Environ. Sci. Technol. 1994, 28, 231-237
Use of Cationic Surfactants To Modify Soil Surfaces To Promote Sorption and Retard Migration of Hydrophobic Organic Compounds Julla Wagner, Hua Chen, Bruce J. Brownawell,t and John C. Westall'
Department of Chemistry, Oregon State University, Corvallis, Oregon 97331-4003 Cationic surfactants can be used to modify surfaces of soils and subsurface materials to promote adsorption of hydrophobic organic compounds (HOC). Batch and column experiments were performed to investigate this phenomenon with the cationic surfactant dodecylpyridinium (DP), a series of chlorobenzenes as representative HOC, and a low organic carbon aquifer material (Lula). The adsorption isotherm of DP was highly nonlinear; at micromolar concentrations, DP was adsorbed strongly but not irreversibly; at millimolar concentrations, adsorption was relatively weak. Distribution ratios of the chlorobenzenes varied nonlinearly with DP loading. The elution of chlorobenzenes from columns packed with DP-treated aquifer material was examined; a transport model based on the results of the batch experiments and the local equilibrium assumption yielded an acceptable approximation for the coelution of DP and HOC from the column. It is concluded that treatment of surfaces with cationic surfactants shows promise as a means of promoting HOC sorption in a variety of treatment processes.
Introduction Cationic surfactants can be used to modify the surfaces of soils and subsurface materials to promote sorption of hydrophobic organic compounds (HOC) to the modified surface (1-7). The sorption of HOC to a soil surface reduces the mobility of the HOC in the subsurface environment. While this process will occur naturally as surfactants are introduced into the environment, it might also be a basis for remedial action at contaminated sites. For example, cationic surfactants could be injected into an aquifer downgradient from a source of contamination to provide a temporary barrier against migration. Alternatively, they could be used in conjunction with in-situ bioremediation to eliminate the HOC-the surfactant could retard a contaminant such as trichloroethene (TCE) long enough for a relatively slow process such as reductive dehalogenation to be effective. Related applications include the control of vapor emissionsfrom in-situ biofilters and the increase in retention time of HOC in slurry bioreactors. Issues related to potential field applications are discussed further in the final section of this paper. While this study deals with the immobilization of HOC through strongly sorbing cationic surfactants, other researchers have discussed mobilization of HOC through the use of more weakly sorbing anionic or nonionic surfactants, primarily through partition into micellar suspensions or alteration of surface tension at the aqueousnonaqueous interface (8). In this paper, we report on batch and column experiments designed to examine the feasibility of using cationic surfactants to modify the surface of a low organic carbon
aquifer material to promote the adsorption of HOC. Several questions arise with respect to this application: (i) Will the cationic surfactant adsorb strongly enough that it will not itself migrate? (ii) Will the cationic surfactant adsorb so strongly that it will be impossible to disperse it in the subsurface environment? (iii) Will the cationic surfactant promote the adsorption of the HOC to a sufficient degree? (iv) How well can column behavior be predicted from batch behavior? (v) Is it possible to engineer a system to determine the amount of cationic surfactant that must be applied to obtain a preselected retardation of HOC? These questions are addressed in this paper. Experimental Approach. The cationic surfactant used in this study was dodecylpyridinium (DP), which was selected for the ease of analysis by UV spectrophotometry and its resistance to biodegradation in slurries of soils and subsurface materials. Furthermore, in a limited number of experiments, we have found that its behavior resembles that of other cationic surfactants. Other criteria would have to be considered if field studies were planned, as discussed later. The hydrophobic organic compounds selected were chlorobenzene homologs. These compounds are moderately hydrophobic (3.38 < log KO, < 5.03) (9). The work was carried out in three phases. In phase I, we performed batch experiments to determine the adsorption isotherm of DP on the aquifer material and the distribution ratios of each HOC on the DP-treated aquifer material. The distribution ratios of each HOC on the DPtreated aquifer material were determined as a function of DP loading: K = CHOC(S)/CH~C(W) (1) where K is the distribution ratio (L k g l ) of an HOC at a particular DP loading, CHOC(S)is the amount of HOC on the sorbent per mass of sorbent (mol kg'), and CHOC(W) is the equilibrium concentration of HOC in solution (mol L-1).
In phase 11, we examined the breakthrough of chlorobenzenes on columns with untreated sorbent and with sorbent uniformly treated with DP. For the DP-treated sorbent, the sorbent was first equilibrated with 10p M DP and then maintained in that condition by the addition of 10 pM DP to all solutions flowing into the column. These conditions are not those that one would expect to encounter in a field application, but they are useful in evaluating the applicability of data generated in batch experiments to column experiments and the validity of the local equilibrium assumption in these systems. From the distribution ratio determined in batch experiments (eq l), the retardation factor for column experiments can be calculated:
+ Present addreas: Waste Management Institute, Marine Sciences Research Center, SUNY, Stony Brook, NY 11794-5000.
R = 1 + Kp,((l - ci)/ei) (2) where R is the retardation factor, ps is the density of the solid (g ~ m - ~and ) , ci is the porosity (pore volume in packed bed divided by total volume occupied by packed bed).
0 I994 American Chemlcal Society
Environ. Sci. Technol., Vol. 28, No. 2, 1994 281
0013-938X/94/0928-0231$04.50/0
The transport equation used for phase I1 of the work is the one-dimensional advection-dispersion model for steady flow through a homogeneous porous medium with local equilibrium: (3) where C is the concentration in the aqueous phase, t is time, D is the dispersion coefficient, R is the retardation factor (eq 2), z is the distance along the column, and u is the flow velocity. In phase 111, we investigated the breakthrough of chlorobenzenes on columns that were initially uniformly coated with DP, but when the chlorobenzene pulse was applied to the column, the supply of DP to the column was switched off, and the DP was allowed to desorb as the chlorobenzenes passed through the column. This design is closer to what would actually be used in the field-a pulse of the cationic surfactant would be applied to treat the surfaces of aquifer particles, but the cationic surfactant would not be injected into the aquifer over an extended time.
Methods Materials. The sorbent was Lula N6 (Lula), a low organic carbon aquifer material (fraction of organic carbon, f O c = 0.2 g of C/kg), obtained from the R. S. Kerr Environmental Research Laboratory, US. Environmental Protection Agency, Ada, OK. The material is essentially an iron oxide-coated silica sand. The fraction that passed through a 250-pm sieve was used in all experiments for this study. Bouchard (6)reported some properties of the Lula fraction that passed through a 250-pm sieve. Stauffer (10) reported an analysis for Lula ground to pass through a 1-mm sieve. The cationic surfactant was dodecylpyridinium (DP, molar mass 248 g mol-l). Labeled N-[1-14Cldodecylpyridinium bromide, as described by Brownawell et al. (II), was obtained from Pathfinder Laboratories Inc. (St.Louis, MO). Unlabeled dodecylpyridinium chloride monohydrate was used as received from Aldrich. Chlorobenzeneswere obtained from Ultrascientific and American Tokyo Kasei. The compounds used for the study were 1,2-dichlorobenzene (DCB), 1,2,3-trichlorobenzene (TCB), 1,2,3,4-tetrachlorobenzene(TeCB), and pentachlorobenzene (PeCB). The compounds used for internal standards were 1,4-dichlorobenzene(1,CDCB) and 1,2,4,5tetrachlorobenzene (1,2,4,5-TeCB). All aqueous solutions were prepared with deionized water obtained from a Millipore Milli-Q system. The electrolytes CaCl2 (Baker) and Ca(N03)2 (EM Science) were used as received. "Non-spectro" hexane (Burdick & Jackson) was used for preparing chlorobenzene calibration standards and for extracting chlorobenzenesfrom aqueous samples. High-purity acetone (High Purity Chemical, Portland, OR) was used for preparing chlorobenzene solutions for spiking into aqueous solutions. Phase I: Batch Experiments. The procedure for determining the adsorption isotherm of DP on Lula aquifer material was described in detail by Brownawell et al. (11). The distribution ratios of chlorobenzenes (DCB, TCB, TeCB, and PeCB) between 0.01 M CaClz and DP-treated Lula aquifer material were determined as a function of DP loading in conventional batch experiments. For each 232 Envlron. Scl. Technol., Vol. 28, No. 2, 1994
DP loading, chlorobenzene distributions were determined for a single total concentration of chlorobenzenes (i.e., full multipoint isotherms were not obtained); other studies have indicated that the HOC isotherms are linear (2,3,5). Equilibration were carried out in 25-mL Corex centrifuge tubes with Teflon-lined caps. To the tubes containing 0.500 g of washed aquifer material were added 10 mL of 0.01 M CaClz with various concentrations of DP and 10 mL of 0.01 M CaC12 containing the chlorobenzenes. The slurries were shaken for 4 h at 500 rpm on a New Brunswick G-24 shaker at 25 OC and then centrifuged for 30 min at 25 OC at 1OOOOg. The solution was then analyzed for DP, and the solution and the sorbent were analyzed for chlorobenzenes. The amount of DP associated with the sorbent was determined from the difference between the initial aqueous DP concentration and the equilibrium concentration of DP in solution. After centrifugation, 10-mL aliquots of the supernatant were collected and analyzed for DP with a Hewlett-Packard 8452A diode array spectrophotometer and thermostated cell at 25 "C. Chlorobenzenes were stripped from the DP samples with an airstream prior to analysis. The overall precision for aqueous samples ranged from 2 % relative standard deviation (RSD) to 5 % RSD. Both the solution and the sorbent were analyzed for chlorobenzenes. The concentrations of chlorobenzenes in solution were determined by extracting the chlorobenzenes into hexane containing the internal standards. The amounts of chlorobenzeneson the sorbent were determined by extracting the chlorobenzenes into acetone and then into hexane. To facilitate the separation of the acetone from the hexane, deionized water was added to each tube; then the tubes were centrifuged at lOOOOg at 25 "C for 30 min. The concentrations of chlorobenzenes in hexane were determined with a Hewlett-Packard 5880A gas chromatograph equipped with a J&W Scientific DB-5 fused-silica capillary column (30 m X 0.32 mm i.d.) and an electron capture detector. The overall precision for aqueous and sorbent samples ranged from less than 1% RSD to 5 % RSD for both types of samples. Phase 11: Column Characteristics. The properties of the column are summarized in Table 1. Stainless steel and low dead-volume unions were used throughout. In one experiment (the determination of breakthrough curves of chlorobenzeneson untreated aquifer material), a Kontes Chromaflexglass column with Teflon end fittings was used. However, the all-stainless-steel system was found to be superior to the glass-Teflon system in reducing adsorption of these moderately hydrophobic compounds to surfaces of the apparatus. All columns were dry packed with Lula aquifer material according to a method described by Snyder and Kirkland (12). C02 was passed through the packed columns to flush the air out of the pore space before the columnswere wetted with 0.01 M CaC12. All eluents were filtered through 0.20.4-pm filters and degassed by sonication under vacuum before use. The holdup volume of the column was determined with a NO3- tracer. Fractions of the effluent from the UV detector were collected, and the mass was determined to allow the true flow rate to be determined. No independent measurements (e.g., 3H breakthrough) were obtained to substantiate the assumption that NO3- is indeed an ideal nonsorbing tracer for this aquifer material.
Table 1. Physical Properties of Column Packed with Lula Aquifer Material symbol’
descriptiona
L
D
column length internal diameter sorbent mass holdup volume dead volume flow rate dispersion coefficient
vi
interstitial volume
di
MU Vm vd F C
experimentalb Determined Directly 5.000 f 0.001 1.000 f 0.001 6.84 f 0.01 1.73 0.14 0.14 f 0.01 3.99 f 0.03 0.0003 f 0.0001 Derived 1.59 f 0.14
*
v m - vd
X VS
T
U PB PO
units
5.00 1.00 6.84 1.71 0.14 3.99 0.0002d
cm cm g mL mL mL h-1 cm2 s-1
1.57
mL
empty column volume = 3.14(di/2PL volume of sorbent
3.927 f 0.008
3.927
mL
2.34 f 0.14
2.36
mL
porosity = Vi/X residence time = Vi/Fc X (3600 s h-1) flow velocity =L/7 bulk density = MdX sorbent density
0.40 f 0.04
0.40
1435 f 120
1417
8
0.0035 f 0.0003
0.003
cm s-1
1.742 f 0.003
1.742
g mL-1
2.92 f 0.17
2.90
g mL-1
=x-vi
ci
modelc
Mu/Vs 0 Nomenclature follows IUPAC chromatography conventions (18). Estimated uncertainty in experimentally determined value is given. Values used in the models. As described in the text, several different values for D were used. D = 0.0002 for model of HOC breakthrough on untreated aquifer material and for model of HOC breakthrough on aquifer material equilibrated with 10 pM DP (Figure 3). D = 0.02 for DP transport (Figures 4 and 5). D = 0.002 for HOC transport through column with time-varying DP (Figure 5).
0”
-I
-5
L -7
F -6
-5
-4
-3
-2
Figure 1. Adsorption isotherm of Nclodecylpyridinium on Luia aquifer Experiments conducted material (0.005 kg/L) In 0.01 M CaCI2. (0) with 14C-iabeledDP and direct determinatlon of DP In solution and on the sorbent. (0) Experiments conducted with unlabeled DP and determination of DP in solution by UV spectrophotometry and DP on the sorbent by dlfference. The dashed line is the empirically adjusted isotherm used in the transport model In Figure 4. The two symbols (0)on the isothermdeslgnate the DP concentration at which the phase I1 experiment (10 pM, Figure 3) was carried out and the initial DP concentration for the phase I11 experiment (3 mM, Figures 4 and 5).
Chlorobenzene Breakthrough Curves. To determine the breakthrough of HOC on untreated aquifer material, the column was equilibrated with 0.01 M CaClz. Then a pulse of 0.01 M CaClz spiked with chlorobenzenes was injected, followed by more 0.01 M CaC12. The flow rate was 4 mL h-l throughout. To determine the breakthrough of HOC on aquifer material treated with DP, the sorbent in the colum was first loaded with DP by pumping 10 pM DP in 0.01 M CaClz through the column. According to the batch isotherm, 4.3mmol k g l DP on the surface is in equilibrium
with 10p M DP in solution. Then a pulse of chlorobenzenes (DCB, TCB, TeCB, and PeCB) in 0.01 M CaClz with 10 pM DP was applied to the column. Finally, the chlorobenzenes were allowed to desorb as chlorobenzene-free 0.01 M CaClz with 10 pM DP was passed through the column. Phase 111. First, the aquifer material in the column was loaded with DP by pumping a solution of 0.01 M CaClz with 3 mM DP through the column. Next, 20 mL of 0.01 M CaC12 (without DP) was pumped through the column. Then, a pulse of chlorobenzene-spiked 0.01 M CaClz (without DP) was pumped through the column for 75 h. Finally, 0.01 M CaClz without chlorobenzenes and without DP was pumped through the column to allow the chlorobenzenes to desorb.
Results and Discussion Phase I: Adsorption Isotherm of DP. The adsorption isotherm of DP on Lula aquifer material in 0.01 M CaClz (Figure 1)is similar to that obtained by Brownawell et al. (11)for DP on Lula aquifer material in 0.1 M NaC1. The slope of the log-log plot is approximately 0.6 in the low concentration range; on alinear scale (not shown), the isotherm has an almost step-function appearance. In the high concentration range, near the critical micelle concentration, the isotherm levels off at approximately the cation-exchange capacity of the sorbent. The fOcof untreated Lula is about 0.02%; near the plateau of the isotherm, the amount of DP on the surface is about 40 mmol/kg, or fOc= 1% Thus, the f O c of the sorbent can be increased by a factor of about 50 by loading with DP. A significant feature of this isotherm is its nonlinearity: at 3 mM DP in solution, the distribution ratio of DP is approximately 10 L k g l , while at 10 pM DP in solution, the distribution ratio is approximately 5000 L kgl. Thus,
.
Environ. Scl. Technol., Vol. 28, No. 2, 1994 233
1500
I
cn
Table 2. Dependence of K of Chlorobenzenes on DP 9 Loading of Lula Aquifer Material. compound 1,S-DCB 1,2,3-TCB 1,2,3,4-TeCB PeCB
1000
Y
--1
U
\
500
Y
0
I
I
I
I
0.0
0.4
0.8
1.2
foc
/
%
Flguro 2. Partition constant of chlorobenzenes between 0.01 M CaCi2 and Lula aquifer material (0.025 kg/L) with different loadings of DP: (O)DCB,(A)TCB, (O)TeCB,(O)PeCD. Thesoliilineswerecalculated from the polynomials in Table 2.
the lower the concentration of DP in solution, the more tenaciously it adheres to the sorbent. This behavior is ideal for establishing absorbent barriers by injecting a pulse of cationic surfactant into the ground. At high concentrations the surfactant would be mobile and would disperse to cover a zone, while at low concentrations it would be immobile and provide a coating to particles to retard migration of HOC. Mechanism for DP Adsorption. With the data that are available, it is difficult to say too much about the extent of adsorption of DP to the different fractions of the sorbent (silica sand, iron oxide coating, 1:l clay, 2:l clay). Brownawell et al. (11)have shown that the adsorption of DP to Lula is independent of pH over the range 4-7. This pH independence seems to suggest an ion-exchange mechanism and a significant role of the clay fraction, but the capacity of the Lula material for DP is greater than one might expect considering the small fraction of clay in the Lula material. Thus, while adsorption to the clay fraction [and the appealing simple models for adsorption of organic cations of clays (13)]is undoubtedly important, it may not account for all of the DP adsorption. Further experiments with the pure fractions of Lula would be necessary to substantiate this speculation. Distribution of Chlorobenzenes on DP-Treated Aquifer Material. The distribution ratios of the chlorobenzenes between the DP-treated aquifer material and 0.01 M CaC12 are shown in Figure 2 as a function of DP loading. Third-order polynomials were fit to these data for incorporation in the transport model of chlorobenzenes on DP-treated aquifer material. The coefficients for these polynomials are given in Table 2. A high degree of uncertainty is associated particularly with the zero-order term, since the extent of sorption to the untreated aquifer material is so low. Therefore, the zero-order terms of the polynomials in Table 2 were set equal to the values of K determined from a visual fit of the experimental breakthrough curves for the untreated sorbent. Values of the distribution coefficients calculated from the polynomials are represented by the solid lines Figure 2. At surfactant concentrations above the CMC of DP (not shown), the distribution ratios of the chlorobenzenes decreased with 234
Envlron. Scl. Technol., Vol. 28, No. 2, 1994
a0
a1
0.2 0.7 2.3 9
0.0915 0.1178 0.1244 1.194
a2
a3
0.0255 -0.0002 0.1261 -0.0011 0.2844 -0.0022 0.5428 -0.0046 a K = ao + UlCDp(s) + a z C ~ p ( s + ) ~U & ~ p ( s ) ~with , &p(S) in units of mmol/kg and K in units of L/kg. The coefficients are reported with several digits to allow the function to be calculated accurately; the number of digits does not reflect the precision of the coefficients. Values for a0 were set equal to column K values determined from breakthrough of chlorobenzenes on untreated sorbent (Table 3); values of al,az, and a3 were then determined by a weighted nonlinear least-squares fit of the data in Figure 2. Uncertainty in these parameters is estimated to be a few percent, based on the reproducibility of the data in Figure 2.
increasing DP concentration, as would be expected in the presence of micelles. The values of log K predicted for the HOC from Kow(9) and foc by the equation of Schwarzenbach and Westall (14) are lower than the values found here. This result indicates that DP accommodatesthe chlorobenzenesmuch better than natural organic matter, per carbon atom. The correlation developed by Schwarzenbach and Westall (14) included all of the compounds used in this study, and the sorbents were a variety of environmental materials including soils, sediments, aquifer materials, and activated sludge. The Lula aquifer material used in this study has a relatively low clay content and a relatively low organic carbon content. For aquifer materials with higher clay contents, we would expect stronger adsorption of the cationic surfactant (11) and stronger immobilization of the HOC. For aquifer materials with higher organiccarbon contents, the effect of the cationic surfactant would be less pronounced. Breakthrough of Chlorobenzenes. The breakthrough curves of the chlorobenzenes through the untreated and treated aquifer material are shown in Figure 3. The lines in Figure 3 were calculated from a finite difference approximation of eq 3. The values of K and D that were used in the model for the untreated sorbent were determined from the experimental breakthrough curves for the untreated sorbent by visual fit; the other parameters were taken from Table 1. These values of K are listed in Table 3 along with those determined in batch experiments; they are not significantly different. The value of the apparent dispersion coefficient, D, was set to 0.0002 cm2 s-l; the value of D estimated from the nitrate breakthrough was 0.0003 f 0.0001 cm2 s-l. For the treated sorbent, the polynomials in Table 2 were used to determine the values of K for the models of the breakthrough curves of chlorobenzenes on DP-treated sorbent. The value of D that was used for the model of the untreated sorbent was also used for the model of the DP-treated sorbent. This experiment has important implications for both applications and theory. First, note that the amount of DP adsorbed was only 4.3 mmol k g l ; this amount is equivalent to a DP fOc of about 0.09% compared to the natural fOc of 0.02 5%. As seen in Figure 2, much higher loadings are possible. Even at this low amount of DP, retardation factors for the HOC are approximately three
1.2 1 -
1.2
1 .o
1 .o
8 0.8
u , u
0.6 0.4 0.2 0.0 0
0
0
4
8
12
10
20
30
40
50
I
I
1 .o
0.6
0.4
0.2
Pore Volume
Pore Volume 1.2
-
20 0
16
\ V
0.8
0.0
0
25
8 0.8
, u 0
50
75
100
125
150
175
Pore Volume
0.6
1.2
0.4 0.2
1 .o
0.0 0
20
40
60
0
80
50
100 150 200 250
Pore Volume Pore Volume Figure 3. Breakthroughcurves of the chlorobenreneson(0)untreated Lula aquifer material and (A)Lula aqulfer material equilibrated wlth 10 pM DP. The solld llnes were calculated from a finite difference approximationto the advectlon-dlsperslonequation, with the parameters in Table 1 and the coefficients in Table 2.
0 0
\ 0
0.8 0.6
0.4 0.2
0.0
Table 3. Retardation Factors, R, and Distribution Ratios, K,of Chlorobenzenes between 0.01 M CaCl2 and Lula Aquifer Material untreated sorbent DP-treated sorbenta batch Kb polynomial Ke batch K b column Kc (Lkgl) Rf (L k g l ) Rd ( L k g l ) (Lkgl) DCB