Environ. Sci. Technol. 2006, 40, 7488-7493
Modeling Iron Binding to Organic Matter† T I P H A I N E W E B E R , * ,‡,§ T H I E R R Y A L L A R D , ‡ EDWARD TIPPING,| AND M A R C F . B E N E D E T T I * ,§ Institut de Mine´ralogie et Physique des Milieux Condense´s, Universite´s Paris 6 et 7, UMR 7590 CNRS and IPGP, 140 rue de Lourmel, 75015 Paris, France, Laboratoire de Ge´ochimie des Eaux, Universite´ Paris 7, IPGP, UMR CNRS 7154, 2 place Jussieu, Case postale 7052, 75251 Paris, Cedex 05, France, and Centre for Ecology and Hydrology, Lancaster LA1 4AP, United Kingdom
The aim of the present work is to model iron speciation during its interaction with natural organic matter. Experimental data for iron speciation were achieved with an insolubilised humic acid used as an organic matter analogue for 30 µM to 1.8 mM total iron concentrations and 2 e pH e 5.5. IHA was found to be able to impose its redox potential to the solution and therefore the Fe(II)/Fe(III) ratio. Model VI and the NICA-Donnan model have been adjusted to experimental results of acid-base titrations, total iron measurements, and redox speciation in solution. They both describe well pH and concentration dependence of iron adsorption. For high iron concentration, Fe(III) solution activity is limited by precipitation of a poorly ordered Fe oxyhydroxide with a higher solubility (log Ks ) 5.6-5.7) than ferryhydrite described in the litterature.
elements in aquatic systems including iron itself. Furthermore, interactions with natural organic acids are known to modify the stability and surface reactivity of iron hydroxides, and iron nanooxides have been found to be intimately associated with organic macromolecules in natural environments (4). The interaction of iron with humic substances affects not only the metal but also the organic matter. Microbial iron reduction controls NOM degradation in both freshwaters and sediments, iron oxides interactions with NOM have a catalytic effect upon phenolic compounds polymerization (5), and iron binding may alter humic substance aggregation and adsorption onto surfaces. Moreover, in view of the high concentration of iron in natural systems, compared to those of trace metals, competitive binding of iron onto humic substances may affect the binding, and then the fate, of other metals. A predictive modeling of Fe(III) speciation involving water, mineral, and organic compounds is therefore needed for a full understanding of natural systems chemistry. Recent modeling efforts (6, 7) have provided generic Fe binding parameters toward NOM analogues (i.e., humic and fulvic acids). Nevertheless these models are calibrated with a limited database with poor statistics and which covers a limited pH range. Moreover two major aspects of NOM’s interaction with iron, i.e., its effect on iron oxides solubility and iron redox speciation, were not explicitly considered in these previous studies. One of the major difficulties in iron speciation studies and especially during Fe interaction with NOM is the determination of dissolved iron concentrations. Hence the objective of this work was to use an insolubilized humic acid as a NOM analogue. Model VI and the NICA-Donnan model were assessed by comparing the calculated and measured effects of pH and organic matters concentration on the speciation of iron.
Introduction
Experimental Section
Iron, a major component of the Earth’s crust, plays a dynamic and central role in all earth surface systems. Understanding its speciation and the various biogeochemical interactions of its different forms remains an important issue in environmental science (1). Iron species act as essential micronutrients for various biota including phytoplankton and as energy shunts in microbial organic matter reduction and oxidation (2). In addition the high surface area and reactivity of the iron nanooxide phases also control the adsorption and, thus, the fate of toxins such as trace metals, organic ligands, and vital nutrients such as phosphorus. Both dissolved and particulate iron species are known to significantly interact with natural organic matter (NOM) (1). Organic particles such as humics are ubiquitous in most natural environments (fresh waters, soils, sediments, ...), and their colloidal size and high chemical reactivity make them extremely important in controlling the mobility and bioavailability of micronutrients and pollutants (3). NOM provides alternate binding sites for nutrients and/or trace
Reagents. All solutions were prepared using milli-Q (Millipore) water. Humic acid was purchased from Aldrich (sodium salt, cat. H1,675-2, lot 31620-021, ash content 23.1%). Chemicals (Prolabo) were reagent grade or better, and total iron content was less than 10 ppm (except in iron salts). Methods. Aldrich humic acid was insolubilized by heating its calcium salt at 330 °C for 1 h according to the procedure of H. Seki and A. Suzuki (8). It was then purified with a 2:1 mixture of HF:HCl at 70 °C in order to destroy clay minerals and afterward in pure HCl for 2 h at 70 °C. This last step, allowing oxide dissolution, was repeated twice to completely remove all impurities. Absence of clays and iron oxides after the purification process was checked using electron paramagnetic resonance (EPR) spectroscopy (9) and Fourier transformed infrared spectroscopy (FTIR). Proton titration was performed using a computer controlled titrimeter (10) in a thermostated vessel (25 °C) under agitation. The pH was measured using a pH Metrohm glass electrode and a Metrohm calomel reference electrode. Titrations were carried out with suspensions of 2 g/L HA and 2.2 g/L IHA, respectively, in a 0.1 M KNO3 background electrolyte using CO2-free NaOH (0.1 M, titridose) and ultrapure HNO3 (0.14 M) under nitrogen bubbling. The pH electrodes were calibrated by both buffer calibration (pH 4.008, 6.986, and 9.993 Reagecon buffers) and by performing a blank titration of the background electrolyte prior to the titration experiments. After adition of acid or base, the rate of drift for both electrodes was measured over 1 min, and
† This article is part of the Modeling Natural Organic Matter Focus Group. * Address correspondence to either author. Phone: 33-(0)1-4427-57-05 (T.W. and M.F.B.); Fax: 33-(0)1-44-27-60-38 (T.W. and M.F.B.); E-mail:
[email protected] (T.W.);
[email protected] (M.F.B.). ‡ Universite ´ s Paris 6 et 7. § Universite ´ Paris 7. | Centre for Ecology and Hydrology.
7488
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 24, 2006
10.1021/es0607077 CCC: $33.50
2006 American Chemical Society Published on Web 10/19/2006
readings were accepted when the drift was less than 0.25 mV‚min-1. For each data point, the maximum time for monitoring the pH drift was 20 min. The extent of H+ binding to humic acids is calculated by substracting the volume of acid or base required to decrease or increase the pH of the blank electrolyte solution from the one needed for an equivalent volume of the humic acid’s suspension. Changes in the ionic strength during the titration were explicitly taken into account for each experimental data point. Since potentiometric titrations only allow the location of the relative position of the charge (Q) versus pH curves, humic acids were assumed to be neutralized at the lowest pH of the titration experiments for graphical representation of experimental data. Mixtures of 2 g/L of this insolubilized humic acid (IHA) were reacted with solutions of known total iron concentrations (1.8 × 10-3 M, 1.8 × 10-4 M, 5 × 10-5 M, and 3 × 10-5 M as Fe(NO3)3) in a 0.1 M KNO3 background electrolyte. The pH of the experiments, ranging from 2 to 5.5, was adjusted with known volumes of 1 and 0.1 M KOH solutions after a first 24 h equilibration step under horizontal shaking at room temperature. Samples were then equilibrated at their final pH with a horizontal shaker during 24 more h. The pH values of the samples were checked after the whole equilibration process. The solutions were then centrifuged during 30 min at 4000 rpm. The supernatant was removed and filtered onto 0.2 µm minisart-GF filter units to perform chemical analysis. Remaining IHA was also recovered and dried in an oven at 60 °C. The dried samples were used for spectroscopic analysis. The adsorption equilibrium has been assumed to be achieved in 24 h because 1 more week of contact between IHA and the iron solution at the final pH did not lead to any change neither in iron concentration in the supernatant nor in its redox speciation. Furthermore experiments were made with and without light shielded, and light has been shown to have no effect during the interaction between iron and IHA. Iron concentrations in the supernatants were determined after appropriate dilutions by graphite furnace atomic absorption spectrometry (GFAAS) at 302.1 nm, and the detection limit was 3 × 10-8 M. The total organic carbon (TOC) was determined using a Shimadzu TOC analyzer that utilized a high-temperature oxidation procedure with IR detection. The detection limit was 0.1 mg‚L-1, and the precision ranged from 5 to 10%. Redox speciation of iron in solution was determined using the ferrozine method adapted for small volumes of natural samples (11). This method allows sequential determination of Fe(II) and Fe(III) concentrations on a single aliquot by spectrophotometric analysis of the Fe(II)-ferrozine complex before and after reduction by hydroxylamine. It was chosen because of its low detection limit (10-7 M), the stability of the complex over a wide range of pH (12), and because it does not interfere even with a high level of natural organic matter (11). The solid phase (iron adsorbed onto IHA) was analyzed using FTIR measurements conducted in transmission mode in the 250-4000 cm-1 range on KBr pellets (about 1 mg of the sample, precisely weighted, pressed in 300 mg of dry KBr) using a Magna 560 Nicolet spectrometer. Redox Data Analysis. In solution iron can be found at the +II and +III oxidation states. The half-reaction involving the Fe3+/Fe2+ couple is written as
Fe3+ + e- ) Fe2+
(1)
and the equilibrium potential of a solution where the two species co-occur is given by the expression (13)
pE ) pE°Fe′′+/Fe2+ + log
aFe3+ aFe3+ ) 13.0 + log aFe2+ aFe2+
(2)
The only specie in our batch experiments able to impose a redox potential low enough to reduce Fe(III) is the IHA. NOM is known to be redox reactive, capable of reducing elements such as soluble Fe(III) species (2, 14). Humic acids are both proton and electron acceptors. Indeed, the proton dependent redox reactions of a humic acid can be described by the general reaction
Aoxx- + ne- + qH+ ) Ared(x+n-q)-
(3)
where Aox and Ared represent the oxidized and reduced form of the humic acid, respectively. This leads, for the solutions where a humic acid enforced the redox potential, to a conditional potential (14):
pEc ) log Kc - (q/n) pH
(4)
It has been hypothesized that quinone moieties in natural organic matter are the functional groups responsible for the electron-transfer reactions because they form semiquinone free radicals (2). For quinone moieties the q/n ratio equals 1 since the redox reaction can be written as
The positive correlation observed by Scott et al. (2) between Fe(III) reduction and free radical concentration using EPR spectroscopy offers additional support for this hypothesis. On the other hand, HA redox titrations by Struyk and Sposito (14) revealed a q/n ratio of 0.3 and are not supportive of HA quinones being the exclusive agent for electron transfer. Speciation Modeling. Around 5‰ of the IHA was dissolved during the experiment process. Quantities of dissolved humic acid (referred as DHA) were estimated as twice the TOC values expressed in gram per liter. DHA was supposed to have the same properties as IHA upon proton and iron binding. The iron speciation was simulated with two advanced models for ion binding to humic substances, e.g., model VI (15) and NICA-Donnan (16). These two models have proven their ability to successfully describe competitive ion binding onto natural organic matter in various environments although model VI assumes a discrete distribution of affinity constants and NICA a continuous one. Fittings have been performed using pH, total iron concentration, iron(II) concentration in the supernatant, IHA quantities, and DHA concentration as input data for parameters adjustment. In the pH range of our experiments, iron(III) aquo ions (Fe3+) and its first hydrolysis product (Fe(OH)2+) coexist in the solution, and the ligand exchange rate of (Fe(OH)2+) is very slow compared to its adsorption rate onto IHA (17). Thus this two iron species can compete with each other in addition to proton, for binding by humic substance. Moreover Tipping et al. (7) found that FeOH2+ is the principal form of iron bound to humic acid in the pH range 4-9. In addition, the solution activity of Fe3+ is limited by the precipitation of its oxy-hydroxides according to the reaction
Fe3+ + 3H2O ) Fe(OH)3 + 3H+
Ks-1
(6)
Oxyhydroxides freshly precipitated in our experiments should be poorly crystallized products, e.g., ferryhydrite, according to Schwertmann et al. (18). NICA-Donnan Model. Calculations were made using the ECOSAT code (19) that includes speciation with inorganic VOL. 40, NO. 24, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
7489
complexes and the NICA-Donnan model to describe humic substances complexation of proton and iron. The NICADonnan model has been extensively discussed elsewhere (16, 20). It accounts for competition between proton and metal ions, ion specific nonideality, variable stoichiometry, and electrostatic effects. First of all the Fe(III) complexation parameters can be adjusted with the two experiments where Fe(II) concentration is negligible. The four following complexation reactions are to be taken into account for iron(III) adsorption onto humic substances:
RCOO- + Fe3+ ) [RCOO-Fe(III)]2+ Φ-O- + Fe3+ ) [Φ-O-Fe(III)]2+
K ˜ 1,Fe K ˜ 2,Fe
(7)
[RCOO-Fe(III)(OH)]+ + H+ K ˜ 1,FeOH (9)
Φ-O- + Fe3+ + H2O )
K ˜ 2,FeOH (10)
The two first reactions represent Fe3+ complexation onto carboxylic-like and phenolic-like sites, respectively. The other two reactions account for the complexation of the hydrolysis species Fe(OH)2+. The adjusted parameters were then fixed, and only Fe(II) complexation parameters were let free for the others experiments modeling. Humic ion-binding model VI (7, 15) describes humic substances with 8 proton-dissociating groups, divided into two types (A and B), that can also provide monodentate, bidentate, and tridentate binding sites for metal ions. Metal binding is quantified with average intrinsic equilibrium constants (KMA, KMB) for each metal, with a parameter (∆LK1) that takes into account the variation of the constants around the mean and with a further parameter, ∆LK2, that characterizes the tendency of the metal to interact with “softer” ligand atoms (N and S). Electrostatic effects on ion binding are dealt with an empirical submodel that combines molecular charge and ionic strength. Model VI is implemented within the WHAM6 software (21).
Results Titrations of HA and IHA are shown in Figure S1 where the organic matter’s negative charge is plotted as a function of pH. The intrinsic affinity distribution (Figure S1) can be approximated by taking the first derivative of the proton adsorption isotherm (22). It reflects a bimodal distribution consisting of two major broad peaks centered around pK ) 4.5 and 8.0 characteristic of carboxylic (S1) and phenolic (S2) type of groups, respectively. In the NICA-Donnan model, each distribution is characterized by the peak position (median affinity K ˜ H) and the width of the distribution (m) that reflects the apparent heterogeneity of the humic acid (20). Since only one titration with a fixed ionic strength was made, we did not adjust the electrostatic parameters and applied those adjusted by Milne et al. (23) for both titrations. The model fits shown in Figure S1 are in good agreement with the experimental data. The parameters for the distribution are given in Table 1. The two log K ˜ H values (4.52 and 8.0) are consistent with the assignment of both types of groups. The ability of modeling both HA with a unique set of median affinity constants for both HA and IHA demonstrates that the insolubilization process does not affect much the reactivity of the humic substance. The values of m (
