II. AQUEOUS SOLUBILITY AND DISTRIBUTION COEFFICIENTS

Interaction of Oil Residues in Patagonian Soil



147



vent effects on sorption equilibrium of hydrophobic organic chemicals by organoclays; and evaluation of the NAPL compositional changes in partitioning

coefficients [12]. In principle, an organic cosolvent could be effectively used for

estimation of the aqueous concentration of complex systems, such as the oil residual in soils.

In basic research, the enhancement of the solubilization of nonpolar solutes

in water by organic cosolvents has been reported to follow a log-linear model:

log S m ϭ log S w ϩ σf c



(6)



where S m is the solubility of the solute in the mixed solvents (cosolvent and

water), S w is the aqueous solubility, σ is the cosolvency power, and f c is the

volume fraction (0 Յ f c Յ 1) of the cosolvent in the solvent mixture. Measurement

of the mixed-solvent solubility (S m ) at various cosolvent fractions f c provides a

set of data that can be plotted on a log-linear scale to determine the slope (σ) and

the y-intercept, S w. The y-intercept is equal to the predicted solute concentration in

pure aqueous solution (no cosolvent).

In this research, the prediction of aqueous concentrations using cosolvent mixtures has been extended to the measurement of poorly soluble compounds found

in the aqueous phase of complex mixtures. In this case, the presence of one

component in water phase should necessarily be affected by the presence of the

others. Components will be removed according to their solubility in the specific

cosolvent, which is influenced by molecular weight, functional groups, and polarity of the cosolvent.

According to Rao’s solvophobic theory, the sorption coefficient K m of a hydrophobic organic compound (HOC) decreases exponentially with increasing

volume of the cosolvent ( f c ) in a binary solvent mixture:

ln



΂ ΃



Km

ϭ Ϫaασf c

Kw



(7)



where K w is the equilibrium sorption coefficient from water (L kgϪ1 ), K m is the

equilibrium sorption coefficient from mixed solvent (L kgϪ1 ), a is the empirical

constant accounting for water–cosolvent interactions (note that for water–methanol a ϭ 1, implying ideal water–cosolvent interactions), α is the empirical constant accounting for solvent–sorbent interactions, and σ is the cosolvency power

of a solvent for a solute accounting for solvent–solute interactions. At a given

temperature, the parameter σ is dependent only on the sorbate and solvent properties and not on the sorbent characteristics. The value of σ for a sorbate estimated

from data for different sorbents (soils, sediments) is expected to be constant if

the model assumptions are valid.

Equations (6) and (7) are strictly valid for only one solute, not for a mixture

of solutes of varied polarities; however, in this work the applicability of the model



TM



Copyright n 2003 by Marcel Dekker, Inc. All Rights Reserved.



´

Nudelman and Rıos



148



is tested considering the oil residual as only one solute. The aqueous concentration and the distribution coefficients in this case are global values and therefore

account for the interactions among the components in the mixture and for the

overall interactions of each of them with the mineral matrix. When the product

ασ is small, the Eq. (7) can be expressed as

K m ϭ K w Ϫ mf c



(8)



where m ϭ K w ασ. This linear approach was also tested for treating the experimental data; in all cases, the best adjustment of the experimental information

with the equations was examined.

Contaminated soil samples, the product of oil spills in six different locations

in the environs of Comodoro Rivadavia, were obtained. The oil spills are of

different ages, crude oil sources, and environmental exposure conditions. In all

cases, except for samples 1 and 6, fertilization of the affected areas was carried

out to improve the general conditions of the land, to accelerate the biodegradation

processes, and to favor reforestation of species adapted to the zone. Table 5 summarizes some properties of the samples.

Figure 4 shows illustrative examples of the equilibration test for samples 1

and 5. The log oil residual aqueous concentration is plotted as a function of the

cosolvent fraction. The data indicate a good linear correlation, which shows good

agreement with Eq. (7). Table 6 compares the measured aqueous concentrations

to those calculated by Eq. (7). The values of σ glo (the subscript glo is used to

indicate a global behavior) correspond to the slopes of the straight line and represent the cosolvency power of the solvent for each sample. The standard deviations

for the calculated log S w values are given in Table 6 together with other statistical

parameters. The relative goodness of the regression adjustment is shown by the



TABLE 5 Description of Oil-Contaminated Soil Samples



Sample

1

2

3

4

5

6

a

b



TM



Landscape



Description



Meadow

Coastal area

Depression

Creek

Arid plateau

Meadow



Prairie

Barren soil

Barren soil

Open shrub

Shrub steppe

Prairie



Extract 1: 5 wt/wt.

25°C.



Copyright n 2003 by Marcel Dekker, Inc. All Rights Reserved.



Oil spill

age

Conductivity

Total oil Clay

(years) (µS cmϪ1 ) a,b pH a (wt%) (wt%)

Ͼ10

10

6

3

3

2



9364

1633

646

618

387

426



7.6

7.4

8.0

7.4

7.6

6.8



25.8

16.6

8.7

8.6

9.3

16.1



33

22

9

8

12

16



Interaction of Oil Residues in Patagonian Soil



149



FIG. 4 Log of the oil residual aqueous concentration (mg LϪ1 ), as a function of the

cosolvent fractions, for samples 1 (diamonds) and 5 (squares).



r 2 coefficients and the validity of the plotting pattern by means of the critical

values of F.

For the oldest samples (1, 2, and 3) the aqueous concentrations calculated

according to the theory are higher than those measured, while the calculated value

for the youngest samples (4, 5, and 6) is in all cases smaller than the experimental.

The error in the determinations is approximately constant for the range of σ glo

(0.92–1.25). A good correlation exists between f c and solubility in cosolvent mixtures (0.928 Յ r 2 Յ 0.999), and the logarithmic model seems to be a good representation of the experimental data for f c Ն 0.2.



TABLE 6 Equilibrium Aqueous Concentrations, Global Cosolvent

Power σ glo , and Statistical Regression Values

Equilibrium aqueous

concentration

Sample

1

2

3

4

5

6



TM



Exptl.

(mg LϪ1 )



Calcd.

(mg LϪ1 )



Standard

deviation

of log S w



σ glo



r2



Critical

value of F

(%)



114.1

136.2

64.0

64.5

103.8

188.0



157.3

254.5

72.8

35.3

57.8

131.4



0.039

0.011

0.075

0.054

0.022

0.017



0.78

0.64

0.74

1.25

1.08

0.92



0.987

0.999

0.950

0.928

0.989

0.999



0.64

1.58

2.50

3.68

0.56

1.81



Copyright n 2003 by Marcel Dekker, Inc. All Rights Reserved.



´

Nudelman and Rıos



150



For contaminated samples 1, 2, and 3, the oil residuals contain a smaller proportion of water-soluble components when compared to the extrapolation of solubility for different cosolvent fractions. This could be interpreted by assuming

that the cosolvent mobilizes the liquid-phase hydrophobic components that are

not really available in the water phase. In the case of contaminated samples 4,

5, and 6, the oil residuals contain a bigger proportion of water-soluble components as compared to the extrapolated solubility to noncosolvent fractions. Although the solubilization cosolvent power is good (0.92 Յ σ glo Յ 1.25), it is not

possible to reproduce the aqueous concentration value by extrapolation, probably

because the oil residuals should possess important hydrophilic global properties.

The reported values of cosolvent power for PAHs in soils vary between 1.63

and 9.09 when methanol is used as cosolvent; in our case the values were in the

range 0.64–1.25, indicating a smaller solvent effect. This agrees with the high

PAH hydrophobicity, compared to the lower hydrophobicity of hydrocarbon mixtures in oil residuals. The value of σ for naphthalene in methanol–water mixtures

was estimated from Nzengung to be 8.95, and it is independent of the sorbent.

But Lane shows that the σ-values were not consistent for individual compounds

in different soil samples.

Table 7 shows the results obtained by applying solvophobic theory to the

calculation of the distribution coefficients K d . Although, Rao’s solvophobic theory is based on the equilibrium sorption coefficient, desorption experiences have

been carried out in this work. And a low hysteresis effect has been considered,

due to the probable linearity of the isotherms, in case adsorption on the mineral

surface was the dominant process. Table 7 shows the measured coefficients and

the distribution coefficients calculated by application of both the logarithmic

model, Eq. (7), and the linear approach, Eq. (8), according to the best adjustment

and the interaction parameter α app (the subscript is used to indicate that total

interactions are taken into account). Smaller differences between the measured



TABLE 7 Distribution Coefficient K w , Cosolvent Power σ app , Coefficient α app , and

d

Statistical Regression Values

Kw

d

Sample

1

2

3

4

5

6



TM



Model



Exptl.



Calcd.



r2



σ app



α app



Critical

value of F

(%)



Logarithmic

Logarithmic

Linear

Linear

Linear

Linear



2913

1337

1387

1348

878

901



2272

1017

896

1346

1035

971



0.993

0.996

0.978

0.942

0.961

0.960



2.31

2.32

2.27

2.14

1.95

1.73



0.97

1.22

0.37

0.46

0.52

0.57



0.33

3.81

1.03

2.95

1.95

1.28



Copyright n 2003 by Marcel Dekker, Inc. All Rights Reserved.



Interaction of Oil Residues in Patagonian Soil



151



FIG. 5 Log K d (L kgϪ1 ) as a function of cosolvent fraction for sample 1 (triangles) and

sample 2 (squares).



and calculated K w were found when Eq. (8) was used for the samples 4, 5, and

d

6, and for the samples 1 and 2 when Eq. (7) was applied.

Figure 5 shows that samples 1 and 2 give a good correlation of log K m with

d

f c , as predicted by application of the solvophobic theory, while, as shown in

Figure 6, samples 4 and 5 exhibit a linear correlation. Although the logarithmic

equation, Eq. (7), could strictly be replaced by the linear approximation, Eq. (8),

when the product ασ is very small (usually Ͻ0.1), in the present case Eq. (7)

correlates the experimental data better for all cases in which ασ Ͻ 1 (the differ-



FIG. 6 K m (L kgϪ1 ) as a function of cosolvent fraction for sample 4 (triangles) and sample

d

5 (squares).



TM



Copyright n 2003 by Marcel Dekker, Inc. All Rights Reserved.



152



´

Nudelman and Rıos



ence between e ασ and 1 ϩ ασ is under 0.6 for samples 3–6, while it is 6.7 and

13.1 for samples 1 and 2, respectively).

An approximate estimation of the individual α- and σ-values can be carried

out as follows: The cosolvency power σ depends on the solute–solvent interactions and can be estimated by applying Eq. (6) to the oils extracted from samples

1–6, respectively. Table 7 shows the σ app-values thus determined (the squares of

the calculated linear regression coefficients for this equation were 0.956 Ͻ r 2 Ͻ

0.999). Taking into account that for the present work a ϭ 1, from the calculated

K d values the product ασ can be estimated and, therefore, the values of α app

calculated. The values of α app (which account for the solvent–sorbate (soil) interactions) are smaller than unity, as shown in Table 7.

According to theory, a value of α near 1 would indicate that the properties of the

sorbent are independent of the changes in the composition of the water: cosolvent phase. In studies on sorption of hydrophobic organic compounds by soils

from solutions containing varying fractions of organic cosolvent, α Ͻ 1 has usually been obtained, which indicates that sorption from solvent mixtures was

greater than that predicted from increased solubility alone. This behavior has

been attributed to the swelling of soil organic matter in solvents. The gel swelling

of sorbent organic matter results in enhanced permeation of compounds, leading

to greater sorption. In our case, according to the available information a similar

conclusion cannot be validated, since Patagonian soils are poor in organic matter.

More investigations are necessary to draw sound conclusions with respect to α appvalues.



B. Effects of Spill Age

The equilibrium aqueous concentration, C w , of the contaminated soil samples

e

normalized by hydrocarbon percentage are shown in Figure 7 as a function of

the age of the spills. The values are between 130 mg LϪ1 and 1,100 mg LϪ1. When

the rate of the degradative processes decreases with time, the concentrations of

the nonpolar components tend to become constant while those of the polar ones

decrease due to solubilization. Therefore, a decrease in aqueous solubility is expected with age, as observed in Figure 7. Those residuals that are the most aged

have a superior hydrophobic behavior due to the loss of polar components.

The distribution coefficients K w (L kgϪ1 ) are shown in Figure 8 as a function

d

of the age of the spills. The values are between 900 L kgϪ1 and 10,000 L kgϪ1.

Figure 8 shows that aged oil residuals exhibit a similar behavior, an increase of

sorption with time, since they are from different crude oil sources and environmental exposure conditions [13,14]. The oil residual could be formed by recalcitrant original components, particularly the resins and the asphaltic fraction

[15,16] and by the products of their successive transformations.



TM



Copyright n 2003 by Marcel Dekker, Inc. All Rights Reserved.



Interaction of Oil Residues in Patagonian Soil



153



FIG. 7 C w , equilibrium aqueous concentration (mg LϪ1 ) as a function of the age of the

e

spill (years).



For the interpretation of the hydrosolubility time dependence, the ratio

(Aliph ϩ Aro)/(Pol ϩ Asph) could be used. The ratio (Aliph ϩ Aro)/(Pol ϩ

Asph) for the case of regional crude oils are 4.59 Ϯ 1.08 and for the degraded

environmental samples are 1.03 Ϯ 0.31 (age, 2–3 years) and 2.31 Ϯ 0.48 (age,

6–57 years). It can be observed that the ratio (Aliph ϩ Aro)/(Pol ϩ Asph) for

crude oils indicates a high content of aliphatic and aromatic components. In the



FIG. 8



TM



K w , distribution coefficients (L kgϪ1 ) as a function of the age of the spill (years).

d



Copyright n 2003 by Marcel Dekker, Inc. All Rights Reserved.