1817
Anal. Chem. 1986, 58, 1817-1822
Ultrasonic Solvent Extraction of Trihalomethanes from Granular Activated Carbon Katherine T. Alben* a n d J o a n
H.Kaczmarczyk
Wadsworth Center for Laboratories and Research, New York State Department of Health, Albany, New York 12201
Trlhakmethanes (THMs) adsorbed from aqueous sokrtbn onto granular actlvated carbon (GAC) were recovered by lowfrequency sonlcatlon In methanol for 30 min and determined by analysts of methanol extracts using packed column gas chromatography wtlh electron capture detection. Analytical recoverles averaged 94 % for four trlhalomethanes from granular activated carbon samples Immersed in aqueous 80lutions of known trlhalomethane composltbn. No slgnlflcant difference was found in recovery of trlhalomethanes from repllcafe treatment plant granular actlvated carbon samples extracted 24 h afler colledlon and extracted afler 18 months' storage in sealed containers at 5 OC. Detection limits using an electron capture detector are defined by the cutoff concentration corresponding to a postlive response on the linear cailbratlon curve. Absolute detection llmits averaged 93 pg: this amount Is equivalent to 8 pg of trlhalomethane/g of actlvated carbon, for a solvent to sample slze ratio of 0.27 L/g and for 3 pL of methanol extract analyzed. Quanttlatbn limits averaged 3-fdd greater than detectlon Ihntls. The feaslMiity of direct analysts of THMs adsorbed to GAC thus provldes a means to verify THM-GAC surface concentrattons estimated from mass balance cakulatbns, uslng concentrations of THMs In Influent and effluent water samples.
Solvent extraction of granular activated carbon (GAC) samples has been widely practiced, using either a Soxhlet apparatus or in-column elution to remove organics adsorbed from raw, potable, and waste waters (1-6). However, application of this procedure has led to acknowledgment of its limitations, in terms of less than quantitative recoveries of model compounds (7,8). Nonetheless, useful applications of GAC solvent extraction methods remain; recent examples include identification of produds of surfacecatalyzed reactions between various oxidants and adsorbed organics (9-11)and comparative analyses to profile the distribution of organics as a function of bed depth within a GAC adsorber (12,13). Our current interest in solvent extraction of GAC samples is to directly determine concentrations of organics adsorbed and/or desorbed at sequential bed depths in a GAC pilot column (14).Given a direct means of analysis, concentrations of organics on the adsorbent would not have to be deduced from discrete measurements of fluctuating concentrations in influent and effluent water samples (combined with estimates of the volume of water treated and the mass of adsorbent in the bed). In this paper, we present results for the ultrasonic methanol extraction of trihalomethanes (THMs) from GAC; analytical recoveries and precision for solvent extraction of THMs from GAC samples are compared to similar data acquired for liquid-liquid extraction of THMs from water samples. The feasibility of quantitative solvent extraction of THMs from treatment plant GAC samples is suggested by favorable recoveries of volatile organics eluted by carbon disulfide from GAC traps (15-17'). Selection of methanol as the solvent permits use of the electron capture detector for gas chroma0003-2700/86/0358-1817$01.50/0
tographic analysis, without concentration of the sample extract to enhance analytical sensitivity. Methanol also has the advantage that it is miscible with residual water trapped in the pores of GAC samples from a pilot column or treatment plant. However, the relatively high boiling point of methanol compared to the THMs precludes use of the Soxhlet apparatus. Ultrasonic solvent extraction was chosen as an alternative, since numerous reports have been made that it results in higher recoveries of organics in much less time than required for Soxhlet extraction (18-23).
EXPERIMENTAL SECTION Chromatographic Analysis. Samples are analyzed for THMs on a Perkin-Elmer Sigma 1 gas chromatograph equipped with a Model 4990 autosampler and 63Nielectron capture detector. THMs are separated on a 2-mm4.d. X 1.8-m-longcolumn packed with 80/100 mesh Chromosorb 101, operated isothermally at 165 "C, with the injector and detector blocks maintained at 225 "C and 300 "C, respectively. The carrier gas (argon-5% methane, 99.999% purity) is supplied at 11mL/min to the column and at 34 mL/min as makeup to the detector. Although our instrument configuration includes moisture traps on carrier gases, no oxygen traps are used. Typical elution times are 3.21 min for chloroform, 6.15 min for dichlorobromomethane, 9.62 min for 2-bromo-lchloro-propane (the internal standard), 11.17 min for chlorodibromomethane, and 21.17 min for bromoform; total run time is 30 min/sample. Standards are prepared in hexane or methanol, using chemicals of reagent grade purity. Data Reduction. Quantitativeanalysis is carried out following injection of a set of standards and samples; results in terms of integrated peak areas are saved in data files in the Sigma 1 memory for later access by the quantitative program, written in BASIC. The original version of the program occupied 570 words of the 6K of RAM in the Sigma 1; modifications to reference standard curvea to an internal standard, and to eliminate the effect of drift in the electron capture detector's response (24), have expanded the program to 1369 words. A copy of the quantitation program is available from the author upon written request. Coefficients of a linear standard curve are calculated for each of the trihalomethanes injected (3 rL) as mixed standards at 0.1, 0.2,0.5,1.0, and 2.5 mg/L. The program then determines THM concentrations in individual samples, after requesting as input the sample file name, the volume of solvent used for extraction, and the sample volume (for water samples) or sample weight (for GAC samples). In each sample file, a quality control check is performed on detection of the internal standard (at 10 mg/L), and deviations are flagged in the report. Sample files are transferred to cassette tape for storage. Recovery Studies. Results in this paper are based on 1 2 X 40 mesh Calgon Filtrasorb 400, to permit application of the method developed to various pilot column samples. Virgin GAC samples are spiked indirectly with THMs by immersing the GAC in a water of known THM composition and allowing adsorption to occur. A THM-spiked water (concentration Co = 500-1000 pg/L) is prepared by injecting a known amount of the THMs of interest, dissolved in methanol, through the septum of a stock bottle (1.9 L) filled with THM-free water. The THM-free water is taken from a 15-cm-0.d. X 72-cm-longfixed-bed GAC contactor. The THM concentration of the spiked water is determined analytically; this water is used t o fill sample bottles (volume V, F 0.155 L) containing a preweighed amount of GAC (dryweightDo= 100 mg). Sample bottles are kept sealed at room 0 1986 American Chemical Society
1818
ANALYTICAL
CHEMISTRY, VOL.
58, NO. 8, JULY 1986
Table I. Summary of THM Recoveries from Spiked GAC Samples concn range, mg/g 1.0-1.6 0.2-0.8 0.1-0.8 0.2-0.8
av
CHC13 102 f 88 f 77 f 102 f
17 6 17 33
92 f 18
CHBrC12
recovery, 70 CHBrzCl
18 21
111 f 4 102 f 5 77 f 18 88 & 17
94 f 16
94 f 11
100 f 104 f 81 f 93 f
16
7
temperature for 1-7 days; at the stop time, the water in a particular sample bottle is analyzed for the residual THM concentration (Cf,pg/L), and the GAC is analyzed for the adsorbed THM concentration (Qexp,mg/g). The percent recovery is calculated by normalizing the experimental value, Qerp,to the theoretical value, Qth, for THMs adis used sorbed, where Qth = 10-3(C0- Cf)Vo/Do;a factor of to convert from micrograms to milligrams. Solvent Extractions of GAC and Water Samples. Prior to extraction, wet GAC samples are suctioned briefly in a Millipore filtration apparatus fitted with a 0.45-pm regenerated cellulose membrane filter. The sample ( ~ 0 . 2 g5 wet weight) is transferred to a culture tube (27 mL), which is filled with reagent grade methanol, capped, immersed in the water bath of a Branson B220 ultrasonic cleaner, and sonicated for 30 min at a nominal frequency of 55 kHz and applied power of 125 W. Samples are left to stand at room temperature for approximately24 h until the GAC settles. Methanol (=2 mL) containing the extracted THMs is transferred to a crimp-capvial, which is placed on an autosampler for analysis. If the GAC is a synthetic sample spiked with THMs in the laboratory, the dry sample weight for quantitation is already known. However, if the GAC is a treatment plant sample, the dry weight for quantitative analysis is obtained after solvent extraction, by drying the GAC at 110 "C for 30 min. To assist in determining recoveries of GAC-solvent extraction 7 spiked with THMs are also procedures, water samples ( ~ 2 mL) solvent extracted using reagent grade hexane (2 mL) and following procedures described elsewhere (25). The hexane layer with extracted THMs is recovered after phase separation from water and handled in the same manner as methanol-THM extracts. RESULTS AND DISCUSSION Analytical Recoveries of Solvent-Extracted T H M s from Spiked GAC and Water Samples. Ultrasonic solvent extraction of the THMs from GAC samples is concluded to result in satisfactory recoveries, as indicated by the results summarized in Table I. Recovery data were acquired from four separate experiments, one at a high concentration range and three a t a low concentration range. Of these experiments, the first two involved spiking a stock bottle, then using this water to fill a set of smaller bottles containing the preweighed GAC. For the last two experiments, the smaller bottles containing the GAC were spiked directly, thus avoiding transfer of spiked water: contrary to our expectations, the precision and accuracy of these experiments were generally worse. The overall recovery of THMs from the combined experiments was 94 & 14%. In retrospect, the concentration range investigated, approximately 0.10-1.60 mg/g of GAC, is appropriate primarily for GAC adsorption of CHC1, from potable water: isotherms obtained in our laboratory for CHC13 adsorption on Calgon F400 indicate these surface concentrations correspond to solution-phase concentrations of 0.8-43 kg/L (14). Moreover, for pilot column GAC samples and influent concentrations of 16.5 i 5.3 wg of CHCl,/L, surface concentrations determined by ultrasonic extraction ranged from 0.05 to 1.09 mg/g (14). For CHBrCl,, maximum GAC surface concentrations found on potable water treatment-GAC samples exceed 0.1 mg/g; concentrations of CHBr,Cl and CHBr3 are lower (14). T o be representative of environmental GAC samples containing the bromine-substituted THMs, these recovery studies
CHBr3
THMs (av)
N
19
104 f 11 101 f 6 78 f 18 93 f 22
18 18 18 18
* 13
94 f 14
4
104 f 109 f 76 f 88 f 94
5 7 21
CHBrCl2 I)
I
It
, 8 I
I
I
I
1
I
CONTACT TIME (hrs) Figure 1. GAC surface concentrations reached at time t , relative to the average surface concentration from data obtained at t L 24 h. Results are based on experiments 1-4 described in Table I: to avoid compressing the time scale, results from experiment 1 at 168 h arre not shown, although relative surface concentrations remain at 1.O.
should therefore be extended down to 0.01 mg/g. One might also question what the recoveries obtained represent with respect to the GAC structure and its pore size distribution; a typical batch of Calgon F400 may have as much as 65% of its total volume (0.94 cm3/g) associated with micropores ranging in diameter from l to 1000 8, (26-28). As described in the Experimental Section, recoveries for this paper were determined at different solution-phase-adsorbent-phase concentrations by stopping individual bottles a t the same time, in groups of two or three: stop times ranged from 4 to 168 h for the first experiment in Table I, but from 0.5 to 54 h for the remaining three experiments. It is possible that recoveries may deteriorate with increasing time allowed for diffusion to micropores, which are generally regarded to have a higher energy of adsorption than macropores (28). As shown in Figure 1, adsorption of the THMs from water is essentially complete a t 24 h, possibly earlier because of a gap in data collection between 6 and 24 h: based on Figure 1, recovery data collected from 24 h on is assumed to describe extraction of the THMs distributed at equilibrium within the GAC pore structure. In further support of this interpretation, a plot of recoveries vs. contact time in Figure 2 indicates below average recoveries were obtained at 0.5,25, and 48 h, whereas above average recoveries were obtained a t relatively long contact times, e.g., 54, 76, and 168 h. These data do not indicate a continual decrease in recovery associated with in-
ANALYTICAL CHEMISTRY, VOL. 58, NO. 8, JULY 1986
1819
Table 11. Summary of THM Recoveries from Spiked Water Samples CHC13
CHBrClz
recovery. % CHBr2Cl
CHBr3
THMs
N
77 f 11
79 f 17
81 f 15
82 f 18
80 f 15
23
total
70 f 10 30 f 6 100 f 7
70 f 7 27 & 6 97 f 5
72 f 11 29 f 7 101 f 5
72 f 11 29 f 7 101 f 5
71 f 10 29 f 6 100i6
18
25
(a) 4 mL of hexane (b) 2 mL of hexane total
88 f 6 19 f 5 107 f 5
87 f 6 19 f 5 106 f 4
95 f 8 20 f 6 115 f 4
94 f 8 20 f 6 114 f 5
91 f 7 20 i 6 110 f 5
18
10 10 5 25 20
1 mL 1 mL 1 mL 1 mL 1 mL
82 f 2 78 f 2 93 82 71 f 2
88 f 2 85 f 2 92 86 73 f 2
91 f 2 88 f 4 97 90 78 f 3
95 f 2 90 f 2
89 f 2 85 i 5 94 f 3 86 f 4 77 f 7
range, rglL
sample size, mL of HzO
15-400
25
2 mL of hexane
100-400
25
(a) 2 mL of hexane (b) 2 mL of hexane
concn
100-400
0.1-200 0.1-200 2-34 2-34 19-90
solvent
pentane isooctane methylcyclohexane methylcyclohexane pentane
__
......................- ._ -.- -..- - __. - -. .
I
6o
of of of of of
t
I, 0
I
I
I
50
I
100
I
I
1
150
CONTACT TIME (hrs) Figure 2. Recoveries of THMs in ultrasonic solvent extracts as a function of contact time, t , albwed for THM adsorption: (- -) CHCI,, (-) CHBrCI2, (---) CHBr2CI (---) CHBr3. The Solid double 1- (=) indicates the average THM recovery, R; upper and bwer dashed double llnes (= = =) indicate warning limits based on twice the standard deviation (f2a) from the average.
-
creasing contact time up to 1 week. Therefore, ultrasonic solvent extraction of THMs from GAC does not appear to depend on their adsorption in micropores vs. macropores.
Criteria for Determining Recoveries of Ultrasonic Methanol Extraction of THMs from GAC. Criteria developed during this study as the basis for recovery data should also be mentioned. The f i t criterion was to quantitate GAC and water samples within the range of linearity defined by THM standards. The second criterion was to account for incomplete recovery of THMs extracted from water samples when determining Qth,the theoretical THM concentration of a spiked GAC sample (cf. Experimental Section). With regard to the fiit criterion, quantitative results were generally obtained by regression analysis of THM standards from 0.1 to 5.0 mg/L. Thus, it was possible to analyze for THMs in water samples (25mL extracted in 2 mL of hexane) from 8 to 400 MIL, and in GAC samples (100 mg extracted in 27 mL of methanol) from 0.027 to 1.35 mg/g. For CHBrCl,, however, saturation of the electron capture detector limits its range of linearity to 2.5 mg/L maximum (3pL injected): for this study the CHBrCl, concentrations of some samples (-5%) were determined from nonlinear calibration curves; additional experience has since been acquired using different sample size to solvent volume ratios for extraction of samples known to
86 f 4
ref
31 31 32 32 30
have a high concentration. From Table I it is also apparent that the first recovery experiment, run at concentrations from 1.0 to 1.6 mg/g of GAC, involved quantitation of some samples beyond the upper limit of quantitation defined by the standards. This problem was avoided in later experiments. An additional problem encountered was incomplete recovery of THMs from water samples, which particularly affected determination of Co, the initial concentration of the spiked water used to load THMs onto the GAC: an accurate value of Co is essential to calculate Qth (cf. Experimental Section). In general it was noted that if Co was determined experimentally,values were obtained that averaged 68 f 10% of those expected. Thus, experiments were performed to determine analytical recoveries for solvent extraction of the THMs from water a t concentrations from 15 to, 400 pg/L; results correspond to the first experiment listed in Table 11. The overall recovery of the four THMs in a single solvent extraction from spiked water was 80 f 15%; this value is comparableto an average recovery of 77 f 7% ,calculated from Glaze and Lin's results on liquid-liquid extraction of THMs from water (29). In subsequent experiments, over a 100-400 pg/L concentration range, it was found that a two-step extraction procedure results in quantitative recoveries (cf. Table II). Alternatively increasing the solvent to water volume ratio results in a high recovery on the fiit extraction, but introduces other problems: a larger multiplication factor is used to convert from the THM concentration in the extract to the concentration in the original water sample (e.g., 174 vs. 80); the lower limit for quantitation increases (in this case for 0.1 mg/L as the lowest standard, from 8 to 17.4pg/L). However, a main conclusion from these various experiments is that losses of THMs by volatilization from spiked water samples have had a negligible effect on analytical recoveries. In effect, recovery of THMs from water by solvent extraction appears to be limited by their partitioning between water and hexane. Hexane:water partition coefficients can be calculated from our THM recovery data in Table I1 for a single-step hexane extraction (15-400pg/L). Results in Table I11 show that these partition coefficients are qualitatively consistent with values published in the literature, but may be biased low. According to Barbari et al. (36), partition coefficientsshould increase with decreasing solvent molecular weight. Our data for hexane do not fit between the data reported for undecane and pentane. It remains of interest if in our extractions we are achieving ideal partitioning: for this study samples were mixed with little or no headspace; other authors contend volatilization losses are negligible once solvent is added, and that a headspace improves mixing of the solvent when shaken with water (6, 37). In summary, if values for Q6 are based on analyses of water samples following a single solvent extraction procedure, then
1820
ANALYTICAL CHEMISTRY, VOL. 58, NO. 8, JULY 1988
Table 111. Partition Coefficients
solvent
volume ratio (water:solvent)
pentane pentane pentane
hexane methylcyclohexane isooctane octanol octanol octanol undecane
10 20 20 12.5 25 10
CHC13
CHBrClz
CHBr2Cl
CHBr,
46 49 44 42 113 35 79.4 79.4
81 54 60 47 134 57
73 71 63 53 225
123 75 57
ref 31 30 30 32 31 33 34 35 36
55.3
Table IV. Analytical Precision for Solvent-Extracted GAC and Water Samples and for THM Standards
concn
sample
CHC13
0.1 mg/L = 8 pg/L = 0.027 mg/g 0.2 mg/L = 16 pg/L = 0.054 mg/g 0.5 mg/L = 40 pg/L = 0.135 mg/g 1.0 mg/L = 80 pg/L = 0.27 mg/g 2.5 mg/L = 200 pg/L = 0.675 mg/g 15-400 pg/L 6-46 pg/L 0.2-1.5 mg/g 0.2-0.8 mg/g 0.05-0.3 mg/g
standard standard standard standard standard spiked water finished drinking water spiked GAC (batch) spiked GAC (individual) pilot column GAC
7.2 4.5 2.6 1.7 0.3 11.7 11.0 2.5 8.8 9.0
relative mecision, % CHBrClz CHBr2Cl CHBr3
results have to be corrected for incomplete recovery of the THMs. For this paper recovery factors were used to correct C,, and C,, and thus to calculate values for Qth: in this context recoveries of THMs by ultrasonic solvent extraction of GAC samples were found to be quantitative.
Recovery of THMs from GAC Samples following Long-Term Storage. In general, field samples of GAC are extracted in our laboratory within 24 h after collection. However, it was of interest to determine if significant changes (e.g., surface reequilibrations, diffusion to micropores, biodegradation) take place when samples are stored. Figure 3 presents two sets of data for the distribution of THMs on a GAC pilot column that had been in a potable water treatment plant: the second set of resdta was obtained nearly 18 months after the first, during which time the GAC samples had been stored wet a t 5 "C, in glass bottles sealed with Teflon-lined screw caps. Application of Student's t test (38)to paired data points (N = 15) indicated the difference in the two sets of data was not significant, It1 = 1.774 < 2.145, with a probability of 195% (14 df).
Analytical Precision and Corresponding Detection Limits. Results are given in Table IV for the precision of THM analyses based on solvent extracts of water and GAC samples. In each case the precision is given as the pooled standard deviation from the average of N paired samples (38). In the case of water samples, precision for spiked and actual water samples is comparable and is in agreement with values reported in the literature (25, 29). Precision for ultrasonic solvent extracts of GAC samples approaches instrumental precision for analysis of THM standards; as noted previously, precision is significantly better for batch-spiked GAC samples than for individually spiked samples. However, this level of precision (1.8% average) is not obtained for pilot column samples: in this case, the precision (9.0%) is considered to reflect the difficulty of obtaining identical 1 W m g GAC samples from a given bed location when approximately 1-2 g is taken at a time (for real-time sampling; 10-20 g for postrun column sectioning-cf. ref 13 and 14). In Table V, instrumental detection limits, based on analytical precision for analysis of standards, are compared to
THMS
N 8 8 8 8 8 74 146 30 36
9.3 3.7 2.4 2.1 0.5 11.3
7.2 2.9 1.7 1.8 0.4 12.7
4.3 2.7 1.9 1.3 0.2 13.5
7.0 3.4 2.2 1.7 0.4 12.3
1.8 12.4
2.0 13.8
2.0 13.0
1.8 14.3
1-
0,6
I
m m E
\
-
f
0.3-
U
0.0
0
I
I
'
I
"
I
'
I
"
'
8
'
'
I
(
16
BED SECTION
Flgure 3. Comparison of CHCI:, concentration in GAC samples extracted within 24 h of sample coUection (0)and 18 months after sample collection (0). Samples are from bed sections spaced approximately 5 cm apart. Flow of water was in the direction of increasing bed section number: the GAC column had been in operation for 60 weeks.
method detection limits (29,39),calculated from the analytical precision for solvent extraction of water and GAC samples. In both cases the detection limit, CDL,is given by
for n measurementsof a signal N above background N b , where s is the standard deviation at zero concentration and t is the Student factor at n -1 for a particular confidence level. For instrumental detection limits s is the analytical precision for analysis of standards; for method detection limits, s is the precision based on analysis of solvent-extracted water and GAC samples. For chromatographic analyses, determination of s is complicated by the fact that integration systems set up for normal data acquisition do not give a measure of true background N b at the zero concentration limit. Integration systems detect peaks based on changes in slope above a
ANALYTICAL CHEMISTRY, VOL. 58, NO. 8, JULY 1986
1821
Table V. Comparison of Quantitation Limits, Instrumental Detection Limits, and Method Detection Limits for Solvent Extracts of GAC and Water Samples CHClS quantitation limit mg/L pg/L equiv mg/g equiv instrumental limit" mg/L pg/L equiv mg/g equiv method detection limita pg/L, spiked water mg/g, spiked GAC
CHBrClz
0.1
CHBrzCl
CHBr3
THMs
0.1 8.0 0.027
0.1 8.0 0.027
0.1 8.0 0.027
0.1
0.012 (0.043) 0.96 (3.4) 0.003 (0.012)
0.009 (0.056) 0.72 (4.5) 0.002 (0.015)
0.008 (0.015) 1.2 0.004
0.010 (0.031)
0.80 0.003 0.79 0.005
0.76 0.007
0.85 0.008
0.91 0.008
0.83 0.007
8.0 0.027
0.010
N
8.0 0.027 8
0.92 (2.5) 0.003 (0.008) 74 36
"Based on analytical precision of standards (instrumental limit) and samples (method limit), (cf. ref 27 and 36). However, for CHBrCl,, CHBr,Cl, and CHBr3 the instrumental limit is actually defined by the cutoff concentration (given in parentheses) on the linear standard curve for an instrumental response above zero, rather than by analytical precision.
Table VI. Characteristics of Average THM Calibration Curves Based on External Standards (Absolute, N = 26) and Internal Standards (Relative, N = 8), with Experimental Uncertainties Given as 1 Standard Deviation, for Data Acquired over 2 Years CHC13 range mg/L ng regression coeff (absolute) intercept, a. (cts) slope a, (cts/(mg/L)) corr coeff regression coeff (rel) intercept, a. (-) slope, al (l/(mg/L)) corr coeff
CHBrClz
CHBrzCl
CHBr3
0.1-2.5 0.3-7.5
0.1-2.5 0.3-7.5
0.1-2.5 0.3-7.5
0.1-2.5 0.3-7.5
2.1 f 1.0 21.1 f 4.3 0.998 f 0.002
-4.5 f 3.6 115 f 28 0.9993 f 0.0008
-3.2 f 3.0 85.4 f 18.5 0.9996 f 0.0006
-0.1 f 0.7
0.04 f 0.02 0.42 f 0.02 0.9996 f 0.0007
-0.10 f 0.05
-0.08 f 0.04 1.72 f 0.07 0.9998 f 0.0002
-0.009 f 0.004 0.61 f 0.02 0.9999 f 0.0001
threshold set by the analyst. Typically all data at background level are rejected, and integrated peak areas are given only for signal above background. One approach is to obtain s from a linear regression of precision on concentration: s is then determined from the intercept at zero concentration. Glaser et al. recommend a linear regression of relative precision on the inverse of concentration, which leads to an estimate of s from the slope (39). In evaluating our own data, we did find the latter method to give consistently high correlation coefficients (averaging 0.97 f 0.01 for all four THMs), which were not characteristic of direct linear regressions of precision on concentration. By use of criteria established by Glaser et al., detection limits in Table V are based on 99% confidence limits. In general, method detection limits for ultrasonic solvent extraction of GAC samples (0.007 mg/g average) are somewhat greater than GAC method detection limits calculated from instrumental precision (0.003 mg/g average). Method detection limits for solvent extracts of water samples (0.83 pg/L average) are comparable to those based on instrumental precision (0.92 pg/L). However, for CHBrC12,CHBr2Cl,and CHBr3 the actual detection limit is determined by the cutoff concentration for a positive detector response and the fact that their standard curves have a negative intercept (Table VI). Therefore, average THM detection limits are 2.5 pg/L and 0.008 mg/g for solvent-extractedwater and GAC samples, respectively. We note that, on the average, these detection limits are a factor of 3 below our typical quantitation limits for water and GAC samples. Moreover, the amount analyzed at the detection limit is the same for both types of samples, equivalent to injection of a 0.031 mg/L standard, or for 3 pL, equivalent to 9 3 pg. From this it is apparent that better THM-GAC detection limits (than 0.008 mg/g) could be obtained with less dilution, i.e., a lower solvent volume to sample size ratio than 0.27 L/g. Thus far our efforts have focused
2.34 f 0.14 0.9996 f 0.0003
29.9 f 6.3 0.9997 f 0.0008
on keeping pilot column GAC samples, especially those nearing saturation at 1.0 mg of CHC13/g (14), below the upper quantitation limit defined by the maximum THM standard analyzed.
CONCLUSION Sonication in methanol is appropriate for extraction of THMs from GAC, and this procedure can be applied to pilot column and treatment plant GAC samples, for a direct determination of THMs adsorbed from water to the GAC surface. In further treatment plant-pilot column work, it would be interesting to spiked labeled trihalomethanes into the influent, to determine if the coadsorptionof complex mixtures of organics interferes with extraction of trihalomethanes from GAC samples. In further laboratory experiments, it would be interesting to determine if organics more strongly adsorbed than the THMs can also be extracted by sonication in methanol, possibly using higher power input than that used in this study. Registry No. CHCl,, 67-66-3; CHBrCl,, 75-27-4; CHBrzCl, 124-48-1; CHBr,, 75-25-2; H20, 7732-18-5; C, 7440-44-0.
LITERATURE CITED Braus, &my; Mlddleton, Francis, M.; Walton, Graham Anal. Chem. 1951. 23. 1160-1164. Mlddleton,'Francis M.; Pettit, H. H.; Rosen, Aaron A. Proc. Ind. Waste Conf. 17th 1963. 122, 454-460. Keith, Lawrence H.; Garrison, Arthur W.; Allen, F. R.; Carter, M. H.; Floyd, T. L., Pope, J. D.; Thruston, A. D. I n Identlflcation and Analysis of Or@nlc Pollutants in Water Keith, Lawrence H., Ed.; Ann Arbor Sclence: Ann Arbor, MI, 1976; pp 329-373. Kleopfer. Robert D. In IdentMcation and Analysls of Organic Pollutants in Water; Keith, Lawrence H., Ed.; Ann Arbor Science: Ann Arbor, MI, 1976; pp 399-416. Dunlap, W. J.; Shaw, D. C.; Scalf, M. R.; Cosby, R. L.; Robertson, J. M. I n Identification and Analysis of Organic Pollutants in Water; Keith, Lawrence H., Ed., Ann Arbor Science: Ann Arbor, MI, 1976; pp 453-476. Standard Methods for the Examination of Water and Waste Water, 15th ed.;American Public Health Association, American Water Works
Anal. Chem. 1986, 58, 1822-1826
1822
Association, Water Pollution Control Federation: Washington DC, 1980. (7) Chrlswell, Colin D.; Ericson, Rhonda L.; Junk, Gregor A,; Kenneth W.; Fritz, James S.; Svec, Harry J. J. A m . Water Assoc. 1977, 69, 669-674. (8) Van Rossum, Peter; Webb, Ronald G. J . Chromatogr. 1978, 750. 381-392. (9) Chen, Abraham S.; Larson, Richard A,; Snoeyink, Vernon L. Environ. Sci. Technol. 1982, 16, 268-273. (10) McCreary, John J.; Snoeyink, Vernon L.; Larson, Richard A. Environ. Sci. Technol. 1982, 16. 339-344. (11) Voudrlas, E. A.; Dielmann, L. M.; Snoeyink, Vernon L.; Larson, Richard A.; McCreary, John J.; Chen Abraham S. C. Water Res. 1983, 17, 1107-1 114. (12) Coyle, Gerry T.; Maloney, Stephen W.; Gibs, Jacob; Suffet, Irwin H. I n Water Chlorination : Environmental Impact and Health Effects, Joky, Robert L., Brungs, William A., Cotruvo, Joseph A,, Cumming, Robert E., Mattice, Jack S., Jacobs, Vivian A,, Ed.; Ann Arbor Science: Ann Arbor, MI, 1983; Vol 4, pp 421-443. (13) Alben, Katherine T.; Shpirt, Eugene Environ. Sci. Technol. 1983, 17, 187- 192. (14) Alben, Katherine T.; Shpirt, Eugene; Kaczmarczyk, Joan H. Proc. AWWA Annu. Conf., Conf. Proc. Am. Water Works Assoc. 1984,
1555-1571. (15) Grob, Kurt J. Chromatogr. 1973, 8 4 , 255-273. (16) White, Lowell D.; Taylor, David G.; Mauer, Patricia A,; Kupel, Richard E. Am. Ind. Hyg. Assoc. J. 1970, 31, 225-232. (17) Coleman, W. Emile; Melton, Robert G.; Slater, Robert W.; Kopfler, Frederick W.; Voto, Stephen J.; Allen, Wendy K.; Aurand, Theresa A. Proc. Am. Water Works Technol. Conf. V I I , Am. Water Works Assoc, 1980, 93-111. (18) Sawicki. E. Health Lab. Sci. 1975, 72. 407-414. (19) Golden, C. Sawicki. E. Anal. Letf., PartA 1978, A l l , 1051-1062. (20) Swanson, Donald H.: Walling, Joseph F. Chromatogr. Newsl. 1981, 9 , 25-26. (21) Cooke, N. E.; Gaikwad, R. P. Can. J . Cbem. Eng. 1983, 67, 697-702. (22) Jackson, W. Roy; Larkins, Frank P.; Thewlis, Paul; Watkins, Ian Fuel 1983, 62, 606-607. (23) Koh, Tee-Siaw Anal. Chem. 1983, 55, 1814-1815.
(24) Alben. Katherine T.; Kaczmarczyk, Joan H. J. Chromatogr. 1986, 357 , 497-500. (25) Fed. Regist. 1979, 4 4 , 68624. (26) Activatd Carbon Product Bulletin: Flkrasorb 300 and 400 Grsnular Activated Carbons for Potable Water Treatment; Calgon Corp.: Plttsburgh, PA, 1976. (27) Fisher, James L., Calgon Corp., Pittsburgh, PA, private communication to K. Alben, June 13, 1985. (28) Basic Concepts of Adsorption on Activated Carbon: Calgon Corp.: Plttsburgh, PA, 1985. (29) Glaze, William H.;Lin, '2.4. Optlmization of Liquid-LiqvM Extraction Methods for Analysis of Organics ln Water, National Technlcal Information Service: Springfield, VA, Oct 1963. (30) Glaze, William H.; Rawley, R.; Burleson, J. L.; Mapel, D.; Scott, D. R. I n Advances in the Identlkatlon and Analysis of Organlc Pollui%nts in Water; Keith, L. H., Ed.; Ann Arbor Science: Ann Arbor, MI, 1981; Vol. 1, pp 267-280. (31) Richard, John J.; Junk, Gregor A. J. Am. Water Works Assoc. 1977, 69,62-64. (32) Mleure, James P. J. Am. Water Wwks Assoc. 1977, 69, 60-62. (33) Veith, Gilbert D.; Macek, K. J.; Petrocelli, S. R.; Carroll, John I n Aquatic Toxic&@; Eaton, J. G., Parrish, P. I?.,Hendricks, A. C., Eds.; ASTM: Philadelphia, PA, 1980; pp 116-129. (34) Banerjee, Sufi; Yalkowsky, Samuel H.; Valvani, Shri Environ. Sci. Technol. 1980, 14, 1227-1229. (35) Hansch, Corwin; Quinlan, John E.; Lawrence, Gary L. J. Org. Chem. 1968, 33, 347-350. (36) Barbari, Timothy A.; King, C. Judson Envlron. Sci. Technol. 1982, 76, 624-627. (37) Junk, Gregor A.; Ogawa, I.; Svec, Harry J. I n Advances in the Identification and Analysis of Organic Pollutants; Keith, Lawrence H., Ed.; Ann Arbor Sclence: Ann Arbor, MI, 1981; pp 281-292. (38) Dixon, Wilfrid J.; Massey, Frank J. Introduction to Statistlcal Ana/@; McGraw-Hill: New York, 1957; pp 109, 124-127, 384. (39) Glaser, John A.; Foerst, Dennis L.; McKee, G. Ed.; Quave, Stephen A,; Eudde, William L. fnviron. Sci. Techno/. 1981. 75, 1426-1435.
RECEIVED for review November 25, 1985. Accepted March 18, 1986.
Magnitude of Artifacts Caused by Bubbles and Headspace in the Determination of Volatile Compounds in Water James F. Pankow Water Research Laboratory, Department of Chemical, Biological, and Environmental Sciences, Oregon Graduate Center, 19600 N . W. Von Neumann Dr., Beauerton, Oregon 97006
The formation of bubbles in a water sample or the presence of headspace above a water sample wlU cause losses of volatlle analytes. Bubbles may occur when ground waters high in dlssdved gases are brougM to the M a c e . Equations are derived to predict the magnitudes of both types of artlfacts. When 1.0 atm is the lowest pressure that a sample sees, the crlterlon for bubble formatbn wly be pb > 1.0 pw, where p,, Is the equivalent In situ parHal pressure (atm)of the dissolved gas and p w Is the vapor pressure of water. I f bubbles form, the percent error (worst case) at 20 OC (293 K) WW be 100(exp[-1024 H (293 K) K, (293 K) @, 0.97711 - 1). H (293 K) (atm-m3/mol) and K , (293 K) (Watm) are the Henry's law constants of the analyte and bubble-forming gas, respectively, at 293 K. The percent error due to a headspace at 20 OC will be -100(41.6H (293) VdV,)/(41.6 H (293) V , / V , 1). V , (mL) Is the volume of the headspace, and V , Is the volume of the sample. Bubbles and headspace can in some cases cause problematic artifacts; in many cases, however, the arklfads they cause wlll be smaH for many of the commonty determined organlc compounds.
-
-
+
The need to determine organic compounds accurately in water samples is growing at a rapid pace. As such, the attention given to sample acquisition, handling, and storage is
increasing (I). With volatile compounds, there has been particular interest in the artifacts that result if (1)bubbles form in the water during sampling and/or (2) headspace is present in the sample vial after sampling. The former may occur when ground water high in dissolved gases is removed from the overlying hydrostatic pressure and brought to the surface. For volatile compounds, both bubble formation and the presence of headspace w ill cause a negative bias since they cause losses. While some uncertainty in the concentration of a contaminant will not greatly alter the perception of the potability of a given water, artifacts must nevertheless be kept within limits. Firstly, measured levels must be compared in a confident manner to specific water quality standards. Secondly, it is frequently important to know whether the water quality is indeed improving or worsening. Finally, when the processes controlling the fates of contaminants are studied, artifacts can become very troublesome when trying to assess the relative importance of various natural fate processes such as volatilization, biodegradation, sorption, and dispersion. Although it is not always possible to offer simple equations for the expected magnitudes of artifacts, it can be done for the cases when volatile compounds partition to bubbles and headspace. The equation for when bubbles form due to sample depressurization has not yet been presented, and is derived here. Although the equation for headspace-related
0003-2700/86/0358-1822$01.50/00 1986 American Chemical Society