Titled fiber Bragg grating (TFBG) sensors have at-tracted considerable attention in the past decade be-cause they offer potentialities without equivalent forany other kind of fiber gratings. A single TFBG in-deed provides temperature-insensitive surroundingrefractive index (SRI) [1], axial strain [2], or bendingmeasurements [3]. Recently, new application fieldshave been proposed based on nanoscale-coatedTFBGs. Covering TFBGs with specific coatings in-deed allows one to tailor their spectral characteris-tics or to develop biochemical sensors based onsurface plasmon resonances (SPR) when metallicor functionalized overlays are employed [4,5].
TFBGs belong to the short-period gratings family(grating periodicity ∼500nm) and are characterizedby a refractive-index modulation blazed with respectto the fiber axis. In TFBGs, both a core mode reso-nance and several cladding-mode resonances appearsimultaneously in transmission [1]. This has severaladvantages. The cladding modes are sensitive toexternal perturbations (e.g., temperature, strain, re-fractive index, bending), while the core mode is onlysensitive to temperature and strain. In practice, thetemperature sensitivity of the cladding modes issimilar to that of the core mode. The temperatureinfluence can thus be removed from a comparisonbetween the shifts of the core mode and the selectedcladding modes [2]. Temperature-independent sen-sors can therefore be realized without requiring anyadditional device for compensation.
Up to now, two main techniques have been pro-posed to demodulate the transmitted spectrum of a
TFBG subject to a perturbation. The first one, basedon global monitoring of the area delimited by all thecladding modes in transmission, is particularlysuited for SRI measurements [1]. At this level, amonitoring of spectra partitions depending on thecladding-mode order provides a valuable tool forsensitivity tailoring [6]. An alternative methodologyrelies on monitoring selected cladding-mode reso-nance shifts with respect to the Bragg wavelength.Such a demodulation technique has been demon-strated not only for refractometry [7] but also forstrain measurement purposes [2].To correctly operate, these techniques require
spectral measurements of a few tens of nanometers.Within the context of quasi-distributed sensing, thismakes the wavelength division multiplexing (WDM)of TFBG sensors interrogated by such techniquesdifficult. For this reason, TFBG quasi-distributedsensing is mainly based on parallel network config-urations. However, it is impossible to simultaneouslyaddress all the sensors using a tunable or broadbandlaser source as the different TFBGs transmittedspectra overlap. Time division multiplexing (TDM)is required, which increases the complexity and totalresponse time. In addition, because they are based onaccurate spectral investigation, the aforementionedtechniques prevent real time interrogation.Hence, although TFBGs have demonstrated a ca-
pability to be used for many sensing applications, theproposed demodulation techniques severely limittheir use in industrial applications. To solve this lim-itation, an interesting solution consists in stronglyreducing the useful bandwidth of the signal providedby the TFBG sensors. This would allow managingparallel TFBG sensor networks, where each gratingprovides a narrowband signal carrying the sensedinformation without overlap with signals comingfrom other elements. In this case, commercialWDM components and intensity-based measure-ments can be readily used for real time networkinterrogation, enabling TFBG exploitation at theindustrial level.We demonstrate in this paper that a fiber-grating-
based hybrid configuration working in reflectionmode can be efficiently used for this purpose. Itenables the encoding of the sensed parameter in anarrow spectral range and provides a simple de-modulation based on intensity measurements. It alsooffers the possibility to select the spectral range foreach TFBG, allowing tailoring of the sensitivitythrough the selection of the proper cladding modes,taking into account their correct spectral position forWDM demodulation. In the remainder of the paper,we first present the main characteristics of the pro-posed hybrid configuration and then we focus on itsperformances in the frame of SRI sensing. SRI sen-sing in liquids is indeed of high importance, not onlyto characterize their density but also in practical ap-plications, such as chemical species concentrationsmonitoring, pH monitoring, and liquid level sensing.
2. Gratings Characteristics
TFBGs were written into hydrogenated standardsingle-mode fiber using a frequency-doubled argon-ion laser and uniform phase masks. To generatethe tilted refractive-index modulation, the phasemask was placed on a rotating stage so that it canbe tilted in the plane perpendicular to the laserbeam. For the experiments reported here, small tiltangles were used so that both cladding modes andcore mode resonances appear in transmission whilea residual core mode resonance is present in reflec-tion at the right side of the spectrum, as shownin Fig. 1.
To modify the SRI around the TFBGs, we usedboth water–sugar solutions and a set of Cargille oilswhose refractive indices around 1550nm are knownwith an accuracy of the order of 10−3 in terms ofrefractive-index unit (RIU). To keep the strain onthe TFBGs constant during the experiments, theTFBGswere attached to amicroscope slide and smallquantities of liquids with various refractive indiceswere deposited on the TFBGs. Note that the fiberis bare at the grating location, which requires theTFBGs to be handled with care.
The core mode resonance wavelength, λBragg, andthe wavelength at which the discrete coupling tothe ith cladding mode occurs, λcoupling;i, are givenby the following set of equations [8]:
λBragg ¼ 2neff ;coreΛg
cos θ ; ð1Þ
λcoupling;i ¼ ðneff ;clad;i þ neff ;coreÞΛg
cos θ ; ð2Þ
where neff ;core and neff ;clad;i are the effective refractiveindices of the core mode and the ith cladding mode,respectively. Λg denotes the nominal grating periodand is such that Λg ¼ Λ cos θ, where Λ is the period-icity along the fiber axis. Each cladding-moderesonance is therefore characterized by its own effec-tive refractive index and will respond differently to a
Fig. 1. Spectra of a 5° TFBG versus SRI measured at 25 °C.
SRI variation. As shown in Fig. 1, the cladding-moderesonance strengths progressively decrease whenSRI increases. This results from the fact that, whenthe SRI rises and reaches neff ;clad;i, the correspondingith cladding mode becomes weakly guided, reducingits resonance amplitude. When the SRI is equal toneff ;clad;i, the ith cladding is no longer guided butbecomes radiated. Hence, as the SRI grows, aprogressive smoothing of the transmitted spectrumis obtained, starting from the short wavelengths.Figure 1 also shows that the cladding-modes reso-nances extend on nearly 60nm.To get a narrowband reflected signal, each TFBG
was followed by a uniform FBG. The refractive-indexmodulation of the latter was chosen to yield an∼100% reflectivity. As the FBG reflection coverssome TFBG cladding-mode resonances, it presentsa modulation depending on the optical losses dueto TFBG cladding-mode couplings. Hence, any per-turbation that affects the TFBG cladding-mode cou-plings, such as the SRI, bending, or torsion, modifiesthe optical power carried out by the reflected signal.Figure 2 illustrates how the reflection spectrum
from a 1mm long FBG located behind a 5° TFBG(same TFBG as in Fig. 1) evolves when the sensoris immersed in liquids with various SRI.As the SRI increases, the optical losses due to
cladding-mode couplings tend to disappear, leadingto a reflected spectrum where the narrowband anddeep attenuation notches in the TFBG transmittedspectrum are replaced by a broadband low-levelattenuation that gradually extends across the fullspectrum. As a result, the reflected power from thesensor monotonically decreases for a certain rangeof SRI. Therefore, the SRI can be unambiguouslyobtained from the reflected power monitoring withinthe FBG bandwidth. Compared with [9], wherepower measurement was first proposed to demodu-late the TFBG transmitted spectrum for SRI sensingpurposes, our configuration does not require spectralanalysis and works in reflection. Also, the signifi-
cantly reduced useful bandwidth per sensor offersstraightforward multipoint sensing capability.
3. Sensor Implementation
Figure 3 presents the practical implementation ofthe proposed parallel network for multipointsensing. The light from an ASE source coveringthe Cþ L bands is launched through an optical cir-culator and a splitter to a parallel network of opticalfibers. Each fiber contains a TFBG followed by a uni-form FBG so that each sensor provides a reflectedsignal whose bandwidth is delimited by the FBG.In this way, each reflection from the parallel networkprovides a narrowband signal plus the residualTFBG Bragg reflection at longer wavelengths. Thelatter can be used for thermal compensation whenrequired. In the parallel network, the operatingwavelength λ0 and bandwidth of the FBGs as wellas the tilt angle θ of the TFBGs were designed toenable a correct spectral allocation with a goodSRI sensitivity and to avoid the overlap betweenthe different channels of the network. As it influ-ences the sensing performances, this design is ad-dressed in more detail in Section 4. Behind thecirculator, a WDM tunable bandpass filter allowsthe selection of the different gated reflected signalsso that they can be processed independently by apowermeter. The bandwidth of the filter was chosenslightly higher than that of the FBGs to cover all thereflected cladding-modes resonances.
4. Experimental Results
Since each cladding mode presents its own SRI sensi-tivity, the central wavelengths λ0 and bandwidths ofthe FBG mirrors dictate the SRI range to which thereflected spectrum is sensitive [6]. In our experi-ments, we fixed the length of the mirrors to 1mmso that the main reflection band has a full width ofabout 3nm and covers several cladding-moderesonances. The wavelength spacing between twoadjacent cladding-mode resonances is indeed of theorder of several hundreds of picometers. The effectof the central wavelength on the SRI sensitivitywas analyzed for the 5° TFBG used above. For thatpurpose, the TFBG was spliced to different uniform
Fig. 2. Reflected spectrum evolution for a 5° TFBG followed by a1mm long FBG centered around 1515nm when the SRI evolves.
Fig. 3. Schematic of the experimental setup in which a TFBGparallel network works in reflection.
FBGs characterized by the same reflectivity but bydifferent centralwavelengths. For each configuration,the reflected power wasmeasured as a function of theSRI. Figure 4 compares the evolution against SRI ofthenormalized reflectedpower in theFBGbandwidthfor different operating wavelengths λ0 of the mirrorwith respect to the Bragg wavelength of the TFBGλB. This normalized power is defined as the ratiobetween the reflected power measured at a givenSRI and the one of a reference reflected spectrum(SRI ¼ 1:33). One can see that all the curves mono-tonically decrease as a function of the SRI value.However, they present different dynamics so thatthey are characterized by their own sensitivity tothe SRI, defined as the variation of the normalizedreflected power per SRI unit. In the following, we talkabout mean sensitivity when the complete investi-gatedSRI range is considered (the sensitivity is there-fore averaged) and about maximum sensitivity whenit is computed in a limited SRI range where the var-iation of the reflected power per SRI unit ismaximumand relatively constant.When looking to the obtainedresults, it is worth noting that the SRI sensitivityobtained in a narrow window of several nanometersis substantially different from the one linked to a glob-al power spectrum measurement, as reported in [9].This comes from the fact that each cladding modeis characterized by its own effective refractive indexand exhibits itsmaximumSRI sensitivity for SRI val-ues approaching its effective refractive-index value.Hence, as expected from [1,7], the longer wavelengthspresent their maximum sensitivity at higher SRI val-ues (>1:40), and the smaller ones exhibit their max-imum sensitivity at smaller SRI values (<1:40). Thisstrong difference in behaviors weakens when thewhole TFBG spectrum is processed at once: it leadsto an important averaging effect in the SRI sensitiv-ity, which is avoided when a couple of cladding-modesresonances is analyzed independently. Hence, for agiven tilt angle, the position of the interrogationwindow with respect to the TFBG Bragg wavelength
is a parameter of high importance that can be used totailor the SRI sensitivity. In particular, for wave-length ranges matching the middle of the cladding-modes spectrum (λB − λ0 around 20nm in our case),a good mean SRI sensitivity is obtained in the SRIrange between 1.33 and 1.44. It has been computedequal to about 0.09 per 0.01 RIU, while the maximumsensitivity reaches 0.15 per 0.01 RIU in the range of1:42–1:44. To maximize the SRI range for which agiven sensor can operate, the mirror should thus bepreferentially located in the middle of the cladding-modes spectrum. To keep this feature in a parallelnetwork while avoiding the overlap between the re-flected spectra, one can play on θ to shift the claddingmodes [1].
Figure 5 depicts, for a 3nm wide mirror centeredaround 1525nm, the effect of θ on the SRI sensitivityof the reflectedspectrum.ThedifferentTFBGsused toobtain these data were characterized by similar clad-ding-modes-resonances strengths in the1525nmwin-dow.Theywerewritten throughdifferent orientationsof the phasemaskwith the sameoptical power but dif-ferent inscription times. The resonances were similarin strengthbutnot inorder, as increasing the tilt anglemodifies the cladding-modes couplings and shiftstheir resonance wavelengths toward longer wave-lengths. Figure 5 confirms that θ can be used in addi-tion to λ0 to tailor the SRI sensitivity. As the tilt angledecreases, the sensitivity to higher SRI valuesincreases. The ranges of maximum sensitivity lie be-tween 1.40 and 1.44, while theminimum sensitivitiesare obtainedat the endsof the investigatedSRI range.
Based on these considerations, we have developeda parallel network with eight sensors. The wave-length spacing between two adjacent mirrors wasset to ∼10nm to avoid any overlap even with second-ary lobes. Care was also taken in the wavelengthallocation to avoid interferences of the useful signalswith the TFBGs’ Bragg reflections. The tilt angle andcladding-modes resonances of each TFBG were ad-justed with respect to their corresponding mirror
Fig. 4. (Color online) Evolution of the reflected power in an∼3nm window as a function of the SRI for a 5° TFBG (Braggwavelength λB ∼ 1543nm) at different operating wavelengths λ0.
Fig. 5. (Color online) Evolution of the reflected power in an∼3nmwindow as a function of the SRI for different TFBGs around1525nm.
to get a good mean SRI sensitivity. Prior to its use inthe network, each sensor was calibrated so that themean SRI sensitivity is known for every sensor.Figure 6 presents the global reflected spectrum fromthe parallel network where the different TFBGswereplaced in different SRI environments. To obtain thisspectrum, the tunable filter and the powermeter ofthe experimental setup presented in Fig. 3 werereplaced by an optical spectrum analyzer with theresolution in wavelength set to 0:1nm. One cansee that the gratings were designed so that thereis no overlap between the different signals. TheBragg wavelengths of the two last TFBGs (TFBGs7 and 8 in Fig. 6) fall outside the wavelength rangeof the optical source and are consequently not visiblein Fig. 6.Figure 7 shows a typical power measurement
obtained on a given TFBG via the bandpass filter.The error bars represent the maximum error onthe SRI value for a repeatability test conducted in10 measurements. For this sensor, a mean sensitivity
of 55 μW=0:01 RIU was obtained in the range of1:33–1:45. The maximum sensitivity has been com-puted equal to about 120 μW=0:01 RIU in the rangeof 1:41–1:43. Taking into account all the sensors, themaximum error on the SRI value was computed ofthe order of 10−3.
Let us also add that the sensitivity can be furthertailored via the FBG bandwidth as it modifies thenumber of cladding-mode resonances that arecomprised in the reflection band.
Finally, the temperature influence on the hybridconfiguration was also investigated in the range of30 °C–120 °C. Figure 8 shows that the reflected powerfrom a hybrid configuration made of a 5° TFBG im-mersed in SRI ¼ 1:00 and followed by a uniformFBG centered around 1519nm (λB − λ0 ¼ 24nm) re-mains stable when the temperature evolves. Themaximum difference between the measured powershas been computed equal to 6 μW and is thus equiva-lent to an error on the determination of the SRI valueof about 10−3. Our experiments thus confirm that thepower measurements are inherently insensitive totemperature. This results from the fact that a TFBGand its corresponding FBG are closely spaced (lessthan several centimeters) and that they exhibit thesame temperature sensitivity. As they are placed inthe same temperature environment, the relativepositions between the TFBG cladding-modes reso-nances and the FBG reflected spectrum do notchange, which do not modify the reflected power.Hence, the SRI measurements are temperature in-sensitive and the information about temperaturecan be deduced from the TFBG Bragg reflection.Let us alsomention that the design of the parallel net-work takes into account the possibility that the differ-ent sensors experience different thermal conditions.The wavelength spacing between the sensors’elements is indeed sufficiently high to avoid any over-lapping due to different thermal behaviors.
Fig. 6. Total reflected spectrum from the parallel networkmeasured with an optical spectrum analyzer.
Fig. 7. Evolution of the reflected power as a function of the SRIfor a 5° TFBG followed by a 3nm wide FBG centered around1519nm (λB − λ0 ¼ 24nm).
Fig. 8. Temperature influence on a hybrid configuration for a 5°TFBG immersed in SRI ¼ 1:00 and followed by a 3nm wide FBGcentered around 1519nm (λB − λ0 ¼ 24nm).
We have presented a novel architecture for multi-point sensing with TFBGs that offers fast, simple,and cost-effective measurements. Our solution usesa uniform FBG behind every TFBG so that wave-length-gated reflected signals are obtained and inde-pendently demodulated by power measurements.The gratings’ characteristics were tailored to allowproper allocation of the wavelength range as wellas good SRI sensitivity. A temperature-insensitivebehavior and a maximum error of 10−3 on the SRIvalue have been reported.
Christophe Caucheteur is supported by the FondsNational de la Recherche Scientifique (F.R.S.-FNRS).The Belgian authors acknowledge the financial sup-port of the Attraction Pole Program of the BelgianScience Policy, IAP6/10.
References
1. G. Laffont and P. Ferdinand, “Tilted short-period fiber-Bragg-grating induced coupling to cladding modes for accuraterefractometry,” Meas. Sci. Technol. 12, 765–770 (2001).
2. C. Chen and J. Albert, “Strain-optic coefficients of theindividual cladding modes of a single mode fiber: theoryand experiment,” Electron. Lett. 43, 21–22 (2006).
3. S. Baek, Y. Jeong, and B. Lee, “Characteristics of short-periodfiber Bragg gratings for use as macro-bending sensors,” Appl.Opt. 41, 631–636 (2002).
4. D.Paladino,A.Cusano, P. Pilla, S.Campopiano,C.Caucheteur,and P. Mégret, “Spectral behavior in nano-coated tilted fiberBragg gratings: effect of thickness and external refractiveindex,” IEEE Photon. Technol. Lett. 19, 2051–2053 (2007).
5. Y. Shevchenko and J. Albert, “Plasmon resonances ingold-coated tilted fiber Bragg gratings,” Opt. Lett. 32, 211–213 (2007).
6. C. Caucheteur, D. Paladino, P. Pilla, A. Cutolo, S. Campopiano,M. Giordano, A. Cusano, and P. Mégret, “External refractiveindex sensitivity of weakly tilted fiber Bragg gratings withdifferent coating thicknesses,” IEEE Sens. J. 8, 1330–1336(2008).
7. C. Chan, C. Chen, A. Jafari, A. Laronche, D. J. Thomson, andJ. Albert, “Optical fiber refractometer using narrowband clad-ding-mode resonance shifts,” Appl. Opt. 46, 1142–1149 (2007).
8. T. Erdogan and J. E. Sipe, “Tilted fiber phase gratings,” J. Opt.Soc. Am. A 13, 296–313 (1996).
9. T. Guo, C. Chen, A. Laronche, and J. Albert, “Power-referencedand temperature-calibrated optical fiber refractometer,” IEEEPhoton. Technol. Lett. 20, 635–637 (2008).
