Spectroscopic Study into Lanthanide Speciation in Deep Eutectic Solvents.

Deep eutectic solvents are a new class of green solvents that are being explored as an alternative for used nuclear fuel and critical material recycling. However, there is a paucity of knowledge regarding metal behavior in them. This paper explores the underlying chemistry of rare-earth elements in choline chloride-based deep eutectic solvents by using a multi-technique spectroscopic methodology. Results show that speciation is highly dependent on the choice of the hydrogen-bond donor. Collected EXAFS data showed Ln3+ coordination with ethylene glycol and urea in their respective solvents and coordination with chloride in the lactic acid system. Generalized coordination environments were determined to be [LnL4-5], [LnL7-10], and [LnL5-6] in the ethylene glycol, urea, and lactic acid systems, respectively. Collected UV/vis spectra for Nd3+ and Er3+ showed variations with changing solvents, showing that Ln-Cl interactions do not dominate in these systems. Luminescence studies were consistent, showing varying emission spectra with varying solvent systems. The shortest luminescent lifetimes were observed in the choline chloride-ethylene glycol deep eutectic solvent, suggesting coordination through O-H groups. Combining all collected data allowed Eu3+ coordination geometries to be assigned.

Introduction

Actinide and lanthanide coordination behavior in acid solutions, organic solvents, and molten salts is relatively well understood, particularly with regard to their use in used nuclear fuel treatment and critical materials recovery. However, though these solvent media are proven to be effective, their application in industrial settings is still hindered by issues such as high energy requirements, safety concerns, and cost. In an effort to overcome these issues, ionic liquids (ILs) have been explored as an alternative, due to their many beneficial properties.[1,2] However, their potential was never realized due to issues such as cost and toxicity. Deep eutectic solvents (DES’s) are a new class of solvents that are now garnering attention in the metal recovery sector, as they share many of the positive attributes of ILs, while being cheaper and having the potential to be classed as green solvents with low toxicities. These solvents have the capacity to solve the issues surrounding conventional recovery processes.[3−8] DES’s have many favorable characteristics for use as solvents in metal recovery processes, such as low volatility, a wide liquid range, low melting points, low/non-flammability, a large electrochemical window, low toxicity, and the potential to be tailored to specific uses based on judicious choice of constituents.[9,10] There are a number of defined types of DES’s, of which this study focuses on type III, those composed of a quaternary ammonium salt (QAS) and a hydrogen-bond donor (HBD).

Rare-earth elements (REE’s) are critical materials in the energy sector.[11,12] In the nuclear industry, REE’s are found in relatively large amounts in used nuclear fuel. They require separation and removal to enable the recycling of nuclear materials due to their high neutron-absorption cross-sections. REE’s, alongside other materials, are defined as critical due to their economic and strategic importance coupled with a substantial risk to their supply and inability to be replaced with more accessible alternatives. This necessitates the development of technologies to maintain a security of supply for them, often involving novel developments due to the use of secondary resources, such as waste electronics. Understanding of REE speciation in DES’s is fundamentally important for the development of new, sustainable, and safe recovery processes. An example of this is the work conducted by Choppin and colleagues into the complexation of REE’s and actinides by polyaminocarboxylate ligands, which paved the way for next-generation separation processes being developed by the nuclear industry for waste reprocessing.[13−15] In addition, the recovery of REE’s is vital to address the environmental and energy challenges in other energy sectors, which are only going to be exacerbated by the growing and ever-increasingly connected global population.[12,16−21]

The REE coordination number (CN) and geometry are dominated by ligand size and therefore steric effects around the metal center. This allows for high CN complexes to form as the REEs are relatively large. In aqueous systems, nine-coordinate species of the tricapped trigonal prismatic geometry dominate, but some heavier lanthanides do exhibit eight-coordinate structures, where one of the capping water molecules has been lost.[22] The presence of other ligands in aqueous systems can result in higher CN’s, such as the 12-coordinate [Ln(NO3)6]3+ complexes (Ln = La, Ce, Nd, Pr).[23] Lower coordinate complexes are rare in aqueous systems but can be formed in other solvents with particularly bulky ligands.

A major challenge when using novel solvents such as DES’s in metal recovery processes is the paucity of knowledge regarding metal behavior in them. Such processes often rely on differences in metal speciation between phases to drive them. This is something that is well-understood in ILs but has not enabled their scale-up progress for use in industrial separation processes as solvents,[24] a major reason for this being their cost, and questions surrounding their green credentials, particularly as they are commonly dissolved in organic solvents when they are used.[25] The green nature of DES’s is heavily dependent on the chosen constituents, but it is well-established that they can be composed of “green” compounds. With this in mind, one of the largest obstacles in their use and scale-up progress for REE recovery will be this understanding of metal behavior in them. Published works that present promising REE dissolution and/or separation are unable to provide a definitive explanation of the underlying chemistry driving the process.[26−29] A comprehensive paper that details this behavior is not only needed to understand these processes but also to catalyze their scale-up progress through enabling the intelligent design of them.

In this paper, we report on the coordination environment of a range of REEs (Nd3+, Eu3+, Dy3+, Er3+, and Yb3+) in a set of DES’s. A multi-technique spectroscopic approach is taken, using a synergistic combination of an extended X-ray absorption fine structure (EXAFS), UV/vis absorption, and time-resolved luminescence (TRLS) spectroscopies. Choline chloride (ChCl) was selected as the QAS due to its ubiquity of use in DES’s, and ethylene glycol, urea, and lactic acid were selected as as HBD’s. These HBD’s are historically well-represented in the literature and have been shown to exhibit similar properties when compared to other polyols, amides, or carboxylic acids when used as an HBD. Following this, we envisage that the REE coordination environment will show similarities when switching similar HBD’s, for example, ethylene glycol with glycerol, enabling the data collected here to be relevant to other type III DES’s. The determined coordination environment of the REE’s is discussed with reference to what we know about the molecular structure of DES’s.

Results and Discussion

EXAFS Spectroscopy

Data were collected for Nd3+, Eu3+, Dy3+, Er3+, and Yb3+ in the ChCl–ethylene glycol and ChCl–urea DES (Figures S1–S3). In the ChCl–lactic acid, DES data were only collected for Nd3+ and Eu3+ due to their low solubilities and thus reduced data quality. The extraction of the EXAFS function and subsequent Fourier transform gave radial distribution functions with a large peak at ca. 2 Å, followed by weaker peaks at greater R values (Figure ). Fits were conducted by only considering O and Cl atoms in the first coordination sphere, with the addition of C atoms for subsequent shells if appropriate.

Figure 1

LIII-edge EXAFS spectra and fits in the R space for REEs dissolved in the choline chloride–ethylene glycol, −urea, and −lactic acid DES. Please refer to the Supporting Information for fitting windows.

In the ethylene glycol DES, acceptable fits were achieved by only including oxygen atoms in the inner coordination sphere for all REE’s (Tables , S1, Figures S4–S13). This is consistent with the expectation that Ln’s coordinate with oxygen more readily than chloride. The CN was 8 for all the Ln’s tested, except for Nd3+ which had a CN of 10. The second peak between 2.8 and 3 Å was fit with Ln-C scattering paths, producing acceptable fits for all data sets apart from Yb3+. This leads to the conclusion that the coordinating ligand is ethylene glycol, as seen in this DES previously with Ni2+, where Ni2+ is hexa-coordinated to three bidentate ethylene glycol molecules.[32] It has been reported that in this DES, the ethylene glycol is less able to disrupt the electrostatic interaction between the N+ and Cl– of the ChCl, meaning that the ethylene glycol is bound less strongly to the QAS.[33]

Table 1

Coordination Numbers and Bond Lengths Derived From EXAFS Data Fits for REEs in Choline Chloride–Ethylene Glycol, −Urea, and −Lactic Acid DES’sa

	ChCl–ethylene glycol				ChCl–urea				ChCl–lactic acid
Ln3+	Ln–O		Ln–C		Ln–O		Ln–C		Ln–Cl
	Nb	r/Å	Nc	r/Å	Nb	r/Å	Nc	r/Å	Nc	r/Å
Nd3+	10/9	2.53	10	3.48	10/8	2.44			10/6	2.7
Eu3+	8/9	2.5	8	3.47	9/8	2.4	9	3.54	10/5	2.62
Dy3+	8/9	2.43	8	3.35	9/7	2.35	9	3.58
Er3+	8/9	2.4	8	3.94	8/7	2.31	8	3.42
Yb3+	8	2.35			8/6.5	2.26	8	3.44

Detailed EXAFS fitting parameters can be found in the Supporting Information (Tables S1–S3).

Numbers to the left of the slash are CNs derived from EXAFS fits. Numbers to the right of the slash are CNs based on the most-likely case based on bond distances.

Coordination numbers for carbon atoms are those derived from EXAFS fits, and they are likely an overestimation.

The reported Ln–O bond lengths produced from the fits in the ChCl–EG DES can be correlated with reported species in the literature. This is important as EXAFS data give much more accurate bond distances than CNs, which is of particular importance with lower quality data sets. Single-crystal X-ray diffraction data for Nd3+ coordinated by polyethylene glycol, Cl–, H2O, and NO3– ligands have CN’s ranging from 9 to 11.[34] The average Nd–O distance to alcoholic O atoms varies from 2.45 to 2.52 Å. Our determined Nd–O distance of 2.53 Å lies just outside this range. However, when looking at a larger sample of Nd-(poly)ethylene glycol complexes, the average Nd–O distance to alcoholic O atoms varies from 2.46 to 2.58 Å. Our determined distance of 2.53 Å lies in the middle of this range. This is suggestive of a 10-coordinate complex, which would agree with a bidentate ethylene glycol ligand.

Table S11 shows a collection of Ln–O bond distances for Ln coordination complexes with (poly)ethylene glycol ligands from the Cambridge Crystallographic Data Centre (CCDC).[34−53] These data are in good agreement with our determined bond lengths for Nd3+, Er3+, and Yb3+ as 10- and 8- coordinate complexes. However, the distances for Eu3+ and Dy3+ are longer in our EXAFS fits than in the collected data when only considering the Ln–O distances for alcoholic O atoms. These longer distances may correlate with nine-coordinate complexes. In a comprehensive study of Ln–O bond distances in Ln–dimethylpropyleneurea (DMPU) complexes, Lundberg et al. used EXAFS and crystallographic data to correlate the ionic radius with CN.[54] This was done by subtracting the ionic radius of a coordinated oxygen atom (1.34 Å) from the Ln–O bond distance. Repeating this methodology with our data gives ionic radii that generally correlate well with the fit CNs but do make a definitive assignment difficult (Table S9). Assuming a bidentate ethylene glycol ligand, the data correlate. If not, a more sensible CN assignment would be nine, nine, nine, nine, and eight for Nd3+, Eu3+, Dy3+, Er3+, and Yb3+, respectively.

Fits of the data in the ChCl–urea DES show some similarities with those of the ChCl–ethylene glycol DES. Acceptable fits were produced by fitting two shells to the data, the first containing O atoms and the second containing C atoms (Tables and S12). The CN was more variable in this solvent, with values of 8, 9, and 10 being fit. This is consistent with a reduction in the ionic radius with increasing Ln3+ mass and a monodentate coordinating ligand. This is likely to be the amidic oxygen of the urea moiety, as it has been reported that the main interactions between ChCl and urea do not involve the carbonyl group, leaving it free to interact with the Ln3+ center.[10] It is technically possible that the alcoholic oxygen atom of the choline is involved; however, its interaction in the bulk DES structure means that the more likely coordinating moiety is urea. The Ln–O distances are shorter when urea is the HBD when compared with ethylene glycol. This is due to the monodentate nature of the urea ligand causing less steric hindrance than the bidentate ethylene glycol moiety.

Repeating the process for correlating the ionic radius with CN for the ChCl–Urea–Ln3+ system gives ionic radii that are lower than those from the fits (Table S10). This would give CNs of eight, eight, seven, seven, and seven for Nd3+, Eu3+, Dy3+, Er3+, and Yb3+, respectively. However, this could be expected, as urea is a smaller ligand than DMPU, which is a larger, more space-demanding ligand.[54] Xiong et al. reported a nine-coordinate heteroleptic Eu3+ complex produced via ionothermal synthesis in the ChCl–ethyleneurea DES.[55] The three coordinating ethyleneurea ligands had Ln–O distances of 2.37, 2.38, and 2.41 Å, respectively, matching with the 2.4 Å distance produced from our EXAFS fits. This structure is encouraging as to the veracity of our fits; however, the lack of structural data for Ln3+ complexes associated with the DES does need to be addressed further.

A survey of the CCDC for Ln3+ complexes containing urea and derivatives shows a variety of CNs and Ln–O bond lengths (Table S12).[55−71,74] There are very few homoleptic complexes reported, making direct comparisons between our data and the crystal structures difficult. The homoleptic [Dy(Urea)8]3+ complex has an average Ln–O bond distance of 2.37 Å, slightly longer than that in our proposed [Dy(Urea)9]3+ complex.[67] However, the heteroleptic [Dy(Urea)4(H2O)4]3+ complex, also eight-coordinate, has an average Ln–O bond distance of 2.29 Å, which is shorter than that in our data.[72] The only other homoleptic urea complex for the Lns studied in this work in the CCDC is [Yb(Urea)6]3+, with a shorter average Ln–O bond length of 2.20 Å when compared with the 2.26 Å refined in this work. In contrast to this, the eight-coordinate [Yb(propyleneurea)8] complex has an average Ln–O bond distance of 2.40 Å, much longer than that proposed in our eight-coordinate structure, likely arising from steric effects.[56] The only 10-coordinate Nd3+ complex in the CCDC is [Nd(Urea)2(H2O)(NO3)3].[73] The average Ln–OUrea distance in this complex is 2.44 Å, the same as in our EXAFS data. It is likely that our fitted CNs are not perfect, which prevents us from reporting them with confidence. However, our data do have some consistency with the previously published literature, and this report can act as a guide for what CNs to expect in such media.

Fits containing O atoms did not produce acceptable results for Nd3+ and Eu3+ in the ChCl–lactic acid DES. Instead, acceptable fits were obtained for a first coordination shell comprised of Cl– ligands for both Ln’s (Figures S4–S7, Tables , S3). This was unexpected, as Ln’s are well known to form complexes with carboxylic acids in aqueous and non-aqueous media.[74,75] However, even though Ln’s tend to prefer O-donor ligands to Cl– ligands, this may not be true in DES’s. When comparing the extracted radial distribution functions for Nd3+ and Eu3+ in each solvent, it is clear that the first peak is at a further R in the acidic DES than the other two. This means that the first coordination sphere is further away, suggestive of Cl– atoms in that shell, as Ln–Cl distances are generally larger than Ln–O distances.

The nature of the interactions between the QAS and HBD is less understood in this solvent than in the ChCl–ethylene glycol and urea DES. However, molecular modeling has shown that there are major interactions between the lactic acid and choline moieties in the ChCl–lactic acid DES. The collected data actually suggested a repulsive interaction between the lactic acid and Cl–.[76] These interactions result in a Cl– ion that is less strongly interacting with the other DES molecules than in the ethylene glycol and urea cases, leaving it free to coordinate with dissolved metal species in preference to the HBD. This also goes some way in explaining the observed low solubility of Lns in the ChCl–lactic acid DES, as formation of [LnCl]3– complexes may not be favorable.

Based on a literature survey, it seems that the fitted Ln–Cl distances for Nd and Eu correlate quite well with CN’s of six and five, respectively (Table S13).[77−102] This is much lower than the fitted CN of 10 for both lanthanides. Homoleptic Nd3+- and Eu3+-chloro complexes tend to be six-coordinate octahedra, with average Ln–Cl distances of 2.73 and 2.69 Å, respectively. There are no examples of homoleptic Nd3+- or Eu3+-chloro complexes in the CCDC with a CN above six. Heteroleptic chloro complexes can show higher CNs, accompanied by an increase in the average Ln–Cl distance. Interestingly, for six-coordinate Nd3+- and Eu3+-chloro complexes, heteroleptic Ln–Cl distances are shorter than those for homoleptic complexes (Table S24).

Based on this, further attempts were made to include oxygen atoms in the fits, but this always resulted in unacceptable fitting parameters. There is a Nd3+ center coordinated by seven Cl– ligands reported by Shan et al. with Ln–Cl distances from 2.7 to 2.9 Å, showing the high variance in the Ln–Cl distance in some Ln–chloro complexes.[103] All the Cl– ligands in this crystal structure are bridging between two or three metal centers, which is in some ways similar to in a DES, where they will be interacting with the Ln3+ ions and other DES components. An 11-coordinate Eu3+ center in a metal–organic framework was reported with two coordinating Cl– ligands at very short Ln–Cl distances of 2.15 and 1.92 Å, respectively.[104] This does show that highly coordinated Eu3+ centers can have unexpectedly short Ln–Cl distances, though this is an exception rather than a norm.

There are no homoleptic Ln-chloro complexes in the CCDC database with a CN higher than six. When considering our data and the literature, it is highly likely that our CNs are too high and that the Ln3+ centers are coordinated by five–six chlorides. However, with the current lack of research into the metal coordination behavior in and structure of carboxylic acid DESs, it is not possible to define the exact nature of the complexes present. This demonstrates the challenges with determining solution state structures when compared with crystal structures, as their disordered nature makes them more difficult to resolve. Solution state species can be much more variable in bond lengths and angles than crystal structures; however, it has been shown that REE coordination complexes show a good consistency between both types of structure.[22] The inclusion of La3+ into this work would have been valuable as it is the largest of the lanthanides and can exhibit higher CNs. Based on general trends in coordination chemistry along the lanthanide series, it is likely that the coordination structure of La3+ in ChCl–Urea and ChCl–lactic acid would be similar to that of Nd3+ as the coordinating species is monodentate. It is possible that slightly higher CNs would be seen in ChCl–ethylene glycol due to its bidentate nature. This warrants further examination, ideally including both spectroscopic and computational methodologies.

UV/Vis Absorption Spectroscopy

Interpretation of Ln3+ UV/vis absorption spectra is challenging, primarily owing to the Ln contraction and subsequent weak interactions of valence orbitals with those of ligating species. The observed transitions in these spectra are due to f–f electronic transitions, where the energies of the f-orbital microstates are non-degenerate due to electron repulsion, spin–orbit coupling, crystal-field effects, and the Zeeman effect.[105] These absorption bands are often quite weak, with molar extinction coefficients less than 1 L mol–1 cm–1. Nevertheless, useful information can still be gleaned from the analysis of such spectra. Particular attention in this section will be given to Nd3+ and Er3+, as they have easily accessible absorption spectra. Collected UV/vis absorption spectra and extracted data are shown in Figure and Tables S4–S8, respectively. Oscillator strengths (f) of each transition were calculated using eq , where ε is the molar absorption coefficient (L mol–1 cm–1), and ν is the wavenumber (cm–1). Conversion of ε from kg mol–1 cm–1 to L mol–1 cm–1 was done using literature values for DES densities.[9,106,107]

Figure 2

UV/vis spectra of Nd3+ and Er3+ in choline chloride-based DESs.

Nd3+ UV/vis spectra show up to 16 peaks rising from the 4I9/2 ground state, which vary in shape and intensity based on the HBD (Figure ). This suggests that the Nd3+ coordination environment is different in each solvent, consistent with the EXAFS data analyses. The hypersensitive transition, 4I9/2 → 4G5/2 (ca. 580 nm), reduces in oscillator strength with the HBD in the order lactic acid > urea > ethylene glycol (Tables S4–S6). A reduction in the intensity of a hypersensitive band can be correlated with an increase in symmetry around the central Nd3+ ion. The 4I9/2 → 2G7/2 transition, which lies in the same region of the spectrum, is not visible in the urea and ethylene glycol DES but is easily discernible in the lactic acid DES.

Figure 3

Luminescence spectra of EuCl3 (0.02 mol kg–1) dissolved in choline chloride-based DESs with ethylene glycol, urea, and lactic acid as HBDs.

The two next most intense peaks in the Nd3+ spectra correspond to the 4I9/2 → 4F7/2, 4S3/2 (ca. 745 nm) and 4I9/2 → 4F5/2, 2H9/2 (ca. 800 nm) transitions. These are not considered hypersensitive transitions, but they do show a large degree of variance between the three solvents. Of particular note is the 4I9/2 → 4F5/2, 2H9/2 transition in choline chloride–urea, which is more intense than the hypersensitive 4I9/2 → 2G7/2 transition. This is not the case in the other two solvents. The slight changes in peak positions, intensities, and splitting are all evidence for different Nd3+ coordination environments in each solvent.

The UV/vis spectra of Er3+ are remarkably similar in both the ChCl–ethylene glycol and urea DES. In general, there are subtle differences in the peak position and intensity in each solvent, suggestive of a different coordination environment. Er3+ exhibits two hypersensitive transitions in absorption spectra, 4I15/2 → 4G11/2 (ca. 380 nm) and 4I15/2 → 2H11/2 (ca. 520 nm). These are by far the most intense transitions in the collected spectra, and decrease in the oscillator strength/intensity with the HBD in the order urea > ethylene glycol is seen. This is the same order as Nd3+, again showing an increasing symmetry around the central Er3+ atom in the same order.

Luminescence Spectra

Europium

Luminescence spectra were collected for Nd3+, Eu3+, and Dy3+ in the ChCl–ethylene glycol, −urea, and −lactic acid DES. Attempts were made with solutions of Er3+ and Yb3+ but were unsuccessful due to equipment limitations. The luminescence spectra of Eu3+ in the choline chloride-based DES are shown in Figure . For comparative purposes, the spectra have been normalized as per the convention that the total integrated intensity of the 5D0 → 7F1 transition is equal in each spectrum. There are obvious differences between the spectra, particularly the hypersensitive 5D0 → 7F2 transition and the 5D0 → 7F4 transition and more subtle differences between the other peaks. These differences are telling of different Eu3+ coordination environments in each DES, which agrees with the results of the EXAFS fitting.

The luminescence spectrum of Eu3+ in the ChCl–ethylene glycol DES is dominated by the hypersensitive 5D0 → 7F2 transition, with an intensity three times that of the 5D0 → 7F1 transition (Table ). This is indicative of a geometry without an inversion center. The 5D0 → 7FJ (J = 1, 3, 4) peaks are also clearly visible in the spectrum. The 5D0 → 7F0 peak is slightly obscured by a shoulder on the larger 5D0 → 7F0 peak, making it difficult to explicitly confirm that it is a single peak. However, the EXAFS spectroscopy data suggest the presence of one Eu3+ species, with eight or nine coordinating oxygen atoms. Of the possible coordination geometries, the only ones that allow a 5D0 → 7F0 transition are the bicapped trigonal prism (CN = 8, C2v symmetry) and the capped square antiprism (CN = 9, C4v symmetry).[105,108,109] As the spectra were collected at room temperature in the liquid state, the vibronic coupling of the Eu3+ complex prevented the resolution of the 2S+1LJ sublevels. Low-temperature studies would help to alleviate this problem and allow for a more confident coordination assignment, assuming no change in the coordination geometry upon freezing.

Table 2

Peak Maxima and Relative Integrated Intensities for Luminescence Emission of Eu3+ Dissolved in Choline Chloride–Ethylene Glycol, −Urea, and −Lactic Acid DESsa

	ethylene glycol		urea		lactic acid
transition	λ/nm	relative intensity	λ/nm	relative intensity	λ/nm	relative intensity
5D0 → 7F0	581.12	0.06	578.92	0.10	580.84	0.08
5D0 → 7F1	592.94	1.00	593.21	1.00	592.94	1.00
5D0 → 7F2	616.00	3.21	613.26	3.41	614.91	5.41
5D0 → 7F3	651.10	0.15	651.37	0.21	651.92	0.22
5D0 → 7F4	697.03	1.68	700.30	2.14	699.21	1.56

The 5D0 → 7F1 integrated intensities were normalized to make them equal in all spectra. Relative intensity refers to the integrated intensity of each 5D0 → 7FJ transition divided by that of the 5D0 → 7F1 transition.

The emission spectrum of Eu3+ in the ChCl–urea DES shows five peaks, corresponding to the transitions 5D0 → 7FJ (J = 0, 1, 2, 3, 4). The presence of the 5D0 → 7F0 peak shows that the coordination geometry of the Eu3+ complex is Cn, Cnv, or Cs.[110] The EXAFS fit of eight–nine coordinating oxygen atoms would give us expected geometries of either the tricapped trigonal prism, capped square antiprism, or bicapped trigonal prism. The symmetry groups of these geometries are D3h, C4v, and C2v, respectively. As 5D0 →7F0 transitions are not observed for D3h complexes, it is logical to assign this complex to the C4v or C2v symmetry groups. A distorted capped square antiprismatic Eu3+ center was reported for the [Eu(2,2′-biphenyldicarboxylate)(NO3)(ethyleneurea)3] complex produced from the ChCl–ethyleneurea DES.[55] Additionally, as there is only one observable 5D0 → 7F0 peak, we can be confident that there is only one species present, which agrees with the EXAFS fitting. The 5D0 →7F1 peak appears to be composed of two transitions, which would agree with a geometry of C3 or higher symmetry, though it must be noted that the resolution is too low to show the individual peaks. A very intense 5D0 → 7F4 peak is present in the spectrum, showing a much higher intensity than that of the 5D0 → 7F1 transition. This enhanced intensity of the 5D0 →7F4 transition has been reported previously for the nine-coordinate capped square antiprismatic Eu3+ complex [Eu(DOTA)(H2O)]−.[111,112]

Eu3+ in the ChCl–lactic acid DES has the most intense 5D0 → 7F2 peak in its luminescence spectrum, indicating that the Eu3+ ion has the lowest symmetry in this solvent. The large intensity of this hypersensitive peak is interesting, as it could be taken as evidence for a lack of an octahedral [EuCl6]3- species, which exists in other high-chloride media and would fit with the Eu–Cl distance.[113] If this species was indeed present, then the spectrum would be dominated by the 5D0 → 7F1 transition.[105,114−117] It is possible that the species present could be a distorted octahedron, but the 5D0 → 7F1 still tends to dominate the spectra of such species.[24] A distorted capped octahedral geometry was reported for Eu3+ in a β-diketonate complex, which showed an intense 5D0 → 7F2 transition.[21] This is an example of a high-symmetry complex showing such a transition. It is possible that there are other, non-coordinating DES constituents contributing to a large distortion of the [EuCl6]3– octahedron, but we are unable to state this definitively. It is also possible that the extended H-bonding network of the DES and interactions between Cl– and other DES constituents cause this, but this is currently speculative only. Regardless of the true nature of the Eu3+ coordination environment, we can confidently say that it is highly unlikely to be a 10-coordinate. The 5D0 → 7F0 peak is observed, again identifying Cn, Cnv, or Cs as the possible symmetry groups for this complex. Interestingly, this peak in the spectrum looks like it may be composed of two separate peaks, which would indicate the presence of multiple species. However, the addition of other scattering paths to the EXAFS fit resulted in poor fits, and the luminescence lifetime (see the Supporting Information) is well-fitted with a mono-exponential function.

Structural identities of Eu3+ in each DES are not strictly identified from the EXAFS studies alone. The most likely coordination geometries around Eu3+ are the bicapped trigonal prism and capped square antiprism for the ethylene glycol and urea systems, respectively. Though a distorted octahedral geometry is suggested as most likely in the ChCl-EG system, there is more ambiguity with this than with the other solvents. DES constituents at further distances will be important in defining the true coordination structure, but these were not resolved in the collected EXAFS data. Though we have tried to define specific coordination polyhedra, it must be noted that these are not necessarily a strict reflection of reality.

Dysprosium and Neodymium

Luminescence spectra of Dy3+ are not as informative as those of Eu3+ with regard to the coordination geometry.[118] The spectra of Dy3+ in the choline chloride–ethylene glycol and −urea DES (Figure ) are similar, with slight differences in the peak position and intensity (Table ). These subtle peak differences are characteristic of different Dy3+ coordination environments. As the EXAFS fits of Dy3+ and Eu3+ produced similar CN’s in the choline chloride–ethylene glycol and −urea DES, we can say it is likely that they have the same coordination geometry. This goes some way to explaining the observed spectral differences in each DES. Only two peaks are observed for Dy3+ in both DES’s, corresponding to the 4F9/2 → 6H13/2, 6H11/2 transitions. The 4F9/2 → 6H15/2 transition, usually observed at ca. 480 nm, was not present in our collected spectra. It is not clear why this is the case, but this, combined with the intense 4F9/2 → 6H13/2, suggests that this medium is effective at quenching the yellow luminescence of Dy3+.

Figure 4

Luminescence spectra of DyCl3 (0.02 mol kg–1) dissolved in choline chloride-based DESs with ethylene glycol and urea as HBDs.

Table 3

Luminescence Transition Data for DyCl3 Dissolved in Choline Chloride-Based DESs With Ethylene Glycol and Urea as HBDs

HBD	transition	λmax./nm	integrated intensity
ethylene glycol	4F9/2 → 6H13/2	573.6	11.35
	4F9/2 → 6H11/2	661.9	0.91
urea	4F9/2 → 6H13/2	574.7	13.94
	4F9/2 → 6H11/2	663.8	1.27

Neodymium samples were excited at 355 nm, and luminescence spectra were collected by monitoring the 700–990 nm region. Unfortunately, the ICCD detector available to us is not able to detect further into the N-IR region, meaning we were unable to detect the 4F3/2 → 4I11/2, 4I13/2 emission bands. The well-known 4F3/2 → 4I9/2 transition (ca. 870 nm) is present in the Nd3+ spectra in each DES, alongside another transition at ca. 800 nm, which has been assigned to the 2H9/2,4F5/2 → 4I9/2 transition, also seen in the UV/vis absorption spectra (Figure , Table ). There are differences in the emission spectra based on the HBD, alluding to a change of the Nd3+ coordination geometry, which is in agreement with both the EXAFS fits and UV/vis data. Due to our experimental limitations, it is not possible to go beyond this simplistic interpretation of the spectra.

Figure 5

Luminescence spectra of NdCl3 (0.02 mol kg–1) dissolved in choline chloride-based DESs with ethylene glycol, urea, and lactic acid as HBDs.

Table 4

Luminescence Transition Data for NdCl3 Dissolved in Choline Chloride-Based DESs with Ethylene Glycol and Urea as HBDs

HBD	transition	λmax./nm	integrated intensity
ethylene glycol	2H9/2, 4F5/2 → 4I9/2	801.7	2.69
	4F3/2 → 4I9/2	868.7	17.26
urea	2H9/2, 4F5/2 → 4I9/2	801.1	4.12
	4F3/2 → 4I9/2	866.8	20.55
lactic acid	2H9/2, 4F5/2 → 4I9/2	803.9	3.84
	4F3/2 → 4I9/2	870.3	21.55

Luminescence Lifetime

Luminescence lifetime spectra were collected for Nd3+ and Eu3+ in each of the three DES’s, and spectra for Dy3+ were collected in the DES with ethylene glycol and urea as HBD’s. Spectra were collected at concentrations of 0.01, 0.015, and 0.02 mol kg–1 and the fit in origin (Figures S14–S16, Table ). All collected data were best fit by a mono-exponential decay, indicating the presence of one species. Lifetimes for Eu3+ and Nd3+ increased according to the changing HBD in the order ethylene glycol > lactic acid > urea, with Dy3+ increasing in the order ethylene glycol > urea. Ln3+ exhibiting the shortest lifetimes in the choline chloride–ethylene glycol DES is consistent with coordination through the O–H groups of the ethylene glycol molecule. This is due to efficient luminescence quenching via O–H vibrations. These data show that the HBD has a large effect on luminescence lifetime and is indicative of a dependence of the coordination environment on the HBD, which is consistent with our discussion thus far.

Table 5

Luminescence Lifetimes for Ln3+ in Choline Chloride-Based DES. Nd3+: λEx. = 355 nm, Monitor the 4F3/2 → 4I9/2 Emission. Eu3+: λEx. = 416 nm, Monitor the 5D0 → 7F2 Transition. Dy3+: λEx. = 435 nm, Monitor the 4F9/2 → 6H13/2 Transition

	HBD	[Ln3+]/mol kg–1	lifetime/μs	R2
Eu3+	ethylene glycol	0.020	71.6	0.9989
		0.015	103.5	0.9979
		0.010	137.9	0.9992
	urea	0.020	467.3	0.9992
		0.015	523.6	0.9989
		0.010	617.3	0.9992
	lactic acid	0.020	308.6	0.9997
		0.015	386.1	0.9997
		0.010	427.3	0.9995
Dy3+	ethylene glycol	0.020	7.2	0.9998
		0.015	7.3	0.9996
		0.010	7.3	0.9995
	urea	0.020	20.7	0.9995
		0.015	20.7	0.9996
		0.010	20.5	0.9993
Nd3+	ethylene glycol	0.020	79.8 × 10–3	0.9992
		0.015	78.4 × 10–3	0.9995
		0.010	77.5 × 10–3	0.9993
	urea	0.020	197.6 × 10–3	0.9999
		0.015	198.8 × 10–3	0.9998
		0.010	202.4 × 10–3	0.9998
	lactic acid	0.020	157.2 × 10–3	0.9998
		0.015	156.0 × 10–3	0.9997
		0.010	151.5 × 10–3	0.9989

The Eu3+ lifetime shows an inverse correlation with [Eu3+]. This is seen in all DES’s studied but is not observed for Nd3+ or Dy3+. This may be a result of Eu3+ complexes reabsorbing emitted light through the inner filter effect, but there is no evidence to suggest that Eu3+ absorbs in the same region as its emission spectrum. However, the lifetime reductions are significantly larger when ethylene glycol is the HBD when compared with urea and lactic acid, which are similar (Table ). This may mean that there is another quenching process occurring, a possibility is some form of the collisional quenching process.[116] The markedly lower viscosity of choline chloride–ethylene glycol (36 cP), when compared with that of choline chloride −urea (996.8 cP) and choline chloride–lactic acid (1245.4 cP), may offer an explanation for this hypothesis.[9,106,107]

Table 6

Reduction of Eu3+ Luminescence Lifetime with Increasing [Eu3+]. The Longest Lifetime was Arbitrarily Set at 100%, and all Others are Compared to This Value

	relative luminescence lifetime/%
[Eu3+]/mol kg–1	ChCl–EG	ChCl–U	ChCl–LA
0.02	51.9	75.7	72.2
0.015	75.1	84.8	90.4
0.01	100.0	100.0	100.0

Conclusions

EXAFS spectroscopy has been used to determine the coordination environment of Ln3+ (Ln = Nd, Eu, Dy, Er, Yb) in ChCl-based DES’s. These data, combined with UV/vis absorption and TRLS data, show the important relationship between the Ln3+ coordination environment and HBD (Table ). All collected spectroscopic data suggest the presence of one species for each Ln3+ in each DES, and we have attempted to identify the complexes present based on the current understanding of the bulk molecular structure of the studied DES and literature data. The determined REE speciation illustrates the complexity of the metal behavior in DESs, whereby the intermolecular interactions between DES constituent molecules are very important in complex formation and must be considered alongside the stability of the formed complex.

Table 7

Identified Metal Species for REEs (Nd, Eu, Dy, Er, and Yb) in Choline Chloride-Based DES with Ethylene Glycol, Urea, and Lactic Acid as HBDs

	identified Ln3+ species
HBD	Nd	Eu	Dy	Er	Yb
ethylene glycol	[Nd(HBD)4.5–5]	[Eu(HBD)4–4.5]	[Dy(HBD)4–4.5]	[Er(HBD)4–4.5]	[Yb(HBD)4]
urea	[Nd(HBD)8–10]	[Eu(HBD)8–9]	[Dy(HBD)7–9]	[Er(HBD)7–8]	[Er(HBD)7–8]
lactic acid	[NdCl6]	[EuCl6–5]

The complexes determined for Ln3+ in ChCl–EG and ChCl–urea are consistent with the literature, based on Ln–O distances from previously published single-crystal XRD data. However, there is still some room for debate about exact CN’s and coordination geometry. This is starker for the complexes in the ChCl–lactic acid DES, where Ln–Cl distances are very similar to those found in the [LnCl6]3– octahedra. None of those Ln–Cl distances were measured for complexes associated with the DES in any way, where the complex intermolecular interactions could have unexpected effects. With this in mind, we recommend that future work in this area combines more spectroscopic studies of a wider range of Ln3+—DES systems with computational studies. We believe that this work helps to lay the foundations of these future studies, hopefully leading to the design of DES-based separation processes which can take advantage of the relative interaction strengths between DES moieties and dissolved REE cations.

Experimental Section

Research Design

This work employs a set of complementary spectroscopic techniques to gain a deeper insight into the speciation of REE’s in DES’s than could be obtained using these techniques individually. The EXAFS spectroscopy data tell us which atoms are in the coordination sphere of the central atom and at what distance. However, this does not give us information about the specific coordination geometry. The use of UV/vis absorption and time-resolved luminescence spectroscopies fills in this gap, providing information about how complex symmetry changes are when changing the DES and enabling the elucidation of coordination geometries. A range of REE’s were chosen (Nd, Eu, Dy, Er, and Yb), allowing for an understanding of how speciation changes across the row.

Reagents

LnCl3 (Ln = Nd, Eu, Dy, Er, Yb, anhydrous, ≥99.9%), LnCl3.6H2O (Ln = Nd, Eu, Dy, Er, Yb; ≥99.9%), choline chloride (98%), ethylene glycol (anhydrous, 99.8%), and urea (99.5%) were purchased from Sigma-Aldrich. l-lactic acid (anhydrous, 98%) was purchased from Alfa Aesar. All anhydrous chemicals were transferred to an Ar atmosphere glovebox [(O2), (H2O) ≤ 5 ppm] for use and storage. Choline chloride and urea were dried in a vacuum oven (24 h, 80 °C) before being transferred into the glovebox.

Deep Eutectic Solvent Production

All DES production was carried out in the glovebox. Choline chloride was mixed with either ethylene glycol, urea, or lactic acid at molar ratios of 2, 2, and 1 (HBD/ChCl), respectively.[9,30] These mixtures were then heated with stirring at 60 °C until a clear homogenous liquid was formed, which typically took ca. 30 min. Care was taken to minimize the heating time during DES production, as the ChCl–lactic acid DES is known to degrade through an esterification reaction.[30]

EXAFS Spectroscopy Measurements

EXAFS spectroscopy measurements were performed at the Pohang Accelerator Laboratory (PAL) using beamlines 6D, 7D, and 8C. Data were collected at room temperature in the fluorescence mode, using the L-III edge of Nd, Eu, Dy, Er, and Yb. LnCl3 salts were dissolved in the DES in a glovebox [(Ln3+) = 0.1 mol kg–1] before being transferred into sealed sample holders for analysis. Collected spectra were processed using the Demeter software suite.[31] Specifically, normalization was carried out using the Athena software package, and subsequent data fits were carried out using the FEFF database in the Artemis software package. The coordination number for each fitted shell was set, and the amplitude reduction factor was allowed to refine based on the model.

UV/Vis Spectroscopy

Stock solutions of NdCl3 and ErCl3 were made in choline chloride–ethylene glycol and −urea DES’s [(Ln3+) = 0.2 mol kg–1]. A stock solution of NdCl3 was also made in choline chloride–lactic acid [(Ln3+) = 0.06 mol kg–1]. Aliquots of each stock were taken and diluted appropriately for measurement. All sample preparation was done in the glovebox, and samples were only removed from the glovebox for measurement in lidded cuvettes. UV/vis absorption spectroscopy data were recorded using a Zeiss MCS 600 spectrometer.

Time-Resolved Luminescence Spectroscopy

Luminescence emission and lifetime were collected at room temperature for LnCl3 (Ln = Nd, Eu, Dy) in the studied DES [(Ln3+) = 0.01, 0.015, 0.02 mol kg–1]. Excitation wavelengths of 416 nm (pulse energy of 1.6 mJ), 453 nm (1.54 mJ), and 355 nm (4.9 mJ) were used for Eu, Dy, and Nd, respectively. A pulsed Nd:YAG laser at 355 nm (Continuum, Surelite) and a wavelength-tunable optical parametric oscillator (OPO, OPOTEK, Vibrant B) were used for TRLS studies. The laser pulse energy was measured using an energy meter (Coherent, EPM 2000 with a J25LP-MB detector), and the gated intensified charge-coupled device (ICCD, Andor, DH-720/18U-03 iStar 720D) was used to collect the time-resolved luminescence spectra. A more detailed explanation of the TRLS experiments is included in the Supporting Information.