Effects of HA on contaminant transport
The breakthrough curves of pyrene and tetracycline in the presence and absence of HA are compared in Figure 1. Note that in the presence of HA in the mobile phase, the solution-phase concentration of tetracycline or pyrene, C, is a bulk concentration for both freely dissolved contaminant and contaminant sorbed to dissolved HA. For both pyrene and tetracycline, the breakthrough data can be well described with the two-site nonequilibrium transport model, and the fitted parameters are summarized in Table 1. The transport data for neither contaminant can be described with the equilibrium transport model (Eqn. 1; see Supplemental Data, Fig. S5).
Figure 1. Breakthrough curves of pyrene (a) and tetracycline (b) in the presence and absence of humic acid (HA). Column and influent properties are summarized in Table 1. The dashed lines were plotted by curve-fitting experimental data with the two-site nonequilibrium transport model (Eqns. 5 and 6). PV = pore volumes.
Download figure to PowerPoint
In the absence of HA, breakthrough of pyrene occurred at 13 pore volumes. The percentage of breakthrough—C/C0 (where C0 is the total contaminant concentration in the influent; Table 1)—increased quickly and reached more than 80% after 40 pore volumes. In the presence of 20 mg C/L HA in the influent, the transport of pyrene was inhibited: the R value increased by 18% (Table 1). Nonetheless, in the presence of 50 and 80 mg C/L HA, the transport of pyrene was facilitated, and the R value decreased by more than 33 and 41%, respectively.
The breakthrough curves for tetracycline exhibit shapes markedly different from those for pyrene. In the absence of HA, tetracycline breakthrough occurred after 10 pore volumes. However, as the flow continued, the C/C0 ratio of tetracycline increased much more slowly compared with that of pyrene, and after 40 pore volumes the C/C0 ratio was less than 40%. This indicates that, at a given pore volume, a larger fraction of tetracycline than pyrene was retained by the porous medium. More important, comparing the breakthrough profiles between tetracycline and pyrene indicates that the transport properties of tetracycline deviate much more significantly from the equilibrium-based transport model (Eqn. 1) 31, 32. This can be understood further with the fitted parameters in Table 1, in that for tetracycline the fraction of the nonequilibrium-sorption site (that is, 1−f) is much larger, and the kinetic constant associated with the nonequilibrium site (α) is smaller. (The slower kinetics of tetracycline can possibly be linked to its greater polarity 33.) Figure 1 also shows that HA has a distinctively different effect on the transport of tetracycline. Contrary to the case for pyrene, the effect of HA on the transport of tetracycline was consistent: with the increase of HA concentration, the breakthrough curve shifts consistently to the left. Thus, for tetracycline, transport is always enhanced by the presence of HA.
Effects of HA on contaminant sorption
A comparison of the fitted R values in Table 1 indicates that low HA concentrations in the influent enhance sorption of pyrene to sand, which leads to inhibited transport, but high HA concentrations inhibit sorption of pyrene, which leads to enhanced transport; nonetheless, even low HA concentrations in the influent inhibit adsorption of tetracycline on sand, so tetracycline transport is always enhanced in the presence of HA. These assumptions were verified with the batch sorption data of pyrene and tetracycline to sand in the presence and absence of HA (Fig. 2). Note that, in this set of experiments, pyrene and tetracycline molecules are distributed among four phases (refer to the schematic illustration in Fig. 3), including molecules freely dissolved in solution (phase 1), molecules sorbed to dissolved HA (phase 2), molecules sorbed to sand (phase 3), and molecules sorbed to HA that is adsorbed on sand (phase 4). In Figure 2, the aqueous-phase concentration (C) is the bulk contaminant concentration of phases 1 and 2, and the sorbed-phase concentration (q) is the bulk concentration of phases 3 and 4.
Figure 2. Sorption isotherms of pyrene (a) and tetracycline (b) to sand in the presence and absence of humic acid (HA). The HA concentrations (20, 50, and 80 mg C/L) are total HA concentration in the system. Solid lines were plotted by fitting the data with linear sorption isotherm.
Download figure to PowerPoint
Figure 3. Schematic illustration of contaminant mass distribution among four phases (each indicated with Roman numbers I, II, III, or IV) in a closed system containing sand, aqueous solution, contaminant, and humic acid.
Download figure to PowerPoint
Figure 2 clearly shows that, in the presence of 20 mg C/L HA, sorption of pyrene to sand was considerably enhanced compared with the sorption in the absence of HA. In the presence of 50 or 80 mg C/L HA, sorption of pyrene was significantly inhibited. Nonetheless, HA at all three concentrations inhibited the adsorption of tetracycline on sand. In addition, the adsorption isotherms of tetracycline in the presence of 50 and 80 mg C/L HA nearly overlap, indicating that, once the HA concentration is above a certain level, its adsorption-inhibition effect stops increasing further. In Table 2, the Kd values obtained in the batch sorption experiments are compared with the Kd values calculated from the R values (Table 1) that were obtained by fitting the transport data with the two-site model (Eqns. 5–8). A sensitive analysis of the model parameter R was performed, and the results are shown in Supplemental Data, Fig. S6. In general, the two groups of Kd values agree well for both pyrene and tetracycline.
Table 2. Comparison of experimentally obtained and estimated Kd values of pyrene and tetracycline to sand in the presence and absence of humic acid (HA)
|CHA_total (mg C/L)||Kd (L/kg)a|
|0||5.44 (0.05)||4.81 (0.08)||4.81||14.5 (0.1)||14.9 (0.4)||14.9|
|20||6.43 (0.06)||5.93 (0.08)||5.58||11.8 (0.1)||12.5 (0.4)||14.8|
|50||3.54 (0.08)||3.26 (0.04)||3.31||7.35 (0.05)||6.85 (0.14)||14.5|
|80||3.13 (0.07)||2.73 (0.03)||2.21||7.29 (0.05)||6.72 (0.18)||14.1|
The most striking observation in the present study is that for pyrene a critical HA concentration appears to exist, below which HA enhances sorption of pyrene to sand but above which HA inhibits sorption; for tetracycline, however, HA always inhibits adsorption. This interesting difference in the effect of HA between pyrene and tetracycline can be understood by examining contaminant mass distribution among the four phases mentioned above. The following equations can be derived based on mass balance (see Supplemental Data for detailed derivation)
where f1 through f4 are mass fraction of a contaminant in each of the four phases; ms/Vw (kg/L) is the sand-to-water ratio; Kc-HA (L/kg C) and Kc-sand (L/kg) are the sorption coefficients of a contaminant to HA and to sand, respectively; CHA (kg C/L) is the concentration of HA in the solution; and qHA (kg C/kg) is the concentration of HA adsorbed on sand. In addition, an apparent Kd value, Kd_apparent, can be calculated as (see Supplemental Data)
where Kd_apparent has the same physical meaning as the slopes of the sorption isotherms in Figure 2 (i.e., it is the ratio of q, the bulk concentration of phases 3 and 4, to C, the bulk contaminant concentration of phases 1 and 2). The values of f1 through f4 and Kd_apparent can be calculated once the values of Kc-HA, Kc-sand, CHA, and qHA have been obtained (these values were obtained experimentally in the present study). Note that two underlying assumptions are that the organic C originally in the sand and the HA that adsorbs to the sand function independently and that sorption affinity of pyrene and tetracycline to the adsorbed HA is the same as that to the aqueous-phase HA (that is, Kc-HA).
Figure 4a shows the sorption isotherm of pyrene to HA. The sorption data follow reasonably the linear sorption isotherm q = Kc-HAC, and a Kc-HA value of 104.92 L/kg C can be obtained. This value is consistent with the literature-reported values 26 and with the sorption coefficients determined using the solid-phase dosing method (Supplemental Data, Fig. S7). A Kc-sand value of 4.81 can be calculated for pyrene using the sorption data in the absence of HA in Figure 2a and can be converted to a KOC value of 104.57 L/kg. Even though sorption of pyrene to both HA and sand is controlled primarily by hydrophobic partitioning to natural organic matter, it is possible that sorption to HA is enhanced by π–π electron donor–acceptor interactions resulting from the aromatic nature of HA 34.
Figure 4b shows the sorption isotherm of tetracycline to HA, and an average Kc-HA value of 939 L/kg C can be obtained. This value is comparable with the literature values 24. A Kc-sand value of 14.9 L/kg was obtained for tetracycline with the sorption data in Figure 2b. These values indicate that tetracycline exhibits much weaker (approximately two orders of magnitude) sorption to HA than pyrene does, but tetracycline adsorbs more strongly on sand than pyrene does. As an amphoteric compound, tetracycline has three ionizable functional groups (Supplemental Data, Fig. S8), each with a specific pKa value 5. Within the test pH, tetracycline exists mainly as a zwitterionic species and can interact strongly with the deprotonated sites of HA (mainly carboxylic groups) 19, 24, 35. Furthermore, because of the high content of polar functional groups in HA (phenolic, carboxylic, and alcoholic groups; Supplemental Data, Table S3), strong H-bonding among the hydroxyl, ketone, and amino groups in tetracycline and the respective functional groups in HA likely is an important sorption mechanism 19, 24, 35. Thus, even though KOW (n-octanol–water partition coefficient) of tetracycline is more than six orders of magnitude lower than that of pyrene, it still exhibits moderate sorption affinity to HA. Similarly, because the interaction of tetracycline with the quartz-like sand is likely via surface complexation and ligand exchange 36, 37 rather than hydrophobic partitioning, tetracycline exhibits stronger adsorption to the extremely low-fOC sand than pyrene does.
Figure 5 shows the adsorption isotherm of HA on sand. The adsorption data can be well described with the Langmuir adsorption isotherm
where qmax (kg C/kg) is the maximum monolayer adsorption capacity of HA on sand and b (L/kg C) is the Langmuir adsorption affinity. This is consistent with the literature 38, in that soil minerals provide surfaces on which amphiphilic humic moieties form a membrane-like coating. The values of b and qmax, obtained by fitting the adsorption data, are 1.1 × 105 L/kg C and 1.3 × 10−4 kg C/kg, respectively. Note that other types of DOM might have different b and qmax values but likely would follow the Langmuir-type adsorption isotherm too. The values of CHA and qHA at a given total HA concentration in the system (CHA_total) can then be calculated based on mass balance as well as Equation 14 (see Supplemental Data).
Figure 5. Adsorption isotherm of humic acid (HA) on sand. Solid line was plotted by curve-fitting adsorption data with Langmuir model (Eqn. 14).
Download figure to PowerPoint
In Figure 6, the changes of f1 to f4 with CHA_total are shown for pyrene and tetracycline. An interesting observation is that, for pyrene, the value of f4 first increases with CHA_total and then decreases, but no such trend is observed for tetracycline. The Kd_apparent of pyrene shows a very similar pattern (Fig. 6a); it increases with CHA_total and peaks when CHA_total reaches approximately 10 mg C/L, then starts to decrease with the further increase of CHA_total. However, for tetracycline, Kd_apparent is highest when CHA_total is zero and continues to decrease with the increase of CHA_total (Fig. 6b). These Kd_apparent–CHA_total correlations provide a good explanation for the observed HA effects on transport (Fig. 1) and on sorption (Fig. 2) that differ between pyrene and tetracycline.
Figure 6. Mass fraction of pyrene (a) and tetracycline (b) among four phases in response to different total humic acid (HA) concentration in the system (CHA_total). f1 through f4 are defined in Equations 9 to 12, and Kd_apparent is defined in Equation 13.
Download figure to PowerPoint
The Kd_apparent peak or the critical HA concentration for pyrene exists because of a combined effect: the large Kc-HA to Kc-sand ratio and the Langmuir-type adsorption of HA on sand. When HA is present at low concentrations, corresponding to the linear part of the Langmuir isotherm of HA on sand, a large fraction of HA and hence a significant fraction of pyrene (because pyrene sorbs predominantly to HA) are associated with sand, resulting in a higher Kd_apparent compared with that in the absence of HA. Once CHA_total reaches the critical concentration, the adsorbed concentration of HA on sand (qHA) does not increase appreciably with the further increase of CHA_total; that is, qHA approaches the plateau part of the Langmuir isotherm. Accordingly, the fraction of dissolved HA increases, along with sorbed pyrene, and Kd_apparent starts to decrease.
In Table 2, the Kd_apparent values of pyrene and tetracycline estimated using Equation 13 are compared with the values observed in the batch sorption and transport experiments. For pyrene, the estimated values agree reasonably with the experimentally observed values. Nonetheless, the estimated Kd_apparent values for tetracycline deviate increasingly from the observed values with the increase of CHA_total: according to Figure 6, Kd_apparent should change little within the test concentration range of HA, because the mass fractions of tetracycline associated with HA (f2 and f4) should be insignificant compared with the fraction in the solution (f1) and the fraction adsorbed on sand (f3). This is apparently contradictory to the strong effects of HA on the transport and adsorption of tetracycline (Figs. 1 and 2). A possible explanation is the competitive adsorption between HA and tetracycline for the available adsorption sites (most likely silicon or aluminum hydroxyls) on sand 39, 40. As the concentration of HA increases, an increasing fraction of available adsorption site on sand surface is covered by HA, resulting in inhibited adsorption of tetracycline on sand. The fact that the extent of adsorption-inhibition is similar in the presence of 50 and 80 mg C/L HA (Fig. 2) is consistent with the competitive adsorption theory: because adsorption of HA on sand follows the Langmuir isotherm, once qHA approaches the maximum monolayer adsorption capacity, adsorption of HA, and subsequently its competition effect, stops increasing further. Because competitive adsorption is not accounted for in Equations 9 through 13, these equations cannot accurately quantify the effect of HA on tetracycline adsorption on sand, whereas, for pyrene, competitive sorption of HA is likely negligible, because HA and pyrene sorb to different sites on sand, and Eqns. 9–13 work well.
The findings in the present study indicate that the same type of DOM might have effects on the subsurface transport of ionic, polar, and hydrophilic organic contaminants, such as tetracycline antibiotics, considerably different from those on the transport of nonionic, apolar, and highly hydrophobic organic contaminants, such as high-molecular-weight polycyclic aromatic hydrocarbons. The specific effect of DOM on transport depends on the nature of contaminant–DOM, contaminant–porous medium and DOM–porous medium interactions. In general, DOM mediates the transport of nonionic, apolar, highly hydrophobic contaminants primarily by affecting the partition of contaminants between the mobile phase and the porous medium, whereas the effect of DOM on the transport of ionic, polar, and highly hydrophilic contaminants might also involve the competition of DOM for the available adsorption sites on the porous medium, at least when DOM is present at relatively high concentrations. Furthermore, the DOM-mediated transport of emerging contaminants such as antibiotics cannot be accurately quantified with the conventional models and must be better understood.