Article pubs.acs.org/Langmuir
Sorption of Phthalic Acid at Goethite Surfaces under Flow-Through Conditions K. Hanna,*,†,‡ S. Martin,†,‡ F. Quilès,§,∥ and J-F. Boily⊥ †
Ecole Nationale Supérieure de Chimie de Rennes, UMR CNRS 6226, 11 Allée de Beaulieu, F-35708 Rennes Cedex 7, France Université Européenne de Bretagne, Rennes, France § Université de Lorraine, Laboratoire de Chimie Physique et Microbiologie pour l’Environnement, UMR 7564, Villers-lès-Nancy, F-54600, France ∥ CNRS, Laboratoire de Chimie Physique et Microbiologie pour l’Environnement, UMR 7564, Villers-lès-Nancy F-54600, France ⊥ Department of Chemistry, Umeå University, Umeå SE-901 87, Sweden ‡
S Supporting Information *
ABSTRACT: The objectives of this investigation were to improve our understanding of organic acid transport in porous media by focusing on a model system involving phthalic acid and goethite-coated sand (GCS). This was specifically made by first recalibrating a molecularly sound phthalate surface complexation model to GCS and then applying this model to describe breakthrough curves (BTC) in a GCS packed column. ATR-FTIR spectra of phthalic acid adsorbed at goethite surfaces at pH 3.0 and 6.0 and at loadings from 2.0 to 40.8 μmol/m2 confirmed the coexistence of metal-bonded (MB) and hydrogen-bonded (HB) complexes at low pH and the predominance of HB complexes at high pH. This concept was incorporated into a surface complexation model used to describe BTC at influent pH (pHin) values of 3.0, 6.0, and 7.8. The BTC revealed strongly pH-dependent behaviors. At pHin 3.0, the BTC revealed one front/plateau behavior while at pHin 6.0 two fronts/plateaus occurred. The existence of a second front/plateau led to an overestimation of the sorbed amount compared to that observed in the batch and caused a failure in the prediction of BTC. Additional column investigations suggested that surface loadings of nonspecifically adsorbed complexes could vary with pH and ionic strength and that the two-step breakthrough behavior may have emerged as a result of the formation of surface species of different natures than those during the first step, with the latter even serving as attachment sites corresponding to the second step. These findings call for refinements in current day modeling approaches used in reactive transport studies.
1. INTRODUCTION Interactions between naturally occurring organic acids and environmental particle surfaces, such as those of iron oxides, have attracted much attention in the literature due to their determining impact in natural processes as well as the complexity of the processes involved in the nature of (ad)sorption and dissolution reactions.1−6 Organic ligand adsorption to mineral surfaces can, in this context, affect the fate and transport of contaminants and nutrients as well as play a considerable role in mineral dissolution and biogeochemical redox processes.1 Furthermore, adsorption on iron oxide has been considered to be the main removal mechanism in ironbased treatment technologies (e.g., permeable reactive barrier).1,2 Spectroscopic and theoretical investigations2−6 improved the description of these interactions by identifying organic ligands that are (i) metal-bonded (MB; inner-sphere (IS)), (ii) hydrogen-bonded (HB; direct H-bond to surface (hydr)oxo group), or (iii) bound as outer-sphere (OS; separated by at least one hydration sheath) species, the relative importance of © 2014 American Chemical Society
which is affected by the pH, ionic strength, and mineral surface and ligand structures. Work on low-molecular-weight organic acids has been particularly beneficial along this front.2−6 Phthalic acid (PA; benzene-1,2-dicarboxylic acid) is one of such compounds commonly found in natural settings, and due to its ortho-positioned carboxylate groups, it can form strong seven-membered chelate complexes with dissolved and mineral surface metal species.5,6 The interfacial speciation of this and other related carboxylates promotes MB species under acidic conditions, coexisting with HB/OS which, in turn, can also form under circumneutral to alkaline conditions provided electrostatic interactions are favorable. PA adsorption on goethite (α-FeOOH) has been resolved along such lines and described with a thermodynamic adsorption model that can predict its pH-dependent surface speciation.5 The implementation of this level of knowledge in natural environments, Received: December 27, 2013 Revised: April 16, 2014 Published: May 20, 2014 6800
dx.doi.org/10.1021/la4049715 | Langmuir 2014, 30, 6800−6807
Langmuir
Article
(0.22 μm) and then analyzed by UV−vis spectrophotometry (Cary 5G UV−vis−NIR). In order to test for the impact of the presence of sand or a coating on the reactivity of goethite, kinetic tests were performed with goethite alone, GCS, as well as a sample containing a simple mixture of 1 wt % goethite with sand. When the goethite/solution ratio was kept constant, the kinetic sorption results for all of these systems were highly comparable, suggesting that the coating procedures or the presence of sand did not affect the goethite surface reactivity. Centrifuged wet pastes (10 000g for 60 min) of goethite suspensions equilibrated with PA were also analyzed by ATR-FTIR spectroscopy. They were uniformly applied to a diamond ATR cell. All experimental details and sample preparation for ATR-FTIR analysis are reported in the SI. Chemometric analyses of the spectra were carried out with the multivariate curve resolution-alternating least squares (MCR-ALS) method.18 This approach was specifically used to represent FTIR spectra in terms of spectral components of MB and HB species. 2.3. Breakthrough Column Experiments. Breakthrough column experiments were conducted according to Hanna et al.19 and are detailed further in the SI. Briefly, 72 g of dry GCS was packed into glass chromatographic columns of 2.6 cm internal diameter to give a porous bed length of 9 cm and a uniform density of GCS of 1.50 ± 0.01 g/cm3. The column was then wetted upward with a 0.01 M NaCl solution at a constant flow rate. Once the column became watersaturated, the flow characteristics of the porous bed were determined by a nonreactive tracer experiment. The classical convection dispersion equation (CDE) was used to describe the 1D transport of a nonreactive solute under steady-state water flow in a saturated column. The fit of the bromide elution curve provided estimations of (i) the volumetric water content (θ) which is in agreement with the value determined by weighing and (ii) the dispersion coefficient (D) that characterizes flow homogeneity. Values of θ, D, Darcy velocity (q), and water velocity (v) are given in the SI. The dispersivity λ was ∼130 μm, close to the particle grain size (100−150 μm). Furthermore, as the Péclet number (Pe = vL/D, where L is the bed length) was greater than 500, a convective regime predominated in the column. After water saturation, solution containing 1 mM PA was continuously injected in the column at the same constant flow rate under an O2- and CO2-depleted atmosphere. The flow-through experiments were duplicated for each goethite column. Two slower flow rates (0.1 and 0.5 mL/min) were also used to ensure greater column residence times. Solute and dissolved iron concentrations in the collected fractions were measured respectively by UV−visible spectrophotometry and inductively coupled plasma−atomic emission spectroscopy. More details on these experiments can be found in the SI. PA adsorbed amounts in the column were calculated by integrating the area above the curve, and the results were confirmed further by organic solvent extraction on the solid, as described in our previous work.19 2.4. Modeling. The time-resolved PA adsorption data were used to derive kinetic constants. Pseudo-first-order and second-order models as well as an intraparticle diffusion model were tested to obtain kinetic data. The most comprehensive results were obtained with the pseudofirst-order model whereby the relationship ln(Q/Q − Qt) = kt enabled the extraction of the kinetic constant k and where Q and Qt (μmol/ m2) are loadings at equilibrium and at time t, respectively. A thermodynamic adsorption model predicting batch PA adsorption data on synthetic goethite was developed from the original model of Boily et al.6 However, it uses the protonation model of Gaboriaud and Ehrhardt14 for which the basic charging properties of the goethite particles under study were developed and therefore different sets of constants for PA binding and interfacial capacitance values. Charges on the adsorbates were distributed among the 0 (H+, OH−, MB PA), 1 (HB PA), and 2 (Na+, Cl−) planes of the three-plane model (TPM). A system of equations representing mass action, mass balance, charge balance, and electrostatic equations was solved numerically using a trust region reflective algorithm and a gradient estimation obtained by finite differencing in the computational environment of Matlab (The Mathworks, Inc.). Equilibrium constants for phthalate adsorption were
however, requires the validation of such models to column flow-through systems emulating convective−dispersive flow. Previous studies reported important disparities in terms of sorption loadings under batch and column dynamic experiments, even when carried out with identical compounds and solid samples.7−9 Although many factors have been reported to cause discrepancies between the results of the two methods, the main reasons are still not well resolved and therefore call for further investigation. In this study, we implement a thermodynamic adsorption model predicting the molecular-scale speciation of PA for predicting transport in a flow-through column packed with goethite-coated sand (GCS). The interest in goethite (αFeOOH) lies in its great stability in low-temperature natural environments10 as well as its high reactivity and high specific surface area. Although it is commonly studied as aqueous suspensions in the laboratory, it is more commonly present as coatings for less reactive soil particles, such as silica sand,11 in nature. GCS is therefore a highly suitable packing medium for column studies and for porting knowledge to transport studies aimed at predicting contaminant transport in nature.12 Although PA transport in water-saturated hematite columns has already been studied,13 no attempt has yet been made to correlate its macroscopic transport attributes to molecular-level information. The present work thus combines information from timeresolved and equilibrium batch adsorption data as well as from vibrational spectroscopy to describe the breakthrough behavior of PA in a GCS-packed column. Experiments at different flow rates and column residence times are notably used to evaluate implications of nonequilibrium/kinetic processes in the breakthrough behavior and in relation to the interfacial speciation of phthalate. Breakthrough curves (BTC) were also determined at different influent pH values all under a low (i.e., 0.1 mL/min) flow rate to achieve local equilibrium. Predictions of breakthrough curves are then made using the adsorption models derived from new batch adsorption data on GCS. Implications of molecular-scale mechanisms involving adsorption under batch in contrast to convective−dispersive flow are then discussed in the latter part of this work.
2. MATERIALS AND METHODS 2.1. Preparation and Characterization of Solid Samples. All sample preparation and characterization procedures are detailed in the Supporting Information (SI). Briefly, goethite was prepared as described in Gaboriaud and Ehrhardt14 and coated onto sieved Fontainebleau quartz sand (100−150 μm) as previously detailed.15 All synthetic solids were washed to remove soluble Fe and electrolyte ions and stored in an anaerobic N2 (g) chamber at ambient temperature. The possible dissolution of pure Fontainebleau sand was evaluated under our experimental conditions, and all of the measurement tests confirmed that no silica was released from this sand surface. This result contrasts with low-particle-size (
