Anal. Chem. 2007, 79, 4169-4176
Flow Injection Analysis of H2O2 in Natural Waters Using Acridinium Ester Chemiluminescence: Method Development and Optimization Using a Kinetic Model D. Whitney King,*,† William J. Cooper,‡ Steven A. Rusak,*,§ Barrie M. Peake,§ James J. Kiddle,| Daniel W. O’Sullivan,⊥ Megan L. Melamed,† Chris R. Morgan,† and Stephen M. Theberge†
Chemistry Department, Colby College, 5755 Mayflower Hill, Waterville, Maine 04901-8857, Urban Water Research Center and Department of Civil and Environmental Engineering, University of California, Irvine, Irvine, California 92697-2175, Chemistry Department, University of Otago, P.O. Box 56, Dunedin, New Zealand, Department of Chemistry, University of Western Michigan, 3425 Wood Hall, Kalamazoo, Michigan 49008, and Department of Chemistry, U.S. Naval Academy, 572 Holloway Road, Annapolis, Maryland 21402
Chemiluminescence (CL) of acridinium esters (AE) has found widespread use in analytical chemistry. Using the mechanism of the reaction of H2O2 with 10-methyl-9-(pformylphenyl)acridinium carboxylate trifluoromethanesulfonate and a modified flow injection system, the reaction rates of each step in the mechanism were evaluated and used in a kinetic model to optimize the analysis of H2O2. Operational parameters for a flow injection analysis system (reagent pH, flow rate, sample volume, PMT settings) were optimized using the kinetic model. The system is most sensitive to reaction pH due to competition between AE hydrolysis and CL. The optimized system was used to determine H2O2 concentrations in natural waters, including rain, freshwater, and seawater. The lower limit of detection varied in natural waters, from 352 pM in open ocean seawater (mean, 779 pM ( 15.0%, RSD) to 58.1 nM in rain (mean, 6,340 nM ( 0.92%, RSD). The analysis is specific for H2O2 and is therefore of potential interest for atmospheric chemistry applications where organoperoxides have been reported in the presence of H2O2. Chemiluminescence (CL) of acridinium esters (AE) has found widespread use in analytical chemistry and biochemistry. Low detection limits and specificity of acridinium ester probe compounds have led to AE-CL methods being used for many applications, particularly for the analysis of hydrogen peroxide in a diverse range of natural waters 1-3 where H2O2 exists at * Corresponding authors. E-mail:
[email protected]. Tel: +207-859-5755. Fax: +207-859-5760. E-mail:
[email protected]. Tel: +64 3 479 7477. Fax +64 3 479 7906. † Colby College. ‡ University of California, Irvine. § University of Otago. | University of Western Michigan. ⊥ U.S. Naval Academy. (1) Cooper, W. J.; Moegling, J. K.; Kieber, R. J.; Kiddle, J. J. Mar. Chem. 2000, 70, 191-200. (2) Miller, G. W.; Morgan, C. A.; Kieber, D. J.; King, D. W.; Snow, J. A.; Heikes, B. G.; Mopper, K.; Kiddle. J. J. Mar. Chem. 2005, 97, 4-13. 10.1021/ac062228w CCC: $37.00 Published on Web 04/25/2007
© 2007 American Chemical Society
concentrations ranging from subnanomolar in deep oceans 4 to micromolar in rain,5,6 fogwater, and cloudwaters,7,8 and snow and firn.9 Hydrogen peroxide is involved in redox chemistry in natural waters,10-16 in the oxidation of SOX to S(IV) in the atmosphere,17,18 and is implicated in a range of biological processes19-20 along with its most common precursor, superoxide radical, HO2/O2•-. H2O2 is the most stable of the compounds commonly referred to as reactive oxygen species and may serve as a proxy for the timeintegrated production of these species in natural waters. Thus, a useful analytical procedure for H2O2 in natural waters must have a relatively low detection limit and a linear dynamic range that spans 5 or 6 orders of magnitude. Additionally, analysis of H2O2 in natural water samples is often complicated by interference from biogenic natural organic matter, which may be both colored and fluorescent.21 For a number of analytical methods, organoperox(3) Zhang, F.; Zhuang, H. Fenxi Huaxue 1992, 20, 342-344. (4) Johnson, K. S.; Coale, K. H.; Elrod, V. A.; Tindale, N. W. Mar. Chem. 1994, 46, 319-334. (5) Cooper, W. J.; Zika, R. G.; Saltzman, E. S. J. Geophys. Res. 1987, 92, 29702980. (6) Kieber, R. J.; Peake, B. M.; Willey, J. D.; Jacobs, B. Atmos. Environ. 2001, 35, 6041-6048. (7) Anastasio, C.; Faust, B. C.; Allen, J. M. J. Geol. Phys. 1994, 99, 8231-8248. (8) Faust, B. C. Environ. Sci. Technol. 1994, 28, 217A-222A. (9) Bales, R. C.; Losleben, M. V.; McConnell, J. R.; Fuhrer, K.; Neftel, A. Geophys. Res. Lett. 1995, 22, 1261-1264. (10) Abu-Saba, K. E.; Flegal, A. R.; Sedlak, D. L. Mar. Chem. 2000, 69, 33-41. (11) Emmenegger, L.; Scho ¨nenberger, R.; Sigg, L.; Sulzberger, B. Limnol. Oceanogr. 2001, 46, 49-61. (12) Kwan, W. P.; Voelker, B. M. Environ. Sci. Technol. 2002, 36, 1467-1476. (13) McKnight, D. M.; Kimball, B. A.; Runkel, R. L. Hydrol. Process 2001, 15, 979-1992. (14) Rose, A. L.; Waite, T. D. Environ. Sci. Technol. 2003, 37, 4877-4886. (15) Sedlak, D. L.; Voelker, B. M.; Zafiriou, O. C. Mar. Chem. 1995, 50, 93102. (16) Wilson, C. L.; Hinman, N. W.; Sheridan, R. P. Photochem. Photobiol. 2000, 71, 691-699. (17) Seinfeld, J. H.; Pandis, S. N. Atmospheric Chemistry and Physics: From Air Pollution to Climate Change; Wiley-Interscience: New York, 1998. (18) Zuo, Y.; Hoigne´. J. Science 1993, 260, 71-73. (19) Manley, S. L.; Barbero, P. E. Limnol. Oceanogr. 2001, 46, 1392-1399. (20) Palenik, B.; Zafiriou, O. C.; Morel, F. M. M. Limnol. Oceanogr. 1987, 32, 1365-1369.
Analytical Chemistry, Vol. 79, No. 11, June 1, 2007 4169
Figure 1. Mechanism of CL reaction between AE and H2O2. Thermodynamic and kinetic constants for each step of the reaction are provided on the diagram.
ides present in rain and other atmospheric samples22 provide another potential source of positive interference with regard to analyses of H2O2 in these sample types. In contrast to earlier colorimetric and fluorometric methods for the analysis of H2O2 in natural waters,23,24 the AE-CL-based method described in the present study is useful in a wide range of natural water matrixes and is significantly less sensitive to interference from naturally occurring chromophores, fluorophores, and organic peroxides. The mechanism of reaction of AE with H2O2 is shown in Figure 1 .2,25 AE (I) is stable in acidic solutions but will rapidly hydrolyze in base. In the presence of the peroxide anion (HOO-), AE forms an unstable dioxetane compound IV, which decays yielding N-methylacridone (V) and light at a wavelength of 470 nm. We have quantified the kinetic details of this mechanism and coupled this information with the flow dynamics of a flow injection analysis (FIA) system to describe the analytical response of AE for the analysis of H2O2 in a range of aqueous samples. The analytical model was specifically tested against the empirical optimization of H2O2 analysis in three different types of natural water: seawater, freshwater, and rain. EXPERIMENTAL SECTION AE-CL Method. A flow injection analysis instrument with chemiluminescence detection (Waterville Analytical, Waterville, ME) was used for both the mechanistic investigation of the reaction of AE with H2O2 and the determination of H2O2 in seawater (Figure 2). The CL detector was a Hamamatsu HC-135 photon counting PMT (Hamamatsu Corp., Bridgewater NJ) operated at the manufacturer’s recommended voltage (900 V) for optimal signal/noise ratio for nanomolar concentrations of hydrogen peroxide typical in seawater and terrestrial surface water. However, for analysis of rain samples, where micromolar concentrations of H2O2 were observed, it was necessary to reduce the voltage setting to 700 V to avoid PMT saturation. (21) Kowalczuk, P.; Cooper, W. J.; Whitehead, R. F.; Durako, M. J.; Sheldon, W. Aquat. Sci. 2003, 65, 381-398. (22) Finlayson-Pitts, C.; Pitts, J. N. Atmospheric chemistry: fundamentals and experimental techniques; Wiley: New York, 1986. (23) Miller, W. L.; Kester, D. R. Anal. Chem. 1988, 60, 2711-2715. (24) Kieber, R. J.; Heltz, G. R. Anal. Chem. 1986, 58, 2312 - 2315. (25) Kaltenbach, M. S.; Arnold, M. A. Mikrochim. Acta 1992, 108, 205-219.
4170 Analytical Chemistry, Vol. 79, No. 11, June 1, 2007
Figure 2. Flow injection analysis manifold for chemiluminescence analysis. (1) Sample syringe (500 µL), (2) peristaltic pump (flow rates in mL min-1), (3) acid wash loop (500 µL), (4) 10-port, 2-position injection valve, (5) sample loop (500 µL), and (6) flow cell. The syringe1 may be replaced with a line to the pump for continuous sample introduction. For kinetic studies, the carrier and sample lines were both placed in the sample to produce a continuous signal independent of valve position. The details of reagent composition, manifold operations, and flow rates are provided in Table 1.
Solutions were delivered with a Rainin Rabbit Plus peristaltic pump using Tygon pump tubing. All other tubing was 0.76 mm i.d. × 1.59 mm o.d. Teflon with polypropylene fittings. The compositions of the reagents used for kinetic studies and analyses of natural waters are included in Table 1. For mechanistic studies, the AE reagent (1 µM) was prepared by adding 10-methyl-9-(p-formylphenyl)acridinium carboxylate trifluoromethanesulfonate (AE) to MilliQ water.1,2 For analysis of natural water samples, the AE reagent was buffered to pH 3 with 1 mM phosphate buffer to inhibit base hydrolysis of AE.25 The concentration of the AE reagent solution was 1 µM for analysis of seawater and freshwater samples and 10 µM for analysis of rain samples. The MilliQ water used for preparation of reagent and buffer solutions was treated with 3 mg L-1 catalase (Sigma) for at least 30 min to remove trace amounts of H2O2. Catalase added to the reagent and buffer will react with some of the H2O2 in the sample. However, loss of sample H2O2 is insignificant because the half-life of H2O2 in 3 mg L-1 catalase is 15 min and the FIA reaction time is 9 resulting in a first-order decay of the CL signal, as shown in the inset in Figure 3. The decay rate of the CL signal was first order over two or more reaction half-lives. A nonlinear fit to a simple exponential decay was used to determine the AE hydrolysis rate constant as a function of pH (Figure 3). A stop-flow technique was used to determine the decay of the AE-OOH intermediate II. The pH of the buffer (reagent D, Table 1) was adjusted over the range from 9.0 to 11.5. The flow system was configured with the carrier line in the sample to produce a constant CL signal. After establishing the steady-state signal, the instrument pump was stopped while data collection continued (Figure 4, inset). Formation of the intermediate II is fast (1-10 s) relative to the time span of these experiments, which allowed the decay of II to be observed directly as the CL signal over time. The decay of the CL signal was first order over two or more halflives, and a nonlinear fit to a simple exponential decay was used to determine the pH-dependent decay rate of II (Figure 4). The standard FIA setup (Figure 2) was used to measure the CL signal (peak height or area) as a function of H2O2 and AE concentrations. In one set of experiments, the H2O2 concentration was varied from 20 nM to 14 µM using a constant AE concentration of 1 µM. In another set of experiments, the AE concentration was varied from 100 nM to 5 µM using a constant H2O2 concentration of 1 µM. In both sets of experiments, the signal was linearly dependent on concentration (data not shown).
Figure 4. AE-OOH first-order decay rate as a function of reaction pH (open symbols). Inset shows typical first-order CL decay of AEOOH complex at pH 11.3. Solid lines are fits to both data sets. Solid symbols in main graph are the steady-state signals at the start of decay experiments for 1 µM H2O2 and 1 µM AE.
A model of the CL signal was created using Stella Software (ISEE Systems V8.11) running on an Apple microcomputer. The model incorporated the sample loop volume, detector volume, and the flow rates of all reagents and the carrier/sample. To model peak shape, H2O2 in the sample loop was treated as 25 sequential parcels each containing 1/25th of the sample volume. Transit of each parcel in the model was computed from the flow rates of the carrier and reagents and the volume of the detector. Chemical reactions in each parcel were modeled according to the mechanism shown in Figure 1. The formation rate of V in each parcel was assumed to be proportional to the CL flux of that parcel. The detector response was modeled by summing the CL flux from each parcel that was present within the detector volume. Model simulations were run with a time step of 10 ms using the RungeKutta method for numerical integration. Additional details of the model are provided in the Supporting Information. Analysis of Natural Water Samples. Three types of natural water samples were utilized for the present study: (1) Open ocean seawater collected from the surface and from several different depths using 10-L Go-Flo bottles (General Oceanics) deployed on a chemically inert sampling rosette aboard the R/V Tangaroa at 46°38′ S, 178°30′ E in the Pacific Ocean east of New Zealand; (2) terrestrial surface freshwater collected from the Water of Leith, a small surfacewater stream in southern New Zealand (45°52′ S, 170°31′ E), by submerging an acid-washed glass bottle (Schott) and rinsing the bottle three times with the streamwater before collection of the final sample; (3) rainwater collected from the roof of the chemistry building at The University of Otago (45°52′ Analytical Chemistry, Vol. 79, No. 11, June 1, 2007
4171
Table 2. Precision Estimates and Limit of Detection of the FIA-AE-CL Method for Hydrogen Peroxide
sample Milli-Q water (non-catalase treated) (n ) 8) open ocean seawatera 25 Feb 05 (0745 h) 46°38′ S 178° 32′ E (n ) 8) Water of Leithb 19 Nov 03 (1345 h) (n ) 8) rainwaterc 28 October 03 (1020 h) (n ) 8)
mean [H2O2] (nM)
rel std dev (%)
32.3
1.49
1.45
0.779
15.0
0.352
283
0.57
1.61
6,340
0.92
58.1
limit of detection (nM)
a Collected from 500 m depth. b Surface sample from freshwater stream, University of Otago, Dunedin, New Zealand. c Collected on the roof of the Chemistry building, University of Otago.
S, 170°31′ E), using an automated wet-dry precipitation collector (Aerochem Metrics). In every case, the samples were analyzed within 10 min of collection without any filtration or pretreatment steps. Calibration of the instrument response was established separately for each sample type by adding known amounts of H2O2 to the sample water and then analyzing the amended samples. This procedure of standard additions was repeated, utilizing at least five different H2O2 concentrations for each sample type. Concentrations of H2O2 in individual samples were measured repeatedly (n g 8) in order to establish relative standard deviation values and limits of detection for each sample type (Table 2). RESULTS AND DISCUSSION OF MECHANISTIC STUDIES Hydrolysis. Figure 3 shows the results of AE hydrolysis experiments. The hydrolysis rate constants are low in solutions with pH 10 M-1 s-1
rate A ≈ kA[AE] ) kOH
photon flux ) rate B × QYIV )
)
kBRHOOkB[HOO-] ) [AE][HOOH] (7) kA kA
[II] ≈ 5 × 105[AE][HOOH]
(at pH 11.3)
(8)
The formation of II is proportional to both AE and H2O2 concentrations. Furthermore, the concentration of the CL intermediate [II] is much lower than either [H2O2] or [AE] over the range of reagent concentrations used in this work. These results are in agreement with the observed linearity of the CL signal as a function of both reactant concentrations. The lower limit for kB was established by the observed photon flux for the CL reaction. Using known work functions for the PMT, flow cell geometry, flow cell residence times, and reagent concentrations, we calculated a total CL yield of 0.002 detected photons/molecule of H2O2 in a sample. This is not a true quantum yield, but rather the overall efficiency of converting hydrogen peroxide to detected photons. Values of kB can be used to constrain production rates of II and, therefore, IV. The concentration of IV is not measured directly but is calculated from the flux of photons that result from its decay. Under steady-state conditions, the photon flux can be written as
(10)
Values of kB will range from a lower limit of 10 M-1 s-1 if IV has a QY of 1.0 to the upper limit of 3 × 104 M-1 s-1 requiring IV to have a QY of 3 × 10-4. For purposes of modeling the CL reaction, any combination of kB and QY may be used as long as the inequality defined in eq 10 is maintained. An intermediate value of kB of 5 × 103 M-1 s-1 has been selected for all of the present models. Models of the CL Detector Response. The reactions and rates depicted in Figure 1 were incorporated into a numerical model of detector response. The calculated concentrations of H2O2, AE, II, and photon flux for a single parcel of solution are shown in Figure 5. Note that the H2O2 concentration is essentially constant while the AE concentration decreased rapidly due to base hydrolysis. Formation of II begins immediately upon mixing of reagents and reaches a maximum concentration after 11 s at pH 11.3. The time evolution and magnitude of II will depend on the rate of base hydrolysis of AE, RHOO-, and conversion of II to IV, all of which are pH-dependent variables. The actual photon flux from the flow cell is influenced by the rate of the CL reaction, the time evolution of H2O2 entering the flow cell, solution flow rates, and flow cell volume. As described in the Experimental Section, these processes were modeled by dividing the plug of H2O2 arriving at the flow cell into 25 unique parcels and modeling each as a function of time. Figure 6 shows the simulated detector response for different FIA flow rates and flow cell volumes. For a small flow cell volume of 0.5 mL, increasing the reagent flow rate from 0.9 to 4.2 mL/min decreased the peak widths and heights, as shown in peaks A-C (Figure 6). Decreasing peak widths are due to shorter residence time of the sample in the flow cell at higher flow rates. Peak heights also decrease as flow rate increases because maximum concentrations of II in the flow cell are lower at higher flow rates. This can be overcome by increasing the flow cell volume as shown for peaks C-E (Figure 6), which represent flow cell volumes of 0.5, 1.5, and 4.5 mL, respectively, with a fixed reagent flow rate of 4.2 mL/ min. Interestingly, flow rate has less of an influence on peak height when the flow cell volume is large due to the longer residence times of II in the flow cell afforded by the greater flow cell volume. Peaks E and F represent flow rates of 4.2 and 0.9 mL/min, respectively, with a fixed flow cell volume of 4.5 mL. Because the H2O2-AE CL reaction is fairly slow, lasting for greater than 10 s, the integral of the detector response is larger at low flow rates (Figure 7). At low flow rates, the reaction simply spends more time in the flow cell. The pH dependence of the CL integral shown in Figure 7 is due to the opposing effects of AE hydrolysis and formation of IV. Independent of flow rate, the optimal pH for this analytical system is 11.3. Analytical Chemistry, Vol. 79, No. 11, June 1, 2007
4173
Figure 5. Concentrations of reactants and intermediates during CL reaction at pH 11.5, flow cell volume 1.5 mL, flow rate 2.4 mL/min. (A) H2O2 (M) scaled by 10×, (B) AE I (M), (C) AE-OOH intermediate, II (M), (D) photon flux due to decay of AE-OOH (M s-1) scaled by 4×. Maximum AE-OOH intermediate concentration and CL flux was at 13 s. The sample reached the detector at 4 s. The shaded box indicates the time that the reaction mixture was in the detector flow cell.
Figure 6. Computed CL peak shape as a function of reagent flow rate and flow cell volume: (A-C) 0.5 mL flow cell volume at flows of 0.9, 2.0, and 4.2 mL min-1, respectively. (C-E) illustrate the effect of increasing flow cell volume (C, 0.5; D, 1.5; E, 4.5 mL) F has a flow cell volume of 4.5 mL and a flow rate of 0.9 mL/min. All simulations were run at pH 11.3.
For many chemiluminescence systems, it is desirable to maximize peak height as opposed to peak integral. Increasing peak height improves the signal/noise ratio of the detection system. Sample throughput is also improved with a narrow peak shape. Figure 8 shows the simulated influence of reagent flow rate on peak height for three different detector volumes in comparison with the observed behavior for the experimental detector volume of 2.0 mL. At low flow rates, the measurable photon flux, which is proportional to peak height, was limited by the flux of H2O2 into the flow cell. Essentially, the CL reaction rate was faster than the mass-transfer rate of H2O2 into the flow cell. At higher flow rates, the rate of the CL reaction limits the measurable photon flux, and some fraction of H2O2 in the sample leaves the flow cell without reacting. In the present simulated system, peak heights are always low in the 0.5 mL flow cell because the residence time of the solutions is short at high flow rates and it is limited by mass flow of H2O2 into the cell at low flow rates. Using the numerical model of detector response for the H2O2AE system, it is possible to optimize system performance to the specific requirements of the analyst. The greatest photon produc4174 Analytical Chemistry, Vol. 79, No. 11, June 1, 2007
Figure 7. Computed CL peak integrals as a function of reaction pH and flow rate (graph labels, mL/min). Slower flow rates allow more luminescence of the AE-OOH complex before exiting the flow cell. At higher pH the signal is lower due to ester hydrolysis.
Figure 8. Calculated and measured peak heights as a function of reagent flow rate and flow cell volume (inset, mL). Simulations run with a reaction of pH 11.3. At small flow cell volumes the CL reaction increasingly occurs in waste as flow rate increases. As the flow cell volume increases (relative to sample loop size) the system becomes less sensitive to flow rate, under these conditions ester hydrolysis is fast relative to the residence time of the CL mixture in the flow cell.
tion and instantaneous CL flux occurs at a reaction pH of 11.3. A flow cell volume of greater than 1 mL provides sufficient detector residence time for the CL reaction to reach full intensity. Alternatively, the sample and reagents could be mixed 5 s before arriving in the flow cell, allowing a smaller flow cell volume or very fast flow rates to be used. There is no optimal AE reagent concentration for this system. Increasing the AE concentration will increase the CL flux proportionate to the reagent concentration. This may result in better detection limits depending on the sensitivity of the PMT but also leads to increased reagent consumption. RESULTS AND DISCUSSION OF NATURAL WATERS ANALYSIS Optimization of Detector Response to H2O2. Using the results of the mechanistic studies and output from the STELLA
model, the method was optimized for determination of H2O2 concentrations in seawater, freshwater, and rain. The FIA system for these analyses used a relatively high volume flow cell (2 mL), resulting in long residence times of the luminescent solution in the flow cell. This ensured that, at moderate flow rates, the magnitude of the luminescent response was limited by mass transfer of H2O2 to the flow cell and not by the rate of the chemiluminescent reaction. Using the high-volume flow cell and the optimized reaction pH of 11.3, the influences of flow rate on peak integral values and peak heights were examined. In agreement with the modeled response, higher flow rates resulted in narrower response peaks and lower peak integral values due to shorter residence time of the luminescent solution in the flow cell (Figure 7). But the effect of flow rate on the height of the response peaks was negligible over the range of flow rates from 0.7 to 2.3 mL/min because the volume of the flow cell ensured that maximum concentrations of the reactive intermediate II were attained at all flow rates in the range (Figure 6). Flow rates greater than 2.3 mL/min resulted in decreased peak heights. The kinetic model confirmed that the height of the response peaks was limited by the rate of formation of the reactive intermediate II at flow rates greater that 2.3 mL/min. At flow rates lower than 0.7 mL/ min, the CL response was limited by mass transfer of the AE probe I to the flow cell. For analysis of natural water samples, peak integral values were more closely correlated to concentrations of H2O2 in a series of matrix standards than were peak heights. Within the range of flow rates that allowed formation of maximum concentrations of II in the flow cell (0.7-2.3 mL/min), higher flow rates decreased our ability to resolve very low concentrations of H2O2, but resulted in narrower peaks, shorter analysis times, and more linear calibration plots when determining H2O2 concentrations in the range relevant to our natural water samples. However, the lower limit of detection generally improved at lower flow rates. Using optimized PMT settings (see below), the best results for the present studies were achieved using a flow rate of 1.7 mL/min, consistent with the model results shown in Figure 7. The temporal resolution of the data acquisition system, controlled by the “integration time” setting on the PMT, affected the performance of the FIA system. Shorter integration times allowed better temporal resolution but made it difficult to distinguish low levels of luminescence signal from baseline noise. Longer integration times effectively smooth the detector response over time, removing noise from the signal at the expense of temporal resolution. Using the optimized flow rate of 1.7 mL/ min, we found that a PMT integration time of 400 ms provided the best balance between temporal resolution and signal-to-noise ratio. The STELLA model confirmed that 400 ms was less than 5% of the FIA peak width, providing adequate temporal resolution of the luminescence in the flow cell (Figure 6). The flow cell volume was fixed at 2 mL. The volume of the flow cell was constrained by the size of the PMT photocathode situated directly above the flow cell. The volume of the sample loop was 500 µL, which allowed full development of the CL signal from the smallest practical sample volume. Analysis of Seawater. Using the optimized parameters established using the kinetic model, the concentrations of H2O2 were determined in seawater samples collected from the surface
Figure 9. Depth profiles of H2O2 concentrations in seawater from the subtropical convergence zone. Measured concentrations of H2O2 in seawater collected from 500 m depth were typically below 1 nM. The inset shows variation in solar UVB irradiance and hydrogen peroxide concentration in a terrestrial freshwater stream over the course of one diel cycle; diamonds represent the concentration of H2O2; circles represent solar UVB irradiance.
and from several different depths at a site in the Pacific Ocean east of New Zealand. The study site was located within the subtropical convergence zone, where mixing of subtropical and subantarctic water masses results in concentrated biogeochemical activity. Because the main pathways for formation of H2O2 in seawater involve sunlight-driven photochemical reactions, concentrations of the analyte are higher in samples collected from the surface compared to samples collected from greater depths. Our measurements showed maximum concentrations of H2O2 at ∼10 m depth and decreases in the concentration of H2O2 at greater depths (Figure 9). The shallow water depth profiles obtained are similar to those reported for similar oceanographic regions.2,5 Concentrations of H2O2 in the surface seawater ranged from 8.5 to 138.5 nM in the subtropical convergence zone. For seawater samples, the combination of seawater and alkaline buffer caused precipitation of magnesium hydroxide in the flow cell and waste lines. The accumulation of this precipitate in the flow cell caused a decrease in photon detection and can eventually block the flow cell. This was avoided by adding an acid wash loop to the injection valve so that the flow cell was flushed with 0.01 M HCl after each seawater sample. The wash loop was integrated into the system via a 10-port injection valve,2 which conveniently loaded the wash loop during the sample injection interval, enabling the flow cell to be flushed with HCl during the sample load interval. Details regarding the reagents used for measurement of H2O2 concentrations in seawater are provided in Table 1. Analysis of Terrestrial Surface Water. Concentrations of H2O2 were determined in freshwater samples collected from the Water of Leith, a terrestrial surface water stream in Dunedin, New Zealand. Due to the lower pH and buffering capacity of the freshwater samples, the optimum flow cell pH for the chemiluminescent reaction (pH 11.3) was attained using a 0.01 M solution of sodium bicarbonate as the alkaline buffer reagent solution, while a 0.1 M bicarbonate solution was necessary to attain the optimum flow cell pH when analyzing seawater samples (Table 1). Measurements of H2O2 in the streamwater collected over the course Analytical Chemistry, Vol. 79, No. 11, June 1, 2007
4175
Table 3. Sensitivity to Organic Peroxides added peroxide (nM) [H2O2]
[CH3OOH]
0 255 765 1020 1785 2550 15 15 15
0 235 705 940 1645 2350
a
[T-B]a
73 730 7300
measured peroxide (nM) 0 206 690 1196 1840 2533 15.0 16.3 22.7
errorb 6 10 35 60 93 127 0.2 0.2 1.0
tert-Butyl peroxide. b 95% confidence interval.
of 24-h periods showed distinct diel patterns in H2O2 concentrations. The highest levels of H2O2 were observed shortly after solar noon, and the lowest levels were observed at night (Figure 9). These observations are consistent with previously reported photochemical mechanisms of formation for H2O2 in natural waters.5 Analysis of Rain. H2O2 and short-chain organic peroxides are ubiquitous gas-phase species in the troposphere and are formed through gas-phase photochemical reactions.31 H2O2 is readily partitioned into atmospheric aqueous phases due to its large Henry’s law constant.32 The high solubility coupled to the photochemically driven formation pathways results in concentrations of H2O2 in rainwater that are frequently in the micromolar range. In order to maintain the fundamental relationship between the concentration of H2O2 in the sample and the magnitude of the detected luminescent response, it was necessary to increase the concentration of the AE reagent solution from 1 µM for analysis of freshwater and seawater samples to 10 µM for analysis of rain (Table 1). This increase in the concentration of the AE reagent ensured that the concentration of H2O2 in the sample was the limiting factor for formation of II and assured linearity of the relationship between [H2O2]sample and the peak area of the detected luminescent response at micromolar concentrations of H2O2 (Table 2). The measured concentrations of H2O2 in the rain samples ranged from less than 800 nM to greater than 6.3 µM (Table 2). Limit of Detection. For each type of sample, the lower limit of detection was determined as three times the standard deviation of a series of replicate analyses (Table 2). Of our natural water samples, open ocean seawater collected from a depth of 500 m had the lowest concentration of H2O2 ([H2O2]mean ) 779 pM). A series of eight analyses of [H2O2] in the seawater collected from 500 m yielded results with a relative standard deviation of 15.0% and a corresponding limit of detection of 352 pM. Limits of (31) Lee, M.; Heikes, B. G.; O’Sullivan, D. W. Atmos. Environ. 2000, 34, 34753494. (32) O’ Sullivan, D. W.; Lee, M.; Noone, B. C.; Heikes, B. G. J. Phys. Chem. 1996, 100, 3241-3247.
4176 Analytical Chemistry, Vol. 79, No. 11, June 1, 2007
detection for other sample types are shown in Table 2. The constraining factors that limited detection levels in this study were background noise detected by the photomultiplier tube and heterogeneity of the carrier solution, which led to small baseline fluctuations. Specificity. Unlike methods involving colorimetric or fluorometric determinations,23 the response of this chemiluminescence method is affected to a much lesser extent by the range of natural organic material typical in streams, seawater, and rain. Additionally, we tested for interference from organoperoxides by adding known concentrations of tert-butyl peroxide and methyl hydroperoxide to samples. As shown in Table 3 and the Supporting Information, the presence of organoperoxides did not present a significant interference to this method. CONCLUSIONS Chemiluminescence-based measurements using the probe compound 10-methyl-9-(p-formylphenyl)acridinium carboxylate trifluoromethanesulfonate afford the sensitivity and precision necessary for nanomolar-level determination of hydrogen peroxide concentrations in the natural water matrixes tested during this work. However, in order to adapt the AE-CL method for use in a flow injection analysis system, a variety of operational parameters, including reagent flow rates, detector settings, flow cell volume, and pH of the reagents, must be optimized. Using the mechanism of the reaction of H2O2 with AE and a modified flow injection system, the rate constants for the various steps in the luminescence reaction have been evaluated and successfully employed in a kinetic model to optimize the operational parameters for this method of analysis of H2O2 in a range of natural waters using flow injection analysis. ACKNOWLEDGMENT W.J.C. gratefully acknowledges a research reassignment funded by University of North Carolina. Wilmington (JanuaryMay 2003) at the University of Otago and the award of a William Evans Visiting Fellowship from the University of Otago. S.A.R. thanks the University of Otago for the award of a Postgraduate Scholarship, and the officers and crew of the R/V Tangaroa. The authors thank Luc Richard for assistance with collection of seawater samples, and two anonymous reviewers for constructive comments. SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This is contribution 10 from the UCI Urban Water Research Center. This material is available free of charge via the Internet at http://pubs.acs.org.
Received for review November 24, 2006. Accepted March 19, 2007. AC062228W
