MS

Anal. Chem. 2004, 76, 469-473

Analysis of Perchlorate in Water and Soil by Electrospray LC/MS/MS Paul Winkler,*,† Mark Minteer,‡ and Janice Willey§

Time Solutions Corporation, 511 Creekside Court, Golden, Colorado 80403, Analytical Quality Associates, Albuquerque, New Mexico 87123, and General Engineering Laboratories, 2040 Savage Road, Charleston, South Carolina 29417

A method has been developed for the determination of perchlorate in water and soil matrixes using electrospray liquid chromatography/mass spectrometry/mass spectrometry. Perchlorate is quantitated by monitoring the ion signal from mass 83, which is formed by a loss of an oxygen atom from the perchlorate molecular ion. The method was developed to be effective and economical in production laboratory analysis of perchlorate in environmental water and soil samples. Data were gathered to define method sensitivity, performance, selectivity, and robustness. Analyte stability, method susceptibility to interferences, and the reliability of the chlorine isotope ratio as an identification tool were examined. The aqueous method detection limit (MDL) is 0.05 µg/L and was determined using an actual groundwater matrix. The soil MDL is 0.5 µg/kg and was determined using Ottawa sand. The stability study was performed by spiking water samples at 0.25, 10, and 20 µg/L and analyzing them 50 days later. Acceptable recoveries were obtained for all samples. The relative standard deviation (RSD) for the replicate analyses in the stability study indicates that the method is capable of RSD values less than 5% in a relatively clean groundwater matrix. The ionization suppression study was performed by spiking water samples containing 1000 mg/L carbonate, chloride, and sulfate with 0.05 and 0.5 µg/L perchlorate and then measuring the recovery of the spike. The results indicate that the procedure does not have significant suppression effects at the high salt levels tested. Calibration, quality control sample, field sample, and suppression study data were combined to examine isotope ratio reliability. The results of that work show that chlorine isotope ratios can be used to define statistical process control limits for use as an additional analyte identification tool. The presence of perchlorate in the environment is of concern due to possible adverse health effects that can arise at relatively low concentrations.1 The analysis of perchlorate is usually performed by ion chromatography with conductivity detection. * Corresponding author. E-mail: [email protected]. † Time Solutions Corp. ‡ Analytical Quality Associates. § General Engineering Laboratories. (1) U.S.E.P.A. Region 9 Perchlorate Update, June 1999. 10.1021/ac034618d CCC: $27.50 Published on Web 12/03/2003

© 2004 American Chemical Society

EPA method 314.1 is the current approved method for analysis of perchlorate. While this method has been used successfully in the past, a method with improved sensitivity and specificity is required. The EPA is considering a regulatory limit for perchlorate of 4 µg/L, but the required limit may be lower if environmental and health risks warrant.2 The detection limit for perchlorate analysis by IC has been reported to be 0.15 µg/L, but this was determined in clean aqueous matrixes.3 In our experience, a more practical MDL is 4 µg/L, especially when attempting to measure perchlorate in diverse groundwater and surface water matrixes. Because detection by conductivity is not specific to perchlorate but responds to any species with sufficient specific conductivity, the IC method does not provide absolute evidence for the presence of perchlorate. The identification of perchlorate depends solely on observing a peak in the sample chromatogram at the correct retention time. In complex matrixes, it is possible to have a large number of peaks observed in the chromatogram, which increases the likelihood of a false positive result. To improve the specificity of detection, several workers have used electrospray liquid chromatography/mass spectrometry (LC/MS/MS) for the determination of perchlorate in groundwater. Recently a comprehensive review of perchlorate methods, including LC/MS/MS techniques, has been published.4 Magnuson5 and co-workers developed a method that extracts perchlorate from the matrix prior to analysis to avoid suppression. The MDL was calculated to be 0.1 µg/L. While the method did remove perchlorate from the matrix, the procedure is laborious and would be difficult to perform on a large production scale. Koester6 and co-workers developed a method for the analysis of perchlorate directly from water with a calculated MDL of 0.5 µg/L, but this method suffered from matrix suppression, requiring that the method of standard additions be used for quantitation. We have incorporated the use of Dionex On Guard columns for sample pretreatment and a Dionex AG16 anion exchange column for separation. These changes have resulted in a procedure that allows for the routine analysis of a large number of samples. We have collected data for the performance of this (2) (3) (4) (5)

EPA review meeting, Sacramento, CA, February 2001. Dionex Application Note No. 134. Urbansky, E. T. Crit. Rev. Anal. Chem. 2000, 30, 311-343. Magnuson, M. L.; Urbansky, E. T.; Kelty, C. A. Anal. Chem. 2000, 72, 25-59. (6) Koester, C. J., Beller, H. R., Halden, R. U. Environ. Sci. Technol. 2000, 34, 1862-1864.

Analytical Chemistry, Vol. 76, No. 2, January 15, 2004 469

method and have used them to establish quality control limits for the implementation of this method in commercial laboratories. The method has also been applied to the analysis of perchlorate in soil, which is the first reported use of electrospray ionization (ESI)/LC/MS/MS for this matrix. EXPERIMENTAL SECTION Sample Preparation. Aqueous samples were prepared by eluting a 10-mL sample aliquot through Dionex (Dionex, Sunnyvale, CA) silver, barium, and hydrogen On Guard columns. This is referred to as the “standard procedure” in the text below. Variations on this procedure were examined in the context of the ionization suppression study. Soil samples were prepared by leaching 2 g of sample with 20 mL of deionized water. The solution was then prepared for analysis following the aqueous procedure. Sample Analysis. The separation was performed using a 2795 HPLC (Waters, Milford, MA) with a Dionex AG16 anion exchange column. A 95/5 water/50 mM ammonium hydroxide isocratic eluant was run at a flow rate of 0.7 mL/min. The injection volume was 50 µL. Under these conditions, the retention time of perchlorate was ∼4.5 min. A Quattro micro API (Micromass, Manchester, U.K.) triple quadrupole tandem mass spectrometer was used for the detection of perchlorate using Masslynx version 3.5 software. The following conditions were found to provide the optimum signal: ion source temperature 85 °C, desolvation temperature 400 °C, cone gas 100 L/h, desolvation gas 500 L/h, cone voltage 45 V, collision energy 25 eV, collision cell pressure 4 × 10-4 kPa, and capillary voltage 0.2 kV. The instrument was run in the negative ion mode, and each quadrupole was set to unit mass resolution. Two multiple reaction monitoring (MRM) transitions were monitored: m/z 9983 (35Cl) and 101-85 (37Cl). Quantitation was accomplished using an external standard method. Instrument calibration was performed by analyzing standards at 0.05, 0.1, 0.25, 0.5, and 1.0 µg/L. To demonstrate that the instrument was properly calibrated throughout the analysis, a continuing calibration verification standard (CCV) was analyzed every 10 samples. A blank sample was analyzed after every CCV to demonstrate that there was no carryover from the standard. Additional batch quality control included method blanks, matrix spikes, replicate samples, and laboratory control samples. RESULTS AND DISCUSSION The proposed LC/MS/MS analysis was performed by monitoring two multiple reactions. The transition is from the loss of oxygen from the molecular ion of perchlorate, ClO4- f ClO3-. The transition for the 35Cl isotope is m/z 99 f 83 and 101 f 85 for the 37Cl isotope. The 37Cl transition was monitored to provide additional confirmation for the presence of perchlorate. Data were collected to determine soil and water MDLs, levels of suppression from anions common to groundwater, limits of acceptance for isotope ratio results, storage stability, and acceptance criteria for calibrations and quality control samples. The MDLs for soil and water were calculated using the procedure specified by the Environmental Protection Agency (EPA).8 The EPA procedure for determining the detection limit of a method recommends the analysis of seven samples that have 470 Analytical Chemistry, Vol. 76, No. 2, January 15, 2004

Table 1. Water and Soil Method Detection Limit Data matrix

n

spike concn

average recovery (%)

% RSD

calcd MDL

MDL

soil water

7 7

0.5 µg/kg 0.05 µg/L

121 118

6.4 10.0

0.12 0.02

0.5 µg/kg 0.05 µg/L

been fortified at a concentration near the expected detection limit. The MDL is then calculated by multiplying the standard deviation of the results by the appropriate t statistic. The water and soil MDLs were both determined by spiking seven samples; water at 0.05 µg/L and soil at an equivalent concentration of 0.5 µg/kg. The results are shown in Table 1. The calculated MDL in the table is the number that resulted from the calculation of the standard deviation from the seven results. The MDL column is our verified MDL value. Occasionally it is not possible to actually detect an analyte when it has been spiked at the calculated MDL. We have taken the approach that an MDL should be verified by spiking at or near the calculated MDL to demonstrate that it is actually possible to detect the analyte at that concentration. A chromatogram of a 0.05 µg/L calibration standard and an MDL sample spiked at 0.05 µg/L is shown in Figure 1. The data demonstrate that it is possible to detect a sample that has been spiked at 0.05 µg/L. The data also indicate that while the signal for the m/z 99 to 83 transition is clearly visible, the signal-to-noise ratio for the m/z 101 to 85 transition is ∼3:1. Because the m/z 101 to 85 transition must be accurately measured to provide ratio data for analyte confirmation, the achievable MDL for this method is 0.05 µg/L. The recoveries for both soil and water were slightly high but were within 25% of the expected value. The RSDs for both the soil and water MDL measurements were less than 15%, indicating excellent precision for the method at low concentrations. When electrospray ionization is used for quantitative analyses, the effect of ionization suppression must be considered. The presence of common anions in the samples can have serious suppression effects on the perchlorate signal and result in the observed values being biased low compared to the true value. It is important to minimize the effect of suppression by separating the anions from perchlorate chromatographically or removing the anions from the sample matrix prior to analysis. The most severe suppression effects in previous work have been from the presence of sulfate because, at high concentrations, the tail of the sulfate peak coelutes with the perchlorate peak.3 It is therefore important to demonstrate that the analytical conditions can produce accurate results for samples that have high concentrations of dissolved salts present in the matrix. The effect of various anions on the suppression of the perchlorate signal was evaluated by preparing a solution containing 1000 mg/L each of chloride, carbonate, and sulfate. Six sets of three samples were prepared to test different sample preparation procedures. Three sets were spiked at a concentration of 0.05 µg/L, and the other three were spiked at a concentration of 0.5 µg/L. Samples were then prepared for analysis using three procedures: (a) the standard sample preparation procedure, (b) (7) Dionex application notes AN138, AN144, and AN151. (8) United States Code of Federal Regulations, Vol. 40, Part 136, Appendix B.

Figure 1. Selected ion chromatograms for both perchlorate transitions. (a) MRM m/z 99 to 83 for 0.05 µg/L standard, (b) MRM m/z 101 to 85 for 0.050 µg/L standard, (c) MRM m/z 99 to 83 for aqueous MDL spiked at 0.05 µg/L, and (d) MRM m/z 101 to 85 for aqueous MDL spiked at 0.05 µg/L.

Table 2. Average Percent Recovery of Perchlorate in Salt Solution (n ) 3) % recovery perchlorate concn (µg/L)

standard preparation

one barium cartridge

two barium cartridges

0.05 0.5

0 38

115 102

131 104

a procedure using a single barium cartridge, and (c) a procedure using two barium cartridges. The barium cartridge-only experiments were attempted because it was expected that the AG16 column would provide adequate separation of the chloride and carbonate anions from the perchlorate ion, thus minimizing the need for removal of those anions using the hydrogen and silver cartridges. The results of the analyses are shown in Table 2. The perchlorate recoveries for the standard procedure were very low or zero with this matrix, although other analyses have indicated that acceptable perchlorate recoveries can be obtained in cleaner matrixes using this sample preparation procedure. The cause of the poor recovery is not known at this time. However, the results of the single barium cartridge tests indicated that perchlorate could be measured with acceptable recovery from this high-salt matrix. The recoveries for the single barium cartridge and double barium cartridge experiments were comparable, but higher baseline noise was observed in the single cartridge experiments. The recoveries from the double barium cartridge experiments show that the procedure minimizes any suppression effects from high-salt matrixes for the salts tested. After these experiments had been performed, we observed poor recoveries for matrix spike results in some test samples that had high concentrations of bicarbonate when the two barium cartridge procedure was used. A set of experiments was performed to define the suppression from bicarbonate. Two sets of water

samples were spiked at 0.2 and 1.0 µg/L perchlorate. One set had a bicarbonate concentration of 50 mg/L, and the other had a bicarbonate concentration of 500 mg/L. The recoveries for the 0.2 and 1.0 µg/L spike concentrations for the 50 mg/L solutions were both 100% while the recoveries were 60 and 48%, respectively, for the 500 mg/L solution. When a single hydrogen cartridge was added to the preparation procedure, the recoveries for the 0.2 and 1.0 µg/L spikes in the 500 mg/L bicarbonate solution were 124 and 117%, respectively. These data indicate that bicarbonate is a suppressor that must be considered for the analysis but that it can be effectively removed with a hydrogen cartridge. The isotopic ratio of 35Cl to 37Cl can be used to improve the specificity of the method and to provide ultimate confidence that the detected signal was due to perchlorate and not an interfering compound.6 The theoretical ratio of 35Cl to 37Cl for natural isotopic abundances is 3.08,5 and the area ratio for the m/z 83 to 85 peaks should be near this value. By comparing the ratio of the area from the 35Cl MRM to the area from the 37Cl MRM, it should be possible to identify erroneous signals if the ratio is not within an established window around 3.08. Previous workers6 have used a ratio of 3.08 ( 5% for the analysis of perchlorate by LC/MS/MS, but our data indicate that this may not be achievable across a range of concentrations. For example, at lower concentrations, it may be difficult to measure accurate areas for the smaller 37Cl peak. This could result in the incorrect rejection of perchlorate results that appear to be outside a 5% acceptance window. We collected a significant number of results at different concentrations and in different matrixes to calculate statistical limits for the acceptance of ratio data. Data were collected from calibration curves, suppression studies, and field samples, resulting in a data set of 99 points. A plot of the m/z 83 to 85 ratio versus concentration is shown in Figure 2. Using only concentrations at or above the MDL, the average measured ratio was 2.93, which is within 5% of the theoretical natural isotopic abundance. The statistical range of these data was 1.11, the standard deviation was Analytical Chemistry, Vol. 76, No. 2, January 15, 2004

471

Figure 2. Isotopic ratio data.

0.19, and the population showed a 6.5% RSD. While the average ratio was just within 5% of the theoretical value, any negative deviation would result in the rejection of those data, and especially at low concentrations, the 5% window proved to be too restrictive. A control limit for the acceptance of analytical results based on the isotopic ratio was calculated using a limit of 3σ about the mean. The resulting acceptance window for the isotopic ratios ranged from 2.4 to 3.5 (-22 to +14%). It should be noted, however, that this window might be too wide for concentrations above 0.2 µg/ L. The amount of scatter observed above 0.2 µg/L is less than for lower concentrations, and it may be advantageous to apply tighter control limits for higher levels of perchlorate. For example, for a sample result of 0.5 µg/L, a ratio as high as 3.4 was never observed in our data set, but with the proposed window, a result with a ratio of 3.4 would be accepted. The application of separate limits for concentrations below 0.2 µg/L, for concentrations between 0.02 and 0.5 µg/L and for concentrations above 0.5 µg/L may further reduce the possibility of reporting a false positive result. The increased scatter for results below 0.05 µg/L is an additional reason that the MDL of the method was set to 0.05 µg/L even though it was possible to observe a signal for the m/z 99 to 83 transition for concentrations as low as 0.01 µg/L. The stability of the perchlorate ion was investigated to provide data that may be used to establish acceptable holding times for perchlorate samples. A groundwater sample was obtained, and four aliquots each were spiked at 0.25, 10, and 20 µg/L. The samples were stored in polypropylene vials for 50 days at 4 °C and then analyzed. The stability study results, given in Table 3, show average perchlorate recoveries between 95 and 115%. These data indicate that perchlorate samples are stable for at least 50 days without degradation when stored under these conditions. QUALITY CONTROL RECOMMENDATIONS Based upon the experimental results described above, several recommendations for sample holding time, MDLs, instrument 472 Analytical Chemistry, Vol. 76, No. 2, January 15, 2004

Table 3. Storage Stability for Perchlorate in Groundwater spike concn (µg/L) 0.25 10 20

n

measd concn (µg/L)

av %R

% RSD

4 4 4

0.2825 9.8521 19.6585

112.59 98.18 99.66

4.04 2.40 3.61

calibration, batch quality control, and method performance requirements are proposed. Holding Time. From the stability study, we conclude that the 28-day holding time typically applied to unpreserved anion samples will be sufficient for application to perchlorate. While the data demonstrate that perchlorate is stable for up to 50 days, a holding time of 28 days is recommended to be consistent with other EPA anion analysis guidelines. Method Detection Limit. While a signal-to-noise ratio of ∼3:1 for the m/z 99 to 83 transition was observed for standards prepared at 0.02 µg/L, method blank contamination has been observed at about that concentration on several occasions. In addition, the significantly increased scatter in the isotope abundance data at low concentrations (Figure 2) leads to the conclusion that the reliability of the ratio as a detection confirmation tool is unacceptably diminished below 0.05 µg/L. Therefore, a concentration of 0.05 µg/L is proposed as the appropriate MDL for this method. Instrument Calibration. We recommend that calibration be performed using a blank and five standards, with standard concentrations at 0.05, 0.1, 0.25, 0.5, and 1 µg/L. Experience has shown that it is routinely possible to achieve linear regression correlation coefficients (r) of 0.995 or greater. Batch Quality Control. Initial calibration verification and continuing calibration verification acceptance criteria should be set at (15%. CCVs should be analyzed after every 10 samples and

at the end of the analytical run. Method blanks should be acceptable only when less than the MDL value, assuming that the MDL value is chosen with the considerations discussed above. We recommend that a detection limit verification standard prepared at approximately twice the MDL level be analyzed in each batch and that a (30% acceptance criterion be applied. Specifying replicate precision acceptance at e20% RPD and matrix spike recovery at (25% is reasonable for this method. A suppression standard containing 1000 mg/L bicarbonate, carbonate, chloride, and sulfate should be spiked with perchlorate at the midpoint of the calibration curve and analyzed in every batch to demonstrate that the method is adequately free from suppression effects. Method Performance. The transitions associated with both of the major chlorine isotopes should be monitored to take full advantage of the specificity available in this method, and statistical process control should be used to establish acceptance criteria for verification of perchlorate detection. The isotope ratio popula-

tion should include data for field samples, calibration standards, and quality control standards. In our work, a reasonable window of acceptance for the ratio was approximately -20 to +15% from the theoretical value based on natural isotopic abundances. Concentration-dependent limits may be applied to the acceptance criteria for the isotopic ratio to minimize any false positive results at higher concentrations. Sample Preparation. The suppression data indicate that it is not necessary to use a silver On-Guard cartridge, but a hydrogen cartridge and two barium cartridges are necessary to reduce suppression from the matrix. The use of two barium cartridges and a single hydrogen cartridge is recommended as the most versatile and rugged procedure for the analysis of a wide variety of environmental samples. Received for review June 6, 2003. Accepted October 27, 2003. AC034618D

Analytical Chemistry, Vol. 76, No. 2, January 15, 2004

473