. Since the optimization of the nested-loop configuration in G-CRDT (Fig. 3) requires a detailed understanding and control of multiple parameters, it is initially performed without the tapered fiber in the cavity. The nested-loop is optimized again after splicing the tapered fiber, before performing the refractometric measurements. In each case, the optimization is performed successively for the inner and the outer loop. The input power of the CRD pulse to the inner loop, the loss of the inner loop, and the pump current of EDFA are optimized for the inner loop, in order to achieve a constant gain. For these optimized operating conditions, the loss of the outer loop is optimized such that the inner loop continues to lase while the CRD pulse propagating in the outer loop experiences the clamped gain. These optimization procedures are discussed in detailed below.
Fig. 8. Gain at the signal wavelength for different pump currents of EDFA and VOA 1 loss values in G-CRDT.
4.1.1. Optimization of the inner loop in G-CRDT The power of a CRD pulse in a CRD cavity changes with each roundtrip and hence gain clamping is studied as a function of input power. For this, only the inner loop is constructed to start with and a continuous wave (CW) signal at λsig is considered as input. The power of this CW signal represents the average power of the CRD pulse reaching EDFA in the inner loop (after the first 50/50 coupler). The pump current of EDFA is maintained at 190 mA in order to ensure that the inner loop is lasing. Gain clamping conditions, at different values of gain provided by the
Fig. 9. Spectra of the inner loop for pump currents of 100 mA and 160 mA and VOA 1 loss of 0.57 dB.
inner loop, are demonstrated. The loss of the inner loop is varied using VOA 1. Gain at λsig for different values of input power of the CW signal is recorded. The results are shown in Fig. 7. The gain values without 189
K. Sharma et al.
Optics Communications 407 (2018) 186–192
Fig. 11. Full spectra of the inner loop (with input signal on and off) in G-CRDT. Left and right insets show the zoomed version of the spectra at signal (outer loop) and laser (inner loop) wavelengths respectively.
Fig. 10. Spectra of the inner loop with different loss values of the outer loop, varied by VOA 2 in G-CRDT. The inset shows the zoomed version of the spectra.
feedback (i.e. inner loop not lasing hence, no clamping of gain) are also shown for comparison. It is found that, with the inner loop lasing, the gain remains constant for a wide range of input powers. While an increase in VOA 1 loss enhances the gain (as the inner loop is lasing), the range of input powers for which the gain is clamped, reduces. Therefore, VOA 1 can be used to set the gain required to compensate the inherent loss of the CRD cavity but the input power of CW signal (or the average power of the CRD pulse) to the inner loop has to be chosen carefully. All the gain curves in Fig. 7 coincide with the ‘without feedback’ curve when the input power of CW signal (at λsig ) reaching EDFA approaches the power at λlas circulating in the inner loop. At this point, the lasing in the inner loop is disturbed. Hence, the average input power of the CRD pulse should always be smaller than the power at λlas in the inner loop, in order to achieve gain clamping. Gain as a function of pump current is studied next, to show that the gain provided by the inner loop remains constant beyond its threshold pump current and to find the optimized value of the pump current. The gain at λsig is shown as a function of pump current for each VOA 1 loss value in Fig. 8. The input power of CW signal is maintained at −30 dBm. Fig. 8 shows that the threshold pump current, above which the gain at λsig remains constant, increases with the VOA 1 loss. Fig. 9 shows spectra of the inner loop at 100 and 160 mA pump currents for VOA 1 loss of 0.57 dB. As the pump current increases, the power at λlas (1554.1 nm) increases, while the ASE profile and the power at λsig (1550 nm) are clamped at constant values. This confirms that the gain (or inversion) in the EDFA remains constant for λsig considered beyond threshold pump currents, thus achieving gain clamping. However, increase in pump current can potentially lead to larger noise in the system. Thus, to minimize noise, the pump current should be chosen just above the lasing threshold of the inner loop.
Fig. 12. Dependence of the ring-down time and standard error in ring-down time on VOA 2 loss in G-CRDT.
the gain at λsig . Left and right insets show the zoomed version of the spectra at λsig and λlas respectively. When the loss of the outer loop is larger than the gain at λsig , the CRD pulse does not affect the lasing in the inner loop, but gets amplified. This would imply that there could be a range of values of VOA 2 loss, for which the above condition is satisfied. However, with the increase in VOA 2 loss, the ring-down time is expected to decrease, which is not desirable. On the other hand, operating closer to the lasing threshold (with VOA 2 loss very close to the gain experienced by the propagating CRD signal) would result in instabilities in the outer loop and lead to a larger error in ring-down time. In order to optimize the VOA 2 loss, 𝜏 and Δ𝜏 are measured for different values of VOA 2 loss and the results are shown in Fig. 12. Note that, the gain-clamped conditions are maintained throughout this experiment. When VOA 2 loss increases from 3.05 to 3.5 dB, Δ𝜏 reduces by an order of magnitude whereas 𝜏 reduces by a factor of half. Also, the number of effective round-trips reduces from ∼7.9 to ∼4.2. This would result in Δ𝛼 changing from 1.34 × 10−2 Np to 7.34 × 10−3 Np (calculated using (3)). The exact value of VOA 2 loss can be decided based on the requirement of Δ𝛼 for the intended application.
4.1.2. Optimization of the outer loop in G-CRDT After optimizing the average input power of CRD pulse, loss of the inner loop, and the pump current of EDFA, the loss of the outer loop has to be optimized. If the loss of the outer loop becomes smaller than the gain provided by EDFA at λsig , the outer loop starts lasing and disrupts the lasing in the inner loop. For this analysis, both the loops are connected together as shown in Fig. 3 and CRD pulse is given as input. Fig. 10 shows the spectra of the inner loop at center wavelength corresponding to λlas (inset shows the zoomed version of the spectra) as a function of loss of the outer loop (controlled by VOA 2). A loss corresponding to 3.02 dB in the outer loop is smaller than the gain provided by EDFA at λsig , thus disrupting the lasing in the inner loop. At 3.03 dB, the loss and gain are comparable, whereas at 3.04 dB, the inner loop starts lasing at λlas , since the loss of the outer loop is greater than the gain provided by EDFA at λsig . Thus, the loss in the outer loop needs to be optimized carefully such that it does not lase and disrupt lasing in the inner loop. Fig. 11 shows full spectra of the inner loop with input signal (CRD pulse) on and off, when loss of the outer loop is maintained larger than
4.2. Refractive index sensing For performing RI sensing, tapered fibers of length 3 mm and diameter ∼5 μm are spliced in all the three CRD cavities. The nestedloop of G-CRDT is optimized according to the procedures mentioned in Section 4.1. The optimized values of the input average power of CRD pulse to the outer cavity (output of the pulsed laser source), pump current, VOA 1, and VOA 2 loss used for G-CRDT experiment are −10 dBm, 165 mA, 10.5 dB, and 4.1 dB respectively. Note that the inherent loss of the outer cavity has changed with addition of the fiber taper. Hence, these values are slightly different from those discussed in section 4.2, where the optimization is performed without the fiber taper. A few μl of sugar solution is poured onto the taper surface, ensuring that the taper waist and transition regions are fully immersed inside the 190
K. Sharma et al.
Optics Communications 407 (2018) 186–192
RI, pre-calibration of the sensor system is required. To estimate DL of the system, a smaller range of RI (1.322–1.33) is chosen in which the sample loss versus RI curve follows a linear function (shown in the inset of Fig. 14 for all the three techniques). The slope values of these linear curves are used to calculate DL in the next section. The change in ringdown time versus RI is shown in Fig. 15 for all the three configurations of CRDT. The dependence of ring-down time with RI is found to be a rational function since 𝜏 is inversely proportional to the loss [27]. In Fig. 14, the sample loss versus RI curves for all the three techniques are expected to overlap as the sample loss corresponds to the loss induced by the sample in a single round-trip. However, we find a deviation for the G-CRDT case. This discrepancy can be explained as follows. Even after using the same fabrication parameters, the actual diameter of the tapered fiber used for G-CRDT experiment was found to be slightly smaller (∼4.5 μm) than that used (∼5 μm) in the C-and A-CRDT experiments. This results in a different performance for the GCRDT, as evidenced by the smaller number of pulses than A-CRDT in Fig. 13 and a different fitting function for the sample loss versus RI curve in Fig. 14. The slope of the sample loss vs RI curve in RI range 1.322–1.33 is found to be 5.8 Np/RIU and 6.2 Np/RIU for C- and ACRDT respectively while it is 30.1 Np/RIU for the G-CRDT. Smaller taper diameter results in poor confinement of mode which leads to larger 𝜂𝑐𝑙𝑎𝑑 . This increases the overlap of sample with the evanescent field and hence, the loss. Therefore, tapered fibers with smaller diameter are more sensitive to RI change. The variation of 𝜂𝑐𝑙𝑎𝑑 with different taper diameters has been studied earlier by our group [21,27]. It is to be noted that in an actual sensing experiment, a differential measurement for sample concentration (or refractive index) would be made. Hence, the absolute value of sample loss would not matter.
Fig. 13. Time-domain traces captured by oscilloscope with water as a sample for C-, Aand G-CRDT with tapered fiber of length 3 mm and diameter ∼5 μm.
4.3. Detection limit Fig. 14. Change in sample loss with refractive index in C-, A- and G-CRDT. The experimental data points are fitted with quadratic function. The data points corresponding to a smaller range of RI (1.322–1.33) are fitted with a linear function (shown in inset) to determine DL for all the three techniques.
The standard error in ring-down time (Δ𝜏) is calculated using a set of 50 data points, after averaging 1024 ring-down traces in the oscilloscope for each datum. This Δ𝜏 is not influenced by the fact that the taper diameter used for G-CRDT is slightly different from that used in C- and A-CRDT as it represents the noise of the whole system. Then, Δ𝛼 is calculated using (3). The DL calculated at RI of 1.322 (using (4)) for all the three techniques is given in Table 1 along with 𝜏, Δ𝜏, Δ𝛼 and the slope values. The standard error in ring-down time in C-CRDT is 4.66 ns, the lowest among all, but associated with a very small ring-down time (0.092 μs). On the other hand, the ring-down times in the A- and G-CRDTs are longer but at the expense of larger Δ𝜏 − 450 ns and 50.5 ns respectively. Comparing Δ𝜏 in A- and G-CRDTs, the latter has approximately 10 times smaller Δ𝜏 due to gain clamping. The detection limits achieved at a refractive index of 1.322, with tapered fibers of length 3 mm and diameter 5 μm in C- and A-CRDT are 3.91 × 10−3 RIU and 1.32 × 10−3 RIU respectively. DL achieved in G-CRDT at the same refractive index, with a tapered fiber of length 3 mm and diameter 4.5 μm is 1.03 × 10−4 RIU. This improvement in DL in G-CRDT is contributed primarily due to the reduction in Δ𝜏, coupled with an increase in the slope of sample loss versus RI curve because of the use of tapered fiber with a slightly different diameter. If tapered fibers with identical dimensions were used in all the three CRDTs, then the slope would remain the same and the best DL (5.15×10−4 RIU) would be still be achieved by G-CRDT. The DL would be much worse for all the three cases, when calculated at higher RI values as the sample loss increases and the ring-down time reduces.
Fig. 15. Change in ring-down time with refractive index in C-, A- and G-CRDT. The experimental data points are fitted with rational function.
solution. The tapered fiber is rinsed with water twice before being used with another sample. The time-domain traces are recorded with the oscilloscope with 1024 averages for each ring-down trace as shown in Fig. 13. These timedomain traces are fitted with an exponential function to extract the ringdown time. Then, the sample loss in Nepers (Np) is calculated using (1) and (2). The change in sample loss versus RI is plotted in Fig. 14 for C-, A-, and G-CRDTs. The sample loss versus RI curves follow a quadratic function (indicated by the fit shown in Fig. 14). This is due to the fact that 𝜂𝑐𝑙𝑎𝑑 shows a quadratic behavior for this large range of RI (1.316– 1.344) [27] and sample loss is directly proportional to 𝜂𝑐𝑙𝑎𝑑 [29]. Due to non-linear dependence of the sample loss on RI in this large range of
5. Conclusion The performance and DL of conventional, amplified, and gainclamped CRDTs have been compared experimentally by measuring RI of sugar solutions using tapered fibers as sensing heads. It is found that Δ𝜏 in conventional CRDT is minimum but with very small 𝜏 value. 𝜏 is 191
K. Sharma et al.
Optics Communications 407 (2018) 186–192 Table 1 Comparison of conventional, amplified, and gain-clamped CRDTs. Parameter
Conventional CRDT
Amplified CRDT
Gain-clamped CRDT
Ring-down time (𝜏) (μs) Standard error in ring-down time (Δ𝜏) (ns) Minimum detectable change in loss (Δ𝛼) (Np) Slope (Np/RIU) Detection limit at RI of 1.322 (RIU)
0.092 4.66 2.27 × 10−2 5.8 3.91 × 10−3
4.22 450 8.17 × 10−3 6.2 1.32 × 10−3
2.45 50.5 3.09 × 10−3 30.1 1.03 × 10−4
increased by two orders of magnitude in the case of amplified CRDT, but at the expense of an increased Δ𝜏. On the other hand, gain-clamped CRDT is found to have 10 times lower Δ𝜏 than amplified CRDT in addition to the increased 𝜏. In terms of complexity of the techniques, conventional CRDT is simpler as compared to the other two techniques. The gain of the EDFA and the total cavity loss in amplified CRDT has to be optimized every time, based on the inherent loss of the cavity. Gainclamped CRDT is by far the most complex technique, as parameters like input power of CRD pulse, pump power, cavity loss of the inner and outer loops have to be optimized before performing the measurements. The procedure for optimizing these parameters in gain-clamped CRDT is discussed in detail. The best DL achieved in refractometric sensing is 1.03 × 10−4 RIU at RI of 1.322 with a 3 mm long and 4.5 μm diameter tapered fiber in gain-clamped CRDT. The minimum detectable loss that can be detected by the system is 3.09 × 10−3 Np. These values can be improved by using components with smaller loss — for instance, a 99.9/0.1 coupler instead of a 99/01 coupler; and also by fabricating tapered fibers with lower insertion loss. Use of a low noise EDFA would help in achieving smaller values of Δ𝜏 in amplified CRDT, further improving minimum detectable loss and DL. Signal to noise ratio can be improved in all the three techniques by using a trans-impedance amplifier at the output of the photo-detector. From our study, gain-clamped CRDT appears to be the most sensitive technique among all three, which can be used for RI measurements with advantages such as long effective path lengths, good sensitivity, and insensitivity to the intensity fluctuations of the light source.
[7] Z. Tong, M. Jakubinek, A. Wright, A. Gillies, H.P. Loock, Fiber-loop ring-down spectroscopy: a sensitive absorption technique for small liquid samples, Rev. Sci. Instrum. 74 (2003) 4818–4826. [8] H. Waechter, J. Litman, A.H. Cheung, J.A. Barnes, H.P. Loock, Chemical sensing using fiber cavity ring-down spectroscopy, Sensors 10 (2010) 1716–1742. [9] G. Stewart, K. Atherton, H. Yu, B. Culshaw, An investigation of optical fibre amplifier loop for intra-cavity and ring-down cavity loss measurements, Measure. Sci. Technol. 12 (2001) 843–849. [10] G. Stewart, K. Atherton, B. Culshaw, Cavity-enhanced spectroscopy in fiber cavities, Opt. Lett. 29 (2004) 442–444. [11] G. Stewart, P. Shields, B. Culshaw, Development of fibre laser systems for ring-down and intracavity gas spectroscopy in the near-IR, Meas. Sci. Technol. 15 (2004) 1621– 1628. [12] N. Ni, C.C. Chan, Improving the measurement accuracy of CRD fibre amplified loop gas sensing system by using a digital LMS adaptive filter, Meas. Sci. Technol. 17 (2006) 2349–2354. [13] N. Ni, C.C. Chan, T.K. Chuah, L. Xia, P. Shum, Enhancing the measurement accuracy of a cavity enhanced fiber chemical sensor by an adaptive filter, Meas. Sci. Technol. 19 (2008) 115203:1–115203:6. [14] K. Atherton, G. Stewart, B. Culshaw, Gas detection by cavity ring-down absorption with a fibre optic amplifier loop, Proc. SPIE 4577 (2002) 25–31. [15] J. Meng, Z. Weigang, Z. Qi, L. Yaping, L. Bo, Investigation on an evanescent wave fiber-optic absorption sensor based on fiber loop cavity ring-down spectroscopy, Opt. Commun. 283 (2010) 249–253. [16] A. Yolalmaz, F.H. Sadroud, M.F. Danışman, O. Esenturk, Intracavity gas detection with fiber loop ring down spectroscopy, Opt. Commun. 396 (2017) 141–145. [17] S. Silva, O. Frazão, Multimode interference-based fiber sensor in a cavity ring-down system for refractive index measurement, Opt. Laser Technol. 91 (2017) 112–115. [18] N.L.P. Andrews, J. Litman, D. Stroh, J.A. Barnes, H.P. Loock, Near-infrared absorption detection in picolitre liquid volumes using amplified fibre loop ring-down detection, Opt. Fiber Technol. 19 (2013) 822–827. [19] D. Wu, Y. Zhao, Q. Wang, SMF taper evanescent field-based RI sensor combined with fiber loop ring down technology, IEEE Photonic. Tech. L. 27 (2015) 1802–1805. [20] X. Liu, Q. Wang, C. Li, C. Zhao, Y. Zhao, H. Hu, J. Li, Refractive index sensor based on fiber loop ring-down spectroscopy, Instrum. Sci. Technol. 44 (2016) 241–248. [21] K. Sharma, S. Liang, S.U. Alam, S. Bhattacharya, D. Venkitesh, G. Brambilla, Fiberbased cavity ring-down technique for refractive index sensing at 1953 nm using tapered fibers, IEEE Sensors Lett. 1 (2017) 1–4. [22] P. Wang, G. Brambilla, M. Ding, Y. Semenova, Q. Wu, G. Farrell, High-sensitivity, evanescent field refractometric sensor based on a tapered, multimode fiber interference, Opt. Lett. 36 (2011) 2233–2235. [23] G. Gagliardi, H.-P. Loock, in: Cavity-Enhanced Spectroscopy and Sensing, vol. 179, Springer-Verlag, 2014. [24] S. Silva, R. Magalhães, R.A. Pérez-Herrera, M. Lopez-Amo, M.B. Marques, O. Frazão, Fiber cavity ring down and gain amplification effect, Photonic Sens. 4 (2016) 324– 327. [25] M. Zirngibl, Gain control in erbium-doped fibre amplifiers by an all-optical feedback loop, Electron. Lett. 27 (1991) 560–561. [26] H. Okamura, Automatic optical loss compensation with erbium-doped fiber amplifier, J. Lightwave Technol. 10 (1992) 1110–1116. [27] K. Sharma, D. Venkitesh, S. Bhattacharya, B. Srinivasan, G. Brambilla, Non-linear behavior of ring-down time in cavity ring-down spectroscopy with tapered fibers, in: Conference on Lasers and Electro-Optics, OSA Technical Digest (online, Optical Society of America, 2016, paper JW2A.12. [28] A. Leung, P.M. Shankar, R. Mutharasan, A review of fiber-optic biosensors, Sensors Actuators B 125 (2007) 688–703. [29] Z. Wei, Z. Song, R. Song, X. Zhang, Z. Meng, Measurement of the optical absorption coefficient for liquid based on optical microfiber, Optik 125 (2014) 2880–2884. [30] K. Zhou, D.J. Webb, C. Mou, M. Farries, N. Hayes, L. Bennion, Optical fiber cavity ring down measurement of refractive index with a microchannel drilled by femtosecond laser, IEEE Photonics Technol. Lett. 21 (2009) 1653–1655. [31] C.M. Rushworth, D. James, J.W.L. Lee, C. Vallance, Top notch design for fiber-loop cavity ring-down spectroscopy, Anal. Chem. 83 (2011) 8492–8500. [32] G. Brambilla, Optical fibre nanotaper sensors, Opt. Fiber Technol. 16 (2010) 331– 342.
Acknowledgments Funding: This work was supported by the People Programme (Marie Curie Actions) of the European Unions Seventh Framework Programme FP7/2007–2013/ under REA grant agreement n0318941. The authors thank B. Srinivasan, M. Srivastava, Y. Panbiharwala, and J. Barnes for helpful discussions; M. Turvey, Dr. P. Hua, and Dr. P. Smith for their help in measuring refractive indices using refractometers; Y. Panbiharwala for proofreading the paper. They also thank Dr. J. He, Dr. S. Ganapathy, V. Mittal, H. A. Vasant, M. F. Namiq, M. B. Alonso, S. Liang, N. Thipparapu, K. Bottrill, F. H. Suhailin, S. Jain, and Dr. A. Masoudi for providing several components. References [1] J.J. Scherer, J.B. Paul, A. O’Keefe, R.J. Saykally, Cavity ringdown laser absorption spectroscopy: history, development, and application to pulsed molecular beams, Chem. Rev. 97 (1997) 25–51. [2] S.O. Silva, R. Magalhães, M.B. Marques, O. Frazão, New advances in fiber cavity ring-down technology, Opt. Laser Technol. 78 (2016) 115–119. [3] G. Berden, R. Peeters, G. Meijer, Cavity ring-down spectroscopy: experimental schemes and applications, Rev. Phys. Chem. 19 (2009) 565–607. [4] J.M. Herbelin, J.A. McKay, M.A. Kwok, R.H. Ueunten, D.S. Urevig, D.J. Spencer, D.J. Benard, Sensitive measurement of photon lifetime and true reflectances in an optical cavity by a phase-shift method, Appl. Opt. 19 (1980) 144–147. [5] A. O’Keefe, D.A.G. Deacon, Cavity ring-down optical spectrometer for absorption measurements using pulsed laser sources, Rev. Sci. Instrum. 59 (1988) 2544–2551. [6] R.S. Brown, I. Kozin, Z. Tong, R.D. Oleschuk, H.P. Loock, Fiber-loop ring-down spectroscopy, J. Chem. Phys. 117 (2002) 10444–10447.
192
Contents lists available at ScienceDirect
Optics Communications journal homepage: www.elsevier.com/locate/optcom
Comparison of detection limit in fiber-based conventional, amplified, and gain-clamped cavity ring-down techniques K. Sharma a,b,1 , M.I.M. Abdul Khudus a,c , S.U. Alam a , S. Bhattacharya b , D. Venkitesh b, *, G. Brambilla a a b c
Optoelectronics Research Centre, University of Southampton, Southampton SO17 1BJ, UK Department of Electrical Engineering, Indian Institute of Technology Madras, Chennai - 600036, India Department of Physics, Faculty of Science, University of Malaya, 50603 Kuala Lumpur, Malaysia
a r t i c l e
i n f o
Keywords: Cavity ring-down technique Detection limit Evanescent wave sensing Optical fiber sensors Refractive index Tapered fiber
a b s t r a c t Relative performance and detection limit of conventional, amplified, and gain-clamped cavity ring-down techniques (CRDT) in all-fiber configurations are compared experimentally for the first time. Refractive index measurement using evanescent field in tapered fibers is used as a benchmark for the comparison. The systematic optimization of a nested-loop configuration in gain-clamped CRDT is also discussed, which is crucial for achieving a constant gain in a CRDT experiment. It is found that even though conventional CRDT has the lowest standard error in ring-down time (𝛥𝜏), the value of ring-down time (𝜏) is very small, thus leading to poor detection limit. Amplified CRDT provides an improvement in 𝜏, albeit with two orders of magnitude higher 𝛥𝜏 due to amplifier noise. The nested-loop configuration in gain-clamped CRDT helps in reducing 𝛥𝜏 by an order of magnitude as compared to amplified CRDT whilst retaining the improvement in 𝜏. A detection limit of 1.03 × 10−4 RIU at refractive index of 1.322 with a 3 mm long and 4.5 μm diameter tapered fiber is demonstrated with the gain-clamped CRDT. © 2017 Elsevier B.V. All rights reserved.
1. Introduction Cavity ring-down technique (CRDT) is a popular time-domain technique for measurements that require high sensitivity [1]. It has many significant industrial and bio-medical applications [2]. This technique is known for its advantages such as long effective path lengths, insensitivity to intensity fluctuations of the light source and above all, immunity to electromagnetic interference of the sensing set up [3]. CRDT was first used in 1980 to measure the reflectivity of highlyreflective mirrors [4] and later, for measurements of small cavity loss in free-space cavities [5]. Compared to free-space cavities, fiber-based cavities — consisting of a fiber loop (known as conventional CRDT) make the arrangement compact and portable, in addition to being alignment-free [6,7]. However, these cavities are affected by large inherent loss due to the insertion loss of components such as couplers and sample holders [8]. A suitable fiber amplifier was included in the fiber cavity to compensate this inherent loss and achieve long ring-down times and hence, improve detection limit [9]. An erbium
doped fiber amplifier (EDFA) was used in this particular demonstration of amplified CRDT since the wavelength of operation was in 1530– 1560 nm range. However, the pulse-to-pulse fluctuations in the gain and amplified spontaneous emission (ASE) noise of EDFA resulted in a reduction of the detection limit [10]. Post-processing techniques such as digital and adaptive filtering were suggested to remove the impact of ASE noise [11–13]. A nested-loop configuration (known as gainclamped CRDT) was also suggested to provide a constant (clamped) gain for all pulses, using a laser loop instead of an amplifier in the fiber cavity [9,14]. Conventional [2,6,7,15,16], amplified [9–13,17], and gain-clamped CRDTs [8,9,14,18] have been implemented in the past for various applications, albeit independently. However, to the best of our knowledge, there has been no systematic comparison of parameters such as minimum detectable change in loss and detection limit, between the three techniques through experiments. In this paper, refractive index (RI) measurements of sugar solution are performed at a wavelength
* Corresponding author.
E-mail addresses: [email protected] (K. Sharma), [email protected] (M.I.M. Abdul Khudus), [email protected] (S.U. Alam), [email protected] (S. Bhattacharya), [email protected] (D. Venkitesh), [email protected] (G. Brambilla). 1 Permanent address: Department of Electrical Engineering, Indian Institute of Technology Madras, Chennai - 600036, India. http://dx.doi.org/10.1016/j.optcom.2017.09.017 Received 9 June 2017; Received in revised form 3 September 2017; Accepted 5 September 2017 Available online 22 September 2017 0030-4018/© 2017 Elsevier B.V. All rights reserved.
K. Sharma et al.
Optics Communications 407 (2018) 186–192
Fig. 1. Schematic of the experimental set-up of fiber-based C-CRDT. Fig. 2. Schematic of the experimental set-up of fiber-based A-CRDT. An erbium doped fiber amplifier (EDFA) compensate the intrinsic loss of the cavity.
of 1550 nm using uncoated tapered fiber as an intra-cavity element in the three techniques. These measurements are used as a benchmark to quantify and compare the relative performance and detection limits of the three techniques. Conventional and amplified CRDTs have been used in the past by others at a wavelength of 1550 nm [19,20] and by our group at a wavelength of 1953 nm [21], for refractometric sensing using tapered fibers in independent demonstrations. The evanescent field in tapered fibers helps in achieving good sensitivity [22], which is further accentuated when the tapered fiber is included as an intracavity element. Time-domain refractometric measurements using the evanescent field in tapered fibers thus combine the advantages of fiberbased CRDT and that of tapered fibers. A nested-loop configuration is typically used in a gain-clamped CRDT for achieving a constant gain. Balancing the two loops for an optimal operation — especially for performing a CRDT sensing experiment is extremely challenging [14]. Some of the issues of the nested-loop are discussed in the literature [8,14,18] but a detailed procedure for its optimization is not available. In this paper, the systematic optimization of the nested-loop in gain-clamped CRDT is also discussed in detail.
by the amplifier. In this configuration, the equation of ring-down time is modified as [9] 𝜏=
In pulsed CRDT, light from a pulsed laser source is coupled into a high-finesse cavity. The time-dependent output of the cavity is monitored using a photo-detector and an oscilloscope to determine the cavity loss [1]. In a conventional fiber-based CRDT (referred to as C-CDRT henceforth), the cavity is constructed using two 99/01 couplers and a sample head (tapered fiber in this case), as shown in Fig. 1. A pulse is launched into the cavity using the first coupler. The power of this cavity ring-down (CRD) pulse decreases exponentially in each roundtrip, based on the total loss of the cavity. The second 99/01 coupler is used to extract output of the CRD cavity. The time taken for the output of the cavity to decay to 1∕𝑒 of its initial value is referred to as the ringdown time (𝜏), which is related to the total cavity loss (𝛼) through [6] 𝑡𝑟 𝑡𝑟 = , 𝛼 𝛼𝑐 + 𝛼𝑠
(2)
where G is the single pass gain of EDFA. The total loss in the cavity (𝛼) is controlled by adjusting the gain, G (by varying pump power of EDFA). In order to maximize the ring-down time, the gain has to be adjusted such that almost all the loss except the sample loss is compensated. A variable optical attenuator (VOA) can be used to finely balance the gain and loss of the cavity. This is a sensitive adjustment, because an exact match of gain and the total cavity loss would lead to lasing in the cavity, which has to be avoided. A careful adjustment of G can significantly increase the ring-down time and the effective number of round-trips, which in turn improves the detection limit [24]. However, the EDFA induces pulse-to-pulse gain fluctuations since the power of the CRD pulse progressively decreases with each round-trip. Consequently, the noise due to ASE also fluctuates, thus degrading the performance of the system. CRDT in the gain-clamped configuration is used to maintain the constant gain experienced by the CRD pulse in each round-trip by replacing the amplifier shown in Fig. 2 by a laser loop. The schematic of experimental setup of the gain-clamped CRDT (referred to as G-CRDT henceforth) is shown in Fig. 3. It consists of two nested-loops - an outer loop (with solid lines) that constitutes the CRD cavity and an inner loop (with dotted lines) which is a laser loop that provides a stabilized gain to the CRD pulse propagating in the outer loop. The outer and inner loops are inter-connected using two 50/50 couplers, with an EDFA in common. The lasing wavelength (λlas ) propagating through the inner loop and the CRD pulse propagating at signal wavelength (λsig ) in the outer loop share the same EDFA, thus enabling the gain experienced by the CRD pulse to be clamped by λlas . An optical band pass filter (BPF 1 – with center wavelength at λlas ) is used in the inner loop to ensure that λsig ≠ λlas , to avoid cross-talk between the two loops [18]. A variable optical attenuator (VOA 1) is used to carefully adjust the gain in the EDFA at λsig such that it can compensate the inherent loss of the outer loop. The output of a pulsed laser source is coupled into the outer loop using a 99/01 coupler. The other components in this loop are tapered fiber, a second 99/01 coupler to extract the output of the CRD cavity, a band pass filter (BPF 2) and a variable optical attenuator (VOA 2). BPF 2 (with center wavelength at λsig ) ensures that the wavelength propagating in the outer loop correspond to only λsig and VOA 2 is used to adjust the cavity loss of the outer loop, such that the outer loop does not lase. The outer loop has to be maintained just below the lasing threshold at λsig and the inner loop above the lasing threshold at λlas [14]. When the inner loop is lasing, the gain (or inversion) in the EDFA remains constant i.e. gain clamping occurs at all wavelengths except for λlas . This provides a constant gain to the CRD pulse in the
2. Theory
𝜏=
𝑡𝑟 𝑡𝑟 = , 𝛼 𝛼𝑐 + 𝛼𝑠 − 𝐺
(1)
where 𝑡𝑟 is the round-trip time of the cavity. 𝛼𝑐 refers to the inherent loss of the cavity, which includes the loss of couplers, splices, tapered fiber, and the absorption loss due to solvent in case of a liquid sample. 𝛼𝑠 is the loss due to the sample absorption and is referred to as the sample loss. The envelope of the output pulses of the cavity is fitted with an exponential function to extract 𝜏. For a given cavity, 𝑡𝑟 and 𝛼𝑐 are known; 𝛼𝑠 is then calculated using (1). The effective number of round-trips — defined as the ratio (𝜏/𝑡𝑟 ) [23], is limited by the inherent loss of the cavity in the case of C-CRDT. A schematic of the experimental setup for amplified CRDT (referred to as A-CRDT henceforth) is shown in Fig. 2. An EDFA is additionally included in the cavity to compensate for the inherent cavity loss. A band pass filter (BPF) is used to filter the out-of-band ASE generated 187
K. Sharma et al.
Optics Communications 407 (2018) 186–192
Fig. 4. Schematic of a tapered fiber showing waist, transition and un-tapered fiber regions.
3. Experimental set-up In the set-up used for C-CRDT (Fig. 1), the 99% ports of the two couplers are spliced in the CRD cavity along with a tapered fiber. The 1% ports of the two couplers are used to inject pulses from a pulsed laser source (1550 nm, average power: ∼10 dBm, repetition rate: 200 kHz, and pulse width: ∼5 ns) and to extract the output. The round-trip time (length) of the cavity is 41.6 ns (8.32 m). The output pulses are detected with a photo-detector (Thorlabs DET01CFC, bandwidth: 1.2 GHz) and recorded with an oscilloscope (LeCroy WaveSurfer 44Xs, bandwidth: 400 MHz). It is to be noted that a slower detector can also be used for detecting the nano-second pulses used in the experiments. In the A-CRDT set-up (Fig. 2), a home-made EDFA, a tunable BPF (TECOS WTFM-1550-S-FA; bandwidth: 0.4 nm), and VOA are included along with two couplers with split ratios of 99:01 and a tapered fiber. The round-trip time (length) of the cavity is 324 ns (64.8 m). The repetition rate of the pulsed source is reduced to 10 kHz (average power: ∼0 dBm) to account for increase in length of the cavity due to additional components. The BPF is tuned to the signal wavelength (1550 nm). In the G-CRDT set-up (Fig. 3), two 50/50 couplers are used to couple light between the inner and outer loops. BPF 1 (TECOS TFM-1550-S-FA; bandwidth: 2 nm) and BPF 2 (TECOS WTFM-1550-S-FA; bandwidth: 0.4 nm) are used to choose the inner loop lasing wavelength (λlas = ∼1554.1 nm) and outer loop signal wavelength (λsig = 1550 nm) respectively. The output spectra of the inner and outer loops are monitored using an OSA (Yokogawa AQ6370). The repetition rate of the pulsed source is 10 kHz with average output power of −10 dBm. The round-trip time (length) of the outer loop is 360 ns (72 m). The wavelength of the CRD pulse from the pulsed laser source is maintained at 1550 nm for all the experiments. Schematic of the tapered fiber used in the experiments is shown in Fig. 4. A standard single mode fiber (Corning-SMF28e) is adiabatically tapered using the modified flame brushing technique to result in a reduced diameter fiber, with uniform waist in between two transition regions [32]. The transition regions convert the core mode of the standard single mode fiber to the cladding mode of the taper waist, with the confinement provided by the cladding-ambient interface [32]. During fabrication, fiber is fixed on two computer-controlled translation stages using magnetic clamps and a micro-heater is used to heat the fiber being stretched until the desired diameter is obtained. An image of a tapered fiber (5 μm diameter), taken using a microscope (Nikon Eclipse LV100) with a 50x objective is shown in Fig. 5(a). Image of an un-tapered fiber with 125 μm diameter taken using the same objective lens is shown in Fig. 5(b) for comparison. The loss of fibers is monitored in real-time during taper fabrication. The fabricated tapered fibers have an excess loss smaller than 0.4 dB. Sample solutions with different RI (1.316–1.344) are prepared by dissolving sugar in water with concentrations between 0% and 20%. RI of these solutions is measured using a standard refractometer (Metricon Corporation Model 2010 Prism Coupler) at 1550 nm.
Fig. 3. Schematic of the experimental set-up of fiber-based G-CRDT. The inner loop (with dotted lines) is designed to stabilize the gain whereas the outer loop (with solid lines) functions as the CRD cavity.
outer loop, irrespective of the input power [25]. If a small input signal is given to the inner loop, it gets amplified using the available gain at that wavelength in the EDFA, without disturbing the lasing action. To achieve a clamped gain in a G-CRD experiment, the nested-loop must be operated under the following conditions: (1) the input power of CRD pulse to the inner loop should be low enough to avoid any gain competition with λlas ; (2) the pump current in the inner loop must be maintained above the lasing threshold; (3) the loss in the outer loop must be larger than the gain available at λsig in the EDFA; (4) the wavelengths of the two loops need to be optimized. Hence, the optimization process relies on tuning of the parameters such as the input power of CRD pulse, the EDFA gain, the pump current of the EDFA, the inner and outer loop loss, and operating wavelengths of the inner and the outer loops. The possible wavelengths of operation of the two loops depend on the gain saturation characteristics of the EDFA and its optimization is discussed in [18,26]. In all three configurations of CRDTs, a tapered fiber is spliced in the CRD cavity, and is immersed in a sample solution. Increase in the concentration of the sample solution increases the RI of the sample surrounding the tapered fiber. This enhances the fraction of power in the evanescent field (𝜂𝑐𝑙𝑎𝑑 )–defined as ratio of power propagating outside the fiber and the total power—thus increasing the overlap of sample with the evanescent field [21,27]. When a sample (sugar solution in this case) absorbs at the wavelength of propagation, it induces loss in the light propagation path [28]. Increase in the overlap between the sample and the evanescent field, therefore, augments the loss induced by the sample [29], which in turn reduces the ring-down time [27]. This reduction in ring-down time helps in measuring the RI of the sample. The performance metric of a refractometric sensor is its detection limit (DL), defined as the minimum RI that can be accurately measured by the system. It is calculated using a method similar to the one used in [30,21]. For a CRDT experiment, the minimum detectable change in loss (Δ𝛼) that can be accurately measured by the system can be calculated as [31] Δ𝜏 , (3) 𝜏2 where Δ𝜏 is standard error in 𝜏; quantified as standard deviation of the ring-down time [30,21]. DL can be calculated as
Δ𝛼 = 𝑡𝑟
4. Results
Δ𝛼 , (4) 𝑆 where S is the slope of sample loss versus RI curve. DL can be improved either by reducing Δ𝜏 or by increasing 𝜏. DL is calculated for all the three techniques separately. DL =
4.1. Optimization of the three cavities In C-CRDT (Fig. 1), the inherent loss of the cavity should be small enough such that sufficient number of round-trips can be achieved to 188
K. Sharma et al.
Optics Communications 407 (2018) 186–192
Fig. 5. Optical microscope images of (a) tapered and (b) un-tapered fibers.
Fig. 6. Change in ring-down time and cavity loss with pump current of EDFA in A-CRDT.
Fig. 7. Variation of gain at the signal wavelength with change in the input signal power for different VOA 1 loss values in G-CRDT. Gain clamping is observed for different magnitudes of gain, albeit for different ranges of input powers.
increase the effective path length and ring-down time. The inherent loss for our set up is measured to be 0.5 dB and hence, the effective number of round-trips is ∼7. In A-CRDT (Fig. 2), the pump current of EDFA can be adjusted to achieve a desired gain such that it can almost compensate for the inherent loss in the cavity. Fig. 6 shows the change in ring-down time and cavity loss with pump current of EDFA. Large pump current increases gain in the cavity, reduces the total cavity loss and thus increases the ring-down time [10]. If the pump current is too high, the gain in the cavity overcomes the loss and the cavity starts to lase. Hence, the pump current (limited to 120 mA in this experiment) should be maintained below the lasing threshold of the cavity. The VOA is be used to fine-tune the cavity loss. The effective number of round-trips in this case is expected to be ∼40. Since the optimization of the nested-loop configuration in G-CRDT (Fig. 3) requires a detailed understanding and control of multiple parameters, it is initially performed without the tapered fiber in the cavity. The nested-loop is optimized again after splicing the tapered fiber, before performing the refractometric measurements. In each case, the optimization is performed successively for the inner and the outer loop. The input power of the CRD pulse to the inner loop, the loss of the inner loop, and the pump current of EDFA are optimized for the inner loop, in order to achieve a constant gain. For these optimized operating conditions, the loss of the outer loop is optimized such that the inner loop continues to lase while the CRD pulse propagating in the outer loop experiences the clamped gain. These optimization procedures are discussed in detailed below.
Fig. 8. Gain at the signal wavelength for different pump currents of EDFA and VOA 1 loss values in G-CRDT.
4.1.1. Optimization of the inner loop in G-CRDT The power of a CRD pulse in a CRD cavity changes with each roundtrip and hence gain clamping is studied as a function of input power. For this, only the inner loop is constructed to start with and a continuous wave (CW) signal at λsig is considered as input. The power of this CW signal represents the average power of the CRD pulse reaching EDFA in the inner loop (after the first 50/50 coupler). The pump current of EDFA is maintained at 190 mA in order to ensure that the inner loop is lasing. Gain clamping conditions, at different values of gain provided by the
Fig. 9. Spectra of the inner loop for pump currents of 100 mA and 160 mA and VOA 1 loss of 0.57 dB.
inner loop, are demonstrated. The loss of the inner loop is varied using VOA 1. Gain at λsig for different values of input power of the CW signal is recorded. The results are shown in Fig. 7. The gain values without 189
K. Sharma et al.
Optics Communications 407 (2018) 186–192
Fig. 11. Full spectra of the inner loop (with input signal on and off) in G-CRDT. Left and right insets show the zoomed version of the spectra at signal (outer loop) and laser (inner loop) wavelengths respectively.
Fig. 10. Spectra of the inner loop with different loss values of the outer loop, varied by VOA 2 in G-CRDT. The inset shows the zoomed version of the spectra.
feedback (i.e. inner loop not lasing hence, no clamping of gain) are also shown for comparison. It is found that, with the inner loop lasing, the gain remains constant for a wide range of input powers. While an increase in VOA 1 loss enhances the gain (as the inner loop is lasing), the range of input powers for which the gain is clamped, reduces. Therefore, VOA 1 can be used to set the gain required to compensate the inherent loss of the CRD cavity but the input power of CW signal (or the average power of the CRD pulse) to the inner loop has to be chosen carefully. All the gain curves in Fig. 7 coincide with the ‘without feedback’ curve when the input power of CW signal (at λsig ) reaching EDFA approaches the power at λlas circulating in the inner loop. At this point, the lasing in the inner loop is disturbed. Hence, the average input power of the CRD pulse should always be smaller than the power at λlas in the inner loop, in order to achieve gain clamping. Gain as a function of pump current is studied next, to show that the gain provided by the inner loop remains constant beyond its threshold pump current and to find the optimized value of the pump current. The gain at λsig is shown as a function of pump current for each VOA 1 loss value in Fig. 8. The input power of CW signal is maintained at −30 dBm. Fig. 8 shows that the threshold pump current, above which the gain at λsig remains constant, increases with the VOA 1 loss. Fig. 9 shows spectra of the inner loop at 100 and 160 mA pump currents for VOA 1 loss of 0.57 dB. As the pump current increases, the power at λlas (1554.1 nm) increases, while the ASE profile and the power at λsig (1550 nm) are clamped at constant values. This confirms that the gain (or inversion) in the EDFA remains constant for λsig considered beyond threshold pump currents, thus achieving gain clamping. However, increase in pump current can potentially lead to larger noise in the system. Thus, to minimize noise, the pump current should be chosen just above the lasing threshold of the inner loop.
Fig. 12. Dependence of the ring-down time and standard error in ring-down time on VOA 2 loss in G-CRDT.
the gain at λsig . Left and right insets show the zoomed version of the spectra at λsig and λlas respectively. When the loss of the outer loop is larger than the gain at λsig , the CRD pulse does not affect the lasing in the inner loop, but gets amplified. This would imply that there could be a range of values of VOA 2 loss, for which the above condition is satisfied. However, with the increase in VOA 2 loss, the ring-down time is expected to decrease, which is not desirable. On the other hand, operating closer to the lasing threshold (with VOA 2 loss very close to the gain experienced by the propagating CRD signal) would result in instabilities in the outer loop and lead to a larger error in ring-down time. In order to optimize the VOA 2 loss, 𝜏 and Δ𝜏 are measured for different values of VOA 2 loss and the results are shown in Fig. 12. Note that, the gain-clamped conditions are maintained throughout this experiment. When VOA 2 loss increases from 3.05 to 3.5 dB, Δ𝜏 reduces by an order of magnitude whereas 𝜏 reduces by a factor of half. Also, the number of effective round-trips reduces from ∼7.9 to ∼4.2. This would result in Δ𝛼 changing from 1.34 × 10−2 Np to 7.34 × 10−3 Np (calculated using (3)). The exact value of VOA 2 loss can be decided based on the requirement of Δ𝛼 for the intended application.
4.1.2. Optimization of the outer loop in G-CRDT After optimizing the average input power of CRD pulse, loss of the inner loop, and the pump current of EDFA, the loss of the outer loop has to be optimized. If the loss of the outer loop becomes smaller than the gain provided by EDFA at λsig , the outer loop starts lasing and disrupts the lasing in the inner loop. For this analysis, both the loops are connected together as shown in Fig. 3 and CRD pulse is given as input. Fig. 10 shows the spectra of the inner loop at center wavelength corresponding to λlas (inset shows the zoomed version of the spectra) as a function of loss of the outer loop (controlled by VOA 2). A loss corresponding to 3.02 dB in the outer loop is smaller than the gain provided by EDFA at λsig , thus disrupting the lasing in the inner loop. At 3.03 dB, the loss and gain are comparable, whereas at 3.04 dB, the inner loop starts lasing at λlas , since the loss of the outer loop is greater than the gain provided by EDFA at λsig . Thus, the loss in the outer loop needs to be optimized carefully such that it does not lase and disrupt lasing in the inner loop. Fig. 11 shows full spectra of the inner loop with input signal (CRD pulse) on and off, when loss of the outer loop is maintained larger than
4.2. Refractive index sensing For performing RI sensing, tapered fibers of length 3 mm and diameter ∼5 μm are spliced in all the three CRD cavities. The nestedloop of G-CRDT is optimized according to the procedures mentioned in Section 4.1. The optimized values of the input average power of CRD pulse to the outer cavity (output of the pulsed laser source), pump current, VOA 1, and VOA 2 loss used for G-CRDT experiment are −10 dBm, 165 mA, 10.5 dB, and 4.1 dB respectively. Note that the inherent loss of the outer cavity has changed with addition of the fiber taper. Hence, these values are slightly different from those discussed in section 4.2, where the optimization is performed without the fiber taper. A few μl of sugar solution is poured onto the taper surface, ensuring that the taper waist and transition regions are fully immersed inside the 190
K. Sharma et al.
Optics Communications 407 (2018) 186–192
RI, pre-calibration of the sensor system is required. To estimate DL of the system, a smaller range of RI (1.322–1.33) is chosen in which the sample loss versus RI curve follows a linear function (shown in the inset of Fig. 14 for all the three techniques). The slope values of these linear curves are used to calculate DL in the next section. The change in ringdown time versus RI is shown in Fig. 15 for all the three configurations of CRDT. The dependence of ring-down time with RI is found to be a rational function since 𝜏 is inversely proportional to the loss [27]. In Fig. 14, the sample loss versus RI curves for all the three techniques are expected to overlap as the sample loss corresponds to the loss induced by the sample in a single round-trip. However, we find a deviation for the G-CRDT case. This discrepancy can be explained as follows. Even after using the same fabrication parameters, the actual diameter of the tapered fiber used for G-CRDT experiment was found to be slightly smaller (∼4.5 μm) than that used (∼5 μm) in the C-and A-CRDT experiments. This results in a different performance for the GCRDT, as evidenced by the smaller number of pulses than A-CRDT in Fig. 13 and a different fitting function for the sample loss versus RI curve in Fig. 14. The slope of the sample loss vs RI curve in RI range 1.322–1.33 is found to be 5.8 Np/RIU and 6.2 Np/RIU for C- and ACRDT respectively while it is 30.1 Np/RIU for the G-CRDT. Smaller taper diameter results in poor confinement of mode which leads to larger 𝜂𝑐𝑙𝑎𝑑 . This increases the overlap of sample with the evanescent field and hence, the loss. Therefore, tapered fibers with smaller diameter are more sensitive to RI change. The variation of 𝜂𝑐𝑙𝑎𝑑 with different taper diameters has been studied earlier by our group [21,27]. It is to be noted that in an actual sensing experiment, a differential measurement for sample concentration (or refractive index) would be made. Hence, the absolute value of sample loss would not matter.
Fig. 13. Time-domain traces captured by oscilloscope with water as a sample for C-, Aand G-CRDT with tapered fiber of length 3 mm and diameter ∼5 μm.
4.3. Detection limit Fig. 14. Change in sample loss with refractive index in C-, A- and G-CRDT. The experimental data points are fitted with quadratic function. The data points corresponding to a smaller range of RI (1.322–1.33) are fitted with a linear function (shown in inset) to determine DL for all the three techniques.
The standard error in ring-down time (Δ𝜏) is calculated using a set of 50 data points, after averaging 1024 ring-down traces in the oscilloscope for each datum. This Δ𝜏 is not influenced by the fact that the taper diameter used for G-CRDT is slightly different from that used in C- and A-CRDT as it represents the noise of the whole system. Then, Δ𝛼 is calculated using (3). The DL calculated at RI of 1.322 (using (4)) for all the three techniques is given in Table 1 along with 𝜏, Δ𝜏, Δ𝛼 and the slope values. The standard error in ring-down time in C-CRDT is 4.66 ns, the lowest among all, but associated with a very small ring-down time (0.092 μs). On the other hand, the ring-down times in the A- and G-CRDTs are longer but at the expense of larger Δ𝜏 − 450 ns and 50.5 ns respectively. Comparing Δ𝜏 in A- and G-CRDTs, the latter has approximately 10 times smaller Δ𝜏 due to gain clamping. The detection limits achieved at a refractive index of 1.322, with tapered fibers of length 3 mm and diameter 5 μm in C- and A-CRDT are 3.91 × 10−3 RIU and 1.32 × 10−3 RIU respectively. DL achieved in G-CRDT at the same refractive index, with a tapered fiber of length 3 mm and diameter 4.5 μm is 1.03 × 10−4 RIU. This improvement in DL in G-CRDT is contributed primarily due to the reduction in Δ𝜏, coupled with an increase in the slope of sample loss versus RI curve because of the use of tapered fiber with a slightly different diameter. If tapered fibers with identical dimensions were used in all the three CRDTs, then the slope would remain the same and the best DL (5.15×10−4 RIU) would be still be achieved by G-CRDT. The DL would be much worse for all the three cases, when calculated at higher RI values as the sample loss increases and the ring-down time reduces.
Fig. 15. Change in ring-down time with refractive index in C-, A- and G-CRDT. The experimental data points are fitted with rational function.
solution. The tapered fiber is rinsed with water twice before being used with another sample. The time-domain traces are recorded with the oscilloscope with 1024 averages for each ring-down trace as shown in Fig. 13. These timedomain traces are fitted with an exponential function to extract the ringdown time. Then, the sample loss in Nepers (Np) is calculated using (1) and (2). The change in sample loss versus RI is plotted in Fig. 14 for C-, A-, and G-CRDTs. The sample loss versus RI curves follow a quadratic function (indicated by the fit shown in Fig. 14). This is due to the fact that 𝜂𝑐𝑙𝑎𝑑 shows a quadratic behavior for this large range of RI (1.316– 1.344) [27] and sample loss is directly proportional to 𝜂𝑐𝑙𝑎𝑑 [29]. Due to non-linear dependence of the sample loss on RI in this large range of
5. Conclusion The performance and DL of conventional, amplified, and gainclamped CRDTs have been compared experimentally by measuring RI of sugar solutions using tapered fibers as sensing heads. It is found that Δ𝜏 in conventional CRDT is minimum but with very small 𝜏 value. 𝜏 is 191
K. Sharma et al.
Optics Communications 407 (2018) 186–192 Table 1 Comparison of conventional, amplified, and gain-clamped CRDTs. Parameter
Conventional CRDT
Amplified CRDT
Gain-clamped CRDT
Ring-down time (𝜏) (μs) Standard error in ring-down time (Δ𝜏) (ns) Minimum detectable change in loss (Δ𝛼) (Np) Slope (Np/RIU) Detection limit at RI of 1.322 (RIU)
0.092 4.66 2.27 × 10−2 5.8 3.91 × 10−3
4.22 450 8.17 × 10−3 6.2 1.32 × 10−3
2.45 50.5 3.09 × 10−3 30.1 1.03 × 10−4
increased by two orders of magnitude in the case of amplified CRDT, but at the expense of an increased Δ𝜏. On the other hand, gain-clamped CRDT is found to have 10 times lower Δ𝜏 than amplified CRDT in addition to the increased 𝜏. In terms of complexity of the techniques, conventional CRDT is simpler as compared to the other two techniques. The gain of the EDFA and the total cavity loss in amplified CRDT has to be optimized every time, based on the inherent loss of the cavity. Gainclamped CRDT is by far the most complex technique, as parameters like input power of CRD pulse, pump power, cavity loss of the inner and outer loops have to be optimized before performing the measurements. The procedure for optimizing these parameters in gain-clamped CRDT is discussed in detail. The best DL achieved in refractometric sensing is 1.03 × 10−4 RIU at RI of 1.322 with a 3 mm long and 4.5 μm diameter tapered fiber in gain-clamped CRDT. The minimum detectable loss that can be detected by the system is 3.09 × 10−3 Np. These values can be improved by using components with smaller loss — for instance, a 99.9/0.1 coupler instead of a 99/01 coupler; and also by fabricating tapered fibers with lower insertion loss. Use of a low noise EDFA would help in achieving smaller values of Δ𝜏 in amplified CRDT, further improving minimum detectable loss and DL. Signal to noise ratio can be improved in all the three techniques by using a trans-impedance amplifier at the output of the photo-detector. From our study, gain-clamped CRDT appears to be the most sensitive technique among all three, which can be used for RI measurements with advantages such as long effective path lengths, good sensitivity, and insensitivity to the intensity fluctuations of the light source.
[7] Z. Tong, M. Jakubinek, A. Wright, A. Gillies, H.P. Loock, Fiber-loop ring-down spectroscopy: a sensitive absorption technique for small liquid samples, Rev. Sci. Instrum. 74 (2003) 4818–4826. [8] H. Waechter, J. Litman, A.H. Cheung, J.A. Barnes, H.P. Loock, Chemical sensing using fiber cavity ring-down spectroscopy, Sensors 10 (2010) 1716–1742. [9] G. Stewart, K. Atherton, H. Yu, B. Culshaw, An investigation of optical fibre amplifier loop for intra-cavity and ring-down cavity loss measurements, Measure. Sci. Technol. 12 (2001) 843–849. [10] G. Stewart, K. Atherton, B. Culshaw, Cavity-enhanced spectroscopy in fiber cavities, Opt. Lett. 29 (2004) 442–444. [11] G. Stewart, P. Shields, B. Culshaw, Development of fibre laser systems for ring-down and intracavity gas spectroscopy in the near-IR, Meas. Sci. Technol. 15 (2004) 1621– 1628. [12] N. Ni, C.C. Chan, Improving the measurement accuracy of CRD fibre amplified loop gas sensing system by using a digital LMS adaptive filter, Meas. Sci. Technol. 17 (2006) 2349–2354. [13] N. Ni, C.C. Chan, T.K. Chuah, L. Xia, P. Shum, Enhancing the measurement accuracy of a cavity enhanced fiber chemical sensor by an adaptive filter, Meas. Sci. Technol. 19 (2008) 115203:1–115203:6. [14] K. Atherton, G. Stewart, B. Culshaw, Gas detection by cavity ring-down absorption with a fibre optic amplifier loop, Proc. SPIE 4577 (2002) 25–31. [15] J. Meng, Z. Weigang, Z. Qi, L. Yaping, L. Bo, Investigation on an evanescent wave fiber-optic absorption sensor based on fiber loop cavity ring-down spectroscopy, Opt. Commun. 283 (2010) 249–253. [16] A. Yolalmaz, F.H. Sadroud, M.F. Danışman, O. Esenturk, Intracavity gas detection with fiber loop ring down spectroscopy, Opt. Commun. 396 (2017) 141–145. [17] S. Silva, O. Frazão, Multimode interference-based fiber sensor in a cavity ring-down system for refractive index measurement, Opt. Laser Technol. 91 (2017) 112–115. [18] N.L.P. Andrews, J. Litman, D. Stroh, J.A. Barnes, H.P. Loock, Near-infrared absorption detection in picolitre liquid volumes using amplified fibre loop ring-down detection, Opt. Fiber Technol. 19 (2013) 822–827. [19] D. Wu, Y. Zhao, Q. Wang, SMF taper evanescent field-based RI sensor combined with fiber loop ring down technology, IEEE Photonic. Tech. L. 27 (2015) 1802–1805. [20] X. Liu, Q. Wang, C. Li, C. Zhao, Y. Zhao, H. Hu, J. Li, Refractive index sensor based on fiber loop ring-down spectroscopy, Instrum. Sci. Technol. 44 (2016) 241–248. [21] K. Sharma, S. Liang, S.U. Alam, S. Bhattacharya, D. Venkitesh, G. Brambilla, Fiberbased cavity ring-down technique for refractive index sensing at 1953 nm using tapered fibers, IEEE Sensors Lett. 1 (2017) 1–4. [22] P. Wang, G. Brambilla, M. Ding, Y. Semenova, Q. Wu, G. Farrell, High-sensitivity, evanescent field refractometric sensor based on a tapered, multimode fiber interference, Opt. Lett. 36 (2011) 2233–2235. [23] G. Gagliardi, H.-P. Loock, in: Cavity-Enhanced Spectroscopy and Sensing, vol. 179, Springer-Verlag, 2014. [24] S. Silva, R. Magalhães, R.A. Pérez-Herrera, M. Lopez-Amo, M.B. Marques, O. Frazão, Fiber cavity ring down and gain amplification effect, Photonic Sens. 4 (2016) 324– 327. [25] M. Zirngibl, Gain control in erbium-doped fibre amplifiers by an all-optical feedback loop, Electron. Lett. 27 (1991) 560–561. [26] H. Okamura, Automatic optical loss compensation with erbium-doped fiber amplifier, J. Lightwave Technol. 10 (1992) 1110–1116. [27] K. Sharma, D. Venkitesh, S. Bhattacharya, B. Srinivasan, G. Brambilla, Non-linear behavior of ring-down time in cavity ring-down spectroscopy with tapered fibers, in: Conference on Lasers and Electro-Optics, OSA Technical Digest (online, Optical Society of America, 2016, paper JW2A.12. [28] A. Leung, P.M. Shankar, R. Mutharasan, A review of fiber-optic biosensors, Sensors Actuators B 125 (2007) 688–703. [29] Z. Wei, Z. Song, R. Song, X. Zhang, Z. Meng, Measurement of the optical absorption coefficient for liquid based on optical microfiber, Optik 125 (2014) 2880–2884. [30] K. Zhou, D.J. Webb, C. Mou, M. Farries, N. Hayes, L. Bennion, Optical fiber cavity ring down measurement of refractive index with a microchannel drilled by femtosecond laser, IEEE Photonics Technol. Lett. 21 (2009) 1653–1655. [31] C.M. Rushworth, D. James, J.W.L. Lee, C. Vallance, Top notch design for fiber-loop cavity ring-down spectroscopy, Anal. Chem. 83 (2011) 8492–8500. [32] G. Brambilla, Optical fibre nanotaper sensors, Opt. Fiber Technol. 16 (2010) 331– 342.
Acknowledgments Funding: This work was supported by the People Programme (Marie Curie Actions) of the European Unions Seventh Framework Programme FP7/2007–2013/ under REA grant agreement n0318941. The authors thank B. Srinivasan, M. Srivastava, Y. Panbiharwala, and J. Barnes for helpful discussions; M. Turvey, Dr. P. Hua, and Dr. P. Smith for their help in measuring refractive indices using refractometers; Y. Panbiharwala for proofreading the paper. They also thank Dr. J. He, Dr. S. Ganapathy, V. Mittal, H. A. Vasant, M. F. Namiq, M. B. Alonso, S. Liang, N. Thipparapu, K. Bottrill, F. H. Suhailin, S. Jain, and Dr. A. Masoudi for providing several components. References [1] J.J. Scherer, J.B. Paul, A. O’Keefe, R.J. Saykally, Cavity ringdown laser absorption spectroscopy: history, development, and application to pulsed molecular beams, Chem. Rev. 97 (1997) 25–51. [2] S.O. Silva, R. Magalhães, M.B. Marques, O. Frazão, New advances in fiber cavity ring-down technology, Opt. Laser Technol. 78 (2016) 115–119. [3] G. Berden, R. Peeters, G. Meijer, Cavity ring-down spectroscopy: experimental schemes and applications, Rev. Phys. Chem. 19 (2009) 565–607. [4] J.M. Herbelin, J.A. McKay, M.A. Kwok, R.H. Ueunten, D.S. Urevig, D.J. Spencer, D.J. Benard, Sensitive measurement of photon lifetime and true reflectances in an optical cavity by a phase-shift method, Appl. Opt. 19 (1980) 144–147. [5] A. O’Keefe, D.A.G. Deacon, Cavity ring-down optical spectrometer for absorption measurements using pulsed laser sources, Rev. Sci. Instrum. 59 (1988) 2544–2551. [6] R.S. Brown, I. Kozin, Z. Tong, R.D. Oleschuk, H.P. Loock, Fiber-loop ring-down spectroscopy, J. Chem. Phys. 117 (2002) 10444–10447.
192