Chapter 3  Sorption of hydrophobic organic compounds (HOC).

Introduction

Tables for Chapter 3 may be found at this link.

Hydrophobic organic compounds (HOC) are common contaminants in soils and sediments. HOC include aromatic compounds in petroleum and fuel residue, chlorinated compounds in commercial solvents and chemicals no longer produced in the United States, for example DDT (Luthy et al, 1997). For hydrophobic organic compounds, sorption is a critical process controlling the fate of these chemicals in ground water. It is important to understand and be able to predict the transport and fate of HOC for choosing effective remediation procedures. Sandy glacial aquifers are an important source of ground water. The Long Island sandy upper glacial aquifer is a typical example of glacial aquifer. The problem of HOC contamination of the sediments is especially important for Long Island, because most drinking water passes through this aquifer.

Hydrophobic organic compounds are exchanged between dissolved and sorbed phases in the subsurface. Water flows through aquifer sediments and dissolved HOC interacts with sediments and can be sorbed on the sediments (sorbent). The distribution coefficient (Kd) describes the distribution of HOC between water phase and surface of sediments.

Kd = C sediment / C water (mol*g-1/mol*ml-1)                                                             (eq. 1)

Where Csediment is the concentration of HOC sorbed on the sediments and Cwater is the concentration of HOC in the water phase.

The Kd is dominantly a function of the abundance organic matter of the solid (foc), (Schwazenbach et al, 1993).

Kd = Coc * foc / Cw                                                                                                          (eq. 2)

Where Coc is the concentration of sorbate associated with organic carbon and Cw is the concentration of nonpolar organic compounds in solution (Schwazenbach, et al., 1993). The ratio of Coc to Cw is the carbon normalized partitioning coefficient Koc. The Koc describes the distribution or partitioning of HOC between two immisible solutions (organic matter and water) (Schwazenbach et al,1993).

Koc = Coc / Cow                                                                                                                 (eq. 3)

Different studies in the literature provide the range of the predicted relationships between Koc and the compounds octanol-water coefficient (Kow is a measure of HOC hydrophobicity). Three equations, that represent the range of Koc predicted, were used to estimate the expected Koc for HCB.

log Koc = 0.82*log Kow + 0.44  (Schwazenbach et.al., 1993)                                                    (eq. 4)

log Koc = log Kow 0.21  (Karicknoff et al., 1979)                                                              (eq. 5)

log Koc = 0.72*log Kow + 0.49 (Schwarzenbach and Westall, 1981)                                                   (eq. 6)

Kd and foc were measured during experiment and experimental Koc then was calculated.

Kd = Koc*foc Koc = Kd / foc                                                                              (eq. 7) 

Comparison of experimental and predicted Koc can give information about HOC sorption, and which sediments characteristics are mainly responsible for the sorption.

On the sediments with significant amount of organic matter, organic matter is mainly responsible for sorption of nonpolar organic compounds. In an environment with a very low fraction of organic matter (foc 0.0002-0.001) association of HOC with mineral surface may be significant  (Schwazenbach et al, 1993). Long Island sediments contain a low fraction of organic carbon. Possibly, both mechanisms: organic material and mineral surface sorption could be significant.

The rate of sorption of organic compounds on non-porous mineral surfaces is faster than rate of sorption on porous surfaces (Huang et al., 1996). Holmen and Gschwend (1997) suggest that the rate of the sorption is controlled by diffusion of HOC through the coatings. In this study we measured sorption after 48 hours (short- term sorption). It has been shown that sorption equilibrium for HOC on sandy aquifer sorbents may take days to years (Ball and Roberts, 1991). The treatment of sediments (sonication e.t.c.) may have a significant effect onto both the availability of sorption sites and the kinetic approach to equilibrium.

In this study we want to determine the following: are coatings important in the sorption of HOC and what is the relative importance organic carbon content and surface area in sorption of HOC in these sediments?

Approach

All sands, chosen for this experiment have coatings. A study conducted in Cape Cod, Massachusetts with aquifer sands similar to Long Islands presented evidence that coatings are responsible for most HOC sorption, because the largest fraction of organic carbon concentrated in the coatings (Holmen and Gschwend, 1997).

For the HOC sorption study, four samples of sands from Long Island were chosen. All samples contained coatings on the grain surfaces. Samples 1 and 5 (from 100-120 and 160-180cm below the surface) are medium-coarse sand from Port Jefferson. Samples 6 and 7 (from 140-150 and 200-220 cm below the surface) are medium-coarse and coarse sand from South Setauket (See methods). As a nonpolar hydrophobic organic compound, Hexachlorobenzene (HCB) was chosen, because of its high binding capacity (log Kow = 5.50). As a result sorption can be followed with a small amount of coatings and HCB is undegradable by microorganisms.

The distribution coefficient (Kd) was measured on: the whole coated sand, coatings separated by sonication, skeletal grains without coatings (naked sand), whole sonicated sand including coatings and HCl-treated sand including coatings but without iron compounds.

 

Results

The Kd values for HCB sorption described with different samples and treatments are summarized in the Table 1 and Figure 1. The Kd for whole sand varies between 2.5 ml/g (sample 7) to 10.7 ml/g (sample 1) (Table 1). The Kd for coatings varies between 396 ml/g (coatings from sample 7) and 1480 ml/g (coatings from sample 1). The Kd for the naked sand is always lower than that observed for the whole sand and the coatings varies between 1.55 ml/g (naked sand from sample 6) and 9.02 ml/g (naked sand from sample 5) (Table 1).


Figure 1. The comparison among the Kd whole sand, the Kd coatings, the Kd naked sand, the Kd sonicated sand but without separation of coatings, the Kd HCl treated sand and the sum of Kd coatings * fraction of coatings and Kd of naked sand * f naked sand.  


A. Sample 1. Coarse sand from Port Jefferson at depth 100-120 cm.


B. Sample 5. Medium-coarse sand from Port- Jefferson at depth 160-180 cm


C. Sample 6. Coarse sand from Port- Jefferson at depth 120-140 cm.


D. Sample 7. Coarse sand from Port-Jefferson at depth 200-220 cm.


The Kds for the sonicated sand (but without separation of the coatings) are always higher than that for unsonicated whole sand were in the range 7.53 ml/g (sample 6) and 35.7 ml/g (sample 5). HCl treatment was done only for samples 1 and 5, and Kd is 34.9 ml/g and 42.8 ml/g (Table 1).

The fraction of the coatings separated by sonication varies from 0.008 to 0.065

(Table 2). The Kd of the whole sand was compared to that of the coatings and naked sand by using mass balance (Table 3).

Kd total = Kd coatings * f coatings + Kd naked sand * f naked sand                                                                  (eq. 8)

The Kd of the whole sand for all samples is always lower than Kd for the whole, sonicated sand (Figure 1), showing that the process of sonication changes Kd significantly. The Kd following sonication of sample 1 was increased by 41.8%, sample 5 Kd was increased by 20.1%, sample 6 Kd was increased by 58.4%, and sample 7 Kd was increased by 73.1% (Figure 1). Probably, sonication exposes or opens up pore space and HOC diffuses faster through the pore space and also the area available for sorption increases. For example, surface of cemented grains may become exposed. Therefore, more organic molecules can be sorbed in a shorter amount of time.

The removal of iron cement by HCl dissolution also increases Kd, consistent with explanation above. Iron compounds cement fine particles together and their dissolution opens up pore spaces.

For samples, which contain a higher fraction of coatings (samples 1 and 5), the Kd for whole sonicated sand is always lower than the estimated Kd (sum of coatings and naked sand Kd). For samples with a low fraction of coatings (samples 6 and 7) calculated whole sand Kd is lower than the Kd for sonicated whole sand. There are several possible explanations for why this is so. First, should be noted that the procedure for sonication varied in the experiments where coatings were separated and in these, in which sorption was measured into the whole sonicated sand. Most important, the separated coatings and naked sand were re-dried before sorption experiment, which may have affected aggregates stability. Another possibility is that sonication of grains with low fraction of coatings changes surface area and pore space in the large grains more significantly than with sonication of the samples with a high coatings fraction, which have a significant surface area available for sorption even without sonication. The mass-balance (Figure 1, Table 3) suggests that in all four samples the major amount of HOC sorbed could be attributed to the coatings. In sample 1, 85% of HOC sorbed on the coatings, in sample 5, 84.5%, in sample 6, 72.8%, in sample 7, 64.7%. Data suggests that the coatings could be the primary sorbent for hydrophobic contaminants like HCB.

Coatings contain chiefly clay minerals (illite, kaolinite and chlorite) with minor amounts of iron compounds and organic matter. For hydrophobic organic compounds, natural organic matter is often the main sorbent phase. Measurements of total organic carbon (TOC) on the whole sand and coatings show that the weight fraction of TOC varies on the whole sand from 0.0002 (sample 6) to 0.00027 (sample 1) (Table 4). TOC on the coatings displayed a wider range; from 0.0021 (sample5) to 0.0076 (sample 1) (Table 4). TOC was estimated on the coatings by multiplying TOC on the coatings by the fraction of the coatings (Table 4). The contribution of the coatings to the foc of the whole sand was calculated as

foc(coatings estimated) = foc(coatings) * f coatings                                                                   (eq. 9)

For samples 1 and 5, the estimated TOC of the coatings contributes a large fraction of the TOC on the whole sand but for samples 6 and 7, the estimated TOC contribution from coatings is significantly lower than the TOC of the whole sand (Table 4). We assumed that some amount of TOC is associated with naked sand grains. TOC of the naked grains was calculated as the difference between the whole sand TOC and the estimated coatings TOC. Naked sand TOC for the samples with a low fraction of coatings (samples 6 and 7) is higher than for the samples with a high fraction of coatings (samples 1 and 5) (Table 4). Probably, coatings in samples 6 and 7 are harder to remove and coating residue is responsible for the high TOC on the naked sand skeletal grains. The comparison between the coatings Kd and fraction of TOC indicates that samples with higher fraction of TOC have higher Kd. The exception is sample 5, which has the lowest fraction of TOC and the highest Kd (Figure 2). We can hypothesize that the natural organic matter of the sediments was predominantly responsible for HCB sorption. Possibly in sample 5, along with fraction of TOC, other factors, for example surface area, may control sorption of HOC.

Figure 2. Correlation between the fraction of total organic carbon (foc) and the Kd of the coatings. Coarse sand form Port Jefferson (samples 1 and 5) and coarse-medium sand from South Setauket (samples 6 and 7).

Surface area controls the sorption of HOC on the sediments with a very low fraction of organic carbon. Surface area measured on the whole sand has shown the following results: sample 5 has the highest surface area at 3.2 m2/g and sample 7 has the lowest, at 1.03 m2/g (Table 5). Whole sand with higher surface area has a higher Kd value.

Figure 3. Correlation between surface area (S.A.) and the Kd of the coated sand. Coarse sand form Port Jefferson (samples 1 and 5) and coarse-medium sand from South Setauket (samples 6 and 7).

Discussion

The coatings appear to be the major sorbent for HOC. Using a mass-balance 85%, 84.5%, 72.8% and 64.7% of the HCB sorption may be contributed by the coatings on the samples 1, 5, 6 and 7 respectively. According to Holmen and Gschwend (1997) the main sorption sites in aquifer sands for HOC occurs within coatings. They suggest this because the fraction of organic carbon on the coatings is higher than on the whole sand. Coatings are described as associations of natural organic matter with iron and aluminum-bearing minerals. Sand from the Pine Barrens (New Jersey), which Holmen and Gscwend used in their experiment, is chiefly quartz-covered with coatings composed of goethite, kaolinite and organic material. Coatings on Long Island sediments have similar composition: clay minerals (Illite, kaolinite and chlorite), organic mater and a minor amount of iron compounds.

The organic carbon normalized partition coefficient (Koc) indicates an organic sorbate distributing itself between two immiscible solutions (organic matter and aqueous solution) (Scwazenbech, et al, 1993). The experimental partitioning coefficient Koc was calculated for HCB (Equation 7). The Kd and the foc were measured for each sample. The measured log Koc for the whole sand lies between 5.10 (sample 5) and 4.02 (sample 7). The log Koc for coatings lie between 5.52 (sample 5) and 5.06 (sample 7). The log Koc for naked sand lie between 5.05 (sample 1) and 3.98 (samples 6 and 7) (Table 6).

The calculation of predicted Koc values are presented in Table 6 and based on the following equations 4, 5 and 6. The Koc is the water-organic mater partitioning coefficient and Kow is octanol-water partitioning coefficient. Log Kow was determined for HCB and equaled 5.50 (Chiou et al., 1982). Predicted log Koc varied between 5.29 and 4.45

(Table 6). Experimental log Koc for the whole sand is closer to the low predicted log Koc (the exception is sample 5, showing a high log Koc= 5.10). Experimental log Koc for the coatings is closer to the high predicted log Koc (the exception is sample 5, showing a high log Koc = 5.52 is higher than high predicted value). The experimental log Koc for the naked sand is closer to the low predicted log Koc (exception is sample 1 log Koc = 5.05) (Table 6). High values log Koc for sample 5 may suggest that not only natural organic matter was responsible for sorption of HOC but also surface area (sample 5 has the lowest fraction of TOC and the highest whole sand surface area). The log Koc for the whole sand is generally lower than the predicted range of Koc (especially for samples 6 and 7) (Table 6). This may indicate that not all organic matter was available for sorption. Holmen and Gschwend (1997) observed that the column Kd values were always much less than measured from batch testing and predicted values. They suggested that association of negatively charged organic matter molecules with positively charged iron oxide minerals formed aggregates, in which such organic matter may not be really available for sorption. During batch test disaggregation of these aggregates may expose organic matter and make them available for sorption. Probably, increasing Kd value after sonication of the whole sand may be explained by the exposure of organic matter, which was unavailable for sorption before sonication. In the experiments with HCB, sorption kinetic limitation for pore diffusion may occur. The experiment was equilibrated for 48 hours (short-term sorption) which may not be sufficient to attain equilibria with less accessible organic matter. We could not do long term equlibria experiment because some leakage from bottles occurred. But it is also possible that some amount of organic matter would be totally unavailable for sorption regardless how long we waited for complete sorption. We concluded that natural organic matter of the soil was predominantly responsible for HCB sorption. This statement is questionable for sample 5, in which surface area may also be a significant factor in the sorption of HOC.

It has been argued that for low organic carbon sorbents like aquifer sands, the mineral surface adsorption of hydrophobic organic compounds becomes increasingly significant (Schwazenbach et al,1993). A surface area normalized partitioning coefficient (Ksa, L/m2) was calculated from our results and compared to Ksa values measured for tetrachlorobenzene with various mineral surfaces (Table 7).

Kd = Ksa * S.A.                                                                                   (eq. 10)

Ksa = Kd/S.A.                                                                                      (eq. 11)

where Kd is the experimentally determined distribution coefficient (ml/g) and S.A. is surface area (m2/g). The predicted Ksa was calculated by using surface area (S.A) and mineral Kd data from Table 11.3 in Schwazenbach, Gschwend and Imboden, (1993).

Porous silica, g -alumina and kaolinite were chosen as sorbent. Given Kd for tetrachlorobenzene were corrected by a factor of 10 because this study was done with HCB and HCB is about 10 times more hydrophobic than tetrachlorobenzene (Kow=4.50). Schwazenbach et al, (1993) have shown that mineral surface sorption of different sorbates is almost directly proportional to Kow of sorbates. We understand that this may represent a rough approximation of Ksa values, but should be useful for comparising the magnitude of Ksa measured and those calculated for organic free minerals. The Ksa estimated for pure minerals varies between 0.18 ml/m2 (g -alumnia) and 4.08 ml/m2 (kaolinite). Experimental Ksa values lies between 2.13ml/m2 (sample 6) and 9.95 ml/m2 (sample 5). Generally, the Ksa predicted is lower than the Ksa experimental. This result is consistent with hypothesis that the sorption of HOC was controlled mainly by the natural organic matter and less so by the surface area.

Conclusions

1. The Kd values for HCB sorption on the coatings are much higher than those of the whole sand. The results suggest that sorption by the coatings likely contributes much of the short-term (24-48 hours) sorption of the whole sand.

2. Sorption of HCB by the coatings is consistent with the amount of organic matter on the coatings. The partitioning of HCB into the organic matter is likely to control sorption. However, some contribution of adsorption of HCB onto bare mineral surface is also possible.

3. Some amount of natural organic matter remains unavailable for sorption of HOC into sand particles during short-term (48 hours) sorption experiment. Possibly, during a longer sorption experiment more organic carbon may become available for sorption. The Koc predicted is higher than those experimentally determined on the whole and the naked sand samples.

4. Different treatments (sonication e.t.a.) affect the sorption capacity of the sand. First, sonication opens up pore space and diffusion of HOC to the sorption sites may occur faster. Second, a larger surface area exposed after sonication and become available for sorption. Third, more of the organic carbon may become available for sorption after sonication.

Return to Table of Contents