JEQ Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Related articles in JEQ
Right arrow Similar articles in this journal
Right arrow Similar articles in Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Web of Science (14)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Williams, L. E.
Right arrow Articles by Melville, J. G.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Williams, L. E.
Right arrow Articles by Melville, J. G.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Williams, L. E.
Right arrow Articles by Melville, J. G.
Related Collections
Right arrow Heavy Metals
Journal of Environmental Quality 32:841-850 (2003)
© 2003 American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America

TECHNICAL REPORTS
Heavy Metals in the Environment

Adsorption and Transport of Arsenic(V) in Experimental Subsurface Systems

L. Elizabeth Williamsa,b, Mark O. Barnett*,a, Timothy A. Kramerc and Joel G. Melvillea

a Dep. of Civil Engineering, 208 Harbert Engineering Center, Auburn Univ., Auburn, AL 36849
b PBS&J, Water Resources, 5665 New Northside Drive, Suite 400, Atlanta, GA 30328
c Dep. of Civil Engineering, Room 205K, WERC, MS 3136, Texas A&M Univ., College Station, TX 77843-3136

* Corresponding author (barnettm{at}eng.auburn.edu)

Received for publication September 10, 2001.

   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 METHODS AND MATERIALS
 MODELING
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
The adsorption and transport of As(V) in a heterogeneous, iron oxide–containing soil was investigated in batch and column laboratory experiments. The As(V) adsorbed rapidly to the soil over the first 48 h, but continued to adsorb slowly over the next several weeks, clearly indicating the potential for rate-limited transport. The equilibrium As(V) adsorption isotherm was markedly nonlinear, further indicating the potential for nonideal transport. A model developed for the adsorption of As(V) to hydrous ferric oxide (HFO) was able to predict the pH-dependent adsorption of As(V) to the soil in batch experiments within 0.116 to 0.726 root mean square error (RMSE). Arsenic(V) was significantly retarded in column transport experiments. The column transport experiments were modeled using the one-dimensional advection–dispersion equation, considering both linear and nonlinear adsorption equilibrium. Although the nonlinear local equilibrium model (NLLE, RMSE = 0.273) predicted the data better than the linear local equilibrium model (LLE, RMSE = 0.317), As(V) breakthrough occurred more rapidly than predicted by either model due to adsorption nonequilibrium. However, due to the presence of an irreversible or slowly desorbing fraction, the peak aqueous As(V) concentration (0.624 mg L-1) and the total amount of As(V) recovered (44%) was lower than predicted based on the two equilibrium models (NLLE and LLE). For the conditions used in this study [1 mg L-1 As(V), pH 4.5 and 9, 0–0.25 mM PO4, 0.53–1.6 cm min-1 pore water velocity], the effect on As(V) mobility and recovery increased in the order pH < pore water velocity < PO4.

Abbreviations: AOD, acid ammonium oxalate reacted in the dark • CBD, citrate–bicarbonate–dithionite • HFO, hydrous ferric oxide • LLE, linear local equilibrium • NLLE, nonlinear local equilibrium


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
METHODS AND MATERIALS
 MODELING
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
ARSENIC FROM BOTH anthropogenic and geologic sources is commonly found in the subsurface. Arsenic is a carcinogen and contributes to other health effects such as diabetes and cardiovascular disease. The use of organic and inorganic As as a pesticide in the USA began in the 1850s and peaked in the 1950s, although its use as a wood preservative has recently been increasing (Welch et al., 2000). Historically, the most significant source of As release into the environment is due to the process of mining and smelting metals (Smith et al., 1998). In addition, the presence of naturally occurring As in ground water is a tremendous public health threat to millions of people of West Bengal, India, and Bangladesh (Chowdhury et al., 1999). Although problems in the USA are not as severe, moderate to high (>50 µg L-1) naturally occurring concentrations of As are commonly found in ground water throughout the western USA (Welch et al., 2000). The cost of removing As from ground water in the USA may increase to more than $4.1 billion per year in the near future (Frey et al., 1998). The mobility of As in soil and ground water is important when evaluating possible environmental and economic effects.

Although As can be found in most natural waters in both organic and inorganic forms with varying oxidation states (+5, +3, 0, -3), As is found predominantly in the inorganic form in oxidation states of +5 or +3 (Sadiq, 1997). In oxidizing conditions, the oxyanion arsenate [As(V)] is the predominate species. A triprotic acid (pKa 2.20, 6.97, 11.5), the thermodynamically stable species of As(V) in most normal soil pH ranges (4 to 8) are H2AsO-4 and HAsO2-4 . The reactive interaction of As(V) with the subsurface media significantly affects the movement of As in soil and ground water. The equilibrium adsorption of As(V) to pure solid phases and soils has been studied extensively as documented in recent reviews (Sadiq, 1997; Smith et al., 1998). These studies have shown that Fe oxyhydroxides strongly interact with dissolved As(V) and that the degree of As(V) adsorption is extremely pH-dependent. Spectroscopic studies have generated molecular knowledge regarding the adsorption of As(V) to minerals and soils for incorporation in mechanistic models (Rietra et al., 1999; Goldberg and Johnston, 2001). However, the ability to predict As(V) adsorption is still quite limited.

In addition to pH, the presence of other ions also affects the adsorption of As(V). One of the most significant of these ions is PO4 (Jain and Loeppert, 2000). Phosphate exhibits similar chemical behavior and is often used in fertilizers in agricultural areas where As may have been used as a pesticide or herbicide (Smith et al., 1998). Surface complexation models have been invoked to model the competitive adsorption of As(V) and PO4 on minerals and soils in batch systems (Manning and Goldberg, 1996a, b). Less attention has focused on the role of PO4 in promoting As(V) transport through porous media. However, Peryea and Kammereck (1997) concluded that PO4 greatly enhanced the downward mobility of As(V) in soil columns. Melamed et al. (1995) concluded that PO4–amended soils exhibited an increase in mobility of As relative to non-PO4–amended soils and noted the potentially important role of physical nonequilibrium on As(V) transport.

Similarly, relatively little attention has focused on adsorption–desorption rates and their effect on As(V) transport, although the adsorption of As(V) to some heterogeneous materials can occur over time spans of weeks or longer (Lombi et al., 1999). Darland and Inskeep (1997a)( b) demonstrated the significant effects of pore water velocity, pH, and PO4 on the transport of As(V) through sand columns. However, these phenomena have not been well studied in flowing systems with heterogeneous (i.e., multicomponent) porous media.

The objective of this paper is to describe the results of a study of the adsorption and transport of As(V) in a heterogeneous subsurface media. Specific objectives of the investigation included (i) determining the primary physical and chemical parameters that influence As(V) adsorption and transport, with a particular emphasis on those processes that produce nonideal (i.e., nonlinear, rate-dependent) transport, and (ii) determining the ability of commonly used models to predict As(V) adsorption and transport in heterogeneous media.


   METHODS AND MATERIALS
TOP
 ABSTRACT
 INTRODUCTION
 METHODS AND MATERIALS
MODELING
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
The soil used in this study (Table 1) was acquired from the Melton Branch watershed on the U.S. Department of Energy Oak Ridge Reservation in eastern Tennessee. It was obtained at a depth of 1.5 m from the C horizon of a weakly developed Inceptisol (Montevallo series: loamy skeletal, mixed, thermic, shallow Typic Dystrochrept) that has weathered from interbedded shale–limestone sequences. The limestone has weathered to massive clay lenses devoid of carbonates, and the more resistant shale has weathered to a highly fractured saprolite. This soil was chosen because it is heavily coated with iron oxides (Arnseth and Turner, 1988), which are known to contribute to the adsorption and transport of subsurface As(V). The soil was air-dried and gently ground with a mortar and pestle to pass a 2-mm sieve. Particle size was measured with sieves and a hydrometer (Gee and Bauder, 1986). The soil pH was determined using 5 mM CaCl2 in a 2:1 solution to solid ratio. The pH of the supernatant was measured using an ion analyzer (EA940; Thermo Orion, Beverly, MA) and combination electrode (8102 ROSS; Thermo Orion). The background concentration of As in the soil measured by Method 3050B (USEPA, 1998) was 2.0 mg kg-1. Citrate–bicarbonate–dithionite (CBD)–extractable iron and manganese oxides and acid ammonium oxalate (AOD)–extractable iron oxides were also measured (Jackson et al., 1986). Total organic carbon (TOC) was measured by combustion on a total carbon analyzer (CN-2000; LECO Corporation, St. Joseph, MI). The mineralogy of the <2-µm clay fraction was identified with X-ray diffraction (DMS 2000; Scintag, Cupertino, CA).


View this table:
[in this window]
[in a new window]
  		Table 1. Soil properties.

 
Batch adsorption rate and isotherm experiments at approximately 25°C were conducted in low density polyethylene bottles containing 20 mL of 0.01 M NaNO3 solution to maintain a constant ionic strength and 0.8 g (rate experiments) or 0.1 to 2.2 g (isotherm experiments) of soil. A 1000 mg L-1 As(V) stock solution was prepared by adding reagent grade As2O5 to a 4 g L-1 NaOH solution according to Clesceri et al. (1989). A small volume of the stock solution was added at the beginning of the experiments to achieve the desired initial concentration (10 mg L-1) with a small volume of 0.1 M HNO3 to neutralize the base in the spike and maintain the soil's natural pH of approximately 4.5 (Table 1). The bottles were mixed using slow end-over-end rotation (approximately 4 rpm) and samples were taken from the vessels at desired time intervals and processed as described below.

To measure the influence of pH and the presence of other ions on As(V) adsorption, adsorption envelope [i.e., percent adsorbed as a function of pH at constant As(V) concentration] experiments were also performed. These experiments were also conducted in low density polyethylene bottles at room temperature (approximately 25°C) with 20 mL of 0.01 M NaNO3 and 0.1 g of soil. A small volume was added from the As(V) stock solution to achieve an As(V) concentration of 1 mg L-1 followed immediately by the addition of 0.1 M NaOH or 0.1 M HNO3 to adjust the pH. Additional NaNO3 (0.1 M final concentration), 10-3 M CO3, or 10-4 M PO4 was added as desired to alter the ionic strength or measure the effects of competing ions on As(V) adsorption. Phosphate was chosen as a competing ion because it shares a similar chemical behavior to that of As(V) and it is readily available in many As-contaminated agricultural areas due to land application of PO4–amended fertilizers. Carbonate was chosen because the CO3 system is one of the most important systems in the environment, its effects on other chemical systems can be profound, and recently its effects on the leaching of As into ground water have been noted (Kim et al., 2000). Samples were mixed using slow end-over-end rotation (approximately 4 rpm) for one week and then processed as described below.

Because the batch results indicated a significant potential for nonideal transport, additional column transport experiments were conducted to measure the effects of the most significant variables identified in the batch experiments (pH, time, and PO4 concentration). These experiments were conducted in a 1-cm-diameter glass column at room temperature and constant ionic strength (0.01 M). Two grams of soil were dry-packed to a depth of 1.7 cm, and the column was flushed from the bottom using a 0.01 M NaNO3 solution at a pH of 4.5 (the soil's natural pH) or 9.0 until air spaces were no longer visible. A square wave input pulse of 1 mg L-1 As(V) in a 0.01 M NaNO3 background matrix was introduced to the column at a pore water velocity of either 0.53 or 1.6 cm min-1 using high performance liquid chromatography pumps. After a breakthrough period of approximately 875 pore volumes, the inlet solution was switched back to a solution of identical composition but without As(V). Throughout the experiment, effluent samples were collected with a fraction collector and analyzed for As(V) and pH. The column properties were measured physically yielding a porosity of 0.43 and a bulk density of 1.5 g cm-3. The dispersion coefficient (D) was measured at a pore water velocity of 0.53 cm min-1 by fitting the one-dimensional advective–dispersive equation to the breakthrough curve of a nonreactive tracer (Cl-) by adjusting the parameter D using CXTFIT (Parker and van Genuchten, 1984). The nonreactive tracer exhibited a sharp breakthrough yielding a dispersion coefficient of 0.14 cm2 min-1 and a column Peclet number (vL/D, where v is the pore water velocity and L is the column length) of 6.4, indicating a low potential for nonideal transport due to physical conditions (e.g., preferential flow paths) inside the columns.

The samples were filtered using a 0.45-µm syringe filter (Acrodisc 4559; Pall Corporation, Ann Arbor, MI), acidified with 5% HNO3 and capped pending analysis. The use of a 0.45-µm filter may have allowed As(V) adsorbed to colloidal soil material to pass through the filter, potentially overestimating the true dissolved As(V) concentration. Before the samples were acidified, the pH was measured using an ion analyzer (EA940; Thermo Orion) and combination electrode (8102 ROSS; Thermo Orion). Samples were analyzed for As concentration using graphite furnace atomic absorption spectrometry (3110 and HGA-600; PerkinElmer, Wellesley, MA). An electrodeless discharge lamp (EDL System 2; PerkinElmer) was used for As analysis to improve sensitivity and linearity due to the greater spectral purity of their emission relative to a standard hollow cathode lamp. Standards were tested periodically (e.g., every 10 samples) throughout the analysis period to ensure the accuracy of the sample results. The detection limit was determined to be approximately 5 µg L-1 (three times the standard deviation of a blank). Samples were diluted gravimetrically, if necessary, to ensure As measurements of <100 µg L-1, the maximum linear range of the analysis. Samples were analyzed in triplicate and were considered accurate when the relative standard deviation of the replicates was <5%. The adsorbed As(V) concentration was calculated from the difference in the concentration in solution before and after the experiments. Blank samples were included without soil to verify the initial As(V) concentration and that the As(V) did not adhere to the reaction vessels. Blank samples with no added As(V) were also included to verify that natural background As was not desorbing from the soil.


   MODELING
TOP
 ABSTRACT
 INTRODUCTION
 METHODS AND MATERIALS
 MODELING
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Both equilibrium adsorption modeling and coupled adsorption–transport modeling were conducted. The emphasis on the modeling was to determine the agreement between commonly used predictive models and the experimental results rather than simple curve fitting (i.e., simple isotherms were fit to the data while the remainder of the modeling was predictive). Batch equilibrium isotherms were modeled using two conventional nonelectrostatic models. The experimental data was fit with both linear (Kd) and Freundlich isotherms to describe the data quantitatively by the equation:

[1]
where q and C are the equilibrium adsorbed (mg kg-1) and aqueous (mg L-1) As(V) concentrations, respectively, K is the adsorption capacity parameter (Kd for linear and Kf for Freundlich), and n is the adsorption intensity parameter. The linear isotherm is a special case of the Freundlich isotherm where n = 1. The model was fit to the data using linear regression with a linearized version of Eq. [1] with K (linear and Freundlich isotherms) and n (Freundlich isotherm only) as fitting parameters. Distribution coefficients (Kd) were also calculated for individual experimental points from the ratio of adsorbed to aqueous As(V) concentration.

As this soil is heavily coated with Fe oxides (Arnseth and Turner, 1988), the pH-dependent As(V) adsorption was modeled using the generalized two layer adsorption model developed by Dzombak and Morel (1990) for the adsorption of cations and anions onto hydrous ferric oxide (HFO). The model was used to predict the pH-dependent partitioning of As(V) on the soil by assuming that HFO governed the adsorption of As(V) to these materials. The concentration of HFO in the soils was estimated from both the AOD-extractable (amorphous and poorly crystalline) and CBD-extractable Fe (crystalline) contents (Jackson et al., 1986). The model was chosen because it is widely used by the USEPA (1999) to describe and quantify the adsorption behavior of contaminant metals and metalloids. This approach has recently been shown to work well for predicting As(V) adsorption to aquifer sediments (Welch and Lico, 1998) and a soil waste (Lumsdon et al., 2001) and for predicting U(VI) adsorption to this same soil (Barnett et al., 2002). The As(V) surface complexation reactions used in the model are:

[2]

[3]

[4]

[5]
where {equiv} Fe - represents surface-bound Fe. Because of the chemical similarity of PO4 to As(V), the surface reactions used to describe the interactions of PO4 with surface {equiv} Fe - groups are analogous to the equations used to describe the interactions of As(V) with {equiv} Fe - (Eq. [2]–[5]) with P replacing As. The equilibrium constants, however, are different, reflecting the relative strength of the PO4 and {equiv} Fe - surface bonds. The mass balance and mass action equations involved in the model were solved using MINTEQA2 (Version 4.0) and its standard thermodynamic database, which includes Eq. [2] through [5]. Parameters such as specific surface area and site concentration were taken from Dzombak and Morel (1990).

Column experiments were modeled using the one dimensional advection–dispersive reaction model:

[6]
where {rho}b is the bulk soil density (mass, M, length, L-3), t is the time (T), {theta} is the volumetric water content (L3 L-3), C is the solute concentration (M L-3), D is the hydrodynamic dispersion coefficient (L2 T-1), x is the distance along the column (L), {upsilon} is the average Darcy velocity (L T-1), and q and C are as described above.

A common assumption in modeling the transport of reactive contaminants is the local equilibrium assumption, which assumes that the adsorption process is rapid compared with transport via dispersion and advection (Brusseau, 1998). If adsorption equilibrium is obtained, the relationship between the instantaneous rate of change of the adsorbed and aqueous concentrations can be obtained by differentiating Eq. [1] with respect to time to obtain:

[7]

Equations [6] and [7] are coupled partial differential equations that represent the reactive transport model assuming local equilibrium. If the adsorption isotherm is linear (i.e., n = 1), Eq. [6] and [7] can be solved analytically and the resulting model is termed the linear local equilibrium model (LLE). If the adsorption isotherm is nonlinear (i.e., n != 1), Eq. [6] and [7] must be solved numerically and the resulting model is termed the nonlinear local equilibrium model (NLLE). Both the LLE and NLLE models were applied using the Multireaction Transport Model (MRTM) (Selim et al., 1990), which solves the equations numerically. To validate the correct implementation of MRTM, the numerical solution of the LLE model was compared with an analytical solution.

The fits of the models to the experimental data were quantified by calculating the root mean square error (RMSE) between the model predicted and experimental data points. The RMSE is calculated from:

[8]
where nd is the numbers of data points, np is the number of adjustable parameters (zero when used in a purely predictive manner), i is an index, C is measured aqueous concentration, C is the predicted aqueous concentration, and Co is the initial aqueous concentration (all in mg L-1). The RMSE is a measure of the error between the predicted and measured values expressed as a fraction of the initial concentration (e.g., the closer to zero the RMSE, the better the fit of the model to the data).


   RESULTS AND DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 METHODS AND MATERIALS
 MODELING
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
To examine As(V) adsorption dynamics and to determine the time frame when adsorption equilibrium was approached, batch adsorption rate experiments were conducted at the soil's natural pH of 4.5. Long-term (28-d) adsorption rate experiments exhibited a biphasic pattern (Fig. 1) , with a rapid period of adsorption occurring over the first 24 to 48 h, followed by a slower period of adsorption over the next several weeks. The As(V) concentration in solution (Fig. 1a) continued to decrease for up to three weeks. The distribution coefficient (Kd, the ratio of the adsorbed to aqueous concentrations) after three weeks was three to four times the value measured after one week, clearly indicating the potential for rate-limited, nonideal transport. This behavior contrasts with that exhibited by As(V) in adsorbing to ferrihydrite and other pure solid oxyhydroxides (Grossl and Sparks, 1995; Raven et al., 1998). However, other investigators have concluded that the rapid adsorption of As(V) onto Fe oxyhydroxides is followed by a slower process of surface diffusion to sites within colloidal aggregates (Fuller et al., 1993) or possibly further surface reactions (Grossl et al., 1997). However, despite the slow decrease in As(V) in solution, its effect on the adsorbed As(V) concentration was minimal after 48 h (Fig. 1b). Therefore, additional batch adsorption equilibrium experiments were conducted after at least 48 h of equilibration time.



View larger version (15K):
[in this window]
[in a new window]
  		Fig. 1. Adsorption of As(V) as a function of time (40 g soil L-1, pH 4.5, 0.01 M NaNO3). (a) Aqueous-phase and (b) solid-phase concentration.

 
After establishing a minimum equilibrium time, a batch adsorption isotherm was measured at the soil's natural pH. Arsenic(V) adsorbed strongly and nonlinearly to the soil at pH 4.5 (Fig. 2) . The nonlinear relationship indicates that there is a decrease in the As(V) adsorption capacity of this soil with an increase in the surface coverage. These data also indicate that As(V) subsurface transport would be significantly retarded due to interactions with the solid phase. Although the data were clearly not linear, a linear isotherm was applied to the model because, despite its limitations, linear adsorption is a commonly used approximation in reactive solute transport modeling (Brusseau, 1998; Bethke and Brady, 2000). The linear isotherm was best described with an equilibrium constant Kd = 345 ± 33 L kg-1. The Freundlich isotherm fit the data more accurately and resulted in the following relationship parameter estimates: Kf = 278 ± 1 L kg-1 and n = 0.32 ± 0.02. These parameters were used subsequently in predicting the outcome of As(V) transport experiments.



View larger version (21K):
[in this window]
[in a new window]
  		Fig. 2. Adsorbed solid versus aqueous As(V) concentration (pH 4.5, 0.01 M NaNO3).

 
In addition to time and As(V) concentration, several other parameters have been shown to influence As(V) adsorption, including pH, ionic strength, and the presence of competing anions. Adsorption pH envelopes [percent adsorbed versus pH at constant As(V) concentration] were measured as a function of initial As(V) concentration, ionic strength, and competing ions (Fig. 3) to identify the variables that were most likely to influence As(V) transport. The results graphically demonstrate the importance of pH on the adsorption process (Fig. 3a), with higher adsorption occurring at lower pH. From pH 3 to 7 the percentage of As(V) adsorbed decreased slightly from approximately 95 to 85%. As the pH increases from 7 to 10 the percentage of As(V) adsorbed dropped dramatically, decreasing to approximately 40 to 50% between pH 9 and 10. This behavior is typical of anion adsorption onto variably charged surfaces and results from the pH-dependent surface charge and aqueous speciation of As(V). At lower pH values (pH < 7) As(V) exists predominately as an anion in the form H2AsO-4 and is attracted to the positively charged soil surfaces (e.g., Fe oxides). At high pH values (pH > 7), As(V) exists as an anion in the form HAsO2-4 and the Fe oxide surfaces become increasingly negatively charged. The repelling negative charges between the soil particle and the As(V) ion help explain the decrease in As(V) adsorption with an increase in pH. These results indicate that pH would have a very strong effect on As(V) transport, with distribution coefficients decreasing by almost one order of magnitude in moving from pH approximately 7 to 9. Laboratory measurements of decreased adsorption and increased mobility of As with increasing pH are consistent with observations in As-contaminated ground water (Mariner et al., 1996).



View larger version (24K):
[in this window]
[in a new window]
  		Fig. 3. Adsorption at constant As(V) concentration [5 g soil L-1, 0.01 M NaNO3, and 1 mg As(V) L-1 unless otherwise noted]. (a) Effect of pH; (b) effect of ionic strength (I), As(V) concentration, and CO3; (c) effect of PO4.

 
Increasing the ionic strength from 0.01 to 0.1 M did not greatly affect As(V) adsorption (Fig. 3b). However, the increase did extend the 85 to 95% As(V) adsorption range from pH 7 (at 0.01 M ionic strength) to pH 8 (at 0.1 M ionic strength). This is a commonly recognized phenomenon (Manning and Goldberg, 1997; Smith et al., 1999) for the adsorption of oxyanions at high pH onto variably (and increasingly negatively) charged surfaces. The increase in ionic strength lessens the electrostatic repulsion of oxyanions for negatively charged surfaces by shielding the anion from the surface charge (Hiemstra and Van Riemsdijk, 1999). Decreasing the As(V) concentration by an order of magnitude also had a relatively minor effect on As(V) adsorption (Fig. 3b). The decrease in concentration resulted in an increase in the As(V) adsorbed at low pH values. Between pH 3 and 7, adsorption in the system with 0.1 mg L-1 As(V) was within the range of approximately 90 to 100%, whereas with a system concentration of 1.0 mg L-1 As(V) the range of adsorption was approximately 85 to 95% for the same pH values. Thus the effect was relatively minor because the degree of adsorption was already relatively strong (approximately 90%) in this region.

The effect of CO3 on As(V) adsorption was examined because carbonate is one of the most ubiquitous and important aqueous anions in the environment. The presence of CO3 decreased the extent of As(V) adsorption slightly, although the effect was minor in comparison with experimental variability (all replicates shown). Although CO3 has been shown to play an important role in mobilizing As from sulfide minerals (Kim et al., 2000), its effect on adsorption in this study was relatively small in keeping with other studies showing the effect of CO3 on As(V) adsorption to Fe minerals (Meng et al., 2000).

Phosphate has been previously shown to influence As(V) adsorption (Reynolds et al., 1999), and the addition of 0.1 mM PO4 greatly reduced the As(V) adsorption capacity of the soil (Fig. 3c). At low pH values (pH < 7), without the addition of PO4, As(V) adsorption was >85% with distribution coefficients > 1000 L kg-1. At low pH values (pH < 7) with the addition of 0.1 mM PO4, As(V) adsorption decreased to approximately 25 to 50% with distribution coefficients ranging from approximately 70 to 200 L kg-1. At higher pH values (pH > 7) As(V) adsorption decreased from >40% with distribution coefficients of approximately 150 to 1000 L kg-1 to <20% with distribution coefficients of <50 L kg-1 on the addition of 0.1 mM PO4. Phosphate was able to effectively compete with As(V) for adsorption sites and significantly decrease adsorption.

As this soil is heavily coated with Fe oxides, the pH-dependent As(V) adsorption was modeled using the generalized two-layer adsorption model developed by Dzombak and Morel (1990) for the adsorption of cations and anions onto hydrous ferric oxide (HFO). This model was used to predict the pH-dependent partitioning of As(V) on the soil by assuming that HFO governed the adsorption of the As(V) on these materials. The model was used to predict the As(V) adsorption envelopes for all of the experimental data sets except for the experiment with CO3 (because there are no equilibrium constants for the adsorption of CO3 to HFO available in the model). The concentration of HFO in the soils was estimated from both the AOD-extractable (amorphous and poorly crystalline) and CBD-extractable (crystalline) Fe contents (Jackson et al., 1986). For example, 5 g soil L-1 (0.1 g per 20 mL) with a CBD-extractable Fe concentration of 25.8 g kg-1 (Table 1) yields 0.205 g HFO L-1 using the formula Fe2O3·H2O for HFO recommended by Dzombak and Morel (1990). Using the AOD-extractable Fe concentration to estimate the HFO content (Fig. 3) resulted in the model typically underpredicting the extent of adsorption. In contrast, using the CBD-extractable Fe concentration to estimate the HFO content (Fig. 3) resulted in the model consistently overpredicting the extent of adsorption and the pH at which adsorption would begin to decrease. These results indicate that the AOD-extractable Fe oxide concentration underestimates the HFO content or that other components of the soil significantly contribute to the adsorption of As(V). These results also indicate that the CBD-extractable Fe oxide concentration overestimates the HFO content of the soil.

Although the model correctly predicted the competitive effect of PO4 on As(V) adsorption (Fig. 3c), the model significantly overpredicted the magnitude of the effect when using the AOD-extractable Fe oxide concentration as an estimate of the HFO content and underestimated the magnitude of the effect when using the CBD-extractable Fe concentration to estimate the HFO content. The differences in the effect of adding PO4 to the model suggests that the HFO content is somewhere in between the AOD- and CBD-extractable Fe oxide concentration. In most cases, the agreement between the experimental data and model-predicted results is reasonable given the inherent assumptions in the modeling and suggests that the As(V) adsorption capacity for heterogeneous, multicomponent materials may be modeled with knowledge of the soil's Fe oxide content and pH within the RMSE shown. Overall, the model using the AOD-extractable Fe content (RMSE 0.116–0.381) predicted the pH-dependent adsorption of As(V) to the soil better than when using the CBD-extractable Fe oxide content (RMSE 0.199–0.726) to estimate the HFO content.

The batch data indicate several complicating factors that may influence As(V) adsorption and transport in flowing systems. These include: nonlinear adsorption, adsorption nonequilibrium, and the influence of chemical parameters (pH and the presence of PO4) on the transport of As(V) in the subsurface. Column experiments were then undertaken to ascertain the effects of these parameters in flowing systems.

At the soil's natural pH of 4.5, the breakthrough curve was asymmetrical (Fig. 4) , which is indicative of both nonlinear and rate-limited adsorption (Brusseau, 1998). As predicted by the batch results, the breakthrough curve demonstrates that As(V) exhibits significant retardation due to chemical interactions with subsurface materials. Detectable As(V) concentrations were not observed in the column effluent until almost 200 pore volumes. The transport experiment also illustrated another phenomenon not uncovered in the batch experiments, the presence of an irreversibly adsorbed and/or slowly desorbing fraction. After peaking at a relative As(V) effluent concentration of 0.624 after the completion of the As(V) input pulse, the As(V) concentration began to decrease as the influent concentration was switched to an As(V)-free solution. After >1100 pore volumes of desorption, the relative As(V) concentration was approximately 0.06 and decreasing very slowly, exhibiting a tailing effect. At this point, numerical integration of the effluent curve indicated that the amount of As(V) recovered was only 44% of the total As(V) input. The low recovery suggests that a significant fraction of the initial As(V) has effectively irreversibly adsorbed to the soil. The presence of a irreversible or slowly desorbing fraction has also been detected at As-contaminated sites (Kuhlmeier, 1997a,b).



View larger version (26K):
[in this window]
[in a new window]
  		Fig. 4. Relative effluent As(V) concentration versus pore volumes eluted (pH = 4.5, 0.01 M NaNO3, Co = 1 mg L-1, pore water velocity 0.53 cm min-1).

 
The As(V) breakthrough was predicted using the measured isotherm (Fig. 2) and two adsorption equilibrium models. Both models assume adsorption equilibrium is instantaneously maintained between the fluid and the solid phase. The LLE model predicted a similar peak As(V) concentration to that of the experimental data. However, the predicted peak breakthrough did not occur until approximately 600 pore volumes later than observed. The model also did not exhibit a tailing effect, as was observed in the experiment breakthrough curve. The model also overpredicted the amount of As(V) available for recovery, as 100% recovery was exhibited in the model-predicted breakthrough curve whereas only 44% recovery was obtained in the experimental breakthrough curve. Quantitatively, the model predicted the experimental data to RMSE = 0.317. The variations between the experimental data and the model-predicted data could be attributed to violations of the underlying assumptions of the model: (i) the nonlinearity of the isotherm and (ii) rate-dependent adsorption equilibrium. This pore water velocity (0.53 cm min-1) results in a hydraulic residence time of 3.21 min, which is too fast to allow equilibrium between As(V) and the soil based on the kinetic data (Fig. 1). The shortcomings of this model are noteworthy as the model is commonly used in predicting the migration of As(V) (Carrillo and Drever, 1998) and other contaminants in the subsurface (Bethke and Brady, 2000).

The NLLE model using the nonlinear Freundlich isotherm parameters (Kf = 278 L kg-1 and n = 0.32) predicted the data more accurately (RMSE = 0.273) than the LLE model. Again, the experimental breakthrough occurred more rapidly than predicted, though not to the same extent as with the LLE model. The predicted peak relative As(V) concentration was approximately 0.95 and occurred approximately 100 pore volumes after the experimental breakthrough peak. The NLLE model thus predicted the time to peak breakthrough (as measured in dimensionless pore volumes) significantly better than the LLE model. Although the NLLE model predicted a tailing effect similar to the experimental data, it still overestimated the percentage of As(V) available for recovery. Both the NLLE and the LLE models assume complete reversibility of the adsorption reaction. However, the LLE model did not exhibit the same tailing effect that the NLLE model and experimental data exhibit. The differences in the two models indicate that the linearity or nonlinearity with which the isotherm data can be described plays a key role in predicting breakthrough curves, even in the absence of rate-dependent adsorption.

The presence of an irreversible or slowly desorbing fraction produced an interesting phenomenon in comparison with the predicted results assuming adsorption equilibrium (Fig. 4). The presence of rate-limited adsorption theoretically promotes greater mobility of subsurface contaminants compared with those that rapidly achieve adsorption equilibrium, as the initial breakthrough curve confirms. Arsenic(V) concentrations appeared in the column effluent faster than predicted by both models. However, due to the presence of an irreversible or slowly desorbing fraction, the peak aqueous As(V) concentration and the total amount of As(V) recovered was lower than predicted based on the two equilibrium models. This may have important implications to subsurface As(V) contamination. Although subsurface As(V) may reach a given point of interest (e.g., a site boundary) faster than predicted by the models, the overall mobility in terms of maximum concentrations and total recovery may be lower than predicted by the models.

Increasing the pore water velocity by a factor of three from 0.53 to 1.6 cm min-1 (decreasing the column residence time from 3.2 to 1.1 min) significantly increased As(V) mobility (Fig. 5) , further evidence of adsorption nonequilibrium. Detectable As(V) effluent concentrations occurred after approximately 15 pore volumes, an order of magnitude faster than at the lower pore water velocity of 0.53 cm min-1. The relative effluent concentration at approximately 875 pore volumes peaked at 0.850. Significantly, the irreversible or slowly desorbing fraction and the associated degree of tailing also decreased. After approximately 1800 pore volumes, the effluent As(V) concentration was below the detection limit and the quantity of As(V) recovered was 76%. A decrease in pore volumes required for breakthrough, an increase in relative peak breakthrough concentration, and an increase in As(V) recovery occurred at higher pore water velocity. The increased pore water velocity also had a marked effect on breakthrough curve asymmetry, which is indicative of adsorption nonequilibrium.



View larger version (23K):
[in this window]
[in a new window]
  		Fig. 5. Effect of varying physical (pore water velocity) and chemical (pH and PO4) conditions on As(V) transport (0.01 M NaNO3 and Co = 1 mg L-1).

 
As predicted from the batch data (Fig. 3), an increase in pH from 4.5 to 9.0 significantly increased the mobility of As(V) (Fig. 5) as measured by the initial breakthrough, peak breakthrough concentration, and total As(V) recovery. At the same pore water velocity, As(V) breakthrough commenced at <15 pore volumes at pH 9 compared with >200 pore volumes at pH 4.5. In addition, the peak relative concentration increased to 0.757 and the total recovery increased to >65%. The degree of tailing also decreased, as after 2000 pore volumes the effluent As(V) concentration was less than the detection limit (<5 µg L-1). The increased pH results complement the results of the batch experiments in showing that an increase in pH results in a decrease in the total degree of adsorption. These results also extend the results of the batch experiments to show that a decrease in the amount of adsorption results in a corresponding decrease in the amount of irreversible or slowly desorbing As(V).

As in the batch experiments, PO4 had the most dramatic effect on As(V) adsorption and the resulting mobility and total recovery of As(V) (Fig. 5). A detectable concentration of As(V) appeared in the effluent almost immediately, and the relative As(V) effluent peak concentration at the end of the input pulse was 0.969 compared with 0.624 in the same experiment without PO4. After the As(V)–PO4 input pulse, the As(V) rapidly desorbed. After desorbing for approximately 50 pore volumes the relative As(V) effluent concentration was less than 0.10. After 1200 pore volumes the relative effluent As(V) concentration was lower than the detection limit (<5 µg L-1) and exhibited no tailing effect. The As(V) recovery of this column was 92%, which is the highest of the column experiments conducted, as the addition of PO4 decreased the irreversible or slowly desorbed fraction, which greatly increased the mobility of As(V).

These PO4 results are significant. In the absence of PO4, As(V) would be considered a relatively immobile anion. In the presence of 0.25 mM PO4, it would be considered a relatively mobile anion. Although this level of PO4 is higher than in many subsurface environments, PO4 is ubiquitous in the natural environment. In addition, there are a number of agricultural areas where both As(V) and PO4 have been added to the subsurface as a pesticide and fertilizer, respectively (Smith et al., 1998).


   CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 METHODS AND MATERIALS
 MODELING
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
The results of this paper illustrate several important aspects regarding the mobility of As(V) in the subsurface. There is a significant potential for both nonlinear and rate-dependent (i.e., nonideal) adsorption–desorption of As(V). Adsorption of As(V) to this iron oxide–containing media continued over a time scale of weeks and the adsorption isotherm was markedly nonlinear (n = 0.32). Both of these phenomena could result in greater mobility of As(V) than predicted from simple linear partitioning approaches. However, the presence of an irreversible or slowly desorbing fraction acted to counteract the increased mobility that was otherwise predicted. Chemical parameters such as pH and the presence of PO4 also significantly affected As(V) mobility. A commonly used surface complexation model for As(V) adsorption to HFO predicted the pH-dependent adsorption of As(V) to this soil within an RMSE of 0.116 to 0.726, indicating the potential for such models to provide initial estimates of pH-dependent As(V) adsorption in the absence of site-specific data. Using the AOD-extractable Fe concentration as an estimate of the HFO content typically resulted in underpredicting the extent of adsorption, while using the CBD-extractable Fe concentration to estimate the HFO content consistently resulted in overpredicting the extent of adsorption. Arsenic(V) was significantly retarded in soil columns. Although the nonlinear equilibrium model (RMSE = 0.273) predicted the data better than the linear model (RMSE = 0.317), As(V) breakthrough occurred more rapidly than predicted by either model due to adsorption nonequilibrium. For the parameters examined in this study, the effect on As(V) mobility and recovery increased in the order pH < pore water velocity < PO4. These results illustrate the complex hydrogeochemical factors (e.g., pH- and rate-dependent adsorption) that would need to be incorporated into any comprehensive modeling approach to describe As(V) transport in the subsurface.


   ACKNOWLEDGMENTS
 
The authors acknowledge the comments of two anonymous reviewers that significantly improved the manuscript. L. Elizabeth Williams was supported in part by a graduate teaching assistantship from the Auburn University Department of Civil Engineering and by a graduate research assistantship sponsored by the Strategic Environmental Research and Development Program (SERDP) under the direction of Ms. Cathy Vogel and Dr. Andrea Leeson.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 METHODS AND MATERIALS
 MODELING
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 


Related articles in JEQ:

This Issue in Journal of Environmental Quality

JEQ 2003 32: 745-750. [Full Text]  



This article has been cited by other articles:


Home page
 Soil Sci.Home page
H. Zhang and H. M. Selim
Modeling the Transport and Retention of Arsenic (V) in Soils
Soil Sci. Soc. Am. J., August 22, 2006; 70(5): 1677 - 1687.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Related articles in JEQ
Right arrow Similar articles in this journal
Right arrow Similar articles in Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Web of Science (14)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Williams, L. E.
Right arrow Articles by Melville, J. G.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Williams, L. E.
Right arrow Articles by Melville, J. G.
GeoRef
Right arrow GeoRef Citation
Agricola
Right arrow Articles by Williams, L. E.
Right arrow Articles by Melville, J. G.
Related Collections
Right arrow Heavy Metals

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Crop Science
Journal of Natural Resources
and Life Sciences Education		 Vadose Zone Journal
Soil Science Society of America Journal Journal of Plant Registrations The Plant Genome