Project Topics
Get Complete Project Material File(s) Now! »
Thiol implementation in PHREEQC-Model VI and the models used
A new model was developed to implement the thiol groups in PHREEQC-Model VI. PHREEQC-Model VI described humic substances as a set of discrete functional sites that can be divided into groups of weak and strong sites (Tipping, 1998). Weak sites are usually assumed to be carboxyl groups, whereas strong sites are generally assumed to consist of phenolic and N-containing sites. In the original Model VI, the binding of metals by humic substances occurs through eight discrete sites: four weak sites, named A sites and four strong sites, named B sites. In the present study, to implement the thiol group, we added four thiol groups, named S sites. The abundances of the type A, B and S sites are named nA, nB and nS (mol g-1), respectively. The intrinsic proton dissociation constants for the type A, B and S sites and their distribution terms are pKA, pKB, pKS, ΔpKA, ΔpKB and ΔpKS, respectively. The fractions of sites that can make bidentate sites and tridentate sites are named fB and fT and are equal to 0.5 and 0.065, respectively (Tipping, 1998). The abundances calculated for the 84 sites (monodentates, bidentates and tridentates) are given in Annexe II 1 and 2.
The proton association/dissociation equations and calculations of pK (equilibrium constant) for the 12 groups (carboxylic, phenolic and thiol) are described in the Annexe II 2. The protonation/deprotonation of the bidentates and tridentates are described as the decomposition of both protonation/deprotonation of the monodentates and the associated pK. For example, for the bidentate Ha_ab, the reaction and pK are: Ha_abH2 = Ha_abH- + H+ (Eq. 2).
H-HA parameters
The protonation/deprotonation parameters for the A, B and S sites are presented and compared to Tipping’s parameters (Tipping, 1998) for Model VI (Table I. 2). The variation in the carboxylic and phenolic abundances might be explained by the implementation of thiol groups (nS). For a same total abundance of site (dependent only on the type of humic substance), three different abundances (nA, nB and nS) are defined here versus two abundances (nA and nB) in Model VI. Figure I. 2 showed that the HA titrations performed in this study are within the range of the HA titration values compiled by Milne et al. (2001); notably, the obtained pKA and ΔpKA are close to those of Tipping’s parameters (Tipping, 1998). The most noticeable difference occurred for pKB which was explained by the fact that in Model VI, the thiol and phenol groups were grouped together and considered as site B.
Data from the literature showed that the pK of the phenol ligands seems to depend on the carbon radical to which the hydroxyl (OH) is bound. For example, the pK of hydroxybenzene, (OH- bound to one benzoic cycle) is 9.98 (at 25°C and IS = 0 mol L-1) versus 7.21 for nitrophenol (OH bound to one benzoic cycle, which is itself bound to NO2) (at 25°C and IS = 0 mol L-1). As the molecular structure of OM is complex and heterogeneous, the form/structure of the carbon radical of phenol cannot be identified. However, the fitted pKB of 7.11 obtained in this study was consistent with the pKA of phenol.
The H+ dissociation constant for the thiol groups, pKS = 5.82 ± 0.05, was lower than pKB, suggesting that thiol groups are more deprotonated at acidic pH than phenolic groups. For a simple organic ligand (aliphatic or aromatic) containing thiol groups, the pK varied from 5.2 to 13.24 (Figure I. 8). This pKA range correlates with the molecular weight of the molecules: pKA decreases with increasing molecular weights (Figure I. 8). Moreover, for aromatic molecules containing one thiobenzene, the increasing molecular weight of the radical associated with the aromatic ring is correlated with the decreasing pKA. Based on this dataset, the low pKS (5.82) obtained for the deprotonation of the thiol sites can therefore be justified by the high molecular weight and aromaticity of HA (Figure I. 8). The distribution term of pKS, ΔpKS, was high (6.12 ± 0.12), suggesting that the thiol pK were distributed over a large pK range. Humic acids are not only macromolecular but also supramolecular moieties (Pédrot et al., 2009), i.e. not only formed with high weight aromatic molecules but also lower weight molecules. The correlation between pK for the thiol group and the molecular weight of thiol-containing organic molecules might therefore explain this high distribution (ΔpKS). Several simulations were performed to test the influence of ΔpKS on the fitting of the As(III)- HA binding parameters. Variations in ΔpKS did not produce any variation for the As(III) concentrations bound to HA (Table I. 5).
As(III)-HA binding parameters
There is no consensus regarding the mechanisms involved in the binding of As(III) by HA. Buschmann et al. (2006) and Lenoble et al. (2015) proposed that this binding occurs through the complexation of As(III) by carboxylic and phenolic groups. These functional groups are the most abundant in OM and they are able to complex many cations such as Fe(II, III), REE, Al, Mg, etc. (Buffle et al., 1998; Milne et al., 2003; Sposito, 1986; Tipping and Hurley, 1992). However, the direct complexation of As(III) species by HA carboxylic groups has not been supported so far by any spectroscopic data. At any rate, the log K of As(III) binding to simple organic ligands is low – As(III)-catechol with log K = -6.89 and As(III)- pyrogallol with log K = -6.32 (Martell and Smith, 1989), indicating that this complexation, if any, should be of minor importance. By contrast, recent spectroscopic studies suggested two new binding mechanisms. The first one consists of an indirect mechanism in which As(III) is bound to OM via Fe (Buschmann et al., 2006; Hoffmann et al., 2013; Ko et al., 2004; Lin et al., 2004; Liu and Cai, 2010; Warwick et al., 2005). The second consists of a direct mechanism in which As(III) is bound to OM via thiol functional groups (Hoffmann et al., 2014; Langner et al., 2011b). Arsenic(III) has high affinity for S containing-ligands. The stoechiometry of the formed As-thiol organic molecules are either 1:1 (i.e. thiol in peat and HA, Hoffmann et al., 2012) or 1:3 (i.e. cystein Spuches et al., 2005a) depending on the ligand involved. In this study, two models were designed to test the reality of these complexes, i.e. the monodentate (1:1) model, the so-called Mono model, and the tridentate (1:3) model, the Tri model. Simulations of published datasets with the binding parameters established using the experimental data of this study demonstrated that the Mono model well reproduced more datasets than the Tri model (weighted RMSE = 0.86 and 1.22 for the Mono and Tri models, respectively). Considering the datasets of PAHA and AHA without H2AsO3 – species, the mean RMSE was lower with the Mono than the Tri model. The Mono model is in accordance with the binding mechanisms proposed by Hoffmann et al. (2012), i.e. the formation of monodentate complexes. Hoffmann et al.’s (2012) spectroscopic study demonstrated that only one S is located in the vicinity of As(III) in their HA sample (0.5 < CN (coordination number) < 1.5 at 2.29 – 2.34 Å). The fact that neither the Mono model nor the Tri model was successful in fitting these experimental data (Hoffmann et al., 2012) could be explained by the experimental conditions used by the authors. In order to meet the requirement for the spectroscopic analyses, high As(III) and HA concentrations had to be used in the experiments. These high amounts of As(III) could promote the formation of arsenite polymers. This hypothesis is supported by the presence of As in the vicinity of the bound As(III) in the EXAFS fitting of HA data by Hoffmann et al. (2012) (0.3 < CN < 0.5 at 2.63 – 2.67 Å). This model, which only considered As(III), therefore overestimated the bound As(III) concentrations. The same overestimation was obtained for Warwick et al.’s (2005) dataset. The experimental conditions of this dataset ([As(III)] = 2 – 42 mg L-1 and [HA] = 1.5 g L-1) were within the same range as those of Hoffmann et al. (2012) ([As(III)] = 4.1 mg L-1 and [peat] = 4.5 g L-1) and it is likely that As(III) polymers were formed during these experiments.
Moreover, the sorption isotherm of Warwick et al. (2005) exhibited two sorption increase/decrease steps (Figure I. 4c), a feature that could not be explained. The differences between the experimental and modeled data for the peat dataset of Hoffmann et al. (2012) can be explained by the nature of the peat used, as it is a specific OM formed in very specific conditions and this could thereby influence its composition and surface reactivity. The Mono model also failed to reproduce SRHA datasets (Kappeler, 2006; Lenoble et al., 2015). Since the occurrence of H2AsO3 – was expected for most data, only five points of both datasets could be used for fitting. The RMSE depended on the number of extrapolated points. For large datasets, the RMSE is expected to be lower than for small datasets. Moreover, a high discrepancy was observed between Kappeler (2006) and Lenoble et al. (2015) datasets.
Lenoble et al. (2015) showed that between 30 to 80% of As(III) was bound to SRHA versus 0.11 to 23.9% for Kappeler (2006) for equivalent experimental conditions (at pH = 8.4, DOC = 50 mg L-1, [As(III)]tot = 0.134 μmol L-1, As(III) bound = 8.87% and pH = 8, DOC = 15 mg L-1 and [As(III)]tot = 0.16 μmol L-1, As(III) bound = 38%, respectively). So far, we have no explanations for these observed differences. The RMSE were high for both Mono and Tri models. These RMSE corresponded to the average of the RMSE calculated for the 5 tested nS in the calculation of which Stot had the same values for each PAHA and each AHA samples, (%S = 2.33% and 4.2% for PAHA and AHA, respectively). However, considering the date of the various published studies, HA were probably provided from different lots. Moreover, for PAHA, the purification had probably modified Stot and thiol % in the HA sample.
Therefore, modeling calculations should only be considered as estimations and they had to be improved with the true Stot and thiol %.
Setup of the binding experiments
A standard batch equilibrium technique was used. Three series of Fe(II)-HA complexation experiments were conducted in triplicate. Firstly, the pH was monitored and kept constant during 24 h with a multi-parameter Consort C830 analyzer combined with an electrode from Bioblock Scientific (combined Mettler InLab electrode). Calibrations were performed with WTW standard solutions (pH = 4.01 and 7.00 at 25°C). The accuracy of the pH measurements is ± 0.05 pH units. An isotherm adsorption experiment was carried out relative to the increasing Fe(II) concentration (0.61 to 8.55 mg L-1). The average concentration of dissolved organic carbon (DOC) was 48.9 mg L-1. The pH was fixed at 5.9 with ultrapure HCl and NaOH. Secondly, a pH sorption edge experiment was performed over a pH range from 1.95 to 9.90 with DOC and Fe(II) concentrations of 48.8 and 3.03 mg L-1, respectively. Finally, a pH sorption-edge experiment was carried out over a pH range from 2.95 to 8.89 with DOC and Fe(II) concentrations of 76.5 mg L-1 and 3.15 mg L-1, respectively.
The [Fe(II)]tot, pH and DOC concentrations used in these experiments are representative of the concentrations that can be found in wetland waters (Dia et al., 2000; Olivie-Lauquet et al., 2001; Ponnamperuma, 1972; Reddy and Patrick, 1977). The ionic strength of all experiments was fixed at 0.05 M with NaCl electrolyte solution. Experimental solutions were stirred for 24 h to reach equilibrium. At equilibrium, 15 mL of solution was sampled and ultrafiltrated at 5 kDa (Vivaspin VS15RH12, Sartorius) by centrifugation at 2970 g for 30 min. under N2 atmosphere. Ultracentrifugation cells were previously washed with 0.15 N HCl and Milli-Q water to obtain a DOC concentration below 1 mg L-1 in the ultrafiltrate.
Chemical analyses
Dissolved organic carbon (DOC) concentrations were determined using an organic carbon analyzer (Shimadzu TOC-V CSH). The accuracy of the DOC measurements was estimated at ± 5% for all samples using a standard solution of potassium hydrogen phtalate. Iron concentrations were determined by ICP-MS using an Agilent Technologies 7700x atRennes 1 University. The samples were previously digested twice with 14.6 N HNO3 at 90°C, evaporated to complete dryness and then resolubilized with HNO3 at 0.37 mol L-1 to avoid any interferences with DOC during the analysis. ICP-MS analyses were carried out introducing He gas into collision cell to suppress any interference from Ar. The iron interference (40Ar16O/56Fe) was properly reduced by using He gas into collision cell to reach a low detection limit for Fe analysis (LD Fe: 0.07 μg L-1) (Instrumental and data acquisition parameters can be found in ANNEXE 1). Quantitative analyses were performed using a conventional external calibration procedure (7 external standard multi-element solutions – Inorganic Venture, USA). A mixed solution of rhodium-rhenium at a concentration level of 300 ppb was injected on-line with the sample in the nebulizer. This solution was used as an internal standard for all measured samples to correct instrumental drift and matrix effects.
Calibration curves were calculated from the intensity ratios between the internal standard and the analyzed elements. A SLRS-5 water standard was used to check the accuracy of the measurement procedure, and the instrumental error on the Fe analysis is < 5%. Chemical blanks of Fe were below the detection limit (0.07 μg L-1), and were thus negligible.
Table of contents :
Chapitre I : Complexation de l’arsenite par les groupements thiols de la matière organique
1 Introduction
2 Experimental, analytical and modeling methods
2.1 Reagents and materials
2.2 Experimental setup
2.3 Chemical analyses
2.4 Determination of the PHREEQC-Model VI binding parameters
2.4.1 Thiol implementation in PHREEQC-Model VI and the models used
2.4.2 Electrostatic model
2.4.3 Fitting the binding parameters
2.4.4 Dataset from the literature
3 Results
3.1 S(-II) grafting and titration
3.2 Adsorption isotherms
3.3 H-HA model
3.4 As-HA model
3.5 Simulations with the Mono and Tri models
4 Discussion
4.1 H-HA parameters
4.2 As(III)-HA binding parameters
4.3 Implications of the direct binding mechanism evidenced
Conclusion and perspectives
Chapitre II : Complexation du Fe(II) par les substances humiques : apport de la modélisation
1 Introduction
2 Experimental, analytical and modeling methods
2.1 Reagents and materials
2.2 Setup of the binding experiments
2.3 Chemical analyses
2.4 Determination of the PHREEQC-Model VI binding parameters
2.4.1 PHREEQC-Model VI
2.4.2 PHREEPLOT
2.5 LFER linear free energy relationship
3 Results
3.1 Experimental results
3.1.1 Adsorption isotherm
3.1.2 pH sorption edge
3.2 Model results
3.2.1 Adsorption isotherm
3.2.2 pH adsorption edge
3.3 Fe(II) speciation onto HA binding sites
3.3.1 Adsorption isotherm
3.3.2 pH adsorption edge
4 Discussion
4.1 Validation of the set of binding parameters
4.2 Fe(II) speciation onto HA binding sites
4.3 Environmental implications
Conclusions
Chapitre III : L’As(III) est-il capable de former des complexes ternaires avec la matière organique via Fe(II) et Fe(III) ionique?
1 Introduction
2 Experimental section
2.1 Experimental setup
2.2 Chemical analyses
2.3 Modeling
2.3.1 Model description
2.3.2 Binding parameters and modeling strategy
3 Results
3.1 As(III)-Fe(II)-AH experimental and modelling data
3.2 As(III)-Fe(III)-AH experimental and modeling data
4 Discussion
4.1 Monodentate or bidentate Fe(II)-HA sites: which ones complex As(III) most efficiently?
4.2 Instructions to better model As(III)-Fe(II)-HA interactions
4.3 Interpretation of As(III)-Fe(III)-AH data
4.4 Environmental implications
Conclusions et perspectives
1 Conclusions
1.1 La complexation directe d’As(III) par la MO est-elle possible?
1.2 La MO est-elle réellement un fort complexant du Fe(II)?
1.3 Est-il possible de former des complexes ternaires As(III)-Fe(II, III) ionique-MO?
1.4 Dynamique de l’As en solution dans les zones humides
2 Perspectives
2.1 Perspectives analytiques
2.2 Modèle de complexation à la surface des AH
2.3 Compétition ou formation d’autres complexes ternaires?
ANNEXES
