Ž .Sensors and Actuators B 62 2000 49–54www.elsevier.nlrlocatersensorb

Developing stable optical fiber refractometers using PMDI withtwo-parallel Fabry–Perots

Yu-Lung Lo ), Hsin-Yi Lai, Wern-Cheng WangDepartment of Mechanical Engineering, National Cheng Kung UniÕersity, Tainan 70101, Taiwan

Received 31 May 1999; accepted 20 September 1999

Abstract

The paper presents a new approach to developing an optical fiber refractometer. The objective of the study is to come up with arelatively inexpensive but reliable optical refractometer that can be used to measure the change of refractive index in a resolution of 10y5

and to work in a dynamic range up to 6=10y3 at a DC frequency of up to 100 Hz. It is known that the phase modulations of opticalfiber sensors are very sensitive to external disturbances, especially to the effects of thermal drifts or vibrations. A cancellation techniqueto compensate the effect of variation on a PZT stack is proposed in this paper to stabilize the system. Two parallel Fabry–Perot sensingcavities corresponding to two path-matching cavities for read-out systems are employed to form path-matching differential interferome-

Ž .tries PMDI . One Fabry–Perot cavity is used as sensing head, and the other as reference sensor. As a result, the experimental data showthat the change of refractive index of a so designed sensing system can be kept in at the level of 10y4 without any serious variations evenfor a 3-h long-term monitoring. Accordingly, the proposed new system can be easily implemented and used as a long-term monitoringsystem in a medical care environment. q 2000 Elsevier Science S.A. All rights reserved.

Keywords: Refractometer; White-light interferometry; Optical fiber sensors

1. Introduction

The refractive indices of liquids are conventionallymeasured by using an Abbe refractometer andror some

w xsimilar devices 1 . However, in the past decades, severaldifferent interferometries have been proposed and imple-

w xmented for use as refractometers 2,3 . Moosmuller andw xArnott 4 introduced a folded Jamin interferometer that

consists of two optical elements. This system was foundnot quite sensitive to the variation in movement. In addi-tion to traditional bulk optics, integrated optical sensors

w xwere also proposed 5 using the principle of evanescentwave sensing in an integrated Mach-Zehnder interferome-ter.

Research and development in optical fibers has gener-ated lots of research interest in recent years. This isattributed to the fact that sensors possess several advan-tages over conventional electrical transmitters. Thesedistinct advantages include inherent immunity to electro-magnetic interference, safety in hazardous or explosive

) Corresponding author. Tel.: q886-6-275-7575; fax: q886-6-235-2973; e-mail: loyl@mail.ncku.edu.tw

environments, high sensitivity, and long-distance remotemeasurements. The miniature size, intrinsic safety, andease of installation of fiber-optic based sensors make thesystem ideal for applications in various engineering areasincluding numerous in-line chemical, food, beverage, ormedical analysis and monitoring operations. Moreover,some researchers further employed the aforementionedoptical fiber sensing techniques to develop refractometers.

w xTakeo and Hattori 6,7 proposed refractometers by mea-suring the attenuation of light guided by optical fiberswithin the liquid. They also applied the refractometer thusbuilt to monitor the state of a skin hydration. Recently,

w xAsseh et al. 8 presented a fiber Bragg grating refractome-ter using an evanescent field refractive index fiber sensorthat comprises a 42-mm Bragg grating in an etched fiber.However, none of them have come up with a refractometerthat can be used for a long-term monitoring. To theauthor’s knowledge, this paper presents a newly developedfiber-optic based refractometer based on PMDI, also espe-cially on developing a highly stable measurement system.The theory and procedure of an optical fiber sensingsystem developed to implement the phase-modulated sta-ble optical fiber sensing systems are detailed. The environ-

0925-4005r00r$ - see front matter q 2000 Elsevier Science S.A. All rights reserved.Ž .PII: S0925-4005 99 00368-8

( )Y.-L. Lo et al.rSensors and Actuators B 62 2000 49–5450

mental factors that affect the sensitivity of the proposedoptical fiber sensing system and the associated compensa-tion techniques are also addressed.

w xRao and Jackson 9 developed a single-mode fiber-based ultrahigh pressure remote sensor based upon a Fizeaucavity using a dual-wavelength coherence reading tech-

w xnique and a built-in temperature compensator. Lo 10 andw xLo et al. 11 presented several cancellation methods for

compensating the variation of intensity in quadrature sig-w xnals for Fabry–Perot sensors in PMDI 12 , and for Bragg

grating sensors, respectively.This paper proposes a system that can be used to

construct a high stability and high sensitivity optical fiberrefractometer based on a white-light interferometry. Thechange of refractive index in a mixture of de-ionized waterand ethanol is used to illustrate the effectiveness of theproposed system by comparing the results with the mea-surement from conventional Abbe refractometer. It showsa good correlation between them. Based on a proposedcancellation technique, disturbances induced by thermaldrifts or vibrations in the read-out system are compensated.As a result, the optical fiber refractometer can be used tomeasure the change of refractive index in a resolution of10y5 and the dynamic range of 6=10y3 at the DCfrequency of up to 100 Hz. The estimated standard devia-tion is 2.33=10y5 for the experiment on de-ionized waterwith ethanol solutions.

2. Optical fiber refractometers

The optical setup arrangement and the implementationof a demodulation scheme for measuring the change ofrefractive index for use in long-term monitoring of anongoing system are detailed in this section. Fig. 1 showsthe structure of a two-beam interferometry used in thesensors and in the read-out systems for PMDI. The coher-ent length of the light source, L , is chosen to be less thanC

the cavity lengths. If the length of sensing and read-out

Fig. 1. Configuration of PMDI.

Fig. 2. Schematic diagram of setup.

cavities are denoted as L and L , respectively, L yLS R S R

is adjusted to be less than the coherent length. By doingw xso, the intensity, I, can be estimated as 12 :

22I(AqB cos 2k L yL exp ys L yLŽ . Ž .Ž .0 S R S R

1Ž .

where A and B are constants, s is the spectral width ofthe light source that is one half the distance between twopoints of which the spectral density equals the ratio of thepeak value to e, and k is the wavenumber. It is obvious0

from the above equation that the intensity, I, reaches themaximum degree of visibility while L (L . This is theS R

condition for a path-matching in PMDI.Fig. 2 illustrates a schematic diagram of the proposed

sensing system based on the technique of PMDI withcompensations. The main source of phase noise is thermaldrifts or induced vibrations on the PZT stack. The PZTstack is used for demodulating the phase signal in the

w xsingle-channel phase tracker 13 . To compensate the effectof variations in read-out system, two-parallel Fabry–Perotsare implemented in PMDI. One is used as the sensor forsensing the change of refractive index, and the other isused as the reference sensor formed by an in-line fiber

Ž .etalon ILFE . As illustrated in Fig. 2, two read-out cavi-ties are aligned together with the PZT stack to form theread-out systems. This arrangement is specifically de-signed to compensate the effect of thermal drift and vibra-tion on the PZT stack.

( )Y.-L. Lo et al.rSensors and Actuators B 62 2000 49–54 51

Fig. 3 shows a reference ILFE sensor that is composedof a hollow core fiber fused in between two standard

w xsingle mode optical fibers 14 . The air gap between theinterfacial surfaces of two fibers approximately between30 and 400 mm acts as a low finesse Fabry–Perot cavity.Since an ILFE sensor has low thermal apparent strain,thermal insensitivity makes it very suitable for compensat-ing thermal drifts andror vibrations on a PZT stack. Thus,the phase change of an ILFE Fabry–Perot reference sensorin PMDI shown in Fig. 2 can be expressed as:

4pDf ( n L yn L qD LŽ .ILFE air ILFE air PZT_I PZT_I

l0

4p( yD L 2Ž .Ž .PZT_I

l0

where l is the wavelength of the launching light, the0

refractive index of air, n , is equal to 1, L and Lair ILFE PZT_I

are the cavity lengths of an ILFE sensor and the corre-sponding read-out cavity, and D L is the variation inPZT_I

length caused by PZT thermal drifts or vibrations. InPMDI, the cavity lengths of an ILFE sensor and thecorresponding read-out cavity in L and L are ofILFE PZT_I

path-matching. It is worth noting that an ILFE sensor isthermally stable and isolated. Therefore, the variation fromthe PZT stack of read-out cavities is the only effect thatneeds to be considered.

The sensing head, another Fabry–Perot cavity, is formedby two optical fibers and placed on the V-groove substrate,as shown in Fig. 4. To eliminate the thermal drifting effectin the sensing head, the V-groove substrate is made ofceramic or glass. The coefficient of thermal expansion isthen limited to 5=10y6r8C. While the refractive index ofthe measurand is changed in the sensing head, the corre-sponding phase change in Fig. 2 can be estimated by:

4pDf ( n qDn L yn L qD LŽ . Ž .S S S S air PZT_S PZT_S

l0

4p( Dn L yD L 3Ž .Ž .S S PZT_S

l0

where n is the original refractive index in a sensing head,S

Dn is the change of refraction index, L is the cavityS S

length of the sensing head, and D L is the variation inPZT_S

length introduced by the effects of thermal drifts andror

Fig. 3. ILFE sensor.

Fig. 4. Sensing head.

vibrations in PZT. Here, n L and L are in optical-S S PZT_SŽ .path balance in terms of PMDI. As compared with Eq. 2 ,

Ž .it can be seen that D L and D L in Eqs. 2 andPZT_I PZT_SŽ .3 are variations caused by the thermal drifts androrvibrations from the PZT stack. This implies that D LPZT_I

Ž .is equal to D L . As the phase change in Eq. 2 , it isPZT_SŽ .subtracted from that of Eq. 3 , the effects of thermal

andror vibration on the PZT stack can be cancelled fromeach other. Consequently, the result of cancellation can beobtained by estimating:

4pDf yDf ( Dn L . 4Ž . Ž .S ILFE S S

l0

Thus, the phase change induced by the thermal driftsandror vibrations in PZT is removed in essence. Thestability of the sensing system in PMDI can be improvedby using two parallel Fabry–Perot sensors for a subtractionoperation. It is noteworthy that the two cavity lengths intwo parallel Fabry–Perot cavities, L and L , are re-S ILFE

quired to be kept constant to prevent signal errors fromcontaminating in the refractive index measurements. If theresolution of 10y5 is expected in an optical fiber refrac-tometer, the coefficient of thermal expansion for the sub-strate of a sensing head will have to be around 5=

10y6r8C. This is estimated at the level of temperaturevariation within 18C.

3. PMDI technique with two-parallel Fabry–Perot cavi-ties

Two-parallel Fabry–Perot cavities in PMDI are de-signed to stabilize a sensor system by subtracting thereference signal from a sensing one. However, it becomesa critical task to come up with a schema so that the cavitylengths of parallel Fabry–Perots can be designed and

w x w ximplemented without cross-talk. Lo et al. 15 and Lo 16have recently proposed a novel method to allow the lengthof Fabry–Perot cavities to be designed quantitatively inseries or parallel without cross-talk by using a spectrumanalysis technique. The condition required for none cross-

w xtalk between two-parallel Fabry–Perots 16,17 is given as:

< <L yL 4L 5Ž .S2 S1 c

( )Y.-L. Lo et al.rSensors and Actuators B 62 2000 49–5452

where L and L are the cavity lengths of two-parallelS1 S2

Fabry–Perots, respectively. It is interesting to note that inorder to eliminate the effect of cross-talk, the difference inthe lengths of two-parallel Fabry–Perot sensors cannot beless than the coherence length of a broadband source.

4. Experimental setup and results

ŽAs illustrated in Fig. 2, the broadband light FWHM ;. Ž50 nm of the pigtailed super luminescent diode Model:

.MRV MREDSP010-1 has a nominal wavelength of 1.3mm and the optical power is 150 mW. The receiverŽ .Model: New Focus 2011 has a noise-equivalent power

'Ž .NEP -1 pWr Hz . The light travels through a fiber-optic coupler to two parallel Fabry–Perot cavities. OneFabry–Perot cavity is used as the sensing head for sensingthe change of refractive index, and the other, ILFE, is usedas a reference sensor. The reflected lights of the twoparallel Fabry–Perot cavities again travel through thefiber-optic coupler to form low finesse Fabry–Perot cavi-ties with a mirror. The mirror is bonded to a PZT stackthat is placed on a linear transition plate. The PZT stack isused to provide a phase carrier for demodulation. Thetranslation plate is used to adjust the optical path of theread-out cavities to match with that of the sensing headand the ILFE. The cavity lengths of the read-out systemcorresponding to the sensing head and the ILFE are 133and 100 mm, respectively. The length difference is 33 mm,which is longer than the coherence length of SLD broad-

w xband light source of 20 mm 16 . This configuration isused to test the measurement capacity of the sensing headfor the change of refractive index by driving the PZT stack

w xwith a ramp function in a single-channel phase tracker 13for demodulation. This scheme is based on an electronic

w xfeedback phase nulling technique 13 using a AD639trigonometric simulator chip. By driving a PZT stack witha 6-kHz sinusoidal carrier frequency and selecting 800 Hz

Žas the cutoff frequency of the low-pass filter usually

Fig. 5. Maximum detectable signal.

Fig. 6. Minimum detectable signal.

.around 0.2 time carrying frequency , the sensitivity of asensing system can be improved in a low frequency re-sponse. Subsequently, several tests are conducted to char-acterize the performance of demodulation in the single-channel phase tracker. By driving a PZT from DC to 500Hz in various depths of modulation, the behavior of thedemodulation system can be characterized. Fig. 5 showsthe detectable maximum phase shift as a function of thesystem frequency. The minimum detectable phase shift asa function of the system frequency is given in Fig. 6.These figures show that the minimum detectable change ofrefractive index is approximately around 10y5 in a dy-namic range up to 6=10y3 at the frequency responsefrom DC to 100 Hz.

In order to verify the stability of the compensatedsensing system, an external loading is applied on the PZTstack to simulate thermal drifts andror vibrations. Both ofthe original and the compensated phase signals are recordedby detectors. The results are plotted and given in Fig. 7. As

Ž .it can be seen that the original phase signal I is fluttered1

by the simulating thermal drifts andror vibrations on theŽ .PZT stack. The lower trace I y I , as illustrated in Fig.1 2

7, is the compensated phase signal. It can be seen that theeffect of variation due to thermal drifts andror vibrationsis eliminated on the PZT stack. Therefore, the phase

Fig. 7. Simulating variation on a PZT.

( )Y.-L. Lo et al.rSensors and Actuators B 62 2000 49–54 53

Fig. 8. Long time monitoring.

variations caused by thermal drifts andror vibrations onthe PZT stack can be compensated by the proposed cancel-lation technique effectively.

A long-time test is also conducted by inserting a sens-ing head into the de-ionized water in an isothermal envi-ronment. Fig. 8 shows that the variation of the compen-sated signal in the sensing system is limited to be withinthe range of 10y4 in terms of the change of refractiveindex as compared to that of original signal without com-pensation for a test period of 3 h. The fluctuation can beexplained as the electric noise effects in circuit. It can beseen that the variation of the original signal without com-pensations is uncontrolled. In addition to the phase varia-tion associated with thermal drifts andror vibrations onthe PZT stack, a phase noise from the broadband sourcecan be explained as a main source that causes the variationof the sensing system. Nevertheless, as compared to thelong coherence light source, the phase noise from thebroadband source is much smaller since the optical path-

Ž .imbalance in Eq. 1 is small while the path-matchingcondition is satisfied. Therefore, for long-term monitoring,the use of a white-light interferometry does not requirehigh stability of a light source that is generally a necessityin an interferometry based on a long coherence source.

Fig. 9. Comparisons between Abbe and optical fiber refractometers.

A mixture of de-ionized water and ethanol solution isselected for testing the implemented optical fiber refrac-tometer. The change of refractive indices for various mix-tures of de-ionized water and ethanol is measured. Theresults are given in Fig. 9. A linear regression, as repre-sented by the solid line in Fig. 9, returns an excellentcorrelation between the change of refractive index and themixing composition. The standard deviation of the measur-and, s , is equal to 2.33=10y5. As compared with experi-mental data conducted by using the traditional Abbe re-fractometer, slight differences exist. These differences arecaused by different light sources used in both systems. Thelaunching light used in an Abbe refractometer is around638 nm in wavelength which is somewhat different fromthe wavelength of the light source used in our proposedsystem.

5. Conclusions and discussions

An optical fiber-based refractometer that measures thechange of refractive index in a resolution up to 10y5 forthe dynamic range of 6=10y3 in DC frequency of up to100 Hz is developed in this paper. A cancellation methodis proposed and implemented for use in compensating theeffect of thermal drifts andror vibrations on a PZT stackso that the long-term monitoring of the change of refrac-tive index can be accomplished. The use of two parallelFabry–Perots in PMDI to allow the cancellation of phasechange is proved to be efficient in compensating phasevariations on the PZT stack driven by a function generator.

It should be noticed that the cancellation technique isapplied to compensate the thermal drifts andror vibrationson a PZT stack in the read-out system only. Therefore, asensing head packaged by a low coefficient of thermal

Fig. 10. Alternation in the sensing head.

( )Y.-L. Lo et al.rSensors and Actuators B 62 2000 49–5454

expansion substrate is usually needed. To compensate thethermal expansion on the substrate of the sensing head,Fig. 10 shows an alternation for an optical fiber refrac-tometer. Also, by inserting de-ionized water in the cavityof the ILFE, the thermal-optic effect in de-ionized watercan be minimized.

Because the distinct features such as small cavity insensing head and high robustness are very attractive, thedevice is very suited for various applications where smallsize, high mechanical stability, and small analyte volumesare substantially required.

Acknowledgements

The authors would like to acknowledge Dr. J.S. Sirkisfor his support on hollow-core fibers in the Department ofMechanical Engineering at the University of Maryland,College Park, USA.

References

w x1 R.S. Longhurst, Geometrical and Physical Optics, 3rd edn., Long-mans, London, 1962.

w x2 D.J. Webb, R.P. Tatam, D.A. Jackson, A novel interferometricŽ .liquid refractometer, 60 1989 3347–3348.

w x3 K.I. Fujii, E.R. Williams, R.L. Steiner, D.B. Newell, A new refrac-tometer by combining a variable length vacuum cell and a double-pass Michelson interferometer, IEEE Trans. Instrum. Meas. 46Ž .1997 191–195.

w x4 H. Moosmuller, W.P. Arnott, Folded Jamin interferometer: a stableŽ .instrument for refractive-index measurements, Opt. Lett. 21 1996

438–440.w x5 B. Maisenholder, H.P. Zappe, R.E. Kunz, M. Moser, P. Riel, Optical

Refractometry Using a Monolithically Integrated Mach-Zehnder In-terferometer, Transducers ’97, International Conference on Solid-State Sensors and Actuators, Chicago, 1997, pp. 79–80.

w x6 T. Takeo, H. Hattori, Optical fiber sensor for measuring refractiveŽ .index, Jpn. J. Appl. Phys. 21 1982 1509–1512.

w x7 T. Takeo, H. Hattori, Skin hydration state estimation using a fiber-Ž .optic refractometer, Appl. Opt. 33 1994 4267–4272.

w x8 A. Asseh, S. Sandgren, H. Ahlfeldt, B. Sahlgren, R. Stubbe, G.Edwall, Fiber optical bragg grating refractometer, Fiber Integr. Opt.

Ž .17 1998 51–62.w x9 Y.J. Rao, D.A. Jackson, Prototype fiber-optic-based ultrahigh pres-

sure remote sensor with built-in temperature compensation, Rev. Sci.Ž .Instrum. 65 1994 1695–1698.

w x10 Y.L. Lo, Intensity compensation for in-fiber Fabry–Perot sensors inŽ . Ž .high disturbed environments, IEE Electron. Lett. 34 4 1998

394–395.

w x11 Y.L. Lo, J.F. Huang, M.D. Yang, Intensity effects in bragg gratingsensors scanned by a tunable filter, Smart Structures and MaterialsSPIE Conference, Sensory Phenomena and Measurement Instrumen-

Ž .tation for Smart Structures and Material 3670 1999 65–73.w x12 P.G. Cielo, Fiber optic hydrophone: improved strain configuration

Ž .and environmental noise protection, Appl. Opt. 18 1979 2933–2937.

w x13 A.D. Kersey, R.P. Moeller, T.A. Berkoff, W.K. Burns, Singlechannel phase-tracker for the open loop fiber optic gyroscope, SPIE

Ž .1585 1991 198–202.w x14 J.S. Sirkis, T.A. Berkoff, R.T. Jones, H. Singh, A.D. Kersey, E.J.

Ž .Friebele, M.A. Putnam, In-line fiber etalon ILFE fiber-optic strainŽ .sensors, J. Lightwave Technol. 13 1995 1256–1263.

w x15 Y.L. Lo, M.H. Tsai, C.C. Tsao, Spectrum analysis in cross-talk ofseries Fabry–Perot sensors in path-matching differential interferome-

Ž . Ž . Ž .try PMDI , Opt. Laser Technol. 30 6–7 1998 395–401.w x16 Y.L. Lo, Study of cross-talk on parallel Fabry–Perot sensors in

Ž .path-matching differential interferometry PMDI , accepted in Opt.Lasers Eng., 1999.

w x17 J.L. Brooks, R.H. Wentworth, R.C. Youngquist, M. Tur, B.Y. Kim,H.J. Shaw, Coherence multiplexing of fiber-optic interferometric

Ž .sensors, J. Lightwave Technol. LT-3 1985 1062–1072.

Yu-Lung Lo received the B.S. degree from the National Cheng KungUniversity in 1985, and the M.S. and Ph.D. degree from Smart Materialsand Structures Research Center at the University of Maryland, CollegePark, in 1992 and 1995, respectively, all in mechanical engineering. After

Ž .his graduation, he joined Industrial Technology Research Institute ITRIat Opto-Electronics and Systems Laboratories working on fiber opticsmart structures. He has been a member of the faculty of the MechanicalEngineering Department at the National Cheng Kung University since1996. His research interests are in the areas of experimental mechanics,fiber-optic sensors, smart structures, optical techniques in precision mea-surements, electronic packaging, and MEMS. He has authored over 40technical publications and filed three patent disclosures. Dr. Lo is amember of SPIE.

Hsin-Yi Lai received his Ph.D. from the University of Wisconsin atMadison in 1984. He joined North Carolina A&T State University as afull-time faculty member right after he completed his terminal degree in1984. Dr. Lai is currently working as a full professor in the Departmentof Mechanical Engineering at National Cheng-Kung University in Tainan,Taiwan. He had experiences in the application of modeling techniques forvarious engineering problems in system analysis, design, manufacturingand control. In the past 5 years, Dr. Lai has participated in 12 fundedresearch projects and published over 40 technical papers in the areas ofhis expertise. Dr. Lai had served as the Section Chairman, ProgramChairman, Program Secretary, and Vice Chairman in Dixie Section ofASME Region IV. He has also worked as the Director of NASA Centerof Research Excellence in North Carolina A&T State University during1990–1993. Dr. Lai’s current research interest includes High SpeedDynamic Systems-Smart Sensing, Measurement and Control, and Designfor Industrial Automation.

Wern-Cheng Wang is a graduate student in the Department of Mechani-cal Engineering at the National Cheng Kung University, Tainan, Taiwan,ROC.