Anal. Chem. 1995, 67,4387-4395
Determination of N-Nitrosodimethylamineat Part-per-TrillionLevels in Drinking Waters and Contaminated Groundwaters Bruce A. Tomkin$,* Wayne H. Griest, and Cecil E. Higgins Organic Chemistry Section, Chemical and Analytical Sciences Division, Oak Ridge National Laboratory, P.0. Box 2008, Oak Ridge, Tennessee 37831-6120
The carcinogen N-nitrosodimethylamine(NDMA) may be quantitated routinely at ultratrace (ng/L) levels in drinking water or contaminated groundwater. The aqueous sample is passed through a preconditionedEmpore CISmter disk to remove neutral nonpolar species and then extracted continuously overnight with highest purity dichloromethane. The latter is then concentrated to 1 mL, and a large aliquot (up to 200 pL) is loaded onto a dual-stage carbon sorbent trap, after which the solvent is removed with ultrapure helium. The concentrated residues are then injected onto a gas chromatographic column using a short-path thermal. desorber. NDMA is selectively detected using a chemiluminescent nitrogen detector (CLND) operated in its nitrosamine-selectivemode. The reporting limit for this procedure, evaluated using two independent statistically unbiased protocols, is 2 ng of N D W L . A related procedure, employing an automatic sampler instead of the short-path thermal desorber, provides convenient analysis of heavily contaminated samples and exhibits a reporting limit (same protocols cited previously) of 110 ng of N D W L . When the two methods are used together in a “two-tiered” protocol, NDMA concentrations spanning 4 orders of magnitude (ng/L to p g / L levels) may be measured routinely. The law-levelprocedure employingonly the short-paththermal desorber was applied successfully to three sources of drinking water, where NDMA concentrations ranged between 2 and 10 ng of NDWL. The two-tiered protocol was applied to a series of contaminated groundwaters whose NDMA concentrations ranged between approximately 10-7000 ng of NDWL. The results agreed with those obtained from an independent collaboratinglaboratory, which used a different analytical procedure. N-Nitrosodimethylamine (NDMA, CAS Registry No. 62-759) is a contaminant and oxidative degradation product of unsymmetrical dimethylhydrazine (UDMH),’ a component of rocket fuel. Prior waste management and general processing procedures, now considered both outdated and ineffective,introduced both UDMH and NDMA into the groundwater supply.2 NDMA is also a well(1) Fine, D. H.; Rounbehler, D. P.; Rounbehler, A; Silvergleid, A; Sawicki, E.; Krost, IC; DeMarrais, G. A Environ. Sci. Technol. 1977,11, 581-584. (2) Gregory B. Mohrman, Program Manager Rocky Mountain Arsenal, Commerce, CO, private communication, 1994.
0003-2700/95/0367-4387$9.00/0 0 1995 American Chemical Society
known carcinogen3 that is present in nitrite-cured meats,* and cooked foods? Because NDMA is a volatile species, it has also been detected in the atmosphere surrounding UDMHproducing facilities,lJO as well as those of the rubber, leather, metal, chemical, and miniig industries.” The U S . Environmental Protection Agency (EPA) has promulgated a regulatory standard for NDMA in surface water qualified for drinking water use of 0.7 ng of NDMA/L12 -more than 10-fold lower than the drinking water standard currently employed in Ontario, Canada.13-15 Existing procedures for determining NDMA in aqueous samples typically employ dichloromethane extraction followed by concentration to a final volume of 1mL and gas chromatographic analysis of a 2-pL aliquot of concentrate using either a nitrogenphosphorus detector (NPD),l6-l8 mass spectrometric detec~
~~~~
(3) Magee, P. N. The Experimental Basis for the Role of Nitroso Compounds in Human Cancer. Cancer Surv. 1989,8, 208-239. (4) Osterdahl, B.-G. Food Addit. Contam. 1988,5, 587-595. (5) Scanlan, R. A; Barbour, J. F.; Hotchkiss, J. H.; Libbey, L. M. Food Cosmet. Toxicol. 1980,18, 27-29. (6) Scanlan, R A; Barbour, J. F.:Chappel, C. I. J. Agvic. Food Chem. 1990,38, 442 -443. (7) Hotchkiss, J. H.; Havery, D. C.; Fazio, T. J. Assoc. Oft:Anal. Chem. 1981, 64, 929-932. (8) Gavinelli, M.; Fanelli, R; Bonfanti. M.; Davoli, E.; Airoldi, L. Bull. Environ. Contam. Toxicol. 1988,40, 41-46. (9) Hotchkiss, J. H. Preformed N-Nitroso Compounds in Foods and Beverages. Cancer Sum. 1989,8, 295-321. (10) Fine, D. H.; Rounbehler, D. P.; Pellizzari, E. D.; Bunch, J. E.; Berkley, R. W.; McCrae, J.; Bursey, J. T.; Sawicki, E.; Krost, K.; DeMarrais, G. A Bull. Environ. Contam. Toxicol. 1976,15, 739-746. (11) Tricker, A. R.; Spiegelhalder, B.; Preussmann, R Environmental Exposure to Preformed Nitroso Compounds. Cancer Surv. 1989,8, 207-239. (12) Water Quality Standards; Establishment of Numeric Criteria for Priority Toxic Pollutants; States’ Compliance. Fed. Regist. Dec 22, 1992,246, 60848-60918. (13) Taguchi, V. Y.; Jenkins, S. W. D.; Wang, D. T.; Palmentier, J.-P. F. P.; Reiner, E. J. Can. J. Appl. Spectrosc. 1994,39,87-93. (14) Jobb, B.; Hunsiger, R.; Meresz, 0.; Taguchi, V. Removal of N-Nitrosodimethylamine (NDMA)ffomthe Ohsweken (Six Nations) Water Supply. Interim Report Uuly 1993); Queen’s Printer for Ontario, Canada: Ontario, 1993;PIBS 2687. (15) Taguchi, V. The Determination of N-Nitrosodimethylamine (NDMA) in Water by Gas Chromatography-High Resolution Mass Spectrometry (GC-HRMS); Environment Ontario (Canada), Laboratory Services Branch, Quality Management Office: Ontario, 1994; Method catalogue code NDMA-E3291A, originally approved Feb 19, 1993; revised and approved Apr 7, 1994. (16) Method 8070, Nitrosamines by Gas Chromatography. Test Methods for Evaluating Solid Waste, Volume 1B: Laboratory Manual Physical/Chemical Methods, SW-846, 3rd ed.; U S . Environmental Protection Agency: Washington, DC, proposed update I, Nov 1990. (17) Method 607, Nitrosamines. Code of Federal Regulations: Protection of the Environment, Part 1 3 6 Title 4 0 US. GPO: Washington, DC, Oct 1984. (18) The Determination of N-Nitrosodimethylamine and N-Nitrosodi-n-propylamine in Water by Gas Chromatography, Ve’ersionNo. 2. DataChem Laboratories: Salt Lake City, UT, June 1990.
Analytical Chemistry, Vol. 67, No. 23, December 1, 1995 4387
t ~ r , or ~ chemiluminescent ~ . ~ ~ , ~ ~ nitrogen detector (CLND),1,111,21-23 Such a protocol does not permit detection of NDMA at the desired health-based criterion unless high-resolution mass spectrometric (HRMS) detectors are e m p l ~ y e d . ~ ~The J j additional needed sensitivity can be achieved by adding a short-path thermal desorption unit (SPTD)24-27 as an accessory to the injection port of the gas chromatograph employed. The analytical procedure described in this work employed an initial solid-phase extraction of groundwater samples with a preconditioned Empore CIS disk, used to remove interfering neutral species prior to continuous overnight extraction of the filtrate with highest purity dichloromethane.2s The resulting extract is concentrated to a final volume of 1 mL; the residues from a 200-pL aliquot of the concentrate are injected onto a gas chromatographic column using the short-path thermal desorber. NDMA is collected at the head of the column near ambient temperature (without cryotrapping) , eluted in the conventional manner, and detected using a CLND. The reporting limit for this procedure, 2 ng of NDMA/L of groundwater, was calculated using two statistically unbiased procedure^.^^^^^ The certified range for this procedure is 2-40 ng of NDMA/L of groundwater. A related procedure permitted unattended analysis of samples for NDMA concentrations exceeding 100 ng of NDMA/L of groundwater. The sample preparation is the same as that described above; however, 2-pL aliquots of dichloromethane concentrates are injected onto the gas chromatographic column using an automatic sampler. EXPERIMENTALSECTION
Reagents, Standards, and Samples. Sample Collection and Storage. Groundwater samples should be collected in commercially available precleaned 1-L narrow-neck amber glass bottles, stored at 4 “C, and extracted within 7 days after sampling.29 NDMA Standard Solutions. Stock solutions of NDMA, 5000 pg/mL in methanol, are purchased from NSI Environmental Solutions, Inc. (Research Triangle Park, NC). Aliquots are diluted sequentially to final concentrations ranging between 2 and 240 ng of NDMA/mL in highest purity dichloromethane (GC2grade, ~~
~
~
~~
(19) Plomley, J. B.: Koester, C. J.; March. R. E. Anal. Chem. 1994,66, 4437-
4443. (20) The Determination of N-Nitrosodimethylamine in Water by Gas Chromatography/Mass Spectrometry Selected Ion Monitoring, Method UM34, Version 1; DataChem Laboratories: Salt Lake City, UT, Oct 1992. (21) Fine, D. H.; Rounbehler, D. P.; Sawicki, E.: Krost, K Enuiron. Sci. Technol. 1977,11, 577-580. (22) Fine, D. H.; Rounbehler, D. P. 5 Chromatogr. 1975,109, 271-279. (23) Fine, D. H.; Rufeh, F.; Lieb, D.: Rounbehler, D. P. Anal. Chem. 1975,47, 1188-1191. (24) Manura, J. J.; Hartman, T. G. Am. Lab. 1992,24, May. (25) Manura, J. J. LC-GC 1993,11, 140-146. (26) Short Path Thermal Desorption Technical Bulletin: Mod$cafions of the Hewlett-Packard 5890 GC Split/Splitless Capillaly Injection Port for Use With the Short Path Thermal Desotption System, No. 2; Scientific Instrument Services: Ringoes, NJ, Jul 1991. (27) Short Path 77aermal Desorption Technical Bulletin: Detection of Volatile Organic Compounds in Liquids Utilizing the Short-Path Thermal Desorption System, No. 8 Scientific Instrument Services: Ringoes, NJ, Sep 1991. (28) Method 3520A. Continuous Liquid-Liquid Extraction. Test Methods for Evaluating Solid Waste, Physical/Chemical Methods, SW-846, 3rd ed.; Final Update 1.U S . Environmental Protection Agency: Washington, DC, Jul1992; Revision 1. (29) Program Managerfor Rocky Mountain Arsenal: Chemical Quality Assurance Plan. Version I, Sep 1993. (30) Appendix B to Part 136-Definition and Procedure for the Determination of the Method Detection Limit-Revision 1.11.Code of Federal Regulations: Protection of the Environment, Parts 100-149; Title 40: U S . GPO: Washington, DC, Revised Jul 1, 1990.
4388 Analytical Chemistry, Vol. 67, No. 23, December 1, 1995
Burdick & Jackson, Muskegon, MI, or equivalent). Fresh standards should be prepared every 60 days. Caution: NDMA is a listed regulated OSHA carcinogen. Current regulations stipulate that neat NDMA must be dispensed only by individuals with specific training, and then only in contained areas. Dilute solutions of NDMA may be handled with the normal care and caution due any hazardous chemical. NDMA is a listed waste (no. P082) on the US. Environmental Protection Agency’s “Plist.31 Consult the local Environmental Protection Officer for guidance concerning disposal of standards. Preparation of Blanks and Spiked Samples for Method Certijication and Quality Assurance. Aliquots (1-L) of synthetic groundwater are prepared from HPLC-grade water (Burdick & Jackson, J. T. Baker (Phillipsburg, NJ), or equivalent) which is fortified to 100 mg/L each in sulfate and chloride, as described e l s e ~ h e r e . ~ ~ These are fortified to concentrations ranging between 2’and 40 ng of NDMA/L using a master spiking solution (0.1pg of NDMA/ mL of water). (Note: High-purity water generated from typical “on-demand systems contains NDMA and may not be substituted for bottled HPLC-grade Water produced from ondemand deionized/reverse-osmosis systems having a final W treatment may be substituted.15) Apparatus. Initial Sample Preparation and Extraction. Aqua ous samples are filtered through Empore Cis filter disks (47 mm diameter, J. T. Baker or Varian Sample Products Division, Harbor City, CA). An all-glass funnel/support assembly (DELTAWARE, with a 1-L filtering flask and a 300-mL reservoir, handles 47 mm diameter disks, part no. KT 9382547 or equivalent) is used to support the disk and hold the filtrate. The disks are preconditioned with two 10-mL portions each of methanol and water. Each aliquot remains on the disk for 1 min prior to removal. Do
not permit the disk to become dry during the conditioning sequence. The overnight continuous extraction employs the usual glass apparatus (extractors, flasks, concentrator tubes, etc.) and dichloromethane volume described in ref 28. Sorbent Trap. The sorbent trap (3.0 mm i.d. x 100 mm glasslined stainless steel desorption tube, Scientific Instrument Services, Ringoes, NJ) contains 125 mg each of Carbotrap and Carbotrap C (Supelco, Inc., Bellefonte, PA). The two sorbents are packed into the trap using the “tapfill” method. Small plugs of clean silanized glass wool separate the sorbents into distinct beds and seal the ends of the trap. The sorbent trap should be conditioned using the short-path thermal desorber (described below) using repeated 5min desorptions at 250 “C, with ultrapure helium (research grade, 99.999% or better) carrier gas (3 mL/ min) flowing from Carbotrap to Carbotrap C, until the resulting gas chromatogram is free of contaminant, and a stable baseline is achieved in either the “total nitrogen” or “nitrosamine-selective” mode of the CLND (described below). The injection needle employed (25 mm length, 0.63 mm o.d., 0.12 mm id., Scientitic Instrument Services) fits on the Carbotrap ~
~~
(31) Part 261-Identification and Listing of Hazardous Waste, and Paragraph 261.33-Discarded Commercial Chemical Products, Off-SpecificationSpecies, Container Residues, and Spill Residues Thereof. Code of Federal Regulations: Protection of the Environment, Parts 260-299 Title 40; U S . GPO: Washington, DC, Revised Jul 1, 1994. (32) Fiddler, W.; Pensabene, J. W.: Doerr, R. C.: Dooley, C. J. Food Cosmet. Toxicol. 1977.15, 441-443. (33) Kimoto, W. 1.; Dooley. C. J.: Carre. J.: Fiddler, W. Water Res. 1980,14, 869-876. (34) Gough, T. A.; Webb, K S.: McPhail. M.F. J. Cosmet. Toricol. 1977,15, 437-440.
C end of the trap during desorption. A matched pair of graphite and vespel/graphite needle seals are installed in the needle body such that the vespel/graphite seal always faces an end of the sorbent trap. Short-Path Themal Desorber. The short-path thermal desorber (SFID),Model 'I'D-1 (Scientiiic Instrument Services, Ringoes, NJ) , was positioned on its mounting plate over the injection port of a Hewlett-Packard Model 5890 Series I1 gas chromatograph. A carrier gas solenoid switching valve was installed in the helium line of the gas chromatograph, as recommended in ref 26. Graphite and vespel/graphite seals are installed in the "top tube" of the desorber such that the vespel/graphite seal always faces the sorbent trap. Dichloromethane aliquots are typically dried at room temperature on the sorbent trap mounted into the short-path thermal desorber using ultrapure helium (25-30 mL/min) for 2 min. The residues are then desorbed for 5 min at 150 "C using ultrapure helium (3 mL/min). ChemiluminescentNitrogen Detector (CLND). An Antek Model 705D pyrochemiluminescent nitrogen detectorN/@ ' 99058, Antek Instruments, Inc., Houston, TX) was retrofitted to the gas chromatograph described above. The high-temperature furnace was mounted on a new detector base controlled by the CLND console and not by the gas chromatograph. Vacuum was obtained with a "silent" M2 noise-filtered single-stage pump with vacuum capability of 15 in. of mercury at 1 ft/min (Antek). Researchgrade (99.999%)oxygen passed through both the furnace ("pyro" oxygen) and the ozone generator ("ozone" oxygen) located within the CLND console. When new sorbent traps were conditioned, the CLND was operated frequently in the total nitrogen mode, to evaluate the cleanliness of the trap. The factory-optimized operating parameters were as follows: furnace temperature, 900"C; detector base temperature, 280 "C; photomultiplier tube voltage and gain, 750 V and high; photomultiplier attenuation and output gain, 16 and 50; pyro oxygen (through the furnace) flow rate, 80 mL/min; ozone oxygen (passing through the ozone generator) flow rate, 5 mL/min. These settings collectively permitted the greatest sensitivity for nitrogen and the greatest selectivity for nitrogenover-carbon. When NDMA was determined in dichloromethane concentrates, the CLND was operated exclusively in the nitrosamineselective mode. The factory-optimized parameters were as follows: furnace temperature, 250 "C; photomultiplier voltage and gain, 750 V and high; photomultiplier attenuation and output gain, 8 and 50; pyro oxygen flow rate, 30 mL/min; ozone oxygen flow rate, 5 mL/min. Gas Chromatograph. The Hewlett-Packard Model 5890 Series I1 gas chromatograph equipped with a split/splitless injector both supports and is physically interfaced to the SFTD and the CLND. One of the existing data interface boards was replaced with a Hewlett-Packard analog input board so that the output from the CLND could become part of the overall system network. The integrator automatically starts whenever the column oven temperature program begins. The gas chromatographic parameters and conditions were optimized as follows: injector, equipped with "double gooseneck liner, 150 "C; injector purge valve, OFF at 0.00 min and ON at 5.00 min; guard column, 0.53 mm i.d. x 5 m, deactivated and
uncoated fused silica column (Restek Corp., Bellefonte, PA) connected to the analytical column using a Universal Press-Tight connector (Restek); analytical fused-silica capillary column, Rtx200 (Crossbond biiluoropropylmethyl), 3.00 pm film thickness, 0.53 mm i.d. x 30 m (Restek). Other columns evaluated were purchased from either Restek Corp. or J&W Scientific (Folsom, CA). (Note: The polyimide coating must be removed from approximately 25 mm of the outlet end of the column, where the column enters the CLND furnace. This is achieved conveniently using either a match or cigarette lighter.) Carrier gas is ultrahighpurity helium, 4 mL/min (13.0 cm/s). Column oven temperature is increased linearly from 35 "C (hold for 5 min) to 175 "C (hold for 5 min) at 6 "C/min. Automatic Sampler. When the high-level procedure was used, a Hewlett-Packard Model 7673A automatic sampler replaced the SPTD and injected 2 pL on column. The gas chromatographic operating parameters are the same as those described above with the exception of the purge valve settings, which become OFF at 0.00 min, ON at 2.00 min, and OFF at 30.00 min. Procedure. Groundwater samples are filtered through the conditioned Empore disk under vacuum and then extracted continuously overnight (18 h) using highest purity dichloromethane (Burdick & Jackson GC2 grade or equivalent, described above) as described in ref 28. The dichloromethane extract is reduced to a final volume of 1 mL using KudernaDanish concentration followed by nitrogen blowdown. A 1WpL aliquot of the dichloromethane concentrate is added slowly to the Carbotrap C end of the dual sorbent trap, allowing sufficient time for the liquid to soak into the trap bed. The trap is then mounted into the SFTD (Carbotrap C end is connected to the top tube without the needle attached) and dried with ultrahighpurity helium for 2 min at 25-30 mL/min. This procedure is repeated, leaving the residues from 200 pL of concentrate on the trap. The trap is then disconnected from the top tube and inverted. The needle is connected to the Carbotrap C end of the trap, and the remaining Carbotrap end is screwed into the top tube. The residues present on the trap are then desorbed directly from the sorbent trap onto the guard and analytical gas chromatographic columns, both maintained at 35 "C, using a 150 "C desorption temperature maintained for 5 min and a helium desorption flow rate of 3 mL/min. The column oven temperature program commences immediately after completion of desorption. RESULTS AND DISCUSSION
Method Optimization. The isolation of NDMA from aqueous solutions using the continuous extraction procedure described in US. EPA Method 352OAz8represents not only the most conservative analytical approach but also the one that generates the greatest volume of hazardous chemical waste (spent dichloromethane). Procedures employing either solid-phase extraction with reversed-phase sorbent or purge-and-trap technologies are largely ineffective. Very recent p a p e r ~ ' ~ J jdescribe .~j the successful extraction of NDMA from drinking waters using Ambersorb 572 as the extraction medium; however, drinking waters contain lower concentrations of organic materials than drinking waters, and the reported recovery of this extraction did not exceed 30%. Pending additional evaluation of Ambersorb 572 for the samples (35) Jenkins, S. W. D.; Koester, C. J.; Taguchi, V. Y.; Wang, D. T.; Palmentier, J:P. F. P.; Hong, K P. Enuiron. S a . Pollut. Res. Int., in press.
Analytical Chemistry, Vol. 67, No. 23, December 1. 1995
4389
in question, the extraction procedure detailed in US.EPA Method 3520A was employed; the used dichloromethane was recycled. This should be considered an interim solution at best, since the US.Environmental Protection Agency has targeted the use of dichloromethane for reduction.36 Because the residues concentrated from 40-60 mL of dichloromethane (approximately 100 times the normal loading) would be loaded onto the trap, it was imperative to use the purest dichloromethane available in this work. The factory specifications for Burdick & Jackson GC2 dichloromethane suggested that this solvent grade would be satisfactory in the current procedure without additional in-house pudication. Pesticide-grade dichloromethane from two reputable manufacturers were evaluated, and both exceeded the nominal background levels observed from the GC2 grade solvent. The groundwater samples expected from certain defense sites are typically contaminated with substantial quantities of neutral species, particularly diisopropylmethane phosphonate @IMP), which would be extracted by US.EPA Method 3520A and which may interfere with the determination of NDMA. Nonpolar neutrals such as DIMP may be removed easily from samples prior to the continuous extraction using a preextraction step, in which the groundwater is passed through a preconditioned C I Empore ~ filter disk. The extraction efficiency for DIMP under these conditions exceeds 98%. NDMA is far more soluble in water than DIMP, and therefore passes through the disk unextracted. The choice of trap components employs 250 mg of total mass carbonaceous sorbent, permitting the addition of 100 pL of dichloromethane without loss of liquid. The bed design permitted most of the dichloromethane extract to be concentrated on a layer of Carbotrap C, which exhibits a low surface area and would permit efficient desorption of NDMA. A layer of Carbotrap, which features a l@fold greater surface area, collects any material that breaks through the Carbotrap C bed. More active sorbents such as Carbosieve SI11 sorb dichloromethane and NDMA strongly and therefore are not acceptable trap components. Another consideration was a trap containing either Carbotrap or Carbotrap C, but not both. In both cases, NDMA could be collected and desorbed easily, but the analyte peak shape was invariably "better" (more symmetrical, narrower peak width, and taller) when both sorbents were used in the trap as compared to just one. For that reason, a dualcomponent rather than a single-component trap was selected. The desorption conditions were optimized for five parameters: (a) drying time; (b) drying flow rate; (c) desorption temperature; (d) desorption time; and (e) desorption flow rate. In general, a 2-min drying time with the helium flow rate fixed at approximately 25-30 mL/min appeared sufficient to remove dichloromethane from up to 100 pL of extract spiked onto the trap. Longer drying times did not improve the removal of dichloromethane; higher flow rates could not be achieved using the SPTD and might actually strip volatile NDMA from the trap. The recovery of NDMA from the trap improved as the desorption time was held constant (5 min) and the desorption flow rate of helium carrier gas was reduced stepwise from 25-30 to 10 mL/ min and finally to 3 mL/min. Flow rates less than 3 mL/min might actually be preferred; however, it became very difficult to monitor such slow desorber flow rates conveniently using the existing thermal desorber equipment (rotameter ball). The (36) Ho, J. S.; Tang, P. S.; Eichelberger. J. W.; Budde, W. L.]. Chromatogr. Sci. 1995,33, 1-8.
4390 Analytical Chemistry, Vol. 67, No. 23,December 7, 1995
NDMA response increased steadily as the desorption temperature was reduced from 300 to 150 "C but dropped sharply at 100 "C. For that reason, the optimized thermal desorption temperature was maintained at 150 "C. The overall desorption efficiency was easily measured by first spiking the trap with a substantial mass of NDMA (normally 8000-12000 pg) and then desorbing it twice using the optimized conditions. The peak area for NDMA was measured both times. The measured desorption efficiency, 95-98%, was deemed satisfactory for this analysis. The known thermal lability of NDMA also influenced the contiguration and temperature of the gas chromatographic injector port. To minimize thermal degradation, a double-gooseneck linerj7 and low injector temperature (150 "C) were employed. The gas chromatographic analytical column must be capable of permitting efficient collection of the analyte at or near ambient temperature and separating NDMA from other components in the extract. NDMA is a volatile but high-boiling species (151 "C) that was collected efficiently at a column oven temperature of 35 "C; cryotrapping with liquid nitrogen was not necessary. Of four thickfilm phases evaluated, Restek Rtx 200 (Crossbond trifluoropropylmethyl, 3.0 pm) consistently produced acceptable peaks for NDMA (symmetrical with narrow peak width) and exhibited the necessary thermal stability to be compatible with the SPTD. Other columns evaluated either did not exhibit sufficient resolution between the solvent peak and NDMA (DB5,1.5 pm film thickness and Rtx 65,l.O pm film thickness) or did not exhibit the required thermal stability (Stabilwax, 2.0 pm film thickness). Literature procedures have employed either the NPD, CLND, or high-resolution mass spectrometer (HRMS) for quantitation of NDMA. The applicability of each was weighed against the expected properties of the samples to be analyzed. Procedures employing the NPD would respond to residual DIMP in the extracts and would also exhibit a detection limit ranging between 42 and 150 ng NDMA/L, or approximately 60-200 times the desired level of 0.7 ng of NDMA/L. For these reasons, procedures employing the NPD were set aside. Methods employing HRMS detection of NDMA using selected-ion monitoring and a nonstandard ionization modelq would be capable of providing specificity for NDMA even in the presence of part-per-million quantities of DIMP and other potential contaminants. The sensitivity of these procedures would permit quantitation at or near the 0.7 ng of NDMA/L 1 e ~ e l . l ~35 However, the cost of the HRMS system is normally beyond the budgets of most laboratories and could be justified only if many hundreds or even thousands of samples were to be analyzed. It was strongly preferred that the analytical instrument employed be less expensive than a HRMS system. For all of these reasons, analytical procedures employing high-resolution mass spectrometric detection of NDMA were also set aside. Taguchi et al. l3 explained that procedures employing low-resolution mass spectrometry, which are presently quite common, are also unacceptable. Both chlorobenzene and ethylbenzene co-elute with NDMA and its perdeuterated analog, NDMA-d6, on certain gas chromatographic columns. Both species also exhibit a minor ion at m / z 74 in their mass spectrum (identical to the parent ion of NDMA), which might interfere in the quantitation of NDMA if these additional species are present in sufficient quantities. The CLND exhibits a nitrogen-to-carbon selectivity greater than that of either the Hall ~
~~
~
~~
(37) Restek Chromatography Products 1994/1995, p 73
or nitrogen-phosphorus detector as well as excellent sensitivity (20 pg of nitrogen, correspondingto 106 pg of NDMA) , wide linear range (4 orders of magnitude), and easy operation?* Such a detector is normally not available as a standard option on most commercially available gas chromatographs, but it can be installed at modest cost. Such features, taken together, make procedures employing the CLND an attractive alternative to those that use HRMS. Even with a CLND “on line”, the health-based criterion of 0.7 ng of NDMA/L could not be achieved under typical analytical conditions which feature a 1-L groundwater sample, a final dichloromethane extract concentrate volume of 1mL, and a 2-pL injection of extract onto the gas chromatographic column. With the SPTD employed as the injector, a 100-fold greater sample volume than normal could be injected, and the health-based criterion could be satisfied. Because the SPTD provides analyte m m not concentration for analysis, the chromatographic peaks obtained are independent of the volume of extract spiked onto the trap. In this work, 50-pL aliquots of NDMA standards prepared in dichloromethane were used to calibrate the detector, while 200 pL of sample extract concentrates was employed for analysis. Greater sample extract aliquots were not employed because it was desirable to retain a portion of each extract for archival or reanalysis. Method Cedcation and the Statistical Determination of the Reporting Limit Using the SPTD as the Injector. The reporting limit for this procedure was evaluated initially using the protocol established by the US.Armyz9for calculating the method reporting limit (MRL). This procedure requires a target reporting limit VRL), which was set at 2 ng of N D W L (believed to be achievable) or approximately three times the health-based criterion of 0.7 ng of N D W L . The first part of the protocol evaluated the detector response and calibration curve expected. Briefly, a calibration curve was prepared on two consecutive days from standards ranging in concentration from 2 to 160 ng of NDMA/mL of dichloromethane. In all cases, a 50-pL aliquot was spiked onto the sorbent trap and dried, and the residues were thermally desorbed as described above. The resulting data represented the range of 100-8000 pg of NDMA detected by the CLND. A spreadsheet statistical program, available from the U S . Army, calculates lack of fit based on a 95% significance level for (a) a linear model with zero intercept, (b) a linear model with nonzero intercept, and (c) an intercept statistically different from zero. Taken together, the statistical tests demonstrated that a linear model with zero intercept was appropriate for the calibration data. The second part of the protocol evaluated the analytical methodology itself and calculated the MRL. Two sets of spiked synthetic groundwater samples, analyzed on consecutive days, reflecting concentrations of NDMA ranging between 1x TRL and 20x TRL (Le., 2-40 ng of NDMA/L) were subjected to the extraction procedure. Aliquots of the final dichloromethane concentrate (200 pL from a final volume of 1mL) were processed using SPTD followed by gas chromatography and detection of NDMA using the CLND. The resulting integrated peak areas were converted to mass of NDMA (in pg) using the calibration data to determine the “found mass of NDMA in each case. These found masses of NDMA were compared to the target masses of NDMA, which ranged between 400 and 8000 pg of NDMA and (38) Britten, A J. Res. Dev. 1989,77-80.
A
I
0
C
1
I
I
10
20
30
Time (min.)
Figure I. Detection of NDMA in (A) synthetic groundwater sample fortified to 10 ng of NDM#L, (B) synthetic groundwater sample fortified to 2 ng of NDM#L, and (C) synthetic groundwater blank.
1
2 3 4 5 6 7 8 9 Target Mass (picograms) Thousands Figure 2. Graphical determination of the method reporting limit (MRL) using all of the method certification data. (-) = linear regression curve based on the found NDMA masses; (- -) = lower confidence limit; (- - -) = upper confidence limit: 0 = found NDMA mass: 0 = MRL.
0
1
are covered by the calibration range described above. Figure 1 shows the respective chromatograms for a synthetic groundwater sample fortified to 10 ng of NDMA/L (Figure lA),a synthetic groundwater sample fortified to 2 ng of N D W L (Figure lB), and a synthetic groundwater blank (Figure 1C). All of the method certification data are displayed in Figure 2, while the calculation of the MRL is depicted graphically in the enlargement given in Figure 3. The MRL value is located by the following four-step procedure: (1) calculate and plot a regression line, representing the found vs target concentrations, with a p propriate two-sided 90%confidence limits for a predicted observation; (2) locate the intercept of the upper 90% predictive confidence limits with the y-axis (found concentrations); (3) draw a horizontal line from this intercept until it intersects the lower 90%predictive confidence limits; and (4) draw a vertical line from the intercept described in (3) to the x-axis (target concentrations). This intersection with the x-axis is the M E . Additional details describing the calculation of the MRL are presented elsewhere.39 The slope of the certification line given in Figure 2 (0.499) may be taken as the overall recovery of NDMA using the procedure (Le., 50%recovery). In this particular case, the calculated MRL was 348 pg of NDMA, as shown in Figure 3. This value may be (39)Tomkins, B. A; Memweather, R; Jenkins, R A; Bayne, C. K /. h o c . Ofi Anal. Chem. Int. 1992,75,1091-1099.
Analytical Chemistry, Vol. 67, No. 23, December 1, 1995
4391
1000 I
E
800
0
.-0
ea f
600
c
I
Table 2. Calculation of the Method Detection Limit using Replicate Water Samples Containing 10 ng of NDMLVL, Short-Path Thermal Desorber Used as the Injector
400
SD, best 7 samples MDL, 7 samplesb(ng/L) SD, all 8 samples MDL, 8 samplesc
5.43 6.47 5.77 6.40 7.07 5.61 6.49 5.30 0.59 1.9 0.63 1.9
n
: 200
L
U
0 Target Mass (picograms)
Figure 3. Enlargement of the method certification curve showing the calculation of the MRL. (-) = linear regression curve based on the found NDMA masses; (- -) = lower confidence limit; (- - -) = upper confidence limit; 0 = found NDMA mass; (- - -) = steps used to calculate the MRL. Table 1. Comparison of Target and Found NDMA Masses on the Two Days of Certification Using the Short.Path Thermal Desorber as the Injector”
concn (ng/L)
target mass NDMA (pg)
2 4 10 20 40
400 800 2000 4000 8000
a
calculated NDMA (ng/L)“
1
-2
target NDMA
replicate no.
found mass of NDMA day 1 day 2 151 402 1059 2005 4071
288 411 993 2006 3973
Values not corrected for recovery.
readily converted to approximately 2 ng of N D W L of synthetic groundwater knowing (a) the NDMA mass, (b) the aqueous sample volume (1 L prepared, 0.9 L analyzed), (c) the dichloromethane concentrate final volume (1 mL), and (d) the volume of (c) used for the determination (200 pL). Table 1 presents the target and found masses observed on both days of certification.
348 pg of NDMA 1000pL of concentrate 0.9 L of sample 200 pL of concentrate ng Of NDMA = 1.9 of N D W L = MRL 1000 pg of NDMA As part of the proposed analytical procedure, several additional samples were analyzed during certification: (a) a “trap blank to ensure that the trap was free of NDMA; (b) a QA/QC test whereby 50 pL of a standard, 16 pg of N D W p L in dichloromethane (total loading 800 pg of NDMA) was spiked on the trap and desorbed to test for response well above the instrument noise; (c) a dichloromethane concentrate blank to ensure that the dichloromethane did not contribute to the measurement of NDMA; (d) a synthetic groundwater blank to ensure that the HPLC-grade water did not contribute to the measurement of NDMA; and (e) a QA/QC test whereby 50 pL of a standard containing 2 pg of N D W p L in dichloromethane (total loading 100 pg of NDMA) was spiked onto the trap and desorbed to test for response near the instrumental detection limit. In general, the trap, dichloromethane, and synthetic groundwater contained no detectable NDMA, as expected, and the minimum quantity of NDMA which could be observed (S/N -3) was 100 pg. Note that this figure, even though it is estimated without statistical validation, was close to the calculated NDMA sensitivity given in the manufacturer’s 4392 Analytical Chemistty, Vol. 67, No. 23, December 7, 7995
3 4 5 6 7 8
Values not corrected for recovery. Student’s t-value (one-tailed) at 99%confidence for seven samples (six degrees of freedom) = 3.143. Student’s t-value (one-tailed) at 99% confidence for eight samples (seven degrees of freedom) = 2.998.
literature (viz.,106 pg of NDMA), and it therefore provided some reasonable assurance that the CLND was operating at or near its maximum sensitivity. The detection limit was also determined using the method detection limit WDL) defined by the US.Environmental Protection Agency.30 Briefly, eight (minimum seven required) 1-L synthetic groundwater samples were fortified to 5x TRL (Le., 10 ng of N D W L ) and analyzed as described above. The standard deviation, which was calculated for both seven and eight replicates, was then multiplied by the appropriate one-tailed Student’s t statistic for 99% confidence. The resulting value is the MDL, which in this case was approximately 2 ng of N D W L , as shown in Table 2. These calculations also satisfy an additional criterion given in ref 29, viz.,the magnitude of the MRL (1.9 ng of NDMA/ L, discussed previously) should be equal to or larger than that of the MDL (1.9 ng of N D W L , given in Table 2). Method Certification and the Statistical Determinationof the Reporting Limit Using the Automatic Sampler as the Injector. To evaluate the performance of the high-level procedure for determining NDMA in heavily contaminated groundwaters, the short-path thermal desorber was removed, replaced with the HP 7673A automatic sampler, adjusted to inject 2-pL volumes of either standards or extract concentrates on cohmn. The evaluation of the calibration curve, using the lack of fit and zero intercept tests described above, required testing standards containing 50-2000 ng of NDMA/mL over 2 consecutive days. Once again, a linear calibration curve was obtained. The statistical evaluation showed no significant lack of fit when using a linear model with intercept, but did indicate the presence of a nonzero intercept. The MRL was then evaluated as described previously, using synthetic groundwater samples that had been fortified to 100-2000 ng of NDMA/L. The target and found NDMA concentrations from this method certification are presented in Table 3. When the MRL was calculated using all of the data values given in Table 3, a concentration of 272 ng of NDMA/mL of extract, corresponding to 302 ng of NDMA/L of groundwater was obtained, considerably greater than the desired 100 ng of NDMA/ L. The NDMA concentration in the original groundwater sample is calculated knowing the extract concentration (measured) and
Table 3. Comparison between the Target and Found Concentrations of NDMA When the Automatic Sampler Is Used for Sample Injection
target NDMA concn (ng/L)
found NDMA concn (ng/L)O day 1 day 2 67 147 376 745 1324
100 200 500 1000 2000
62 120 357 796 1401
i\
f
Q
R
Values not corrected for recovery. Table 4. Calculation of the Method Detection Limit Using Replicate Spiked Synthetlc Groundwater SampIes Containing 500 ng of NDMA/L, Automatic Sampler Used as the Injector
replicate no.
calculated sample concn (ng/L)a
sample SD, best 7 samples MDL, best 7 samplesb(ng/L) sample SD, all 8 samples MDL, all 8 samplesC
369 336 316 348 315 363 389 387 27 84 29 87
1
“L 3 4 5 6 7 8
Values not corrected for recovery. Student’s t-value (one-tailed) at 99%confidence for seven samples (six degrees of freedom) = 3.143. Student’s t-value (one-tailed) at 99% confidence for eight samples (seven degrees of freedom) = 2.998.
the volumes of the aqueous sample and dichloromethane concentrate (0.9 L and 1 mL, respectively), since
272 ng of NDMA 1mL of concentrate m L of concentrate 0.9 L of groundwater 302 ng of NDMA/L = MRL The data set was then truncated by removing the found concentrations observed at 2000 ng of NDMA/L, as recommended in ref 29. The MRL calculated using the truncated data set was 103 ng of N D W m L of extract, corresponding to 114 ng of NDMA/L of groundwater, was obtained. This latter value was close enough to the TRL to be useful; further truncation of the data set was unwarranted. The estimated overall method recovery, based on the slope of the MRL certification graph, was 71%. The MDL was calculated using eight replicate spiked synthetic groundwater samples (minimum seven required) fortified to 5x TRL (500 ng of N D W L ) and the same protocol described previously. The resulting data, presented in Table 4, demonstrate that the calculated MDL, 87 ng of NDMA/L, satisfied the criteria for acceptance presented previously. Determination of NDMA in Three Sources of Drinking Water. Four replicate 1-Lbottles from each of three independent sources of drinking water were submitted for determination of NDMA. Because the anticipated NDMA concentration was expected to be well below 100 ng/L (probably below 10 ng/L), the short-path thermal desorber was employed exclusively for the determinations. A 1-L synthetic groundwater sample fortified to 10 ng of N D W L was analyzed as described both to assure
I
I
1
I
0
10
20
30
Time (mln.)
Figure 4. Detection of NDMA in two sources of drinking water, (A) and (B). The corresponding synthetic groundwater blank (C) exhibits a smaller concentration of NDMA than in either source.
quality and to evaluate the method recovery. The method of external standards was employed to quantitate NDMA in the drinking waters. Normal sample throughput is approximately 10 samples/day, including blanks and fortified groundwater (QA/ QC) samples. The chromatograms presented in Figure 4 generally show a number of substantial peaks, most of them arising from the concentrated dichloromethane as well as NDMA. Repeated tests using small alkylated nitrosamines, viz.,N-nitrosoethylmethylamine, N-nitrosodiethylamine, N-nitrosodi-n-propylamine, and Nnitrosodi-n-butylamine, demonstrated that these authentic nitroso compounds exhibited retention times that were different than those of the extraneous peaks observed. To date, none of these products have interfered with the determination of NDMA but have largely precluded the use of surrogate compounds. In many instances, the calculated NDMA concentrations for drinking water, given in Table 5, represented a value below the MRL of 2 ng of NDMA/L. For this reason, these data carry a “J” (interpret as “estimated”) qualifier. The typical observed relative standard deviation (RSD) for each source was approximately 1020%. Determination of NDMA in Contaminated and Treated Groundwaters. The groundwater from at least one defense site has become severely contaminated over time, with concentrations of NDMA reaching as high as 5 ppb (5000 pg/LI2. Samples of treated and untreated groundwaters were “split” and analyzed both by the procedures described herein and a commercial analytical laboratory. Initially,all of the groundwater samples were analyzed using the high-level procedure and the automatic sampler to identify and quantitate strongly contaminated samples. Those samples whose apparent NDMA concentrations were below the MRL (i.e.,
