Anal. Chem. 1995, 67, 2650-2654
Development of a Submicrometer Optical Fiber Oxygen Sensor Zeev Rosenzweig and Raoul Kopelman* Department of Chemistry, The University of Michigan, Ann Arbor, Michigan 48 109
A submicrometer optical fiber oxygen sensor has been fabricated, based on the fluorescence quenching of tris(1 ,10-phenanthroline)ruthenium(II) chloride in the presence of oxygen or dissoived oxygen. The Ru compound has been incorporated into acrylamide polymer that is attached covalently to a silanized optical fiber tip surface by photoinitiated polymerization. Leaching of the sensing reagent from the polymer host matrix has been minimized by the optimization of the ratio between the acrylamide monomer and the cross-linker, N,N-methylenebisacryIamide. The sensor is fully reversible and highly reproducible. A standard deviation of -2% for 10 consecutive fluorescence measurements has been observed for several oxygen concentrations. The sample volume required for measurements is 100 fL. An absolute detection limit of 1 x 10-17 mol is achieved. This is an improvement by a factor of lo6 as compared to other existing optical fiber oxygen sensors. The determination of molecular oxygen levels in solution is of great importance in environmental, biomedical, and industrial analyses. The development of in situ and real-time measurements using on-line fiber-optics systems is a possible alternative to the current environmental monitoring systems. In the past few years, a number of works have been published describing the use of optical fibers in monitoring 02 as well as dissolved oxygen in water. In most of these investigations, absorption or fluorescence-type indicator^^-^ have been immobilized in an organic polymer matrix and incorporated with optical fibers. The quenching of fluorescence is then correlated with the oxygen concentration. Luminescent transition metal complexes, especially those based on Ru, represent an important class of fluorescence indicators used in these sensors. Ru compounds show great photostability and have long lifetimes and high quenching efficiencies suitable for oxygen sensing. The photophysics and photochemistry of Ru compounds have been extensively studied by Demas et a1.8 An optical fiber sensor with a Ru complex as (1) Pal, T.;Jana, N. R.; Das, P. K Analyst 1991,116, 321-322. (2) Baldini, F.; Bacci, M.; Cosi, F.; Del Bianco. A,; Scheggi, A. M. OFS Proc., Opt. Fiber Sens. Cont, 8th 1992,325-328. (3) Goswami. K; Kleiner, S. M.: Tokar, J. M. Proc. SPIE Int. SOC. Opt. Eng. 1988,990, 117-126. Chemical, biomedical and environmental applications of fibers. (4) Peterson, J. I.; Fitzgerald, R. V.; Buchhold, D. K Anal. Chem. 1984,56, 62-67. (5) Liberman, R. A.: Blyler, L. L.: Cohen, L. G. OFS Proc.. OFS Tech. Dig.Ser. 1988,2,346-348. (6) Surgi, M. R.OFS Proc., OFS Tech. Dig.Ser. 1988,2, 349-351. (7) Klimant, I.; Belser, P.: Wolfieis, 0. S. PYOC.EUY.Conf Opt. Chem. Sens. Biosens., 1st (Graz, Austria) 1992,130. (8) Demas. J. N.; Degraff, B. A. Anal. Chem. 1991,63, 829-837.
2650 Analytical Chemisfry, Vol. 67, No. 75,August 7, 1995
oxygen level indicator has been developed and applied successfully in serum? In another application, an optical fiber oxygen sensor, again based on the fluorescence quenching of a Ru complex, was used for oxygen monitoring in immobilization matrices.1° Rubased indicators can also be a part of a sensor array, designed for simultaneous measurement of pH, oxygen, and carbon dioxide." Wolfbeis et al. have reported the development of the first optical fiber oxygen sensor for the determination of the biochemical oxygen demand in yeast cells.12 The measurement of oxygen levels at submicrometer spatial resolution is of great importance in studying not only biological systems but also fundamental phenomena like electrode performance or corrosion. However, true microresolution cannot be achieved with the current sensor technology, because the smallest core size of optical fiber oxygen sensors reported to date is about 100 pm, with a typical response time of 1-10 s.13 In this paper we describe the fabrication and the analytical properties of a submicrometer optical fiber oxygen sensor. Like the optical fiber sensors described above, it is based on the fluorescence quenching of a Ru complex and therefore can be applied to the measurements of oxygen levels in simple water solutions as well as more complicated biological and technological environments. There are two major diffkulties in making submicrometer optical fiber sensors. One is the attachment of the sensing molecules to the optical fiber surface. The attachment of chemical reagents by membrane or tubing or any other mechanical method makes only very limited use of the fiber end surface and signscantly increases the sensor size, which translates into low signals and long response times. The immobilization of the sensing reagent by photoinitiated polymerization, according to the method suggested by Walt et al.,14 allowed us to fabricate submicrometer optical fiber sensors. In the photoinitiated polymerization approach, the polymer serves to increase the surface area and the number of molecules available for sensing. The response time of the sensor is faster since no membrane is used. The manufacturing of the submicrometer optical fiber oxygen sensor is simple with respect to other sensors, since the Ru compound, tris(l,l@phenanthroline)rutheniumOI) chloride, also serves as the initiator for photopol~merization.~~ The ~~
~
~~
~
(9) Singer, E.; Duvenech, G.;.I Ehrat. M.; Widmer, H. M. Sem. Actuators 1994 A 4 2 (1-3), 542-548. (10) Kohls, 0.;Andres. K. D.; Scheper, T. Biofomm 1994,17 (3). 44-51. (11) Walt, D. R: Bamard, S. M.; US. Patent 5,244,636, 1994. (12) Preininger, C.; Klimant, I.; Wolfbeis, 0. S. Anal. Chem. 1994,66, 18411846. (13) Shahrirari. M. R; Ding, J. Y.; Tong, J.; Sigel, G. H., Jr. Chemical, biomedical and environmental applications of fibers. Proc. SPIE-lnt. SOC.Opt. Eng. 1993,224-240. (14) Munkholm, C.; Walt, D. R.; Milanovich, F. P.; Klainer. S. M. Anal. Chem. 1986,58, 1427-1430. (15) Iwai, IC;Uesugi, Y.; Sakabe. T.; Hazama, C.; Takemura, F. Polym. J. 1991, 23 (6), 757-763.
0003-2700/95/0367-2650$9.00/0 8 1995 American Chemical Society
other difficulty in making submicrometer optical fiber sensors is that the smallest size optical fibers commercially available are in the range of 50 pm (cladding diameter). This difficulty is overcome by pulling the optical fiber to produce a tip with a core size down to 0.1 pm. As for submicrometer pH ~ e n s o r s , the ~~J~ fabrication of submicrometer optical fiber oxygen sensors is realized by forming fiber tips and then incorporating the dye into a polyacrylamide matrix which is attached covalently to the silanized fiber tip surface via photoinitiated polymerization. There are significant differences between the miniaturized pH and oxygen sensors. First, the oxygen sensing reagent, which is an inorganic complex of Ru, shows higher photostability as compared to pH dyes. On the other hand, we found that under excitation wavelength of 488 nm and a dye concentration of loL4 M, the fluorescence intensity of tris(l,l@phenanthroline)rutheniumaI) chloride is 3 times lower than that of acrylamide fluorescein (the pH indicator). As a result, we had to improve the sensitivity of our detection system in order to measure fluorescence spectra from submicrometer oxygen sensors. Second, the Ru complex is physically trapped in the polyacrylamide matrix, while the pH sensing reagent (fluorescein amine) is copolymerized with the acrylamide monomer. Physical immobilization is a more general technique, having the advantage of not being associated with chemical modification and potentially the spectral properties of the sensing reagent. Therefore, the conditions for the sensor fabrication have to be optimized to avoid leaching of the sensing reagent from the host matrix. This is a major concern, especially with submicrometer sensors, because of the small number of molecules available for sensing. The optimization of the host matrix pore size, photostability, reversibility, response time, linear dynamic range, and detection limit performance of our miniaturized oxygen sensor, the smallest oxygen sensor to date, are discussed in this paper. EXPERIMENTAL SECTION Fiber Tip Fabrication. The fabrication of optical fiber tips has been previously described by Betzig et al.lB The apparatus we used for fiber tip forming is described e1se~here.l~ Briefly, it consists of a P-87 micropipet puller (Sutter Instrument Co.) and a 25 W CO2 infrared laser (Synrad Co.). Single mode optical fibers were purchased from 3M Co., with core diameters of a few micrometers and cladding diameter around 100 pm. The end of the optical fiber is soaked in methylene chloride for 10 s. The jacket is then removed with forceps from 3 to 5 cm of the optical fiber where the tip is about to be formed. The tip is produced by drawing the optical fiber in the puller while heating it with a laser beam power of 10 mW. The laser beam is focused on the optical fiber. The beam diameter at the focal point is around 1 mm. A pretension force of 4 N is applied along opposite directions of the fiber prior to heating. When exposed to the laser beam, the fiber is heated and instantly separated to form a pair of 0.1-0.5 pm pulled tips. A vacuum chamber with a base pressure of 10-7 Torr, equipped with an aluminum coil, is employed for coating the pulled fiber tips. Only the fiber tip sides are coated with aluminum, leaving the end face as a transmissive aperture. The selective (16) Tan, W.; Shi, Z.; Smith, S.; Bimbaum, D.; Kopelman, R. Science 1992,258, 778-781. (17) Tan, W.; Shi, 2. Y.; Kopelman, R Anal. Chem. 1992,64,2985-2990. (18) Betzig, E.; Trautman, J. IC;Hanis, T. D.; Weiner, J. S.; Kostenak, R L. Science 1991,251,1468-1470. (19) Tan, W. Ph.D. Thesis, University of Michigan, Ann Arbor, MI, 1993.
Ar ion laser coupler
i
at 488 nm
1
Figure 1. Submicrometer optical fiber oxygen sensor apparatus. The instrument consists of Ar ion laser as a light source, optical fiber oxygen sensor, fluorescence microscope, and two detection systems: (a) spectrograph and charge-coupled device (CCD) camera for spectroscopic measurements and (b) photomultiplier and photon counter for fluorescence intensity measurements. AT-PC microcomputer is used for data acquisition and analysis.
coating is achieved by positioning the pulled tips at an angle of 30" with respect to the aluminum coating source. The specific angle of rotation is crucial for the achievement of perfect tips and has to be optimized for each coating chamber. Sizes and shapes of the fiber tips are determined by scanning electron microscopy. Fabrication of a Submicrometer Optical Fiber Oxygen Sensor. The optical fiber pulled tip is silanized by dipping the tip for 1 h into a 20 mL water solution containing 2% [y-(methacryloxy)propyl]trimethoxysilane. After silanization, the tip is inserted into the polymerization solution while an Ar ion laser beam at 488 nm is used for photopolymerization. The solution for the photopolymerization contains 35%acrylamide, 2%-7% N,Nmethylenebisacrylamide as cross-linker, 0.1 M triethylamine as electron donor, and 2 x M tris(1,lGphenanthroline)ruthe nium(II) chloride, all dissolved in a phosphate buffer at pH 6.5. When a laser power of 10 mW is coupled into the opposite end of the optical fiber, the fabrication of the polyacrylamide dye-doped sensor is concluded in 1-3 min. The optical fiber sensor is rinsed in a water solution for 48 h to release all the dye molecules that were not trapped inside the polymer matrix. The submicrometer optical fiber oxygen sensor is then ready for use. Detection System. An inverted Olympus microscopebased optical fiber sensor setup is describe in Figure 1. An Ar ion (Spectra Physics) laser at 488 nm is used as a light source. A high-precision single mode optical fiber coupler (Newport) is used to direct the laser beam into the fiber sensor tip. The sensor reagent is excited by the incident beam, and the fluorescence light as well as the excitation light is collected by a 2Ox objective lens. The fluorescence is transmitted by a dichroic mirror, while the excitation light is reflected and stopped. For obtaining spectra, the fluorescence signal is collected by a set of two flat field camera lenses (Nikon) to match thefnumber of the spectrograph grating. A 150 mm spectrograph (Acton) with three mirrors and 600 grooves/" grating is used to disperse the fluorescence signal. The influence of the dichroic mirror on the spectra, collected with the spectrograph, is neglected because of the large Stokes shift in our system. A 488 nm laser beam is used for excitation while the emission peak is at 610 nm. A liquid nitrogen-cooled charge coupled device (CCD) camera (Princeton Instruments, Model LNCCD1024E/1) is used for spectral data collection. This experiAnalytical Chemistry, Vol. 67, No. 15, August 1, 1995
2651
Photographs of optical fiber oxygen sensors (a) Sensor prepared from an unpulled, 100 flm core, muhimode fiber (b) Sensor prepared from an unpulled, single mode, 3-5 pm core, optical fiber (c) Sensor prepared from a pulled, submicrometer optlcal flber t ~ pThe scale bars represent (a) 25, (b) 10, and (c) 10pn Figure 2.
mental arrangement is used for sensors larger than 1pm, It is not as useiitl for fluorescence measurements from submicrometer optical fiber sensors. That is because of the significant optical loss associated with the large number of optical elements in this detection system. Data analysis of fluorescence from submicrometer optical fiber sensors is based on total intensity measure ments. A photomultiplier (Hammamatsu R928), installed on top of the fluorescencemicroscope, is used for the intensity measure ments. A photon counter (EG&G photon counter, Model 1109, and discriminator control, Model 1121) is utilized for data acquisition. While this experimental arrangement is not suitable for spectroscopic measurements, it is much simpler with respect to the optical system and therefore leads to a 10 times better detection limit. A compatible PC microcomputer is used for data acquisition and analysis. Reagents. Tris(l,lOphenanthrolne)ruthenium~9chloride, acrylamide. and Nflmethylenebisacrylamide were purchased from Aldrich Chemical Co. The fiber silanization reagent, [y-(methacryloxy)propyllhimethoxysilane, was obtained from Sigma. All chemicals were used as received.
2.1 0
2
4
6
8
10
TIME (days) Figure 3. Leaching of dye from an optical fiber oxygen sensor. The dye is physically trapped in an acryiamide polymer that is anached covalently to the optical fiber end. The polymerization mixtures contain 35% acrylamide monomer, 2 x IO-' M tris(1,lo-phenanthroline). ruthenium(l1) chloride, and 2%, 3%. 5%, and 7% cross-linker N , N methylenebisacrylamide.
RESULTS AND DISCUSSION Submicrometer Optical Fiber Oxygen Sensor. Three
optical fiber oxygen sensors are shown in Figure 2: 100pm sensor F i e Za), 3 pm sensor (Figure Zb), and 0.8 p m pulled optical fiber sensor (Figure 2c). As mentioned previously, the sides of the pulled fiber are coated with aluminum, the flat end of the tip is left bare, and a tiny aperture is formed. When a 488 nm laser beam is coupled into the aluminumcoated fiber tip, a bright spot of fluorescence at the tip is observed (Figure Zc). The emission peak is at 610 nm. The polymer tip itself is not seen because of the limited spatial resolution (1pm) of the fluorescence micro scope. The actual size of the pulled sensor at F i r e 2c has been determined with an environmentalscanning electron microscope to be 0.8 pm. Careful observation of the sensor with high magnikation (6ox microscope objective) reveals a homogeneous fluorescence from most of the polymer tip. Based on this finding. we conclude that the thickness of the sensor is similar to the size of the polymer tip. The sensor size might vary as a result of swelling if the sensor is employed in organic solvents or if it is excited with a UV light source. For our conditions, 488 nm excitation and application to water solutions, no swelling effects are observed. 2652
Analytical Chemistry, vel. 67, No. 15, August
1, 1995
leaching of Dye from the Polyacrylamide-Based Optical Fiber M e n Sensor. As mentioned above, in our oxygen sensor, the tris(l,lOpbenanthrolne)Nthenium(II) chloride dye molecules are physically trapped in the acrylamide matrix, and leaching of the dye from the polymer matrix is a major concern. The leaching rate depends on the average pore sue in the host matrix. The microstructure of acrylamide polymers has been a subject for intensive investigations. Using transmission electron microscopy KEM). Ruche1 et al?O found a parabolic dependence between the cross-linker concentration and the average pore size. Ahern et al.2' employed surface-enhanced Raman scattering (SERS) to verify this dependence. While there is no agreement as to the exact structural details, both studies have shown that a minimum pore sue of 10 i 6 nm is realized when using a composition of 35%-40%acrylamide monomer and 5%noss-linker. Actual leaching measurements from our optical fiber oxygen sensor are shown in Figure 3. The polymerization solution for (20) (a) Ruche!, R: Bwer, M.D.Anal Biochcm. 1975.68.415-428,ib) Ruche], R: Steere, R L:Erbe. E. F.J Chronrotogr 1978.166.563-575. (21) Ahem. A M.:Garrell, R L Langnruir 1988.4,1162-1168.
2000
1
I
0 ppm 0 , -
p w
n
I
I
I
1500
U
1000
0
2
500
0 680
600
620
640
860
WAVELENGTH (nm) Figure 4. Fluorescence spectra from a 2 pm optical fiber oxygen sensor at 0, 4, 8, and 40 ppm dissolved oxygen; 1 ppm is the equivalent of 1 mg/L or 3 x M dissolved oxygen.
the fabrication of the sensor consists of 35% acrylamide, 0.1 M triethylamine, 2 x M tris(l,l@phenanthroline)ruthenium(Ir) chloride, and different percentages of the cross-linker NJVmethylenebisacrylamide. The measurements have been performed in air-saturated water solutions. The experimental results are in agreement with the predictions from refs 21 and 22. It can be seen that a more rapid degradation of the sensor occurs when 2%cross-linker is used. Stability is maximized at 5%-7% crosslinker. A cross-linker percentage of 5%has been chosen for the fabrication of most sensors. That is because the permeability of the polymer matrix to oxygen decreases with the pore size of the polymer matrix. Photostability of Optical Fiber Oxygen Sensors. The photostability of optical fiber sensors is a well-known concern. It is probably the main reason why optical fiber fluorescence sensors have not yet been commercialized. Several mechanisms are involved in the photodegradation of polymer-based optical fiber fluorescence sensor. One is direct photobleaching, in which the fluorophore is consumed upon illumination. The rate of photobleaching increases nonlinearly with the number of exciting photons. In our experimental arrangement, the Ar ion laser excitation beam is carefully overlapped spatially with an He-Ne laser beam. The He-Ne laser beam is used to align and optimize the light coupling into the optical fiber. The sensor is therefore exposed to the excitation light only during the actual data collection. This has been achieved by installation of an electronic shutter that is triggered by the CCD controller 20 ms prior to the exposure of the CCD chip and is closed 10 ms after the exposure has been terminated. A typical exposure time for the CCD chip is 0.1-1 s. The second photobleaching mechanism is unique to polymer hosts. Contaminants trapped in the polymer matrix or even the polymer matrix itself may induce nonradiative transitions in the dye molecules and therefore reduce the effective quantum efficiency. A 2-fold decrease in the fluorescence signal of dye molecules when trapped in an organic polymer matrix is a common observation. This problem may be avoided by trapping the fluorescence dye in an inorganic sol-gel matrix.13 Another
photodegradation process results from self-photobleachingwhen interactions between the dye molecules lead to self-quenching. We found that the fluorescence signal from our sensor is sharply decreased upon illumination if the Ru dye concentration in the polymerization mixture exceeds 5 x M. For our application, the polymer host remains attractive since the size of the sensor is fully controlled via photoinitiated polymerization. As previously mentioned, the Ru dye is highly photostable. We observed a 6% variation in the fluorescence intensity during 1 h of constant illumination. This should be contrasted with our short exposure time (0.1-1 s) per measurement. For an exposure time of 0.5 s, 1 h illumination time means 7200 independent measurements. Performance of Submicrometer Optical Fiber Oxygen Sensor in Water Solution. A detailed discussion of the predicted performance of an optical fiber oxygen sensor based on fluorescence quenching is given by Wolfbeis.22The analytical range of an oxygen optode is governed by the respective quenching curve and the Stern-Volmer constant. The variation in the fluorescence intensity as a function of the dissolved oxygen concentration is given by the Stern-Volmer equation,
where l o is the fluorescence intensity of the sensor dipped in a nitrogen-saturated solution, I, is the fluorescence intensity of the sensor in a given dissolved oxygen concentration,and KSVis the Stern-Volmer quenching constant. In principle, higher quenching constants result in a better accuracy at low levels of oxygen because of a larger relative signal change per oxygen concentration interval. However, high quenching constants result in a more limited linear dynamic range. We found that the linear range of our sensor is between 0 and 12 ppm dissolved oxygen; 1 ppm dissolved oxygen is defined as 1mg of molecular oxygen dissolved in 1L of water solution. The fluorescence spectra obtained from a 2 pm (0.d.) optical fiber oxygen sensor at different dissolved oxygen concentrationsare shown in Figure 4. It can be seen that the relative change in fluorescence between 4 and 8 ppm is comparable with the change in fluorescence between 8 and 40 ppm dissolved oxygen. The sensitivity factor, Zo/Zf(Oz) = 3.2, where Zf(02) is the fluorescence intensity from oxygen-saturated solution, is comparable with those of other oxygen optodes using organic polymer mat rice^.^^^^^ A Stern-Volmer analysis of a submicrometer optical fiber oxygen sensor is shown in Figure 5. KSVis found to be 5420 M-l, with a correlation coefficient of 0.996 in the linear range of the sensor. The accuracy of the oxygen optode is governed by the uncertainties in the determination of I, (nitrogen saturated solution), KSV,and I,. An accurate determination of ZO is essential for obtaining a sufticiently precise calibration curve and KSVvalue. We found a standard deviation of -2% for 10 consecutive measurements at 4 ppm dissolved oxygen. As expected, the standard deviation is increased at lower oxygen levels when two large fluorescence intensities are sub tracted from each other to obtain a small intensity difference. The limit of detection COD) of the sensor, which is defined as 2 times the standard deviation (2SD), is about 0.3 ppm dissolved oxygen. In terms of the oxygen concentration, this LOD is comparable with those of other oxygen optodes. The absolute detection limit
~
(22) Wolfbeis, 0.Fiber Optic Chemical Sensors and Biosenson; CRC Press: Boca Raton, FL, 1991;Vol. 2, p 19.
(23) Hauser, P. C.; Tan, S. S. Analyst 1993,118 (8), 991-995. (24) Murray, J. U S . Patent 4,752,115, 1988.
Analytical Chemistry, Vol. 67,No. 75,August 7, 7995
2653
3r
I
I
I
K,, = 5420 M” 2.5
n
1.2 1
0.8
-
2 -
so
0.6 0.4
-
0.2 4
0
. 1
0 0.5 0
I
1
,
5
10
15
20
30
40
TIME (minutes)
20
DISSOLVED OXYGEN @pm> Figure 5. Stern-Volmer plot of a fiber-optic dissolved oxygen, 0.8 pm sensor. KSV = 5420 M-l, with a correlation coefficient of 0.996 between 0 and 12 ppm.
is 1 x mol. This absolute detection limit is improved by 6 orders of magnitude as compared to those of any other optical fiber oxygen sensor. The high reversibility of our miniaturized oxygen sensor is demonstrated in Figure 6, where Z/ZO is plotted at different oxygen levels. The sensor shows 99%recovery after an oxygen-saturated solution is replaced with an oxygen-free water solution. It should be noted that a fast change in the oxygen level is not easily achieved experimentally. In our experiment, we applied a constant flow of oxygen above the water sample and allowed 5 min to achieve equilibrium. Based on the Einstein equation for diffusion, the response time of our sensor is predicted to be in the millisecond range. Our measurements in the gas phase (data not shown) show that the response time is
