An Autonomous Sensor and Telemetry System for Low-Level pCO

Anal. Chem. 1999, 71, 154-161

An Autonomous Sensor and Telemetry System for Low-Level pCO2 Measurements in Seawater Mary Beth Tabacco,† Mahesh Uttamlal,‡ Mariann McAllister, and David R. Walt*

Max Tishler Laboratory for Organic Chemistry, Department of Chemistry, Tufts University, 62 Talbot Avenue, Medford, Massachusetts 02155

The measurement of low-level dissolved CO2 using a fiberoptic sensor is described. The sensor, based on the Severinghaus CO2 electrode principle, consists of a CO2sensitive bicarbonate buffer solution containing the pHsensitive fluorescent dye carboxy-SNAFL-1 immobilized at the end of an optical fiber using a gas-permeable membrane. The sensor is used in a ratiometric mode and has a reversible working dynamic range between 200 and 1000 ppm pCO2 and a sensitivity (1 ppm. Results are presented for the sensor calibration, effects of temperature, and response time characteristics. An integrated measurement system with electrooptic and data acquisition modules coupled to a satellite transmission system was tested in Vineyard Sound, MA, and data are presented that demonstrate continuous monitoring of pCO2 in surface seawater. The accurate monitoring of low-level pCO2 (0-1000 ppm) is important in many systems. For example, CO2 is used as an aerial fertilizer in greenhouses; CO2 enrichment from ambient levels (345 ppm) to 1000 ppm can improve tomato yield by 35%.1 Similarly, the production or disappearance of CO2 is a key parameter in assessing the performance of various fermentation and bioreactor processes in the biotechnology industries. Therefore, robust sensing technology for the fast and accurate determination of lowlevel CO2 is highly desirable. This paper describes the development of a pCO2 sensor and the related enabling technology to monitor seasonal changes in surface CO2 in ocean waters over extended periods of time. At present, pCO2 seawater measurements are obtained by research ships using water-sampling techniques. Such an approach is expensive and provides low spatiotemporal resolution due to the limited numbers of samples that can be taken. These measurements are important to understand global changes in the environment brought about by the burning of fossil fuels and the destruction of rain forests.2 Specifically, wide-reaching, long-term monitoring of pCO2 is required to complete models of ocean-atmosphere coupling and to balance the global CO2 budget.3 * Corresponding author (e-mail) [email protected]). † Permanent address: Echo Technologies, Inc., 451 D S., Boston, MA 02210; (e-mail) [email protected]. ‡ Permanent address: Department of Physical Sciences, Glasgow Caledonian University, City Campus, Cowcaddens Rd., Glasgow G4 0BA, Scotland, UK; (e-mail) [email protected]). (1) Hand, D. Grower 1985, 104 (3), 31. (2) Sarmiento, J. L. Chem. Eng. News 1993, 70 (22), 30.

154 Analytical Chemistry, Vol. 71, No. 1, January 1, 1999

While CO2 in the gas phase is usually determined using IR measurements, dissolved CO2 is most easily measured using chemical sensors. Most chemical sensors for dissolved CO2, including the one described here, are based on the principles behind the Severinghaus electrode.4 The electrode consists of a pH electrode in contact with a bicarbonate buffer solution trapped on the electrode surface by a gas-permeable membrane, such as PTFE or silicone rubber. For optical sensors, the pH electrode is replaced with an absorbance- or fluorescence-based pH-sensitive indicator coupled to an optical fiber. The sensor (optical or electrochemical) measures the pH of the HCO3- solution which is in equilibrium with CO2 outside the membrane, i.e.: membrane

Ka

K

CO2 798 CO2 + H2O 798 H2CO3 798 K2

HCO3- + H+ 798 CO32- + 2H+ The external CO2 concentration is related to the internal [H+] by5

h3 + nh2 - (K1aT + Kw)h - 2K1K2aT ) 0

(1)

where n is the concentration of sodium ions in the internal solution, h ) [H+], KW ) h[OH-], a is the total analytical concentration of carbon dioxide, i.e., a ) [CO2]aq + [H2CO3], K1 ) KKa ) hb/a, and K2 ) hc/b (b ) [HCO3-] and c ) [CO32-]). The literature has many reports of fiber-optic CO2 sensors. Some are absorbance-based6,7 and others fluorescence-based;8-11 however, most are suitable only for high CO2 levels (0.01-1 atm) and only a few report ppm range sensitivity. We previously reported using inner filter effects to enhance sensitivity; however, even with this method the maximum resolution was (7 ppm.12 DeGrandpre13 reported the most sensitive optical sensor for dissolved CO2. It operates in the 0-1000 ppm CO2 range and has (3) Sundquist, E. T. Science 1993, 259, 934. (4) Severinghaus, J. W.; Bradley, A. F. J. Appl. Physiol. 1956, 13, 515. (5) Jensen, M. A.; Rechnitz, G. A. Anal. Chem. 1979, 51, 515. (6) Vurek, G. G.; Peterson, J. I.; Goldstein, S. R.; Severinghaus, J. W. Fed. Proc. Am. Soc. Exp. Biol. 1982, 41, 1484. (7) Mills, A.; Chang, Q.; McMurray, N. Anal. Chem. 1992, 64, 1383. (8) Munkholm, C.; Walt, D. R. Talanta 1988, 35, 109. (9) Uttamlal, M.; Walt, D. R. Bio/Technology 1995, 13, 597. (10) Mills, A.; Chang, Q. Analyst 1993, 118, 839. (11) Zhujun, Z.; Seitz., W. R. Anal. Chim. Acta 1984, 160, 305. (12) Walt, D. R.; Gabor, G.; Goyet, C. Anal. Chim. Acta 1993, 274, 47. (13) DeGrandpre, M. D. Anal. Chem. 1993, 65, 331. 10.1021/ac980513c CCC: $18.00

© 1998 American Chemical Society Published on Web 12/03/1998

an accuracy of (0.8 ppm. Unlike most sensors that are fixed reagent sensors, his sensor operates by the constant replacement of the sensing solution at the distal end of the fiber via a fluid pumping system. This paper describes the development of a robust, fixed-reagent fiber-optic chemical sensor (FOCS) for low-level dissolved CO2 which is capable of operating for many months without user intervention. The sensor is based on the Severinghaus principle and incorporates the fluorescent pH indicator, 5′ (and 6′)carboxyseminaphthofluorescein (c-SNAFL-1), immobilized in a polymer film at the distal end of an optical fiber. The sensor is used in a ratiometric mode to account for system instabilities. We present calibration data, response time characteristics, and temperature effects and assess its performance for the determination of pCO2 in surface seawater. In addition, the results of an initial at-sea deployment are reported.

(cross-linker), 0.5 mL pH 7.3 phosphate buffer, and 30 mg of benzoin ethyl ether (photoinitiator). This solution was degassed with argon, and 100 µL was placed on a microscope slide and covered with a cover slip. The slide was exposed to longwavelength UV light for 5 min. The slide was then immersed in distilled water. The fragile polymer film was removed carefully from the glass, placed in the indicator solution, and left 24 h before use. Sensor Preparation and Calibration. The sensor was prepared by cutting a 3-mm-diameter disk using a small cork borer and mechanically fixing it to the distal end of a 400-µm-diameter optical fiber using a 10-µm-thick PTFE membrane (Goodfellows Corp., Berwyn, PA) (Figure 4). The sensor was stored in 0.67 M NaCl. Calibrations were performed in 0.67 M NaCl through which CO2 gas/balance in N2 (Northeast Airgas, Nashua, NH) was bubbled.

MATERIALS AND METHODS Materials and Reagents. Unless otherwise stated, all chemicals were AnalaR grade (or their equivalent) and were used without further purification. All solutions were prepared from distilled water. Instrumentation. Excitation and emission spectra were performed on a fiber-optic double-monochromator fluorescence system described previously.8 Calibrations and other continuous on-line measurements were performed on a custom-built portable fluorometer (Steve Brown Engineering, Livermore, CA) interfaced to an IBM-compatible PC. The sensor optoelectronic interface consists of a custom-designed, compact, hermetically sealed fluorometer which uses a light-emitting diode (LED) for excitation, dichroic and band-pass filters for separating and detecting the emitted light, and photodiode and lock-in amplifier detection electronics. This system was configured with a 485 nm with 22nm band-pass excitation filter (Omega, Brattleboro, VT), and the emission filters were 540 nm with 30-nm band-pass and 630 nm with 30-nm band-pass. The extended band-pass dichroic had a wavelength cutoff of 505 nm. The integration time for each measurement was 2 s. Data acquisition rate, filter switching, lamp, and photodetector were software controlled. The portable fluorometer was modified for at-sea tests by incorporation of an Onset Computer TT8 data logger and Persistor (Peripheral Issues) flashcard with extended memory. Power was provided by a marine battery recharged using two 10-W solar panels (Atlantic Solar Products). Line-of-sight communication was possible using a set of spread spectrum transceivers (Xetron Corp.), and data were telemetered using a SEIMAC PTT transmitter via the ARGOS satellite system. Indicator Solution Preparation. The indicator solution was prepared as follows: a 100 µM c-SNAFL-1 (Molecular Probes Inc., Eugene, OR) solution was prepared in water or in 0.67 M NaCl. The solution was bubbled with 10% CO2 in N2 for 10 min to generate carbonic acid. The final pH of the solution was adjusted to 8.2 using 1 M NaOH. The solution was stored at 4 °C in the dark. Indicator Solution Support Preparation. The indicator solution support was poly(N-vinylpyrrolidone). The polymer was prepared by the photopolymerization of N-vinyl-2-pyrrolidone monomer stock solution. The stock solution contained 0.5 mL of N-vinyl-2-pyrrolidone (monomer), 10 µL ethylene dimethacrylate

RESULTS c-SNAFL in Aqueous Solution. c-SNAFL is a fluorescent pH indicator dye and was first reported by Whitaker et al.14 A full study of the properties of the indicator are described elsewhere,12,15,16 and are summarized in Table 1. An important feature of the dye is that it possesses two emission peaks when excited at 488 nm. These peaks are centered at 540 and 620 nm and result from the protonated and deprotonated forms of the dye, respectively. This dual-wavelength feature of the dye makes it particularly

suitable for use in a ratiometric mode which accounts for system instabilities, such as photobleaching and lamp intensity fluctuations. There is also an isosbestic point at 625 nm (Ex ) 488 nm) which can be used for ratioing and to assess photobleaching of the dye. The fluorescence intensity of c-SNAFL-1 results from the following equilibrium established in aqueous solution: Kin

HIn y\z H+ + Inwhere HIn and In- are the protonated and deprotonated forms of the of the dye, respectively, and Kin is the acid dissociation constant for the dye in its ground electronic state:

Kin ) hd/dh

(2)

where d ) [In-] and dh ) [HIn]. Also (14) Whitaker, J. E.; Haugland, R. P.; Prendergast, R. P. Anal. Biochem. 1991, 194, 330. (15) Szmacinski, H.; Lakowicz, J. R. Anal. Chem. 1993, 65, 1668. (16) Mordon, S.; Devoisselle, J. M.; Soulie, S. J. Photochem. Photobiol. B. Biol. 1995, 28, 19.

Analytical Chemistry, Vol. 71, No. 1, January 1, 1999

155

Table 1. Physical and Chemical Properties of Carboxy-SNAFL-1 (at 21 °C)

refs 12-14

in H 2O

in 0.67 M NaCl

sensor (at 12 °C) (calcd)

481 nm (acid) 510 nm (acid) 539 nm (base) 542 nm (acid) 615 nm (isos) 623 nm (base) 7.75 6.13 0.09 0.0192 3.78 -22.6

545 632 618 8.01 5.41 0.145

542 624 620 7.67 5.27 0.129

545

property absorbance emission (Ex ) 488 nm) pKin I(543/623)(max) I(543/623)(min) ∆pH/∆T/°C-1 ∆H°/kcal mol-1 ∆S°/cal mol-1 K-1 pK1 [HCO33-]/µM

610 7.98 5.54

6.53 158

dhT ) d + dh

(3)

Figure 1. Emission spectra of carboxy-SNAFL-1 in distilled water (Ex ) 488 nm). The maximums centered at 540 and 620 nm arise from the protonated and deprotonated forms of the dye, respectively.

where dhT is the total indicator concentration. Equation 3 can be rewritten in the form of the HendersonHasselbach equation:

pH ) pKin - log

() dh d

(4)

Let fluorescence intensity, I ) dh and Io ) dhT, i.e., the fluorescence intensity when the indicator is fully undissociated. Then eq 4 becomes

pH ) pKin - log

( ) I Io - I

(5)

or, in terms of I

I)

1 +(10

Io pKin-pH -1 )

+B

(6)

In this equation we introduce an additional term B, which is the background fluorescence of the system and is an experimentally derived value. I can also be the ratio of the fluorescence intensities of the protonated and isosbestic forms, i.e., I ) I545/Iisos. Figure 1 shows the emission spectra of c-SNAFL-1 in distilled water (25 °C) at various pH values using 488-nm excitation light. Figure 2 shows the corresponding ratio vs pH plot. Also, shown in Figure 2 is the calibration performed in 0.67 M NaCl solution. Experiments were carried out in 0.67 M NaCl because it is necessary to balance the osmotic pressure of the indicator solution in the sensor to that of the test solution. The osmotic pressure of seawater is equivalent to 0.67 M NaCl. The solid lines in Figure 2 are theoretical fits of eq 5. The results for Io, pKin, and B are summarized in Table 1 together with typical values reported elsewhere. Table 1 shows that increasing the ionic strength decreases pKin17 and the maximum fluorescence intensity also decreases with increasing ionic strength. 156 Analytical Chemistry, Vol. 71, No. 1, January 1, 1999

Figure 2. Titration curves for carboxy-SNAFL-1 in distilled water and in 0.67 M NaCl. The solid lines are the theoretical curve fits using eq 6. The pKa for the indicator increases with increasing ionic strength.

The indicator solution for the sensor was prepared by making a 100 µM c-SNAFL-1 solution in 0.67 M NaCl containing bicarbonate ions (∼150 µM). The pH of this indicator solution to varying pCO2 is shown in Figure 3. CO2 was bubbled through the indicator solution, which was held at room temperature. Sensor Preparation. The intended oceanographic application for the CO2 sensor necessitated several critical design considerations dealing primarily with chemical and mechanical stability and longevity. Ideally, the sensor design would allow for unattended operation (no intermittent calibration) for periods in excess of six months. The sensor used in our study (Figure 4a) consists of a 400-µm-diameter single-core fiber in a PEEK housing (Oxford Electrodes, Abington, UK). The PEEK material is chemically stable and mechanically robust and can be precision machined. There is a cavity around the fiber that contains the indicator solution (∼50 µL total volume). A disk of poly(N-vinylpyrrolidone) (17) Albery, W. J.; Uttamlal, M. J. J. Appl. Electrochem. 1994, 24, 8.

Figure 3. Emission spectra (Ex ) 488 nm) of the pCO2 indicator solution at various pCO2 tensions. The inset shows the corresponding calibration curve using the ratio of the emission intensities at 542 and 625 nm. The experiment was performed at 21 °C.

As shown in the Figure 4 inset, the fiber interrogates a relatively small region in the center of the NVP membrane. CO2 in seawater diffuses through the entire membrane and may even penetrate somewhat into the reservoir solution. The response time is determined by how long it takes for CO2 to equilibrate with the solution in front of the fiber. The additional membrane volume, outside of the interrogation region, assists in this equilibration by providing a pathway for lateral diffusion of CO2 and dye within the NVP membrane. As discussed in a subsequent section, the rather slow response of the sensor is a consequence of the relatively thick NVP layer as well as its porosity. Dye from the reservoir is continually replenishing the indicator in the NVP pores so that photobleaching of the dye in front of the fiber does not compromise the signal. The reservoir solution contains ∼4000× as much dye as is contained in the interaction region.18 The operation of the sensor is based on the Severinghaus electrode principle. When carbon dioxide crosses the membrane, the pH of the indicator solution is affected as given by the Henderson-Hasselbach equation:

pH ) pKin - log(aT/b)

(7)

where aT ) KHKMpCO2, KH/mol dm-3 atm-1 is Henry’s constant, and KM is the membrane constant. Substituting eq 7 into eq 5 and rearranging gives

I)

Io 1 + (10

pKin-pK1+log(aT/b) -1

)

(8)

for a > 0 and

I)

Figure 4. Sensor construction. The sensor consists of a 400-µmdiameter single core fiber in a PEEK housing. There is a cavity around the fiber which contains the indicator solution (∼50 µL total volume). A disk of N-vinylpyrrolidone polymer, presoaked in indicator solution, is sandwiched between the fiber surface and an outer gas-permeable membrane (10-µm-thick PTFE).

(NVP) polymer, ∼100 µm thick serves as the sensing membrane. After screening numerous hydrogel candidates, this polymer was chosen because of its hydrolytic stability in seawater. Acrylatebased polymer systems were found to hydrolyze slowly, causing the equilibrium pH of the indicator solution to change and making the sensor insensitive to CO2. The polymer is presoaked in indicator solution and sandwiched between the fiber surface and an outer gas-permeable membrane (10-µm-thick PTFE), thereby providing a fixed optical path length interaction region. The porous polymer provides a pathway for the natural convection of the indicator solution.

1 + (10

Io pKin-pHo -1 )

(9)

for a ) 0. Where pHo is the pH of the indicator solution at zero CO2. After the sensor is first prepared, it is stored in 0.67 M NaCl solution for several days and allowed to equilibrate with the external solution before use. During this time, water vapor crosses the membrane until the osmotic pressure on both sides of the membrane is matched. Figure 5 shows the sensor calibration. The dissolved CO2 concentration was determined using Henry’s constant.19,20 We assume aT to be equal to the dissolved CO2 concentration in the bulk solution, i.e., KM ) 1. The sensitivity of the sensor is ∼(1 ppm. The theoretical equation relating the fluorescence intensity ratio to pCO2 can be tested on the data and is shown by the solid curve in Figure 5. From this analysis, the values for the constants in eq 6 can be determined and are summarized in Table 1. The pHo value is calculated from eq 5 using pKin while Io is derived (18) The interrogation volume is calculated using a cylindrical volume and does not include the dispersion angle of light exiting the fiber. Also, the membrane volume is not corrected for the porosity, or water content, of the membrane. These parameters would have opposite effects. (19) Cox, J. D.; Head A. J. J. Chem. Soc., Faraday Trans. 1962, 58, 1839. (20) Markham, A. E.; Kobe, K. A. J. Am. Chem. Soc. 1941, 63, 449.

Analytical Chemistry, Vol. 71, No. 1, January 1, 1999

157

Figure 5. Sensor calibration. Solid line is the theoretical curve fit to the data using eq 8.

from the above analysis. There is good agreement between experimental and theoretically derived data. Effect of Temperature on the Sensor Response. The temperature of ocean water is subject to change on a daily basis and therefore it was necessary to determine whether these changes would affect the sensor response. In the waters off the coast of New England, water temperature varies in the range 5-23 °C. CO2 solubility is also affected by temperature.21 This latter effect is eliminated by performing all experiments in CO2-free 0.67M NaCl in a controlled-temperature bath. Temperature changes can affect the sensor response in three ways. First, temperature may affect the fluorescence differently at the two excitation wavelengths. For example, increasing temperature reduces the quantum efficiency of most molecules and could reduce the fluorescence intensity of one transition relative to the other. Second, temperature increases cause a decrease in pKin and at a given pH the ratio increases with increasing temperature. Finally, the pH of the bicarbonate buffer solution is temperature dependent with the pH decreasing with increasing temperature. The overall effect of temperature on the sensor response is a combination of all three processes. Using the fundamental thermodynamic relationships ∆G° ) -RT ln K and ∆G° )∆H° - T∆S° it has been shown that for a system containing one indicator dye and one principal pH buffer22

d log(IB/IA) dT

-1

)

∆H°buffer - ∆H°ind 2.303 R

(10)

where IA and IB are the fluorescence intensities of the acid and base forms, respectively. According to eq 10, a plot of log(IA/IB) vs T-1 should yield a straight line of slope (∆Hobuffer - ∆Hoin)/ 2.303R. Figure 6 shows a plot of log(IA/IB) vs T-1, which, as predicted, yields a straight line. These results suggest that for an accurate determination of pCO2 an independent measurement of temperature must be made. Response Time Characteristics. The development of performance criteria for new chemical sensors is by necessity (21) Morrison, T. J.; Billett, F. J. Chem. Soc. 1952, 3819. (22) Hafeman, D. G.; Crawford K. L.; Bousse, L. J. J. Phys. Chem. 1993, 97, 3058.

158 Analytical Chemistry, Vol. 71, No. 1, January 1, 1999

Figure 6. Effect of temperature on the sensor response. The experiment was performed at 12 °C in N2-saturated solutions. Plot shows the corresponding log(ratio 610/545) vs 1/T according to eq 10.

application specific. For example, a process control application (fermentor or bioreactor) might require faster response times than the present oceanographic application. Oceanographic pCO2 data will be used to better understand the global CO2 budget, and therefore, the sensor must be able to monitor slowly changing conditions with good resolution and accuracy. Response times of minutes to hours are acceptable for this application. The response time characteristics for the Severinghaus pCO2 electrode have been described by several workers,4,23-25 Much of these previous studies dealt with response time characteristics where concentration step changes were relatively large. In a previous paper, we described the response time characteristics of a pCO2 sensor and showed that they are similar to those of the Severinghaus electrode. The theory predicts that small step changes at lowlevel pCO2 exhibit much longer response times than large step changes because, at low levels, a large proportion of the CO2 crossing the membrane is consumed by the reaction with H2O, CO32-, HCO3-, and the pH indicator dye before equilibrium across the membrane is established. For large step changes, only a small fraction of the permeating CO2 is used in this process resulting in a much shorter response time. Figure 7 shows the response time profiles for step changes in pCO2 in 0.67 M NaCl at 12 °C. The results are summarized in Table 2 and are consistent with the theory described above with small step changes having longer response times than large step changes. The response times are also very much longer than those of high-level pCO2 sensors, again, consistent with theory. It is important to note, however, that the sensor responds immediately to even a small change in CO2. The response time of the sensor can be improved by adding carbonic anhydrase26 to the indicator solution. This enzyme catalyzes the CO2 hydration reaction, which is the rate-limiting step in the sensor response. For the seawater experiments, we did not use any carbonic anhydrase as we felt if would denature during long deployments. (23) Van der Schoot, B.; Bergveld, P. Anal. Chim. Acta 1984, 166, 93. (24) Ross, J. W.; Riseman, J. H.; Krueger, J. A. Pure. Appl. Chem. 1973, 36, 473. (25) Donaldson, T. L.; Palmer, H. J. AIChE J. 1979, 25 (1), 143. (26) Merz, K. M., Jr. J. Am. Chem. Soc. 1989, 111, 5636.

Figure 7. Response time characteristics for step changes in pCO2. Experiment performed in 0.67 M NaCl at 12 °C. The data also show the long-term stability of the sensor and precision of ∼1.9 ppm at a concentration of 200 ppm. Table 2: Response Time Characteristics (12 °C)a

a

∆CO2 ([CO2]o ) 200 ppm)

t90/min

150 300 600

130 100 60

The starting [CO2] for each step change was 200 ppm.

pCO2 in Seawater. There is clear evidence from environmental monitoring that a large proportion of the CO2 produced by the burning of fossil fuels has a substantial, but not quantified, ocean sink. The extent to which this occurs and effects on the system brought about by climatic change are not fully understood.27 This lack of understanding is due in part to the lack of extended time series data. The monitoring of pCO2 in surface seawater has been achieved using titrimetry,28 coulometry,29 gas chromatography,30 and IR spectrometry.31 Several FOCSs for seawater have been reported,10,11,32 but most do not exhibit the sensitivity required for monitoring the relatively small changes seen in ocean waters. DeGrandpre demonstrated an absorption-based fiber-optic sensor system with excellent reported accuracy of (2 ppm in the laboratory, but the system requires a fluid handling system which may not be appropriate for extended, autonomous operation.32 The integrated sensor system was deployed in the Atlantic Ocean on a discus-type buoy located ∼0.3 km offshore (Woods Hole, MA) as shown in Figure 8. The fiber-optic cable was guided through an Extren tube extending off the side of the buoy to provide a rigid support, and when in place, the sensor was ∼2 m (27) Sarmiento, J. L. U. S. JGOFS News 1995, 6 (2), 4. (28) Dyrssen, D. Acta Chem. Scand. 1965, 19, 1265. (29) Johnson, K. M.; King, A. E.; Sieburth, J. McN. Mar. Chem. 1985, 16, 61. (30) Weiss, R. F. J. Chromatogr. Sci. 1981, 19, 611. (31) Dickson, A. G.; Goyet, C. Handbook of Methods for the Analysis of the Various Parameters of the Carbon Dioxide System in Seawater; ORNL CDIAC report; Oak Ridge National Laboratory; 1994; Vol. 2, p 5.1. (32) DeGrandpre, M. D.; Hammar, T. R.; Wallace, D. W. R.; Wirick, C. D. Limn. Oceanogr. 1997, 42 (1), 21.

Figure 8. Fiber-optic pCO2 sensor system deployed on a discus buoy in Vineyard Sound, MA, ∼0.3 km offshore, 41°31′50′′ N, 70°38′26′′ W.

below the sea surface. A light baffle was installed at the end of this tube to eliminate intense scattered sunlight in the upper water column. A TT8 data logger is integrated into the instrument and used to control power up, timing, and data acquisition parameters. A platform transmitter terminal (PTT) is integrated into the electronics, and data are telemetered every 90 s via the ARGOS satellite system. This data transmission protocol allows data to be sent from the sensor system to a central station where it can be accessed via the Internet. There is approximately a 2-3-h delay time between data transmission and availability of data to the user. Therefore, a spread spectrum transceiver set was installed and used for line-of-sight communication with the sensor system. The transceivers provide real-time, two-way communication with the system. This capability is particularly useful during emplacement on a buoy at sea using a ship or small craft and permits verification of system status and adjustment of parameters such as gain, phase, and signal integration times (prior to leaving the vicinity of the buoy). After 8 months of characterization in the laboratory, the CO2 measurement system was deployed from July 2 to August 20, 1997. A subset of the collected data is shown in Figure 9, where the ratio, S1/S2, corresponding to the fluorescence intensity from the individual photodetector channels is plotted as a function of time. Figure 9 suggests that the sensor is identifying diurnal variations in pCO2 arising from changes in surface seawater temperature and from biological activity. Based on our laboratory calibration data, the mean CO2 concentration is ∼380 ppm during the first two weeks of the test. The apparent drift upward in the signal ratio starting around hour 320 is likely related to microbial fouling, which changes the CO2 concentration in the microenvironment around the sensor. Observation of the sensor tubing after retrieval showed substantial fouling of the outer Extren guide tube and moderate fouling of the fiber cable and sensor housing. Water samples were taken periodically from August 4 to August 20, 1997 and were subsequently analyzed at Woods Hole Oceanographic Analytical Chemistry, Vol. 71, No. 1, January 1, 1999

159

Figure 9. pCO2 collected during the summer of 1997. Diurnal variations in pCO2 are evident.

Institution. Total alkalinity and total CO2 were determined by potentiometric titration using a method derived from one first described by Dyrssen28 and later modified by Bradshaw et al.33 The automated titration was performed in a closed cell maintained at constant temperature (25 ( 1 °C). The ionic strength of the hydrochloric acid solution (0.1 N) was adjusted with NaCl to better approximate seawater. The precision of the measurement is estimated to be better than 0.15%. These analyses give a mean seawater concentration of ∼357 ppm. CONCLUSION This work has demonstrated the feasibility of using a fiberoptic sensor and data acquisition system for unattended monitoring of pCO2 in seawater. The fiber sensor is robust enough to survive a range of weather and wave conditions (after being operational in the laboratory for eight months). Furthermore, the ability to telemeter data using the ARGOS satellite system and to control system parameters remotely supports the concept of using such a system for long-term deployment on either a stationary buoy or a drifting, expendable buoy. Improvements in the sensor response time would provide better temporal resolution which might be required for application in a more dynamic environment such as coastal waters or tidal basins. As mentioned in a previous section, inclusion of an enzyme, such as carbonic anhydrase, has been shown to significantly decrease response time. The fiber-optic system measurements could be further enhanced by real-time seawater temperature measurement and correction of sensor response, improving the temperature stability of the electronics (e.g., the LEDs and certain components on the lock-in amplifier boards are temperature sensitive), and development of more refined algorithms to correct for both effects. Problems associated with microbial fouling are being addressed by coating all support tubing with antifoulant (33) Bradshaw, A. L.; Brewer, P. G.; Shafer, D. K.; Williams, R. T. Earth Planet. Sci. Lett. 1981, 55, 99.

160 Analytical Chemistry, Vol. 71, No. 1, January 1, 1999

paint, and a controlled-release antifouling sleeve will be evaluated in the future. The system is presently being tested under the more challenging conditions of the open ocean, and for an extended period of time (4 months) in the Sargasso Sea as part of the Bermuda Testbed Mooring Program. This test program is also significant in that data taken as part of the Bermuda Atlantic Time Series (BATS) program will be available for comparison. The BATS time series data are probably the best available for intercomparison, but even its limited scope underscores the need for wide-reaching, extended time series pCO2 measurements. ACKNOWLEDGMENT The authors thank the National Science Foundation, Grant OCE-9102670, for financial support. We also thank Mr. Tony Schanzle of JAS Research, Inc. for engineering support; Dr. Ed Sholkovitz and Mr. Dave Hosom of Woods Hole Oceanographic Institution (WHOI) for technical support; Dr. Catherine Goyet and Mr. Greg Eischeid of WHOI for analyzing the water samples; and Dr. Dana Kester of the University of Rhode Island for earlier field support. LIST OF SYMBOLS a

[CO2]aq (mol dm-3)

aT

[CO2]aq + [H2CO3] in the indicator solution layer

B

background fluorescence intensity

b

[HCO3-] (mol dm-3)

c

[CO32-] (mol dm-3)

d

[In-] unprotonated dye concentration

dh

[HIn] protonated dye concentration

dhT

total dye concentration

D

diffusion coefficient of CO2 through the membrane

h

[H+] (mol dm-3)

∆H°buffer

dissociation enthalpy for the buffer under standard conditions

∆H°in

dissociation enthalpy for the indicator under standard conditions

K1

)KKa

K2

second acid dissociation constant for carbonic acid

I

fluorescence intensity of protonated form of the indicator dye or I545/Iisos

KW

dissociation constant for water

Kin

acid dissociation constant for the indicator

Ix

fluorescence intensity at wavelength x

KM

membrane constant

Io

maximum fluorescence intensity (ratio)

n

sodium ion concentration in the indicator solution

IA

fluorescence intensity of the acid form of the indicator dye

pHo

pH of indicator solution when [CO2] ) 0

IB

fluorescence intensity of the base form of the indicator dye

K

equilibrium constant for CO2/H2CO3 system

Received for review May 11, 1998. Accepted October 7, 1998.

Ka

first acid dissociation constant for carbonic acid

AC980513C

Analytical Chemistry, Vol. 71, No. 1, January 1, 1999

161