Quantitative Determination of Total Molar Concentrations of

A highly sensitive and quantitative group parameter to determine total molar concentrations of organic micropol- lutants that can bioaccumulate in the...
1 downloads 0 Views 280KB Size
Anal. Chem. 1996, 68, 2916-2926

Quantitative Determination of Total Molar Concentrations of Bioaccumulatable Organic Micropollutants in Water Using C18 Empore Disk and Molar Detection Techniques Willem M. G. M. van Loon, Femke G. Wijnker, Marcel E. Verwoerd, and Joop L. M. Hermens*

Research Institute of Toxicology (RITOX), Utrecht University, P.O. Box 80.176, 3508 TD Utrecht, The Netherlands

A highly sensitive and quantitative group parameter to determine total molar concentrations of organic micropollutants that can bioaccumulate in the lipid phase of aquatic organisms from effluents, surface water, and drinking water has been developed. C18 empore disk was used as a surrogate lipid phase. The partition process between water and C18 empore disk was employed to simulate the bioaccumulation process. After partition extraction of the water sample, the empore disk was extracted with cyclohexane, and total molar concentrations were determined in these extracts using vapor pressure osmometry (VPO) and gas chromatography/ mass spectrometry (GC/MS), respectively. Total molar concentrations bioaccumulated in aquatic biota were estimated from the cyclohexane concentrations. Good accuracy for the total molar determination was obtained using VPO, due to the practically constant molar response factors (43.1 ( 1.7 V/M) for a wide compound range and to excellent additivity of individual compound responses. Satisfying reproducibility (0-8.3%) of VPO was obtained for sample extracts. The detection limit of VPO in cyclohexane extracts corresponded to 0.60 mM in the lipid phase of aquatic biota. A minimal separation GC/ MS system was developed, which enabled highly sensitive and sufficiently accurate total molar determinations. The reproducibility of the GC/MS determination for samples ranged from 0.7 to 22%. The detection limit of GC/MS in cyclohexane extracts corresponded to 0.044 mM in the lipid phase. The determined total molar concentrations in the lipid phase of aquatic biota were in the range of 0.139-168 mM for effluents, 0.26-1.34 mM for surface water systems, and 99%), methanol (J. T. Baker, >99.8%), and acetone (technical grade). All inorganic compounds used were obtained from Merck (purity >99.5%). All organic compounds used (see Table 1) were obtained from Fluka, Riedel de Haen, Aldrich and Janssen Chimica (purity in all cases >97%, in most cases g99%). Suwannee River fulvic acid and humic acid standard materials were purchased from the International Humic Substances Society (IHSS; Golden, CO). Empore disks (C18 solid phase, 47 mm diameter, 90% C18, 10% Teflon fibers; carbon content of C18, 17%) were obtained from J. T. Baker. Home-made (RITOX) disk holders (see Figure 1) were used. They were constructed from stainless steel, and thin (0.6 mm) stainless steel wires were used for the grid. Due to internal disk holder dimensions (14.0 mm i.d.; internal width, 5-6 mm), the disk can move slightly in the holder, and consequently, the grid does not cover a part of the disk surface. Magnetic stirrers (Janke & Kunkel, Staufen, FRG) and homemade (RITOX) stirring tables, glass test tubes (4 mL) with glass stoppers, and calibrated glass vials (Alltech, Catalog No. 95010) of 1 mL were used. Instrumentation. Vapor pressure osmometry was performed on a Gonotec (Berlin, FRG) Model Osmomat 070 cell unit and a (25) Ballesteros, M.; Coronel, C.; Galera-Go´mez, P. A. J. Colloid Interface Sci. 1991, 145, 113-118. (26) Burge, D. E. J. Appl. Polym. Sci. 1979, 24, 293-299. (27) Morris, C. E. M. J. Polym. Sci. 1976, 55, 11-16. (28) Morris, C. E. M. J. Appl. Polym. Sci. 1977, 21, 435-448. (29) Myers, M. E.; Swarin, S. J.; Nellis, B. L. Anal. Chem. 1979, 51, 18831885.

Analytical Chemistry, Vol. 68, No. 17, September 1, 1996

2917

Table 1. Molar Response Factors of Nonionic Organic Compounds As Determined by Vapor Pressure Osmometry (VPO) and GC/MSa no.

compound

MWb

bpb (°C)

VPO MRFc (V/M)

GC/MS RMRFd

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42

aniline benzene cyclohexane chlorobenzene 1-chloro-3-nitrobenzene 4-chlorotoluene 2,4-dichloroaniline 1,2-dichlorobenzene 1,3-dichlorobenzene 1,4-dichlorobenzene 1,2-dichloro-4-nitrobenzene 3,4-dichlorotoluene dieldrin 1,4-dimethylbenzene dodecane fluoranthene hexachlorobenzene hexachloroethane γ-hexachlorocyclohexane methylparathion naphthalene octadecane octaethylene glycol monododecyl ether 1-octanol pentachlorobenzene pentachloroethane pentachlorophenoli phenanthrene phenchlorphos glycerol 2,3,4,5-tetrachloroaniline 1,2,3,4-tetrachlorobenzene 1,2,3,5-tetrachlorobenzene 1,2,4,5-tetrachlorobenzene 2,3,4,6-tetrachlorophenoli 2,4,6-trichloroaniline 1,2,3-trichlorobenzene 1,2,4-trichlorobenzene 1,3,5-trichlorobenzene 2,4,6-trichlorophenoli 2,4,5-trichlorotoluene tricosane (C23) av molar response factor no. of data (n) standard deviation (Sx) relative standard deviation (%)

93.1 78.1 84.2 112.6 157.6 126.6 162.0 147.0 147.0 147.0 192.0 161.0 381.0 106.2 170.3 202.3 284.8 236.7 290.8 263.2 128.2 254.5 538.8 130.2 250.1 202.3 266.4 178.2 321.6 92.1 230.9 215.9 215.9 215.9 231.9 196.5 181.5 181.5 181.5 197.5 195.5 324.6

185 80.1 84.2 132 235.5 162 245 180.5 173 132 255.5 196.5

(16.3e) -0.0004 0 9.49 41.7 24.7 43.4 30.6 31.2 33.5 43.8 41.5 43.2 15.8 42.1 45.1 44.4 39.1 43.2 43.9 42.4 41.6 40.0 35.7 44.0 27.9 44.1 45.1 42.8 (0.0h) 39.6 44.6 42.6 44.1 39.9 43.1 44.7 40.2 41.8 44.3 46.2 45.2 39.4j 38 8.2 21

ndf nd nd 0.56 0.44 0.76 0.71 nd nd 0.93 0.44 0.89 1.62 0.76 1.04 1.28 1.54 0.89 1.23 0.55 0.79 1.8 0.00g 0.60 1.34 0.75 nd 1.18 1.30 nd nd 1.00 nd 0.99 0.53 0.85 0.90 nd 0.92 nd 0.94 2.36 1.00 30 0.43 43

138.5 216 344 322 186 323 340 218 317 194.5 277 162 309.5 340 356 290 295 254 246 244.5 310 262 218.5 213.5 208 246 229 380

a For VPO, absolute molar response factors (MRFs, V/M) are given. For GC/MS, relative molar response factors (RMRFs) as compared to 1,2,3,4-tetrachlorobenzene are given. b Data obtained from Handbook of Chemistry and Physics, 61st ed.; CRC Press: Boca Raton, FL, 1980-1981. c Molar response factor. d Relative molar response factor; see Experimental Section. e Aniline does not dissolve completely in cyclohexane. f Not determined. g Octaethylene glycol monododecyl ether does not elute from the GC column. h Glycerol does not dissolve in cyclohexane. i Chlorophenols may be partly ionized, dependent on the pH. j For total molar calculations, the average molar response factor for the boiling point range >200 °C, which is 43.1 ( 1.7 V/M, is used.

Model Osmomat 070/090 control unit-B. Osmometric signals were recorded on a flat-bed recorder. Gas chromatography/mass spectrometry was performed using a Carlo Erba Instruments (Milano, Italy) system consisting of a Model OC516 on-column control unit and a Model MFC500 gas chromatograph connected to a Model QMD-1000 quadrupole mass spectrometer. Sampling. Five grab samples (2 × 10 L) from effluents, originating from a steel industry, two chemical industries, and a paper mill, were provided by the Institute for Inland Water Management and Waste Water Treatment (RIZA) in March and April 1995. Grab samples (2 × 10 L) of surface water systems were collected from the Rhine River at Lobith (sampling date, April 25, 1995), from the Meuse River at Luik (April 25, 1995), from Lake Markermeer (April 5, 1995), and from the North Sea at Scheveningen (April 5, 1995). Drinking water was kindly 2918 Analytical Chemistry, Vol. 68, No. 17, September 1, 1996

supplied by the water works of the city of Amsterdam (GWA; June 12, 1995). Sewage was collected from the sewage works of the city of Utrecht. Water samples (10 L) were conserved by the addition of 5 mL of a 1 mg/mL silver nitrate solution immediately after sampling and were processed immediately at the laboratory. Duplicate samples were stored at 4 °C in the dark and were used to refresh the water samples after 7 days. The water samples were characterized by determining the pH and the DOC concentration. Cleaning Procedures. All glassware was thoroughly cleaned by vortexing with methanol, acetone, and cyclohexane. The stainless steel disk holder and the attached metal wire were thoroughly cleaned by ultrasonic extractions in methanol, acetone, and cyclohexane. A metal spatula for all disk manipulations was kept ultraclean by storage in cyclohexane and ultrasonic cleaning.

Empore disks (47 mm diameter) were thoroughly cleaned prior to use with the following procedure. First, a disk was soaked with ∼10 mL of cyclohexane in a 5 cm diameter closed glass dish for 7 days. Every day, the dish was gently shaken for 60 min, and the cyclohexane was refreshed thereafter. Second, the precleaned disk was placed on a 1 L filtration flask (Millipore, Catalog No. XX1504700), and the disk was eluted with cyclohexane and methanol as follows. Succesively, 10 aliquots of 5 mL of cyclohexane were put onto the filter, the disk was allowed to soak for 5 min, and the aliquot was drawn through the disk by vacuum as slowly as possible. The disk was then dried with vacuum for 10 min. Four aliquots of 5 mL of methanol were used for washing as described above, and the filter was dried for 15 min. Finally, the disk was washed with one aliquot of cyclohexane, and the disk was dried for 30 min to remove all residual solvent. Partition Extraction Procedure. The pH of unfiltrated and conserved 10 L effluent samples was adjusted to 7.5; the pH of surface water samples was not adjusted. Empore disks of 13 mm diameter (40 mg) were cut out of a freshly cleaned part of a 47 mm diameter disk using an ultraclean cutting device. A 13 mm diameter blank disk was codetermined for every 47 mm diameter disk used. The 13 mm diameter C18 empore disk was inserted into the disk holder, the disk holder was attached via a stainless steel wire to the screw cap of the sample bottle, and the disk holder was inserted immediately into the water sample. The partition extraction was performed while stirring the sample at 300 rpm, at a room temperature of 22 °C,3 and under dark conditions to prevent algal growth. After 7 days, the 10 L sample was replaced with a fresh 10 L sample. After 14 days, the empore disk was removed from the disk holder and placed on a tissue. Adhering particulate matter was carefully removed with a tissue, and the moist empore disk was inserted immediately in a 4 mL test tube with 2 mL of cyclohexane, which was stored at ∼6 °C. The extraction time was 1 day; the test tube was shaken approximately every hour manually during daytime. After 1 day, the disk was removed from the extract, and the latter was frozen at -20 °C prior to analysis. Before analysis, a total molar determination of the extract using GC/MS was performed to determine if it could be analyzed with VPO without further concentration (see GC/MS section). In that case, the internal standard, 2,4,5-trichlorotoluene (20.42 nmol in 5 µL of cyclohexane), was added to the extract before analysis. If a concentration step was necessary, the 2 mL extract was evaporated under nitrogen as carefully as possible to ∼200 µL, and then 2,4,5-trichlorotoluene (2.063 nmol in 5 µL of cyclohexane) was added. The final extract volume was determined by weighing (Fcyclohexane ) 0.7785 kg/L), and ∼100 µL of the extract was put in a syringe for VPO determination immediately. The remaining 100 µL was used for GC/MS determination. Safety Considerations. The large 10 L glass bottles have to be put in a stable position on the magnetic stirrers using appropriate stirring tables. The bottles are relatively fragile and have to be handled with care. The temperature of chilled water samples has to reach room temperature (∼22 °C) before the bottle is closed, in order to avoid glass breakage. Characterization of Partition Extraction Procedure. The effect of the addition of silver nitrate to the water samples was investigated by duplicate determination of sewage effluent with and without silver nitrate. The purity of cyclohexane was determined by GC/MS. The purity of C18 empore disk was

checked for every 47 mm diameter disk by extracting a blank 13 mm diameter (40 mg) disk with 2 mL of cyclohexane, followed by evaporation to 200 µL, addition of the internal standard (2.063 nmol of 2,4,5-trichlorotoluene in 5 µL of cyclohexane), and GC/ MS analysis. The recoveries of the direct disk extraction with cyclohexane were investigated by extracting 10 L of ultrapure water spiked with 1,4-dichlorobenzene (300 µg/L), 1,2,3-trichlorobenzene (60 µg/L), 1,2,3,4-tetrachlorobenzene (30 µg/L), and pentachlorobenzene (9 µg/L) with a 13 mm diameter C18 empore disk for 14 days. The disk was extracted two times with 2 mL of cyclohexane for 1 day, and finally the disk was extracted ultrasonically with 2 mL of cyclohexane for 30 min. The 2 mL extracts were diluted to 10 mL and spiked with the internal standard, hexachlorobenzene (515 ng). The four chlorobenzenes and the internal standard were determined using GC-ECD, and the cumulative percentages of the chlorobenzenes in the three extracts were calculated. The recoveries of the evaporation procedure were determined using the chlorobenzene mixture in cyclohexane obtained as described above. Three aliquots of 2 mL were evaporated to ∼200 µL as carefully as possible, and these concentrates were diluted back immediately to 2 mL. The four chlorobenzenes and the internal standard were determined using GC-ECD as described above, and the recoveries of the four chlorobenzenes in the evaporation procedure were calculated. An apparent distribution coefficient (Kd) of a nonylphenolpoly(ethylene glycol) mixture (10 mg/L; 10 L) was determined by a standard partition extraction, followed by disk extraction with 2 × 2 mL of methanol. The equilibrium concentrations in water and methanol were determined using UV spectrometry at 254 nm. The C18 empore disk volume was calculated as follows:3 from the weight of the disk (40 mg) and the carbon content (17%), the amount of chemically bound octadecane was calculated (7.168 mg); assuming that the density of silica-bound octadecane is equal to that of liquid octadecane (0.78 g/mL), the volume of the C18 liquid phase was calculated (Vdisk ) 7.168/0.78 ) 9.19 µL). The concentration on the empore disk was calculated from the molar amount in methanol (Am) and Vdisk (Cdisk ) Am/Vdisk), and the Kd was calculated as Cdisk/Ca. Total Molar Determinations Using VPO. The operating temperature of the osmometer cell was set at 37 °C. The signal stabilization time was set at 8 min. Two syringes were filled with cyclohexane and used to set the zero point of the osmometer. One syringe was used for cyclohexane extracts. First, the VPO zero point was set by the repeated determination of pure cyclohexane and resetting the VPO signal until stability (0.000 ( 0.001 V). Second, the sample extract was introduced (sample volume, g100 µL), and again the zero point was set by the repeated determination of pure cyclohexane until stability (0.000 ( 0.001 V). Finally, the sample was measured (n ) 4); the first measurement was omitted, and the average of the last three determinations was calculated. The total molar concentration in the cyclohexane extract (Cextr) was calculated using eq 2 (S is the VPO signal (V); 44.1 is the average molar response factor):

Cextr ) (S/44.1) × 1000

(mM)

(2)

The total molar concentration of organic compounds on the C18 empore disk (Cdisk) was calculated using eq 3 (Vextr is the volume Analytical Chemistry, Vol. 68, No. 17, September 1, 1996

2919

of cyclohexane extract, 2000 µL; Vdisk is the calculated volume of the C18 phase, 9.19 µL):3

Cdisk ) (CextrVextr)/Vdisk

(mM)

(3)

The total molar concentration of organic micropollutants in the lipid phase of aquatic biota was estimated using eq 4 (CL is the lipid-normalized concentration):3

CL ) 0.198Cdisk

(mM)

(4)

Characterization of VPO. The effect of the cell temperature on the VPO sensitivity was determined using a 10 mM 1,2,3,4tetrachlorobenzene standard at cell temperatures of 30-75 °C (interval, 5 °C). The stability of the osmometer over a period of ∼50 days was determined from the VPO signal of a 10 mM 1,2,3,4tetrachlorobenzene standard at days 1-5 (n ) 5) and at days 4146 (n ) 5). Molar response factors for VPO were determined for 38 compounds (see Table 1). The average molar response factor for relatively nonvolatile compounds (bp > 200 °C) was calculated. The repeatability of VPO was derived from the determinations of molar response factors and samples. The osmometric blank signal, its corresponding 95% confidence interval, and the corresponding instrumental detection limit were determined for a cyclohexane disk extraction volume of 2 mL, which was concentrated to 200 µL. The detection limit of the total molar concentration in biota was calculated from the instrumental detection limit. Total Molar Determinations Using GC/MS. GC conditions used were as follows: on-column injection, 5 µL; secondary cooling time, 10 s; retention gap, deactived fused silica (J&W); 5 m length × 0.32 mm i.d.; connected with a glass press fit union to a column (J&W), 5 m length × 0.32 mm i.d.; film thickness, 0.1 µm; carrier gas, helium; linear gas velocity (determined with 0.1 µL of cyclohexane at 100 °C), 80 cm/s; inlet pressure, ∼5-10 kPa; GC/ MS interface temperature, 250 °C; temperature program, 40 °C (∼2 min; from end of solvent peak) to 290 °C (0 min) at 30 °C/ min. MS conditions used were as follows: ion source temperature, 175 °C; ionization mode, EI; ionization current, 150 µA; electron energy, 70 eV; scan range, m/z 34-500 (full-scan mode); cycle time, 0.5 s (0.05 s interscan delay); solvent delay, ∼1.5 min (until the cyclohexane solvent peak has passed; the solvent peak was monitored conveniently using the ion source pressure). The MS was tuned using perfluorotributylamine and the masses m/z 69 (relative intensity, 100), 264 (8-9), 502 (2), and 614 (0.5). First, 2,4,5-trichlorotoluene (10 ng) was injected in triplicate to test the sensitivity and its variation of the GC/MS system and to determine the percentage of the total ion current (%TIC) of the fragment ions m/z 194, 196, 159, and 161. Second, the sample was determined in duplicate. The total signal of the sample was determined by integrating the TIC area above the signal treshhold, starting from the last cyclohexane contaminant (at ∼1.5 min) to the last eluted compound; the internal standard area was substracted from this signal. For surface water samples, the blank signal (residual cyclohexane and air background) was determined by injecting 5 µL of cyclohexane and integrating the TIC in the same retention time range as for the sample; this blank signal was subtracted from the sample signal. For effluent extracts, the blank signals were relatively low and, therefore, were not determined. If the internal standard was separated from the 2920

Analytical Chemistry, Vol. 68, No. 17, September 1, 1996

sample compounds, it was integrated directly. However, if the internal standard coeluted, then a specific m/z value (e.g., m/z 194, 196, 159, or 161) was integrated, and the TIC was reconstructed using the %TIC of the ion used. The total molar amount in the cyclohexane extract was calculated using eq 5 (Aextr is the total molar amount in the extract; Stot is the integrated area of the total sample TIC; Sis is the integrated area the internal standard TIC; Sis,corr is the corrected internal standard area (Sis,corr ) Sis × 1.064); Ais is the added amount of internal standard):

Aextr ) [(Stot - Sis)/Sis,corr]Ais

(mol)

(5)

The cyclohexane concentration was then calculated using eq 6:

Cextr ) (Aextr/Vextr) × 1000

(mM)

(6)

The concentrations in the C18 empore disk and the internal concentration in the lipid phase of biota were then calculated using eqs 3 and 4, respectively. Characterization of GC/MS. The transmission of the minimal separation GC system for nonionic surfactants was investigated using octaethylene glycol monododecyl ether (injected amount, ∼50 ng) and a nonylphenol-poly(ethylene glycol) mixture (injected amount, ∼50 ng). Relative molar response factors (RMRFs) of 30 compounds (see Table 1) were determined. The RMRF is defined in eq 7 (Si is the TIC area of compound i; Ai is the molar amount (∼0.4 nmol) injected; Stcb is the TIC area of 1,2,3,4-tetrachlorobenzene, which is added as the internal standard; and Atcb is the injected molar amount (∼0.4 nmol) of tetrachlorobenzene):

RMRF ) (Si/Ai)/(Stcb/Atcb)

(7)

2,4,5-Trichlorotoluene was tested for use as an internal standard. The repeatability of GC/MS was derived from the determinations of the relative molar response factors and the samples. The linear dynamic GC/MS range was determined for the internal standard 2,4,5-trichlorotoluene (injected amount, 1-500 ng). The instrumental detection limit of the GC/MS group determination was estimated from a drinking water extract (Amsterdam, The Netherlands). The detection limit of the total molar concentration in biota was calculated from the instrumental detection limit. Investigation of Potential Interferences. The possible interference of water in the VPO determination of cyclohexane extracts was estimated by vortexing 2 mL of cyclohexane with 2 mL of water for 5 min, presumably leading to a cyclohexane phase saturated with water. The cyclohexane layer was separated, concentrated under nitrogen to 200 µL, and determined by VPO. The potentail interference of inorganic salts in the VPO determination was investigated by performing a standard partition extraction of an artificial inorganic river water matrix (see below) without organic contaminants and by VPO analysis of the cyclohexane extract. In addition, the seawater results were used to evaluate the effect of salts on the VPO signal. The potential interference of reference aquatic humic and fulvic acids by partitioning onto empore disk was investigated using a water sample containing 5.4 mg/L fulvic acids and 0.6 mg/L humic acids in an artificial inorganic river water matrix (1 g of NaHCO3,

200 mg of KHCO3, 2 g of CaCl2‚2H2O, and 1.8 g of MgSO4‚7H2O in 10 L of water). The pH is of the water sample was adjusted to 7.0. A partition extraction with a 13 mm diameter empore disk was performed, and after 14 days, the disk was extracted with 5 mL of methanol/water (85/15 v/v; pH 7.0).30 The humic substances content of the extract was determined using UV spectrometry at 360 nm and was quantified using calibration curves of the separate aquatic humic and fulvic acids in methanol/water (85/15 v/v; pH 7.0). The apparent distribution coefficient, Kd, of the mixture of aquatic humic substances was calculated as Cdisk/ Ca. RESULTS AND DISCUSSION Partition Extraction Procedure. (a) Sample Pretreatment. The filtration of the water samples was omitted in order to (a) eliminate the risk of adsorption losses of hydrophobic organic compounds onto the microfilter31 and (b) prevent a timeconsuming filtration step. However, since bacteria and algae remain in the water sample, alternative conservation procedures were essential to prevent sample degradation. Silver nitrate was selected as the conserving agent, since it a strong bactericide at low concentrations (1-10 µg of Ag+/L)32 and since it is not reactive toward organic compounds. It appeared that, in the unconserved sewage samples, a strong growth of brown algae had occurred after 14 days. Little or no visual and odorous changes of the conserved sewage samples were observed during the partition extractions. The total molar signals with and without silver nitrate as determined by VPO (0.049 ( 0.005 V and 0.050 ( 0.006 V, respectively) surprisingly did not differ significantly, which suggested that the bioaccumulatable total molar concentrations in sewage were not influenced by the degradation processes observed. However, GC/MS analysis revealed the disappearance of one major peak in the extract and the appearance of another major peak, thus showing a significant change in the sample composition. Considering these results, chemical conservation of water samples with silver nitrate appeared to be effective. An additional conservation measure used was to refresh the water samples after 7 days. A final conservation measure was to perform the partition extraction under dark conditions, thus preventing algal growth and photolytic degradation. The pH of a water system is an important factor in the bioconcentration of acidic (e.g., chlorophenols)33 and basic compounds. The pH value of a surface water sample does not need adjustment, since the bioconcentration processes of organic micropollutants occur at this pH. However, the pH of effluent samples has to be adjusted to that of Dutch river water systems (on average, 7.5),34 in which bioconcentration processes will actually take place. (b) Partition Extraction System. It is of great importance for a universal group parameter for organic micropollutants that the extraction system (C18 empore disk and cyclohexane) is highly pure and does not introduce contaminants. The cyclohexane appeared to be very pure. However, at the beginning of this study, the C18 empore disks used appeared to be highly contaminated. Therefore, a thorough cleaning procedure was developed, which (30) Saleh, F. Y.; Chang, D. Y. Sci. Tot. Environ. 1987, 62, 67-74. (31) Abdel-Moati, A. R. Water Res. 1990, 24, 763-764. (32) Woodward, R. L. J. Am. Water Works Assoc. 1963, 55, 881-886. (33) Saarikoski, J.; Lindstro ¨m, R.; Tyynela¨, M.; Viluksela, M. Ecotoxicol. Environ. Saf. 1986, 11, 158-173. (34) De Groot, R., WRK, Nieuwegein, personal communication, 1995.

Table 2. Extraction and Evaporation Recoveries (%) of 1,4-Dichlorobenzene (DiCB), 1,2,3-Trichlorobenzene (TriCB), 1,2,3,4-Tetrachlorobenzene (TeCB), and Pentachlorobenzene (PeCB) from Empore Disk (40 mg) Using 2 mL of Cyclohexane and Evaporation to 200 µL DiCB cumulative recovery of extraction 1a cumulative recovery of extraction 2a cumulative recovery of extraction 3b evaporation recoveryc total recoveryd

TriCB

TeCB

PeCB

>95 ( 0.5 96.4 ( 0.2 96.8 ( 0.1 97.0 ( 0.1 99%), it gave very stable VPO signals. For example, n-hexane, which usually contains 5-15% of other hexane isomers, appeared to give less stable VPO signals than cyclohexane.35 However, it was found that cyclohexane from Merck (purity, >99.5%) gave less stable VPO results than the cyclohexane from Baker (purity, >99%). So, the suitability of a particular solvent or solvent grade for VPO determinations cannot be predicted from its purity alone and has to be determined emperically. An additional advantage of cyclohexane is that extracts can be frozen at 4 °C, giving an excellent sample conservation. The optimization of the osmometer cell temperature showed that the VPO sensitivity slowly increases with the cell temperature; the temperature dependence of the molar response factor (MRF) of 1,2,3,4-tetrachlorobenzene appeared to be MRF (V/M) ) 0.627T (°C) + 22.1. An optimum temperature range of 35-60 °C is recommended by the manufacturer. However, at higher cell temperatures (>50 °C), the Teflon plungers of the VPO syringes softened, leading to leakage. Also, at a low cell temperature of ∼37 °C, the best stability of the VPO signal of cyclohexane was obtained. Based on these considerations, a cell temperature of 37 °C was selected. The absolute molar VPO response and the 95% confidence interval of a 10 mM 1,2,3,4-tetrachlorobenzene standard was 45.2 ( 0.6 V at days 1-5 and 43.1 ( 0.6 V at days 42-46. The average molar response factors differ significantly, but the relative difference is quite small (∼5%). Subsequent experiments also showed that molar response factors for a wide range of compounds are quite constant (see next paragraph). It was concluded, therefore, that absolute VPO signals can be used without calibration. Molar response factors (MRFs) for a wide range of common organic micropollutants are shown in Table 1. The relationship between the boiling point and the molar response factor of a compound is shown in Figure 2. It appears that, for compounds with bp > 200 °C, the molar response factors are quite constant (43.1 ( 1.7 V/M, see Table 1). This average molar response value was used for the calculation of total molar concentrations in organic extracts. However, at lower boiling points, the molar response factors decrease sharply and, e.g., for benzene (bp 80.1 °C), become zero. This phenomenon can be explained from the principle of VPO: relatively volatile compounds weaken the vapor pressure temperature effect and consequently show a decreased VPO response. It appeared in a survey that we performed on 17 (35) Verbruggen, E. M. J., RITOX, Utrecht, personal communication, 1993.

2922 Analytical Chemistry, Vol. 68, No. 17, September 1, 1996

Figure 2. Relationship between the boiling point and the VPO molar response factor as determined for 38 nonionic compounds.

effluent and 15 surface water samples (data not reported here) that, in most samples, most of the extracted compounds are present in the boiling point range >200 °C, and so fairly accurate total molar concentrations are obtained. This phenomenon can partly be explained by the fact that the partition coefficients (Kd) of these more volatile compounds onto empore disk are generally much lower than those for higher boiling compounds, leading to relatively low quantitative contributions of these compounds in the biomimetic extracts. For example, the Kd on C18 empore disk of chlorobenzene (the latter shows a strongly decreased VPO response factor; MRF ) 9.5 V/M, see Table 1), is ∼60 times lower than the Kd of 1,2,3,4-tetrachlorobenzene (MRF ) 44.6 V/M, see Table 1), which may represent a compound with an average boiling point. In addition, these relatively volatile compounds may be removed from the water system more rapidly due to volatilization processes in water purification systems and during river water transport. If these relatively volatile compounds, however, are still present in significant amounts in a cyclohexane extract, an underestimation of the total molar sample concentration will occur. This systematic error cannot be quantified using VPO alone, since the exact composition of the volatile fraction (bp < 200 °C) is not known; therefore, no corrections for MRFs can be made. Alternatively, such extracts could be determined more accurately using GC/MS and volatile solvents (e.g., pentane). Using such a solvent, the volatile fraction may be determined in a sufficiently large boiling point range, and compound-specific MS molar response factors (see GC/MS Procedure, below) can be used. The excellent additivity of the VPO responses of individual compounds into a total molar signal in several test mixtures has already been reported.3 The repeatability of VPO for standard solutions of test compounds (see Table 1) was in the range of 0.1-2%. The repeatability for effluent and surface water extracts was in the range of 2.4-8.3%. If a cyclohexane extract is concentrated, a coconcentration of cyclohexane contaminants occurs, and a blank signal was observed in the VPO determinations. The blank signal and the 95% confidence interval of 2 mL of cyclohexane concentrated to 200 µL was determined to be 0.019 ( 0.006 V. The solvent blank signal should be redetermined for every new combination of solvent, solvent volume, and evaporation procedure. Since this

Table 3. Bioaccumulatable Total Molar Concentrations of Organic Micropollutants from Selected Effluent, Surface Water, and Drinking Water Samplesa sample

DOC (mg/L)

pH

signal VPO (V), (RSD) (%)

10.0

0.109 (1.4)

signal GC/MS [Stot - Sis]/Si,cor, (RSD) (%)

Cextr (mM)

Cdisk (mM)

Clipid (mM)

2.52a 3.86a