HPLC Determination of Cyanuric Acid in Swimming Pool Waters Using

The two methods allowed fast separation and detection of the stabilizer in 4 (phenyl) and ... Neither one of the two methods required the use of sampl...
0 downloads 0 Views 78KB Size
Download PDF
Anal. Chem. 2001, 73, 3358-3364

HPLC Determination of Cyanuric Acid in Swimming Pool Waters Using Phenyl and Confirmatory Porous Graphitic Carbon Columns Ricardo Cantu´,†,* Otis Evans,† Fred K. Kawahara,‡ Larry J. Wymer,† and Alfred P. Dufour†

National Exposure Research Laboratory and National Risk Management Research Laboratory, Office of Research and Development, United States Environmental Protection Agency, 26 West Martin Luther King Drive, Cincinnati, Ohio 45268

The chlorinated salts of cyanuric acid have found an important role in recreational swimming pool waters across the United States. Upon application to pool water, they can (1) release disinfectant chlorine or (2) stabilize the free available chlorine by acting as chlorine reservoirs in the form of cyanuric acid, preventing the photolytic destruction of residual chlorine by sunlight. Recommended levels of the cyanuric acid stabilizer are in the 10-100 mg/L concentration range according to the National Swimming Pool Foundation (San Antonio, TX). Two isocratic HPLC methods with UV detection (213 nm) employing phenyl and porous graphitic carbon (PGC) columns and phosphate buffer eluents (pH 6.7 and pH 9.1, respectively) were developed to accurately measure cyanuric acid in swimming pools. The two methods allowed fast separation and detection of the stabilizer in 4 (phenyl) and 8 (PGC) min. Both methods offered practical sensitivities with method detection limits of 0.07 (phenyl) and 0.02 mg/L (PGC). Neither one of the two methods required the use of sample cleanup cartridges. They exhibit chromatograms with excellent baseline stability enabling low-level quantitation. Most important, the PGC column had a useful lifetime of five months and 500 sample analyses/column. Eleven pool water samples were fortified with 4.8-50.0 mg/L stabilizer, and the average recovery was 99.8%. Finally, statistical analysis on the relative precisions of the two methods indicated equivalence at the 0.05 critical level. The sodium salts of dichloroisocyanuric acid (CAS 2893-789) and trichloroisocyanuric acid (CAS 87-90-1) are used as disinfectants, algaecides, fungicides, and bacteriostats.1,2 These commercial salts are used widely in swimming pools due to their capacity to convert into the disinfectant cyanuric acid (CAS 1080-5) in solution. The behavior of cyanuric acid (CA) with * Corresponding author: (phone) (513) 569-7883; (fax) (513) 569-7757; (email) [email protected]. † National Exposure Research Laboratory. ‡ National Risk Management Research Laboratory. (1) Registration Eligibility Document Facts. Chlorinated Isocyanurates; EPA Document No.738-F-92-010, Washington, DC, 1992. (2) Status of Pesticides in Registration, Re-registration, and Special Review (Rainbow Report); Office of Pesticide Programs; USEPA, Washington, DC,. 1998.

3358 Analytical Chemistry, Vol. 73, No. 14, July 15, 2001

chlorine in solution has been thoroughly studied3-8 and is explained by equilibria that occur in solution with the formation of N-chloro species. These N-chloro species reversibly hydrolyze to HOCl to form chlorine reservoirs that are not subject to photodegradation. The National Swimming Pool Foundation (NSPF) recommends levels of the CA stabilizer in the 10-100 mg/L concentration range.9 The addition of chlorinated cyanurates must be monitored closely to avoid the buildup of unhealthy concentrations of CA.9,10 Hence, there is a need for effective methods to accurately measure the stabilizer in swimming pool waters to avoid excessive level where, currently, dilution is used to reduce it.9 Cyanuric acid’s carbon nitrogen bonds can undergo base-catalyzed hydrolysis,11 but making the water sufficiently basic in order to reduce CA levels is not an option because chlorine disinfection is ineffective above pH 8 and the alkalinity of the solution irritates the skin. Existing methodologies for CA analysis have utilized highperformance liquid chromatography (HPLC).12 A recent survey of these methods showed preference for reversed-phase separation in HPLC using silica columns and phosphate buffers that were preferred because of short analysis times.12 Subsequently, we developed a rapid and simplified HPLC method for CA analysis in water using an octadecyl column and phosphate buffer eluents having pHs in the range of 7.2-7.4.12 This led to the straightforward development of an application method for CA analysis in swimming pool waters after testing several silica-based columns for enhanced resolution from interferences. Since there are many compounds that absorb UV light below 220 nm, a single HPLC method with UV detection is not sufficient (3) O’Brien, J. E. Ph.D. Dissertation, Harvard University, 1972. (4) Petritsi, I. Ph.D. Dissertation, Yale University, 1964. (5) Matte D.; Solastiouk, B.; Deglise, A. M. X. J. Can. Chem. 1990, 68 (2), 307-313. (6) Solastiouk, B.; Deglise, A. M. X. J. Can. Chem. 1988, 66 (2), 2188-2193. (7) Matte, D.; Solastiouk, B.; Deglise, A. M. X. J. Can. Chem. 1988, 67, 786791. (8) The Effect of Cyanuric Acid, a Chlorine Stabilizer, on Trihalomethane Formation; EPA Document No. 600/D-84-167; Cincinnati, OH, 1984. (9) Certified Pool-SPA Operator; Kowalsky, L., Ed.; National Swimming Pool Foundation, San Antonio, TX, 1992. (10) Hammond, B. G.; Barbee, S. J.; Inoue, T.; Ishida, N.; Levinskas, G. J.; Stevens, M. W.; Wheeler, A. G.; Cascieri, T. Environ. Health Perspect. 1986, 69, 287-292. (11) Smolin, E. M.; Rapoport, L. In Chemistry of Heterocyclic Compounds; WileyInterscience: New York, 1959; Vol. 13, Chapter 10. (12) Cantu´, R.; Evans, O.; Kawahara, F. K.; Shoemaker, J. A.; Dufour, A. P. Anal. Chem. 2000, 72 (23), 5820-5828. 10.1021/ac001412t Not subject to U.S. Copyright. Publ. 2001 Am. Chem. Soc.

Published on Web 06/08/2001

to fully characterize the CA structure (λmax ) 213 nm). There is always the possibility of coeluting interferences. Although mass spectrometry is an impressive tool for structural determination, it is expensive compared to UV spectroscopy. In the absence of MS detection, EPA methods, such as EPA method 531.1,13 have relied upon corroboration of the results using a different type of separation column for analyte confirmation. We have followed this approach by developing a second application method using a porous graphitic carbon (PGC) column as a confirmatory tool. The present work (1) develops experimental conditions for two HPLC methods for CA measurement using silica and PGC columns and discusses their chromatographic characteristics, with emphasis on the ruggedness of the PGC method, (2) applies both methods for the measurement of CA in swimming pool waters, while statistically evaluating method equivalency using the F-test, and (3) tests them for accuracy by fortifying several different matrices with CA over a broad concentration range. This is the first time that two different methods have been developed to certify each other unequivocally for CA analyte measurement and confirmation in swimming pool waters. EXPERIMENTAL SECTION Reagents and Swimming Pool Water Samples. HPLC-grade methanol was acquired from Fisher Scientific (Fairlawn, NJ). Cyanuric acid (98%) was obtained from Aldrich Chemical (Milwaukee, WI). A stock solution of 240 mg/L CA was prepared by dissolving the solid in deionized water with vigorous stirring and gentle heating. The CA stock and buffer solutions were filtered with a 0.45-µm cellulose acetate filter. The standard solutions were prepared in the 1-100 mg/L concentration range. A pH Field System (Thomas Scientific, Swedesboro, NJ) was used to measure the pH with 0.01-unit resolution. The swimming pool water samples were collected in 125-mL, wide-mouth high-density polyethylene bottles (Nalgene, Rochester, NY) and stored at 5 °C. Two to three milliliters of the samples were filtered using a 25-mm-diameter 0.2-µm cellulose disk equipped with a 0.8-µm prefilter (A. Daigger, Lincolnshire, IL) to remove debris and other insoluble particulates. A volume of 20 µL was introduced into the HPLC and analyzed in triplicate. Initial analyses required removal of anionic interferences using a strong-anion-exchange (SAX) resin. Oasis MAX cartridges, 3 cm3/60 mg containing a polymeric-based quaternary ammonium salt (CH2+NR3) were purchased from Waters (Milford, MA). The cartridges were cleaned with 0.05 M formic acid. The sample was loaded onto the resin and allowed to pass through by gravity filtration, eluting CA quantitatively in a minimum volume of deionized water. The eluate was analyzed by HPLC. HPLC System. A model 200B liquid chromatograph (Bioanalytical Systems (BAS), W. Lafayette, IN) equipped with a Rheodyne (Rhonert Park, CA) model 7125 injector employing a 20-µL sample loop, built-in single-channel UV-visible detector, and software (Chromgraph Report, version 1.2) was used for all HPLC analyses. The octyl column was purchased from Fisher Scientific (Fairlawn, NJ). The octadecyl, hexyl-phenyl, and ethyl-phenyl (PH-3) were acquired from Phenomenex (Torrance, CA). The (13) Graves, R. L. Measurement of N-Methylcarbamoyloximes and N-Methylcarbamates in Water by Direct Aqueous Injection HPLC with Post Colum Derivatization. In Methods for the Determination of Organic Compounds in Drinking Water, Supplement III. EPA/600/R-95/131 USEPA, National Exposure Research Laboratory, Cincinnati, OH 45268; 1995.

cyano and amino columns were purchased from Agilent Technologies (Wilmington, DE). All of the columns were 25 cm × 0.46 cm with a 5-µm particle size. Column pressures when a mobile phase composed of 95:5 (%, v/v) phosphate buffer (pH 7.3)/methanol at 0.8 mL/min and oven temperature at 50 °C was used were 800900 (octyl), 1600-1800 (octadecyl), 1400-1500 (hexyl-phenyl), 1300-1400 (ethyl-phenyl, cyano), and 1200-1300 psi (amino). The phosphate buffer was composed of 0.006 25 M KH2PO4 and 0.0125 M K2HPO4. PGC columns, 10 cm × 0.30 cm and 5-µm particle size, were acquired from Phenomenex and Hypersil (Bellefonte, PA). Statistical Methods. Equivalence between the respective methods was evaluated by applying F-tests for the equality of the variance among the triplicate measurements for each sample analyzed. In addition, the combined within-pool variances of the two methods were calculated as the mean variance over all swimming pools (excluding any pools in which CA concentrations were below the detection limit) and compared with an F-test. Because the variances are not believed to be constant among the various samples, Satterthwaithe’s approximation14 for the appropriate degrees of freedom associated with the F-statistic was employed for evaluation of the combined within-pool variances. The precision of the recoveries of the methods was also evaluated, given that the standard deviations appear to be proportional to sample means for each method.15 Under the assumption that the log variance is constant from sample to sample for each method, the pooled log variance over all samples comprises an estimate of this variance for the respective method. Pooled log variances for the two methods were compared via an F-test. RESULTS AND DISCUSSION Development of Method 1 for Cyanuric Acid Analysis (Phenyl Method). Silica columns with numerous stationary phases (e.g., octadecyl, octyl, phenyl, amino, cyano, etc.) have been preferred in HPLC methodologies because they present high efficiencies (high plate number) and short equilibration times.16 Recently, we developed a method for CA analysis in water using an octadecyl column and a phosphate buffer eluent at pH 7.3.12 Using this method, we established the strong dependency of CA sensitivity with HPLC-UV and UV techniques based upon eluent pH and solution pH, respectively. A compromise was needed between retention and sensitivity to prevent CA keto-enol tautomerism (peak splitting), low sensitivity, and elution at the chromatographic void volume. However, our method, when applied to pool waters, sometimes resulted in a few pool water samples exhibiting coelution of other unidentified anionic interferences with CA. This situation was eliminated by passing the sample through a commercial cartridge containing a SAX resin. However, the use of the cleanup cartridge added expense and time to the overall analysis. In an effort to possibly avoid the cleanup step, we investigated several other silica-based columns for enhanced selectivity and resolution. Figure 1 compares these columns for their ability to resolve the CA peak from the nitrate ion. The nitrate ion, a strong absorber (14) Satterthwaite, F. E. Biometrics Bull. 1946, 2, 110-114. (15) Bartlett, M. S. Biometrics 1947, 3, 39-52. (16) Snyder, L. R.; Kirkland, J. J.; Glajch, J. L. Practical HPLC Method Development, 2nd ed.; John Wiley & Sons: New York, 1997.

Analytical Chemistry, Vol. 73, No. 14, July 15, 2001

3359

Figure 1. Resolution of cyanuric acid from the nitrate ion using different silica columns. Solution of 10 mg/L cyanuric containing 5 mg/L nitrate ion. See Experimental Section (HPLC System) for HPLC column characteristics and conditions.

Figure 2. Chromatograms showing column efficiencies, retention, and sensitivity of cyanuric acid using the phenyl and PGC methods. A total of 200 ng of cyanuric acid was injected on column using both methods.

of UV light, was chosen for predicting interferences because it is frequently encountered in water and can reach concentrations of up to 50 mg/L.17 As shown in Figure 1a, the previous octadecyl column was not capable of completely resolving the CA peak from the nitrate ion peak. The resolution factor (Rs) for this column is 0.9 where Rs ) ∆tr/wav and ∆tr is the difference in retention times, and wav is the average width measured at the baseline of the peak. For the less alkylated octyl column (Figure 1b), Rs increases slightly to 1.2. A more substantial increase in resolution took place when the two phenyl columns with hexyl (Figure 1c) and ethyl (Figure 1d) groups bonded to the silica support were compared.

Surprisingly, the ethyl-phenyl column gave superior resolution with Rs ) 2.1. On the other hand, the columns containing cyano (Figure 1e) and amino groups (Figure 1f) did not produce practical resolution and were eliminated from further consideration. On the basis of these comparisons, the ethyl-phenyl column (PH-3 designation by the manufacturer) was selected for additional study and is referred to as the phenyl column, hereafter. Method 1 was then developed using the phenyl column and a phosphate buffer. The chromatographic parameters were optimized to attain retention of CA at 3 min (Figure 2). The method detection limit (MDL) of CA was determined to be 0.07 mg/L according to the equation MDL ) t(n-1, 1-R)0.99)(S), where t(n-1, 1-R)0.99) is the Student’s value appropriate for a 99% confidence

(17) Johnson, M.; Melbourne, P. Analyst 1996, 121, 1073-1078.

3360 Analytical Chemistry, Vol. 73, No. 14, July 15, 2001

Table 1. HPLC-Phenyl Method for Cyanuric Acid Analysis stationary phase column dimension oven temperature eluent buffer composition buffer pH flow rate pressure UV detector MDL analysis time

silica ethyl phenyl (PH-3)a (5-µm particle size) 25 cm (length) × 0.46 cm (width)b 35 °C 95:5 phosphate buffer/methanol (%, v/v) 0.005 M KH2PO4 + 0.002 M K2HPO4 6.7 1 mL/min. 1600-1900 psi 213 nm 0.07 mg/L 4 min

a Column name as described by manufacturer (Phenomenex). Equipped with Security Guard Column of 4 × 3 mm dimensions (Phenomenex).

b

Table 2. HPLC-PGC Method for Cyanuric Acid Analysis stationary phasea column dimensions oven temperature eluent buffer composition buffer pH flow rate column pressures UV detector MDL analysis time a

Porous graphitic carbon (Hypercarb)a (5-µm particle size) 10 cm (length) × 0.30 cm (width) 35 °C 95:5 phosphate buffer/methanol (%, v/v) 0.050 M K2HPO4 9.1 0.8 mL/min 2500-3300 psi 213 nm 0.02 mg/L 8 min

Column name as described by column supplier (Phenomenex).

level and a standard deviation estimate with n - 1 degrees of freedom and S is the standard deviation.20 The final optimized method parameters are summarized in Table 1. The calibration curve of standard solutions in the 1-100 mg/L range yielded the equation y ) 4695x - 1054 with the coefficient of determination (r2) ) 0.9999. Development of Method 2 for Cyanuric Acid (PGC Method). Method 2 utilized a PGC column and was developed based on the theory for enhancing chromatographic retention of aromatic organic compounds using the hexagonal structure of graphite as reported by Kawahara and Michalakos.18,19 Optimization of parameters with a phosphate buffer resulted in retention of CA near 7 min (Figure 2) and an MDL20 of 0.02 mg/L. This enhanced retention established sufficient chromatographic separation to eliminate void volume interferences, and thus, the use of SAX cartridges was not needed. Table 2 summarizes the final optimized parameters for the method. The calibration curve yielded the equation y ) 10754x - 2833 with r2 ) 0.9999. The steeper slope of this calibration curve is indicative of the higher sensitivity of method 2 using peak areas. (18) Kawahara, F. K.; Michalakos, P. M. Ind. Eng. Chem. Res. 1997, 36, 1584. (19) (a) Kawahara, F. K.; Michalakos, P. M. Preliminary findings were presented at the I & EC Special Symposium, American Chemical Society, held at Atlanta, GA, Emerging Technology in Hazardous Waste Management VI, September 19-21, 1994; Vol. I, p 314. (b) And at the 21st Annual RREL Research Symposium, April 4-6, 1995, at the Westin Hotel, Cincinnati, OH. Abstracts are published in EPA/600/R-95/012, pg 207-210. (20) Glaser, J. A.; Foerst, D. L.; McKee, G. D.; Quave, S. A.; Budde, W. L. Environ. Sci. Technol. 1981, 15 (12), 1426-1435.

The increased sensitivity using the PGC method resulted from using the alkaline phosphate buffer at pH 9.1, which promotes more complete CA ionization. Theoretically, 99% of CA is ionized in solution at this pH to provide an enolate aromatic ion, which is the most sensitive chromophore for UV detection.12 Figure 2 shows the increased CA detection sensitivity as a function of peak area (broad peak), not peak height. The increased peak height is favored when a phenyl column is used (method 1) because of higher column efficiencies that give rise to narrower and sharper peaks, characteristic of many silica stationary phases. When column efficiencies were measured, the phenyl column yielded 10000 ( 20 plates compared to just 1800 ( 400 plates for the PGC column. Nevertheless, due to the lack of coeluting interferences, the PGC column provides adequate peak width and improved sensitivity. It is interesting to note the ruggedness of the PGC column after five months of usage. It was able to function for a total of 500 analyses. Each analysis averaged about 10 min. These analyses included a variety of samples containing CA such as swimming pool waters and biological samples. The samples were introduced into the HPLC using different conditions (e.g., buffer pH, oven temperature, flow rates, etc.) during the development of the PGC method. Subsequent usages of the PGC column became impractical due to peak broadening (plate loss), and peak splitting (not shown) possibly attributed to keto-enol tautomerism. The PGC method has been adapted for the analysis of CA in body fluids for human exposure studies.21 It has proved beneficial as a tool to monitor the long-term stability of CA in complex biological matrixes. It has also been useful for establishing the effectiveness of various stabilizers (e.g., sulfuric acid, perchloric acid, metaphosphoric acid, etc.) for CA preservation after sample collection and to monitor percent recoveries during extraction procedures.21 The effective application of the method and sample cleanup procedures were recently presented.21 Analysis of Cyanuric Acid in Swimming Pools. Cyanuric acid was measured by applying the two described methods to 11 swimming pool water samples. These samples were obtained from separate pools in northern Kentucky and southwestern Ohio. The chromatograms of representative samples are shown in Figure 3 (phenyl method 1) and Figure 4 (PGC method 2). Table 3 contains the measured CA levels. Samples 2, 6, 7, 8, and 10 were within the 30-50 mg/L stabilizer levels recommended by the NSPF.9 Samples 1 (22.2 ( 0.5 mg/L) and 4 (20.7 ( 0.7 mg/L) fell below this range, but they occurred above the lower NSPF limit of 10 mg/L. Sample 3 with 5.0 ( 0.1 had insufficient quantities of CA to effectively disinfect pool water, while samples 5 and 9 were below detectability using either method. Sample 11 with 77.6 ( 0.4 mg/L CA quantities was above the NSPF recommended 30-50 mg/L levels but still below the upper maximum of 100 mg/L. The phenyl method (pH 6.7) offered advantages, such as baseline stability, higher column efficiencies, and short analysis times (Figure 3). The baseline stability combined with the high column efficiencies of the phenyl column resulted in narrow and sharp CA peaks. Thus, it allows sufficient (1) resolution from void (21) Cantu´, R.; Evans, O.; Kawahara, F. K.; Magnuson, M. L.; Shoemaker, J. A.; Wymer, L. J.; Behymer, T. D.; Dufour, A. P. In Book of Abstracts, Part 1, 221st National Meeting of the American Chemical Society, San Diego, CA, April 1-5, 2001; American Chemical Society, Washington, DC, 2001.

Analytical Chemistry, Vol. 73, No. 14, July 15, 2001

3361

Figure 3. Representative chromatograms of the HPLC analyses of cyanuric acid in swimming pool water samples using phenyl method 1. See Table 1 for HPLC for conditions and Table 3 for quantitative results.

Figure 4. Representative chromatograms of the HPLC analyses of cyanuric acid in swimming pool water samples using the PGC method. See Table 2 for HPLC conditions and Table 3 for quantitative results.

volume interferences and (2) reproducible and accurate results. All of the pool samples showed other peaks eluting at or near the chromatographic void volume at 2 min. Late-eluting peaks with retention times greater than 3 min (CA peak) were only observed in the isolated case of sample 8 with a peak at 3.5 min (not shown). However, the area of this late-eluting peak was less than 1% the total area of the CA peak. Analysis time was just 4 min, making this method fast. The PGC method 2 (pH 9.1) provided effective confirmation of the results obtained from the first method and corroborated their accuracy. The PGC method offered baseline stability and important chromatographic advantages such as enhanced retention, sensitivity, and durability. The enhanced baseline stability

along with the greater retention time of CA at 7 min gave superior resolution from other compounds in the aqueous matrix; thus, it possibly avoids the need for sample cleanup cartridges (Figure 4). It also resulted in precise and accurate quantitation of the stabilizer. None of the swimming pool water samples showed any late-eluting peaks. Method 2 was at least 3 times more sensitive as judged from the MDL determinations. Statistical Validation of Precision and Accuracy for both HPLC Methods. From a statistical standpoint, individual F-tests for the relative variances of the HPLC methods revealed no significant differences in precision with regard to the measurement of CA in any single water sample. In addition, the overall precision, represented by the average variance among all pool samples, indicated no significant difference as shown in Table 3 (P all

3362 Analytical Chemistry, Vol. 73, No. 14, July 15, 2001

Table 3. Measurement of Cyanuric Acid in Swimming Pool Water Samples Using HPLC with Statistical Testing for Precision and Accuracy sample (pool no.)

phenyl mean ( s1a (mg/L)

PGC mean ( s2 (mg/L)

Fc ratiob s12/s22

P0c

22.2 ( 0.5 40.2 ( 1.6 5.0 ( 0.1 20.7 ( 0.7 absent 37.9 ( 0.9 47.6 ( 1.8 40.2 ( 1.0 absent 43.6 ( 0.9 77.6 ( 0.4 37.2 ( 1.0

22.5 ( 0.5 39.7 ( 0.5 4.9 ( 0.1 20.0 ( 0.4 absent 36.7 ( 0.7 46.4 ( 0.6 39.0 ( 1.0 absent 42.9 ( 0.7 75.2 ( 1.6 36.2 ( 0.8

1.09 9.22 1.63 2.43

0.4783 0.0978 0.3805 0.2916

1.40 8.45 1.07

0.4162 0.1059 0.4841

1.38 15.32 1.66d

0.4194 0.0613 0.2174

1 2 3 4 5 6 7 8 9 10 11 average

fortificatione (mg/L) actual measured 19.0 17.5 4.8 9.3 24.8 28.2 34.3 31.3 50.0 13.1 21.3

19.1 18.3 5.3 9.4 25.9 27.3 31.8 30.6 47.3 13.0 21.1

% recovery (error) 100.3 104.6 109.2 100.6 104.6 97.1 92.8 97.7 94.5 98.9 98.8 99.8

a s, standard deviation of triplicate analyses. bF is calculated from s and s ; the larger variance is always the numerator. c Probability of the c 1 2 F-value under the null hypothesis that s12 ) s22. Po < 0.05 indicates a significant difference at the 0.05 critical level. dF-Test for equality of pooled variances of method 1 and method 2 based on Satterthwaite’s approximation. Degrees of freedom are 8.7 for the numerator (method 1) and 8.1 for the denominator (method 2). eSamples were fortified above native concentrations of cyanuric acid found in the respective pool samples and determined by the PGC method.

Table 4. Long-Term Stability of Cyanuric Acid in Swimming Pool Water Samples and Reproducibility of the PGC Method day 53a measurements (PGC column usage 100 analyses)b

day 1a measurements (PGC column usage 500 analyses)b

sample (pool no.)

mean ( s1c (mg/L)

% RSDd

mean ( s1c (mg/L)

pool 1 pool 2 pool 3 pool 4 pool 5 pool 6 pool 7 pool 8 pool 9 pool 10 pool 11 average

22.3 ( 0.4 40.3 ( 0.4 5.0 ( 0.1 20.1 ( 0.3

1.9 0.9 2.3 1.3

37.4 ( 0.4 46.4 ( 0.9g 16.7 ( 0.8h

1.0 2.0 4.6

42.0 ( 0.5 74.5i 37.2 ( 1.0

1.3 2.7

22.5 ( 0.5 39.7 ( 0.5 4.9 ( 0.1 20.0 ( 0.4 stabilizer absent 36.7 ( 0.7 46.4 ( 0.6 39.0 ( 1.0 stabilizer absent 42.9 ( 0.7 75.2 ( 1.6 36.2 ( 0.8

% RSD

Fc ratioe s12/s22

P0f

2.2 1.3 2.0 2.0

1.48 2.20 2.62 2.76

0.4028 0.3130 0.2761 0.2660

1.9 1.3 2.6

4.29 2.30 1.58

0.1889 0.2687 0.3874

1.6 2.1 2.2

2.00

0.3335

1.66j

0.2174

a Samples were analyzed on day 1 with an aged PGC column and 53 days later with a newer PGC column. The results confirmed the long-term stability of cyanuric acid in water reported earlier for up to 43 days12 but now in swimming pool water samples with the exception of sample 8. bTotal analyses were calculated after the measurements reported in this table. Each analysis averaged ∼10 min. cs, standard deviation based on a sample size of three. dRSD, relative standard deviation. eFc is calculated from s1 and s2; the larger variance is always the numerator. Degrees of freedom (df) are 2 for numerator and denominator for individual pools, except for pool 7 where the numerator df ) 1. fProbability of the F-value under the null hypothesis that s12 ) s22. Po < 0.05 indicates a significant difference at the 0.05 critical level. gMean and s of two samples. hPool 8 sample is thought to have degraded between time of analysis with used and new columns. iMean of single sample; standard deviation not calculable. jExcludes pool 11. F-Test for equality of pool variances based on Satterthwaite’s approximation. df are 10.3 for the numerator (new column) and 5.1 for the denominator (used column).

0.05). This statistical analysis provides verification and confidence in the comparability of the results of the two methods. The water samples were fortified above native concentrations of CA found in the respective pool samples to further evaluate method performance. The fortified concentrations and the percent recoveries of these spikes are listed in Table 3. The mean recovery was 99.8% with 95% confidence limits of 96.8-102.7%. The samples were reanalyzed again 53 days later by using the PGC method with a different PGC column. Such analyses proved useful for establishing the long-term stability of CA in swimming pools in agreement with the previously measured stability in Analytical Chemistry, Vol. 73, No. 14, July 15, 2001

3363

deionized water.12 In addition, method ruggedness was demonstrated from the reproducibility of the measurements of the chlorine stabilizer after this elapsed time, as shown in Table 4. All samples were stable for this period of time with the exception of sample 8, which exhibited a marked decline in the concentration of the stabilizer, this being attributed to degradation of the sample during storage. Nevertheless, the pooled variances of the raw recoveries among all samples did not differ significantly whether or not this sample is included. Neither were there any significant differences in variance for any individual pool sample. Similarly, pooled log variances between the new and used column analyses were not significantly different (F ) 1.32 with 15 and 16 df, p ) 0.29), being 6.85 × 10-5 and 9.04 × 10-5 for the used and new columns, respectively. Except for the results from pool 8, mean recoveries from the new and used columns also showed no statistically significant differences. Last, it is interesting to mention that the use of octadecyl and PGC columns has been suggested and applied by Hennion and co-workers,22,23 who measured CA in swimming pool waters and in river water spiked with CA using different conditions. However, their HPLC chromatograms showed severe baseline distortion, suggesting the need for sample cleanup prior to attempting any quantitative work. Our methodology, which was developed and applied independently, offers superior baseline stability with horizontal baselines in the chromatograms enabling reproducible and more accurate quantitation of CA in complex swimming pool water matrixes. CONCLUSION Individual HPLC methods using phenyl and PGC columns were used to measure and to confirm CA in swimming pools. The methods rapidly separated and detected the stabilizer in 4 (phenyl) (22) Hennion, M.-C. J. Chromatogr., A 2000, 885, 83. (23) Guenu, S.; Hennion, M.-C. J. Chromatogr., A 1994, 665, 249.

3364

Analytical Chemistry, Vol. 73, No. 14, July 15, 2001

and 8 (PGC) min, respectively. The selectivity of both columns resulted in enhanced resolution from other sample impurities, thus avoiding the use of sample cleanup cartridges. Both methods offered enhanced sensitivities of at least 2 orders of magnitude below the lower recommended levels of CA as established by the NSPF (i.e., an MDL of 0.02 mg/L vs the 10 mg/L set as the lower limit by the NSPF). The methods provided chromatograms with excellent baseline stability enabling easy quantitation. In addition, the PGC column showed good ruggedness as it exhibited high tolerances against hydrolysis or dissolution of the stationary phase during a period of 5 months. Finally, statistical analysis of the relative precision of the two methods indicated equivalence at the 0.05 critical level. Fortified pool water samples were recovered at 99.8%, further increasing confidence in method performance. ACKNOWLEDGMENT The authors are grateful to Billy Potter, Jody A. Shoemaker, and Jean W. Munch of the National Exposure Research Laboratory for valuable discussions on EPA methodology. Also thanks to Matthew L. Magnuson from the National Risk Management Research Laboratory for HPLC method validation. The Chief of the Chemical Exposure Research Branch, Thomas D. Behymer, is deeply appreciated for his technical help. The manuscript was presented at the 220th National Meeting of the American Chemical Society, Division of Analytical Chemistry, held at the Washington Convention Center, Washington, DC, August 19-23, 2000. This paper has been reviewed in accordance with the U.S. Environmental Protection Agency’s peer and administrative review policies and approved for publication. The mention of trade names or commercial products does not constitute endorsement or recommendation by the U.S. Environmental Protection Agency. Received for review December 4, 2000. Accepted May 3, 2001. AC001412T