Environ. Sci. Technol. 1993, 27, 2831-2843
Coal Tar Dissolution in Water-Miscible Solvents: Experimental Evaluation
Catherine A. Peters' and Richard G. Luthy
Department of Civil Engineering, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213
* To whom correspondence should be addressed at her present
address: Environmental and Water Resources Engineering, Department of Civil and Environmental Engineering, The University
of Michigan, Ann Arbor, Michigan 48109-2125.
that could be applied either in an in situ injection/recovery
system or in an aboveground treatment operation. The
primary objective of this work was to investigate the extent
to which organic water-miscible solvents increase the
solubility of coal tar and its constituent compounds. This
work was part of a larger project, presented in Luthy et
al. (7),aimed at investigating the feasibility of in situ
solvent extraction for remediation of coal tar contaminated
sites. Other aspects of the project included examination
of the mass transfer limitationsto insitu solvent extraction
of contaminated soils (8) and large-scale subsurface
modeling of an in situ solvent extraction process to explore
deployment options and estimate cleanup times (9).
The primary challenge in studying coal tar NAPLs is
that they are mixtures of hundreds of compounds, primarily polycyclicaromatic hydrocarbons (PAHs). It would
be an insurmountable task to completely characterize the
equilibrium-phase compositions of coal tar/solvent/water
mixtures by measuring and describing the partitioning of
every individual compound. The experimental data
required to calibrate models describing phase equilibria
for a mixture increases very sharply as the number of
components in the mixture increases. Even for a ternary
mixture, the experimental effort required is almost 1order
of magnitude larger than that needed for a binary mixture
(IO). Furthermore, of the many constituent compounds
that make up coal tar, only a portion can be identified and
quantified through chromatographic methods. The challenge, then, is to adequately describe the dissolution
behavior of a multicomponent mixture such as coal tar,
which itself cannot be fully characterized, without the
burden of enormous data requirements. The approach
used in this investigation involves a simplification, referred
to here as the pseudocomponent simplification, in which
coal tar is treated as a single component in a system with
two other components: solvent and water. Coal tar
solubility was explored by studying the equilibrium phase
compositions of two-phase liquid mixtures of coal tar,
solvent, and water for several water-misciblesolvents. The
coal tar pseudocomponent simplification allows the composition of the two immiscibleliquid phases to be described
in terms of volume fractions of only three components.
This simplification facilitates experimental analysis and
data representation using ternary phase diagrams and
makes thermodynamic modeling tractable (II), as is
presented in a forthcoming paper (12).
This paper addresses four specific objectives. First,
detailed composition analyses of the coal tar used throughout this project are presented. Second, the validity of the
pseudocomponent simplification for semi-empirical thermodynamic modeling is explored using composition analyses of coal tar before and after extraction and using
measurements of the partitioning behavior of three PAH
compounds. Third, an estimate is made of the bulk
solubility of coal tar in water, which serves as a baseline
for comparison with experiments using solvents and
indicates the possible extent of groundwater contamination
in terms of all constituent compounds. Finally, phase
0 1993 American Chemical Society
Environ. Sci. Technol., Voi. 27, No. 13, 1993 2831
Coal tar, a dense nonaqueous phase liquid (NAPL), is a
common subsurface contaminant at sites of former manufactured gas plants. A proposed remediation technology
is water-miscible solvent extraction, which requires understanding of the effect of water-miscible solvents on the
solubility of coal tar. This study investigated this effect
and the extent to which multicomponent coal tar could be
represented as a pseudocomponent in thermodynamic
modeling. The coal tar used in this study showed a
predominance of polycyclic aromatic hydrocarbons with
no single compound accounting for more than 4% (wt).
The bulk solubility of coal tar in water was estimated to
be 16 mg/L using composition data and Raoult's law
assumption for aqueous solubility. For three solvents,
n-butylamine, acetone, and 2-propanol, equilibrium phase
compositionsof two-phase coal tar/solvent/water mixtures
were experimentally determined using radiolabeled materials and are presented as ternary phase diagrams.
Results showed n-butylamine to be a good water-miscible
solvent for coal tar dissolution. The validity of thermodynamic modeling of coal tar as a pseudocomponent was
explored by examiningthe liquid-liquid solute partitioning
of naphthalene, phenanthrene and pyrene and by assessing
the effect of solvent extraction on coal tar phase composition. It was found that coal tar partitions as a pseudocomponent in systems with appreciable solvent, but not
in systems with only coal tar and water.
Introduction
Today there is growing concern about nonaqueous phase
liquids (NAPLs), a class of subsurface contaminants that
are immiscible in water (I). Coal tar is a NAPL that is
denser than water and often very viscous. Subsurface
contamination with coal tar exists today as a result of
uncontrolled disposal of process residuals at former
manufactured gas plant (MGP) sites. The manufactured
gas industry ended during the 1950sdue to the widespread
use of natural gas and the exploitation of petroleum.
Groundwater contaminationat MGP sites persists decades
later because of the slow, continuous dissolution of
constituent compounds from subsurface coal tar (2-5).
There are as many as 1000MGP sites in the United States
and likely more (6). Numerous MGP site investigations
have verified the presence of coal tar and subsequent
groundwater contamination,but cleanup efforts have been
only sparsely applied. Conventionalremediation methods,
such as direct coal tar pumping or groundwater pumpand-treat, have proven to be of limited practical use to
achieve low residual concentrations, as is discussedin detail
elsewhere (5, 7).
The use of water-miscible solvents for the extraction of
coal tar from contaminated soils is a soil treatment option
0013-936X/93/0927-2831$04.00/0
equilibria of coal tar/solvent/water systems based on
experimental measurements of water and solvent partitioning are presented in the form of ternary phase
diagrams.
nent. The premise that the composition of the dissolved
coal tar is similar to that of the undissolved coal tar, means
that
Theory
Pseudocomponent Simplification. The characterization of phase equilibria of complex mixtures can be
accomplished in a number of ways. Petroleum products
are often characterized using boiling point curves in which
the mixture is thought of as a continuum of infinitesimal
fractions of pseudocomponents (13). Researchersstudying
coal tar or other mixture NAPLs have, for the most part,
described dissolution in terms of individual compounds
(14-19). This approach is necessary when assessing
groundwater contamination because cleanup standards
are specified for individual compounds, usually those on
the priority pollutant list. The individual compound
approach was not useful for this project since the objective
was to assess the bulk solubility of coal tar based on the
representation of the entire coal tar mixture as a pseudocomponent.
For most water-miscible solvents and over a large range
of compositions, mixtures of coal tar, solvent, and water
will separate into two immiscible liquid phases, which are
referred to here as the “coal tar phase” and the “solvent/
water phase”. The coal tar phase consists primarily of
undissolved coal tar and small amounts of solvent and
water incorporated into this organic phase. The solvent/
water phase consists of solvent, water, and dissolved coal
tar. The coal tar pseudocomponent comprises all the
constituent compounds in coal tar, which can be thought
of as everything in the system except the solvent and the
water.
Experimental data are used to determine parameters of
a semi-empiricalthermodynamic model of ternary liquidliquid equilibrium (LLE) (11,12). Fitted model parameters for the coal tar pseudocomponent can be thought of
as describing the equilibrium behavior of a “coal tar
molecule”, representing the collective behavior of all the
constituent PAH compounds. Conceptually, the pseudocomponent representation is valid because of the
chemical similarity of the PAH compounds in coal tar
relative to the two other components of the system. That
is, the molecular interactions between two coal tar
constituents are much more similar than the molecular
interactions between either of these compounds and
solvent or water. Strictly speaking, for model parameters
to truly represent the collective thermodynamicproperties
of the coal tar pseudocomponent, the coal tar component
in each phase must be identical in composition, Le.,
comprised of the same distribution of constituent compounds. The extent to which this is true for a given coal
tar/solvent/water system depends on the degree of similarity of the partitioning behavior of the individual
compounds. Considerthe mixing of coal tar with a solvent/
water solution with the subsequent formation of a twophase system in equilibrium. Each coal tar constituent
compound, i, partitions into the solvent/water phase to an
extent described by its concentration, Ctw [mg/Ll. Normalizing C;w to the total concentration of dissolved coal
tar constituents, CC:”, results in a term that is indicative
of the abundance of i relative to the total pseudocompo2832 Envlron. Sci. Technol., Vol. 27, No. 13, 1993
where the superscripts sw and ct denote the solvent/water
and coal tar phases, respectively. Rearranging eq 1gives
the ratio of the concenthe partition coefficient, Kctlawi,
tration of i in the coal tar phase to the concentration in
the solvent/water phase:
nct
F n c t
for all i
Thus, an important implication is that for a given mixture
the partition coefficients of all coal tar constituent
compounds must be similar to each other and similar to
the overall partitioning of the coal tar pseudocomponent.
Semi-empirical thermodynamic modeling of ternary coal
tar/solvent/water systems is strictly valid only for systems
for which eq 2 is true, and the extent to which predictions
can be made in composition regions beyond where experimental data were used for calibration is determined
by the extent to which eq 2 is true for a wide range of
system compositions.
This concept was explored experimentally using the
solvent n-butylamine, which has been identified as a good
water-miscible solvent for coal tar dissolution (7). First,
solute partitioning tests were done for three coal tar
constituent compounds: naphthalene, phenanthrene, and
pyrene, which represent compounds with a range of
aqueous solubilities. Second, the effect of extraction with
n-butylaminelwater solutions on the composition of the
coal tar phase was studied using quantitative chromatographic analyses of extracted coal tar samples.
Coal Tar Solubilityi n Water. The aqueous solubility
of constituent compounds from coal tar into water has
been discussed in recent years (14-19) in an effort to
understand groundwater contamination at coal tar sites.
It has been found (e.g., ref 15) that predicting aqueous
solubilities of coal tar compounds using the Raoult’s law
assumption of ideality and an approximation for pure
liquid aqueous solubilities based on heats of fusion offers
reasonable agreement with experimental measurements.
This approach was adopted for this work for the purpose
of estimating the bulk solubility of coal tar as a pseudocomponent.
Equilibrium solute partitioning in liquid-liquid systems
is characterized by equal fugacities of the compound in
each liquid phase (20). If it is assumed that the solute
behaves ideally in both the aqueous and coal tar phases,
the equilibrium relation for solute i becomes (21)
xp = xictx iWL
(3)
where x p is the mole fraction of solute i in the water phase,
xpt is the mole fraction of i in the coal tar phase, and
xy‘ is the mole fraction equivalent of the aqueous
solubility of pure liquid i. Assuming that the aqueous
phase is sufficiently dilute such that the volume of the
solution is approximately equal to that of pure water (ZZ),
the aqueous concentration expressed as a mole fraction is
proportional to mass concentration. An equation similar
to eq 3 can be written
cp = "FtS?
(4)
where Cy is the mass concentration of i in the water phase
[mg/Ll, and S? is the aqueous solubility of pure liquid i
[mg/Ll. Coal tar constituent mole fractions were computed by x p = ( w t % j / l O O ) (MWdMWi), where wt% j is
the weight percent of i in the coal tar, MWCtis the average
molecular weight of the coal tar, and MWi is the molecular
weight of compound i.
For many coal tar compounds, SF is a hypothetical
quantity since these compounds are solids in the pure
state at ambient temperatures. An expression relating
SF to Sy, the pure solid aqueous solubility at the system
temperature, is derived from the thermodynamicsof solidliquid equilibrium where the standard state in the liquid
phase is defined as the pure subcooled liquid at the
temperature of the solution (20). Applying this relation
to eq 4 results in
(5)
where the term on the right is the ratio of the pure
component fugacities in the subcooled liquid and the solid
states. Fugacity ratios are often available in the literature
(e.g., ref 23) or can be approximated by an expression that
accounts for the free-energy change between the liquid
and the solid state (20),using a constant entropy of fusion
for organic compounds (24). The bulk coal tar dissolved
concentration is the sum of the concentrations of all the
dissolved species, i.e., C: = Cin4_1Cy,where m is the total
number of compounds in coal tar. Substituting for CT
from eq 5
m
an expression is derived to compute the solubility of bulk
coal tar in terms of individual component properties and
their relative abundances.
Methods
Coal Tar, The coal tar used for all laboratory experiments was a sample of the free-flowing liquid tar residing
in the subsurface at the former manufactured gas plant
site in Stroudsburg, PA. Several published reports
describe the chemical and physical characteristics of this
coal tar (25-28). A t this site, it is possible to collect liquid
coal tar by pumping from a NAPL pool in a stratigraphic
depression in the confining layer. Coal tar was collected
(7, 11) from one of several existing wells that had been
installed for remediation of the site through coal tar
pumping. This coal tar sample was well-suited for
laboratory investigations because it is a thin liquid and is
virtually free of dirt and water bubbles.
Materials. The n-butylamine, acetone, and 2-propanol
solvents were ACS grade from Fisher Scientific Co.
Deionized water was used for solvent/water solutions.
Solute partitioning and coal tar/solvent/water phase
equilibria were determined using radiolabeled techniques.
This provided a means of observing the behavior of
individual compounds or system components without the
need for chromatographic methods, which are difficult
and time-consuming for coal tar because of the large
number of similar hydrocarbon compounds. 14C-Labeled
naphthalene, phenanthrene, and pyrene were obtained
from Amersham Corp., with specific activities ranging from
30 to 60 pCi/mg and chemical purities of >98%. The 14Clabeled solutes were stored in stock solutions by rinsing
the glass ampules with methanol. Stock solution concentrations ranged from 20 000 to 60 000 dpm/pL (dpm
refers to disintegrations per minute in which 2.22 x lo6
dpm = 1pCi). This was sufficiently concentrated so that
when used to prepare coal tar radioactive stock solutions,
methanol accounted for less than 1% of the total system
volume. With this small amount, there was little concern
for cosslvency effects on constituent solubilities (29).
14C-Labeled solvents, n-butylamine, acetone, and 2-propanol, were purchased from the Sigma Chemical Co. in
specific activities ranging from 2.8 to 8.2 mCi/mmol with
chemical purities of >98%. Stock solutions were prepared by diluting the radioactive materials in pure,
unlabeled solvents, resulting in concentrations ranging
from 90 000 to 200 000 dpm/pL. Tritiated (3H)water was
purchased from NEN Research Products of DuPont Co.
with a specific activity of 25 mCi/g (5.5 X 1O1o dpm/g). A
tritiated water stock solution was prepared by diluting by
a factor of 100.
Coal Tar Composition Analysis. Composition analyses that were performed on the Stroudsburg coal tar
sample included analysis of volatile aromatic compounds,
chromatographic analyses of polycyclic aromatic hydrocarbons (PAH), and a molecular weight determination.
Number-average molecular weight determinations were
done by Galbraith Laboratories, Knoxville, TN, on two
replicate coal tar samples using vapor pressure osmometry.
This procedure (30) is based on the relationship between
the vapor pressure of a solution relative to that of pure
solvent and the molar concentration of solute in the
solution. A Knauer-Dampfdruck osmometer was used
with toluene as the solvent.
Volatile aromatic compounds were extracted with
methanol according to EPA method 8020. The methanol
extract was analyzed accordingto EPA method 5030,using
helium for purging and as the carrier gas in the gas
chromatograph (GC). The GC had a 105-m VOCOL
capillary column by Supelco Co. The injection temperature was 240 "C, and the carrier gas flow rate was 6 mL/
min. The temperature program was 10 min at 45 "C, to
200 "C at 4 "C per min, and held for 8 min. Detection was
achieved by photoionization, at a temperature of 250 "C.
Quantification was done using fluorobenzeneas a surrogate
standard compound. A five-point calibration was done
over the 10-100 ppb range.
The primary chromatographic analysis of PAH compounds was done in collaboration with the Coal Science
Division, Pittsburgh Energy Technology Center (PETC),
Pittsburgh, PA. A 10-mgsample of coal tar was dissolved
in methylene chloride to make a l-mL solution. A sample
of this was injected into a 5988A Hewlett-Packard gas
chromatograph-mass spectrophotometer (GCIMS) with
a 35 m X 0.2 mm SB-Phenyl-5 column with 0.3 pm film
thickness. The injection temperature was 300 "C, the
helium carrier gas had a linear velocity of 32 cm/s at 30
"C, the ionization potential was 70 eV, and the voltage
was 2551 V. The program was 3 min at 30 "C, to 320 "C
at 4 "C per min, and 5 min at 320 "C. Thirty-five
compounds were identified by mass spectra and retention
Environ. Sci. Technol., Vol. 27, No. 13, 1993
2833
indices from the literature. Fourteen compounds were
quantitatively determined using individually run calibrations. Benzo[blthiophene was added to the coal tar as an
internal standard to check instrument response. For the
remaining 21 identified compounds, an average response
factor was calculated from the 14 calibrated runs and used
to convert peak area into mass injected to estimate the
weight percents.
Supplementary analysis was done in collaboration with
the Analytical Section of the Research and Development
Division of Texaco, Inc., Beacon, NY. Prior to GC/MS
analysis, the coal tar was fractionated using ASTM method
D2007, in which characteristic groups were quantified in
weight percent using clay-gel adsorption chromatography.
This step was done to make a crude separation of the
PAH compounds residing in the aromatic fraction from
the other coal tar compounds, such as heterocyclics,
oxygenated compounds, and very high molecular weight
substances. This facilitated the GC/MS analysis, however,
the lack of specificity in the operational definitions
prescribed by this procedure limits the usefulness of the
analysis, and the crude separation procedure may have
caused some inaccuracies in the quantification step. The
aromatic fraction of the coal tar was diluted 1 0 0 ~in
methylene chloride and injected into a GC/MS with a 30
m X 0.25 mm SPB5 capillary column. The program was
5 min at 100 "C and then up to 300 "C at 5 "C per min.
Mass spectral analysis was by positive electron ionization,
between 3 and 250 amu. Approximate weight percentages
were estimated for the groups of compounds identified
using a single internal standard, assuming identical
responses and using relative peak areas.
GUMS analyses of coal tar samples that had been
extracted with n-butylaminelwater solutions were also
performed by the PETC laboratory. Coal tar samples were
equilibrated with n-butylaminelwater solutions in 500mL separatory funnels using a coal tar-to-solvent/water
solution volume ratio of 1:4. For two of the samples, the
coal tar was extracted once. The solvent/water solutions
used were 20% (vol) n-butylamine/80% water and 40%
(vol) n-butylamine/60% water. For the third sample, the
coal tar was extracted twice sequentially, each time with
fresh 40 5% n-butylamine/water solution, to observe a trend
in composition change upon sequential extraction. Over
a 24-h period, the mixtures were gently agitated intermittently to prevent emulsion formation and then settled
for 24 h to allow phase separation. Samples of the
extracted coal tar phases were collected from the bottom
of the separatory funnels. GC/MS analyses were performed in the manner described above for the PETC
analysis of the original coal tar. The amounts of dissolved
n-butylamine, estimated using thermodynamic LLE model
predictions, were found for all three extracted samples to
be approximately 5% of the weight (11). The GC/MS
results were corrected to represent weight percent relative
to the coal tar portion only, allowing an assessment of the
change in mass distribution of PAH compounds.
Solute Partitioning in Coal Tar/Water Systems.
For each of the three solutes (naphthalene, phenanthrene,
and pyrene), coal tar was spiked with several microliters
of radioactive solute stock solution giving 100 000-900 000
dpm/mL. Common procedures for LLE experiments
involve mixing the two liquids in a vial and then agitating,
settling, and analyzing each phase. Variations on this
approach were tried for coal tadwater systems, but these
2834
Environ. Scl. Technol., Vol. 27, No. 13, 1993
were found to produce erroneously high solute concentration measurements in the aqueous phase, possibly due
to erroneous sampling of either a floating organic film or
a microemulsion of coal tar micelles in the water (11). A
more successful experimental method was designed to
eliminate the free-flowing coal tar phase. IMPAQ RG20
porous silica gel beads with pore diameters of 200 A,
supplied by the PQ Corp., were used. The beads were
dehydrated by heating at 800 "C for 1h and then cooled.
A 1-mL sample of the coal tar radioactive stock solution
was imbibed into 5 g of beads in a glass vial with a Teflon
septum and mixed on an orbital rotator for 24 h. This
amount of coal tar was sufficient to discolor the beads, but
not enough to saturate them. This provided ample
interfacial surface area for mass transfer and eliminated
the formation of coal tar phase emulsions. The coal tar
imbibed beads were mixed with water in 50-mL centrifuge
tubes, with calcium chloride to aid in the settling of any
particulate matter. Equilibration times ranging from 24
h to 1month were used to test for kinetic hindrances to
solute dissolution from bead pore spaces, but no significant
differences in aqueous concentrations were observed. In
all, four replicate experiments were performed for naphthalene and phenanthrene, and seven were performed for
pyrene. The vials were centrifuged. A several milliliter
sample, taken from the top of the vial through the septum
using a syringe, was passed through a 0.2-pm Teflon filter
and, discarding the first milliliter to precondition the filter,
1-mLvolumes were expressed into 20-mL scintillation vials
containing 15 mL of Ultima Gold, Packard Instrument
co.
The concentration of 14C-labeledsolute in the vial was
measured using a Beckman 5000 TD liquid scintillation
counter. The channel window was set to record events
with pulse heights from 0 to 670. Quenching was corrected
automatically with the H# method and an internally stored
quench curve generated from 14C standard solutions,
obtained from Amersham Corp. Each test vial was counted
twice and for a sufficient time such that the 2a error in
dpm was less than 1% . Measurements were corrected for
background radiation which averaged 40 dpm in the
laboratory. The random coincidence monitoring (RCM)
option was used to indicate samples with high numbers
of light-producing events other than radiation, which was
problematic especially for coal tar containing samples. To
reduce the measurement error caused by these added
counts, the scintillation vials were stored in the dark for
at least 24 h, until the RCM %, the percentage of
nonradiation events relative to total light-producing
events, decreased to less than 1% .
For each experiment, the radioactivity concentration
in the water phase, (dpm/mL)w was measured directly.
The radioactivity concentration in the coal tar phase,
(dpm/mL)Ct,was estimated from knowledge of the total
radioactivity in the system, dpmT, and the volume of coal
tar, VCt, since the amount of dissolved solute is negligible
relative to the undissolved solute. The coal tar/water
partition coefficient, Kctlw,was computed by
Kct,w =
Cct - (dpm/mL)ctN (dpmT)/VCt
- (dpm/mL)" - (dpm/mL)"
(7)
The solute's aqueous solubility, Cw, was computed from
the aqueous concentration of the radiolabeled solute
relative to the total radioactivity, and an estimate of the
total amount of the solute in the system, mT:
c w = (dpm/mL)w
dpmT
1ooomT
The volume fraction of component i in the
vE, is computed by
cy
phase,
(8)
where C W is in milligrams per liter with mT in milligrams.
mT was computed from the wt % of the compound in the
coal tar, the volume of coal tar, and the density of the coal
tar.
Solute Partitioning in Coal Tar/n-Butylamine/
Water Systems, In 50-mL centrifuge tubes, coal tar
spiked with 14C-labeledsolute was added to n-butylamine/
water solution. A range of solvent concentrations were
used. In all cases, the coal tar-to-solvent/water solution
volume ratio was 1:4. For each of the three solutes, at
each solvent concentration, replicate experiments were
performed. The mixture was brought to equilibrium at
25 "C over a period of 24 h. Since vigorous agitation led
to emulsion formation, the vials were agitated intermittently (II). After centrifugation, the solvent/water phase
was sampled with a syringe and expressed through a 0.2pm Teflon filter, discarding the first milliliter to precondition the filter. The filtering step removed any suspended
coal tar, giving a clear filtrate. Coal tar phases were
sampled by decanting the solvent/water phase from the
vial and injecting a syringe into the coal tar. The sampled
coal tar was checked for emulsification by expressing
through a fine needle; emulsions were evidenced by
expulsion of intermittent slugs of solvent/water phase.
Observations from such systems were not used. The coal
tar phase sample was diluted by a factor of 30-40 in
n-butylamine before adding to the scintillation counting
cocktail to lighten its color and reduce quenching for more
accurate liquid scintillation counting. Dilution was also
necessary for solvent/water phase samples that were very
dark due to significant coal tar dissolution. The solute
partition coefficient,Kdlsw,was computed similarly to Kdlw
(eq 7), except that for these experiments (dpm/mL)ct was
measured directly.
CoalTar/Solvent/Water Phase Equilibria. For the
three solvents, n-butylamine, acetone, and 2-propanol,
LLE of coal tar/solvent/water systemswas experimentally
determined for a range of compositions that resulted in
two-phase systems. To determine a single tie line of the
ternary phase diagram, an overall composition of coal tar/
solvent/water was chosen. Parallel experimental systems
were set up: one using 14C-labeledsolvent and one using
tritiated water. For the solvent partitioning experiments,
the solvent/water solution and a spike of 14C-labeled
solvent stock solution were mixed in a 35- or 50-mL
centrifuge tube such that the total level of radioactive
solvent, dpm:, was in the range of (5 X 106)-(9 x 107)
dpm. For the water partitioning experiments, 0.5-1 mL
of tritiated water stock solution was added to the solvent/
water solution resulting in a total water radioactivity,
dpm:, on the order of 5 X lo8 dpm. Coal tar was added,
and the vials were equilibrated, centrifuged, and sampled
in the manner described above. The solvent/water phase
samples for the vials containing 3H had to be diluted
because of the high dpm: needed for these experiments,
For liquid scintillation counting of these samples, the
channel window was set for pulse heights of 0-400. The
H#quench curve was generated from 3Hstandard solutions
from NEN Research Products of DuPont Co.
where (dpm/mL); is the measured concentration of
radioactive i in the cy phase, dpm' is the total radioactivity in the system, and Vi is the total volume of i in the
system in milliliters. The subscript i denotes the component, either water (w) or solvent (s), and the superscript
cy denotes either the solvent/water phase (sw) or the coal
tar phase (ct). The volume fraction of the coal tar
pseudocomponent in each phase is calculated by the
difference from unity:
v f c t = 1- vfs* - VfW*
Throughout, when the symbol ct is used as a superscript
it refers to the coal tar phase, and when used as a subscript
it refers to the coal tar pseudocomponent. Since both
phases were sampled, these calculations do not require
the relative phase volumes, which changed significantly
for high solvent systems.
Each experimentally determined tie line is the result of
a single or duplicate measurements of solvent and water
partitioning. The experiments were designed so that the
relative standard deviations (a,/x) from random error
and vr*, measurements were less than
associated with
2% (see ref 11). Because eq 10 is additive, the relative
standard deviation for random error in
is not approximated by a constant value; the absolute standard
is approximated by uu2 = ( c ~ ,+, ~
deviation for each
a2 )112,assuming Gaussian distributions for error terms.
UG
For very small values of $2, the random errors likely
have skewed distributions, so this analysis is only an
approximation.
Quantitative tie line data were difficult to obtain near
the plait points of the ternary phase diagrams due to
sampling difficulties which resulted because the two liquid
phases were very similar in composition and thus similar
in appearance, and often one of the phases was present in
a very small quantity. For overall compositions thought
to be near the plait point, an additional series of experimental systems was visually examined for heterogeneity
to identify compositions that are conclusively within the
two-phase composition region. Heterogeneity was observed by slight color differencesbetween phases, by light
reflected from the interphase meniscus,and by overturning
the vial to see a thicker coal tar phase clinging to the
glassware. Tests were continued, moving toward the
expected direction of the plait point, until a composition
was found beyond which two phases could not be visually
discerned. This point is referred to here as a two-phase
check point.
Water Solubility in Coal Tar. In experiments where
coal tar is mixed with water containing a spike of tritiated
water, the concentration of radioactivity in the coal tar
phase can, in principle, be used to compute the volume
fraction of water using eq 9. While this method was
successfully applied to systems containing solvents, it was
not practically applied to coal tar/water systems since the
amount of tritiated water in coal tar was at or below the
analytical detection limit due to the low solubility and to
the large dilution necessaryto count coal tar phase samples.
vc
vet
vet
Environ. Sci. Technol., Vol. 27, No. 13, 1993 2835
Naphthalene
1-Phenylnaphthalene
3-Methylphenathrene
2-Methylphenanthrene
2-Methylarithracene
4 H-Cyclopenta [def]phenanthrene
9-Methylphenanthrene
1-Methylphenanthrene
2-Phenylnaphthalene
9-Ethylphenathrene
2-Ethylphenathrene
Fluoranthene
Aceanthrylene
2-Methyl Naphthalene
1 -Methyl Naphthalene
Biphenyl
2-Ethyl Naphthalene
1-Ethyl Naphthalene
2,6-Dirnethylnaphthalene
Acenaphthylene
Acenaphthene
Dibenzothiophene
Phenanthrene
Anthracene
Pyrene
Benzo (c) Phenanthrene
Benzo (a)Anthracene
Chrysene + Triphenylene
Benzo (j) Fluoranthene
+ Benzo (b) Fluoranthene
Benzo (k) Fluoranthene
Benzo (e)Pyrene
Benzo (a) Pyrene
Perylene
lndeno (1, 2, 3-04 Pyrene
Picei-te
1
27
I
I
l
12
l
l
/
18
I
/
-
1
24
I
30
I
36
I
I
42
I
I
I
48
I
I
I
54
I
I
I
60
I
I
I
66
I
I
I
I
72
Time (min.)
Figure 1. Chromatogram of Stroudsburg coal tar with 35 peaks identified.
Table I. Classification of Stroudsburg Coal Tar into
Characteristic Groups Using ASTM Method D2007
wt%
classification
34
17
asphaltenes, the n-pentane insoluble fraction
polar compounds, material retained on adsorbent clay after
percolation of the samples in an n-pentane eluent
aromatics, material that passes through a column of adsorbent clay in an n-pentane eluent but adsorbs on silica gel
saturates, material in an n-pentane eluent that is not adsorbed on either the clay or silica gel
41
8
The amount of tritiated water required for precise
measurement would have made these experiments expensive and would have involved much higher than normal
radioactivity levels employed for routine laboratory work.
It was possible, however, to use information about the
precision of the experimental method to estimate an upper
bound for the water concentration in the coal tar phase.
Results and Discussion
Stroudsburg Coal Tar Composition. Results of the
characteristic group analysis of the Stroudsburg coal tar
are shown in Table I, giving weight fractions and operational definitions. The PAH compounds comprise the
aromatics fraction and are likely also to be present as very
high molecular weight compounds in the asphaltenes
fraction. The results from the PETC chromatographic
analysis of the Stroudsburg coal tar are shown in Figure
1. A total of 280 peaks were found, indicating the
complexity of the mixture. The results of the quantification of compound weight percents from all the chromatographic analyses are shown in Table 11. The compounds that were quantified using calibration standards
2836
Environ. Scl. Technol., Vol. 27, No. 13, 1993
are identified with CS; the others having been estimated
using average response factors. Summing up the weight
percents in Table 11, this composition analysis accounts
for just under half (46%) of the total coal tar weight, for
constituent compounds in the molecularweight range from
78 to 278 that produced peak areas sufficiently large and
distinct. This analysis cannot account for very high
molecular weight material not quantifiable by G U M S
techniques, for trace compounds such as between peaks
9 and 10 in Figure 1, or for compounds that co-elute
resulting in peaks that cannot be separated.
The predominance of PAHs is consistent with analyses
of other coal tars (3, 311, the primary group here being
naphthalenes accounting for 18% of the total weight of
this coal tar. It is noteworthy that no single compound
accounts for more than 4% of the coal tar weight. The
low concentrations of volatile aromatic compounds such
as benzene, xylene, and toluene (BXT), together accounting for just under 1wt %. This is consistent with analyses
of certain other coal tars (3) where the BXT fraction
accounts for about 1wt 5% , and is usually closer to 0.5 wt
%. Volatile aromatics were indeed produced at MGP
facilities, from the volatile fraction of the raw coal and
from the aromatic fraction of certain carburettor oils (3234). However, these compounds may not be present in
significant concentrations in subsurface coal tar because
light fractions may have been recovered as byproducts at
MGP facilities and because the relative dissolution and
migration in groundwater of these compounds is higher
than for other coal tar constituents. The absence of acidextractable organics, such as phenols, is consistent with
what is expected given the method of production that was
used at the Stroudsburg MGP. The process at this plant
employed anthracite coal in a carburetted water gas process, which did not produce large quantities of oxygenated
Table 11. Stroudsburg Coal Tar Composition from GC/MS
Analyses
molecular
weight
wt%
0.050
78
benzene
0.094
92
toluene
0.32
106
m-,p-xyleneb
0.32
106
o-xylene
0.41
106
o-xylene
2.16 (CS)II
128
naphthalene
3.75
142
2-methylnaphthalene
142
3.80 (CS)
1-methylnaphthalene
0.68
152
acenaphthylene
0.50 (CS)
154
biphenyl
1.52 (CS)
154
acenaphthalene
1.84 (CS)
156
2-ethylnaphthalene
0.45 (CS)
156
1-ethylnaphthalene
1.99 (CS)
156
2,6-dimethylnaphthalene
1.4
166
9H-fluorene
0.4
166
1H-phenalene
4.3
170
trialkylated naphthalenes
2.12 (CS)
178
phenanthrene
0.59 (CS)
178
anthracene
1.7
180
methyl-9H-fluorenes
0.2
184
4-methyl-l,l’-biphenyl
0.22 (CS)
184
dibenzothiophene
0.55
192
3-methylphenanthrene
0.43
192
2-methylphenanthrene
0.31
192
2-methylanthracene
0.57
192
4H-cyclopenta[deflphenanthrene
and 9-methylphenanthreneb
192
0.33 (CS)
1-methylphenanthrene
0.3
198
methyldibenzothiophene
0.30
202
fluoranthene
0.29
202
aceanthrylene
0.50
202
pyrene
0.29
204
1-phenylnaphthalene
0.22
204
2-phenylnaphthalene
0.17
206
9-ethylphenanthrene
0.20
206
2-ethylphenanthrene
2.3
206
dimethylphenanthrene
3.9
216
methylpyrene
0.31 (CS)
228
benzo[a]anthracene
0.4
228
acepyrene
0.27 (CS)
228
chrysene and triphenylene b
242
4.4
methylchrysenes
1.8
252
benzopyrenes
0.13 (CS)
278
picene
46.46
total
a CS, quantification was by individually run calibration standards.
* Elute together.
wastes as was often true with the coal carbonization
processes with bituminous coal (3).
The results of average molecular weight determinations
of two replicate coal tar samples were 209 and 211, giving
an average molecular weight for the Stroudsburg coal tar
of 210. This is relatively high compared to the compounds
in Table I1that have been quantified by GC/MS, indicating
that the majority of the mass that has not been accounted
for in the GC/MS analysis lies in the high molecular weight
range. The average molecular weight of the Stroudsburg
coal tar is low relative to the range of average coal tar
molecular weights of 230-1600 from a study of MGP site
tar residues (311, but that study was not limited to freeflowing, liquid coal tars as is present at Stroudsburg.
Additional data that characterize the Stroudsburg coal
tar are a viscosity of 9.93 CP (30 “C), determined by
capillary viscometry, and a density of 0.994 g/mL (30 “C),
determined using hydrometers (11).
Coal tars vary from site to site in terms of composition
and physical properties, resulting from production dif-
Table 111. Solubilities in Water-Phase (Experimental and
Raoult’s Law Prediction), Coal Tar/Water Partition
Coefficients (Experimental and Raoult’s Law Prediction),
and Octanol/Water Partition Coefficients for Three
Selected Solutes
log
C; (mg/L)
1%
(KD)
log
pre(Kctlw) y e Wow)
exptlo
dicted
exptP
dicted ref 23
3.7
3.37
3.81 f 0.03
3.3f 0.2
3.9
naphthalene
5.1
4.57
phenanthrene 0.68 f 0.06 0.16 4.49 f 0.04
5.6
5.18
PFene
0.08f 0.03 0.014 4.8f 0.2
0 Experimental values are shown with 3u (99% confidence)random
error estimates.
ferences as well as environmental factors, as is discussed
elsewhere (3, 11). Nevertheless, the composition of the
Stroudsburg coal tar is typical of liquid coal tars because
the constituent compounds represented are commonly
found in all coal tars and because the PAH portion is
predominantly composed of naphthalenes. This suggests
that the experimentaland modelingresults from this study
may be modestly extended for making predictions about
other coal tar sites having pumpable liquid coal tar.
Coal Tar Solubility in Water. Results of solute
solubility and partitioning in coal tar/water systems are
shown in Table 111. The experimental values for the
aqueous solubilities are shown with 3a (99% confidence)
estimates of the random errors based on repeated measurements, as well as the predicted aqueous solubilities
based on the Raoult’s law approximation (eq 5 ) . Also
reported in Table I11 are the measured coal tadwater
partition coefficientswith their estimated 3arandom errors
estimates (log units). Just as the Raoult’s law assumption
was used to derive an expression for Cy, an expression for
the partition coefficient can be derived (see, e.g., ref 15)
in terms of SF and the molar volume of the organic phase.
The Raoult’s law estimate of the partition coefficient KD
was computed for each solute using the average molecular
weight of 210 and a density of 0.994 g/mL to compute the
molar volume of the coal tar phase. These values and, for
the sake of comparison, literature values of the octanol/
water partition coefficients are included in Table 111.High
correlation between Kcnwand KO,has been demonstrated
by others (14, 17, 19) and is to be expected given the
correlation between partitioning of solutes in different
immiscible organic/water systems as is commonly described as the linear free-energy relationship (35).
Lee et al. (15) found that for several coal tar samples,
and for several PAH compounds, experimentalKdwvalues
were generally within a factor of 2 of predicted KD’s.In
logarithmic units, this corresponds to zt0.30. This agreement implies that even though the coal tar phase is a
complex mixture it can be modeled as ideal in the Raoult’s
law sense which assumes that the molecular interactions
are equivalent to those in a pure organic liquid. For this
investigation, the agreement between experimental and
Raoult’s law predictions are good for naphthalene, but
the measured Ciw and Kctlwof phenanthrene and pyrene
indicate higher aqueous solubilities tha predicted by
Raoult’s law. The difference between log KDand log Kctlw
for pyrene corresponds to a Raoult’s law estimate of the
partition coefficient five times greater than the measured
value. The discrepancies for phenanthrene and pyrene
are not likely due to an inaccurate coal tar molar volume
estimate since this would have resulted in a constant bias
Environ. Sci. Technol., Vol. 27, No. 13, 1993 2837
Table IV. Calculation of the Bulk Solubility of Coal Tar in Water (25 “C) from Estimates of Constituent Solubilities
Using eq 5.
S: (mg/L)
benzene
1.35 X 103
1780
toleuene
2.15 X 10-9
515
m-xylene
3.17 X 10-9
160
p-xylene
3.17 X 103
215
o-xylene
8.12 f 103
220
naphthalene
3.54 x 10-2
31
2-methylnaphthalene
5.55 x 10-2
25
1-methylnaphthalene
5.62 X
28
acenaphthylene
9.39 x 103
16.1
biphenyl
6.82 X 103
7
acenaphthalene
2.07 X 10-2
3.80
2-ethylnaphthalene
2.48 X 10-2
8
1-ethylnaphthalene
6.06 X 10-9
10.1
2,6-dimethylnaphthalene
2.68 X 10-2
1.7
9H-fluorene
1.77 X
1.9
trialkylated naphthalene
5.31 X le2
2.1
phenanthrene
2.50 X 10-2
1.10
anthracene
6.96 X 103
0.045
methyl-9H-fluorenes
1.98 X
1.09
2.28 X 10-9
4-methyl-l,l’-biphenyl
4.05
2-methylanthracene
3.39 f 103
0.03
1-methylphenanthrene
3.61 X 103
0.27
3.12 X 10-9
fluoranthene
0.26
5.20 X 103
pyrene
0.132
benzo[a]anthracene
2.86 X 103
0.011
chrysene
1.24 X
0.002
1.24 X 10-3
triphenylene
0.043
1.50 X
benzopyrenes
0.004
0.42
total
Pure compound aqueous solubilities and fugacity ratios are from ref 23.
X?
in log& for all three solutes. The discrepancies are likely
due to measurement error resulting from aqueous-phase
sampling difficulties in systems which may have an oily
floating phase and tendencies to form microemulsions.
K,, is used in this work solely for comparison with KdlSw,
the solute partition coefficient in coal tar/solvent/water
systems.
The coal tadwater partition coefficients vary over orders
of magnitude, which is indicative of the variation of
aqueous solubilities of the coal tar constituent compounds.
Thus, it is immediately obvious that the constituent compounds do not equally partition to the aqueous phase, and
eq 2 does not hold for a system with only coal tar and
water. This implies that the application of a thermodynamic model to describe the binary LLE of a coal tar/
water system is not strictly valid since the coal tar
component of the equilibrated water phase does not have
the same composition as the coal tar component in the
coal tar phase. A following section discusses the validity
of the pseudocomponent simplification for ternary systems
which include solvent.
Estimation of the bulk solubility of coal tar in water,
Cz,was accomplished using the predicted aqueous solubilities of constituent compounds based on Raoult’s law
(eq 6). Data for this calculation are shown in Table IV,
for compounds which had been quantified (Table 11) and
for which solubility data and fugacity ratios were available
from the literature (23). For co-eluting compounds, e.g.
m- and p-xylene, the weight fractions were taken to be
half of the total for both, effectively using an equallyweighted average aqueous solubilityfor these compounds.
For quantified groups with limited data available, such as
trialkylated naphthalenes, solubility and fugacity ratio
data for a reported compound in the group were used,
2838
Envlron. Scl. Technol., VoI. 27, No. 13, 1993
P/tS)pwe i
1
1
1
1
1
3.53
1.24
1
4.61
2.85
5.05
1
1
6.62
7.94
2.43
5.65
11.5
3.92
1.59
66.2
9.35
7.09
19.8
21.6
189
52.6
32.3
(mg/L)
2.4
1.1
0.51
0.68
1.8
3.9
1.7
1.6
0.70
0.14
0.40
0.20
0.061
0.30
0.27
0.27
0.16
0.024
0.085
0.015
0.0067
0.0091
0.0058
0.014
0.00068
0.00047
0.0028
0.0019
16.3
such as 1,4,54rimethylnaphthalene.These approximations do not contribute significantly to uncertainty in the
estimation of qv The aqueous solubilities of these
compounds sum to 16.3 mg/L, providing an estimate of
CW,. Converting to volume fraction, this corresponds to
uEt = 1.6 X 10-5. Only a portion of the coal tar has been
quantitatively characterized due to analytical limitations;
the sum of the mole fractions used in this calculation is
0.42, and the corresponding sum of the weight fractions
is 0.32. However, the majority of the compounds that are
not analyzable by GC/MS methods are high molecular
weight compounds which have very small aqueous solubilities. Their contribution to the sum in eq 6 is negligible.
For example, pyrene’s aqueous solubility is more than 2
orders of magnitude less than that of naphthalene. Also,
this calculation is limited by the availability of solubility
data for constituent compounds. Again, the more soluble
compounds are the ones for which data are available.
When estimating properties of mixtures, the approach
of selecting a representative compound is often used. The
premise of this method is that there is a single constituent
compound whose property is representative of the bulk
property of the mixture as a whole. While this method is
tempting because of its simplicity, the above calculation
indicates that it is not immediately obvious how a
representative compound would be chosen. Since all the
coal tar constituents are present in small quantities, a
representative compound cannot be chosen on the basis
of predominance. Alternatively, the selection of a representative compound based on a close match with the
number-average molecular weight of coal tar leads to a
low bias for CW,. The compound with available aqueous
solubility data, whose molecular weight most closely
matches 210, is 9,lO-dimethylanthracenewith a molecular
weight of 206 and subcooled liquid solubility of 2 mg/L at
25 "C (23),which is significantly less than 16 mg/L, the
estimated CW,based on eq 6. For a mixture with relatively
evenly distributed weight, the selection of a representative
compound based on molecular weight is inherently biased
low because there is an approximate logarithmic relation
between aqueous solubilities and molecular weight for
nonpolar hydrocarbons, and thus the averages do not
correspond with each other. For purposes of predicting
aqueous solubility, fluorene is more representative of the
Stroudsburg coal tar mixture, with a subcooled liquid
aqueous solubility of 15.1 mg/L and a molecular weight
of 166 (23).
Water Solubility in Coal Tar. The solubility of water
in coal tar is of interest because together with coal tar
solubility in water this completelycharacterizes the mutual
solubility of these two components. This is useful for
calibration of the binary molecular interaction parameters
in a thermodynamic model describing coal tar/solvent/
water phase equilibria ( I I , 2 2 ) . An upper bound estimate
(11). Conversion to mole
of 0.001 was obtained for
fraction gives an upper bound estimate for x$ of 0.01,
which is comparable to water solubilities in other organic
liquids of 0.0030 for water in benzene and 0.0034 in
1-methylnaphthalene (36).
Solvent Selection. A literature survey of organic
compounds used as solvents was conducted to identify an
initial list of 13 water-miscible solvents with suitable
properties for use in a solvent extraction site remediation
system. As described elsewhere (3,
criteria for initial
selection included the following: suitable chemical properties for separation from coal tar and water by distillation,
relatively low volatility and flammability for industrial
safety and handling, commercial availability, and biodegradability. The 13 solvents were evaluated in laboratory
screening tests to assess their capacity to dissolve coal tar
and their effect on the physicochemical properties of the
coal tar phase. As a result of this work, three solvents
were identified for evaluation for potential use in asolvent
extraction coal tar remediation process: n-butylamine,
acetone, and 2-propanol.
Pseudocomponent Simplification. As discussed earlier, the validity of thermodynamic modeling of coal tar/
solvent/water phase equilibria as ternary LLE depends
on the extent to which the composition of the dissolved
coal tar component is the same as that of the undissolved
coal tar. Experimental solute partitioning observations
in coal tarln-butylaminelwater systems are presented in
Figure 2. The data points shown on the ordinate, for 0%
solvent, are the coal tar/water partition coefficient data
(Table 111). With increasing amounts of n-butylamine in
the solvent/water solution, the partition coefficients for
all three solutes approach a comparable value. This
observation indicates that the effect of n-butylamine on
each compound is different. Specifically, with increasing
concentration of solvent the enhancement of the solute
partitioning to the solvent/water phase is greater for
compounds such as pyrene than for compounds such as
naphthalene. This is consistent with what has been
observed in studies of cosolvents on PAH solubilities in
which the cosolvency power, represented by u, the slope
of the log-linear solubility curve, is theoretically predicted
and experimentally verified (29,37)to be proportional to
the logarithm of the solute's KO,. In other words, the
more hydrophobic compounds experience a larger en-
uc
L
8
""""Y
A pyrene
0 phenanthrene
10000
0 naphthalene
0
!
lo!
4
I
0
20
40
60
80
0
volume % n-butylamine in initial
solvent/water solution
Flgure 2. Solute partitioning in coal tar/n-butylamlne/water systems
with 20% (vol) coal tar overall.
hancement of solubility in a given cosolvent solution. In
terms of partition coefficients, this corresponds to a more
significant decrease in K values with increasing solvent
concentration for more hydrophobic compounds. This is,
in fact, what is shown in Figure 2 for K,t/,,. Thus, for
systems with appreciable n-butylamine (>lo% with
respect to the solvent/water solution), the condition for
similar compositions of dissolved and undissolved coal
tar, i.e., eq 2, is satisfied. Since the validation of the
pseudocomponent simplification depends on similarity of
K&,,, values, these results may be extended to other coal
tars with similar constituencies of PAH compounds. Coal
tars that contain acidic compounds, such as phenols, and
basic compounds such as anilines may not be suitable for
this simplification since the solubilities and hydrogenbonding characteristics of these compounds are very
different from those of the neutral fraction (17,35).
The underlying premise of the pseudocomponent simplification was further tested by studying the effect of
n-butylaminelwater extractions on the composition of the
coal tar phase. The weight percents of 27 compounds were
quantified for this analysis. The compositions of the coal
tar phases, corrected for fractions of dissolved n-butylamine, are presented as weight percent distributions over
molecular weight in Figure 3. In Figure 3a, the compositions of coal tar samples that were singly extracted with
either 20% or 40% n-butylaminelwater solutions are
shown relative to the original coal tar composition as
reported in Table 11. In Figure 3b, the compositions of a
coal tar sample that had been sequentially extracted once,
then again with a solvent/water solution of 40% (vol)
n-butylamine are shown relative to the original coal tar.
The total mass accounted for in each of the extracted coal
tar samples was approximately 16%. Note that true
histograms of the coal tar mass distribution over molecular
weight would likely have long tails extending beyond
molecular weight 280.
The changes in compound weight percentages from the
original coal tar and extracted coal tar samples depicted
in Figure 3 range from reductions of 15% to 92 % ,but the
majority of the reductions are about 50%. Given that the
initial weight percentages were already small (<4%),the
reductions are not significant in an absolute sense. The
effect of extracting with a 20% n-butylamine solution or
with a 40% n-butylamine solution is practically the same
Environ. Sci. Technol., Vol. 27, No. 13, I993
2899
Table V.
Experimental Measurements of Component Volume Fractions in Solvent/Water and Coal Tar Phases.
Solvent = n-Butylamine
0.048 (0.014)
0.080 (0.013)
0.065 (0.009)
0.106 (0.011)
0.106 (0.008)
0.162 (0.007)
0.196 (0.007)
0.352 (0.008)
0.052 (0.001)
0.113 (0.001)
0.165 (0.002)
0.300 (0.005)
0.398 (0.006)
0.456 (0.005)
0.500 (0.006)
0.465 (0.007)
0.900 (0.014)
0.959 (0.001)
0.807 (0.013)
0.920 (0.001)
0.770 (0.009)
0.899 (0.002)
0.594 (0.009)
0.910 (0.001)
0.496 (0.006)
0.893 (0.002)
0.382 (0.004)
0.891 (0.001)
0.304 (0.003)
0.887 (0.001)
0.183 (0.003)
0.841 (0.003)
Solvent.= Acet.nna
0.029 (0.001)
0.058 (0.001)
0.073 (0.001)
0.070 (0.001)
0.087 (0.002)
0.090 (0.001)
0.095 (0.001)
0.133 (0.003)
0.012 (0.0002)
0.022 (0.0004)
0.028 (0.0004)
0.020 (0.0004)
0.020 (0.0003)
0.019 (0.0003)
0.018 (0.0003)
0.026 (0.0005)
0.070 (0.001)
0.107 (0.002)
0.144 (0.002)
0.179 (0.003)
0.251 (0.005)
0.002 (0.0004)
0.005 (0.0007)
0.006 (0.0009)
0.007 (0.0009)
0.006 (0.0011)
Solvent = 2-Propanol
0.001 (0.014)
0.041 (0.008)
0.048 (0.007)
0.041 (0.007)
0.075 (0.010)
L1
0.160 (0.002)
0.297 (0.003)
0.485 (0.005)
0.620 (0.006)
0.699 (0.009)
0.839 (0.014)
0.662 (0.007)
0.467 (0.005)
0.339 (0.004)
0.226 (0.004)
0.928 (0.001i
0.888 (0.002)
0.850 (0.002)
0.814 (0.003)
0.743 (0.005)
10 error estimates are indicated in parentheses
molecular weight
Flgure 4. Experimental coal tarlRbutyhminelwater ternary phase
diagram. Error bars on solventlwater phase tie line end points are 30
random error estimates; the errors on the coal tar phase end polnts
are insignificant. The shaded circle is the two-phase check point.
molecular weight
Flgure 3. Weight percent distributions of coal tar samples that have
been extracted with Rbutylaminelwater solutions.
with regard to weight percent distribution. Furthermore,
no significant difference in composition is observed for
the coal tar sample that has been extracted once and then
again. Another important observation is that, in the
molecular weight range of 128-278, the effectof extraction
on weight percent does not vary with molecular weight.
That is, the compounds in this range are being extracted
to the same extent, supporting the solute partitioning
observations shown in Figure 2 for three solutes. The
decrease in weight percentages in the 128-278 molecular
weight range signifies an increase in weight percentages
of compoundsbeyond this range. The cumulative increase
in the unquantified portion of the coal tar is estimated to
be 9% based on a summation of the decreases in mass of
the quantified compounds. Spread over a large number
of compounds, this represents only a slight increase in
relative abundance of each compound.
2840
Environ. SCI. Technol.. Vol. 27. NO. 13. 1993
The primary conclusion to be drawn from this analysis
is that there is not a large change in coal tar composition
uponextractionwith n-butylaminelwater solutions. These
studies indicate that theassumptionof coaltar partitioning
as a single component in coal tarlsolventlwater systems
is plausible.
Coal Tar/Solvent/Water Ternary LLE. For each
of the three solvents, experimental results of coal tar/
solvent/water phase equilibria are presented in Table V
as volume fractions of the three components. The data
are presented as tie lines on ternary phase diagrams in
Figures 4-6, for n-butylamine, acetone, and 2-propanol,
respectively. Inaternarydiagram, theaxisforaparticular
component is a line drawn from the apex (100% of that
component) perpendicular to the opposite triangular face
(0% of that component). The triangular space is divided
intotwo regions representingmixturesthat are completely
miscible and mixtures that separate into two phases. Tie
lines within the immiscible region connect points that
indicate the equilibrium compositions of the two immiscible phases resulting for any overall composition represented by a point on the tie line.
The error bars that are shown in Figures 4-6 are 30
(99% confidence) random error estimates calculated from
acetone
eo
60
FlgunS.
I
Experlmntalcoaitar/aCB1one/watBrternaryphasediagram.
Error bars on soiventlwater phase tie line end points are 30 random
error estimates: the errors on the coal tar phase end point are
insignificant.The shaded circle is the two-phase check point.
pro pa no^
fie
80
(b)
Figure 7. Three-dimensional error space as determined by error bars
for vf, vis. and vf,
precision of ut": diminishes when it is much smaller than
uc" or
As a result, relative errors greater than 100%
result for very small values of $7. In general, this
method provides a means of determining LLE of coal tar/
solvent/water systems with measurements on the order of
volume percents. Below ut": estimates of about 0.05,
order of magnitude precision can be expected. Even with
this level of precision, the data in Table V show that, even
with small amounts of solvent, coal tar solubility is
appreciable relative to its bulk solubility in water.
QualitativeTests for Heterogeneity. The plait point
is the point on a ternary phase diagram where the tie line
end points converge, and the two phases have identical
composition. Qualitative tests performed on compositions
near the plait point were done to gain additional information about the boundaries of the two-phase regions.
The two-phasecheckpoints, shown asgray dots in Figures
4-6, were found to be the same for the acetone and
2-propanolsystems: 10% coal tar, 80% solvent, and 10%
water. It is expected that there are compositions slightly
above this point that are also heterogeneous, but visual
inspection of these systems were not conclusive because
of the small overall volume of coal tar. For n-butylamine,
the region thought to contain the plait point was checked
for heterogeneity, but determination of the two phases
becameimpossiblewithsimplevisualtests. Apoint (shown
in Figure 4) with 80% n-butylamine in the solvent/water
solution, with overall composition of 30% coal tar, 56%
n-butylamine, and 14% water, was found to be two phases.
Although this point is not expected to be near the plait
point, it provides further information about the upper
boundary of the curve. Note that the uppermost tie line
in Figure 4 is not consistent with this observation, since
the solventlwater end point does not extend high enough
to be in line with the two-phase observation. The error
bars calculated based on random experimental error do
not account for this discrepancy, suggesting a systematic
error likely attributable to errors in solvent/water phase
sampling, given the difficulty in distinguishing between
the two phases.
Plait points exist for ternary systems that have only one
partially miscible pair (type I ternary system). In this
.rew.
Figure 6. Experimental coal tarl2-propanoilwater ternary phase
diagram. Error bars on soiventlwater phase tie line end points are 3c
random enw estimates: the errors on the coal tar phase end points
are lnsigniflcant.The shaded circle is the two-phase check point.
the standard deviations in Table V. Only the error bars
for the solventlwater end points of each tie line are
displayed since the error bars for the coal tar phase end
points are insignificant. Since the relative errors for
u t w and
are roughly constant a t 1-2 % ,the size of the
absolute error increases with the magnitude of the ucw or
measurement value. Three-dimensional error display on a ternary phase diagram must be viewed carefully
to accurately visualize the size of the error space. The 3u
error space can be approximated on the ternary phase
diagram by a hexagon bounded by the ends of the three
error bars as shown in Figure 7a. For points with large
error bars in two dimensions but with a small error bar in
the third dimension, the error space is more like a thick
bar (Figure 7b). For example, the solvent/water endpoint
of the bottom tie line on the n-butylamine ternary phase
diagram has a small error in the ucw dimension. The
resulting error space is a horizontal bar roughly parallel
with the tie line. This suggests that the slope of the tie
line has been precisely determined, and the uncertainty
due to random error lies in the position of the end point
along this line.
The representation of coal tar as a single component in
a ternary system allowed the bulk dissolution behavior of
coal tar to be indirectly observed by measuring the
partitioning behavior of the other two components. Since
the estimated error in uf": will always be greater than the
larger of the error in u t w or .Ew.the measurement
ucw
urww
Envlron. Sci. Technol.. VoI. 27. NO. 13. 1993
2841
case, the coal tadwater pair is known to be immiscible, all
three solvent/water pairs are miscible, and the coal tar/
solvent pairs are assumed to be miscible based on visual
observations that fail to identify two phases. For n-butylamine, an attempt was made to verify this assumption
without reliance on visual observations (11). Coal tar/
n-butylamine mixtures were prepared with spikes of
radiolabeled solvent. These tests were done in separatory
funnels to facilitate sampling from the bottom of the vial,
which would contain the coal tar phase if indeed two phases
were present. The measured concentrations of radiolabeled solvent in samples from the top and the bottom of
the vial and the overall concentration were found to be
within experimental error of each other, suggesting singlephase systems. The conclusion can be drawn that coal
tarln-butylamine is a completely miscible pair and that
some water must be present to result in phase separation.
Solvent Effectiveness. The information in a ternary
phase diagram provides a useful metric for assessment of
the effectiveness of a solvent, and it provides necessary
quantitative data for larger scale process modeling (8,9).
The simplest piece of information from the ternary phase
diagram is the vertical height of the two-phase region,
delineated by the end points of the tie lines. The twophase region for n-butylamine is smaller than for acetone
and 2-propanol, indicating that less solvent is required to
completely dissolve coal tar. For example, on the n-butylamine diagram, it is shown that if the overall composition of a mixture is 20 % coal tar, 56% solvent, and 24%
water (i.e., 70% solvent-to-water ratio), then the mixture
is completely miscible. Another indicator of the effectiveness of a solvent is the amount of coal tar dissolved in
the solvent/water phase, as indicated by the position of
the right-side tie line end points, i.e., the solvent/water
rich region. The farther the end points from the solvent/
water edge, the greater the amount of coal tar dissolved
in this phase. For example, with a mixture of 30% coal
tar, 40% n-butylamine, and 30% water, the solvent/water
phase will contain just under 20% coal tar. With asimilar
mixture using 2-propanol, the amount of coal tar in the
solvent/water phase is less than 5 % , and with acetone,
coal tar dissolution is less than 1% . The ternary phase
diagram can also be used to quantify the extent of solvent
dissolution into the coal tar phase, as is indicated by the
slopes of the tie lines. A horizontal tie line, such as is
approximated by the very lowest tie line in the acetone
ternary phase diagram (Figure 5), depicts approximate
equal partitioning of solvent between coal tar and water.
The case of limited solvent dissolution in coal tar is
depicted by a tie line steeply sloped down from right to
left, such as is observed for n-butylamine with increasing
solvent content. The amount of solvent in the coal tar
phase determines the extent to which the physical
properties of this phase are altered, i.e., the change i n
volume, density, viscosity, and surface-wetting properties.
The amount of solvent that remains in the solvent/water
phase determines the solvent-to-water ratio in this phase
and, thus, affects the extent of coal tar dissolution.
Conclusions
The challenge of characterizing the phase equilibria of
a complex mixture was met by representing coal tar as a
pseudocomponent. Phase equilibria of coal tar/solvent/
water systems were experimentally determined and presented as tie lines on ternary phase diagrams. Of the three
2842
Environ. Sci. Technol., Vol. 27, No. 13, 1993
solvents studied, n-butylamine was shown to have the
smallest two-phase region and enhance the solubility of
coal tar to the largest extent. While it is not directly
apparent from the ternary phase diagrams, coal tar
dissolution in solvent/water solutions containing acetone
or 2-propanol is appreciable in terms of being orders of
magnitude higher than the bulk coal tar solubility in water.
The coal tar/solvent/water phase equilibria provide the
necessary equilibrium chemistry for predicting mass
transfer limitations in porous media (8) and in solvent
extraction process modeling (9). Furthermore, these data
are used to determine parameters for a semi-empirical
thermodynamic model describing ternary LLE of highly
nonideal liquid mixtures (11,12).
Acknowledgments
Mr. James Villaume of Pennsylvania Power and Light,
Mr. Curt Kramer of Atlantic Environmental Services, and
Dr. David Nakles and Mr. Robert Weightman of Remediation Technologies Inc. assisted in enabling coal tar
sample collection. Dr. Curt White and Ms. Louise Douglas
of the U S . DOE Pittsburgh Energy Technology Center,
Coal Science Division, and Dr. Edward C. Nelson and Dr.
Ingeborg D. Bossert of Texaco Research Center, Beacon,
NY, arranged for chromatographic analyses. Mr. ZhongBao Liu assisted with solute partitioning measurements.
The authors thank Dr. David Dzombak and Dr. Babu Nott
for their review of this manuscript. The Electric Power
Research Institute was the primary sponsor for this
research project through contract RP 3072-2. Dr. Babu
Nott was the project manager. Additional fellowship
support was provided by the Patricia Harris Government
Opportunities Program.
Literature Cited
Mackay, D. M.; Cherry, J. A. Environ. Sei. Technol. 1989,
23,630-636.
Harkins, S. M.; Truesdale, R. S.; Hill, R.; Hoffman, P.;
Winters, S. US. Production of Manufactured Gases:
Assessment of Past Disposal Practices; Research Triangle
Institute: Research Triangle Park, NC, 1987.
Management of Manufactured Gas Plant Sites; Gas
Research Institute: Chicago, IL, 1987; Vol I: Wastes and
Chemicals of Interest.
Moore, T. EPRI J. 1989, 14, 22-31.
Luthy, R. G.; Dzombak, D. A.; Peters, C. A.; Roy, S.B.;
Nakles, D. V.; Nott, B. Technological Directions for
Remediation of Tar Contaminated Soils a t MGP Sites. To
be submitted for publication.
Salvensen, R. H. In Proceedings of the 5th National
Conference on Management of Uncontrolled Hazardous
Waste Sites; Hazardous Materials Control Research
Institute: Silver Spring, MD, 1984; pp 1-15.
Luthy, R. G.; Dzombak, D. A,; Peters, C. A.; Ali, M. A.; Roy,
S. B. Solvent Extraction forRemediationof Manufactured
Gas Plant Sites; Final report, EPRI TR-101845, Project
3072-2; Carnegie Mellon University: Pittsburgh, PA, 1992.
Roy, S. B.; Dzombak, D. A.; Ali, M. A. Assessment of in situ
solvent extraction for remediation of coal tar sites: Column
studies. Submitted to Water Enuiron. Res.
Ali, M. A.; Dzombak, D. A.; Roy, S.B. Assessment of in situ
solvent extraction for remediation of coal tar sites: Process
modeling. Submitted to Water Environ. Res.
Orye, R. V.; Prausnitz, J. M. Ind. Eng. Chem. 1965, 57,
18-26.
Peters, C. A. Ph.D. Dissertation, Carnegie Mellon University,
Pittsburgh, PA, 1992.
(12)Peters, C. A.; Luthy, R. G. Coal tar dissolution in watermiscible solvents: Semiempiricalthermodynamic modeling.
To be submitted for publication.
(13)Edmister, W. C. Applied Hydrocarbon Thermodynamics,
2nd ed.; Gulf Publishing Co.: Houston, TX, 1988;Vol. 2.
(14) Lane, W. F.; Loehr, R. C. Environ. Sci. Technol. 1992,26,
983-990.
(15) Lee, L. S.;Rao, P. S. C.; Okuda, I. Environ. Sci. Technol.
1992,26,2110-2115.
(16) Mackay, D.; Shiu, W. Y.; Maijanen, A.; Feenstra, S. J.
Contam. Hydrol. 1991,8,23-42.
(17)Picel, K. C.; Stamoudis, V. C.; Simmons, M. S. Water Res.
1988,22,1189-1199.
(18) Groher, D. M. Master's Thesis, Department of Civil
Engineering, Massachusetts Institute of Technology, 1990.
(19) Rostad, C. E.;Pereira, W. E.; Hult, M. F. Chemosphere
1985,14,1023-1036.
(20) Prausnitz, J. M.;Lichtenthaler, R. N.; de Azevedo, E. G.
Molecular Thermodynamicsof Fluid-Phase Equilibria,2nd
ed.; Prentice-Hall, Inc.: Englewood Cliffs,NJ, 1986;Chapter
9.
(21) Banerjee, S.Environ. Sci. Technol. 1984,18,587-591.
(22) Chiou, C. T.; Schmedding, D. W.; Manes, M. Environ. Sci.
Technol. 1982,16,4-10.
(23) Mackay, D.;Shiu, W. Y.; Ma, K. C. Illustrated Handbook
of Physical-Chemical Properties and Environmental Fate
for Organic Chemicals;LewisPublishers: Boca Raton, FL,
1992;Vole. 1 and 2.
(24) Yalkowsky, S.H. Ind. Eng. Chem. Fundam. 1979,18,108111.
(25) Ferry, J. P.;Dougherty, P. J.; Moser, J. B.; Schuller, R. M.
In Petroleum Hydrocarbons and Organic Chemicals in
Ground Water-Prevention, Detection, and Restoration;
NWWAIAPI Houston, TX, 1986,pp 722-733.
(26) Lafornara, J. P.; Nadeau, R. J.; Allen, H. L. In Proceedings
of the 1982 Hazardous Material Spills Conference;Bur.
Explo.: Washington, DC, 1982,pp 37-42.
(27)Unites, D. F.; Housman, J. J. In Proceedings of the 5th
Annual Madison Waste Conference of Applied Research
and Practice on Municipal and Industrial Waste; University of Wisconsin: Madison, WI, 1982,pp 344-355.
(28) Villaume, J. F. Ground Water Monit. Rev. 1985,5,60-74.
(29)Fu, J. K.;Luthy, R. G. J.Environ. Eng. 1986,112,328-345.
(30)Bonnar, R. U.; Dimbat, M.; Stross, F. H. Number-Average
Molecular Weights; Interscience Publishers: New York,
1958.
(31) Chemical andphysical CharacteristicsofTar Samples from
Selected Manufactured Gas Plant (MGP) Sites; Final
report for EPRI Contract RP 2879-12;META Environmental, Inc.: Watertown, MA, 1993.
(32)Rhodes, E. 0.In Chemistry of Coal Utilization;Lowry, H.
H., Ed.; John Wiley & Sons: New York, 1945; Vol. 11,
Chapter 31.
(33)Rhodes, E. 0.In Bituminous Materials: Asphalts, Tars,
and Pitches; Hoiberg, A. J., Ed.; Kreiger Publishing Co.:
Huntington, NY, 1979;Vol. 111, Chapters 1 and 2.
(34) Morgan, J. J. In Chemistry of Coal Utilization; Lowry, H.
H., Ed.; John Wiley & Sons: New York, 1945;Vol. 11,
Chapter 37.
(35) Campbell, J. R.; Luthy, R. G.; Carrondo, M. J. T. Environ.
Sci. Technol. 1983,17,582-590.
(36) Sorensen, J. M.; Arlt, W. Liquid-Liquid Equilibrium Data
Collection;DECHEMA: Frankfort, Germany, 1979;Vol.
5,Part 1.
(37) Morris, K. R.; Abramowitz, R.; Pinal, R.; Davis, P.;
Yalkowsky, S. H. Chemosphere 1988,17,285-298.
Received for review March 18, 1993.Revised manuscript received July 23, 1993.Accepted July 28, 1993.'
@
Abstract published in Advance ACS Abstracts,October 1,1993.
Environ. Sci. Technol., Vol. 27, No. 13, 1993 2843