Environ. Sci. Technol. 2009, 43, 337–342
Use of Multiparametric Techniques To Quantify the Effects of Naturally Occurring Ligands on the Kinetics of Fe(II) Oxidation PRESTON S. CRAIG,† TIMOTHY J. SHAW,† PENNEY L. MILLER,‡ PERRY J. PELLECHIA,† AND J O H N L . F E R R Y * ,† Department of Chemistry and Biochemistry, University of South Carolina, 631 Sumter Street, Columbia, South Carolina 29208, and Department of Chemistry, Rose-Hulman Institute of Technology, 5500 Wabash Avenue, Terre Haute, Indiana 47803
Received July 18, 2008. Revised manuscript received October 24, 2008. Accepted October 24, 2008.
A multifactorial experimental design was employed to quantify and rank the effects of a series of ligands on the rate of Fe(II) (18 µM) oxidation in a system containing chloride, sulfate, carbonate/bicarbonate, fluoride, and natural organic matter (NOM) at pH 8.34 ( 0.13. Several factors and combinations thereof correlated with the rate of Fe(II) oxidation at the 95% level of confidence. Presented in decreasing order of significance, those factors were carbonate/bicarbonate, NOM, sulfate, chloride, the sulfate/fluoride interaction, and fluoride. The center point of the experimental design was repeated with different organic matters substituted, including Nordic Reservoir NOM, fulvic and humic acids; Suwannee River NOM, fulvic and humic acids; and Pony Lake fulvic acid. Despite the widely differing geographical origins of these organic materials, their overall impact on the oxidation rate of Fe(II) was consistent, with the observed rate varying no more than a factor of 2 as a function of different organic matters (on a milligrams of carbon per liter basis). The utility of the pentafactorial response surface model (based on Nordic Lake NOM) to predict Fe(II) oxidation rates was evaluated for different natural water samples, including two seawater and one freshwater.
Introduction Submarine groundwater discharge is an important source of iron for estuarine and coastal mixing zones. Total [Fe] (particulate and dissolved) for groundwater in these regions can range as high as 180 µM with local estimated offshore fluxes up to 2 × 106 mol of Fe day-1 (1-4). Due to the anoxic conditions of these shallow coastal aquifers, reduced iron species [Fe(II)tot] predominate (1-4). Oxidation of aqueous Fe(II) by dissolved oxygen results in the production of several secondary oxidants that may also participate in Fe(II) oxidation, including superoxide, hydrogen peroxide, and the hydroxyl radical (5-12): Fe(II) + O2 h Fe(III) + O2-
(1)
* Corresponding author phone: 803-777-2646; fax: 803-777-9521; e-mail:
[email protected]. † University of South Carolina. ‡ Rose-Hulman Institute of Technology. 10.1021/es802005p CCC: $40.75
Published on Web 12/17/2008
2009 American Chemical Society
2H+
Fe(II) + O2- 798 H2O2 + Fe(III)
(2)
Fe(II) + H2O2 h Fe(III) + HO• + HO-
(3)
Fe(II) + HO• h Fe(III) + HO-
(4)
The rate of Fe(II)tot oxidation in these systems is a function of [O2], ligand, pH, temperature, and availability of radical scavengers (5, 6, 13, 14). Several commonly occurring inorganic anions play a role in moderating Fe(II) oxidation, including carbonate/bicarbonate (15-20), phosphate (5), perchlorate (5, 6, 21, 22), nitrate (5, 6), the halides (5, 6, 15, 16, 18, 19, 21-24), and sulfate (5, 6, 16, 22, 24). The magnitude of their influence on the overall rate of Fe(II) oxidation is a complex function of their association constants with Fe(II) (15, 16, 23, 25), their ability to scavenge HO · and superoxide (16, 17, 24, 26, 27), and the reactivity of their corresponding radicals (e.g., CO3- •) Fe(II) (7, 8, 16, 28, 29). The correlation between the effect of these anions and oxidation suggests that the ligand substitution process reaches equilibrium on a time scale that is rapid compared to Fe(II) oxidation (30, 31). Organic ligands, particularly organic acids, can also have a pronounced effect on the oxidation rate of Fe(II). Generally, complexation by carboxylates increases the rate of Fe(II) oxidation relative to the hexaaquo form [e.g., by a factor of 5 for salicylic acid (18), 2 for aspartic acid (32), and 2 for EDTA (27)]. Rose and Waite (25) recently reported that the majority of complex mixtures of organic ligands derived from the aqueous extracts of leaf litter, differentiated primarily by plant species of origin and leaf processing method, accelerated Fe(II) oxidation. There were marked differences across the family of extracts (differing in rate by nearly an order of magnitude). Although these experiments addressed mixtures of organic ligands, they did not address the effects of these ligands against a changing background of inorganic species, similar to the conditions found during the mixture of freshwater with seawater. In this work, we report the application of a five-factor, five-level Box-Wilson (central composite) experimental design to examine the relationship between initial Fe(II) oxidation rates (initially defined as ln ([Fe(II)]t/[Fe(II)]0) < 0.5) and several ligands that are known to influence Fe(II) oxidation in marine and freshwater systems. Our approach for describing the Fe(II) oxidation process in complex solutions was to employ this design to interrogate a parameter space incorporating the factors Cl-, SO42-, F-, HCO3(representing total carbonate species for the discussion), and natural organic matter (NOM). The ranges of each factor were determined by comparison against the National Assessment of Water Quality Database and the known characteristics of seawater (33, 34). The correlation between the rate of Fe(II) oxidation and the various factors (acting independently and cooperatively) was measured. The absolute quantification of the effects of each ligand or ligand-ligand interaction was also obtained and contextualized for the system (35-37). This multifactorial softmodeling strategy is based entirely on the discovery of correlations between factors and the experimental outcome (38). This is a significant departure from previous hardmodeling approaches that are based on describing the Fe(II) oxidation process according to known metal speciation. The soft-modeling strategy has several advantages. In contrast to the hard-modeling approach, it is sensitive to the presence of complexity, that is, the irreducible interaction of several VOL. 43, NO. 2, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
337
TABLE 1. Concentration Ranges for the Five Factors Used in the Experimental Design factor concentration levels coded factor levels x1 x2 x3 x4 x5
factor (units)
-2
-1
0
1
2
Cl- (mM) SO42- (mM) CO32-/HCO3- (mM) F- (µM) Nordic reservoir organic matter (mg of C/L)
0 0 0.206 0
316 16.4 1.41 38.9
546 28.2 2.28 67.4
776 40.1 3.14 95.8
1092 56.4 4.34 135
0
4.6
8.0
11.3 16.0
FIGURE 1. Fe(II) was rapidly oxidized under the experimental conditions bracketed in this study. Specific experimental conditions shown: pH ) 8.34 ( 0.13; [Cl-] ) 546 mM, [SO42-] ) 28.2 mM, [HCO3-] ) 2.28 mM, [F-] ) 67.4 µM, [NOM] ) 8.0 mg of C/L, n ) 6. factors on the experimental outcome. It is also faster and less expensive than hard models, since the approach can simultaneously measure the impact of many different factors. Most importantly, it enables discovery of new (unpredicted) factors by indicating their significance even if their association constants with Fe(II) are unknown (38).
Experimental Procedures Materials. Sodium chloride (99+%), sodium sulfate (99%), sodium fluoride (99%), sodium bicarbonate (99+%), sodium carbonate (99+%), iron(II) chloride (99%), ammonium acetate (99%), sodium hydroxide (97%), and concentrated hydrochloric acid (ACS grade) were acquired from Fisher Scientific. FerroZine iron reagent (98%) was purchased from Acros. All reagents were used as received. Nordic Reservoir natural organic matter (NOM), fulvic acid (FA), humic acid (HA); and Suwannee River NOM, FA, and HA were acquired from the International Humic Substances Society (IHSS) and used as received (see Table 1 in Supporting Information for elemental composition and Supporting Information for descriptions of the locations and isolation methods for the commercially available materials). Pony Lake FA was isolated from water collected in January 2006 from Pony Lake (77.87° S, 166.80° E), a coastal pond on Ross Island in Antarctica, by the XAD-8 resin method (39, 40). Pony Lake FA is unique in that it is completely autochthonous (no contribution from lignin) (41). All solutions were made in Barnstead E-pure (18 MΩ · cm) water. Experimental Design. The effects of the five factors and their interactions were determined via a circumscribed Box-Wilson experimental design (five-factor central composite with five concentration levels, Table 1). The factor concentration levels were set to bracket the ranges expected in the freshwater/saltwater mixing zone at pH 8.00 (33). The statistical validity of the design was ensured by performing 338
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 2, 2009
six replicate experiments at the center point and three replicate experiments at all other conditions. The overall design was a matrix containing a series of 43 experimental conditions in the defined parameter space, for a total of 132 experiments including replicates (precise conditions and results detailed in Table 2 in Supporting Information). The sequence of experiments was randomized to eliminate timedependent artifacts. Mean experimental pH was held constant at 8.34 ( 0.13 to reduce the effects of the changing activity of hydroxide on Fe(II) oxidation rates (7, 9, 16, 17, 20, 27, 29, 42-45). The correlation between the activity of HO- and kobs for these experiments (r 2 ) 0.234 due to the narrow pH range) is shown in Figure 1 in Supporting Information. The center point of this matrix was used as the basis of comparison for assaying the reactivity of Nordic Reservoir NOM against all other organic matters tested. Other organic matters were substituted for Nordic Reservoir NOM (identical mass of carbon per liter) at the point of comparison in the experimental design. Ionic strength did not correlate with kobs over the parameter ranges investigated in this experiment (Figure 2 in Supporting Information). Iron(II) Oxidation Experiments. A 9.00 mM stock solution of FeCl2 was prepared in boiled, degassed water (99.999% N2). This solution was then stored in the dark with a continuous slow N2 purge (99.999%; F
6.406
chloride sulfate bicarbonate fluoride NOM
β1 β2 β3 β4 β5
Single-Factor Terms 3.684 1.486 -6.916 1.486 30.59 1.486 -3.160 1.486 8.902 1.486
Cl-/SO42Cl-/HCO3Cl-/FCl-/NOM SO42-/HCO3SO42-/FSO42-/NOM HCO3-/FHCO3-/NOM F-/NOM
β12 β13 β14 β15 β23 β24 β25 β34 β35 β45
Interaction Terms -0.179 1.666 -0.0784 1.666 -1.190 1.666 -2.092 1.666 -2.362 1.666 -3.321 1.666 0.926 1.666 0.0643 1.666 -2.115 1.666 0.299 1.666
1.2 × 10-2 2.2 × 10-3 0.51 1.58 2.01 3.97 0.31 1.5 × 10-2 1.61 3.22
0.9145 0.9626 0.4767 0.2118 0.1591 0.0487b 0.5793 0.9693 0.2069 0.0754
Cl-/ClSO42-/SO42HCO3-/HCO3F-/FNOM/NOM
β11 β22 β33 β44 β55
1.623 -0.120 -4.157 -0.370 -3.947
Curvature Terms 1.659 1.659 1.659 1.659 1.659
0.96 5.3 × 10-3 6.27 5.0 × 10-2 5.66
0.3301 0.9423 0.0137b 0.8238 0.0191b
For coded factor levels from Table 1.
b
6.15 21.66 349.90 4.52 35.89
0.0147b
