Partitioning of polycyclic aromatic hydrocarbons ... - ACS Publications

(4) TestMethods for Evaluating Solid Wastes (EPA SW 846. Volume 1A) .... 2-methylnaphthalene. 34. 142.2. 26“. 4.00 (0.2). -3.62 acenapthene. 93. 154...
0 downloads 0 Views 2MB Size
Envlron. Scl. Technol. 1992, 26, 2 104-2 110

least trimodal, with peaks around 55,80,and 95 mg/kg. Samples G and H are intermediate between the distributions of sample F and I. The polymodal nature of the distributions of sample results is most likely the result of several different problems overlapping. The nature of these problems is clearest at higher concentrations. There are a number of sources for these problems. One of these is a failure to control for the linear dynamic range of the instrument. The second is the failure to control for detector drift and buildup of deposits on the detector. Finally, extraction equipment, solvents, and cleanup procedures accentuate or attenuate these underlying problems. This is the subject of future publications. Summary and Recommendation Clearly there are widespread and systematic errors in the analysis for PCBs in soils. It is obvious that most laboratories need to review their PCB analytical procedures. The legal limit for PCBs in solids is 50 pg/g (8), so sample F would be classified as a hazardous waste. Fully, 36 accredited laboratories, just shy of 30%,would have characterized this material as nonhazardous. Hazardous materials laboratory accreditation agencies, where they exist, should work with the laboratories to improve performance, as it is unacceptable to have so many laboratories making so many errors in PCB analysis. Clearly the use of spiked multivial performance evaluation samples is an important tool for improving laboratory performance. Acknowledgments We thank Dr. William Nilsson and Monina Ligao of the Southern California Laboratory for their assistance with the PCB work.

Registry No. Arochlor 1260, 11096-82-5.

Literature Cited Interlaboratory Study on Determination of Polychlorinated Biphenyls in Environmentally Contaminated Sediments. Alford-Stevens, A. L.; Budde, W. L.; Budde, T. A. Anal. Chem. 1985,57, 2452-2457. Bayne, C . K.; Stewart, J. H. Multilaboratory Validation Study of PCBs in Soils Using Soxteca Extraction Technique (Method 3541). In Proceedings of the USEPA’S Sixth Annual Symposium on Waste Testing & Quality Assurance; U.S. Government Printing Office: Weshington, DC, July 1990; pp 11-193-11-132. Kimbrough, D. E.; Wakakuwa, J. Industry-Wide Performance in a Pilot Performance Evaluation Sample Program for Hazardous Materials Laboratories. 1. Elemental Precision and Accuracy. Environ. Sci. Technol., preceding paper in this issue. Test Methods for Evaluating Solid Wastes (EPA SW 846 Volume IA), 3rd ed.; Method 3050; Office of Solid Waste and Emergency Response, U.S. Environmental Protection Agency: Washington, DC, Nov 1986. Bauer, E. L. A Statistical Manual for Chemists, 2nd ed.; Academic Press: New York, 1971; pp 27-70. Eckschlager, K. Errors, Measurements, and Results in Chemical Analysis, 1st ed.; Van Nostrand Reinhold Co: New York, 1969; pp 107-123 (translator, R. A. Chalmers). Experimental Statistics; (Handbook 91); National Bureau of Standards, United States Department of Commerce: Washington DC, 1963; pp T-5, T-7, T-9. Alford-Stevens, A. L. Analyzing PCBs. Environ. Sci. Technol. 1986,20, 1194-1199.

Received for review February 28, 1992. Revised manuscript received June 22, 1992. Accepted July 6, 1992.

Partitioning of Polycyclic Aromatic Hydrocarbons from Diesel Fuel into Water Linda S. Lee,*,+Mats Hagwall,$Joseph J. Delflno,§ and P. Suresh C. Raotvg

Soil and Water Science Department, 2169 McCarty Hall, University of Florida, Gainesviile, Florida 3261 1-0290, Chemical Engineering Department, Royal Institute of Technology KTH, Stockholm, Sweden, and Environmental Engineering Sciences Department, University of Florida, Gainesvlile, Florida 326 1 1 An experimental evaluation of a model derived using Raoult’s law convention for activity coefficients and hypothetical supercooled liquid solubilities is presented for the partitioning of eight polycyclic aromatic hydrocarbons (PAHs), all solids in their standard state, between four diesel fuels and water. The UNIFAC (UNIQUAC functional group activity coefficient) model was used to calculate the likely nonidealities resulting from interactions between components in the diesel fuel. Both the UNIFAC model simulations and the observed diesel-water partition coefficients suggest that, for the PAHs examined, the extent of nonideality is sufficiently small (i.e., errors remain within a “factor of 2”), such that Raoult’s law may be acceptable for most field-scale applications. Introduction Diesel fuel is a complex mixture of the intermediate distillates from crude oil. Diesel fuels are composed of ~

‘Soil and Water Science Department, University of Florida. f Royal Institute of Technology KTH. 8 Environmental Engineering Science Department, University of Florida. 2104

Environ. Scl. Technol., Vol. 26, No. 11, 1992

approximately 40% n-alkanes, 40% iso- and cycloalkanes, 20% aromatic hydrocarbons, and a few percent isoprenoids and sulfur, nitrogen, and oxygenated compounds (1). However, the composition of a specific diesel fuel is dependent on the source of crude oil, the degree of chemical modification such as cracking and re-forming, and the separation method. Diesel fuels may also contain several “quality enhancer” additives, such as diesel ignition improvers, stability improvers, corrosion inhibitors, multipurpose additives, and surfactants (2). Of the components in diesel fuels, compounds of lower molecular mass will tend to evaporate and degrade more readily, leaving the higher molecular mass components; therefore, understanding the release of polycyclic aromatic hydrocarbons (PAHs) from diesel fuel into groundwater and surface waters is important. The primary objective of this study was to measure diesel fuel-water partition coefficients for several PAHs that are crystalline in their pure form and to evaluate the utility of employing Raoult’s law convention for activity coefficients in conjunction with supercooled liquid solubilities to estimate aqueous-phase concentrations of PAHs in equilibrium with diesel fuel.

0013-936X/92/0926-2104$03.00/0

0 1992 American Chemlcai Society

Table I. Selected Physicochemical Properties for the PAHs Investigated compound

mpa ("C)

naphthalene 1-methylnaphthalene 2-methylnaphthalene acenapthene fluorene phenanthrene anthracene fluoranthene

80.2 -22 34 93 116.5 100 216.3 107

Swb(mg/L) 32 27 26c 3.42 1.9

MWa 128.2 142.2 142.2 154.2 166.2 178.2 178.2 202

11 0.07 0.27

log KOwdve 3.35 (0.1) 3.87 (0.2) 4.00 (0.2) 3.92 (0.25) 4.18 (0.20) 4.52 (0.15) 4.5 (0.15) 5.2 (0.2)

10s S,,(mol/L)f -3.05 -3.728 -3.62 -3.98 4.03 -4.5 -4.49 -5.19

aVerschueren (14). bCrystal aqueous solubility at 25 OC (15). CMilleret el. (17). dSangster (18). 'Uncertainty listed by Sangster (18) given in parentheses. 'Supercooled liquid solubility calculated by assuming a constant ASr for PAHs (4, 5). 8Liauid at standard state.

Theory The release of a chemical from an organic liquid phase can be estimated from a liquid-liquid partition coefficient (Kd)for that chemical; Kd is defined as Kd = CdCw (1) where C, and C, are the molar concentrations of the chemical of interest in the organic and aqueous phases at equilibrium, respectively. For a diesel fuel, the partition coefficient will be designated here as Kd,. For liquid-liquid partitioning, thermodynamic equilibrium is defined by the equality of the chemical potentials in the aqueous and organic phases. This equality, in conjunction with the choice of pure (liquid) solute as the standard state and the Raoult's law convention for activity coefficients, results in the following expression at equilibrium: X,Y,* = XWY,* (2) where subscripts o and w denote organic and aqueous phases, respectively; x, and x, are the respective mole fractions of the chemical in the organic and aqueous phase; yo*is the activity coefficient of the chemical in the organic phase in equilibrium with the aqueous phase; and yw* is the activity coefficient of the chemical in the aqueous phase in equilibrium with the organic phase. From eq 2, the molar concentration of a solute in the aqueous phase (C,) can be approximated with the following assumptions: (1)the presence of other components in the aqueous phase is ignored; i.e., yw* is set equal to the aqueous-phase activity coefficient of the solute in equilibrium with the pure solute (7,); (2) the solute behaves ideally in the organic phase; Le., yo* is unity; (3) the of the pure liquid aqueous mole fraction solubility (S), solute is equal to l/y,; and (4) the solution is sufficiently dilute [i.e., moles of the solute are small relative to the total moles of solvent; C, = x,/ V , and Sl/V, = S,, where S1 is the aqueous solubility of the pure liquid solute (in mol/L) and 7, is the molar volume of water]. Application of these four assumptions yields c, = x,s1 (3) Therefore, the partition coefficient (eq 1)for a solute can be approximated as follows: Kd = co/xasl (4) For mixtures comprising a large number of constituents, each contributing a small fraction_to the total, x,/C, can be replaced by the molar volume (Vo, L/mol) of the organic phase. The molar volume can then be approximated by the ratio of the average molecular weight (MW,, g/mol) and density bo,g/L). The resulting expression for Kd is

Taking logarithms of both sides of eq 5, it is evident that

the inverse relationship between log Kd and log Slresults in a unit negative slope and an intercept that is dependent upon the molar volume of the organic phase (i.e., Mw, and Pa):

log Kd = -log Si - log (MWo/PJ (6) Derivation of eq 6 was based on a choice of the pure liquid solute as the standard state. Most of the PAHs investigated in this study are solids in their pure form; therefore, the hypothetical supercooled liquid solubilities of the solid solutes must be employed. The supercooled liquid solubility (SI)of a solute at a given temperature can be calculated directly from the solute's measured heat of fusion (AHf)and melting point (T,)(3) or, alternately, can be estimated by assuming a constant entropy of fusion (ASf = AHf/T,) for the PAHs of interest (4, 5). The utility of the relationship defined by eq 6 was successfully demonstrated for several gasolines by Cline et al. (6). Gasoline is composed of several branched-chain paraffins, cycloparaffins, alkanes, aromatic compounds, and small amounts of various additives. Results presented by Cline et al. (6)revealed that although gasoline is complex in composition, its partitioning behavior was essentially ideal. Note that none of the components investigated by Cline et al. (6)exhibit crystalline structure in their pure form. Here, we will examine the applicability of eq 6 for predicting PAH partitioning between diesel fuels and water. Materials and Methods Chemicals. The PAHs investigated are listed in Table I along with selected physicochemical properties. Standards were purchased from Fisher Scientific in the purest available grade and used without further purification. The solvent used was methylene chloride, Fisher grade Optima. The fuels were obtained from different local service stations in Gainesville, FL. They were dispensed into amber 1-L glass bottles (prerinsed with methylene chloride), dried with nitrogen gas, fitted with Teflon-lined screw caps, and stored in the dark at 4 "C. Batch Equilibration Technique. Two-hundred milliliters of diesel fuel was introduced to the surface of 2000 mL of distilled, deionized, organic-free water (a fuel-water ratio at 1:lO). Glass-stoppered 2-L Pyrex reagent bottles with Teflon-plugged stopcocks positioned approximately 5 cm from the base, designed by Thomas and Delfino (7), were used to equilibrate the diesel fuel-water mixture. The equilibration vessels were then placed in a Magni-Whirl shaker and agitated at room temperature (23 f 1 "C) for 24 h. A 30-min settling period was allowed, followed by withdrawal of the aqueous phase from the bottom of the equilibration vessel with minimal disturbance. Samples were stored in amber glass bottles with Teflon-lined screw caps until analysis. Base-neutral extractable compounds in the aqueous phase were characterized by EPA (Environmental ProEnvlron. Scl. Technol., Vol. 26, No. 11, 1992 2105

tection Agency) method 625. Triplicate 500-mL samples were placed in 1-L glass separatory funnels with Teflon stopcocks. A surrogate standard was inserted and the pH adjusted to >11with sodium hydroxide. The samples were extracted three times with methylene chloride by vigorous shaking for 2 min followed by a 10-min settling time before withdrawal of the methylene chloride. The recovery of the methylene chloride phase was approximately 100%. The extracts were dried through sodium sulfate and concentrated to 8-10 mL using a rotary evaporator and then to 0.9 mL by gentle volatilization with nitrogen gas. A 0.1-mL internal standard solution (naphthalene-de in methylene chloride) was added to the concentrated extracts, transferred to 2-mL crimp-seal vials, and refrigerated until analysis, typically within 3 days. Analytical Method. A Perkin-Elmer Model 8500 gas chromatograph/ion trap detector (GC/ITD) was used. The column was a Hewlett-Packard fused-silica capillary column (length 25 m, internal diameter 0.31 mm) with a 1.03-pm f i i of cross-linked 5% biphenyl methyl silicone. The temperature program for analysis of both the diesel fuel and the aqueous-phase extracts included a 1.5-min hold time at 50 OC, temperature ramping at 20 "C/min to 130 OC followed by a 3-min hold time, and a final temperature ramp at 5 OC/min to 263 OC. For the neat diesel, the final temperature was ramped to 300 OC to reduce accumulation of heavy molecular weight compounds on the GC column. Helium was used as carrier gas at a flow rate of approximately 1.0 mL/min. The ion trap detector was set at an electron energy of 70 eV. The detector scanned from 45 to 450 amu at a rate of 2 scans/s. The electron multiplier voltage was 1650 V, and the transfer line temperature from the GC was 280 O C . A standard calibration file was built for the eight compounds of interest ranging in concentration from 1to 500 ng on column. The calibration plots were linear within the concentration range investigated. The neat diesel samples were analyzed by injection of approximately 1 pL, after diluting with methylene chloride (1:lO) and spiking with an internal standard. The aqueous-phase samples were analyzed by injection of approximately 2 p L of the concentrated methylene chloride extract. All injections were made using the solvent flush technique, which includes thorough rinsing of the syringe with isooctane, drawing air to a mark, and then drawing of the extract prior to injection. This technique dilutes the sample further, but does not affect the analysis since no acquisition is made of the isooctane solvent. Estimation of Average Molecular Weight. The average molecular weight of the diesel fuels was calculated using the simulated distillation method adapted from Gehron (8). In this procedure, the centroid of the chromatogram (Le., when 50% of the area counts had eluted and had been detected) is located and the paraffin peaks closest to the centroid along with their fractions relative to the centroid are used to determine the average carbon number (n). The average molecular weight of the paraffii is then calculated by the empirical expression C,,HZn+*For example, if the centroid is determined to be just between C16and C1,, the average carbon number is 16.5, thus, the average molecular weight is 233. The average molecular weight of the paraffms is assumed to be valid for the d i e d fuel as a whole, since they make up a relatively large part of the total area of the chromatogram. Results and Discussion Diesel-Water Partitioning. Diesel fuel samples collected from five different stations in Gainesville, FL, represented a summer blend for five different brands. 2106

Envlron. Scl. Technol., Vol. 28, No. 11, 1992

cilscj"

117r18

C-p

l0I

c-19

I

1O:OO

15:OO

2O:OO

25:OO

30:OO

35:OO

Time (NIX:SEC)

Flgure 1. Chromatograms from the QC/ITDlnjectlons of neat dlesel fuel (A) and the corresponding aqueous phase (6)of diesel fuel DF 4.

Table 11. Concentration Range of Eight PAHs in Four Neat Diesel Fuels and the Corresponding Aqueous Phase compound naphthalene 1-methylnaphthalene 2-methylnaphthalene acenaphthene fluorene phenanthrene anthracene fluoranthene a

concn (mg/L) neat diesel aqueous phase 350-1500 2000-4000 3500-9000 100-600 350-900 100-1500 100-300 1.5-125

0.08-0.3 0.13-0.17 0.18-0.34 0.004-0.014 0.012-0.026 0.015-0.025 0.0004-0.002 LODa-0.0005

Limit of detection.

Chromatograms from the GC/ITD analyses of a neat diesel fuel and of the corresponding aqueous phase, shown in Figure 1A and B, are representative of all the diesel fuels investigated. The chromatograms for the neat diesel clearly show the homologous series of paraffins (Cl0-C2& which is distinctive for diesel fuel. In the chromatograms for the aqueous phase, peaks for the paraffms, due to their very low aqueous solubilities, should not be detected. Detection of high concentrations of the homologous series of paraffins in the aqueous phase would indicate contamination by the neat fuel (e.g., presence of microemulsions) during withdrawal of the water after shaking (7).Such contamination of the aqueous phase did occur with one of the diesel fuels; therefore, data for this fuel are not reported here. The concentration ranges of the PAHs in the neat diesel and the aqueous phase are shown in Table 11. PAH concentrations in the neat diesel fuels varied considerably between brands, but variations were well within 1 order of magnitude for the PAHs investigated with the exception of fluoranthene. Similar variations were observed in the aqueous-phase concentrations. Assuming that the diesel fuels investigated are representative of most diesel fuels, Table I1 provides a range of maximum values for PAH concentrationsthat might be present in the aqueous-phase

65.5

-

54.5

-

43.5

-

-

Ideal Line

xu 53 3s -5.5 -5 4.5 4 -3.5 -3 3Li -5.5 -5 d.5 A -3.5

-b

log [SI, moles/L] Figure 2. log K, values plotted vs log S, for eight PAHs along with the

Table 111. log Kda Values for Eight PAHs and the Average and Standard Deviation (SD) Observed among Four Diesel Fuels compound naphthalene 2-methylnaphthalene l-methylnaphthalene acenaphthene fluorene phenanthrene anthracene fluoranthene

log Kdw DF 1 D F 2 D F 3 D F 4

av log Kdw

SD (%)

3.67 3.62 3.71 3.71 4.41 4.29 4.44 4.49

3.68 4.42

1.1 1.8

4.31 4.21 4.36 4.33

4.3

1.4

4.35 4.45 4.60 5.15 5.32

4.53 4.48 4.69 6.27 5.29

4.46 4.41 4.61 5.38 4.64

4.68 4.52 4.78 5.36 5.61

4.62 4.55 4.78 5.20 5.60

2.9 1.2

1.9 1.8 7.4

leachate leaving a diesel fuel contaminated area. Average values for the diesel-water partition coefficients for eight PAHs are shown in Table I11 along with the overall average and standard deviation (%) observed among the four fuels. The deviation in log K d w values for the PAHs investigated is less than 3% in most cases. The larger standard deviation (7.4%) observed for the partitioning of fluoranthene is a result of the low K d w value obtained with fuel DF 2. This value is the result of a single measurement; two of the three replicates were below the limit of detection. In Figure 2, the measured log K d w values are plotted against their log S l for the eight PAHs along with the ideal line (solid line) calculated from eq 6 for each diesel fuel. The SIvalues used in Figure 2 were estimated by using the crystal solubilites (S,) given in Table I and assuming a constant ASf of 13.5 eu. This estimation of S,was preferred over a direct calculation from measured AHf and T, values given the similar ASf values reported for the PAH compounds investigated ( 4 , 5 ) and the variation or absence of reported AHf values. The average molecular

Meal line (solid line) calculated from eq 6 for each diesel fuel.

Table IV. Densities (Pdf) Measured at 25 OC and Average Molecular Weights (MW,,) Estimated Using a Simulated Distillation Method Adapted from Gehron ( 8 )for Four Diesel Fuels fuel DF 1 DF 2 DF 3 DF 4 average

Pdf

(g/L)

861 851 868 869 862

MWdf 232 226 225 226 227

weights (MWdf) for the four diesel fuels are shown in Table IV along with their liquid densities (p&) measured at 25 O C . For most PAHs in all four diesel fuels, the log K d w values lie near the ideal line, suggesting that the assumption of ideal behavior may be adequate for describing the partitioning of PAHs from diesel fuel to water. The confidence intervals shown in Figure 2 (bars) were estimated using an error propagation method (9)which incorporates the errors incurred in the analysis of both the neat fuel and aqueous-phase concentrations. Arrowheads reflect the few cases where the propagated error was larger than the average K d w value, as was the case for anthracene and fluoranthene. Note that both compounds were present in small quantities in the neat fuel and/or analytical problems were encountered in detecting small aqueous-phase concentrations. Several factors other than nonideal behavior could result in apparent deviations such as analytical uncertainty in Kdw,as well as errors incurred in the estimations of Sl (i.e., reported S, values and the use of a constant ASf value). Differentiating between nonideal behavior and experimental artifacta will be discussed later. Given the well-established and widely used data base for octanol-water partition coefficients (K,) and similarly in the molar volumes (Le., MWdf/pa) of the diesel fuels investigated, a general correlation between the measured log Kd, and log KO,values is expected. Regression of all Environ. Scl. Technol., Vol. 26, No. 11, 1092 2107

d'

I?

-

I"'

ruxhh -I

4

I

O

1

2

3 -2 Log S,(moles/L)

-1

0

I

3

log [Measured C, ug/L]

noun a.

PAH ~ 0 n - h h n )p n d mhwn w 7. pww h a ~ l u m a t w l h m e m m w e d h m 0 ~ ~ ~ m m 0 l o u a c r c l e l h a c , ~ m t h a 1 : 1 l l n e .Akokrdudsdaretha Flpe4. l D g K , v c l l w s ( O T ~ I ~ h y d o c a r b o n s ~ 00nMenca Intervals for both lhe meaaued and predlctsd ~ ~ ~ x ) n i r a - fmm UNIFAC model calculalbns and U'ta average lDg Kw valw tions. expsrfmentalty daermlnsd by cnw et SI. ( 8 )pbtted agalnst b g S, va1uo8abng wnh tha ideal Ika based on RBOUR'S law. (me rr& data presented in Figure 2 with the exception of the three fraction compOaltlon of m0 gasoline assumad for tha M I F A C rodel caIculal!ms Is shown In lhe Inset.) data points with large uncertainties (arrowheads) yields

the following relationship: log Kd, = 0.93 log K, 0.68

+

9 = 0.89 n = 29 The above empirid relationehip can only be used for fuels with molar volumes simiiar to those employed in this study; otherwise, eq 6 is the preferred approach for estimating log Kdwvalues. Estimation of Equilibrium Aqueous-Phase Concentrations. Reasonable agreement observed in the predicted and measured log Kdr v8 log St relationships for most PAHs (Figure 3) provides a first step in predicting maximum PAH concentrations that may be present in the aqueous leachate leaving a diesel fuel contaminated area. Using Raoult's law and assuming ideal behavior, the concentration of a constituent in the aqueous phase in equilibrium with the organic phase is proportional to the mole fraction of that constituent in the organic phase (see eq 3). Substituting eq 5 into eq 1gives the following equation for the equilibrium aqueous-phase concentrations:

c w = CdfMw&I/Pdf

(7)

where the subeeripts df and w refer to diesel fuel and water, respectively. In Figure 3, PAH concentrations predicted using eq 7 were converted to commonly reported units oCg/L) and plotted againat concentrations measured in the laboratory partitioning studies with the four diesel fuels. Also included in Figure 3 are the confidence intervals for both the measured and predicted concentrations. Measured concentration errors were estimated from the standard deviations observed in triplicate analyses of the aqueous phase; confidence intervals with arrows reflect limits of detection. Similarly, the errors associated with the predicted values were estimated from the standard deviations obtained from triplicate analyses of the neat diesel fuel, Le., the determination of . , C The confidence intervals given for the predicted C, in Figure 3 did not include errors incurred in estimating Mwdfor p,. We consider the correspondence between measured and predicted equilibrium aqueous-phase concentrations shown in Figure 3 to be very good. Assessment of Deviations from Ideal Behavior for Equilibrium Conditions. The relationship between Kdr and St assumed previously (eq 6) was based on the simplifying mumption of ideal behavior (Le., yo* = 1and y,* = .)y Several factors may cause deviations from the 2108 Em*on. Scl. Technol.. Vd. 26.

No. 11, 1992

assumed ideal behavior for diesel-water partitioning of PAHs. For example, negative deviations from the ideal lime could result from the presence of surfactants or emulsions or sufficient nonideality, while positive deviations can be expeded if equilibrium has not been reached, and apparent deviations (positive or negative) can result from uncertainty in parameter estimation. For a mixture which is complex in composition and behaves in a 'nonideal" fashion, the partition coefficient (Kd)between an organic liquid and an aqueous phase can be related to the aqueous solubility of the pure liquid (St) in the following manner (10): log Kd = -1% St - log (MWJPJ - log Y.' + log(Y,*/~.) (8) Comparison of eqs 6 and 8 suggests that any deviations due to nonideal behavior will arise from the last two tema on the right-hand Bide of eq 8. Banerjee (11)observed that the presence of other components in the aqueous phase had a minimal effect on solute activity; therefore, it was assumed that y.*/y, = 1,thus requiring only estimates of yo*. The UNIFAC model (UNIQUAC functional group activity coefficient) proposed by Prausnitz et al. (12)for estimating activity coefficients in liquid-liquid equilibria was employed to estimate yo*values needed in eq 8. In this model, a mixture of dflerent chemicals is treated as a mixture of functional groups constituting the components of the mixture. Interactions between functional groups in the mixture, and the likely nonidealities resulting from such interactions, are calculated in order to estimate the activity coefficient of a chemical for a specified phase. Interaction parametera required in the UNIFAC model were obtained from the most current update (13). Application of the UNIFAC model for assessing the potential for nonideality is presented for a gasoline and diesel fuel. Using the UNIFAC model, activity coefficients (yo*)of several aromatic compounds were estimated for an unleaded gasoline simulated to represent the relative compositions (see inset in Figure 4) reported in Cline et aL (6). The estimated yo*valuea were then used to predict log K, values (shown aa solid triangles in Figure 4) a c cording to eq 8. UNIFAC model calculations for the monocyclic aromatic compounds represented in Figure 4 (compounds 2-5) confum the experimental observations

5

3t

x- I

\

A UNIFAC

~

Diesel Fuel Composition Monocyclic aromatics benzene 4.42E.3 1 toluene 4.22E-2 2 ethylbenzene 5.24E-2 3 m.p,wqlene 7.95E-2 4 bimethylbenzene l.64E-1 Polycyclic aromatics 5 naphthalene 6 methylnaphthalenes 7 acenaohthene 8 fluorene 9 phenanthrene 10 anthracene 11 nuoranthene Alkanes walkane cyclohexane isoalkane aniline

7.36E-2 4.88E-1 2 . 3 4 ~ - 2 -5 3.56E-2 5.18~.3 1.06E-2 3.41 E-3

-4

-3

1

-2

Log S, (moles/L)

7.36E-3 3.03E-3 2.72E-3 7.6E-4

Figure 5. log Kdwvalues for several aromatic hydrocarbons resulting from UNIFAC model calculatlons plotted agalnst log SI values along with the Meal line based on Raoult’s law. (The mole fraction compe sklon of the dlesel fuel assumed for the UNIFAC model calculations Is shown In the Inset.)

of Cline et al. (6)that gasoline-water partition coefficients of several liquid hydrocarbons can be approximated by assuming ideal behavior. However, for compounds with increasingly more aromaticity and that are solids in their standard state (PAH compounds 9-11 in Figure 4), the UNIFAC model predicted some negative deviation from ideal behavior. Partition coefficients for these compounds were not measured by Cline et al. (6)as they are present only in small quantities in gasoline. Compared to gasolines, diesel fuels contain a larger fraction of low-solubilityPAHs. Therefore, it was of interest to see whether the UNIFAC model estimations of yo* for these PAHs resulted in deviations from ideality. The composition of the diesel fuel assumed in the UNIFAC model calculations is shown in Figure 5. The concentrations of the eight PAHs chosen were comparable to those found in the diesel fuels used in this investigation, the concentrations of monocyclic aromatic hydrocarbons used were based on analyses reported by Thomas and Delfino (7), and the mole fraction of water was selected on the basis of the maximum ASTM limiting requirement for diesel fuel (2). To simulate the alkane fraction of the diesel fuel, a representative compound for each alkane (n-, iso-, cycloalkane) was selected (see Figure 5 ) in proportion to those reported by Mackay et al. (1). The UNIFAC model calculations for the yo* values of the PAHs ranged between 0.99 for toluene to 1.16 for fluoranthene. The close proximity of the calculated log Kdwvalues (solid triangles in Figure 5 ) to the ideal line based on Raoult’s law for the simulated diesel fuel suggests that deviations from ideal behavior for PAHs smaller than fluoranthene may be negligible. These calculations suggest that deviations from the ideal line for the larger PAHs noted in Figure 2 cannot be attributed to soluteaolute interactions, lending support to analytical sources of error for the obw ~ e deviations. d Independent assessment of the potential for nonideal behavior emphasizes the need to account for experimental and analytical sources of errors when one is judging whether the deviation noted from the ideal line is indeed the result of nonideal behavior. Summary and Conclusions An experimental evaluation of a model based on ideal behavior was presented for the partitioning of eight PAHs between diesel fuel and water, and the results were com-

pared to data reported earlier for gasoline-water partitioning. The diesel fuel-water partitioning of several PAHs, all solids in their standard state, was well described by employing supercooled liquid solubilities and assuming ideal behavior. Good agreement between the observed partitioning of several PAHs and UNIFAC model calculations for a simulated gasoline and a diesel fuel further suggests that the extent of deviations from ideal behavior may be relatively small. Using Raoult’s law convention for solute activity Coefficients and assuming ideal behavior, the fuel-water partitioning of a compound can be approximated by simply estimating the molar volume of the organic phase. This information along with an estimate of the PAH concentration in the organic phase provides a first step in predicting PAH concentrations that may be present in the aqueous leachate leaving a fuel-contaminated area. Several site-specific hydrogeologic factors might lead to significant mass-transfer constraints for solute partitioning between diesel fuel and water. Such factors include random spatial variability in aquifier hydraulic properties, the patterns of residual fuel entrapment, and the source of fuel contamination (e.g., surface spill vs subsurface leaks). Under nonequilibrium mass-transfer conditions, the concentrations of diesel constituents detected in groundwater are likely to be smaller than those estimated using the equilibrium approach presented here. In contrast, larger concentrations might be observed in the presence of surfactants, emulsifiers, or cosolvents. The concentrations of PAHs in groundwater in equilibrium with diesel fuel, as well as gasoline, estimated by assuming ideal behavior may be considered an upper limit (within a factor of 2) for most field-scale applications. Acknowledgments

We appreciate the technical assistance of Mr. Itaru Okuda in the UNIFAC model calculations. Approved for publication as Florida Agricultural Experiment Station Journal Series No. R-02290. Registry No. Benzene, 71-43-2; toluene, 108-88-3; ethylbenzene, 100.41-4;m-xylene, 108-38-3;p-xylene, 106-42-3;0-xylene, 95-47-6; trimethylbenzene, 25551-13-7; naphthalene, 91-20-3; methylnaphthalene,1321-94-4;acenaphthalene, 83-32-9; fluorene, 86-73-7; phenanthrene, 85-01-8;anthracene, 120-12-7;fluoranthene, 206-44-0; cyclohexane, 110-82-7;aniline, 62-53-3.

Literature Cited Mackay, D.; Shiu, W. Y.; Chau, A.; Southwood,J.; Johnson, C. I. Environmental Fate of Diesel Fuel Spills on Land. Report for Association of American F&ilroads. Department of Chemical Engineering and Applied Chemistry, University of Toronto, 1985. Kirk-Othmer Encyclopedia of Chemical Technology, 3rd ed.; John Wiley & Sons; New York, 1977; Vol. 11, p 682. Valvani, S. C. J. Pharm. Sei. 1980,69, Yalkowsky, S.H.; 912. Yalkowsky, S.H.Ind. Eng. Chem. Fundam. 1979,18,108. Martin, E.; Yalkowsky, S. H.; Wells, J. E. J. Pharm. Sei. 1979, 68, 565. Cline, P. V.; Delfino, J. J.; Rao, P. S. C. Environ. Sei. Technol. 1991, 5, 914. Thomas, D. H.;Delfino, J. J. Ground Water Monit. Rev. 1991, 11, 90. Gehron, M. J. Advanced Mass Spectrometric Methods of Jet Fuel Analyis. Ph.D. Dissertation, University of Florida, 1988. Shoemaker, D. P.; Garland, C. W.; Steinfeld, J. I.; Nibler, J. W. Experiments in Physical Chemistry, 4th ed.; McGraw-Hill Book Co.: New York, 1980; p 46. Chiou, C. T.; Schmedding, D. W. Environ. Sei. Technol. 1982, 16, 4. Environ. Sci. Technol., Vol. 26, No. 11, 1992 2109

Environ. Sci. Technol. 1992, 26, 2 110-2 115

Banerjee, S. Environ. Sci. Technol. 1984, 18, 587. Prausnitz, J.; Anderson, T.; Grens, E.; Eckert, C.; Hisieh, R.; O’Connell, J. Computer Calculations for Multicomponent Vapor-Liquid and Liquid-Liquid Equilibria; Prentice-Hall, Inc.: Englewood Cliffs, NJ, 1980. Hansen, H. K.;Rasmussen, P.; Fredenslund, A. Ind. Eng. Chem. Res. 1991,30,2355. Verschueren, K. Handbook of Environmental Data of Organic Chemicals, 2nd ed.; Van Nostrand Reinhold Co.: New York, 1983. Little, A. D. Reference Constants for Priority Pollutants and Selected Chemicals. Reference 84204, Report to Wald, Harkrader, and Ross, Washington, DC, 1981.

(16) Yalkowsky, S.H.; Orr, R. J.; Valvani, S. C. Ind. Chem. Erg. Fundam. 1979,18, 351. (17) Miller, M.M.; Wasik, P.; Huang, G. L.; Shiu, W. Y.; Mackay, D.Environ. Sci. Technol. 1985, 19, 522. (18) Sangster, J. Phys. Chem. Ref. Data 1989, 18, 1111. Received for review March 18,1992. Revised manuscript received July 14, 1992. Accepted July 16, 1992. This research was partially supported by Electric Power Research Institute Contract RP-2879-7 in a cooperative effort between the Institute of Food and Agricultural Sciences and the Engineering and Industrial Experiment Station at the University of Florida.

Equilibrium Partitioning of Polycyclic Aromatic Hydrocarbons from Coal Tar into Water Linda S. Lee,” P. Suresh C. Rao, and Itaru Okuda Soil and Water Science Department, University of Florida, Gainesvllle, Florida 326 11

Partitioning of several polycyclic aromatic hydrocarbons (PAHs) from eight coal tar samples into water was measured. The measured partition coefficients were used to evaluate a model derived using Raoult’s law convention for activity coefficients and hypothetical supercooled liquid solubilities. Our analysis suggests that the extent of deviations from “ideal” behavior for coal tar-water partitioning is sufficiently small, similar to earlier reports for gasoline-water and diesel-water partitioning of PAHs. The concentrationsof PAHs in groundwater in equilibrium with these complex wastes, estimated from the model presented, may be considered as a reasonable approximation for most field-scale applications. The likely reasons for deviations from the ideal behavior are discussed, as are sources of analytical and computational errors. Introduction In the late 18009 and early 19oos,gas was manufactured from coal and oil for residential, commercial,and industrial uses. Manufactured gas plants (MGPs) were present in most major cities throughout the United States. The gas manufacturing plants generated a variety of process wastes such as tars, spent oxides, ash, sludge, ammonia liquors, and lampblack. The wastes generated from various methods of gas production were similar; however, the specific type and quantity of waste contamination at a given MGP site would be dependent on the feedstock used, the manufacturing process employed, and the time period over which the plant was in operation. In many cases, the wastes were left on-site in pits or containers, placed in nearby ponds or lagoons, or taken to off-site areas for land disposal. Such practices resulted in contamination of soils and groundwater at most former MGP sites. Coal tars make up a large portion of the hydrocarbon wastes generated at MGP sites. Eng and Menzies (1) reported that more than 11billion gallons of coal tar was generated in the United States during the period 1816-1947, but the disposition of several billion gallons is unknown and remains unaccounted. Coal tars are complex mixtures of a large number of hydrocarbons spanning a broad spectrum of molecular weights, with the concentrations of individual constituents varying significantly from one MGP site to another. The coal tar constituents of specific interest in this study are the polycyclic aromatic hydrocarbons (PAHs). These compounds have been detected at former MGP sites and are of particular concern 2110

Envlron. Scl. Technoi., Vol. 26, No. 11, 1992

due to their potential carcinogenic nature (2). Several of these compounds have already been included on the U.S. EPA list of priority pollutants. Near the source of contamination (presence of separate organic phase) at a coal tar disposal site, one of the primary processes controlling the release of organic chemicals is solubility. In the past, it has often been assumed that organic contaminant concentrations in the aqueous phase leaving a coal tar source would be equal to their corresponding pure compound aqueous solubilities. This may be a reasonable estimate if the source of interest was composed of a single contaminant; however, most complex wastes (e.g., coal tar, diesel, and gasoline) consist of mixtures of contaminants. These mixtures may be considered complex on the basis of the number of chemicals that constitute the mixture. On the other hand, complexity of a mixture can be defined by considering how the properties of the mixture deviate from some “ideal” behavior, regardless of the number of components. The former view corresponds to a mixture being complex in composition, whereas the latter implies complexity in behavior. The important point is that a mixture can be complex in composition without being complex in behavior and vice versa. The properties of an organic mixture complex only in composition are determined by the properties of its pure components and their concentrations in the mixture. This implies that the chemicals of interest behave ideally in the matrix containing them. Under these conditions, the concentration of a chemical in the aqueous phase is proportional to the mole fraction of the chemical in the organic phase corresponding to Raoult’s law. With the stated assumptions, the concentrations of a chemical in the aqueous phase in contact with a complex mixture can be predicted using the following simplified expression based on Raoult’s law (3, 4):

c,

= x,s1

(1)

where C, is the chemical’s concentration in the aqueous phase (mol/L) in equilibrium with the organic phase, SI is the aqueous solubility of the pure liquid chemical (mol/L), and x , is the mole fraction of the chemical in the organic phase. For compounds that are solid in their standard state, the hypothetical supercooled liquid solubility (SI)should be used (3). The applicability of Raoult’s law has been shown for several mixtures of organic chemicals, including gasoline ( 4 ) and diesel fuel (31, for predicting aqueous-phase concentrations.

0013-936X/92/0926-2110$03.00/0

0 1992 American Chemical Society