Environ. Sci. Technol. 2001, 35, 1475-1480
A Simple Automated Continuous-Flow-Equilibration Method for Measuring Carbon Monoxide in Seawater H U I X I A N G X I E , † O L I V E R C . Z A F I R I O U , * ,† WEI WANG,† AND CRAIG D. TAYLOR‡ Department of Marine Chemistry and Geochemistry, and Biology Department, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts 02543
A simple, robust, low-maintenance method using airsegmented continuous-flow equilibration was developed and automated to measure carbon monoxide (CO) in natural waters precisely and accurately. Finely regulated flows of CO-free air and of seawater or standard water were pumped into a glass coil, forming discrete gas/liquid segments. The partially CO-equilibrated gas effluent was injected into a Trace Analytical reduction analyzer for CO detection. A semiempirical mass-balance model was established for predicting and optimizing the performance of the CO extractor. The optimized gas and water flow rates were ∼1.2 and ∼14 mL min-1, respectively, giving a response time of less than 15 min and a CO-extraction yield of ∼80%. The analytical blank, precision, and accuracy were, respectively, 0.02 nM, (2.5% (at the ∼1 nM level), and better than 5%. Two extractors can be interfaced to one detector at 4-6 samples per hour for each extractor. Coupled with a continuous surface-water sampler, the system was successfully applied to monitoring the diurnal variation of CO concentration in Sargasso Sea surface waters.
Introduction The surface ocean is supersaturated with carbon monoxide (CO) with respect to its atmospheric concentration (1-3) and hence is a source of atmospheric CO, which plays a key role in regulating the concentration and distribution of hydroxyl radicals in the troposphere (4, 5). The concentration of CO in the surface exhibits large and highly variable diurnal variations (1-3) due to photoproduction and loss by rapid microbial uptake and gas exchange (1, 6). The diurnal cycle of CO is also substantially affected by vertical mixing in the water column (7-10). Due to this large variability and to a factor of 2 uncertainty in gas-exchange coefficients used in estimating air-sea fluxes (2), the major aspects of CO cycling have been explored productively using sampling and analysis methods of moderate reproducibility and little-studied precision and accuracy (although it is noteworthy that recent CO sea-air flux estimates differ by nearly 100-fold (2, 11).) More recently it has been recognized that such shortlived photoproducts as CO have excellent potential for quantitatively studying the couplings among upper-ocean optics and photoprocesses, microbial and chemical loss * Corresponding author phone: (508) 289-2342; fax: (508) 4572164; email:
[email protected]. † Department of Marine Chemistry and Geochemistry. ‡ Biology Department. 10.1021/es001656v CCC: $20.00 Published on Web 02/27/2001
2001 American Chemical Society
FIGURE 1. (A) Gas (CO) extractor. The Teflon Tee has 1/4 in. Swagelok fitting. The glass coil is 6.1 m long, 4 mm i.d., and 6 mm o.d. The gravitational head in the water exit leg of the separator forces gas through the sample loop of the CO analyzer when solenoid valve SV1 (Figure 2) is open. (B) Diagram showing mixing of CO and phase equilibration. processes, upper-ocean mixing dynamics, and gas exchange (7-10). However, realizing this potential depends crucially on the ability to generate high-resolution data sets of wellcharacterized precision and accuracy, so that the discrepancies between modeled and experimentally determined results are identifiable with confidence and sensitivity. There is thus a requirement for analytical methods of demonstrated accuracy which are better characterized yet rapid, convenient, and robust enough to permit generating large data sets at sea, sometimes under adverse conditions (since CO samples cannot be stored). To meet this requirement, we developed an automated gas-segmented continuous-flow-equilibration (CFE) analysis method and successfully applied it to continuously monitoring surface concentrations of CO in the Sargasso Sea. Determination of dissolved trace gases (CO and H2) using the CFE technique was first reported by Setser et al. (12), but no details were given. Here we present detailed information about the structure, operation, and performance of our system and illustrate some of its applications. We also derived an empirical mathematical model for predicting and optimizing the performance of this method. Elsewhere we described a manual headspace analysis approach with welldemonstrated precision and accuracy for analyzing discrete samples, showed that better samplers and sampling are also required to meet the need stated above, and described appropriate equipment and procedures (13).
Experimental Section Principle. A precisely regulated sample water flow (sample or standard) entered a glass coil (6.1 m long, 4 mm i.d., and 6 mm o.d. (custom order, Ace Glass)) via one port of a Teflon Tee (Cole-Parmer). A finely controlled flow of CO-free air was introduced into the other port, forming regularly spaced air-water segments (Figure 1A). As the air-segmented flow traversed the coil, phase equilibration of CO between water and air was facilitated by wall-induced longitudinal mixing VOL. 35, NO. 7, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1475
FIGURE 2. Automated two-coil CFE analysis system. SV stands for solenoid valve and MFC for mass flow controller. Extractors are immersed in water baths (not shown). Details of the CO extractors are shown in Figure 1A. Connecting lines are 316 stainless steel tubing, except the section between SV1 and the extractors, where stainless steel and Peek tubing are used interchangeably. Coils are precleaned by soaking with 0.1 N HCl for ∼24 h and rinsing with distilled water. Coils are kept in dim (UV-free) light. in the water segments (Figure 1B). Partially phase-equilibrated air was collected into a custom-blown glass air-water separator (Anderson Glass, Fitzwilliam, NH) and injected into a CO analyzer. Procedure and Automation. Figure 2 shows a schematic diagram of the automated two-coil system employed on two cruises to monitor diurnal variation of CO concentration in Sargasso Sea surface waters near Bermuda. A newly developed surface-following sampler continuously pumped seawater (∼550 mL min-1) from one or two depths into the coils through 1/4 in. o.d., 316-stainless steel lines. These water streams were subsampled by an IsmaTec precision multichannel miniature peristaltic pump (model 7331-10) injected at flow rates, fw, of ∼14 mL min-1 into the Teflon Tees attached to vertically oriented coils. The coils were immersed in water baths in contact with the ship’s (air-conditioned) laboratory air to buffer temperature variations. Daily changes in temperature were usually within 2 °C. An OMEGA multichannel digital thermometer was used to record the temperatures of the water baths. The (minor) effect of temperature changes on the solubility coefficients of CO was taken into account in calculating CO concentration; a 1 °C change results in 10 psi) ensured against calamitous injection of water due to occasional condensation in the 1/16 in. tubing carrying the gas streams. After one or two cycles, a blocked filter resulted in CO signals so small as to be evidently spurious. The theory of the mercuric oxide reduction detector is described by McCullough et al. (14) and Schmidt and Seiler (15); analytical details are reported elsewhere (13). The RGA3 contained two columns (0.32 cm by 76.8 cm): the first packed with Unibeads 1S (60/80 mesh) and the second with Mole Sieve 5A (60/80 mesh). One sample was analyzed every 12-15 min for each coil; switching between coils was controlled by a solenoid valve (SV1). Immediately after an injection from the first coil, the sample stream from the second coil was switched in line while the air stream from the first coil was blocked by SV1 and forced out of the bubble separator (Figure 1A). About 6-7.5 min later, the sample from the second coil was injected and the air stream from the first coil switched on, completing an analysis cycle and allowing the gas from each coil to continuously flush and overfill the sampling loop in the RGA3 prior to injection. An aqueous CO standard in a sample-water matrix was prepared by continuously purging ∼20 L of untreated surface seawater with 9.755 ( 0.098 ppm gaseous CO standard (Scott Specialty Gases, Inc, certified NIST-traceable). Surface seawater collected in the study area was contained in an opaque stainless steel tank to prevent photochemical CO formation. At ∼100 mL min-1, the CO gas standard supplied ∼7 µmol of CO per day to the standard-water. The biological CO consumption rate was estimated to be only 0.23 µmol of CO per day (∼8 nM standard, biological consumption rate constant 1.5 d-1 (Zafiriou et al. Unpublished data)), so that biological processes were too slow to affect the concentration and stability of the aqueous CO standard. In coastal and estuarine waters, higher gas flow or a more efficient
equilibrator might be required. The standard water was maintained at laboratory temperature, replenished daily, and allowed to preequilibrate for ∼3 h at 250 mL min-1. Solenoid valves SV2 and SV3 switched between unknown water and standard water at programmable intervals, calibrating each coil every 2-3 h. The initial concentration of CO in a water sample, [CO]aq (nM), is given by
[CO]aq ) 103 × βmeqp2/(RT) where meq (ppm) is the atmospheric mixing ratio of CO expected from equilibrium with the sample, which is calculated directly from its peak area relative to the aqueous standard, since the system responds linearly with concentration. β (mL of CO per mL of H2O per atm) is the Bunsen solubility coefficient of CO, a function of temperature and salinity (16); p (atm) is the atmospheric pressure of air corrected for water vapor; R is the gas constant (0.08206 atm L mol-1 K-1); and T is the temperature (K). The system was fully automated by a home-designed LabView-coded program that timed and actuated the switching of all three solenoid valves, the loading and injection of samples into the RGA3, and starting a Chromatopac C-R6A integrator (Shimadzu Corp.). The program also acquired realtime peak information that was subsequently processed by a home-designed Matlab-coded program to derive peak areas. The difference between C-R6A and computer integration values was
