Modeling the Kinetics of Ferrous Iron Oxidation by Monochloramine

reactions involving the amidogen radical that lead to a nonelementary stoichiometry. Using the method of initial rates, the aqueous-phase reactions we...
27 downloads 0 Views 131KB Size
Download PDF
Environ. Sci. Technol. 2002, 36, 662-668

Modeling the Kinetics of Ferrous Iron Oxidation by Monochloramine PETER J. VIKESLAND* AND RICHARD L. VALENTINE Department of Civil and Environmental Engineering, The University of Iowa, Iowa City, Iowa 52242

The maintenance of disinfectants in distribution systems is necessary to ensure drinking water safety. Reactions with oxidizable species can however lead to undesirable disinfectant losses. Previous work has shown that the presence of Fe(II) can cause monochloramine loss in distribution system waters. This paper further examines these reactions and presents a reaction mechanism and kinetic model. The mechanism includes both aqueous-phase reactions and surface-catalyzed reactions involving the iron oxide product. In addition, it considers competitive reactions involving the amidogen radical that lead to a nonelementary stoichiometry. Using the method of initial rates, the aqueous-phase reactions were found to have firstorder dependencies on Fe(II), NH2Cl, and OH- and a rate coefficient (kNH2Cl,soln) of 3.10 ((0.560) × 109 M-2 min-1. The surface-mediated reactions were modeled by assuming the formation of two surface species: >FeOFe+ and >FeOFeOH. Using numerical techniques, combined rate coefficients for the surface-mediated processes were determined to be 0.56 M-3 min-1 and 3.5 × 10-8 M-4 min-1, respectively. The model was then used to examine monochloramine and Fe(II) stability under conditions similar to those observed in distribution systems. Our findings suggest the potential utility of monochloramine as an oxidant for Fe(II) removal in drinking water treatment.

Introduction The maintenance of disinfecting residuals in drinking water distribution systems is one of the primary goals of drinking water treatment. Without a disinfecting residual, it is difficult to prevent microbial outbreaks within the distribution system, and consumer health may be at risk. Historically, disinfectant dosages at the treatment plant were set high enough to consistently maintain a disinfecting residual throughout the distribution system. Recently in the effort to limit disinfection byproduct (DBP) formation, however, many utilities have been forced to decrease the disinfectant concentration, and the potential for outbreak episodes has increased. Management of disinfection practices to maintain a residual sufficient for microbial disinfection while simultaneously minimizing DBP formation is made difficult by abiotic reactions within the distribution system. A substantial research effort has been made to characterize reactions between disinfectants and organic matter (1-4); nevertheless, * Corresponding author present address: Department of Civil and Environmental Engineering, Virginia Polytechnic Institute, 415 Durham Hall, Blacksburg, VA 24061-0246; e-mail: PeterVikesland@ vt.edu. 662

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 4, 2002

considerably less is known about the role of oxidizable inorganic species. Reduced iron, in the form of cast iron pipes and reduced surface scales (e.g., pipe deposits or tubercles), is an important oxidizable inorganic species found within distribution systems that can readily react with free chlorine and monochloramine, the most frequently used distribution system disinfectants. Although reactions between disinfectants and cast iron pipe are possible, the pipes are typically protected from the solution by the surface scale (5). Within the scale, however, oxidizable ferrous iron (Fe(II)) is commonly detected, and it may exert a significant disinfectant demand. Under stagnant conditions, the ferrous iron in these scales can dissolve, leading to soluble ferrous iron concentrations of 0.1-40 mg/L (2-716 µM) (6). Our previous work (7) has shown that Fe(II) and monochloramine readily react under conditions similar to those observed within distribution systems, thereby suggesting the importance of these reactions as a disinfectant loss pathway. In this paper, we present experimental studies to support a hypothesized reaction mechanism. Rate coefficients and reaction dependencies for solution and surface-mediated reactions were determined using a variety of approaches. The resulting model is then used to examine monochloramine stability in the presence of ferrous iron for a number of environmentally relevant reaction conditions. A theoretical presentation is particularly warranted given the difficulty in measuring monochloramine in the presence of ferrous iron.

Materials and Methods All experiments were conducted using deionized water produced by a Barnstead ULTRO pure water system. A Fisher Scientific model 50 pH meter coupled with an Orion Scientific combination reference/analytical electrode was used for all pH measurements. Activity corrections for all ionic species were calculated through the use of the Davies equation (8). Monochloramine stocks and solutions were produced according to published procedures (4, 7). Kinetic experiments were initiated by adding concentrated ferrous sulfate (10 µg/µL) to a sealed vial containing monochloramine. Reaction temperature was controlled by keeping the vials in a thermostated water bath. Periodic samples were taken, and the Fe(II) and/or the NH2Cl concentrations were determined. The dissolved and total Fe(II) concentrations were measured using a modified Ferrozine method (9). In control experiments, the Ferrozine-Fe(II) complex was found to form instantaneously and to be unreactive toward monochloramine (10). It was therefore possible to quench the monochloramine-Fe(II) reactions by simply adding Ferrozine. Monochloramine was quantified via potentiometric titration with phenylarsine oxide (PAO). This method was used because of the interference of Fe(II)/Fe(III) on the standard DPD/FAS method. Experiments to evaluate the accuracy of the PAO titration indicated that Fe(II) concentrations below 358 µM decreased the measured NH2Cl concentration of a control () 24.6 µM NH2Cl) by a maximum of 6.5%. For the Fe(II) and NH2Cl concentrations utilized in these experiments, the interference of Fe(II) on the PAO titrations is expected to be less than 1.0%. In general, rate expressions and kinetic coefficients for these reactions were obtained using measurements of Fe(II) concentration. Monochloramine concentrations were primarily used to verify the model predictions made using the iron oxidation data. 10.1021/es002058j CCC: $22.00

 2002 American Chemical Society Published on Web 01/17/2002

Model Development Our previous work has shown that Fe(II) and monochloramine readily react in solution via a direct interaction between molecular monochloramine and aqueous ferrous iron, Fe(II)soln (7). The reactions are autocatalytic as the iron oxide product of the aqueous-phase reaction accelerates the overall reaction kinetics by enabling the formation of highly reactive Fe(II) surface complexes (Fe(II)surf). The rate-limiting reactions are believed to involve the one-electron reduction of monochloramine. Electron spin resonance evidence indicates that this produces the radical intermediate amidogen (•NH2). This radical either reacts with Fe(II) or is scavenged by reactions with other species. These scavenging reactions result in a nonelementary stoichiometry for this system. At pH values greater than 8.6, a 2:1 stoichiometry (Fe(II) oxidized: NH2Cl reduced) exists, but as the pH is decreased, the stoichiometry approaches 1:1. On the basis of this information, we hypothesize that the general reactive scheme given by eqs 1-6 describes this system: kNH

2Cl,soln

Fe(II)soln + NH2Cl 98 •NH2 + Cl- + Fe(III) (1) + fast

•

Fe(II)soln + NH2 + H 98 Fe(III) + NH3

(2)

kNH

2Cl,surf

Fe(II)surf + NH2Cl 98 •NH2 + Cl- + Fe(III) (3) fast

•

Fe(II)surf + •NH2 + H+ 98 Fe(III) + NH3

(4)

Fe(II)soln + >FeOH a Fe(II)surf

(5)

kscavenge

NH2 + scavengers (e.g., HCO3-, O2) 98 products (6)

where >FeOH denotes an active site on the surface of the ferric iron precipitate produced by the hydrolysis of Fe(III). If amidogen reacts exclusively with Fe(II), an apparent stoichiometry of 2:1 is attained:

2Fe(II) + NH2Cl + H+ f 2Fe(III) + NH3 + Cl-

(7)

If amidogen reacts with radical scavengers and the products do not oxidize Fe(II), then an apparent stoichiometry less than 2:1 is observed. Typical drinking waters contain a variety of potential amidogen scavengers such as carbonate (11) or dissolved oxygen (DO) (12), and reactions with these species can affect the apparent stoichiometry. Assuming that the reactions that occur between monochloramine and ferrous iron are rate limiting (i.e., the amidogen radical is a highly unstable intermediate), we hypothesized that the kinetics of this system were described by the following rate expression:

-

d[Fe(II)tot] ) Θ(kNH2Cl,soln[Fe(II)soln][NH2Cl][OH-] + dt kNH2Cl,surf[Fe(II)surf][NH2Cl][OH-]) (8)

First-order dependencies for NH2Cl and Fe(II) are expected due to the low probability of trimolecular reactions involving Fe(II) and NH2Cl (13). A hydroxide dependency for the aqueous-phase and surface-mediated Fe(II) and NH2Cl interactions is included because it was previously shown that increases in pH accelerate the overall reaction rate (7). Θ is a parameter used to account for the effects of radical scavenging on the overall reaction stoichiometry. Θ can be

FIGURE 1. Determination of reaction order with respect to Fe(II). Errors in kinit values correspond to 95% confidence intervals. [NH2Cl]0 ) 282 µM, [Fe(II)]0 ) 16-179 µM, pH 7.33, µ ) 0.1 M, CT,CO3 ) 5.7 mM, 25 °C, DO < 1 mg/L. rationalized in terms of the relative importance of competing reactions that determine the fate of the amidogen radical. This approach is used in the absence of detailed information concerning the scavenging of this radical and is consistent with that used to describe the effects of hydroxyl radical scavengers in reactions involving ozone (14). In the remainder of this paper, we discuss experiments conducted both to experimentally verify that the assumed reaction orders are correct and to determine kinetic coefficients. The resultant model is then utilized both to describe the potential for monochloramine loss within distribution systems containing Fe(II) and to assess the use of monochloramine as an oxidant for Fe(II) removal during drinking water treatment.

Results and Discussion The solution-phase reaction orders and the rate coefficient, kNH2Cl,soln, were evaluated using the method of initial rates. With this method, it was possible to isolate the homogeneous reaction kinetics (13). This technique utilizes only the kinetic data obtained during the initial 10-30% of the reaction to determine reaction orders and rate coefficients. Under this restriction, monochloramine loss generally never exceeded 10% of the initial value. Considering only the aqueous-phase reactions, eq 8 has the following form:

-

d[Fe(II)soln] ) ΘkNH2Cl,soln[Fe(II)soln][NH2Cl][OH-] (9) dt

Order with Respect to Ferrous Iron (Fe(II)). It was initially assumed that Fe(II) is oxidized by monochloramine in a firstorder reaction, and the collected initial rate data was linearized using the pseudo-first-order expression:

log[Fe(II)soln] ) -kinitt + log[Fe(II)]0

(10)

where [Fe(II)soln] is the measured aqueous-phase iron concentration, [Fe(II)]0 is the initial ferrous iron concentration, and kinit ()ΘkNH2Cl,soln[NH2Cl][OH-]) is the pseudo-first-order rate coefficient obtained from the slope of a first-order plot. For a wide range of initial iron concentrations, the slopes (kinit values) are approximately equal at the 95% confidence level, indicating that the assumed first-order dependency on Fe(II)soln is valid (Figure 1). Order with Respect to Monochloramine. By linearizing the expression for kinit, it was possible to determine the reaction order with respect to monochloramine: VOL. 36, NO. 4, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

663

FIGURE 2. Determination of monochloramine reaction order. Error in slope corresponds to 95% confidence interval. [NH2Cl]0 ) 141, 282, and 704 µM; [Fe(II)]0 ) 23-358 µM; pH 7.33; µ ) 0.1 M; CT,CO3 ) 5.7 mM; 25 °C; DO < 1 mg/L.

FIGURE 3. Hydroxide ion reaction order determination. Errors in kinit values correspond to 95% confidence intervals. [NH2Cl]0 ) 141, 282, and 704 µM; [Fe(II)]0 ) 89.5 µM; pH 6.5-8.5; µ ) 0.1 M; CT,CO3 ) 5.7 mM; 25 °C; DO < 1 mg/L.

TABLE 1. kinit Values as a Function of Solution pHa

the hydroxide and carbonate concentrations, and this may alter the speciation and reactivity of Fe(II) (15-17).

NH2Cl (µM) 141

282

pH

[OH-]

kinit (min-1)

kinit/Θ (min-1)

6.23 6.54 6.72 7.02 7.32 7.58 7.81 8.05 6.40 6.61 6.78 7.01 7.20 7.49 7.70 7.93 6.66 6.84 7.13 7.40 7.65

2.25 × 10-8 4.60 × 10-8 6.96 × 10-8 1.39 × 10-7 2.77 × 10-7 5.04 × 10-7 8.56 × 10-7 1.49 × 10-6 3.33 × 10-8 5.40 × 10-8 7.99 × 10-8 1.36 × 10-7 2.10 × 10-7 4.10 × 10-7 6.65 × 10-7 1.13 × 10-6 6.06 × 10-8 9.18 × 10-8 1.79 × 10-7 3.33 × 10-7 5.92 × 10-7

0.0165 ( 0.0015 0.0244 ( 0.0022 0.0428 ( 0.0048 0.0767 ( 0.007 0.136 ( 0.021 0.421 ( 0.119 0.629 ( 0.293 1.47 ( 1.32 0.0380 ( 0.003 0.0551 ( 0.005 0.0846 ( 0.016 0.150 ( 0.028 0.291 ( 0.051 0.451 ( 0.200 0.861 ( 0.191 2.21 ( 1.20 0.294 ( 0.053 0.488 ( 0.102 0.838 ( 0.153 1.14 ( 1.91 2.08 ( 4.58

0.0154 0.0208 0.0348 0.0578 0.0956 0.280 0.398 0.888 0.0337 0.0461 0.0677 0.113 0.210 0.305 0.558 1.37 0.243 0.385 0.616 0.787 1.36

Using kinit values obtained for a variety of pH values (Table 1), plots of log kinit/Θ versus log [OH-] were produced (Figure 3). The kinit values were normalized to a calculated apparent stoichiometry (Θ ) 0.32 × pH - 0.92; 7) for each pH to eliminate any effects associated with changes in Θ on the observed kinit values. At the 95% confidence level, the calculated slope for each monochloramine concentration is statistically indistinguishable from 1. A parallel figure not normalized by Θ was also produced, and the slopes were found to change insignificantly (data not shown), thereby illustrating that changes in Θ have a relatively minor effect on kinit. The slope of unity indicates that over the tested pH range that the aqueous-phase reaction has an apparent firstorder dependency on the hydroxide ion concentration.

log kinit ) β log[NH2Cl] + log[OH-] + log kNH2Cl,soln + log Θ (11)

Studies examining ferrous iron oxidation by oxygen have shown that the reactions are a complex function of the solution pH (15-18), with the apparent hydroxide ion dependency ranging from zero-order to second-order (16) as the pH varies from slightly less than 2 to greater than 8. This complex pH dependence is hypothesized to occur due to the parallel oxidation of ferrous ion (Fe2+), its hydroxo complexes (Fe(OH)+, Fe(OH)2, Fe(OH)3-), and its carbonate complexes (FeCO3, Fe(CO3)22-, Fe(CO3)(OH)-) (17). Changing the type and number of complexing ligands causes the electronic structure of the Fe(II) atom to be altered, thereby affecting its reactivity. Both hydroxyl and carbonyl ligands donate electron density to the Fe(II) atom, and this donated electron density makes the metal ion a better reductant (19).

For data obtained at a fixed pH, the slope of a plot of log kinit vs log[NH2Cl] was 1.06 ( 0.07, thereby indicating that β, the aqueous-phase reaction order with respect to monochloramine, was one (Figure 2). Order with Respect to [OH-]. Previous results (7) have shown that as the solution pH increases, the rate at which iron is oxidized by monochloramine also increases. This increase could be a result of the effect of pH on the apparent stoichiometry (Θ), or it could be due to the involvement of hydroxide ion or pH-dependent species in the rate-limiting reactions. When the solution pH is increased over the pH range of 6.8-9.5, the apparent stoichiometry increases from approximately 1 to 2 (7). This increase may affect the reaction kinetics since a greater percentage of monochloramine’s oxidizing potential is used to oxidize Fe(II) at higher pH values than at lower pH values. Alternatively, changes in pH affect

Over the pH range of 6.23-8.05, it was found that the monochloramine-ferrous iron reactions are first-order with respect to the hydroxide ion concentration, suggesting that Fe(OH)+ is the reactive species for this pH range. It is also possible, however, that the first-order hydroxide ion dependency may mask a first-order carbonate dependency. Over this pH range, the carbonate concentration is linearly related to the hydroxide ion concentration (20). Preliminary experiments have shown that changes in carbonate concentration dramatically affect the kinetics of the monochloramineferrous iron reactions, thereby indicating the important role of carbonate in these reactions (10). As discussed previously (7), the effect of carbonate is potentially quite complex due to its ability to catalyze monochloramine auto-decomposition (21), to act as a radical scavenger (11), and to form iron complexes (17, 20). In the absence of further information concerning the effects of carbonate, the aqueous-phase

704

a [Fe(II)] ) 89.5 µM, pH 6.2-8.05, µ ) 0.1 M, C 0 T,CO3 ) 5.7 mM, 25 °C, DO < 1 mg/L. [OH-] ) Kw/((10-pH) × γOH) and γOH ) 0.754 (calculated using the Davies equation). Θ ) 0.32 × pH - 0.92 (7). Errors in kinit values correspond to 95% confidence intervals.

664

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 4, 2002

FIGURE 4. Determination of kFe,soln. Error in kFe,soln value corresponds to the 95% confidence interval. [NH2Cl]0 ) 141, 282, and 704 µM; [Fe(II)]0 ) 89.5 µM; pH 6.5-8.5; µ ) 0.1 M (NaClO4); CT,CO3 ) 5.7 mM; 25 °C; DO < 1 mg/L.

FIGURE 5. Comparison of aqueous-phase and dual-phase (aqueous + surface-mediated) model predictions. Solid curves are dualphase model predictions, and dashed curves are aqueous-phase. [NH2Cl]0 ) 282 µM, [Fe(II)]0 ) 94 µM, µ ) 0.1 M, CT,CO3 ) 5.7 mM, 25 °C, DO < 1.0 mg/L. reaction was expressed in terms of first-order dependencies on both [Fe(II)soln] and hydroxide ion concentration. Further experimentation is necessary to better elucidate the complicated roles of carbonate and hydroxide ion. Determination of kNH2Cl,soln. The following rate law describes the aqueous-phase oxidation of ferrous iron by monochloramine:

-

d[Fe(II)soln] ) ΘkNH2Cl,soln[Fe(II)soln][NH2Cl][OH-] dt (12)

Using the collected experimental data in Table 1, kNH2Cl,soln was estimated to be 3.10 ((0.560) × 109 M-2 min-1 by plotting kinit/Θ versus [NH2Cl][OH-] (Figure 4). Surface-Mediated Reactions. The initial rate data discussed in the previous section enabled us to examine the solution-phase reaction kinetics under conditions where the autocatalytic effect of the iron oxide precipitate was minimized. Over longer reaction periods, the rate-enhancing effects of the precipitate need to be considered (Figure 5). A simple fourth-order Runge-Kutta routine (Stella, Version 5; High Performance Systems, Inc., Hanover, NH) was used to evaluate the overall reaction kinetics and obtain rate coefficients describing the surface reaction rates. Iron

FIGURE 6. Comparison of model fits obtained using the full dualphase model (solid lines) to (A) dual-phase model invoking only >FeOFe+ (dashed lines) and (B) dual-phase model invoking only >FeOFeOH (dashed lines). At high pH values the >FeOFe+ model underpredicts Fe(II) oxidation, and at both pH values, the >FeOFeOH model underpredicts Fe(II) oxidation. χ2 analysis of the model fits indicates that the dual-phase model provides a better overall fit: χ2dual-phase ) 0.017, χ2FeOFe+only ) 0.019, χ2FeOFeOHonly ) 0.026 with χ2 ) ∑((experimental data - model prediction)2/(model prediction)). [NH2Cl]0 ) 282 µM, [Fe(II)]0 ) 94 µM, µ ) 0.1 M (NaClO4), CT,CO3 ) 5.7 mM, 25 °C, DO < 1 mg/L. oxidation data obtained for two different monochloramine concentrations (282 and 704 µM), a fixed initial Fe(II) concentration of 94 µM, and a range of pH values was utilized in this evaluation. Final values for the rate parameters and for the pH dependence of Θ were determined via sensitivity analysis (10). The pH dependency of Θ was optimized due to the large error associated with the experimentally determined value. For this optimization, the slope of the Θ regression was varied over the range (slope ) 0.32 ( 0.14) obtained experimentally. Attempts were made to model the surface-mediated processes using reaction terms involving first- and secondorder dependencies on both Fe(II) and monochloramine. As expected, it was found that first-order dependencies on both species provided the best fit to the data. In addition, single reaction terms that were either first-order or second-order with respect to hydroxide ion were also considered; however, these single reaction terms could not adequately account for the effect of changes in pH on the oxidation rate. Therefore, two pH-sensitive surface reaction terms were included, enabling a better overall fit over an extended pH range than either individual term alone (Figure 6). The need to include two surface terms is consistent with previous research that indicated ferrous iron forms two different surface complexes when it adsorbs to an iron oxide surface over a neutral pH range (22-25): VOL. 36, NO. 4, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

665

Ksorb1

Fe(II)soln + >FeOH {\} >FeOFe+ + H+

(13)

Ksorb2

Fe(II)soln + >FeOH + H2O {\} >FeOFeOH + 2H+ (14) Ksorb1 and Ksorb2 are the equilibrium constants for the two surface complexes (>FeOFe+ and >FeOFeOH). After accounting for the formation of two different types of surface ferrous iron, the kinetic expression describing Fe(II) oxidation (eq 8) has the following form:

d[Fe(II)tot] ) Θ(kNH2Cl,soln[Fe(II)soln][NH2Cl][OH-] + dt kNH2Cl,surf1[NH2Cl][>FeOFe+] + kNH2Cl,surf2[NH2Cl][>FeOFeOH])

(15)

Because [>FeOFe+] and [>FeOFeOH] are not experimentally available, this expression was rewritten in terms of the aqueous-phase ferrous iron concentration:

-

TABLE 2. Kinetic Expressions and Model Parameters Describing Iron Oxidation and Monochloramine Reduction Kinetic Expressions -d[Fe(II)tot]/dt ) Θ(kNH2Cl,soln[OH-] + kNH2Cl,surf1Ksorb1RK-1 w [Fe(III)][OH ] + -]2)[Fe(II) kNH2Cl,surf2Ksorb2RK-2 [Fe(III)][OH soln][NH2Cl] w -d[NH2Cl]/dt ) (kNH2Cl,soln[OH-] + kNH2Cl,surf1Ksorb1RK-1 w [Fe(III)][OH ] + - 2 kNH2Cl,surf2Ksorb2RK-2 w [Fe(III)][OH ] )[Fe(II)soln][NH2Cl] d[Fe(III)]/dt ) Θ(kNH2Cl,soln[OH-] + kNH2Cl,surf1Ksorb1RK-1 w [Fe(III)][OH ] + -]2)[Fe(II) kNH2Cl,surf2Ksorb2RK-2 [Fe(III)][OH soln][NH2Cl] w Model Parameters kNH2Cl,soln 3.1 × 109 M-2 min-1 kNH2Cl,surf1Ksurf1R 0.56 M-3 min-1 kNH2Cl,surf2KsurfR 3.5 × 10-8 M-4 min-1 eq describing apparent 0.28 × pH - 0.92 stoichiometry (Θ)

d[Fe(II)tot] ) Θ(kNH2Cl,soln[OH-] + dt kNH2Cl,surf1Ksorb1K-1 w [>FeOH][OH ] + - 2 kNH2Cl,surf2Ksorb1K-1 w [>FeOH][OH ] )[Fe(II)soln][NH2Cl] (16)

where Kw is the hydrolysis constant for water () 10-14.00; 26) and [>FeOH] ) R[Fe(III)]. Because not all of the Fe(III) produced by Fe(II) oxidation is available to form surface complexes (e.g., some of the Fe(III) is internal to the oxide and therefore inaccessible to the solution), it is necessary to invoke a proportionality constant (R) that correlates the number of surface sites to the total amount of Fe(III). Values for R calculated using information from the literature range from 4.0 × 10-3 to 6.0 × 10-2 and depend on the type of oxide and the conditions under which it is formed (27). Unfortunately, the identity of the oxide precipitate may change during the course of these dynamic reactions; therefore, it is not possible to separate out either R or the equilibrium constants from the kinetic coefficients. For this reason, we report combined rate coefficients. Using the Runge-Kutta routine, eq 16 and the corresponding expressions for Fe(III) formation and NH2Cl reduction were evaluated to determine kNH2Cl,surf1Ksorb1R and kNH2Cl,surf2Ksorb2R. These values as well as the value of kNH2Cl,soln and the optimized pH dependence of Θ are tabulated in Table 2. Iron Oxidation Predictions and Model Validation. The validity of the heterogeneous reaction model was examined using experimental data obtained over a range of iron and monochloramine concentrations that greatly exceeded that used in its development. The model is reasonably capable of predicting iron oxidation rates for a [Fe(II)]0 range of 15100 µM and a [NH2Cl]0 range of 70-704 µM. An illustrative example of the model fit is shown in Figure 7. Additionally, not only can the model reasonably predict iron oxidation rates over these ranges but also it can account for changes in solution pH over the pH range of 6.23-8.05 (Figure 8). Typical deviations between model predictions and experimental data are on the order of (10 µM for the fastest experiments at the highest [Fe(II)]0 concentrations; for slower reaction conditions, the deviations are considerably smaller ((1-5 µM). Prediction of Monochloramine Reduction Rates. Considering that the kinetic coefficients discussed in this work were exclusively obtained using iron oxidation data, it is instructive to see how well the model predicts monochlor666

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 4, 2002

FIGURE 7. Ferrous iron oxidation as a function of [Fe(II)]0. Curves are model predictions. [NH2Cl]0 ) 70 µM, pH0 7.33, µ ) 0.1 M, CT,CO3 ) 5.7 mM, 25 °C, DO < 1.0 mg/L.

FIGURE 8. Ferrous iron oxidation as a function of initial pH. (9) pH 6.23, (b) pH 6.54, (2) pH 6.72, ([) pH 7.02, (0) 7.32, (O) pH 7.58, (4) pH 7.81, (]) pH 8.05. Curves are model predictions. [NH2Cl]0 ) 141 µM, [Fe(II)]0 ) 94 µM, CT,CO3 ) 5.7 mM, 25 °C, DO < 1.0 mg/L. amine loss. In all of the tested cases, the model underpredicts monochloramine loss, with this effect most pronounced at low Fe(II)-NH2Cl molar ratios, but nevertheless evident over the entire tested range (Figure 9). Underpredictions in monochloramine loss, even though iron oxidation is reasonably well predicted, suggest that the stoichiometric factor

FIGURE 9. Comparison between experimental data and model predictions of monochloramine reduction and iron oxidation (NH2Cl, b; Fe(II), 2). [NH2Cl]0 ) 253 µM, pH 7.00, µ ) 0.1 M, CT,CO3 ) 5.7 mM, 25 °C, DO < 1 mg/L. Solid curve is NH2Cl prediction, and dashed curve is Fe(II)soln prediction. (Θ) correlating iron oxidation and monochloramine reduction may differ from that used in the model. When the model was developed, it was assumed that for a given pH the same apparent stoichiometry (Θ) exists for all Fe(II)-NH2Cl molar ratios. There is, however, some evidence that the apparent stoichiometry of the Fe(II)-NH2Cl reactions changes as the initial ratio of Fe(II) to NH2Cl is varied (10). Limitations of the Kinetic Model. Although it is relatively successful at predicting iron oxidation rates over the full range of monochloramine and Fe(II) concentrations and pH values tested, the model is most accurate for low initial monochloramine and Fe(II) concentrations and in the middle pH range of 6.8-7.5 (i.e., conditions similar to those found within drinking water distribution systems). At higher pH values and for larger [NH2Cl]0 and [Fe(II)]0, the model formulation employed here may be too simplistic. There is an assumed equilibrium between the ferrous iron present in solution and the ferrous iron complexed at the ferric oxide surface. When the overall reaction kinetics are relatively slow (i.e., for smaller initial reactant concentrations and lower pH values), this assumption is reasonable. When the oxidation reactions are fast, however, the assumption of equilibrium may not be valid, and the model predicts reaction rates that are faster than those observed. To model the data for these short reaction times, it is necessary to determine how quickly the oxide precipitates and is able to catalyze the oxidation reactions. This added mechanistic complexity, although interesting, was beyond the scope of these experiments. At low pH values, where iron oxidation is relatively slow, it may be necessary to account for monochloramine autodecomposition as a parallel monochloramine loss pathway. Monochloramine auto-decomposition, a general term used to describe the complicated series of reactions that occur between monochloramine and its hydrolysis and disproportion products, has previously been shown to be a general acid-catalyzed process (21, 28). At the lower pH values tested, it is reasonable to consider that these reactions could lead to monochloramine loss without Fe(II) oxidation.

Environmental Significance The utility of the Fe(II)-monochloramine kinetic model is 2-fold: (i) It can be used to predict iron oxidation rates when monochloramine is used as an oxidant in primary treatment of drinking water. (ii) It can be applied to predict monochloramine loss kinetics in waters containing ferrous iron.

To achieve the secondary maximum contaminant level of 0.3 mg/L for source waters that contain significant amounts of Fe(II), it is typically necessary to utilize some sort of chemical oxidation process to remove the Fe(II) (29). For this purpose, many drinking water treatment plants use oxidants such as oxygen, potassium permanganate, ozone, or free chlorine to oxidize Fe(II) and cause it to precipitate as iron oxides that can be readily removed. Unfortunately, each of these oxidants has potential drawbacks that can limit their use (30, 31). On the basis of the experimental and model results presented here, it appears that in many instances monochloramine could be used as an alternative to these other oxidants. Not only can monochloramine oxidize Fe(II), but it will form fewer chlorinated byproducts than free chlorine (32). In fact, for water treatment plants that use chloramination for secondary disinfection, the use of chloramines as an oxidant may require only a few plant modifications. This use may be desirable because it can add CT value by taking advantage of the relatively long contact time created in the tanks used for iron oxidation and removal. To illustrate the potential utility of the monochloramineFe(II) model, we used it to predict the kinetics of these reactions under typical drinking water treatment conditions. These predictions utilized a pH range of 7.0-9.0, an Fe(II) concentration of 0.5 mg/L (8.95 µM), and a monochloramine concentration of 3 mg/L as Cl2 (42.3 µM). This monochloramine concentration is slightly lower than the U.S. EPA maximum monochloramine concentration of 4 mg/L as Cl2 (56 µM) and would therefore be used in water treatment (33). The ferrous iron concentration was chosen on the basis that it could also be reasonable. On the basis of the model predictions, if the solution pH is greater than 8.0, the apparent half-life of ferrous iron under these conditions is less than 5 min (Figure 10). Although these kinetics are not as fast as those for either free chlorine or ozone, they are reasonable and suggest that monochloramine could be used to remove Fe(II) under similar conditions. Perhaps more importantly, both monochloramine and ferrous iron are expected to coexist for a significant amount of time at near neutral to slightly acid pH values (e.g., 1 h or more at pH 7 or lower). Additionally, if it is desired to accelerate Fe(II) oxidation by monochloramine, the reaction can be conducted in the presence of an iron oxide-coated media. As shown in a separate paper (34), iron oxides other VOL. 36, NO. 4, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

667

Literature Cited

FIGURE 10. Model simulations of (A) iron oxidation and (B) monochloramine reduction as a function of pH. [NH2Cl]0 ) 3 mg/L as Cl2 (42.3 µM) and [Fe(II)]0 ) 0.5 mg/L (8.95 µM). than the precipitate formed in situ can greatly enhance the overall reaction rate. On the basis of the model predictions, the effects of changes in pH on monochloramine loss are quite complicated (Figure 10B). At low pH values, the initial loss of monochloramine is slow; however, since more monochloramine is lost for a given amount of Fe(II) oxidized (e.g., the apparent stoichiometry is close to one), the total amount of monochloramine lost is greater than at higher pH values. At those higher pH values, where the apparent stoichiometry is closer to 2:1 (Fe(II) oxidized:NH2Cl reduced), the total loss of monochloramine is diminished. Although the initial loss of monochloramine is rapid at these high pH values, there is less total loss, and the residual concentration is larger. These model simulations suggest that a high pH should be used within the distribution system in order to minimize monochloramine loss yet to also maximize its stability.

Acknowledgments We thank Alan Stone and three anonymous reviewers for their constructive comments about the manuscript. The aid of Michael Shoup in conducting the experiments is gratefully acknowledged. This work was supported by a grant from the American Water Works Association Research Foundation (AWWARF) and by the Abel Wolman Doctoral Fellowship of the American Water Works Association (AWWA).

668

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 4, 2002

(1) Reckhow, D. A.; Singer, P. C.; Malcolm, R. L. Environ. Sci. Technol. 1990, 24, 1655-1664. (2) Babcock, D. B.; Singer, P. C. J. Am. Water Works Assoc. 1979, 71, 149-152. (3) Qualls, R. G.; Johnson, J. D. Environ. Sci. Technol. 1983, 17, 692-698. (4) Vikesland, P. J.; Ozekin, K.; Valentine, R. L. Environ. Sci. Technol. 1998, 32, 1409-1416. (5) Sarin, P.; Bebee, J.; Beckett, M. A.; Jim, K. K.; Lytle, D. A.; Clement, J. A.; Kriven, W. M.; Snoeyink, V. L. Proceedings of the American Water Works Association Annual Conference and Exposition, Denver, CO, 2000. (6) Benjamin, M. M.; Sontheimer, H.; Leroy, P. Internal Corrosion of Water Distribution Systems, 2nd ed.; AWWARF and DVGW: Denver, CO, 1996; pp 29-70. (7) Vikesland, P. J.; Valentine, R. L. Environ. Sci. Technol. 2000, 34, 83-90. (8) Snoeyink, V. L.; Jenkins, D., Water Chemistry; John Wiley and Sons: New York, 1980. (9) Roden, E. E.; Zachara, J. M. Environ. Sci. Technol. 1996, 30, 1618-1628. (10) Vikesland, P. J. Ph.D. Dissertation, The University of Iowa, Iowa City, Iowa, 1998. (11) Neta, P.; Huie, R. E.; Ross, A. B. J. Phys. Chem. Ref. Data 1988, 17, 1027-1284. (12) Crowley, J. N.; Sodeau, J. R. J. Phys. Chem. 1989, 93, 4785-4790. (13) Laidler, K. J. Chemical Kinetics, 3rd ed.; Harper Collins: New York, 1987. (14) Hoigne, J. In The Handbook of Environmental Chemistry; Hrubec, J., Ed.; Springer-Verlag: Berlin, 1998; pp 83-141. (15) Millero, F. J. Geochim. Cosmochim. Acta 1985, 49, 547-553. (16) Wehrli, B. In Aquatic Chemical Kinetics; Stumm, W., Ed.; WileyInterscience: New York, 1990; pp 311-336. (17) King, D. W. Environ. Sci. Technol. 1998, 32, 2997-3003. (18) Singer, P. C.; Stumm, W. Science 1970, 167, 1121-1123. (19) Luther, G. W., III. In Aquatic Chemical Kinetics; Stumm, W., Ed.; Wiley-Interscience: New York, 1990; pp 173-198. (20) King, D. W.; Farlow, R. Mar. Chem. 2000, 70, 201-209. (21) Vikesland, P. J.; Ozekin, K.; Valentine, R. L. Water Res. 2001, 35, 1766-1776. (22) Zhang, Y.; Charlet, L.; Schindler, P. W. Colloids Surf. 1992, 63, 259-268. (23) Coughlin, B. R.; Stone, A. T. Environ. Sci. Technol. 1995, 29, 2445-2455. (24) Charlet, L.; Liger, E.; Gerasimo, P. J. Environ. Eng. (N.Y.) 1998, 124, 25-30. (25) Liger, E.; Charlet, L.; Van Cappellen, P. Geochim. Cosmochim. Acta 1999, 63, 2939-2955. (26) Stumm, W.; Morgan, J. J. Aquatic Chemistry: Chemical Equilibria and Rates in Natural Waters, 3rd ed.; John Wiley and Sons: New York, 1996. (27) Cornell, R. M.; Schwertmann, U. The Iron Oxides: Structure, Properties, Reactions, Occurrence and Uses; VCH: Weinheim, Germany, 1996. (28) Valentine, R. L.; Jafvert, C. T. Environ. Sci. Technol. 1988, 22, 691-696. (29) Knocke, W. R.; Conley, L.; Van Benschoten, J. E. Water Res. 1992, 26, 1515-1522. (30) Glaze, W. H. In Water Quality and Treatment, 4th ed.; Pontius, F. W., Ed.; McGraw-Hill: New York, 1990; pp 747-779. (31) Camel, V.; Bermond, A. Water Res. 1998, 32, 3208-3222. (32) Cowman, G. A.; Singer, P. C. Environ. Sci. Technol. 1996, 30, 16-24. (33) Pontius, F. W. J. Am. Water Works Assoc. 1998, 90, 38-50. (34) Vikesland, P. J.; Valentine, R. L. Environ. Sci. Technol. (in press).

Received for review December 29, 2000. Revised manuscript received November 7, 2001. Accepted November 12, 2001. ES002058J