Experimental Methods

Detailed experimental procedures have been previously reported (Ko, 1998; Ko et al., 1998a,b); therefore, they are only briefly described here. Phenanthrene (Aldrich, 99.5+%), naphthalene (Aldrich, 99+%), SDS (Sigma, 99.5+%), and Tween 80 (Aldrich, no purity reported) were used as received; selected physicochemical properties for these compounds are shown in Table 1. Kaolinite, a nonswelling 1:1 layer phyllosilicate clay and common constituent of many subsurface environments, was used as received from Sigma. Solution pH and ionic strength were adjusted as necessary with 0.5 M HC1 and/or 0.5 M NaOH and NaCl, respectively. Aqueous phenanthrene and naphthalene concentrations were quantified by fluorescence (PTI, Inc.) at the excitation/emission wavelengths of 250/364 and 278/322 nm, respectively. A total organic carbon (TOC) analyzer (Shimadzu Model 5050) was used to determine aqueous SDS concentrations and Tween 80 concentrations were determined by UV absorbance at 234 nm.

Table 1. Selected HOC and surfactant properties.a

solubility in

compound

formula

MW

distilled water

log Kûw

CMC"

phenanthrene

CmH.O

178.23

7.2 nM

4.57

-

naphthalene

CioHg

128.17

240 nM

3.36

-

SDS

C12H25S04Na

288.38 .

complete

-

1.45 mMc

Tween 80

C,8S6E20 d

1310

complete

-

9.92 nMe

a HOC data from Karcher (1988). Surfactant data from supplier unless noted. b 0.1 M NaCl. c Israelachvili (1992). d Jafvert et al. (1994). En is n repeating -CH2CH20- groups; S6 is a sorbitan ring.c Pennell et al. (1993).

a HOC data from Karcher (1988). Surfactant data from supplier unless noted. b 0.1 M NaCl. c Israelachvili (1992). d Jafvert et al. (1994). En is n repeating -CH2CH20- groups; S6 is a sorbitan ring.c Pennell et al. (1993).

Surfactant equilibrium isotherms and sorption envelopes on kaolinite were determined in triplicate batch experiments for the appropriate solution chemistry conditions. After equilibration, the solids were separated by centrifugation at 7000 rpm for 30 min and aliquots of the supernatant were taken for analysis. Residual SDS and Tween 80 concentrations (Ssurf, mM) were determined after taking into account dilution factors and system losses, and the sorbed surfactant concentrations (Ssorb, ^mol/g) were calculated by mass balance.

Fluorescence techniques (e.g., Miyagishi et al., 1987, 1994) were used to determine surfactant CMC and micellar partition coefficient (Kmu) values under various solution chemistry conditions (Ko et al., 1998a,b). For some conditions, solubility enhancement experiments were also conducted to compare Kmic values obtained at HOC saturation. To quantify HOC partitioning to sorbed surfactants, two different types of experiments were conducted. In the varying surfactant dose experiments, samples containing kaolinite and the appropriate surfactant were first prepared using the procedure described above for the surfactant sorption tests (i.e., same pH, ionic strength, and surfactant concentrations). After an initial equilibration, HOCs were then added such that the total phenanthrene and naphthalene concentrations were 4.49 and 78 ^M, respectively. After a second equilibration, the samples were centrifuged at 7000 rpm for 30 min to separate the aqueous and solid phases. HOC concentrations in the aqueous phase were determined by fluorescence using external standards in appropriate surfactant solutions, and HOC amounts partitioned to the sorbed surfactants were calculated by mass balance. In addition to the varying surfactant dose experiments, fixed surfactant dose (i.e., varying HOC concentration) experiments were conducted to test the linearity of HOC sorption isotherms for various sorbed surfactant concentrations and to determine whether surfactants sorbed to the same level but under different solution chemistry conditions would have similar HOC partitioning capabilities.

3. RESULTS AND DISCUSSION 3.1 Surfactant Sorption on Kaolinite

SDS sorption on kaolinite was relatively quick ( ' ~6 h; Ko et al., 1998b), and no significant solids effects (i.e., decreasing distribution coefficients with increasing sorbent concentrations) were observed for solid-to-liquid ratios of 1:10 to 1:4 (i.e., 100 to 250 g/L kaolinite; Fig. 1). SDS isotherms exhibited the classical S-shaped curves as previously reported (e.g., Fuerstenau and Wakamatsu, 1975; Holsen et al., 1991; Chandar et al., 1987); those studies generally used positively-charged mineral surfaces (ferrihydrite or alumina), and their sorption isotherms clearly showed three to four distinct regions, indicative of the varying importance of sorption interactions between solid surfaces and surfactant molecules. SDS sorption onto kaolinite, however, appeared to show only three distinct regions for the conditions examined here. The lower amount of SDS sorption in Region I of Fig. 1 generally results because of the electrostatic repulsion between the anionic head group (sulfate ion) of SDS and the overall negatively-charged kaolinite surface; any sorption occurring in Region I can be attributed to hydrophobic interactions between SDS tails and the kaolinite basal plane and/or to anion exchange at the small number of positively charged sites that exist. In Region II, the sharp rise in the isotherm indicates associations between SDS molecules at the surface, presumably through lateral interactions of their hydrophobic tails, and the formation of hemimicelles and/or admicelles. The sorption plateau that occurs in Region III corresponds to either an increase in electrostatic repulsion between the anionic head groups, complete surface coverage, and/or the attainment of a constant surfactant monomer concentration in the aqueous phase, possibly due to incorporation of any additional surfactant oligomers into micelles. In Fig. 1, it is noteworthy that the sorption of SDS begins to level out near its CMC value of 1.5 mM but that the isotherm becomes flat only above the CMC.

Tween 80 sorption to kaolinite also shows a high degree of nonlinearity and an S-shaped curve; however, most of its sorption occurs above the CMC (Fig. 1). This observation is consistent with some previous studies (e.g., Pennell et al., 1993; Edwards et al., 1994; Sharma, 1995) that found nonionic surfactant sorption occurring well above the CMC but contrasts with other studies (e.g., Liu et al., 1992; Brownawell et al., 1997) that found the sorption of nonionic surfactants to plateau near their CMC values. It generally has been thought that surfactant sorption should reach a limiting maximum value at the CMC if the sorbing species are surfactant monomers because the concentration of monomers is constant above the CMC. Although no satisfactory explanation for the contrasting observations above is yet available, a comparison of the quantity of native organic matter present in the sorbents is noteworthy: those studies using sorbents with relatively low organic carbon mass fractions (foc < 0.06%; Pennell et al., 1993; Edwards et al., 1994; Ko et al., 1998a,b) observed sorption above the CMC, whereas the studies using sorbents with much larger amounts of native organic matter (0.76% < foc 3.04%; Liu et al., 1992; Brownawell et al., 1997) did not.

The influence of ionic strength on surfactant sorption is shown in Fig. 2. In general, SDS sorption at 0.1 M NaCl was greater than for no added NaCl, consistent with previous observations (Xu and Boyd, 1995). Increased SDS sorption at the higher ionic strength can be explained by a decrease in the electrostatic repulsion between sorbed SDS molecules as well as between

SDS and kaolinite (i.e., at pH 4.6 both the kaolinite surface and SDS have net negative charges). The initial enhancement of sorption that occurs with increasing NaCl at low SDS concentrations most likely results from a screening effect between SDS and kaolinite that allows SDS molecules to first sorb; enhancement at higher SDS concentrations most likely results from decreasing repulsions between sorbed SDS headgroups when hydrophobic forces become more important. For the nonionic Tween 80 surfactant, isotherms for 0 and 0.1 M NaCl show that differences in sorption are minor for these conditions, consistent with results from Brownawell et al. (1997).

pH effects on SDS sorption to kaolinite are summarized in Fig. 3. A wide range ofpH values at fixed ionic strength were investigated for both sorption versus dose experiments (Fig. 3a) and sorption envelope tests (Fig. 3b). It is clear that SDS sorption decreases with increasing pH over the entire dose region (Fig. 3 a), an observation that is even more dramatic when shown as a sorption envelope (Fig. 3b). This observation is consistent with the idea that a decrease in pH leads to a decrease in the negative charge density on the kaolinite surface; this, in turn, reduces the repulsive force between the surface and the negative head group of SDS molecules, thereby leading to increased SDS sorption. When the solution pH is below the point of zero charge (PZC) of kaolinite (pH 4.1 to 4.3; Ko et al., 1998a), the net surface charge ofkaolinite becomes positive and more SDS molecules can be sorbed due to the electrostatic attraction between the surface and SDS head groups.

A Tween 80 sorption envelope is also shown in Fig. 3b. Tween 80 sorption generally decreases with increasing pH because of the corresponding increase of net negative surface charge on kaolinite above its PZC (i.e., pH-dependent surface sites). For example, hydrogen bonding has been suggested as the mechanism responsible for the sorption of nonionic surfactants on mineral surfaces (Cummins et al., 1992; Brownawell et al., 1997). Depending on the solution pH relative to the PZC, surface hydroxyl groups of kaolinite can be either protonated to positively-charged species or deprotonated to negatively-charged species. Therefore, as the solution pH decreases, favorable hydrogen bonding between Tween 80 oxyethylene groups and sorbed protons on the surface can occur. Conversely, increasing the solution pH results in deprotonation of surface hydroxyl groups, and therefore less hydrogen bonding between Tween 80 and kaolinite and decreased sorption.

Micelar Solubilixation

3.2 HOC Micellar Solubilization

Surfactant titrations of aqueous solutions containing a hydrophobic fluorescent probe result in two distinct fluorescence regions that can provide relatively well-defined surfactant CMC values (Ko et al., 1998b). The influence of ionic strength (as added NaCl) on the CMC of SDS was dramatic (Table 2); as explained by Ko et al. (1998a) and references therein, increases in solution ionic strength lead to stronger bonding energies between SDS molecules and, therefore, lower CMC values. In contrast to ionic strength, no pH effect on the CMC of SDS was evident over the pH range 4 to 10; the latter result is consistent with the low pKa (~2.3)of SDS. For Tween 80, CMC values showed negligible differences under varying ionic strength and pH conditions (Table 2).

Micellar partition coefficient (Kmic) values for phenanthrene and naphthalene below their aqueous solubility limits were determined from experimental fluorescence measurements using nonlinear regression analysis of the following equation:

where is the total aqueous surfactant concentration, and are the HOC-normalized fluorescence intensities in the total system, aqueous phase, and micellar phase, respectively (Ko et al., 1998b). Using the appropriate CMC values for SDS and Tween 80, Kmlc values for phenanthrene and naphthalene were then determined for varying solution chemistry conditions (Tables 2 and 3). The general results show that the more hydrophobic compound, phenanthrene, has a larger partition coefficient than naphthalene and that the nonionic Tween 80 surfactant has larger Kmic values than does the anionic SDS surfactant, agreeing with observations from previous studies (Kile and Chiou, 1989; Nayyar et al., 1994; Park and Jaffe, 1993). As shown in Table 3, Kmic values obtained by fluorescence generally decreased with increasing HOC concentration. Although the deviations were not large and some values had more relative uncertainty associated with them, the decrease appears to be significant at the 95% confidence level.

An increase in the phenanthrene partition coefficient for SDS micelles is observed with increasing ionic strength at a fixed pH of 6 (Table 2). A conceptual model has been proposed to describe the effects of electrolyte addition on the partitioning of nonpolar compounds such as phenanthrene into the core (or deep region within the palisade layer) of ionic surfactant

Table 2. Surfactant critical micelle concentrations (CMC) and micellar partition coefficients ( Kmtc ) for phenanthrene under various solution chemisty conditions.a

SDS

Tween 80

NaCl (M)

CMC (mM)

Km,cxl0 -'(M1)

Koc (L/g)

CMC (jiM)

Kmlcxl0-3 (M1)

Koc (L/g)

0

8.1

1.48 ± 0.12

10.28 ± 0.83

9.92

47.4 ±2.15

66.02 ± 2.99

0.001

8.0

1.67 ± 0.15

11.59 ± 1.04

nd b

nd

nd

0.01

5.6

2.12 ± 0.26

14.72 ±1.81

9.92

54.6 ± 2.23

76.04 ± 3.11

0.1 c

1.5

2.28 ± 0.41

15.83 ± 2.84

9.92

55.0 ± 1.33

76.60 ± 1.85

0.1

1.5

2.17 ± 0.17

15.07 ± 1.18

9.92

48.7 ± 3.43

67.83 ± 4.78

0.1 d

1.5

2.23 ± 0.14

15.49 ± 0.97

9.92

51.3 ± 7.55

71.45 ± 10.52

a Obtained by nonlinear regression analysis of fluorescence data using procedures described by Ko et al. (1998b). Values for Kmc and K,„ are ¿SD. The pH value was 6 unless otherwise noted. b nd, not determined. c pH 4. d pH 10.

a Obtained by nonlinear regression analysis of fluorescence data using procedures described by Ko et al. (1998b). Values for Kmc and K,„ are ¿SD. The pH value was 6 unless otherwise noted. b nd, not determined. c pH 4. d pH 10.

Table 3. HOC partition coefficients to surfactant micelles (A'mic) and sorbed surfactants (Kss).a

SDS

Tween 80

HOC

C, (|iM)

Kmicb(M')

R2

Nc

Kssa(M')

R2

N

Km,c b (M "')

R2

N

Kssd(M-')

R2

N

phenanthrene

1.68

2693142

0.993

15

nd c

634951 1982

0.999

15

nd

2.81

2282 + 610

0.998

15

nd

55011 1 1337

0.996

15

nd

4.49

1635 172

0.998

15

138041621

0.990

27

5150712810

0.999

15

58606 1384

0.971

30

6.73

34671156f

0.998

27

nd

35481 1807f

0.999

15

nd

naphthalene

39

381 15.1

0.999

24

nd

nd

nd

117

315122

0.998

24

525124 s

0.959

27

2754 1408

0.996

18

83231579 s

0.960

30

195

152127

0.999

24

nd

nd

nd

241

24515.1f

0.998

24

nd

2239 ± 289f

0.986

15

nd

a From experiments using fixed HOC concentrations (C,) and varying surfactant concentrations. Values are ± SD. Ionic Strength = 0.1 M NaCl and pH 4.0 (Kmc) or 4.6 (K„). b Determined by nonlinear regression of eq 1 except as noted.c N, number of data points. d Determined by nonlinear regression of eq 2.c nd, not determined.f Obtained by linear regression of solubility enhancement data. 8 Naphthalene concentration was 78 |iM.

a From experiments using fixed HOC concentrations (C,) and varying surfactant concentrations. Values are ± SD. Ionic Strength = 0.1 M NaCl and pH 4.0 (Kmc) or 4.6 (K„). b Determined by nonlinear regression of eq 1 except as noted.c N, number of data points. d Determined by nonlinear regression of eq 2.c nd, not determined.f Obtained by linear regression of solubility enhancement data. 8 Naphthalene concentration was 78 |iM.

micelles (Attwood and Florence, 1983). For example, displacement of the CMC of ionic surfactants to a lower value as a result of electrolyte addition leads to increased partitioning overall because of the increased fraction of surfactant molecules existing in the micellar form (Attwood and Florence, 1983; Israelachvili, 1992); however, this effect does not explain the increase in Kmu observed here because the values have already been normalized for the concentration of micelles present. "Salting out" effects can lead to increases in HOC partition coefficients; however, using a Setschenow constant of 0.28 M ' for phenanthrene (Schwarzenbach et al., 1993), calculations show that the Km,r value should have increased only -6%for the 0.1 M NaCl solution, a value much smaller than the -47% relative difference observed here. Therefore, it appears that the differences in KmiL values for phenanthrene are caused primarily by changes in SDS micellar properties with ionic strength. No significant effects of solution pH were observed for SDS solubilization of phenanthrene (Table 2). Again, this can be attributed to the fact that SDS molecules have strong dissociation characteristics (i.e., low pKa of ~2.3). For Tween 80, little to no effects on phenanthrene solubilization were observed with changing solution chemistry conditions. A few studies, however, have reported increases in the solubilization of organic compounds by nonionic micelles at ionic strength values much higher than the range used here (Attwood and Florence, 1983).

3.3 Equilibrium Partitioning of HOCs to Sorbed Surfactants

Figure 4 shows phenanthrene and naphthalene sorption isotherms to kaolinite covered with varying levels of sorbed surfactant; these levels of surfactant coverage correspond to the different regions existing in the surfactant sorption isotherms discussed earlier (Fig. 1). The linearity of each isotherm was evaluated using Freundlich and linear sorption models. It is apparent from Fig. 4 and Table 4 that HOC partitioning to kaolinite with and without adsorbed surfactants results in linear or near-linear isotherms. As the amount of surfactant adsorbed on the kaolinite surface increased, the sorption of phenanthrene and naphthalene to the solid phase also increased. However, upon normalizing by the amount of sorbed surfactant present, the sorbed surfactant partition coefficient decreased with increasing sorbed surfactant amounts (Table 4).

The effectiveness of a treatment/remediation scheme utilizing surfactants will depend on the distribution of an HOC between immobile and mobile phases, which is commonly quantified by a distribution coefficient (KD). Representative results for KD as a function of aqueous surfactant concentration are shown in Fig. 5. These experiments were conducted by holding total HOC concentrations constant and varying the surfactant doses; Kd values were then determined directly after centrifuging to remove solids. At low aqueous SDS concentrations, Rvalues increased with increasing surfactant concentration (Fig. 5a,b) because the amount of SDS adsorbed to kaolmite increases rapidly in this region (Fig. 1) and because the sorbed SDS is very effective for HOC partitioning (Holsen et al., 1991; Jafvert, 1991). When the aqueous phase SDS concentration reaches its CMC, SDS sorption to kaolinite plateaus and micelles begin competing for HOC molecules, thereby causing a decrease in KD. For Tween 80, all but one data point is above the CMC; thus, the HOCs are partitioning to both surfactant phases (sorbed surfactant and micelles) over the majority of the concentration range examined (Fig. 5c,d). KD values initially increase at the lower concentrations near the CMC because the affinity of sorbed Tween 80 for the two HOCs is greater than that of the micellar Tween 80 in this region. As the amount of micellar Tween 80 becomes larger relative to the sorbed Tween 80, KD values begin to decrease. Note that this decrease in KD occurs even though the Tween 80 sorption to kaolmite has not yet reached its maximum sorption plateau (Fig. 1).

Our results for HOC partitioning in the presence of sorbed surfactant and micelles demonstrate that large differences can exist in the HOC sorption capacity of surfactant aggregates in micellar versus sorbed forms. This can be seen quite readily by calculating Kss values as a function of surfactant dose from the experimental KD values. The distribution coefficient defines the HOC mass balance and can be expressed as:

where Cm:mob (mol/g-kaolinite) and Cmob (mol/L) are the immobile and mobile HOC concentrations, respectively, is the sorbed surfactant concentration (mole/g-kaolinite), and Kmin (L/g-kaolinite) is the HOC sorption constant to the bare kaolinite surface. From the previously determined values for micellar solubilization (Kmic), surfactant distribution (Ssorb and Smic) and HOC sorption to kaolinite (Kmin), Kss values can be calculated for each distribution data point using eq 2. Alternatively, an overall average Rvalue can be calculated by fitting eq 2 to the distribution data using nonlinear regression analysis.

Average Kss values determined by nonlinear regression of eq 2 are shown in Table 3. As expected, these values are intermediate of the ones calculated for each individual data point (Fig. 5). In all cases, Kss values were larger for o

a) Phenanthrene + SDS

OSorbed SDS = 0 □ Sorbed SDS = 0.09 A Sorbed SDS - 13.2 O Sorbed SDS = 27.8

0 05

(e) Phenanthrene + Tween 80

a) Phenanthrene + SDS

OSorbed SDS = 0 □ Sorbed SDS = 0.09 A Sorbed SDS - 13.2 O Sorbed SDS = 27.8

0 045

0.015

(d) Naphthalene + Tween SO OSorbed Tween »0 = 0 563 □ Sorbed Tween 80 = 12.9

O Sorbed Tween >0 - 0 563 OSorbed Tween 80 * 12.9

O Sorbed SDS = 0 □ Sorbed SDS = 0.31 A Sorbed SDS = 13 2

(e) Phenanthrene + Tween 80

O Sorbed Tween >0 - 0 563 OSorbed Tween 80 * 12.9

0 05

O Sorbed SDS = 0 □ Sorbed SDS = 0.31 A Sorbed SDS = 13 2

(d) Naphthalene + Tween SO OSorbed Tween »0 = 0 563 □ Sorbed Tween 80 = 12.9

AQUEOUS HOC CONCENTRATION (jiM)

Figure 4. HOC sorption isotherms on kaolinite with varying concentrations of sorbed surfactant (units of fimol/g-kaolinite). Error bars for some data points are smaller than the symbols. Kaolinite concentration was 100 g/L, except for naphthalene sorption to bare kaolinite (i.e., sorbed SDS = 0), where it was 300 g/L. Adapted from Ko et al. (1998b).

Table 4. Parameters of the HOC sorption isotherms for sorbed surfactants.a

Freundlich Modelc

Linear Modeld

phenanthrene

SDS

Ssorbb (nmol/g) 0.0

Nc 18

Kss (M ') 0.002 ±2E-4f

0.996 ±0.107

R2 0.994

Kss (M1) 0.002 ± 5E-5 f

R2 0.986

0.09

21

51503 12540

0.972 ± 0.067

0.995

51111 ±641

0.991

13.2

18

13149 ±1134

0.992 ± 0.029

0.999

13497 ±584

0.999

27.8

15

11480+1229

0.992 ± 0.026

0.992

11934 ±231

0.999

Tween 80

0.563

15

123027 ±3844

1.083 ±0.035

0.994

162885 ±7994

0.992

12.9

15

85113 ±2201

0.932 ±0.011

0.995

65228 ±1477

0.994

naphthalene

SDS

0.0

15

3.04E-4 ± 8.3E-6'

1.091 ±0.111

0.996

3.64E-4± 1.9E-5 '

0.992

0.31

15

2319 ±229

1.121 ±0.096

0.992

2932 ± 338

0.985

13.2

15

367 ±44

1.033 ±0.015

0.995

385 ±8.7

0.996

Tween 80

0.563

15

8511 ±841

1.091 ±0.209

0.989

7304 ± 772

0.993

12.9

15

1122 ±104

1.107 ±0.125

0.991

977 ± 68

0.982

a From experiments using varying HOC concentrations and fixed sorbed surfactant concentrations (Ssorb). Values for Kss and n are ± SD. Ionic strength = 0.1 M NaCl and pH 4.6. b Maximum sorption plateaus for SDS and Tween 80 were approximately 40 and 29 (imol/g, respectively. c Model parameters determined from nonlinear regression analysis of qH0C = Kss CHOc • d Model parameters determined from linear regression analysis of qHoc = CHOc- e N, number of data points.' Values are Km,„ (LVg-kaolinite).

a From experiments using varying HOC concentrations and fixed sorbed surfactant concentrations (Ssorb). Values for Kss and n are ± SD. Ionic strength = 0.1 M NaCl and pH 4.6. b Maximum sorption plateaus for SDS and Tween 80 were approximately 40 and 29 (imol/g, respectively. c Model parameters determined from nonlinear regression analysis of qH0C = Kss CHOc • d Model parameters determined from linear regression analysis of qHoc = CHOc- e N, number of data points.' Values are Km,„ (LVg-kaolinite).

4 14

4 14

(b) Naphthalene

+ SDS

i

• Kss

/ m. A

A Kd

S

/

2.72

^ Average Ks

, = 525M

A

*

CMC

A

*

0.006

0.003

(d) Naphthalene + Tween 80

0 012

0.006

0.003

45 4.76

(d) Naphthalene + Tween 80

0 03

0.001

0.001

0.01

AQUEOUS SURFACTANT CONCENTRATION (mM)

Figure 5. HOC distribution (Kn) and sorbcd surfactant partition (K„) coefficients. Kaolinite concentration was 100 g/L, Individual K„ values {filled circles) were determined from the Kp values using eq 2 and the micellar partition (Kmu) and kaolinite sorption (Kmin) constants below. Isotherm K„ values (open circles) are from Table 4 (linear values), (a) K^ = I635 M1. Kmm = 0.002 L/g. (b) = 280 M '. Km„ = 0.0003 L/g. (c) Kmi = 51507 M '. Kmm = 0.002 L/g. (d) Kmi = 2496 M1. Kmn = 0.0003 Ug. Adapted from Ko et al. (1998b).

the more hydrophobic HOC (phenanthrene) and for the nonionic surfactant (Tween 80). For both SDS and Tween 80, the average A',, values calculated for phenanthrene and naphthalene were always larger than the Kmic values at equivalent HOC concentrations (Table 3). When all values in Table 3 are considered, only one (i.e., Tween 80 for a phenanthrene concentration of is larger than the corresponding Previous studies have also reported sorbed surfactant partition coefficients that were generally larger than the micellar partition coefficients (Nayyar et al., 1994; Mukerjee et al., 1995; O'Haver and Harwell, 1995; Sharma, 1995; Sun and Jaffe, 1996). No convincing explanation for this observation has yet been advanced; presumably it results from geometric differences between sorbed and micellar surfactant aggregate structures.

Individual values for SDS calculated directly from each HOC distribution data point clearly show two distinct ranges that generally correspond well with the locations of the different surfactant sorption regions (i.e., compare the filled circles in Fig. 5a,b with Fig. 1). For both phenanthrene and naphthalene, as the amount of surfactant sorbed to kaolinite increased, the respective Kss values decreased. In addition, the individual values showed excellent agreement with the isotherm values (filled vs. open circles, respectively, in Fig. 5). For phenanthrene, all individual and isotherm Kss values obtained for SDS were much larger than any of the Kmu values observed (Fig. 5a and Table 3). Although all Kss values for naphthalene and SDS were also larger than the Kmic values, the difference was much smaller; in fact, values approached as sorbed SDS levels increased.

The dependence of Kss on sorbed SDS levels appears to be qualitatively consistent with proposed surfactant structures at mineral surfaces (Fuerstenau and Wakamatsu, 1975; Holsen et al., 1991; Chandar et al., 1987). For example, the configuration of adsorbed SDS in Region I is expected to be different from that in Regions II and III, where micelle-like structures are thought to exist at these relatively higher sorption densities. Apparently, these differences in sorbed surfactant structure that result from regional sorption mechanisms and sorption densities lead to regional differences in Kss values.

In contrast to SDS, calculated individual Kss values for Tween 80 did not show distinct ranges but instead decreased monotonically over the surfactant concentration range examined (Fig. 5c,d). In addition, although the decreasing trend exhibited by individual Kss values (filled circles) agreed qualitatively with that observed for the isotherm Kss values (open circles), there was not the same good agreement in values as was obtained for SDS. This is most likely attributable to the fact that Tween 80 micelles existed throughout the surfactant concentration range studied (e.g., see CMC arrows in Fig. 5); as can be seen in eq 2, calculation of individual Kss values for Tween 80 thus required the additional input values (and associated uncertainties)of Kmu and Smic. Conversely, the close agreement in values for SDS resulted primarily because, with the exception of Region III, HOC partitioning to sorbed SDS did not have to compete with partitioning to SDS micelles over the majority of the SDS sorption regions examined (Fig. 5a,b). Although average Kss values for Tween 80 were larger than corresponding Kmic values, some of the individual and isotherm Kss values actually fell below Kmtc. Whether this proves to be a general observation for nonionic surfactants or is merely a result of the complications from working with Tween 80 will need to be investigated in future studies.

The influence of ionic strength (as added NaCl) on phenanthrene partitioning to sorbed SDS was investigated at pH 6.5 (Table 5 and Fig. 6a). As discussed previously, SDS sorption increases with increasing ionic strength; therefore, one would logically expect phenanthrene sorption to the solid phase (which includes sorbed SDS aggregates) to correspondingly increase with ionic strength. To determine whether SDS molecules sorbed under different ionic strength conditions exhibit different partitioning characteristics, carbon-normalized partition coefficients (Koc) were calculated from the linear distribution coefficients (KD). The organic carbon fraction (foc, % mass of carbon sorbed per sorbent mass) was calculated from SDS sorption results for the same pH and dose (i.e., 2 mM). Kocvalues for sorbed SDS were approximately 10 times greater than those for SDS micelles (Tables 2 and 5), indicating a higher affinity of the sorbed SDS molecules for phenanthrene. Sorbed SDS Koc values increased with increasing ionic strength; although the percentage increase in phenanthrene partitioning was slightly less than that observed for SDS micelles (i.e., 20% versus 30% for a NaCl increase from 0.001 to 0.1 M), the increase was still larger than the estimated 6% increase expected due to salting out.

Phenanthrene sorption isotherms for varying pH conditions at 0.1 M NaCl were observed to be linear (Fig. 6b); therefore, distribution coefficients were determined using linear regression analysis and Koc values were calculated as before (Table 5). values decreased with increasing pH, which can be attributed primarily to decreased SDS sorption at high pH (Holsen et al., 1991). Interestingly, Fig. 7 and Table 5 show that sorbed SDS Koc values can be divided into two distinct regions depending on whether the solution pH is above or below the PZC of kaolinite; Koc values below the PZC (greater SDS sorption) are much greater than those above the PZC (less SDS sorption). This observation contrasts with the results presented earlier where phenanthrene Koc values decreased as the amount of sorbed SDS increased under fixed solution chemistry conditions when the pH was above the kaolinite PZC (Fig. 5 and Table 4). However, those results showed that

Table 5. Phenanthrene distribution (Kn) and organic carbon normalized partition (Km) coefficients to sorbed surfactants on kaolinite for varying solution chemistry conditions. "_

pH

SSorb (fimol/g)

KDb(lJg)

R2

N c

foe" (%)

Kk (L/g)

SDS

3.38

10.51

0.556 ± 0.012

0.976

18

0.1509

368 ± 7.6

3.85

9.99

0.383 ± 0.010

0.992

15

0.1434

264 ± 5.9

4.32

9.80

0.222 ± 0.005

0.988

18

0.1406

155 ± 3.5

5.25

8.07

0.100 ± 0.020

0.999

15

0.1156

86.5 ± 1.4

5.87

7.79

0.098 ± 0.008

0.991

18

0.1122

87.6 ± 7.4

6.50

IS.

= lxlO"3

M

0.84

0.0144 ± 0.002

0.996

18

0.0121

119 ± 14.1

I.S.

=lxlO2

M

0.87

0.0168 ± 0.001

0.999

15

0.0125

135 ± 9.6

I.S.

=1x101

M

3.51

0.0727 ± 0.003

0.983

18

0.0505

144 ± 5.9

7.85

3.26

0.068 ± 0.004

0.992

15

0.0470

145 ± 10.4

8.40

2.91

0.055 ± 0.005

0.997

18

0.0410

131 ± 10.4

9.11

2.62

0.051 ± 0.002

0.991

15

0.0369

138 ± 5.6

Tween 80

4.60

12.90

0.1301 ± 0.0015

0.994

18

0.926

90.9 ± 2.1

6.25

12.13

0.0289 ± 0.0008

0.987

18

0.871

71.2 ± 3.4

7.85

9.47

0.0157 ± 0.0005

0.973

15

0.679

89.6 ± 2.7

a Values for KD and Koc are ± SD. Ionic strength = 0.1 M unless noted. Dose was 2 and 1.527 mM for SDS and Tween 80, respectively. b Obtained from linear regression analysis of experimental isotherms.c Number of data points. d For SDS,/oc was calculated from the theoretical carbon content specified by the chemical formula. For Tween 80, the carbon content was determined by TOC measurements.

higher partition coefficients for sorbed SDS are obtained in Region I of the SDS sorption isotherm whereas the sorbed SDS here exists as surface aggregates such as admicelles or hemicelles for the particular dose used (i.e., 2 mM SDS, which corresponds to sorption in Region II and possibly III as shown in Figs. 2 and 3 a).

To further examine differences in partitioning to sorbed SDS aggregates, phenanthrene sorption tests were conducted on kaolinite loaded with the same amount of SDS under different pH conditions (i.e., q = 1.95 ^mol/g as shown in Fig. 3a). Ideally, phenanthrene isotherms and distribution coefficients should have been identical regardless of pH because of the constant foc value. However, it is obvious from Fig. 8 that the Rvalue for the isotherm below the PZC of kaolinite (i.e., the one conducted at pH 3.2) is much greater than those two conducted above the PZC (i.e., pH 7.8 and 10.1). Koc values calculated from the KD values and sorbed SDS amounts were 435 ± 9.5, 122 ± 2.1, and 150 ± 3.0 L/g for pH 3.2, 7.8, and 10.1, respectively. These values agree well with those obtained using a constant SDS dose of 2 mM (Table 5); additionally, the values fall into the same two regions described above. These results suggest that sorbed SDS aggregates formed at pH values below and above the PZC of a mineral surface will have different HOC partitioning characteristics. A variety of interactions between SDS and kaolinite can occur depending on the surface charge of kaolinite, thus resulting in different sorbed surfactant structures (e.g., SDS monolayers versus bilayers) (Ko et al., 1998a). Consequently, sorbed surfactant partitioning results can thus be interpreted as HOC molecules partitioning to these different surface structures (Ko et al., 1998a).

0.05

0.05

AQUEOUS PHENANTHRENE CONCENTRATION <MM)

Figure 8. Phenanthrene sorption isotherms on kaolinite loaded with the same amount (1.95 pmol/g) of sorbed SDS under different pH conditions. The ionic strength was 0.1 M NaCI, and the kaolinite concentration was 100 g/L. Error bars are smaller than the symbols. Solid lines are the best fits from linear regression and represent the KD values. Adapted from Ko et al. (1998a).

AQUEOUS PHENANTHRENE CONCENTRATION <MM)

Figure 8. Phenanthrene sorption isotherms on kaolinite loaded with the same amount (1.95 pmol/g) of sorbed SDS under different pH conditions. The ionic strength was 0.1 M NaCI, and the kaolinite concentration was 100 g/L. Error bars are smaller than the symbols. Solid lines are the best fits from linear regression and represent the KD values. Adapted from Ko et al. (1998a).

The influence of pH on phenanthrene partitioning to sorbed Tween 80 is summarized in Fig. 9 and Table 5. The Tween 80 dose for these experiments was 1.67 mM, which is well above the CMC value; therefore, Tween 80 micelles contributed to the overall phenanthrene distribution. As can be seen, Kd values decrease with increasing pH. This trend was expected because of the decrease in Tween 80 sorption with increasing pH (Fig. 3b) and the concurrent increase in the number of Tween 80 micelles. Similarly to sorbed SDS, the affinity of sorbed Tween 80 for phenanthrene can be evaluated by calculating Koc values. Because of the presence of micelles, however, a mass balance that includes micelles and sorbed surfactant is required to calculate the partitioning capacity of sorbed Tween 80, expressed previously as Kss in eq 2. KS5 values calculated with this equation were then converted to Koc values using the average carbon fraction of Tween 80 (i.e., 0.548 as determined by TOC analysis). As shown in Table 5, Koc values for sorbed Tween 80 are similar in magnitude to those reported earlier for Tween 80 micelles (Table 2). Koc values for the pH conditions investigated here showed no significant differences, implying that the sorbed Tween 80 aggregates had similar structures and/or physicochemical characteristics.

0.05

0.05

AQUEOUS P HEN ANTHRENE CONCENTRATION (|iM)

Figure 9 Phenanthrene sorption isotherms on kaolinite with sorbed Tween 80 for three pH values at 0.1 M NaCl. Error bars for some data points are smaller than the symbols. Kaolinite concentration was 100 g/L and the Tween 80 dose was 1.67 mM.

AQUEOUS P HEN ANTHRENE CONCENTRATION (|iM)

Figure 9 Phenanthrene sorption isotherms on kaolinite with sorbed Tween 80 for three pH values at 0.1 M NaCl. Error bars for some data points are smaller than the symbols. Kaolinite concentration was 100 g/L and the Tween 80 dose was 1.67 mM.

Depending on the desired treatment methodology and goals, addition of surfactants to a subsurface system should either increase HOC distribution coefficients (i.e., immobilization approach) or decrease them (i.e., mobilization objective as in many SEAR applications). For example, distribution coefficients for phenanthrene and naphthalene to kaolinite are 0.002 and 0.0003 L/g, respectively (Table 4). Therefore, if enhanced mobilization of these HOCs in a similar type of aquifer system was desired, addition of a surfactant would have to bring the distribution coefficients below these values. However, as can be seen in Fig. 5, all distribution coefficients for the surfactant doses investigated here are larger than these values, even when the doses and subsequent aqueous surfactant concentrations are well above the CMC. This observation results from a combination of surfactant sorption followed by HOC partitioning to the sorbed surfactant.

IMPLICATIONS FOR SUBSURFACE REMEDIATION ALTERNATIVES USING SURFACTANTS

A specific numerical example is enlightening. The KD value of naphthalene for an initial Tween 80 dose of 7.63 mM is 0.0075 L/g (i.e., the last point in Fig. 5d corresponding to Ssurf = 5.32 mM; this dose also corresponds to a sorbed Tween 80 concentration of -29 |amol/g as can be seen in Fig. 1). This value greatly exceeds the A^when no Tween 80 is present in the system. In other words, addition of Tween 80 to this model system or a similar type of aquifer system would lead to an increase in naphthalene retardation, not a decrease as would be desired for SEAR applications. Because the plateau for Tween 80 sorption occurs near the above dose (i.e., 770 x CMC, Fig. 1), it is expected that much higher doses would be needed to overcome the enhanced retardation effects caused by Tween 80 sorption to the aquifer matrix. These very high surfactant doses, however, would likely not be practical for real world applications. A more useful approach, therefore, would be to utilize the Tween 80 to increase retardation in this particular system; the optimum dose to use would correspond to a maximum KD value.

In any evaluation of a remediation scheme utilizing surfactants, the effect of dose on HOC distribution coefficients must be quantified. Very often, only one coefficient value for HOC partitioning to sorbed surfactants has been reported in the literature, presumably because the experimental data covers only the sorption regions where the surfactant molecule interactions dominate at the surface (Nayyar et al., 1994; Park and Jaffe, 1993). However, all of the characteristic sorption regions will develop during an in-situ SEAR application as the surfactant front (i.e., mass transfer zone) advances through the porous medium. Therefore, the relative role ofregional HOC partition coefficients to sorbed surfactant should be considered in any remediation process. Finally, the porosity or solid volume fraction for the particular subsurface system must be taken into account when surfactant sorption is quantified.

In addition to the equilibrium aspects discussed above, the rates of all forward and reverse processes, including HOC partitioning to surfactant micelles and sorbed surfactant, and HOC and surfactant sorption to aquifer materials, may need to be quantified (e.g., Deitsch and Smith, 1995; Sahoo and Smith, 1997; Yeom et al., 1995, 1996; Deitsch, following chapter). By taking into account all of these equilibrium and rate processes, numerical models may become effective tools for the screening/design (e.g., Ko and Schlautman, 1998) and/or application (e.g., Smith et al., 1997; Sahoo et al., 1998) of SEAR processes. To provide a truly realistic prediction/analysis of SEAR effectiveness, site-specific data including important HOC and surfactant rate processes must be acquired for input to the models. With the appropriate information and proper analysis, addition of surfactants can then be optimized to achieve increased performance, shorter remediation times, and/or lower overall remediation costs.

Selection of an appropriate surfactant is critical to the ultimate success of SEAR applications. Based on the results presented here as well as from other studies (e.g., Sahoo et al, 1998), it can be seen that surfactants having high HOC solubilization capabilities and low solid phase sorption potentials are desirable for SEAR processes. Unfortunately, these two desirable traits are often mutually exclusive. Therefore, other types of HOC solubilizing/facilitating agents may need to be considered in addition to conventional micelle-forming chemical surfactants for SEAR applications. For example, cyclodextrins have been shown to be effective for removing sorbed HOCs from contaminated aquifer systems because of their negligible sorption loss to the solid phase (Ko et al., 1999, 2000, and references therein); in some cases, the absence of cyclodextrin sorption will more than compensate the fact that they tend to have lower solubilization capabilities than do conventional surfactant micelles. Additionally, naturally-occurring organic materials (NOMs) such as humic and fulvic acids have been shown to be effective for removing sorbed HOCs from saturated aquifer materials (e.g., Johnson and Amy, 1995; Johnson, 2000). Although NOMs also adsorb to aquifer solids, their sorption tends to be less than that for conventional surfactants and, unlike conventional surfactants, the ability of HOCs to partition to sorbed NOMs is less than that for the dissolved NOM constituents (e.g., Schlautman and Morgan, 1993; Hur and Schlautman, 2000). Therefore, even though some fraction of NOMs may be lost to the immobile phase by sorption, overall HOC distribution coefficients will not show the large adverse impacts observed for conventional surfactants in SEAR processes as demonstrated in Fig. 5. Finally, the potential for aquifer plugging by surfactants or other facilitating agents must be considered during the selection process (e.g., Ko et al., 2000). For example, in some cases surfactant flushing of saturated soil or aquifer systems can disperse colloidal particles and thus lead to clogging of the pores, which significantly reduces flow through the contaminated subsurface material (Abdul et al., 1990).

One last consideration during the selection process of a suitable surfactant or other facilitating agent must be an examination of toxicity and biodegradability issues so that no adverse impact on the environment or on human health occurs. For example, upon completion of SEAR, any residual HOCs or surfactants remaining in the aquifer should be easily biodegradable or, at minimum, have a relatively low toxicity. Because NOMs and cyclodextrins are naturally-occurring materials, they may have less of an environmental impact than conventional chemical surfactants and may be more readily acceptable to the public for use in SEAR applications. Bioavailability of the HOCs partitioned to the facilitated phase may need to be considered if biological treatment of the effluent from a SEAR process is desired. For example, recent studies have shown that HOCs solubilized in surfactant micelles may or may not be readily available for biodegradation depending on the specific surfactant types and concentrations used (e.g., Laha and Luthy, 1991; Guha and Jaffe, 1996; Guha et al., 1998; Yeh et al., 1999); for cyclodextrins, a recent report suggests that HOC biodegradation will be enhanced when they are used as solubilizing agents (Wang et al., 1998). Finally, it has been reported that the biodegradation of some nonionic surfactants may actually stimulate HOC biodegradation in contaminated soils (Tiehm et al., 1997).

0 0

Post a comment