Spatially Resolved Analyte Mapping with Extended-Length

sults with a model system (a 48-m-long oxygen sensor) .... (24) Skolnik, M. I. Radar Handbook; McGraw-Hill: New York, 1970. .... (1). Fmax. ) c/(nf. +...
7 downloads 0 Views 199KB Size
Download PDF
AC Research

Anal. Chem. 1998, 70, 1453-1461

Accelerated Articles

Optical Time-of-Flight Chemical Detection: Spatially Resolved Analyte Mapping with Extended-Length Continuous Chemically Modified Optical Fibers Radislav A. Potyrailo† and Gary M. Hieftje*

Department of Chemistry, Indiana University, Bloomington, Indiana 47405

We theoretically evaluate and experimentally verify a novel strategy for spatially resolved analyte mapping over extended remote areas. The approach combines a method for the fabrication of continuous extended-length sensors with optical time-of-flight chemical detection (OTOF-CD). The use of OTOF-CD makes it possible to locate the zones in the fiber where attenuation or fluorescence takes place, to determine the magnitude of these variations, and to relate the magnitude of the variations to the local concentration or concentrations of a single analyte or several analytes. Simulation experiments suggest that OTOF-CD should provide spatial resolution close to its theoretical limit by deconvolution of the returned wave form with all time-dependent experimental variables (laser pulse width, reagent fluorescence lifetime, etc.). The signal-processing technique should be useful for a wide variety of sensors based on absorption, refractive index, or statically and dynamically quenched fluorescence. Experimental results with a model system (a 48-m-long oxygen sensor) compare favorably with those predicted by numerical simulations. Possible experimental difficulties in the realization of these novel sensors are discussed as are ways to overcome them.

Spatially resolved mapping of chemical constituents is an important need in process control, industrial safety, environmental monitoring, and other fields. To be suitable for this purpose, a technology should permit the determination of analyte concentra† Present address: Characterization and Environmental Technology Laboratory, Corporate Research and Development, General Electric Co., Building K-1, P.O. Box 8, Schenectady, NY 12301.

S0003-2700(97)01315-2 CCC: $15.00 Published on Web 03/18/1998

© 1998 American Chemical Society

tion as a function of position along an extended-length sensing element. Fiber optics provide a unique solution to this problem. The use of “distributed” sensors (featuring extended-length continuous chemically sensitive fibers) can offer detection schemes for which there is no counterpart in conventional sensor technologies. A sensor is termed “distributed” here if it operates with a continuous extended-length sensing element and is capable of determining sought-for variations in a parameter along the entire length of the fiber as a continuous function of distance1,2 or in a spatially averaged mode.3,4 Several attempts have recently been reported to develop extended-length continuous sensing elements. For example, in one distributed moisture sensor,5 an optical fiber was held in intimate contact with a hydrogel-coated rod by means of a helically wound thread. Analyte-specific swelling of the hydrogel increased the microbending-induced loss of light within the fiber through an increase in strain caused by the confining thread. Unfortunately, this design suffered from complexity and a limited range of potentially detectable species. An attempt was made also to chemically immobilize reagents on the surface of a commercially available plastic optical fiber.6 Although the technique was useful for modification of short fiber sections with a pH-sensitive dye, it was not attractive for distrib(1) Dakin, J. P. J. Phys. E: Sci. Instrum. 1987, 20, 954-967. (2) Rogers, A. J. Phys. Rep. 1988, 169, 99-143. (3) Lieberman, R. A.; Blyler, L. L.; Cohen, L. G. J. Lightwave Technol. 1990, 8, 212-220. (4) Lieberman, R. A.; Mendoza, E. A.; Ferrell, D. J.; Schmidlin, E. M.; Syracuse, S. J.; Khalil, A. N.; Dergevorkian, A.; Sun, Z.; Gunther, M. Proc. SPIE-Int. Soc. Opt. Eng.1994, 2068, 192-201. (5) Michie, W. C.; Culshaw, B.; Konstantaki, M.; McKenzie, I.; Kelly, S.; Graham, N. B.; Moran, C. J. Lightwave Technol. 1995, 13, 1415-1420. (6) Puschett, J. B.; Matyjaszewski, K. Surface modification of plastic optical fibers. U.S. Patent 5,077,078, 1991.

Analytical Chemistry, Vol. 70, No. 8, April 15, 1998 1453

uted pH sensing over extended lengths because of scattering losses in the modified fibers. In another technique, optical fibers were drawn with a polymer cladding doped with a reagent sensitive to sought-for chemical constituents or properties, including ammonia, oxygen, water vapor, and pH.7,8 Unfortunately, this means of fiber fabrication is restricted to those workers who have the resources required (essentially, a fiber-optic drawing laboratory).9 Overall, a chief problem in designing a continuous chemically sensitive fiber is the absence of a simple and economical method for fabricating the extended intrinsic sensing layer. Alternatively, spatial mapping is possible by multiplexing the outputs of several point sensors.10-13 Unfortunately, these multipoint-detection approaches are limited to spatially discrete measurements. Clearly, spatial mapping with a single continuous optical fiber is a desirable alternative to multiplexed point sensors. The present study utilizes a new strategy for constructing distributed sensors for spatially resolved analyte monitoring. The key difference from other reported technologies is that, in the new method, continuous extended-length chemically sensitive fibers are fabricated by chemical modification of the cross-linked silicone cladding of a commercially available plastic-clad silica (PCS) fiber.14 A change in evanescent-wave absorbance or fluorescence of an indicator or indicators immobilized in the fiber cladding is then related to analyte concentration. In contrast to previously reported fabrication technologies for extended-length sensing elements,5,7 the new strategy significantly simplifies the fabrication process, lowers the construction cost of the sensing element, and permits immobilization of several reagents in the same or in adjacent sections of a single optical fiber. The concept has already been proven by studies on point sensors based on short lengths of modified PCS fibers. These sensors exhibited improved sensitivity,15 expanded dynamic range,16 and better longterm stability14 than sensors reported in the past. These advantages were attributable to the unique properties of the silicone cladding material used as an immobilization matrix.15 Earlier, a fiber-optic time-of-flight spectrometer was reported in which a sample cell was illuminated with short pulses of polychromatic light; the transmitted light was then directed onto a fast detector through a 0.5-1.1-km-long multimode optical fiber.17,18 The arrival time of the photons at the detector was a function of wavelength because of the spectral dispersion of the fiber material. The device therefore served as a replacement for a dispersive spectrometer and was given the appellation “time(7) Blyler, L. L., Jr.; Lieberman, R. A.; Cohen, L. G.; Ferrara, J. A.; MacChesney, J. B. Polym. Eng. Sci. 1989, 29, 1215-1218. (8) Mendoza, E. A.; Sorenson, J.; Iossi, A.; Sun, Z.; Robinson, D.; Lieberman, R. A. Proc. SPIE-Int. Soc. Opt. Eng. 1996, 2836, 242-249. (9) Lieberman, R. A. Sens. Actuators B 1993, 11, 43-55. (10) Murphy, V.; MacCraith, B. D.; Butler, T.; Mcdonagh, C.; Lawless, B. Electron. Lett. 1997, 33, 618-619. (11) Kharaz, A.; Jones, B. E. Sens. Actuators A 1995, 47, 491-493. (12) Browne, C. A.; Tarrant, D. H.; Olteanu, M. S.; Mullens, J. W.; Chronister, E. L. Anal. Chem. 1996, 68, 2289-2295. (13) Bownass, D. C.; Barton, J. S.; Jones, J. D. C. Opt. Lett. 1997, 22, 346-348. (14) Potyrailo, R. A.; Hieftje, G. M., Pittsburgh Conference 1995; Paper 1022. (15) Potyrailo, R. A.; Ray, S. J.; Burden, D. L.; Hieftje, G. M., Pittsburgh Conference 1996; Paper 1080. (16) Potyrailo, R. A.; Hieftje, G. M., FACSS XXII Annual Meeting, 1995; Paper 335. (17) Whitten, W. B.; Ross, H. H. Anal. Chem. 1979, 51, 417-419. (18) Whitten, W. B.; Ross, H. H. Anal. Chem. 1980, 52, 101-104.

1454 Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

of-flight optical spectrometer” by analogy with the mass spectrometric device. Another optical time-of-flight technique was used for diagnosis of fiber-optic communication networks19 and later for spatially resolved distributed sensing.2,8,20 As is detailed below, in this technique, the arrival time of photons at the detector is a function of distance from the launch end of the fiber to the position where analyte detection occurs. When a backscattered signal is measured, the technique is known as optical time-domain reflectometry.21 Because backscatter and back-propagated fluorescence are different, it is useful to generalize this type of sensor as an optical time-of-flight device. The present study first evaluates theoretically the capabilities of absorption- and fluorescence-based distributed chemical sensors that utilize optical time-of-flight chemical detection (OTOF-CD). Second, experimental verification is accomplished with a model system, an oxygen sensor with a 48-m-long continuous chemically sensitive fiber modified with a tetraphenylporphyrin (TPP) fluorophore. The only TPP-based oxygen sensor reported to date utilized the fluorophore adsorbed onto porous polystyrene beads.22 Unfortunately, in that point sensor, the relative humidity of the gas samples affected the fluorescence signal strongly unless the sensing element was held at an elevated temperature. We recently demonstrated15 that incorporation of TPP into the silicone fiber cladding results in immunity of its fluorescence to changes in relative humidity. This paper reports the first experimental demonstration of an OTOF distributed fluorescent sensor based on a dynamically quenched fluorophore.23 PRINCIPLE OF SPATIALLY RESOLVED DISTRIBUTED ANALYTE MEASUREMENTS Choice of Detection Method. To achieve spatial resolution, the same principles of signal generation and processing can be used as in radar measurement24 and fault finding in fiber-optic communication networks.19 In fact, these methods have already been employed for spatially resolved sensing of physical parameters such as magnetic, electric, and acoustic fields, temperature, strain, stress, pressure, displacement, ionizing radiation, and others.2,20 Alternative methods for distributed, spatially resolved measurement of physical parameters include optical time-domain reflectometry (based on Rayleigh,25 Raman,26 and Brillouin27 scattering, fluorescence,28 polarization,29 and attenuation30), transmissive frequency-modulated carrier wave,31 optical frequency(19) Barnoski, M. K.; Jensen, S. M. Appl. Opt. 1976, 15, 2112-2115. (20) Kersey, A. D. In Fiber Optic Sensors: An Introduction for Engineers and Scientists; Udd, E., Ed.; Wiley: New York, 1991; pp 325-368. (21) Barnoski, M. K.; Rourke, M. D.; Jensen, S. M.; Melville, R. T. Appl. Opt. 1977, 16, 2375-2379. (22) Barnikol, W. K. R.; Gaertner, T.; Weiler, N. Rev. Sci. Instrum. 1988, 59, 1204-1208. (23) Potyrailo, R. A.; Hieftje, G. M., Pittsburgh Conference 1997; Paper 921. (24) Skolnik, M. I. Radar Handbook; McGraw-Hill: New York, 1970. (25) Boiarski, A. A.; McGinniss, V. D. Distributed fiber optic sensor using clad material light backscattering. U.S. Patent 5,191,206, 1993. (26) Ho¨bel, M.; Ricka, J.; Wuthrich, M.; Binkert, T. Appl. Opt. 1995, 34, 29552967. (27) Bao, X.; Webb, D. J.; Jackson, D. A. Can. J. Phys. 1996, 74, 1-3. (28) Dakin, J. P. Temperature measuring arrangements. U.K. Pat. Appl. GB 2156513A, 1985. (29) Rogers, A. Electron. Lett. 1980, 16, 489-490. (30) Yataghene, A.; Himbert, M.; Tardy, A. Rev. Sci. Instrum. 1995, 66, 38943900. (31) Zheng, G.; Campbell, M.; Wallace, P. Appl. Opt. 1996, 35, 5722-5726.

between the location l of interest along the fiber and the time t required for the light pulse to propagate forward and backward to this location is given by

t ) l(nf + nb)/c

Figure 1. Principle of spatially resolved measurements with an OTOF distributed chemical sensor. λ1 and λ2 are the probe and fluorescence emission wavelengths, respectively, ∆lmin is the spatial resolution of the sensor as described by eq 7, BS is a beam splitter, MMD is a multimode distribution of propagating light in the fiber core and the corresponding evanescent wave in the cladding, the open circles are scattering centers in the fiber core, and the filled circles represent excited fluorophores in the cladding.

domain reflectometry,32 and optical amplification by a counterpropagating pump pulse.33 A critical analysis of these and less common techniques shows that optical time-domain reflectometry is particularly attractive for distributed chemical monitoring. Advantages of an analogous OTOF technique include the possibility of quantitative measurements, operation with multimode fibers, simple signal formation and processing methods, and inexpensive optoelectronic components. Principle of Operation of OTOF-CD. When a pulse of light at probe wavelength λ1 is launched into a sensing fiber, analytical information can be derived from the portion of light that is returned to the launch end of the fiber (see Figure 1). The analytical signal might be the Rayleigh backscatter at wavelength λ1 or fluorescence emission at a different wavelength λ2. In addition, temperature-dependent Raman scattering can be monitored,2 omitted from this discussion for brevity. The Rayleigh and fluorescence signals can be modulated by variations in analyte concentration at the core-cladding interface. The Rayleigh scattering usually originates from microscopic inhomogeneities in the fiber material. These inhomogeneities cause spatial fluctuations in the refractive index of the fiber core and generate reflections of light guided in the fiber. A portion of the scattered light is recaptured by the fiber numerical aperture in the reverse direction. In a chemical sensor made from a multimode optical fiber with an analyte-permeable cladding, analyte detection based on Rayleigh backscatter can utilize variations in either the refractive index or the absorption coefficient of the fiber cladding. These variations modulate the intensity of the backscatter signal because of losses in the evanescent field. If the fiber cladding is doped with a fluorophore, the evanescent wave will excite the fluorophore at the corecladding interface. Half of the fluorescence emission recaptured by the fiber numerical aperture will be guided toward the launch end of the fiber. To obtain information about the location of an analyte, that is, where a change in the back-propagated radiation occurs (as measured from the launch end of the fiber), the level of the returned signal is monitored as a function of the time delay between the launched and returned light pulse. The relation (32) Kingsley, S. A.; Davies, D. E. N. Electron. Lett. 1985, 21, 434-435. (33) Valis, T.; Turner, R.; Measures, R. M. Appl. Opt. 1989, 28, 1984-1990.

(1)

where c is the velocity of light in vacuum and nf and nb are the refractive indices of the fiber core at the wavelengths of the forward (excitation wavelength) and backward (analytical wavelength) propagated pulses, respectively.34 To increase the signal level and to improve the signal-to-noise ratio (S/N), repetitive light pulses should be used. However, only a single pulse should propagate in the fiber at a given time to avoid signal overlap. Thus, the maximum pulse repetition rate Fmax is limited by the length L of the fiber and is given by

Fmax ) c/(nf + nb)L

(2)

The returned signal in an OTOF measurement system was first formulated for Rayleigh backscatter19,35,36 and then for Raman backscatter.37 If the relationship is also generalized for fluorescence detection, the back-propagated impulse response from a distance l along the fiber excited by a δ-pulse is given by

P(l) ) rPoS(l) exp{-

∫ [R (z) + R (z)] dz} l

0

f

b

(3)

where Po is the excitation power in the input pulse coupled into the fiber, r is the ratio of the transmitted to the reflected optical power in the beam splitter (cf. Figure 1), S is a constant that depends on the local numerical aperture (NA) of the fiber, the fluorescence quantum yield of the immobilized fluorophore (in a fluorescence-based sensor), or the Rayleigh and Raman scattering parameters of the fiber core and cladding (in a scattering-based sensor), and Rf(z) and Rb(z) are the attenuation coefficients of the forward and backward traveling pulses, respectively. To a first approximation,20,38 Rf ) Rb ) R, but in practice, Rf and Rb may be different in multimode fibers because the mode distribution of the back-propagated light may not be equal to that of the incident light.36 In fluorescence-based sensors, both Rf and Rb have two components, one each for the excitation and emission wavelengths. The response of an OTOF chemical sensor based on attenuation detection is a convolution of the ideal impulse response described by eq 3 and the temporal profile Rex(t) of the excitation pulse width τex. By using a transformation of variables l f t given by eq 1, the response Patt(t) of an attenuation-based OTOF sensor can be expressed as

Patt(t) ) P(t) X Rex(t)

(4)

(34) (35) (36) (37)

Malitson, I. H. J. Opt. Soc. Am. 1965, 55, 1205-1209. Personick, S. D. Bell Syst. Technol. J. 1977, 56, 355-366. Healey, P. J. Phys. E: Sci. Instrum. 1986, 19, 334-341. Dakin, J. In Optical Fiber Sensors: Systems and Applications; Culshaw, B., Dakin, J., Eds.; Artech House: Norwood, MA, 1989; Vol. 2, pp 575-598. (38) Farries, M. C.; Fermann, M. E.; Laming, R. I.; Poole, S. B.; Payne, D. N. Electron. Lett. 1986, 22, 418-419.

Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

1455

where X denotes convolution. Similarly, the response Pfl(t) of a fluorescence-based OTOF sensor is given by

Pfl(t) ) P(t) X Rex(t) X Rem(t)

(5)

where Rem(t) is the impulse response of the immobilized fluorophore (i.e., an exponential decay having a time constant equal to the fluorescence lifetime τem of the fluorophore). In a sensor based on a statically quenched immobilized fluorophore, τem is constant. The output signal of such a sensor is a function of the attenuation coefficient of the immobilized fluorophore. It should be clear that this functional dependence is very similar to that of an attenuation (absorption-based) sensor. A sensor based on dynamic fluorescence quenching of an immobilized fluorophore does not exhibit a change in attenuation coefficient upon exposure to a quencher. Rather, the fluorescence lifetime τem and emission intensity I are a function of the quencher concentration [Q] as described by the Stern-Volmer equation:

Io/I ) τoem/τem ) 1 + Ksv[Q]

(6)

where Io and τoem are the emission intensity and fluorescence lifetime of the fluorophore in the absence of the quencher, respectively, and Ksv is the Stern-Volmer quenching constant. Spatial Resolution of Distributed OTOF Sensor. In a distributed sensor, spatial resolution is of primary importance. The spatial resolution should be as great as possible to provide the maximum number of resolved local variations of analyte concentration along the fiber length and to make the distributed sensor as cost-effective as possible compared to point sensors. The maximum spatial resolution is provided by the ideal impulse response of the continuous sensor expressed in eq 3. In principle, this ideal impulse response P can be obtained by measurement of Patt, Pfl, Rex, and Rem and subsequent deconvolution according to eqs 4 and 5. However, several practical limitations often make deconvolution difficult, noise usually being the most serious one.39 Thus, the spatial resolution ∆lmin of a detection system (cf. Figure 1), if deconvolution is not used, is the smallest resolved distance between two fiber locations l1 and l2 given by20

∆lmin ) cτb/(nf + nb)

(7)

where τb is the width of the back-propagated pulse. The number of sensing regions N resolved with such a distributed sensor of length L is given by

N ) (nf + nb)L/cτb

(8)

Assuming that no appreciable pulse broadening takes place in the sensing fiber, the spatial resolution of an attenuation-based sensor is limited by the width of the excitation pulse τex. In contrast, in fluorescence-based sensors, spatial resolution is limited by the convolution of Rex (the temporal profile of the excitation pulse) and Rem (the impulse response of the immobilized fluorophore). Thus, to improve spatial resolution, it is desirable to use (39) Demas, J. N. Excited-State Lifetime Measurements; Academic Press: New York, 1983.

1456 Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

a fluorophore with a short fluorescence lifetime. This approach can be used with a sensor that utilizes static quenching. Indeed, if a fluorophore with a subnanosecond lifetime is used,40,41 then for an excitation source with a pulse width τex of several nanoseconds, τex ≈ τb. Unfortunately, this strategy does not always work with a sensor that utilizes a dynamically quenched fluorophore. The sensitivity of such a sensor is usually lower when a fluorophore of very short lifetime is used. From eqs 3-6, several variables can be exploited to evoke a sensor response Patt or Pfl. Examples include a change in the numerical aperture of the fiber caused by swelling of a plastic cladding upon exposure to a solvent42 or a change in temperature.43 In addition, Rayleigh scattering can be modulated by a change in the refractive index of the fiber cladding as a function of temperature.25 However, for chemical sensing, the most promising approach is to modify the analyte-permeable polymer cladding with an absorbing or fluorescent reagent. This approach leads to a potentially broad range of detectable analytes. In this case, the sensor signal is modulated by changes in the attenuation coefficient of the dye-doped fiber cladding8 or by variations in the fluorescence lifetime of an immobilized reagent. This latter technique is the one utilized in the present study. EXPERIMENTAL SECTION Reagents and Materials. Tetraphenylporphyrin and toluene were obtained from Aldrich (Milwaukee, WI). Multimode PCS optical fiber of 200-µm core diameter and 50-µm-thick silicone cladding was purchased from Fiberguide Industries (Stirling, NJ). Dry nitrogen was supplied from boil-off of a local liquid nitrogen storage facility. Compressed dry oxygen was obtained from Air Products and Chemicals (Allentown, PA). Chemical Modification of Fiber Cladding. A 48-m-long fiber segment was coiled onto a 14-cm-diameter metal drum by means of a custom-built device. The optical fiber was immersed for 20 min in a 2 µM solution of TPP in toluene, rinsed in toluene to remove any surface residue, and dried. Experimental Setup. A schematic diagram of the instrument constructed for spatially resolved chemical monitoring is shown in Figure 2. Light from a Q-switched frequency-doubled (532nm) Nd:YAG laser (13-ns fwhm, 20-Hz repetition rate, Quantel International YG580 Series) was launched into the fiber through a beam splitter (BS) by means of a microscope objective (L1) with a numerical aperture of 0.4. The average laser power transmitted through the fiber was 50 µW. Back-propagated fluorescence was collected with the same objective (L1), redirected through the beam splitter, and focused with a lens (L2) onto the entrance slit of a monochromator (Instruments SA, model HSA-320). A longpass, 550-nm-cutoff optical filter (F) was used to block excitation light from entering the monochromator. Fluorescence was monitored at 720 nm with a photomultiplier tube (PMT; Hamamatsu R928). A small portion of the laser light was directed to a fast (40) Berlman, I. B. Handbook of Fluorescence Spectra of Aromatic Molecules; Academic Press: New York, 1971. (41) Flanagan, J. H.; Romero, S. E.; Legendre, B. L., Jr.; Hammer, R. P.; Soper, S. A. Proc. SPIE-Int. Soc. Opt. Eng. 1997, 2980, 328-337. (42) Bu ¨ rck, J.; Sensfelder, E.; Ache, H.-J. Proc. SPIE-Int. Soc. Opt. Eng. 1996, 2836, 250-260. (43) McGinniss, V. D.; Whitmore, R. S., Jr.; Kingsley, S. A. Thermal refractive materials for optical sensor application. U.S. Patent 5,052,820, October 01, 1991.

Attenuation- and Fluorescence-Based Distributed Sensors of Type I. The returned wave form, computed from eqs 3 and 4 for a distributed sensor of type I, is plotted in Figure 3A as a function of the elapsed time or equivalent fiber length. The deconvoluted wave form is also plotted in Figure 3A to illustrate the smoothing effect of convolution. The different rate of the exponential signal decay in sections I-V is due to the different cumulative two-way attenuation along the fiber; this behavior is described by the integral in the exponent of eq 3. The optical attenuation at any point l along the fiber can be calculated as the slope of the logarithm of the exponentially decaying wave form at that point as Figure 2. Schematic diagram of experimental setup for spatially resolved measurements with the OTOF distributed chemical sensor: M, mirror; BS, beam splitter; L1 and L2, lenses; F, filter; PD, fast photodiode; DO, digitizing oscilloscope; PMT, photomultiplier tube.

photodiode (PD) by a mirror (M). The returned wave forms were recorded at 2-ns sampling interval with a two-channel digitizing oscilloscope (DO; 500 MHz, Tektronix, model TDS 520); the oscilloscope was triggered by the output from the fast photodiode. Each monitored wave form was averaged over 2000 laser pulses, transferred to a Macintosh computer, and analyzed using commercial software packages. For spatially resolved measurements, the modified fiber was coiled into 13 equal-length segments, each of about 15-cm diameter. Fiber segments under test were placed in gas flow cells so they could be independently exposed to various gas mixtures. The rest of the fiber was kept in air.

RESULTS AND DISCUSSION Analysis of Computer-Simulated Wave Forms. For computer simulations, the distributed sensors modeled by eqs 3-6 were divided into two types. Type I included attenuation-based sensors and sensors based on statically quenched fluorophores, since the mathematical description of the response of both sensors is similar. Type II pertained to sensors based on dynamically quenched fluorescence. Conditions For Computer Simulations. Parameters employed for the numerical simulations are summarized in Table 1. The attenuation coefficients used in the simulations are typical for fibers that have chemical reagents immobilized into their cladding.7,8,14-16 The fluorescence lifetime of the statically quenched fluorophore was selected within the range of common UVvisible40 and near-infrared41 fluorophores and to be smaller than the temporal (or spatial) step size of the simulation experiments (cf. Table 1). Simulation parameters for a sensor based on dynamically quenched fluorescence corresponded to those obtained experimentally with an immobilized TPP fluorophore.15,23 The excitation pulse width was selected within the range of reported OTOF systems (from single to hundreds of nanoseconds).20,36 To demonstrate the general theoretical characteristics of an OTOF chemical sensor, differences in the propagation of forward and backward guided modes, differing attenuation of different modes propagating in the fiber, and mode-coupling effects have all been intentionally omitted. All computer simulations were performed with Mathematica software (Wolfram Research, Inc. Champaign, IL).

10 ∂{logPI(l)}/∂l ) - 2R(l)

(9)

if measured in decibels, where PI is the returned wave form. The regions of higher attenuation are characterized by a greater slope in the returned wave form, as predicted by eq 9. The relation between the slope and the analyte-dependent attenuation coefficient suggests a simple method of quantifying analyte concentrations. The slope can be calculated for each point l on the fiber according to eq 9, and its value can be plotted as a function of distance along the fiber. This method is illustrated in Figure 3B. The transformed wave form after deconvolution with Rex (the temporal profile of the excitation pulse) is also plotted for comparison. Figure 3B demonstrates the convenience of using this method to quantify the analyte concentration along the fiber length. If the fiber attenuation in the absence of the analyte is known, it is possible to calibrate fiber attenuation directly in terms of analyte concentration. Figure 3B also shows the advantages of deconvolution in obtaining the best possible spatial resolution. A disadvantage of attenuation- and fluorescence-based distributed sensors (of type I) is that an increase in analyte concentration at a specific location along the fiber causes a rise in fiber attenuation. This local increase in optical attenuation leads to a drop in light intensity available for propagation further down the fiber. The dynamic range DR of the distributed sensor will therefore be limited by the highest analyte concentration detectable along the fiber or by the maximum fiber length as follows:

DR - M >

∑R l

i i

(10)

i

where M is the margin at the far end of the fiber required to produce a desired S/N,44 Ri is the attenuation coefficient of the fiber over a length li, and ∑ili ) L. With a typical value of dynamic range of 30 dB45 and a 10-dB margin,44 the maximum fiber length will be limited to 100-200 m for a 0.1-0.2-dB/m fiber attenuation, typical in distributed sensors.8,46 Fluorescence-Based Distributed Sensors of Type II. Sensors based on immobilized dynamically quenched fluorophores are free from this limitation of type I sensors. However, obtaining the (44) Nakamura, K.; Uchino, N.; Matsuda, Y.; Yoshino, T. IEICE Trans. Commun. 1997, E80-B, 528-534. (45) Tateda, M.; Horiguchi, T. J. Lightwave Technol. 1989, 7, 1217-1224. (46) Potyrailo, R. A.; Hieftje, G. M. Optical time-of-flight chemical detection (OTOF-CD) : absorption-modulated fluorescence for spatially resolved analyte mapping in a bidirectional distributed fiber-optic sensor. Submitted to Anal. Chem.

Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

1457

Table 1. Parameters for Numerical Simulations for OTOF-Based Distributed Sensors Type I. Sensors Based on Detection of (i) Attenuation and (ii) Statically Quenched Fluorescence of an Immobilized Fluorophore attenuation coefficient of fiber, dB/m -0.02 (I and V),a -0.2 (II and IV), -0.1 (III) fluorescence lifetime, ns 0.5 Type II. Sensors Based on Detection of Dynamically Quenched Fluorescence of an Immobilized Fluorophore unquenched fluorescence lifetime, ns 10 quenching constant, KSV, atm-1 0.6 partial pressure of the quencher, atm 1.0 (I),a 0.21 (II), 0.5 (III) attenuation coefficient of fiber, dB/m -0.1 excitation pulse profile excitation pulse fwhm, ns convolution operation deconvolution operation temporal step size in the numerical simulations, ns spatial step size in the numerical simulations, m a

General Parameters Gaussian 10 Fourier transform inverse Fourier transform 1 0.1

Sensing regions of the fiber. See Figures 3 and 4.

Figure 3. Results of simulation experiments with an OTOF distributed attenuation-based sensor and an OTOF distributed sensor based on a statically quenched immobilized fluorophore (both type I sensors). (A) Wave form calculated according to eqs 3 and 4 using parameters of Table 1 (solid line) and deconvoluted with the temporal profile of excitation pulse Rex (dashed line). (B) First derivatives of the logarithm of the wave form (solid line) and when deconvoluted with Rex (dashed line). The sharp spikes are artifacts caused by derivatization of the wave form. The off-scale peaks are due to the large slope of the original wave form. I-V designate the sensing regions as described in Table 1.

impulse response from a measured signal PII is more complicated due to the variable impulse response of the immobilized fluorophore Rem over the fiber length when exposed to different concentrations of a quencher. Here we introduce a method to retrieve information about the analyte-modulated fluorescence lifetime of an immobilized fluorophore at any location along the fiber. The method begins with the intensity profile of the recorded wave form, an example of which is shown in curve B of Figure 4. Figure 4 pertains to a distributed oxygen sensor, with the parameters of Table 1 and exposed to three regions of different oxygen concentrations as shown in Figure 4A. The number of regions used in this example is sufficient to demonstrate the signal-processing principle and in practice can be easily extended to additional (or even continuous) zones. The localized fluorescence lifetime of the immobilized fluorophore is extracted 1458 Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

sequentially from each of the regions of the sensor, starting from either end of the fiber (from the launch end in this example). In step one of the method, the initial returned wave form PII (Figure 4B) is deconvoluted with the excitation pulse profile Rex and corrected for the exponential decay in fluorescence intensity as a function of fiber length. This decrease in intensity originates from the attenuation of the fiber (see eq 3). In distributed sensors of type II, this attenuation is constant along the fiber length. Thus, the fluorescence intensity at each point l along the fiber is normalized by the factor exp(-Rl). The result is illustrated in Figure 4C. In step two, the wave form is deconvoluted with Rem (the impulse response of the immobilized fluorophore of fluorescence lifetime τem) when the fiber is exposed to an unknown oxygen concentration in region I. This deconvolution, performed recursively with different values of τem, eventually yields a step profile of the signal in region I of the wave form (Figure 4D). This value of τem can be used to determine the oxygen concentration in region I. In step three, the contribution of region I is removed from the wave form shown in Figure 4D. The resulting wave form is then convoluted with the lifetime of the immobilized fluorophore calculated above for region I; this process restores the effect of region I on the rest of the wave form. The result is shown in Figure 4E. Steps two and three are repeated to find τem of region II (Figure 4F) and of region III (Figure 4G and H). In step four, oxygen concentrations are calculated according to eq 6 using the values of τem computed for each region; these concentrations are plotted as a function of fiber length (Figure 4A). Thus, although the OTOF wave form of a type II sensor is generated in terms of fluorescence intensity, the information extracted from it is the fluorescence lifetime of the immobilized fluorophore in each region. This lifetime is immune to variations in the wave form amplitude caused by photobleaching of the immobilized fluorophore or other factors; the lifetime information is extracted from the shape of the profile rather than from its magnitude as in sensors of type I. Analysis of Experimental Wave Forms. The initial OTOFCD experiments were aimed at determining the optical attenuation of the chemically modified fiber and the contributions of static fluorescence quenching to the total fluorescence signal. The

Figure 5. Back-propagated fluorescence intensity from the distributed oxygen sensor as a function of elapsed time (or equivalent fiber length) for the entire fiber exposed to nitrogen, air, and oxygen. The spike at 480 ns (48 m) is due to Fresnel reflection from the far end of the fiber.

Figure 4. Results of simulation experiments with an OTOF distributed sensor utilizing a dynamically quenched immobilized fluorophore (type II sensor). (A) Oxygen concentration profile used for simulations. (B) Wave form simulated according to eqs 3 and 5 using parameters of Table 1. (C) Result of deconvolution of the wave form in (B) with the temporal profile of excitation pulse Rex followed by baseline correction of the exponentially decaying fluorescence intensity to compensate for constant fiber attenuation. (D) Result of deconvolution of the wave form in (C) and Rem (the impulse response of the immobilized fluorophore). The fluorescence lifetime τem is matched to that of region I. (E) Result of removing the contribution of region I from the wave form in (D) and convolution with the impulse response of the immobilized fluorophore Rem of fluorescence lifetime τem (I). (F) Result of deconvolution of the wave form in (E) and Rem of the fluorescence lifetime τem matched to that of region II. (G) Result of removing the contribution of region II from the wave form in (F) and convolution with the impulse response of the immobilized fluorophore Rem of fluorescence lifetime τem (II). (H) Result of deconvolution of the wave form in (G) and Rem with the fluorescence lifetime τem matched to that of region III. I-III pertain to the sensing regions as described in Table 1.

Figure 5. The optical attenuation of the modified fiber was calculated according to eq 9 from the slope of the three wave forms over the entire fiber length to be - 0.121 ( 0.001 dB/m (mean ( SD; n ) 3). The slope of the fluorescence wave form of the sensor exposed to different oxygen concentrations was the same, illustrating that only dynamic (collisional) fluorescence quenching47 took place. The slight variations in slope (e.g., at 25 m) are presumably due to imperfections in the fiber or cladding. These imperfections lead to local variations in numerical aperture and cause corresponding variations in the fraction of fluorescence captured into the fiber core. The change in back-propagated signal is in agreement with earlier reports on telecommunicationgrade fibers.48 To investigate detection sensitivity and reproducibility along the fiber length, the data were plotted according to eq 6 as a function of the elapsed time or equivalent fiber length (Figure 6). The relative standard deviation (RSD) of the measurements worsened from 0.5 to 2.5% RSD with distance from the launch end of the fiber because of the fiber attenuation. The detection limit for the distributed sensor (at a S/N ) 3) was calculated from the slope of the calibration curve at the lowest oxygen concentration. The detection limit steadily degraded from 1 to 3% oxygen as a function of fiber length. Earlier we showed15 that incorporation of TPP into the silicone cladding of a 1.5-mlong PCS fiber and use of a yellow light-emitting diode for fluorescence excitation resulted in a 0.5% detection limit for oxygen. The poorer detection limit of the distributed sensor reported here is due to the less stable light source and lower level of generated fluorescence. The only earlier use of TPP for oxygen sensing, obtained with the fluorophore adsorbed onto porous polystyrene beads,22 did not cite a value for the detection limit. The 1-s response time was the same as reported previously15 for our 1.5-m-long sensor used for spatially averaged oxygen detection.

results of these experiments, performed by exposing the entire fiber sequentially to nitrogen, air, and oxygen, are depicted in

(47) Hercules, D. M. In Fluorescence and Phosphorescence Analysis; Hercules, D. M., Ed.; Wiley: New York, 1966; pp 1-40. (48) Di Vita, P.; Rossi, U. Electron. Lett. 1979, 15, 467-469.

Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

1459

Figure 6. Sensitivity of the distributed oxygen sensor to oxygen and air calculated according to eq 6. The profiles are the mean values of five measurements. Every 25th data point is shown with an error bar, which represents one standard deviation.

Figure 8. Comparison of experimental results (data points) and results calculated according to eqs 3 and 5 (solid lines) for spatially resolved oxygen monitoring. See caption of Figure 7 for details.

effects. This deviation is within the expected range. For example, mode-coupling effects probably contribute at least 10% to this discrepancy49 while the excitation pulse of the laser exhibited a 6% deviation from the assumed Gaussian.

Figure 7. Spatially resolved oxygen determinations with the distributed oxygen sensor. Data are shown for four experiments in which the 48-m-long continuous oxygen-sensitive fiber was divided into 13 sensing regions. Several sensing regions were exposed to nitrogen or oxygen, while the rest of the fiber was in air. (A) Regions 2 and 4 in nitrogen; (B) regions 2 and 4 in oxygen; (C) region 11 in nitrogen; (D) regions 11 and 13 in oxygen.

Spatially resolved measurements were performed by exposing segments of the continuous sensing fiber to oxygen and nitrogen while the rest of the fiber was in air. The wave forms recorded in these experiments were normalized by a similar wave form obtained with the entire fiber in air. Results, presented in Figure 7, show a spatial resolution limited only by the fluorescence lifetime of the immobilized fluorophore and the pulse width of the laser. Numerical simulations were performed using the parameters of the immobilized indicator (cf. Table 1) and the 13-ns pulse width of the laser to predict the observed intensity wave form in spatially resolved measurements. Experimental data and numerical simulations agree well (Figure 8). Possible sources of the remaining 20% discrepancy are the non-Gaussian profile of the excitation pulse, pulse broadening in the optical fiber, variable attenuation of different modes propagating in the fiber, and mode-coupling 1460 Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

CONCLUSIONS A theoretical evaluation and experimental findings show the promise of absorption- and fluorescence-based distributed sensors for spatially resolved analyte monitoring by means of OTOF-CD. With this technique, it is possible (i) to locate the zones in the fiber where analyte-induced changes in absorption or fluorescence take place, (ii) to determine the magnitude of these variations, and (iii) to relate the magnitude of the variations to the concentration of an analyte. This technique offers several advantages for distributed chemical monitoring. These advantages include the possibility of quantitative measurements, operation with multimode fibers, simple signal formation and processing, and inexpensive optoelectronic components. Application of such distributed chemical sensors with spatially resolved analyte mapping capabilities is attractive for a variety of practical needs. For example, continuous analyte monitoring with distributed chemical sensors can simultaneously indicate and locate when the accepted level of exposure to toxic or explosive species has been exceeded, can track a source of contamination in an industrial or technological process, can follow the formation and movement of environmental pollutants, and can perform early nondestructive detection and location of corrosion (such as inadvanced structural components of space shuttles, aircraft, and civil engineering structures). Deconvolution is useful to boost the spatial resolution of OTOF distributed sensors. Realization of such high-resolution sensors will require the application of fast sampling rates and total data acquisition times that are long enough to produce high-quality wave forms so the artifacts of deconvolution are reduced. The basic OTOF-CD technique introduced here should be improved to satisfy additional requirements of chemical sensing. These requirements include a uniform signal-to-noise ratio over (49) Ruddy, V.; Shaw, G. Appl. Opt. 1995, 34, 1003-1006.

the length of the sensing fiber, an expanded range of detectable analytes, and the capability for multianalyte determinations. We are currently developing detection techniques to meet several of these needs. ACKNOWLEDGMENT This work was supported in part by the National Institutes of

Health through Grant GM 53560. R.A.P. acknowledges support from a McCormick Science Grant provided by the College of Arts and Sciences, Indiana University. Received for review December 4, 1997. February 25, 1998.

Accepted

AC9713159

Analytical Chemistry, Vol. 70, No. 8, April 15, 1998

1461