Adsorption Kinetics, Thermodynamics, and Isotherm Studies for Functionalized Lanthanide-Chelating Resins

Journal Pre-proofs Adsorption Kinetics, Thermodynamics, and Isotherm Studies for Functionalized Lanthanide-Chelating Resins Jonathan C. Callura, Kedar M. Perkins, John P. Baltrus, Newell R. Washburn, David A. Dzombak, Athanasios K. Karamalidis PII: DOI: Reference:

S0021-9797(19)31005-7 https://doi.org/10.1016/j.jcis.2019.08.097 YJCIS 25353

To appear in:

Journal of Colloid and Interface Science

Received Date: Revised Date: Accepted Date:

5 April 2019 6 August 2019 26 August 2019

Please cite this article as: J.C. Callura, K.M. Perkins, J.P. Baltrus, N.R. Washburn, D.A. Dzombak, A.K. Karamalidis, Adsorption Kinetics, Thermodynamics, and Isotherm Studies for Functionalized Lanthanide-Chelating Resins, Journal of Colloid and Interface Science (2019), doi: https://doi.org/10.1016/j.jcis.2019.08.097

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

© 2019 Published by Elsevier Inc.

Adsorption Kinetics, Thermodynamics, and Isotherm Studies for Functionalized LanthanideChelating Resins Jonathan C. Callura,a Kedar M. Perkins,b John P. Baltrus,c Newell R. Washburn,b David A. Dzombak,a and Athanasios K. Karamalidis*a,d a.

Carnegie Mellon University, Department of Civil and Environmental Engineering, Pittsburgh, PA, USA. b. Carnegie Mellon University, Department of Chemistry, Pittsburgh, PA, USA. c. U.S. DOE National Energy Technology Laboratory, Pittsburgh, PA, USA. d. Pennsylvania State University, Department of Energy and Mineral Engineering, University Park, PA 16802, USA. E-mail: [email protected]

ABSTRACT Conventional ion exchange resins are widely utilized to remove metals from aqueous solutions, but their limited selectivity precludes dilute ion extraction. This research investigated the adsorption performance of ligand-functionalized resins towards rare earth elements (REE). Functionalized resin particles were synthesized by grafting different ligands (diethylenetriaminepentaacetic dianhydride (DTPADA), phosphonoacetic acid (PAA), or N,N-bis(phosphonomethyl)glycine (BPG)) onto pre-aminated polymeric adsorbents (diameter ~0.6 mm). Lanthanide uptake trends were evaluated for the functionalized resins using batch adsorption experiments with a mixture of three REEs (Nd, Gd, and Ho at 0.1-1,000 mg/L each). Resin physical-chemical properties were determined by measuring their surface area, ligand concentrations, and acidity constants. The aminated supports contained 4.0 mmol/g primary amines, and ligand densities for the functionalized resins were 0.33 mmol/g (PAA), 0.22 mmol/g (BPG), and 0.42 mmol/g (DTPADA). Kinetic studies revealed that the functionalized resins followed pseudo-second order binding kinetics with rates limited by intraparticle diffusion. Capacity estimates for total REE adsorption based on Langmuir qMax were 0.12 mg/g (amine; ≈ 0.77 µmol/g), 5.0 mg/g (PAA; ≈ 32.16 µmol/g), 3.0 mg/g (BPG; ≈

1

19.30 µmol/g), and 2.9 mg/g (DTPADA; ≈ 18.65 µmol/g). Attaching ligands to the aminated resins greatly improved their REE binding strength and adsorption efficiency.

INTRODUCTION Rare earth elements (REE), a group of 15 metals comprised of lanthanides plus scandium and yttrium, have been identified by government agencies, internationally, as critical materials due to their importance to developing technologies as well as their supply chain risks [1,2]. Mineral ores remain the primary source of REE, though alternative sources and recycling could be used to augment supplies and mitigate price volatility [3]. International efforts to identify potential REE sources have accelerated dramatically since China, the world’s leading producer, began limiting exports around 2009 [4]. Many industrial wastes contain REE at trace levels [5–7], but other metals present at much higher concentrations make the separation and recovery of dissolved lanthanides from these complex mixtures challenging. There is a growing body of research devoted to the development of new materials and processes that could alter the economic viability of resource recovery from waste streams by improving upon the selectivity of conventional REE separation processes. Solid phase extraction (SPE) is an increasingly prevalent approach for recovery of lanthanides from fluids, though adsorption dynamics are highly dependent on solution chemistry and operating conditions [8]. Benefits of SPE over other mainstream REE extraction processes (e.g., liquidliquid extraction using solvents and ionic liquids) include the ability to produce high purity REE concentrates (>99.99%) in a single pass, process scalability, compatibility with continuous-flow operations (e.g., adsorption columns), and the ability to regenerate and reuse materials [9].

2

Researchers have generated SPE materials, which target metals using a variety of ligand-functionalized supports, including chitosan [10–16], silica [17,18,27,28,19–26], activated carbon [29,30], diatomite [31], magnetite [18,32], and β-cyclodextrin [33,34] and their extraction performance was summarized in recent reviews [8,35,36]. Most SPE studies with novel materials involve nano- or micro-scale materials. Nanomaterials have some desirable properties, such as their high surface area to volume ratio, but they are typically limited to batch applications and present problems with respect to process scaling. This research expands our previous work with ligand-functionalized adsorbent particles [21,37] by studying the performance of chemically modified resins (d ~0.6 mm) grafted with REE-selective ligands. Particles with large diameters, like those used in this study, facilitate continuous-flow processes, such as fixed-bed or fluidized-bed adsorption columns, though trade-offs are to be expected with regards to potentially slower uptake kinetics and lower adsorption capacity on an adsorbent mass basis. Before large-scale REE recovery systems can be developed, it is necessary to understand the performance trends and binding mechanisms of functionalized resins. Process design parameters are dependent on fundamental adsorbent properties, such as kinetics, capacity, thermodynamics, and selectivity. In this work, we experimentally investigated the former three property types. The overall goal for this work was to characterize the REE adsorption performance of aminated resins grafted with three different REE-chelating ligands. Specific objectives were to determine REE uptake kinetics, adsorption capacity, and thermodynamic parameters for lanthanide binding to large-scale functionalized resins. Experimental results were compared to modeling results based on speciation of the metals/ligands and solution phase ligandlanthanide complexation.

3

MATERIALS AND METHODS Functionalization A polystyrene-divinylbenzene copolymer resin particle (d ~0.6 mm) containing primary amine surface groups was selected for this study due to its favorable performance in screening tests, as well as its high porosity and demonstrated chemical/thermal stability. Information about the preparation methods for the aminated resins has not been made publicly available by the manufacturer at this time. Three ligands were employed for this study due to their documented affinity for REE: phosphonoacetic acid (PAA), N,Nbis(phosphonomethyl)glycine

(BPG),

and

diethylenetriaminepentaacetic

dianhydride

(DTPADA) [9,38,39]. The synthesis scheme used to produce these resins has been reported previously [21,37]. Aminated resins were dried overnight to remove moisture (T ~60°C), mixed with dimethylformamide (DMF), propylphosphonic anhydride (T3P), 4-(dimethylamino)pyridine (DMAP), and the target ligand (PAA, BPG, or DTPADA) in a one-pot reaction. After mixing for 24 hours at room temperature under inert atmosphere, the resins were transferred to centrifuge tubes and washed with DMF at room temperature, followed by hot water (T ~70°C), with excess fluid decanted between wash steps. Resins were then dried (T ~60°C) and subjected to an additional cleaning phase by mixing with 0.75 N trace metal grade nitric acid (HNO3) overnight, followed by rinsing with deionized water. This cleaning step was previously shown to improve and stabilize performance by washing away excess unbound reactants [21]. All functionalized resin particles were dried overnight (T ~60°C) prior to use in adsorption and characterization tests. Expected ligand configurations on the aminated resin surfaces and a summary of the functionalization process are presented in Figure 1. The anhydride rings of DTPADA

4

hydrolyze upon exposure to water [40], resulting in the form shown here, though the ‘DTPADA’ label is used throughout this paper to maintain consistency with our previous work. Likely structures for REE-DTPADA complexes on particle surfaces are shown in Figure S1 of the Supporting Information.

Figure 1. Summary of functionalization procedure and expected configurations of surfacetethered REE-chelating ligands shown with residual unreacted primary amines. This synthesis method was adapted from our previous work with functionalized silica particles [21,37]. (DMF = dimethylformamide; T3P = propylphosphonic anhydride; DMAP = 4(dimethylamino)pyridine; HNO3 = nitric acid) Some variability in the functionalized resins’ REE adsorption performance was observed for different small batches prepared for preliminary experiments. Larger batches of functionalized resins were synthesized (several grams) and treated consistently to minimize this batch-to-batch variation and ensure consistent results throughout this work. Additionally,

5

each batch was tested by preparing an adsorption pH edge (pH range ~ 2-8) to check for consistent adsorbent performance before further experimentation. Characterization XPS for Quantification of Ligand Loading X-ray photoelectron spectroscopy (XPS) was used for quantification of ligand loading. The XPS measurements were carried out with a PHI 5600ci instrument using a monochromatic Al Kα X-ray source. An electron flood gun was used to prevent sample charging during the analysis. The pass energy of the analyzer was 58.7 eV, the acquisition area had a diameter of ~800 µm, and the scan step size was 0.125 eV. Binding energies were corrected for charging by referencing to the C 1s peak at 284.8 eV. Atomic concentrations were calculated from the areas under individual high-resolution XPS spectra using manufacturer-provided sensitivity factors. These sensitivity factors are used to scale the measured peak areas for each elemental transition so that meaningful atomic concentrations can be obtained, and they reflect the specific geometries and fixed operating conditions inherent to the manufacturer’s instrument design. The C 1s and N 1s spectra were resolved into their individual components using CasaXPS software. Separate XPS analyses were carried out at two different spots on each of two sample adsorbent particles and the reported results are the averages of the values measured at those two spots. More details for data analysis on the XPS results are provided in the Supporting Information. BET for Determination of Specific Surface Area The Brunauer–Emmett–Teller (BET) surface area and Barrett-Joyner-Halenda (BJH) pore size were determined for the adsorbent particles using nitrogen adsorption data collected via a Micrometrics Gemini VII surface area analyzer. Prior to measurement, samples were degassed under vacuum for 24 hours at room temperature.

6

Adsorbent Titrations for Acid/Base Characteristics Acid-base adsorbent titrations were performed to study particle surface chemistry using a Metrohm 798 MPT Titrino autotitrator. Solutions containing 50 mL 0.5 M NaCl and 100 mg of resin particles were prepared and acidified with HCl prior to titration to ensure comparable pH starting points for each resin. Titrants were added to continuously-mixed solutions at variable rates (<100 µL/min) according to the slope of the pH vs. volume added curve, with the minimum flow rates administered in high slope regions. The base titration was performed with 0.1N NaOH in 0.5 M NaCl to maintain a constant ionic strength. After pH 12 was reached, the autotitrator was flushed with 0.1N HCl in 0.5 M NaCl and the titration was reversed by adding HCl until pH reached ~2. Slopes of the pH-volume curves were analyzed to determine equivalence points (identified by local slope maxima) and pKA value estimates (determined by finding the pH values at half equivalence points). Other Characterization Methods Images of the resin surfaces were collected using a scanning electron microscope (SEM) at 1000x, 2500x, 10000x, and 25000x magnification levels. Additional characterization methods were employed, but their results were inconclusive due to complications related to the relatively large particle diameters used in this study (d ~0.6 mm). Zeta-potential was investigated using electrophoretic mobility measurements, but the mass of the resin beads was too high to allow particle movement through the media. Attenuated total reflectance−Fourier transform infrared spectroscopy (ATR-FTIR) was used in our previous research to confirm ligand tethering onto smaller silica particles [37], but spectra for the functionalized resins in this study did not contain identifiable peaks in the expected ranges, likely due to scattering of the infrared beam by the large particles. Further XPS tests were also conducted to study the REE complexation mechanism, but REE detection limits were

7

elevated by interferences from other elements, preventing detailed analysis of the lanthanide coordination structures. REE Adsorption Experiments All samples for adsorption experiments were prepared and capped under ambient atmosphere without purging CO2 to study the performance of the resins under realistic operating conditions. The maximum concentration for dissolved CO2 in the samples is <0.65 mg/L (<1.5x10-5 M; assuming atmospheric concentration of 415 ppm and temperature of 20°C). For comparison, REE concentrations ranged from 0.1 to 1,000 mg/L (~1.93x10-6 M to 1.93x10-2 M), so the formation of REE-carbonates that would result from atmospheric CO2 dissolution is expected to be insignificant under these experimental conditions. Adsorption kinetics studies were carried out by reacting mixtures containing 0.5 M NaCl, 100 µg/L each of Nd, Gd, and Ho (300 µg/L total REE, prepared by dissolving Ln(NO3)3 salts), and 10 g/L functionalized resins at a fixed temperature (20°C) and pH value. These specific REE were selected to represent a range of molecular weights (i.e., Nd = light REE, Gd = middle REE, and Ho = heavy REE) and atomic sizes. High salt concentrations were selected to simulate the ionic strength of potential aqueous feedstocks to which these adsorbents could be applied (e.g., brines or various industrial wastes). Resins were preequilibrated for three hours with 0.5 M NaCl prior to reaction with REE-containing solutions to ensure particle hydration and focus on the kinetics of REE uptake. Reaction times ranged from 5 minutes to 4 hours. Solution phase rare earth element concentrations were determined by inductively coupled plasma mass spectrometry (ICP-MS) before and after reaction, and removal efficiencies were calculated by differences between these values. Thermodynamic properties of the resins were evaluated by conducting equilibrium adsorption experiments at varied temperatures. Batch samples were prepared with varied

8

concentrations of Nd, Gd, and Ho at equal mass ratios (0.3-300 mg/L total REE) in 0.5 M NaCl and 10 g/L resins at a fixed pH value. Samples were reacted for 24 hours by mixing at 300 revolutions per minute at three different temperatures (20°C, 60°C, 100°C). Reaction temperature was controlled by a heated stir plate with magnetic stir bars to agitate the samples, which maintained a constant sample temperature and mixing rate for the duration of each experiment. Blank samples (0.5 M NaCl + REE + magnetic stir bars) were prepared and analyzed with each set of experiments to ensure that the magnetic stir bars did not interact with REE in the samples. All other sample preparation and processing was identical to room temperature experiments. Equilibrium adsorption isotherm studies were conducted by preparing batch samples with varied REE concentrations in 0.5 M NaCl at fixed pH (6.5 for amine, PAA, BPG, or 2.5 for DTPADA) and fixed temperature (20°C, 60°C, or 100°C). Additionally, equilibrium adsorption pH edge studies were conducted at different REE loadings for BPG and DTPADA to determine the influence of adsorbate concentrations on pH trends. Individual batch samples containing a mixture of three REE in 0.5 M NaCl and 10 g/L resins were prepared, and their pH values were modified by adding either HCl or NaOH. Recovery of adsorbed REE and regeneration of the functionalized resins were studied through reuse experiments. Samples were prepared with 0.5 M NaCl, 100 µg/L each of Nd, Gd, and Ho (300 µg/L total REE), and 10 g/L functionalized resins at a fixed temperature (20°C) and a range of pH values (spanning from pH 1 to 8). Solution pH was modified prior to reaction by adding small volumes of either 0.5 N HCl or 0.5 N NaOH. Samples were mixed end-over-end (time = 24 hours; temperature = 20°C) before supernatant fluids were sampled to determine the remaining REE content. The equilibrium pH of each sample was recorded after each reaction. The remaining fluids were then decanted, and resins were rinsed with deionized water several times and dried overnight (time = 12 hours; temperature =

9

60°C). Dried resins were then allowed to cool to room temperature, spiked with 0.75 N HNO3 (using the same 10 g/L solids loading as the adsorption phase), and mixed end-overend (time = 24 hours; temperature = 20°C). The HNO3 solution was then sampled and analyzed to determine the mass of REE that was eluted and recovered. The resins were rinsed with deionized water multiple times, dried (time = 12 hours; temperature = 60°C), then used for another adsorption/elution cycle. This process was repeated for a total of three cycles. Modeling of REE Complexation by Solution-Phase Ligands Solution-phase complexation models were applied to evaluate differences in performance between free aqueous-phase ligands and surface-tethered ligands. Modeling of REE-DTPA and REE-PAA complexes in the solution-phase was performed using the U.S. Geological Survey’s PHREEQC program [41,42] implemented in the R programming language [43], as described in the Supporting Information. Surface ligand concentrations were determined by measuring surface concentrations from XPS analysis and equivalent solution-phase concentrations were calculated using the solids loading from adsorption experiments (i.e., surface ligand loading [mmol/g] * adsorbent loading [g/L]).

RESULTS AND DISCUSSION Characterization Ligand Loading, Surface Morphology, and Specific Surface Area Our previous work [37] with functionalized silica (d ~0.1 mm) reported ligand loadings of 0.96 mmol/g (amine), 0.15 mmol/g (PAA), 0.09 mmol/g (BPG), and 0.27 mmol/g (DTPADA). Surface-bound ligand concentrations for the polymer resins (Table 1; additional XPS data are provided in Table S1 of the Supporting Information) are higher than for the silica, despite the resins being ~6 times larger in diameter (d ~0.6 mm). This translates to ligand densities (µmol/m2) roughly two orders of magnitude higher for the polymer resins

10

than for silica and is likely due to the higher degree of amination on the resin surface and greater porosity of the resins, but may also be related to multilayer formation of ligands on the resin surfaces. Amine conversion efficiencies for functionalized resins (6-11%) were slightly lower than the functionalized silica (9-28%), so residual primary amines are expected to be present in addition to the surface-tethered ligands on both functionalized supports. Adsorption performance trends were similar for both functionalized supports (Figure S10 of the Supporting Information). Table 1. Summary of resin characterization results. Ligand loadings were calculated from XPS measurements, BET surface area and BJH pore volumes were determined from nitrogen adsorption tests. Ligand Surface Avg Pore Pore Volume Ligand 3 Loading Area Width (cm /kg) Density Ligand (mmol/g) (m2/g) (nm) Micro Meso Macro (µmol/m2) Amine PAA BPG DTPADA

4.00 0.33 0.22 0.42

14.8 12.7 8.0 11.4

24.5 27.7 30.8 33.4

0.67 0.22 0.14 0.17

6.9 8.6 3.7 3.8

68 74 56 99

281 26 28 37

Attachment of chelating ligands to primary amines on the resin particle surfaces resulted in reduced surface area for the PAA, BPG, and DTPADA functionalized resins. Existing research has also shown that functionalization may decrease particle surface area [25]. Table 1 shows higher average BJH pore sizes for PAA, DTPADA, and BPG functionalized materials relative to the aminated support, which is likely a result of the ligands occupying the smallest micropores and shifting the pore width distributions. This is further evidenced by pore volume distributions for each material, which show that the aminated resins have the highest volume of micropores (d < 2 nm) among the tested materials. Resin surface morphologies are shown in Figure 2. The image in Figure 2A shows the aminated support at low magnification (~2.5x); the other resins all have similar shape, size,

11

and appearance, so they are not shown here. The SEM images in Figure 2B do not show any apparent changes to surface morphology for the functionalized resins relative to the aminated support. All the resins in this study appear to be hard spheres to the naked eye, but surface roughness is visible at greater magnification levels.

(A)

(B)

Amine

BPG

PAA

DTPA DTPADA

1,000x 20 µm

20 µm

20 µm

20 µm

2,500x 10 µm

10 µm

10 µm

10 µm

2 µm

2 µm

2 µm

2 µm

10,000x

25,000x 1 µm

1 µm

1 µm

1 µm

Figure 2. Resin surface images. (A) Photograph of aminated support. (B) SEM images showing the aminated support and functionalized resins at 1000x, 2500x, 10000x, and 25000x magnification levels. Acid-Base Titrations Titration curves (Figure S18A in the Supporting Information) were analyzed to estimate equivalence points for each titration to assist with assessment of the presence of adsorbed

12

ligands. Curve slopes were plotted against pH (Figure S18B in the Supporting Information) to find peaks which indicate titration equivalence points. The corresponding titrant volumes were recorded to find the particle pKA values (pH at ½ equivalence point). Data were truncated between pH 3 and 12 to emphasize the region of interest. Ligand acidity constants were expected to change upon attachment to particle surfaces due to the formation of an amide bond between the carboxyl groups present on each ligand and the primary amines on the resins. It has been reported that substitution of carboxyl groups from DTPA with an amide group results in greater ligand acidity [44,45]. In all cases, the acidity constants for the functionalized resins were different than reference values for the untethered solution-phase ligands (Table S3 of the Supporting Information). Differences between acidity constants for acid vs. base titrations are attributed to hysteresis effects, which were most prevalent for the amine particles. The presence of hysteresis complicates the assessment of surface acidity behavior of the ligand-functionalized resins, but only approximate pKA values were sought to identify types of surface functional groups in this work. While primary amines should only have one pKA value, additional equivalence points suggest the presence of other background functional groups or potential electrostatic effects. In contrast, the BPG and DTPADA functionalized resins had fewer equivalence points than expected. This may be due to a shift of lower pKA values below the studied range (i.e., below pH 2). Since the PAA, BPG, and DTPADA functionalized resins are expected to also have residual primary amines on their surfaces, these values should be taken as an approximation of overall surface functionality. More accurate determination of surface-tethered ligand pKA values via potentiometric titrations would require the implementation of additional experimental controls, as detailed by Lützenkirchen, et al. [46].

13

REE Adsorption Results Kinetics Adsorption is measured as the mass of REE bound to the resins and increases with time until plateauing at the equilibrium value in well-mixed batch systems. Adsorbed phase concentrations (q) were calculated based on differences between initial (C0) and equilibrium solution phase concentrations (Ce), sample volume (V), and resin mass (M), as shown in Equation 1.

𝑞=

(𝐶0 ― 𝐶𝑒)𝑉

(1)

𝑀

These experiments were conducted at relatively low adsorbate concentrations (300 µg/L total REE) considering the expected REE uptake capacity for the functionalized resins. Uptake kinetics are expected to increase with increasing adsorbate concentrations due to sharper concentration gradients, so the rates shown here are conservative estimates. Adsorption kinetics data are commonly described and interpreted using a variety of numerical models. First- and second-order kinetic models (Figure 3A) are among the most routinely utilized in describing the time dependence of adsorption processes [36]. First- and pseudo-second-order kinetics are defined in Equations 2 and 3, where k1 is the first order rate constant, C is concentration at a time ‘t’, and C0 is the initial solution phase concentration. ―𝑑𝐶 𝑑𝑡

= 𝑘1𝐶 ⇒ 𝑙𝑛𝐶 = ― 𝑘1𝑡 + 𝑙𝑛𝐶0 ⇒ 𝐶 = 𝐶0𝑒 ― 𝑘1𝑡

―𝑑𝐶 𝑑𝑡

= 𝑘2𝐶2 ⇒

1 𝐶

=

1 𝐶0

+ 𝑘2𝑡 ⇒ 𝐶 = 1

1 𝐶0 + 𝑘2𝑡

(2)

(3)

While these models provide useful tools for assessing kinetic trends, they do not reveal information about the nature of rate-limiting steps. Mechanistic intraparticle diffusion models have been developed to describe processes of adsorption [47]. By plotting adsorption

14

data as a function of the square root of time and conducting a piecewise linear regression (Figure 3B), it becomes possible to observe linear regions associated with different stages of intraparticle diffusion of adsorbate [16,48–50]. Equation 4 defines the rate (kid,n) for each linear region, where the intercept (Cid) indicates boundary layer effects. The linear region with the lowest slope, and thus the highest intercept value, is the rate limiting step. The intraparticle diffusion rate constants were also normalized by their respective volume fractions (e.g., kid,1 [µg cm-3 min-1/2] = kid,1 [µg g-1 min-1/2] / macropore volume [cm3 g-1]) to compare the resins without the influence of varying pore volumes. 𝑞𝑡 = 𝑘𝑖𝑑,𝑛𝑡0.5 + 𝐶𝑖𝑑

(4)

Adsorption was most rapid for the BPG resins, followed by PAA, then DTPADA, as defined by their respective rate constants (Table 2). The aminated resins did not adsorb significant amounts of REE, due to the solution pH (5.3). Our previous work with silica supports [21] revealed that uptake onto the aminated particles was negligible below pH 8. These experiments targeted pH values below 8 to minimize the potential for precipitation of REE as hydroxide species which occurs in basic conditions.

15

Figure 3. (A) Kinetics data for adsorption of total REE (Nd, Gd, Ho) on functionalized resins, fit with first- and second-order models. (B) Weber-Morris plots for intraparticle diffusion modeling of the kinetics data. Relative standard deviations for each data point ranged from 0.2-4.8% with a mean of 1.3%. (C0 = 100 µg/L each Nd, Gd, Ho, 300 µg/L total REE; NaCl = 0.5 M; Resins = 10 g/L; Temperature = 20°C; pH = 5.3 (Amine), 5.4 (PAA), 5.6 (BPG), 2.3 (DTPADA)) Table 2. Kinetic modeling parameters for adsorption on functionalized resins Surface Ligand

First-Order Second-Order Intraparticle Diffusion (min-1) (g mg-1 min-1) (µg g-1 min-1/2) (µg cm-3 min-1/2) k1 R2 k2 R2 kid,1 kid,2 kid,3 kid,1 kid,2 kid,3 PAA 0.0128 0.83 0.31 0.96 4.71 1.65 0.23 63.65 191.9 1,045 BPG 0.0151 0.88 0.49 0.99 3.77 0.49 67.32 132.4 DTPADA 0.0088 0.90 0.10 0.99 3.73 2.00 0.40 37.68 526.3 2,353

Nash et al. [51] previously reported that the formation of REE-DTPA complexes in solution proceeds as a pseudo-first-order process with an observed rate constant of approximately 600-2,400 min-1, depending on pH and ligand concentration. While Nash et al. performed their experiments in lactate media, rather than sodium chloride, their reactions used a similar ionic strength (I = 0.3 M). Comparing these estimates with observed rate constants shown in Table 2 suggests that the complexation reaction is not the rate-limiting

16

step. First-order rate constants for REE complexation with solution phase DTPA are ~5 orders of magnitude higher than rates for the functionalized resins, though second-order rate models provided a better fit for the experimental data in this study. Previous research on adsorption mechanisms has shown that y-intercepts of approximately zero for the first linear region of Weber-Morris plots indicate minimal influence from boundary effects at the fluid-particle film interface [50]. The first y-intercept values from our experimental data (Figure 3B) are ~0, so the mixing rate used in these experiments was sufficiently high to negate any significant influence from film diffusion at the particle interface as a rate limiting step. Thus, intraparticle diffusion was the dominant factor in controlling the speed of REE adsorption in the batch system. Micropore diffusion rates were not discernable for the BPG resins, since only two linear regions were observed in the BPG Weber-Morris plot (Figure 3B). This finding is consistent with characterization results which showed that BPG had the lowest portion of micropores of the studied resins, with 80% less micropore volume than the aminated support. Thermodynamic Studies Data for thermodynamic properties such as Gibbs free energy, entropy, and enthalpy assist in the interpretation of adsorption mechanisms and binding strength. These parameters provide insight into the strength and type of bonds present in a chemical system and are derived from equilibrium adsorption isotherm data collected at fixed temperatures. Relationships between enthalpy (ΔH), entropy (ΔS) and Gibbs free energy (ΔG) at standard states are described in Equation 5. The equilibrium constant for adsorption (K0) can be found by taking the intercept of the plot for the natural log of the ratio of adsorbed concentration (q, mg/g) to equilibrium solution phase concentration (Ce, mg/L) vs. adsorbed concentration (i.e., ln(K0) = y-intercept of ln(q/Ce) vs. q) [52,53]. The reaction quotient, Q, is used to calculate

17

the change in Gibbs free energy (ΔG) from the standard state, where T is the reaction temperature (in Kelvin) and R is the gas constant (0.0083145 kJ/K-mol).

𝑙𝑛𝐾0 = ―

∆𝐻0 𝑅𝑇

+

∆𝑆0 𝑅

⇒ ∆𝐺0 = ―𝑅𝑇𝑙𝑛(𝐾0) ⇒ ∆𝐺 = ∆𝐺0 +𝑅𝑇𝑙𝑛𝑄

(5)

Data from thermodynamic tests with the functionalized adsorbents (Figures S14 and S15 of the Supporting Information) were truncated below q values of 1 mg/g to focus on the contribution from complexation of REE by surface-tethered ligands, rather than non-specific adsorption to other, weaker binding sites on the particle surfaces. This range of interest was determined from results presented in the Equilibrium Adsorption Isotherms section of this paper. Thermodynamic parameters were calculated from the truncated datasets to evaluate the influence of temperature on REE adsorption (Table 3). Higher K0 values indicate that adsorption onto the ligand-functionalized resins was more favorable than the unmodified aminated supports.

18

Table 3. Thermodynamic parameters for adsorption of REE mixtures (Nd, Gd, and Ho) from 0.5 M NaCl solutions onto functionalized resins. (pH = 6.5 for Amine, PAA, and BPG or 2.5 for DTPADA) Ligand Element T ln(K0) ΔG° ΔH° ΔS° (°C) Amine Nd (pH = 6.5) Gd

Ho

PAA Nd (pH = 6.5) Gd

Ho

BPG Nd (pH = 6.5) Gd

Ho

DTPADA Nd (pH = 2.5) Gd

Ho

20 60 100 20 60 100 20 60 100 20 60 100 20 60 100 20 60 100 20 60 100 20 60 100 20 60 100 20 60 100 20 60 100 20 60 100

(kJ/mol) (kJ/mol) (J/K-mol) -4.90 -5.28 -3.41 -4.60 -4.97 -3.30 -4.52 -4.59 -3.57 -0.35 -0.89 0.85 -0.72 -0.99 0.53 -0.63 -1.04 0.73 3.17 2.18 1.02 3.46 1.79 2.14 2.48 2.02 1.83 -0.37 -2.17 -2.18 0.29 0.17 -0.70 0.08 0.16 -1.55

11.95 14.63 10.57 11.21 13.76 10.23 11.01 12.72 11.08 0.85 2.46 -2.65 1.76 2.75 -1.66 1.55 2.87 -2.27 -7.74 -6.03 -3.17 -8.44 -4.96 -6.65 -6.05 -5.59 -5.67 0.91 6.00 6.77 -0.70 -0.47 2.16 -0.21 -0.43 4.80

15.90

10.52

13.84

6.30

10.18

-4.28

12.58

37.06

13.41

37.36

14.49

41.32

-24.28

-55.91

-15.86

-27.57

-7.57

-5.41

-21.30

-77.65

-10.79

-33.37

-17.66

-57.15

19

While thermodynamic evaluations of solution-phase lanthanide complexation with the phosphonate ligands tested here are scarce in the literature, DTPA complexes have been extensively studied by other researchers [54–56]. Gibbs free energy for solution-phase REEDTPA complex formation is approximately -125 kJ/mol, while enthalpy is -30 to -50 kJ/mol, and entropy is 250-300 J/K-mol. Negative values for ΔG0 in Table 3 indicate that adsorption is thermodynamically favorable and spontaneous for the BPG and DTPADA functionalized resins. Chelation of lanthanides onto the DTPADA and BPG adsorbents was slightly diminished with the addition of heat, and the enthalpy terms were negative, so adsorption was exothermic. In contrast, adsorption was endothermic for the aminated and PAAfunctionalized resins. Equilibrium Adsorption Isotherms Equilibrium adsorption isotherms were conducted to provide insight into REE binding capacity of the functionalized resins and were fit with Freundlich and Langmuir models (Equations 6 and 7; linearized fitting for the Freundlich and Langmuir models is shown in Figure S8 of the Supporting Information). Relevant parameters for these models include adsorbed REE concentration (q), equilibrium solution-phase REE concentration (Ce), Freundlich constant (KF) which indicates the overall binding strength of an adsorbent, a linearity term (n) which describes the uniformity of binding energies for surface groups, Langmuir constant (KL), and maximum adsorption capacity (qMax). Freundlich:

𝑞 = 𝐾𝐹𝐶𝑒

Langmuir:

𝑞=

1𝑛

1

⇒ log (𝑞) = log (𝐾𝐹) + 𝑛log (𝐶𝑒) (6)

𝑞𝑀𝑎𝑥𝐾𝐿𝐶𝑒 1 + 𝐾𝐿𝐶𝑒

⇒

1 𝑞

1

1

1

= 𝑞𝑀𝑎𝑥𝐾𝐿𝐶𝑒 + 𝑞𝑀𝑎𝑥

(7)

Adsorption isotherms (Figure 4), indicate the relative affinity of each functionalized resin towards REE. The upward trending curves at higher adsorbate loadings in Figure 4A

20

suggest a mixture of strong (ligands) and weak binding sites (primary amines or other surface groups). In the upper concentration ranges, it is likely that the REE-selective ligand binding sites have been saturated, and additional uptake is the result of non-selective processes such as multilayer adsorption, surface precipitation, or binding to residual amines or other surface groups. The absence of a plateau in the amine plot supports this hypothesis, as the aminated resins contain a more uniform distribution of relatively weak binding sites. Therefore, the reasonable operating capacity for these resins is estimated to be <30 mg total REE per gram of resin (this corresponds to fluid concentrations of 300 mg/L total REE in these batch conditions). Freundlich and Langmuir parameters were initially calculated across the full range of concentrations for each functionalized resin, but the fit of each model to experimental adsorption data was unsatisfactory due to the curved shape of the isotherm profiles for PAA, BPG, and DTPADA resins. The Freundlich model provided an accurate description of adsorption onto the aminated supports, but the ligand-functionalized resins appear to contain two classes of binding sites: those with strong affinity for REE (i.e., PAA, BPG, and DTPA) and those with weak affinity for REE (i.e., primary amines, etc.). Therefore, a multi-site model approach was taken to account for both site types. Langmuir parameters (qMax and KL) were calculated from experimental data for the ligand-functionalized resins at lower REE loadings (C0 < 300 mg/L total REE), while Freundlich parameters (1/n and KF) were calculated from the full set of isotherm data from the aminated supports (Table 4). These two models were then added together to calculate the total REE uptake from surface ligands and the other functional groups on the supports for each resin (Figure 4A). Langmuir KL values can be used as an indicator for binding strength, so we conclude from the data in Table 4 that the relative REE affinity of the functionalized resins is much higher than the Amine support.

21

Figure 4. Isotherm plots for equilibrium adsorption of total REE (Nd, Gd, and Ho spiked at equal concentrations) on functionalized resins. (A) Log-log plot overlaid with isotherm models (Strong Sites = Langmuir model for ligand binding; Weak Sites = Freundlich model for uptake by aminated support; Combined = Strong Sites + Weak Sites). (B) Semi-log plot of total REE removal efficiency vs. spiked concentration. Gray shaded region shows approximate site saturation based on ligand loading for PAA, BPG, and DTPADA. (NaCl = 0.5 M; Resins = 10 g/L; pH = 6.5 (Amine, PAA, BPG), 2.5 (DTPADA); Temperature = 20°C; Time = 24 hours; values are shown as mean ± one standard deviation; results for individual elements in the mixture are shown in Figure S9 of the Supporting Information) Table 4. Isotherm model parameters for REE adsorption (mixture of Nd, Gd, and Ho in 0.5 M NaCl) on aminated support and functionalized resins.

Ligand

Amine PAA BPG DTPADA

Experimental pH 6.5 6.5 6.5 2.5

Freundlich (Weak Sites) 1/n 0.9 0.9 0.9 0.9

Langmuir (Strong Sites)

KF (L/g)n qMax (mg/g) KL (L/g) 0.01 0.01 0.01 0.01

5.00 3.01 2.92

0.05 4.66 0.44

22

Comparison of REE Complexation by Surface-Tethered or Solution-Phase Ligands Complexation models for solution-phase ligands were applied to evaluate differences between REE complexation by free aqueous-phase ligands and surface-tethered ligands. Modeling results suggest saturation of the ligands at loadings of ~70 mg total REE per gram of DTPADA resin or ~35 mg total REE per gram of PAA resin (Figure S7 of the Supporting Information). Reasons for differences between these ligands are twofold: ligand loading was ~30% higher for DTPADA than PAA on a molar basis, and nearly one-third of PAAlanthanide complexes take the form of 2:1 ligand:lanthanide chelates in solution at pH 6.5, while 1:1 complexes are the dominant form in DTPA-lanthanide systems [38]. Once tethered to the particle surfaces, PAA was expected to experience a larger drop in performance as a result of its reduced flexibility to form these two-ligand structures. Chemical equilibrium modeling of REE complexation by free ligands in solution was compared to experimental results for REE adsorption on the functionalized resins (Figure S12 of the Supporting Information). The chelation of REE by surface-bound DTPADA is approximated with reasonable accuracy by solution-phase models in lower concentration ranges (R2 = 0.99 below 30 mg/L total REE). At higher concentrations though, experimental results diverge from the solution-phase model, which suggests either an overestimation of ligand grafting onto the resins or a decreased REE affinity by the surface-bound ligand compared to free, untethered ligands in solution. Langmuir qMax values for PAA and DTPADA (Table 4) are approximately one order of magnitude lower than the solution-phase models predict for REE-ligand complexation, though in all cases the functionalized resins performed favorably when compared to the aminated particle surfaces. By translating the maximum adsorption capacity from the Langmuir models to ligand density and assuming 1:1 molar binding between REE and surface-tethered DTPADA, we

23

estimate the surface concentration of DTPADA to be ~ 1.9x10-5 mol/g, rather than the 4.2x10-4 mol/g measured by XPS. Paul-Roth and Raymond [45] previously found that amidation of two of DTPA’s carboxyl branches diminished its complexation with Gd3+ (KGdDTPA

dropped from 1022.46 to 1016.85), though much of the selectivity for Gd3+ over Ca2+ was

retained. Similarly, Sherry et al.[44] showed that amidation of one carboxyl group reduced the stability constants for Gd-DTPA from 1022.26 to 1019.68. A similar amide bond is formed between DTPADA and primary amines on the resin surfaces during functionalization, so it is reasonable to expect a similar decrease in capacity compared to non-substituted DTPA in the solution phase. Solution models for PAA also provided a useful benchmark for evaluating observed surface PAA concentrations. Experimental adsorption results again indicated REE adsorption lower than for REE complexation in solution with an equivalent molar ligand concentration. Unlike DTPA, however, PAA does not exclusively form 1:1 complexes with REE in solution. The reduced mobility of the surface-tethered PAA ligands likely plays a role in their diminished binding capacity compared to solution-phase ligands, as formation of 2:1 complexes (i.e., 2 PAA molecules bound to one REE molecule) may be hindered by ligand spacing on the resin surfaces. Using the same approach as for DTPADA, the Langmuir qMax value for PAA can be converted to an effective ligand concentration of 4.6x10-6 mol/g, rather than the 3.3x10-4 mol/g from XPS measurements.

Effect of pH on REE Binding Capacity of the Ligand-Functionalized Adsorbents

Solution pH is a master variable for the adsorption and recovery of REE by the functionalized resins, affecting both the adsorption capacity and complexation efficiency. Silica adsorbents functionalized with BPG and DTPADA and treated with HNO3 were

24

previously shown to have optimal uptake at pH >3 and pH 2-3, respectively [21]. However, these tests were conducted at low adsorbate concentrations (300 µg/L total REE). Repo et al. [13] found that adsorption pH trends may change as a function of adsorbate concentrations in their study of chitosan functionalized with ethylenediaminetetraacetic acid (EDTA) and DTPA. This was attributed to electrostatic interference between negatively charged ligands and positively charged amines on the supports at pH 3-5, which was prevented or disrupted at higher metal concentrations. A similar result (adsorption only in low pH at low adsorbate loadings) was also observed for DTPA-functionalized silica in our previous work [37]. Figure 5A shows that adsorption pH trends for DTPADA reverse at total REE concentrations of 30 mg/L. While acidic conditions favor uptake on DTPADA-functionalized resins at low REE concentrations, adsorption was observed in the circumneutral range at 30 mg/L REE. This concentration corresponds approximately to the maximum potential loading for these resins (qMax from the Langmuir isotherm model). The BPG resin did not exhibit any notable trend reversal with increasing concentrations (Figure 5A). This result is consistent with the findings of Repo et al. [13] who showed that DTPA, which is a relatively large molecule (compared to both EDTA in their study and BPG in ours), is more likely to undergo crosslinking due to its long, flexible branches. Possible structures for crosslinked DTPA resins are shown in Figure S16 of the Supporting Information. The isotherm data presented in Figure 5B confirm that binding of lanthanides onto the DTPADA resins in circumneutral conditions is minimal at low adsorbate loadings but increases at higher concentrations to converge with the low pH data series above 30 mg/L total spiked REE. While the data presented here show total REE concentrations, it is also interesting to note that the individual element trends change as a function of pH. At pH 2-3,

25

total REE uptake at the highest spiked concentration was 4.4 mg/g with selectivity towards heavy REE (qNd = 0.7 mg/g, qGd = 1.4 mg/g, qHo = 2.2 mg/g), while pH 6-7 resulted in a maximum total uptake of 5.5 mg/g with selectivity towards lighter REE (qNd = 2.5 mg/g, qGd = 1.6 mg/g, qHo = 1.3 mg/g). This translates to separation factors (defined by Equation 8) of αHo/Nd = 3.64 at pH 2.5 and αNd/Ho = 2.26 at pH 6.5. Solution phase DTPA is more selective towards heavy REE, with αHo/Nd ~ 15, and this selectivity is not dependent on pH (determined by comparing stability constants for each REE-DTPA complex from Grimes and Nash [39]). Individual element data that show the intra-REE trends are shown in Figure S13 of the Supporting Information.

𝛼𝐴/𝐵 =

[(𝐶0 ― 𝐶𝑒)/𝐶𝑒]𝐴 [(𝐶0 ― 𝐶𝑒)/𝐶𝑒]𝐵

(8)

Figure 5. (A) Adsorption pH edges with varied adsorbate concentrations. Initial concentrations represent total REE (Nd, Gd, Ho spiked at equal concentrations on mass basis). Dashed lines indicate 100% removal of spiked REE. (B) Isotherm results for DTPADA resins at different fixed pH values. Results are shown as removal efficiency vs. total spiked concentration; gray shaded region indicated expected saturation of surface

26

DTPADA. (NaCl = 0.5 M; Resins = 10 g/L; Temperature = 20°C; Time = 24 hours; values are shown as mean ± one standard deviation) Regeneration and Reuse of Functionalized Resins Durability is an important resin property, as the materials would likely undergo multiple use cycles in a practical REE extraction scheme. Our previous work with silica particles that were functionalized using a similar protocol as described here revealed that material performance changes during the initial use cycles [21]. Figure 6 shows the total REE (sum of Nd, Gd, and Ho, which were spiked as a mixture at 100 µg/L each) adsorption and elution results for each resin across a range of pH conditions (results for each individual REE in the mixture are shown in Figure S17 of the Supporting Information). Adsorbed REE were eluted and recovered from resin surfaces using 0.75 N HNO3 (pH <0.13). These results show that each resin exhibits unique pH trends. Aminated supports did not yield significant REE adsorption below pH 6, and the uptake at pH 7 and above is not expected to be selective, as explained previously in the Equilibrium Adsorption Isotherms section. PAA resins yielded uptake in circumneutral conditions; the trends appear similar to the aminated support in these conditions (i.e., only NaCl and REE, without additional competing ions), but the isotherm trends discussed previously suggest that the presence of ligands results in strong, preferential binding of REE. Both BPG and DTPADA demonstrated ability to adsorb REE at pH ≤4. DTPADA performance in acidic conditions improved after the first use cycle. The changing patterns across each pH edge are likely the result of cleaning off residual reactants (e.g., unbound excess ligands) left over from the resin functionalization process. Each of these trends is consistent with results that we previously reported for functionalized silica particles [21], evidence which suggests that the uptake trends are related to the attached ligands, rather than the aminated supports. The functionalized resins were

27

subjected to wide pH swings, yet they retained the ability to adsorb REE, which indicates durability of the ligand attachment.

Figure 6. Performance of aminated support and functionalized resins through three use cycles. Resins were equilibrated with NaCl solutions containing 300 µg/L total REE (100 µg/L each Nd, Gd, and Ho) at fixed pH. The pH conditions for each adsorption phase are shown on the x-axis, while total REE mass (µg-REE/g-resin) is shown on the y-axis. The total REE mass spiked for each series is indicated by a dashed line. Black diamonds indicate REE adsorption and blue squares indicate REE recovery upon elution with 0.75 N HNO3. (NaCl = 0.5 M; Resins = 10 g/L; Temperature = 20°C; Time = 24 hours)

28

SUMMARY AND CONCLUSIONS The goal of this research was to evaluate the physical/chemical properties and REE adsorption performance for ligand-modified resins. Batch adsorption experiments were conducted to determine kinetic, thermodynamic, and isotherm parameters for REE adsorption on functionalized resins. Rate constants were developed for first-order, pseudo-second-order, and intraparticle diffusion models. Results from these tests and analyses indicate that uptake of lanthanides by the functionalized resins is kinetically limited by intraparticle diffusion. Adsorption capacities were determined through isotherm tests, and uptake/selectivity trends with respect to pH were found to be dependent on adsorbate concentrations. Isotherm results provided evidence for the presence of a mixture of different binding sites on each functionalized resin, with attachment of REE-selective ligands to resin surfaces resulting in improved REE extraction. Characterization tests provided estimates for ligand loading and particle surface area for each resin. The functionalized resins retained their REE-binding ability after multiple use cycles, despite exposure to acids and a wide range of pH conditions. These resins show promise for separations processes due to their affinity for binding REE, though additional work is required to improve ligand grafting efficiency and better understand their REE selectivity and long-term durability. While some degree of intralanthanide selectivity was observed at higher concentrations, the primary utility of the functionalized resins may be in group-wise separation of REE from bulk ions in solutions. These results provide new insight into the binding properties for lanthanides onto surfacetethered chelating ligands. The larger particle sizes used in this research compared to many other studies could facilitate the scale-up and field implementation of the functionalized resins in resource recovery systems targeting REE.

29

CONFLICTS OF INTEREST Portions of this study are included in pending patent US 20170101698 A1.

ACKNOWLEDGEMENTS The authors thank Qingyang Shi for her assistance with sample preparation, Aditya Menon for conducting portions of the BET analysis, and Clint Noack for providing the basis of the code used in speciation modeling. This work was funded in part by the U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy (contract DE-EE0006749). J. Callura was also supported by the National Science Foundation Integrative Graduate Education and Research Traineeship in Nanotechnology-Environmental Effects and Policy (IGERT NEEP) fellowship program (DGE0966227) and a CMU College of Engineering Dean’s Fellowship. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the NSF. D. Dzombak was supported by the Hamerschlag University Professorship.

REFERENCES [1]

B.S. Van Gosen, P.L. Verplanck, R.R. Seal II, K.R. Long, J. Gambogi, Rare-Earth Elements, Chap. O of Critical Mineral Resources of the United States-Economic and Environmental Geology and Prospects for Future Supply, Reston, VA, 2017. doi:10.3133/pp1802O.

[2]

United States Department of Energy, U.S. Department of Energy Critical Materials Strategy, 2010. http://energy.gov/sites/prod/files/piprod/documents/cms_dec_17_full_web.pdf.

[3]

A. Kumari, R. Panda, J. Rajesh Kumar, K. Yoo, J.Y. Lee, Review on

30

Hydrometallurgical Recovery of Rare Earth Metals, Hydrometallurgy. 165 (2016) 2– 26. doi:10.1016/j.hydromet.2016.01.035. [4]

H. Paulick, E. Machacek, The Global Rare Earth Element Exploration Boom: an Analysis of Resources Outside of China and Discussion of Development Perspectives, Resour. Policy. 52 (2017) 134–153. doi:10.1016/j.resourpol.2017.02.002.

[5]

B.W. Stewart, R.C. Capo, B.C. Hedin, R.S. Hedin, Rare Earth Element Resources in Coal Mine Drainage and Treatment Precipitates in the Appalachian Basin, USA, Int. J. Coal Geol. 169 (2017) 28–39. doi:10.1016/j.coal.2016.11.002.

[6]

Y. Liu, R. Naidu, Hidden Values in Bauxite Residue (Red Mud): Recovery of Metals, Waste Manag. 34 (2014) 2662–2673. doi:10.1016/j.wasman.2014.09.003.

[7]

R.K. Taggart, J.C. Hower, G.S. Dwyer, H. Hsu-Kim, Trends in the Rare Earth Element Content of U.S.-Based Coal Combustion Fly Ashes, Environ. Sci. Technol. 50 (2016) 5919–5926. doi:10.1021/acs.est.6b00085.

[8]

S. Iftekhar, D.L. Ramasamy, V. Srivastava, M.B. Asif, M. Sillanpää, Understanding the Factors Affecting the Adsorption of Lanthanum Using Different Adsorbents: a Critical Review, Chemosphere. 204 (2018) 413–430. doi:10.1016/J.CHEMOSPHERE.2018.04.053.

[9]

N. Krishnamurthy, C.K. Gupta, Extractive Metallurgy of Rare Earths, CRC press, 2004.

[10]

K. Inoue, K. Ohto, K. Yoshizuka, T. Yamaguchi, T. Tanaka, Adsorption of Lead(II) Ion on Complexane Types of Chemically Modified Chitosan, Bull. Chem. Soc. Jpn. 70 (1997) 2443–2447.

31

[11]

S. Nagib, K. Inoue, T. Yamaguchi, T. Tamaru, Recovery of Ni from a Large Excess of Al Generated from Spent Hydrodesulfurization Catalyst Using Picolylamine Type Chelating Resin and Complexane Types of Chemically Modified Chitosan, Hydrometallurgy. 51 (1999) 73–85. doi:10.1016/S0304-386X(98)00073-5.

[12]

J. Roosen, K. Binnemans, Adsorption and Chromatographic Separation of Rare Earths with EDTA- and DTPA-Functionalized Chitosan Biopolymers, J. Mater. Chem. A. 2 (2014) 1530–1540. doi:10.1039/C3TA14622G.

[13]

E. Repo, J.K. Warchol, T.A. Kurniawan, M. Sillanpää, Adsorption of Co(II) and Ni(II) by EDTA- and/or DTPA-Modified Chitosan: Kinetic and Equilibrium Modeling, Chem. Eng. J. 161 (2010) 73–82. doi:10.1016/j.cej.2010.04.030.

[14]

F. Zhao, E. Repo, M. Sillanpää, Y. Meng, D. Yin, W.Z. Tang, Green Synthesis of Magnetic EDTA- and/or DTPA-Cross-Linked Chitosan Adsorbents for Highly Efficient Removal of Metals, Ind. Eng. Chem. Res. 54 (2015) 1271–1281. doi:10.1021/ie503874x.

[15]

M.G. Mahfouz, A.A. Galhoum, N.A. Gomaa, S.S. Abdel-Rehem, A.A. Atia, T. Vincent, E. Guibal, Uranium Extraction Using Magnetic Nano-Based Particles of Diethylenetriamine-Functionalized Chitosan: Equilibrium and Kinetic Studies, Chem. Eng. J. 262 (2015) 198–209. doi:10.1016/J.CEJ.2014.09.061.

[16]

A.A. Galhoum, M.G. Mahfouz, S.T. Abdel-Rehem, N.A. Gomaa, A.A. Atia, T. Vincent, E. Guibal, Diethylenetriamine-functionalized chitosan magnetic nano-based particles for the sorption of rare earth metal ions [Nd(III), Dy(III) and Yb(III)], Cellulose. 22 (2015) 2589–2605. doi:10.1007/s10570-015-0677-0.

[17]

R. Ashour, M. Samouhos, E.P. Legaria, M. Svärd, J. Hogblom, K. Forsberg, M.

32

Palmlof, V.G. Kessler, G.A. Seisenbaeva, Å.C. Rasmuson, DTPA-Functionalized Silica Nano-and Microparticles for Adsorption and Chromatographic Separation of Rare Earth Elements, ACS Sustain. Chem. Eng. (2018). doi:10.1021/. [18]

D. Dupont, W. Brullot, M. Bloemen, T. Verbiest, K. Binnemans, Selective Uptake of Rare Earths from Aqueous Solutions by EDTA-Functionalized Magnetic and Nonmagnetic Nanoparticles., ACS Appl. Mater. Interfaces. 6 (2014) 4980–8. doi:10.1021/am406027y.

[19]

E. Polido Legaria, M. Samouhos, V.G. Kessler, G.A. Seisenbaeva, Toward Molecular Recognition of REEs: Comparative Analysis of Hybrid Nanoadsorbents with the Different Complexonate Ligands EDTA, DTPA, and TTHA, Inorg. Chem. 56 (2017) 13938–13948. doi:10.1021/acs.inorgchem.7b02056.

[20]

C.W. Noack, D.A. Dzombak, A.K. Karamalidis, Rare Earth Element Distributions and Trends in Natural Waters with a Focus on Groundwater, Environ. Sci. Technol. 48 (2014) 4317–4326.

[21]

J.C. Callura, K.M. Perkins, C.W. Noack, N.R. Washburn, D.A. Dzombak, A.K. Karamalidis, Selective Adsorption of Rare Earth Elements onto Functionalized Silica Particles, Green Chem. 20 (2018) 1515–1526. doi:10.1039/C8GC00051D.

[22]

E. Juere, J. Florek, D. Lariviere, K. Kim, F. Kleitz, Support Effects in Rare Earth Element Separation using Diglycolamide-Functionalized Mesoporous Silica, New J. Chem. (2016). doi:10.1039/C5NJ03147H.

[23]

D.L. Ramasamy, A. Wojtuś, E. Repo, S. Kalliola, V. Srivastava, M. Sillanpä, Ligand Immobilized Novel Hybrid Adsorbents for Rare Earth Elements (REE) Removal from Waste Water: Assessing the Feasibility of using APTES Functionalized Silica in the

33

Hybridization Process with Chitosan, Chem. Eng. J. (2017). doi:10.1016/j.cej.2017.08.098. [24]

K.Z. Hossain, L. Mercier, Intraframework Metal Ion Adsorption in LigandFunctionalized Mesoporous Silica, Adv. Mater. 14 (2002) 1053. doi:10.1002/15214095(20020805)14:15<1053::AID-ADMA1053>3.0.CO;2-X.

[25]

Y. Hu, E. Drouin, D. Larivière, F. Kleitz, F.-G. Fontaine, Highly Efficient and Selective Recovery of Rare Earth Elements Using Mesoporous Silica Functionalized by Preorganized Chelating Ligands, ACS Appl. Mater. Interfaces. 9 (2017) 38584– 38593. doi:10.1021/acsami.7b12589.

[26]

G.E. Fryxell, H. Wu, Y. Lin, W.J. Shaw, J.C. Birnbaum, J.C. Linehan, Z. Nie, K. Kemner, S. Kelly, Lanthanide Selective Sorbents: Self-Assembled Monolayers on Mesoporous Supports (SAMMS), J. Mater. Chem. 14 (2004) 3356. doi:10.1039/b408181a.

[27]

G.E. Fryxell, S. V. Mattigod, Y. Lin, H. Wu, S. Fiskum, K. Parker, F. Zheng, W. Yantasee, T.S. Zemanian, R.S. Addleman, J. Liu, K. Kemner, S. Kelly, X. Feng, Design and Synthesis of Self-Assembled Monolayers on Mesoporous Supports (SAMMS): The Importance of Ligand Posture in Functional Nanomaterials, J. Mater. Chem. 17 (2007) 2863. doi:10.1039/b702422c.

[28]

W. Yantasee, G.E. Fryxell, R.S. Addleman, R.J. Wiacek, V. Koonsiripaiboon, K. Pattamakomsan, V. Sukwarotwat, J. Xu, K.N. Raymond, Selective Removal of Lanthanides from Natural Waters, Acidic Streams and Dialysate, J. Hazard. Mater. 168 (2009) 1233–1238. doi:10.1016/j.jhazmat.2009.03.004.

[29]

C.M. Babu, K. Binnemans, J. Roosen, EDTA-Functionalized Activated Carbon for the

34

Adsorption of Rare Earths from Aqueous Solutions, Ind. Eng. Chem. Res. 57 (2018) 1487–1497. doi:10.1021/acs.iecr.7b04274. [30]

W. Yantasee, Y. Lin, G.E. Fryxell, K.L. Alford, B.J. Busche, C.D. Johnson, Selective Removal of Copper(II) from Aqueous Solutions Using Fine-Grained Activated Carbon Functionalized with Amine, Ind. Eng. Chem. Res. 43 (2004) 2759–2764. doi:10.1021/ie030182g.

[31]

Q. Zhou, H. Yang, C. Yan, W. Luo, X. Li, J. Zhao, Synthesis of Carboxylic Acid Functionalized Diatomite with a Micro-Villous Surface via UV-Induced Graft Polymerization and its Adsorption Properties for Lanthanum (III) Ions, Colloids Surfaces A Physicochem. Eng. Asp. 501 (2016) 9–16. doi:10.1016/j.colsurfa.2016.04.030.

[32]

S. da N. Almeida, H.E. Toma, Neodymium(III) and Lanthanum(III) Separation by Magnetic Nanohydrometallurgy using DTPA Functionalized Magnetite Nanoparticles, Hydrometallurgy. 161 (2016) 22–28. doi:10.1016/J.HYDROMET.2016.01.009.

[33]

F. Zhao, E. Repo, D. Yin, Y. Meng, S. Jafari, M. Sillanpää, EDTA-Cross-Linked βCyclodextrin: An Environmentally Friendly Bifunctional Adsorbent for Simultaneous Adsorption of Metals and Cationic Dyes, Environ. Sci. Technol. 49 (2015) 10570– 10580. doi:10.1021/acs.est.5b02227.

[34]

F. Zhao, E. Repo, Y. Meng, X. Wang, D. Yin, M. Sillanpää, An EDTA-β-Cyclodextrin Material for the Adsorption of Rare Earth Elements and its Application in Preconcentration of Rare Earth Elements in Seawater, J. Colloid Interface Sci. 465 (2016) 215–224. doi:10.1016/j.jcis.2015.11.069.

[35]

Y. Hu, J. Florek, D. Larivière, F.-G. Fontaine, F. Kleitz, Recent Advances in the

35

Separation of Rare Earth Elements Using Mesoporous Hybrid Materials, Chem. Rec. 18 (2018) 1–17. doi:10.1002/tcr.201800012. [36]

I. Anastopoulos, A. Bhatnagar, E.C. Lima, Adsorption of Rare Earth Metals: A Review of Recent Literature, J. Mol. Liq. 221 (2016) 954–962. doi:10.1016/j.molliq.2016.06.076.

[37]

C.W. Noack, K.M. Perkins, J.C. Callura, N.R. Washburn, D.A. Dzombak, A.K. Karamalidis, Effects of Ligand Chemistry and Geometry on Rare Earth Element Partitioning from Saline Solutions to Functionalized Adsorbents, ACS Sustain. Chem. Eng. 4 (2016) 6115–6124. doi:10.1021/acssuschemeng.6b01549.

[38]

K.L. Nash, F-Element Complexation by Diphosphonate Ligands, J. Alloys Compd. 249 (1997) 33–40. doi:https://doi.org/10.1016/S0925-8388(96)02520-0.

[39]

T.S. Grimes, K.L. Nash, Acid Dissociation Constants and Rare Earth Stability Constants for DTPA, J. Solution Chem. 43 (2014) 298–313. doi:10.1007/s10953-0140139-6.

[40]

C.A. Bunton, N.A. Fuller, S.G. Perry, V.J. Shiner, The Hydrolysis of Carboxylic Anhydrides. Part III. Reactions in Initially Neutral Solution., J. Chem. Soc. (1963) 2918–2926. http://pubs.rsc.org/-/content/articlepdf/1963/jr/jr9630002918 (accessed June 21, 2018).

[41]

D.L. Parkhurst, C.A.J. Appelo, Description of Input and Examples for PHREEQC Version 3—A Computer Program for Speciation, Batch-Reaction, One-Dimensional Transport, and Inverse Geochemical Calculations, in: U.S. Geol. Surv. Tech. Methods, B. 6, Chap. A43, Available Only Https//Pubs.Usgs.Gov/Tm/06/A43/, 2013: p. 497. https://pubs.usgs.gov/tm/06/a43/.

36

[42]

S.R. Charlton, D.L. Parkhurst, Modules Based on the Geochemical Model PHREEQC for Use in Scripting and Programming Languages, Comput. Geosci. 37 (2011) 1653– 1663.

[43]

R Core Team, R: A Language and Environment for Statistical Computing, (2017). https://www.r-project.org.

[44]

A.D. Sherry, W.P. Cacheris, K.-T. Kuan, Stability Constants for Gd3+ Binding to Model DTPA-Conjugates and DTPA-Proteins: Implications for their Use as Magnetic Resonance Contrast Agents, Magn. Reson. Med. 8 (1988) 180–190. doi:10.1002/mrm.1910080208.

[45]

C. Paul-Roth, K.N. Raymond, Amide Functional Group Contribution to the Stability of Gadolinium(III) Complexes: DTPA Derivatives, Inorg. Chem. 34 (1995) 1408–1412. http://pubs.acs.org/doi/pdf/10.1021/ic00110a019 (accessed July 18, 2017).

[46]

J. Lützenkirchen, T. Preočanin, D. Kovačević, V. Tomišić, L. Lövgren, N. Kallay, Potentiometric Titrations as a Tool for Surface Charge Determination, Croat. Chem. Acta. 85 (2012) 391–417.

[47]

W.J. Weber, J.C. Morris, Kinetics of Adsorption on Carbon from Solution, J. Sanit. Eng. Div. 89 (1963) 31–60.

[48]

A.A. Galhoum, K.M. Hassan, O.A. Desouky, A.M. Masoud, T. Akashi, Y. Sakai, E. Guibal, Aspartic Acid Grafting on Cellulose and Chitosan for Enhanced Nd(III) Sorption, React. Funct. Polym. 113 (2017) 13–22. doi:10.1016/j.reactfunctpolym.2017.02.001.

[49]

G. Yang, H. Chen, H. Qin, Y. Feng, Amination of Activated Carbon for Enhancing Phenol Adsorption: Effect of Nitrogen-Containing Functional Groups, Appl. Surf. Sci.

37

293 (2014) 299–305. doi:10.1016/j.apsusc.2013.12.155. [50]

Y.R. Smith, D. Bhattacharyya, T. Willhard, M. Misra, Adsorption of Aqueous Rare Earth Elements using Carbon Black Derived from Recycled Tires, Chem. Eng. J. 296 (2016) 102–111. doi:10.1016/j.cej.2016.03.082.

[51]

K.L. Nash, D. Brigham, T.C. Shehee, A. Martin, The Kinetics of Lanthanide Complexation by EDTA and DTPA in Lactate Media, Dalt. Trans. 41 (2012) 14547. doi:10.1039/c2dt31851b.

[52]

T. Jin, W. Yuan, Y. Xue, H. Wei, C. Zhang, K. Li, Co-Modified MCM-41 as an Effective Adsorbent for Levofloxacin Removal from Aqueous Solution: Optimization of Process Parameters, Isotherm, and Thermodynamic Studies, Environ. Sci. Pollut. Res. (2017). doi:10.1007/s11356-016-8262-0.

[53]

J.B. Alam, A.K. Dikshit, M. Bandyopadhayay, Evaluation of Thermodynamic Properties of Sorption of 2,4-D and Atrazine by Tire Rubber Granules, Sep. Purif. Technol. 42 (2005) 85–90. doi:10.1016/j.seppur.2004.06.006.

[54]

G. Tian, L.R. Martin, Z. Zhang, L. Rao, Thermodynamic, Spectroscopic, and Computational Studies of Lanthanide Complexation with Diethylenetriaminepentaacetic Acid: Temperature Effect and Coordination Modes, Inorg. Chem. 50 (2011) 3087–3096. doi:10.1021/ic200025s.

[55]

G.R. Choppin, M.P. Goedken, T.F. Gritmon, The Complexation of Lanthanides by Aminocarboxylate Ligands--II Thermodynamic Parameters, J. Inorg. Nucl. Chem. 39 (1977) 2025–2030. doi:https://doi.org/10.1016/0022-1902(77)80539-3.

[56]

P. Thakur, J.L. Conca, C.J. Dodge, A.J. Francis, G.R. Choppin, Complexation Thermodynamics and Structural Studies of Trivalent Actinide and Lanthanide

38

Complexes with DTPA, MS-325 and HMDTPA, Radiochim. Acta. 101 (2013) 221– 232. doi:10.1524/ract.2013.2018.

39

Graphical abstract

40