Experimentally determined Henry's law constants for 17

Quantum Mechanical Predictions of the Henry's Law Constants and Their Temperature Dependence for the 209 Polychlorinated Biphenyl Congeners...
0 downloads 0 Views 856KB Size
Environ. Scl. Techno/. 1988, 22,448-453

Experimentally Determined Henry's Law Constants for 17 Polychlorobiphenyl Congeners Frank M. Dunnlvant, John T. Coates, and Alan W. Elzerman" Environmental Systems Englneering, Clemson University, Clemson, South Carolina 26934-09 19

Henry's law constants (HLCs) have been experimentally determined for 17 polychlorobiphenyl (PCB) congeners by a gas-purging technique, with results ranging from 0.3 X lo4 to 8.97 X lod4atmm3/mol and an average value of 3.53 X atm.m3/mol for the congeners studied. The experimental results are in good agreement with previously published values. For the entire group of congeners studied, measured HLCs were not correlated with molecular weight but did appear directly related to chlorine substitution pattern. Increasing HLCs coincided with greater degrees of ortho-chlorine substitution within a molecular weight class. This trend may be useful to relate individual HLCs to chlorine substitution patterns for the remaining PCB congeners. ~~

~

Introduction Polychlorobiphenyl (PCB) concentrations in water, sediment, and biota can be significantly influenced by local emissions or diffuse urban sources. However, atmospheric inputs can also be significant. Atmospheric transport has been considered a major global redistribution process for PCBs (1-4). For example, Eisenreich et al. (5) estimate that up to 85% of the total input of PCBs to Lake Superior could be due to atmospheric sources. Atmospheric sources have also been suggested as the major source of PCBs to Lake Michigan (6). Harvey and Steinhauer (3) regarded wind transport as the major source of chlorinated hydrocarbons (specifically PCBs) to the oceans. Additional examples of atmospheric transport of PCBs to the Great Lakes have been summarized by Simmons (7). Equilibrium partition coefficients, such as Henry's law constants (HLCs), which describe equilibrium partitioning between a liquid and a gas phase, are commonly used in predictions of the fate of chemicals such as PCBs in the environment. Additionally, HLCs are requisite in the development of atmospheric transport models (8, 9). Accurate HLC and other physical/chemical parameter data must be obtained before fate-prediction models can be used with confidence. Unfortunately, the quality and quantity of available data are still insufficient. PCBs normally exhibit relatively low vapor pressures, typically 4 X lo4 to 2.3 Pa at 25 OC (8). Therefore, solely on the basis of vapor pressure, PCBs would not be considered volatile. However, PCBs are relatively insoluble and hydrophobic (log Ko$s of 4-9) ( I O , 11) and, therefore, can exhibit significant volatility from aqueous systems (8). Given two phases, air and water, PCBs will continue to volatilize from or dissolve into liquid until chemical potential (or fugacity) is equalized and equilibrium is attained between the two phases. This equilibrium can be expressed as the Henry's law constant, the partial pressure in the gas phase divided by the concentration in the liquid phase, commonly expressed in units of atm m3/mol. Several techniques have been used to obtain HLC values, including indirect methods such as estimation of HLCs from estimated and measured vapor pressures (12-14) and direct measurement of air- and water-phase concentrations (15). HLCs obtained from indirect methods require validation with direct measurement techniques prior to ac448

Environ. Sci. Technol., Vol. 22, No. 4, 1988

ceptance. Murphy et al. (15) provide the only direct measurements with the techniques mentioned, but the data must be evaluated in light of the potential analytical error resulting from the low concentrations that result in the experiment. Mackay et al. (16) introduced a novel gas-purging technique that allows calculation of HLCs from kinetic data and used the technique to obtain HLCs for several hydrophobic compounds. Karickhoff and Morris (17) and Coates and Elzerman (18) have used this technique to study desorption of PCBs from sediments. Oliver (19) and Hassett and Milicic (20) used a similar gas-purge technique employing a pipet in place of the glass frit used by Mackay et al. (16) to determine HLCs for selected PCB congeners. Hassett and Milicic (20) noted that equilibrium between the purge gas and water may not have been attained in their apparatus. Thus, the HLC reported for 2,2',5,5'-tetrachlorobiphenyl (20) represents a minimum value. The design of the purge vessel described by Oliver (19) may also have resulted in nonequilibrium conditions, but insufficient data were given to determine if equilibrium was established. This investigation was conducted to provide more experimentally determined HLC values that can be used to validate models for prediction of values for the remaining 209 congeners. Experimental determination of all 209 HLCs is not possible due to the difficulty of obtaining the congeners in pure form and, in any event, is not necessary due to the rare occurrence of many congeners in the environment. Seventeen PCB congeners were selected for study on the basis of availability, presence in the environment, and representation of different chlorine substitution patterns. Experimental Procedures Chemicals. All water used in the experiments was double distilled with the last distillation being in an allglass still. Organic solvents (acetone and isooctane) were Baker-Resi analyzed. Blanks confirmed the absence of interfering peaks in the gas chromatography (GC) analysis. Mercuric chloride (HgCl,) used in inhibiting biological degradation was ACS grade (Fisher, Inc). Pure PCB congeners (purity >99+%) were purchased from Ultra Scientific (Hope, RI). Tenax (40-60 mesh) was purchased from Supelco, Inc. Prior to use, Tenax columns were eluted with acetone and isooctane, and the remaining solvent was evaporated in a vacuum desiccator. Equipment. The purge vessels were a modification of those described by Mackay et al. (16) and Karickhoff and Morris (17) and utilized fritted glass gas outlets. For a detailed description of the vessel used here, refer to Coates (21). An oil-free air compressor (ITT Pneumotive, Monroe, LA) provided the air used in purge experiments. Compressed air was passed through a particulate trap (0.01pm), moisture trap (CaS04),activated-carbon column, and finally a Tenax column prior to introduction to the purge vessel. Gas flow was controlled with pressure regulators. A Hewlett-Packard 5880A gas chromatography system equipped with a 63Nielectron capture detector (ECD) was used in PCB quantitation. Congener separation was ac-

00 13-936X/88/0922-0448$01.50/0

0 1988 American Chemical Society

complished with a 30-m DB-5 capillary column (0.25 mm i.d. and 0.25 pm film thickness) from J & W Scientific. Temperature programs varied depending on the specific set of compounds being analyzed. Purge Experiments, Aqueous solutions of PCBs were prepared by adding appropriate amounts of stock congener solution (in isooctane) to a solvent-cleaned empty flask. In order to avoid the presence of PCB microcrystals, target concentrations did not exceed half the aqueous saturation values for any given congener (22). Isooctane was allowed to evaporate and double-distilled water was added to obtain the desired congener concentration, assuming that all of the previously added PCB dissolved. Mercuric chloride (100 mg/L) was added to avoid biodegradation of PCBs (23). Headspace was minimized, and flasks were capped during equilibration to minimize loss of PCBs to the atmosphere. Solutions were prepared and purged in the following groups: (1) dichloro- (DCB), (2) trichloro- and tetrachloro- (TCB), and (3) pentachloro- (PCB) and hexachlorobiphenyl (HCB). Solutions were stirred for at least 2 weeks in a temperature-controlled room at 25 “C prior to gas purging. Purge experiments (run in triplicate) were conducted by placing 500 mL of equilibrated solution in the purge vessel and by using the gas flow to purge the PCBs onto a Tenax column trap. Gas flow was immediately started (1L/min), and samples were taken at varying time intervals determined by the mass of PCB congener expected in the Tenax tube. The trap columns consisted of glass tubes containing glass wool and Tenax (approximately 0.5 g). PCBs were desorbed from the Tenax tube by elution with 3 mL of acetone followed by 5 mL of isooctane. A total of 8 mL of double-distilled water was added to the collection vial, and the solution was shaken for 3 min and allowed to separate. The resulting isooctane solution was injected directly (no cleanup needed) to the GC or diluted in order to quantitate in the linear range of detector response for each congener. Quantitation. Individual PCB congeners were quantified by the GC/ECD system described earlier. Prior to injection, an internal standard was added to the sample. The internal standard varied with each solution being analyzed but was a PCB congener eluting from the GC in a clear area of the chromatogram. The use of internal standards reduced the error of analysis by correcting for injection errors, splitter inconsistencies, and detector fluctuations. Samples were analyzed in a linear range of detector response for each congener and internal standard. Data Handling. HLC values were calculated from purge-rate data on the basis of the model developed by Mackay et al. (16) and utilized by Karickhoff and Morris U7), Coates and Elzerman (19), and others (19, 20). A depletion rate constant (D,) is determined from the removal rate of material by purging, and the HLCs are calculated from the relationship HLC = D,.RTV/G (1) where D, = depletion rate constant, R = ideal gas law constant, T = 298 K, V = aqueous solution volume, and G = gas flow rate. A nonlinear least-squares technique was used to determine D, from plota of cumulative nanograms of PCBs purged from solution versus time. As discussed below, this approach avoids some potential problems associated with determining D,as a first-order rate constant from linearized plots of the log of the fraction remaining versus time, which has been the common approach.

Results and Discussion Evaluation of Experimental Methods. An important

Table I. Comparison of Data Modeling Techniques

PCB congener 4,4’-DCB 2,2’,6,6’-TCB 2,2’,4,4’,5,5’-HCB 2,2’,4,4’,6,6’-HCB

HLC, X104 atm.m3/mol Coates (21) Coatesa (21), recalcd? new data,b method 3 method 3 method 1 1.45 1.48 1.23 1.15

1.58 3.85 1.55 8.35

1.99 5.50 1.31 7.55

a Calculated using fraction remaining in solution versus time (see text). bCalculated using cumulative mass purged (ng) versus time (see text).

factor in considering the purge technique of measuring HLCs is whether bubbles and solution in the column are in equilibrium. As noted above, two previous investigations (19,20) that used the gas-purge technique may not have attained equilibtium of PCBs between the gas and liquid phases in the purge vessel. In order to test for equilibrium conditions in the purge vessels used in this investigation, three different solution column heights in the purge vessels were evaluated to allow different contact times between the solutions and purge gas. Dichlorobiphenyls (DCBs) were used as test compounds since their greater aqueous solubility as compared to that of more chlorinated congeners makes them less likely to reach equilibrium in the gas phase than,higher chlorinated congeners. Purge runs were performed in triplicate with three purge vessels that were identical except for column volume and corresponding column heights of 8,18, and 28 cm. It should be noted that three separate purge vessels were used as opposed to simply varying the aqueous volume in a single full-size vessel. Each vessel was constructed to contain identical “dead volume” or headspace above the aqueous solution. If a single, full-size vessel had been used in the 8-cm purge experiment, the resultant relatively large headspace above the purged solution could have diluted the gas stream concentration and biased the HLC calculation by lowering the observed depletion rate constant of the plot of cumulative mass PCBs purged versus time. HLCs calculated from eq 1(which includes the aqueous volume factor) were similar for the three column heights. Therefore, all three purge vessels were assumed to allow PCB equilibration between the purge gas and solution. The use of a fritted glass or other diffuser in the vessel to obtain many small bubbles appears to be important. The 28-cm (500 mL of solution) purge vessel was used for the remainder of this investigation. Another important consideration in the purging approach is data interpretation. At least three methods of utilizing the data to determine depletion rate constants for use in eq 1 are possible as follows: (1) fraction remaining in solution versus time, (2) fraction purged from solution versus time, and (3) cumulative mass purged from solution versus time. Mathematically, the three approaches should give identical results for D,, and experimentally, rate constants generated by each of these methods were found to be virtually identical when data fit the model well. However, for data sets that did not seem to fit the model when the fraction remaining was plotted versus time due to errors associated with determining the initial concentration, method 3 proved superior. For example, methods 1and 3 above were evaluated with data from Coates (21). Results are given in Table I. Both calculation methods yielded comparable HLC values for congeners 4,4’-DCB and 2,2’,4,4’,6,6’-HCB. However, large differences resulted for congeners 2,2’,6,6’-TCB and 2,2’,4,4’,5,5’-HCB. Visual comparisons of the model fit to Environ. Sci. Technol., Vol. 22, No. 4, 1988

449

PURGE

TIME (MINI

DEPLETION CURVE FOR 2,2’

10

.(I

?c

00

00

-

DCB

110

o.

60

PURGE TIME (MINI DEPLETION CURVE FOR 2 2 ’ 4 4 ’ 5 5

Y

,

,

00

io

PURGE TIME

- HCB

PO

“0

,eo

60

(MIN)

DEPLETION CURVE FOR 2.2’,4 4’,8.6’

- HCB

Figure 1. Depletion curves for selected PCB congeners.

the data suggested a poorer fit with the fraction remaining in solution method (method 1)than the cumulative mass purged method (method 3). Further analysis indicated that the discrepancies between methods 2 and 3 for the data of Coates (21)were related to experimental artifacts. Additional data manipulations indicated that HLCs calculated by method l were strongly dependent on an accurate initial PCB concentration measurement (i.e., method 1 uses a data plot of Ctime/Cinitidversus time), which can be difficult to obtain. When data from Coates (21) were recalculated by method 3, results consistent with this investigation were obtained (see below). Due to the above discussed sensitivity to accurate initial concentration values for method 1,data reported here were obtained by method 3, which utilizes data without transformation. The calculations used in method 3 yield estimates of D, (related to HLC as shown in eq 1)and initial concentration of purgeable PCB in the solution (related to the maximum value of cumulative amount purged). Other potential experimental problems were also evaluated. Hassett and Milicic (20) noted an experimental artifact of the gas-purge technique that could affect HLC calculations when any of the methods discussed above are used. If adsorption of PCBs from the amount initially added to the glass vessel occurs after the solution is added, subsequent desorption of PCBs back into the solution (after depletion in the water has occurred due to purging) will decrease the apparent D,(and HLC). The presence or absence of glass sorption might be evaluated by comparing the depletion rate constants generated from the model for a complete set of data (rise and plateau) to the constants generated from a set containing only the steep rise in the initial portion of the data plot. Note the partial data plot must contain adequate data points, approaching the plateau, to allow accurate modeling with the nonlinear least-squares technique. A purge interval of 40 min was sufficient in most cases for the partial data plot. If a higher D, (or HLC) is obtained from the partial data plot (0450

Environ. Sci. Technol., Vol. 22, No. 4, 1988

40-min purge) than from the complete data set, then glass adsorption/desorption would be indicated. Comparisons of this type were made for all 17 PCB congeners and indicated no significant difference for 12 compounds. The remaining five were inadequately modeled, since the 40min purge run did not allow the plateau to be reached, and thus, insufficient data were available to adequately model the depletion rate curve. On the basis of these results, adsorptioq/desorption from the glass purge vessel was considered to negligible and not to affect the purge rate or D, values obtained. Another characteristic problem of the purge technique remains unresolved, although it is not generally a significant problem. Mackay et al. (16) noted some deviation of the data below the model occurred in the first few minutes of a purge run. The cause may have been related to surface effects or incomplete mixing in the column. In this investigation, when samples were taken in the first 5 min, the mass of PCB collected in the first fraction was lower than the mass collected in the second fraction, which is inconsistent with the model. However, data points after the first 5 min were consistent with the mode), and since each data point was weighted equally in the analysis, the small deviation of the first point became insignificant in the overall analysis. Figure 1contains cumulative mass (ng) purged versus time plots for 2,2’-DCB, 2,2’,5,5’-TCB, 2,2’,4,4’,5,5’-HCB, and 2,2’,4,4’,6,6’-HCB. Note that since cumulative mass purged data were used, replicate plots will not overlap unless the initial PCB concentrations were the same for each replicate. Some replicates were purged days apart, and several plots exhibited a loss of PCBs to the atmosphere, thus stressing the importance of controlling PCB loss due to volatilization. On the basis of the observation, PCB-spiked aqueous solutions of humic acid or sediment should be equilibrated in a tightly closed container with minimal headspace. In addition, the loss of PCBs in these experiments shows the potential for cross-contamination

Table 11. Henry’s Law Constant (HLC) Data Summary‘ HLC, X104 atmm3/mol IUPAC no. 4 9 11 12

15 26 30 40 52 54 53 77 101 104 128 153 155

biphenyl chlorine substitution pattern 2,2’

275 3,3’ 374 4,4’ 2,3’,5 2,4,6 2,2’,3,3’ 2,2’,5,5’ 2,2’,6,6’ 2,2’,5,6’ 3,3’,4,4’ 2,2‘,4,5,5’ 2,2’,4,6,6’ 2,2’,3,3’,4,4’ 2,2’,4,4’,5,5’ 2,2’,4,4’,6,6’

data from this investigation (25 “C) 3.37 (0.267) 3.88 (0.075) 2.33 (0.106) 2.05 (0.086) 1.99 (0.110) 3.25 (0.095) 6.49 (0.246) 2.02 (0.050) 3.42 (0.086) 5.50 (0.230) 4.06 (0.125) 0.94 (0.076) 2.51 (0.163) 8.97 (0.432) 0.302 (0.023) 1.32 (0.107) 7.55 (0.940)

Coatesb (25 O C )

Oliverc (20 OC)

Hassettd (25 “C)

Murphye (room temp)

Westcott’ (25 “C)

Burkhardg (25 “C)

3.1-5.3

5.5 3.27 1.34 0.948 1.09 2.80 3.68 2.00 5.25 18.60 2.56 0.431 3.23 18.30 0.676 1.77 15.50

2.2

1.45 (0.14)

=3 1.2 1.2

0.25

2.6

1.48 (0.07) 1.1-3.5

0.7 1.23 (0.01) 1.15 (0.16)

0.6

5 3.5

Values given in parentheses are standard deviations. Data taken from Coates (21). Data taken from Oliver (19). Data taken from Hassett et al. (20). eData taken from Muruhv et al. (15). /Data taken from Westcott et al. (14). 8Data taken from Burkhard et al. (13).

of samples in the laboratory. Discussion of Henry’s Law Constants. Table I1 contains experimentally determined HLC values for the 17 PCB congeners studied here and summarizes published experimental and estimated HLCs for these compounds. First, the new data will be compared to other experimentally determined HLCs. As discussed earlier, Coates (21) determined HLCs for four of the PCB congeners studied here using identical purge vessels. The data are in close agreement for congeners 4,4’-DCB and 2,2’,4,4’,5,5’-HCB. However, significant differences resulted for 2,2’,6,6’-TCB and 2,2’,4,4’,6,6’-HCB. As discussed above, these contrasting results can be attributed to the method of analyzing the data and difficulties in measuring total initial aqueous PCB concentrations. Once these corrections have been made by utilizing a plot of cumulative amount purged versus time (method 31, the results are more consistent (refer to Table I). Data shown in Table I1 support the conclusions of Hassett and Milicic (20) that PCB equilibrium between the purge air and water was not obtained in their apparatus. Their value was lower than the corresponding value obtained in this investigation. This equilibrium condition was not evaluated by Oliver (191,but as seen in Table 11, his data are also consistently lower than HLC values reported here, which is the expected relationship for a vessel used in the nonequilibrium condition. The set of experimental data from Murphy et al. (15) was obtained by directly measuring phase concentrations (air and water) of a mixture assumed to be at equilibrium. With the exception of 2,2’,3,3’,4,4’-HCB, data derived from the purge technique are surprisingly close to values obtained by Murphy et al. (15),considering the sampling and analytical difficulties associated with the technique used by Murphy et al. Since all 209 PCB congeners are not generally available for experimentation, indirect techniques have been used to estimate HLC values. Westcott et al. (14) estimated HLCs for 2,2’,5,5’-TCB and 2,2’,4,5,5’-PCB using measured vapor pressures for these congeners. The range of values reported by Westcott et al. (14) (given in Table 11) bracket the values obtained in this investigation. Unfortunately the technique of Westcott et al. (14) has the same drawback as other measurement methods, difficulty in ob-

taining data for all 209 congeners. Burkhard et al. (12,13,24) estimated individual vapor pressures and aqueous solubilities, based on structureactivity relationship models that utilized available data, and calculated the corresponding HLCs for all 209 PCB congeners. Data from Burkhard et al. (13) for the 17 congeners studied here are also given in Table 11. Comparisons between Burkhard et al. (13) data and results from this investigation indicate reasonable agreement with the deviation (relative to experimental values reported here) ranging from 1to 238%. In general, the variation was approximately 37 5% , excluding congeners 2,2’,6,6’TCB, 2,2’,4,6,6’-PCB, 2,2’,3,3’,4,4’-HCB, and 2,2’,4,4’,6,6’-HCB. For these congeners the prediction model used by Burkhard et al. (13)appears to overestimate HLCs, indicating that the model may give high estimates for all congeners containing a high degree of ortho-chlorine substitution. The effects of this chlorination pattern may be related to steric hindrance, resulting in lower solute/ solvent interaction and limiting biphenyl rotation as compared to a PCB congener without ortho-chlorine substitution. Burkhard et al. (13)noted an increase in HLCs with an increasing number of ortho-chlorine substitutions within the same molecular weight class, which is also consistent with the data from this investigation. In contrast, Burkhard et al. (24) state that HLCs increase with decreasing ortho-chlorine substitution for each weight class. In addition, it was reported that congeners with the highest degree of meta- and/or para-chlorine substitution possessed the highest HLCs. As noted above, experimental data presented here indicate that HLCs increase directly with increasing ortho-chlorine substitution in a PCB molecular weight class, which is consistent with the data in Burkhard et al. (13)but not with statements in Burkhard et al. (24). Table I11 presents data for PCB congeners from this investigation grouped by degree of chlorination, and therefore by molecular weight, and arranged by decreasing HLC value within each group. The experimental data indicate that as ortho chlorination increases the HLC increases, which is consistent with the values reported by Burkhard et al. (13). The only notable exception is 2,2’DCB, which has a slightly lower HLC than 2,5’-DCB. Environ. Sci. Technol., Vol. 22, No. 4, 1988

451

Table 111. PCB Congeners Arranged by Degree of Chlorination and HLC IUPAC no.

chlorine position

no. of ortho-chlorines

HLC, xi04 atm.m3/mol

Dichlorobiphenyls 9 4

2,5’ 2,2‘

1 2

11 12 15

3.3‘ 394 4,4’

0 0 0

30 26

2,4,6 2,3’,5

54 53 52 40 77

Tetrachlorobiphenyls 2,2’,6,6’ 4 2,2’,5,6’ 3 2,2’,5,5’ 2 2 2,2’,3,3’ 3,3’,4,4’ 0

5.50 4.06 3.42 2.02 0.94

104 101

Pentachlorobiphenyls 2,2’,4,6,6’ 4 2,2’,4,5,5’ 2

8.97 2.51

155 153 128

Hexachlorobiphenyls 2,2’,4,4’,6,6’ 4 2,2’,4,4’,5,5’ 2 2,2‘,3,3‘,4,4‘ 2

7.55 1.32 0.30

~~~

Trichlorobiphenyls 2 1

3.88 3.37 2.33 2.05 1.99 6.49 3.25

However, this exception may actually be related to an effect of substitution at the 5- or 5’-position, since HLCs for all congeners in Table 111 with a chlorine in the 5- or 5’-position are greater than the HLCs for similar congeners without chlorines in 5- or 5’-position. Thus, it may be possible to relate HLCs to chlorine substitution patterns in PCB congeners and make quantitative estimates of HLCs for all congeners on the basis of chlorine substitution pattern. Unfortunately, sufficient detailed congener-specific data are not readily available, and the basis of required modeling procedures are still being developed. Quantitative structure-activity relationships (QSARs) are well established tools in pharmacology and drug development and could be useful for predicting the fate and distribution of a chemical in the environment. Shaw and Connell(25) found that the uptake and bioaccumulation of PCBs were related to both the partition behavior, as measured by K,,, and the adsorption characteristics of the congener, as estimated by relative chromatographic elution times. Chromatographic elution was determined to be dependent on chlorine substitution patterns, and a steric effect coefficient (SEC) was developed that was highly correlated to bioaccumulation (25). However, the SEC parameter does not adequately differentiate congeners containing ortho-chlorine(s) in addition to meta- and para-chlorine(s),and correlations between SECs and HLCs reported here were not significant. Cullen and Kaiser (26) evaluated rotational barriers of PCBs in relation to toxicity of PCBs using the quantum mechanical INDO method (intermediate neglect of diatomic overlap). Although high correlations were found with toxicities of PCBs (26),application of the INDO approach to HLC data did not produce good correlations. Structure-activity relationships have also been used to study mixed-function oxidase (MFO) inducers specific to PCB congeners (27). QSARs have been successfully used by Devillers et al. (28) to predict acutely toxic concentrations for certain groups of aromatic compounds to Photobacterium phosphoreum. Additional examples and discussions concerning QSAR are available in ref 29 and 30. 452

Environ. Sci. Technol., Vol. 22, No. 4, 1988

More HLC data and further development of QSAR models for PCB congeners are required for quantitative estimates of all PCB HLCs. I t is likely that useful correlations will be found between HLCs and chlorine substitution pattern within a molecular weight class, as evidenced in the data in Table 111. Relationships of HLCs between molecular weight classes will probably require additional factors since there is no apparent trend in the data in Table I11 when all 17 HLCs are considered together.

Conclusions The gas-purgingtechnique introduced by Mackay et al. (16) has been shown to be effective for measuring HLCs for 17 PCB congeners. The data presented here comprise the most comprehensive list of experimentally determined HLCs for pure PCB congeners to date. HLC values range from 0.30 X to 8.97 X atmm3/mol, with an average value of 3.53 X atmm3/mol (standard deviation = 2.38 X for the congeners studied. In general, individual values were comparable to previously reported experimental determinations. Experimental values and predicted values from Burkhard et al. (13)are comparable to data reported here, with the exception of congeners such as 2,2’,6,6’-TCB, 2,2’,4,6,6’-PCB, and 2,2,4,4’,6,6’-HCB that have a high degree of ortho substitution of chlorine. Data indicate that PCB HLCs are not directly related to congener molecular weight but increase with increasing degree of ortho chlorination. Parameterization of steric factors as related to chlorine substitution patterns coupled with measured HLC values might be used to estimate HLCs of PCB congeners not included in this investigation. Registry No. 2,2’-CB, 13029-08-8; 2,5-CB, 34883-39-1; 3,3’-CB, 2050-67-1; 3,4-CB, 2974-92-7; 4,4‘-CB, 2050-68-2; 2,3’,5-CB, 38444-81-4; 2,4,6-CB, 35693-92-6; 2,2’,3,3’-CB, 38444-93-8; 2,2’,5,5’-CB, 35693-99-3; 2,2’,6,6’-CB, 15968-05-5; 2,2’,5,6’-CB, 41464-41-9; 3,3’,4,4’-CB, 32598-13-3; 2,2’,4,5,5’-CB, 37680-73-2; 2,2’,4,6,6’-CB, 56558-16-8; 2,2’,3,3’,4,4’-CB, 38380-07-3; 2,2‘,4,4‘,5,5’-CB, 35065-27-1; 2,2‘,4,4‘,6,6‘-CB, 33979-03-2.

Literature Cited Atlas, E.; Giam, C. S. Science (Washington,D.C.)1981,211, 163-165. Nisbet, C. T.; Sarofim, A. F. EHP, Enuiron. Health Perspect. 1972, 1 , 21-38. Harvey, G. R.; Steinhauer, W. G. Atrnos. Enuiron. 1974, 8, 777-782. NAS (National Academy of Science) Polychlorinated Biphenyls; NAS: Washington, DC, 1979; 182 p. Eisenreich, S. J.; Looney, B. B.; Hollod, G. J. In Physical Behavior of PCBs in the Great Lakes; Mackay, D., et al., Ed.: Ann Arbor Science: Ann Arbor, MI, 1983; Chapter 7 , pp ii5-125. Murphy, T. J.; Rzeszutko, C. P. J. Great Lakes Res. 1977, 3, 295-312. Simmons, M. S. In Toxic Contaminantsin the Great Lakes; Nriagu, J. O., Ed.; Wiley: New York, 1984; Chapter 13, pp 267-309. Mackay, D.; Shiu, W. Y . ;Billington, J.; Huang, G. L. In Physical Behavior of PCBs in the Great Lakes; Mackay, D.. et al., Ed.: Ann Arbor Science: Ann Arbor, MI, 1983; Chapter ’4, pp 59-69. Doskey, P. V.; Andren, A. W. Enuiron. Sei. Technol. 1981, 15, 705-711. Rapaport, R. A.; Eisenreich, S. J. Enuiron. Sei. Technol. 1984. 18. 163-170. (11) Woodburn, K. B.; Dochette, W. J.; Andren, A. W. Enuiron. Sei. Technol. 1984, 18, 457-459. (12) Burkhard, L. P.; Andren, A. W.; Armstrong, D. E. Enuiron. Sei. Technol. 1985, 19, 500-507. (13) Burkhard, L. P.; Armstrong, D. E.; Andren, A. W. Enuiron. Sei. Technol. 1985, 19, 590-596.

Environ. Sci. Technol. 1988, 22,453-456

Baxter, R. A,; Gilbert, P. E.; Lidgett, R. A.; Mainprize, J. H.; Vodden, H. A. Sci. Total Enuiron. 1975, 4, 53-61. Burkhard, L. P.; Armstrong, D. E.; Andren, A. W. Che-

Westcott, J. W.; Simon, C. G.; Bidleman, T. F. Enuiron. Sci. Technol. 1981,11, 1375-1378.

Murphy, T. J.; Pokojowczyk, J. C.; Mullin, M. D. In Physical Behavior of PCBs in the Great Lakes; Mackay, D., et al., Ed.; Ann Arbor Science: Ann Arbor, MI, 1983; Chapter 3, pp 49-58. Mackay, D.; Shiu, W. Y.; Sutherland, R. P. Enuiron. Sci. Technol. 1979,13, 333-337. Karickoff, S. W.; Morris, K. W. Enuiron. Toxicol. Chem. 1985,4,469-479. Coates, J. T.; Elzerman, A. W. J. Contam. Hydrol. 1986, 1, 191-210. Oliver, B. G . Chemosphere 1985, 14, 1087-1106. Hassett, J. P.; Milicic, E. Enuiron. Sci. Technol. 1985,19, 638-643.

mosphere 1985, 14, 1703-1716. Shaw, G. R.; Connell, D. W. Enuiron. Sci. Technol. 1984, 18,18-23. Cullen, J. M.; Kaiser, K. L. E. In QSAR in Enuironmental Toxicology;Kaiser, K. L. E., Ed.; Reidel: Boston, MA, 1984; pp 39-66. Clarke, J. U. Chemosphere 1986, 15, 275-287. Devillers, J.; Chambon, P.; Zakarya, D.; Chastrette, M. Chemosphere 1986,15, 993-1002. QSAR in Enuironmental Toxicology;Kaiser, K. L. E., Ed.; Reidel: Boston, MA, 1984. QSAR in Toxicology and Xenobiochemistry; Tichy, M., Ed.; Elsevier: New York, 1985.

Coates, J. T. Ph.D. Dissertation, Clemson University, Clemson. SC, 1984. Mackay,'D.; Mascarenas, R.; Shiu, W. Y. Chemosphere

Received for reuiew January 2,1987. Accepted October 21,1987.

1980,9, 257-264.

Chalcogen Elements in Snow: Relation to Emission Source Kuen Y. Chlou and Oliver K. Manuel" Department of Chemistry, University of Missouri, Rolla, MissouriI 65401 ~

~

~~

We have measured the concentrations of S, Se, and Te in samples of 1986 snow and compared our results with those of earlier measurements. We were unable to find any earlier reports on Te in snow, but values of the Se/S concentration ratio in 1986 are about a factor of 6 lower than that in snow and glacial ice of 800 B.C. Measurements on intermediate samples demonstrate that most of the decline in the Se/S ratio occurred during the past 200 years, i.e., since the start of the Industrial Revolution. This temporal change in values of the Se/S ratio probably reflects a shift in the major emission sources-from natural processes such as volcanism and biomethylation to largescale combustion of fossil fuels.

Introduction Tellurium (Te) and selenium (Se), usually in association with sulfur (S), are widely distributed in environmental materials such as soils ( 1 , 2 ) ,waters ( 3 , 4 ) ,and air (5-7). The concentrations of Se and S have also been reported in snow samples collected in Greenland ice sheets (8), in Japan (9),and in Boston, MA (10). Prior to this study, we were unable to find any report on measurements of Te in snow. We initiated a study of chalcogen elements in snow as part of our investigation into the geochemical cycles of Te, Se, and S. The last element has received considerable attention because sulfur oxides are now recognized as the primary culprit of acid precipitation, accounting for about two-thirds of this acidity (11). Since Te and Se routinely occur in association with S in nature, most processes that release S to the atmosphere are also expected to release Te and Se. The concentration of Te in snow is low, and a new analytical procedure was developed that takes advantage of the high sensitivity of atomic absorption spectrometry for the detection of Te after removal of interfering elements by cation exchange (5). In the work reported here, this analytical procedure was used to obtain new data on the concentrations of Te, Se, and S in snow. Ratios of these elements, e.g., values of Te/S and Se/S, depend on the nature of the emission source and on the atmospheric chemistry of these three members of the Group VIA elements. The advantages of 0013-938X/88/0922-0453$01.50/0

measuring the concentrations of closely related elements in geochemical studies are well recognized and, in relation to this study (7), could extend our insight into the origin of the Te, Se, and S trapped in snow over the historical period recorded in permanent snow fields.

Experimental Section Fresh samples of snow were collected during, or immediately after, snowfalls here at Rolla, MO, during the period of February 10 to March 13, 1986. The network of six sampling sites was set up in such a way that the distance of any site from the University of Missouri was about 2-4 miles. The snow was collected in polyethylene bags, transferred to the laboratory to melt, and then filtered to remove any dust or other debris. Each sample was then acidified by the addition of 1mL of concentrated HN03 per liter of melted snow. Next, the volume was reduced to 50 mL by pumping off water vapor with an aspirator at low temperature, 60 "C. The preconcentrated melt was then heated to near dryness, and 1.5 mL of concentrated HC1 was added. The HC1 reduces Te(V1) to Te(IV), which can be adsorbed onto the cation-exchange column. The acidity of the solution is reduced to 0.05 N HC1 before loading onto a cation-exchange column, 1.2 cm i.d. X 10 cm long. Under these conditions, Te(1V) exists predominately in the form of TeO(OH)+or Te(OH)3+and is retained quantitatively on the cation-exchange column (5). Other interfering anions are not retained on the column, and Se is released in the first 100 mL of 0.05 N HC1 effluent. Te is efficiently eluted from the cation-exchange column with 150 mL of 0.3 N HC1, but most of the interfering metal cations of Cu, Hg, Ni, etc. remain on the column. Yields for the recoveries of Te and Se are each about 92% ( 5 ) . The 0.05 N HC1 portion of the eluant is used for the determination of Se and the 0.3 N HC1 portion is used for the determination of Te. Graphite furnace atomic absorption spectrometry is used for the determinations. Results and Discussion The concentrations of Te and Se in snow samples collected during the period of February 10 to March 13,1986, are shown in Table I. The numerical designations in the

0 1988 American Chemical Society

Environ. Sci. Technol., Vol. 22, No. 4, 1988 453