Rare Earth Elements Microfluidic Study of Synergic Liquid-Liquid Extraction … · 2020-01-27 · 1 Microfluidic Study of Synergic Liquid-Liquid Extraction of Rare Earth Elements
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
Microfluidic Study of Synergic Liquid-Liquid Extraction of Rare Earth Elements
Asmae El Maangar1.2. Johannes Theisen1. Christophe Penisson1. Thomas Zemb2. and Jean-Christophe P. Gabriel1.3.*.†
1 Institut de Chimie Séparative de Marcoule (ICSM). CEA/CNRS/UM2/ENSCM. F-38054. University Grenoble Alpe. CEA. Grenoble2 Institut de Chimie Séparative de Marcoule (ICSM). CEA/CNRS/UM2/ENSCM. F-30207. Bagnols-sur-Cèze3 Nanoscience and Innovation for Materials. Biomedecine and Energy (NIMBE). CEA/CNRS/Univ. Paris-Saclay. CEA Saclay. F-91191 Gif-sur-Yvette
* Corresponding authors: JCP Gabriel ([email protected])† Current address: Energy Research Institute @ NTU (ERI@N). Nanyang Technology University. Singapore.
Figure 1: (a)-(c) raw data of the extraction efficiency versus time obtained in microfluidics for different lanthanides
(La3+. Nd3+. Eu3+. Dy3+. Yb3+) and molar fractions of DMDOHEMA. [HNO3] = 0.3M. T = 25°C. xDMDOHEMA = 0 (a);
0.25 (b); 0.5 (c). Coherence with values obtained via a standard batch method after 1h of extraction is shown
in figure (d).
4
Figure 2: (a)-(e) raw data of the extraction efficiency versus time obtained in microfluidics for different lanthanides
(La3+. Nd3+. Eu3+. Dy3+. Yb3+) and molar fractions of DMDOHEMA. [HNO3] = 3M. T = 25°C. xDMDOHEMA = 0 (a). 0.25
(b). 0.5 (c). 0.75 (d). 1 (e). Coherence with equilibrium (1h) values obtained via a standard batch method is
shown as figure (f).
5
Table 3: the difference of the extraction ratio between batch and microfluidics (%) for different lanthanides (La3+. Nd3+. Eu3+. Dy3+. Yb3+) and molar fractions of DMDOHEMA (0. 0.25. 0.5. 0.75. 1). [HNO3] = 0.3 M. T = 25°C
xDMDOHEMA La Nd Eu Dy Yb0.00 0.0307 0.0543 0.0811 0.0114 0.01720.25 0.0008 0.0000 0.0052 0.0836 0.01440.50 0.0161 0.0141 0.0130 0.0163 0.0043
Table 4: the difference of the extraction ratio between batch and microfluidics (%) for different lanthanides (La3+. Nd3+. Eu3+. Dy3+. Yb3+) and molar fractions of DMDOHEMA (0. 0.25. 0.5. 0.75. 1). [HNO3] = 3 M. T = 25°C
1.4. Free energy of transfer versus acid concentration:
Figure 3 : Free energies of transfer ΔG (kJ/mole) versus HNO3 concentration for HDEHP extractant (xDMDOHEMA=0) T=25°C.
1.5. Precision in determination of free energy of transfer
The free energy of ion transfer between organic and aqueous phases can be determined by measuring the concentrations of rare earth elements in both of them using off-line X-ray fluorescence measurements. In order to calculate these concentrations with their associated errors, a new method was developed which is explained hereafter.
It relies on generating a full spectrum with the best secondary target according to composition. Then, a PLS (Partial least square) algorithm is used to defined a regression matrix from samples of known concentrations.1-5 Thus, concentrations from unknown samples can be extracted thanks to this regression matrix. Moreover, there is a known
6
external constraint that can be used in the determination of the ratio of extracted and non-extracted species: the sum of the atoms present within the two outgoing fluids must be equal to the amount within the incoming fluid. Adding this external constraint means minimizing the uncertainty in the distribution coefficient.
Figure 4: scheme of the different steps of the spectral treatment program to obtain the free energy of transfer and its associated error.
The exact procedure for obtaining the free energy of transfer using the XRF (as described in detail by Penisson)6 is summarized in the Figure 4 and can be divided into 3 main steps:
1. The calculation of the regression matrix from samples prepared with known concentrations in both the aqueous and organic phases: the program provided by Kirsanov et al. was used to define 42 secondary standard samples to be prepared at varying concentrations of five lanthanides: La, Nd, Eu, Dy, Yb and Iron.4 These samples were prepared from 75 mmol/L stock aqueous solutions of corresponding lanthanide nitrates in 3 M nitric acid. Lanthanide concentrations were varied in a range from 0.2 to 12.5 mmol.L-1. The lower limit of the range is chosen at the limit of quantification estimated by the signal-to-noise ratio of the lanthanum peak. The upper limit is defined above the maximum ion concentration of solutions of interest to be studied. This concentration range was relevant and adapted for liquid-liquid
7
extractions studied in the laboratory using microfluidics. The system to study is composed of an aqueous phase in a nitric medium containing La, Nd, Eu, Dy, Yb and Fe at a concentration of 10 mmol.L-1 and an organic phase containing the extractant diluted in Isane IP175. The preparation of stock solutions in the case of organic standards was prepared by liquid-liquid extraction. The aqueous phases were prepared by dilution of nitrates salts of lanthanides and iron weighed whereas the organic phases were prepared by weighing the extracting molecule HDEHP diluted in Isane IP175. The organic phases prepared were contacted for one hour at 25°C in tubes placed horizontally and shaken planetary. Then, the phases were separated by centrifugation at 15025 G for 10 minutes. The values of the extracted cations concentrations were measured by ICP. Finally, the dilutions of the mother phases were carried out with micropipettes. The 84 samples prepared were measured with the SPECTRO XRF spectrometer to obtain the regression matrices which are calculated from the 32 standards.
2. The calculation of the root mean squared error of prediction: The matrices obtained are then tested with the remaining 10 standard samples that are used as independent test sets to assess the predictive power of the calibration model by calculating the Root Mean Square of Error Prediction (RMSEP) which gives information on the measurement error by comparing the concentrations measured by the PLS algorithm with those prepared. It can be calculated as follows:
𝑅𝑀𝑆𝐸𝑃 = 𝑛
∑𝑖 = 1
(𝑐𝑃𝐿𝑆𝑖 ‒ 𝑐𝑝𝑟𝑒𝑝
𝑖 )2
𝑛 Equation 1
Where is a lanthanide concentration value derived from the PLS. is the actual 𝑐𝑃𝐿𝑆𝑖 𝑐𝑝𝑟𝑒𝑝
𝑖
lanthanide concentration and n is the number of samples.
8
Figure 5 : 3D representation of probability with the mass conservation plan
3. The calculation of the free energy of transfer ΔG: A new method was used to calculate the transfer energies from the concentrations in the two extraction phases and their associated errors. The probability of measurement can be represented by a two-dimensional Gaussian function of the corresponding concentrations and standard deviations (Equation 2) which can be traced in three-dimensional space as a function of both the concentration in water and oil (Figure 5).
𝑝(𝐶𝑎𝑞.𝐶𝑜𝑟𝑔) = 1
2𝜋 𝜎𝑎𝑞𝜎𝑜𝑟𝑔 𝑒
‒12
[(𝐶𝑎𝑞 ‒ 𝐶𝑚𝑒𝑠𝑢𝑟𝑒𝑎𝑞
𝜎𝑎𝑞 )2 + (𝐶𝑜𝑟𝑔 ‒ 𝐶𝑚𝑒𝑠𝑢𝑟𝑒𝑜𝑟𝑔
𝜎𝑜𝑟𝑔 )2]Equation 2
Where: is the standard deviation on aqueous concentrations and is the 𝜎𝑎𝑞 𝜎𝑜𝑟𝑔
standard deviation on organic concentrations
Now, since we know that during a liquid liquid extraction the total concentration of ions remains constant, in the defined three-dimensional domain, this property is translated by a plane that the 2D Gaussian should cut at its centre. In reality, the measurement has some uncertainty and the point (𝐶𝑎𝑞. 𝐶𝑜𝑟𝑔) will not coincide exactly to this plane. In this case, the most probable energy of transfer value will be at the intersection between the plane and the 2D Gaussian. To determine the most probable transfer energy, the intersection curve between the plane and the 2D Gaussian is calculated. The maximum probability corresponds to the most probable
9
ΔG. The error is defined by the width at half height. Note that this is the reason why we do not obtain symmetrical error bars as shown in Figure 6 and Figure 7
Overall, this method uses the concentrations of ions in the two phases and the law of conservation of mass to determine the most probable energy of transfer and the associated errors. From this paragraph we can now understand why XRF with the adequate minimization procedure is the most precise method for obtaining chemical potentials. Typically, with 10% uncertainty. Especially that the use of tracers in nanomole quantities are relying on complex safety procedure and absorption problems that are hard to control. On the other hand, the standard ICP-OES requires back-extraction to measure organic phase concentrations, and is therefore relying on yet another chemical step, adding to the experimental error. Only XRF has the dynamic range and selectivity for immediate analysis and even direct coupling whenever possible.
Figure 6 : Sectional view of the Z plane (left) and Intersection curve of the mass conservation plane and two-dimensional Gaussian (right)
Figure 7: Example of ΔG measurement at maximum probability and associated error defined by half width (Orange arrow). microfluidic extraction point 60 minutes. 50% DMDOHEMA / HDEHP. HNO3= 3M
10
The error bars for the ICP measurements correspond to the uncertainty calculated from equation 1. The error in ICP concentrations is estimated by the Error bars of concentrations measured for solution standards. A fixed error of 1% and a proportional error of 3% on the concentration were observed.
𝑢2𝑐(∆𝐺)𝐼𝐶𝑃 = ( ∂∆𝐺
∂𝐶𝑎𝑞0)2𝑢2(𝐶𝑎𝑞0) + (∂∆𝐺
∂𝐶𝑎𝑞)2𝑢2(𝐶𝑎𝑞)
Equation 3
𝑝(𝐶𝑎𝑞.𝐶𝑜𝑟𝑔) = 1
2𝜋 𝜎𝑎𝑞𝜎𝑜𝑟𝑔 𝑒
‒12
[(𝐶𝑎𝑞 ‒ 𝐶𝑚𝑒𝑠𝑢𝑟𝑒𝑎𝑞
𝜎𝑎𝑞 )2 + (𝐶𝑜𝑟𝑔 ‒ 𝐶𝑚𝑒𝑠𝑢𝑟𝑒𝑜𝑟𝑔
𝜎𝑜𝑟𝑔 )2]Equation 4
Where represents the error on the free energy of transfer of the ions from the aqueous 𝑢 𝑐(∆𝐺)𝐼𝐶𝑃
phase to the organic phase, the error on the measurement of the concentration in the 𝑢 (𝐶𝑎𝑞0)
initial aqueous phase before extraction and after extraction.𝑢 (𝐶𝑎𝑞)In the Figure 8(a) below, we focus on the case where the rare earth elements are well extracted. In this delicate case, the aqueous rare earth concentration is close to zero and therefore the ICP errors on ΔG0 diverge and tend towards infinity, while they remain acceptable in the case of XRF. In the case where the efficiency of the extraction is moderate (Figure 8 (b)), we notice that the measurement error increases for the XRF, whereas it decreases for the ICP.
0 0,25 0,5 0,75 1
-35
-25
-15
-5
5
15
25
-35
-25
-15
-5
5
15
25
0 0,25 0,5 0,75 1
𝜟𝐆0 (k
J/m
ol)
xDMDOHEMA
Nd ICP
Nd XRF
b) 0 0,25 0,5 0,75 1
-35
-25
-15
-5
5
15
25
-35
-25
-15
-5
5
15
25
0 0,25 0,5 0,75 1
𝜟𝐆0
(kJ/
mol
)
xDMDOHEMA
Yb ICP Yb XRF
a)
Figure 8: comparison of the free energy of transfer derived from classical Induced Coupled Plasma analysis (ICP) versus the constrained SIMPLEXE method used for exploiting paralel X-ray fluorescence (XRF). Initial organic phase: [DMDOHEMA]tot+[HDEHP]tot =0.9M in Isane. Initial aqueous phase: Ln(III) = 10 mM, [HNO3] = 0.03M, T = 25°C : (a) for neodymium. (b) for ytterbium.
1.6. Comparison microfluidic/literature
Table 5: experimental parameters of the comparative study
Experimental parameters
Microfluidics (This study) J.MULLER 1 J.REY 2.3
11
[HNO3] (M) 3 1 1
Elements 5 Lanthanides + Fe Eu with Lithium Eu
[Elements] (M) 0.01 1E-12 0.05
Extractants DMDOHEMAHDEHP
DMDOHEMAHDEHP
DMDOHEMAHDEHP
[Extractant]tot (M) 0.9 0.6 0.6
In the study of J. Muller,7 europium is more extracted than in our measurements. This difference comes from the addition of lithium salts within the aqueous phase which suppresses electrostatic interactions and then modifies the chemical potential without modifying the acidity of the aqueous phase, thus increasing the extraction performance.In the case of J. Rey et al.’s work.8, 9 the acidity is three times less and europium is five times more concentrated than in our study, respectively. The energy of transfer measures the energy balance between the two extraction phases and does not quantify the extracted concentrations. The initial concentration of ions in the aqueous phase may exceed the capacity of the extractant and the amount of ions in the aqueous phase becomes greater than that in the organic phase. The resulting energy of transfer increases. In the absence of the ionic extractant i.e xDMDOHEMA = 1, the free energy of transfer increases in the case of J. Rey et al. and this is due to the absence of the electrostatic contribution and the DMDOHEMA is known to extract rare earths at high acidity.
Figure 9 it can also be observed that for all three cases the free energy of transfer can be fitted by a quadratic function in x(1-x), which is equivalent, in a first approximation, to the variation of entropy. The entropic coefficient is smaller in J. Rey’s et al., due to the use of lower amount of extractant molecules.
0 0,25 0,5 0,75 1
-10
-5
0
5
10
-10
-5
0
5
10
0 0,25 0,5 0,75 1
∆G0
(kJ/
mol
)
xDMDOHEMAMICROFLUIDIQUE J.REY J.MULLER
Figure 9: Comparison of the extraction system DMDOHEMA / HDHEP [HNO3] 3M. XRF. with the data obtained in the literature in similar but different condition : adding 3 M of Li(NO3)3 in order to kill all electrostatic terms (J. Muller et al.)7 and J. Rey et al.8, 9 for higher load and lower nitric acid concentration. For each case a slight synergy is observed.
12
1.7. Experimental Data
Table 6: Raw data of the extraction kinetics in microfluidics for the different rare earths and molar fractions with
their associated error bars calculated using the Equation 1 for [HNO3] = 0.03 M.
Table 16: Entropy-Enthalpy for the different lanthanides tested and their three physical quantities: ionic radius. ionic
volume and surface charge density.
ElementIonic radius
(nm)Ionic volume
(nm3)Surface charge density (e/nm2) ∆H(kJ/mol) T∆S(kJ/mol)
La 0.105 0.0076 21.6537 -89.1120 -87.0876
Nd 0.099 0.0064 24.3580 -81.7690 -79.2153
Eu 0.095 0.006 26.4523 -81.2380 -78.2414
Dy 0.091 0.0051 28.8289 -78.0900 -73.3745
Yb 0.087 0.0042 31.5408 -75.7640 -68.3405
1. G. B. Dantzig, Programming in a Linear Structure. , Washington DC., 1948.2. H. Abdi, Wiley Interdisciplinary Reviews-Computational Statistics, 2010, 2, 97-106.3. K. H. Angeyo, S. Gari, A. O. Mustapha and J. M. Mangala, Applied Radiation and
Isotopes, 2012, 70, 2596-2601.4. D. Kirsanov, V. Panchuk, M. Agafonova-Moroz, M. Khaydukova, A. Lumpov, V.
Semenov and A. Legin, Analyst, 2014, 139, 4303-4309.5. D. Kirsanov, V. Panchuk, A. Goydenko, M. Khaydukova, V. Semenov and A. Legin,
Spectrochimica Acta Part B-Atomic Spectroscopy, 2015, 113, 126-131.6. C. Penisson, Ph. D., Université de Grenoble Alpes, 2018.7. J. Muller, PhD Thesis, Université Pierre et Marie Curie-Paris VI, 2012.8. J. Rey, M. Bley, J.-F. Dufrêche, S. Gourdin, S. Pellet-Rostaing, T. Zemb and S.
Dourdain, Langmuir, 2017, 33, 13168-13179.9. J. Rey, S. Dourdain, J.-F. Dufrêche, L. Berthon, J. M. Muller, S. Pellet-Rostaing and T.