Linear 2-Ethylhexyl Imidophosphoric Esters as Effective Rare-Earth Element Extractants

: Imidophosphoric organic esters containing phosphoryl groups are potential polydentate ligands and promising extractants of rare-earth elements. For their preparation, a monophosphazene salt [PCl 3 =N − PCl 3 ] + [PCl 6 ] − and short phosphazene oligomers of the general formula [Cl–(PCl 2 =N) n –PCl 3 ] + [PCl 6 ] − , where n = 4–7, were synthesized via living cationic polymerization of Cl 3 P=NSiMe 3 and used as starting compounds. All phosphazenes were reacted with 2-ethylhexanol to obtain the corresponding esters of imidophosphoric acids (EIPAs). The formation of imidophosphoric acids occurs due to the phosphazene-phosphazane rearrangement of –P(OR) 2 =N– or –P(OH)(OR)=N– units, where R = 2-ethylhexyl. The prepared EIPAs were characterized by 1 H, 31 P NMR, and MALDI-TOF analyses and their extractive capacity towards lanthanide ions in aqueous solutions of nitric acid was examined. The EIPAs are mixtures of mono-, di-, and trifunctional compounds of the type H x A, where x = 1–3, which can form chelate complexes of lanthanide ions [Ln(A) z ], where z = 3–6, depending on the chain length. The longer chain EIPAs are more suitable for collective rare-earth elements extraction. A comparison of the extraction properties of the EIPAs with the industrially used polyalkylphosphonitrilic acid (PAPNA) was drawn. selectivity on spacer length between terminal phosphonyl groups and the number of N–P=O fragments in extractant structure.

The first short-chain EIPAs were prepared by reaction between diphenoxyphosphamide (PhO)2P(O)NH2 and phosphorus pentachloride with further successive treating of the resulting products with formic acid and alkali metal alkoxides or phenolates (Scheme 1) [19]. Scheme 1. Imidophosphoric acid ester synthesis via Kirsanov reaction between diphenoxyphosphamide and PCl5 and further treating with formic acid and alkoxides or phenolates.
The simplest synthetic way to EIPAs is alcoholysis of imidodiphosphoric acid tetrachlorides (OPCl2)2NH with NaOR (R = alkyl, aryl) leading to the formation of the corresponding symmetrical esters [7,18]. Short-chain EIPAs can be prepared by direct reaction of trichlorophosphazodichlorophosphonyl (TCDP or dichlorophosphinylphosphorimidic trichloride) Cl3P=N-POCl2 with aliphatic alcohols using amine as an acceptor of HCl similarly as it usually used in aminophosphazene derivatives synthesis [20] or by treatment of the corresponding "pentaesters" with hydrochloric acid (Scheme 2) [21]. EIPAs with medium chain length or longer can be synthesized by the reaction of TCDP higher oligomer homologs with aliphatic alcohols or their alkoxides.  [22,23]. EIPAs also form lanthanide complexes similar to those with other metals (Figure 1b) [6,24]. However, reaction of {OP(OPh)2}2NH with phenylmercury hydroxide PhHgOH leads to formation of covalent N-Hg-Ph bonds along with coordination of one phosphonyl oxygen atom with a mercury atom of a neighboring molecule (in this case, a dimeric association in the crystal structure is observed, Figure 1c) [25]. EIPAs represent a large and diverse class of organophosphorus compounds, which are widely used as REE extractants due to their ability to form metal complexes. Di-(2-ethylhexyl) phosphoric acid (D2EHPA) and 2-ethylhexyl phosphoric acid mono-2-ethylhexyl ester (EHEHPA) are the most thoroughly studied among them [26][27][28][29]. Since REE are poorly stripped from D2EHPA complexes, EHEHPA is considered to be a perspective alternative with greater selectivity to REE and the possibility of stripping at lower acidities. However, EHEHPA has lower extraction efficiency in comparison with D2EHPA. The use of extractant blends of D2EHPA with various agents also demonstrates some improvements in extraction and separation processes [30][31][32] but does not eliminate all drawbacks of the base extractant. Hence, an active search for new highly efficient extractants and accurate study of their extraction properties are still of utmost actuality.
EIPAs containing 2-ethylhexyloxy groups at phosphorus atoms form complexes with various metals and exhibit high selectivity in the separation of REE by extraction methods [33,34]. A study of extraction properties of EIPAs can reveal the effects of their structure, the chain length of various EIPAs homologs and multidentate coordination ability on the extractability of REE from acidic solutions. Molecular weight characteristics of EIPAs can be precisely defined at the preliminary stage due to the synthesis of linear chlorophosphazene salts via living cationic polymerization [35][36][37][38] since the further substitution of chlorine atoms by RO − -units does not change their chain length. The EIPAs chain length control allows studying the dependency of extractive capacity and selectivity on spacer length between terminal phosphonyl groups and the number of N-P=O fragments in extractant structure.
In this paper, we report on the results of the survey conducted to determine structural aspects and extractive capacity of imidophosphoric esters, synthesized by derivatization of linear phosphazenes of various chain lengths.
The initial REE compounds were Y, Pr oxides (>99.5% pure, JSC "Rushim", Moscow, Russia), La, Ce, Nd, Sm, Gd, Dy, Yb nitrates (>99% pure, JSC "Rushim"), Ho chloride (>99% pure, JSC "Rushim"). Analysis of rare-earth elements for purity was carried out by comparing the results of titrimetric and gravimetric methods, as well as on Jobin Yvon 38P plasma spectrometer. Results of inductively coupled plasma spectroscopy analyses showed that the content of impurities in the REE salts did not exceed 0.01%. Corresponding metal oxides and chlorides were converted to nitrate form for extraction studies.
The solvents, dried according to standard procedures, were used in a freshly distilled form. All glassware was dried in an oven or flame-dried under vacuum and cooled in the inert atmosphere before use. All reactions were performed using standard Schlenk techniques.
Equipment. The 31 P, 13 C, and 1 H NMR spectra were recorded on a Bruker AM300 spectrometer at 121, 75, and 300 MHz, respectively. Mass spectra of MALDI-TOF were obtained on the "Bruker Auto Flex II" instrument.
The content of impurities in metal salts was estimated by ICP spectra recorded on the ICP Jobin Yvon 38P plasma spectrometer. The content of functional acidic groups in extractants was determined by potentiometric titration using a pH meter Mettler Toledo S400.
N-(trimethylsilyl)trichlorophosphoranimine Cl3P=NSiMe3 was synthesized as reported previously [40] by direct reaction of PCl5 and hexamethyldisilazane. Yield: 52%. 31  A solution of Cl3P=NSiMe3 (1.08 g, 4.80 mmol) in CH2Cl2 (20 mL) cooled to −55 °C was added dropwise to a solution of PCl5 (2 g, 9.59 mmol) in CH2Cl2 (50 mL) placed in a three-necked flask. The reaction mixture was allowed to warm to room temperature and stirred with a magnetic stirrer. The conversion of monomer to monophosphazene salt was monitored by 31 P NMR spectrometry. After complete conversion, all volatiles were removed under vacuum. The resulting product was obtained as a white-colored crystalline solid. Yield: 2.37 g (93%). 31  Oligodichlorophosphazene [Cl-(PCl2=N)n-PCl3] + [PCl6] − , where n = 4-7, was synthesized via living cationic polymerization of Cl3P=NSiMe3. A solution of Cl3P=NSiMe3 (9.39 g, 48.02 mmol) in CH2Cl2 (30 mL) was added dropwise to a solution of PCl5 (2 g, 9.59 mmol) in CH2Cl2 (50 mL) cooled to 0 °C in a three-necked flask. The reaction mixture was allowed to warm to room temperature and stirred with a magnetic stirrer. After complete conversion, confirmed by 31 P NMR spectroscopy, all volatiles were removed under reduced pressure. Resulting light-yellow transparent viscous liquid was obtained in 6.58 g 91% yield. According to MALDI-TOF, the distribution of phosphazene homologs in the final mixture was as follows: four (24%), five (39%), six (29%) and seven (7%) -P=Nunits with impurities content of not more than 0.5%.
Triphenoxyphosphazodiphenoxyphosphonyl (PhO)3P=N-(PhO)2P=O, was purchased from Elortech, LLC and was used without purification, Tm = 73 °C, 31  . Then freshly distilled water (67.7 mg, 3.76 mmol) was added dropwise. The reaction mixture was stirred at room temperature for 1 h. The solvent was removed under reduced pressure and the resulting mixture was analyzed by 31 P NMR spectroscopy and consisted of Cl3P=N-POCl2 and POCl3. 31  The reaction between [Cl3P=N-PCl3] + [PCl6] − and sodium trifluoroethylate was carried out similarly to the aforementioned partial hydrolysis of monophosphazene salt.

Synthesis of 2-ethylhexyl derivatives of imidophosphoric acids based on TCDP.
A threenecked flask equipped with a magnetic stirrer and a reflux condenser with a CaCl2 tube was charged with a solution of TCDP (10.28 g, 0.038 mol) in dioxane (40 mL) and a solution of 2-ethylhexanol (29.8 g, 0.229 mol) in dioxane (50 mL). The reaction was carried out at 100 °C with stirring for 48 h with HCl purge with dry argon. At the end of the process, the solvent was distilled off using a rotary vacuum evaporator. The resulting liquid mass was dissolved in 150 mL of chloroform and repeatedly washed with distilled water in a separatory funnel. After vacuum distillation, the product was a clear, oily liquid with a yellow hue (21.1 g).
Synthesis of 2-ethylhexyl derivatives of imidophosphoric acids based on oligodichlorophosphazene. A three-necked flask equipped with a magnetic stirrer, a reflux condenser with calcium chloride tube, and under a positive flow of argon was successively loaded with a solution of oligodichlorophosphazene (4.37 g) in dioxane (20 mL) and a solution of 2ethylhexanol (11.41 g) in dioxane (25 mL). The further process was carried out in a similar way as for TCDP. The product obtained was a clear oily yellowish liquid (8.04 g).
Determination of EIPAs and PAPNA extraction characteristics. The REE ions concentration in the initial and equilibrium aqueous phases was determined by the titrimetric method. A 250 mL conical flask was successively loaded with an aliquot of REE solution, 1% aqueous solution of ascorbic acid (2-3 drops), acetate buffer solution (20 mL, pH = 5.2-5.4), 1% aqueous solution of xylenol orange (3-5 drops) and diluted with water to 100 mL. The prepared solution was titrated with ethylenediaminetetraacetic acid disodium salt dihydrate solution (0.05 mol/L) until the color changed from violet to yellow.
Acetate buffer solution was prepared according to the following procedure: acetic acid (32 mL) was mixed with 25% ammonia solution (34 mL). Then it was diluted with distilled water to 1000 mL in a volumetric flask.
To calculate the concentration of metal ions in the aqueous phase the following Equation (1) was used: where C1 is the concentration of metal ions in the aqueous phase, mol/L; C2 is the concentration of ethylenediaminetetraacetic acid disodium salt dihydrate, mol/L; V1 is the volume of an aliquot of the aqueous phase, ml; V2 is the volume of titrant spent on the experiment, ml. The error in determining the concentration of REE ions in aqueous nitric acid solutions by complexometric titration does not exceed 3%.
The equilibrium concentration of REE ions in the organic phase was calculated from the difference in equilibrium concentrations (Equation (2)): where Сorg is the equilibrium concentration of REE in the organic phase, mol/L; Cinitial is an initial concentration of REE in the aqueous phase, mol/L; Сaq is the equilibrium REE concentration in the aqueous phase, mol/L. Extraction efficiency can be estimated as Corg to Сinitial ratio. Conclusions about the formed complexes can be made from the ratios of initial extractant concentration to Corg [28].
The concentration of the extractant was determined by potentiometric titration with a solution of NaOH. A 100 mL glass beaker was successively loaded with an aliquot of the extractant and 50% aqueous solution of isopropanol (50 mL). Electrodes were placed in a prepared solution and titrated using a magnetic stirrer. Based on the obtained titrimetric data, a graph was plotted in E/V coordinates. The calculation of the concentration of extractant was performed according to Equation (3): where Cext is the concentration of the extractant in the initial mixture, mol/l; VNaOH is the equivalent volume of a titrant solution corresponding to a jump in the titration curve, ml; CNaOH-titrant concentration, mol/L; Valiq is the volume of an aliquot of the extractant, ml.
The procedure of REE extraction with EIPA-1. 2 g (3.3 mmol) of EIPA-1 was dissolved in 10.5 mL of RED-3 solvent so that a final concentration of H + ions = 0.3 mol/L. The system was stirred until complete dissolution of extractant. The ratio of aqueous and organic phases has always been 1:1. The extraction was carried out in disposable plastic tubes for 15 min on a shaker in a horizontal position. All REE ions were taken in 0.25-0.3 mol/L concentration and were extracted from 0.1M nitric acid aqueous solutions. After removal from the shaker, a clear separation of the phases was achieved in 5-15 min, depending on the process conditions. Aliquots of the organic and aqueous phases for analysis were taken directly from the tubes.
REE extractions with EIPA-2 and PAPNA were carried out similarly to the aforementioned procedure of extraction with EIPA-1 with all the same concentrations. All quantities (RED-3 solvent, REE ions, nitric acid, water) were calculated based on the initial concentration of H + ions in extractants determined by potentiometric titration.

Results and Discussion
One of the most recognized 2-ethylhexyl derivatives of imidophosphoric acid used as extractant is polyalkylphosphonitrilic acid (PAPNA), but despite its widespread industrial use, generally in Eastern Europe countries, its structure has not been fully established and analytical data is almost absent. Scanty literary data about PAPNA chemical structure varies a lot ( Figure 2) [33,34,41]. Due to these differences in proposed structures, the relationship between the chemical structure of EIPAs and their effectiveness in REE extractive separation remains unclear.
Within our investigation, we carried out the syntheses of EIPAs in order to provide an exhaustive description of their molecular structures and determine their ability to complex formation with REE ions, as well as the possibilities for these ions separation.
Next, starting materials were used for synthesis of the desired EIPAs: TCDP as a model compound and linear oligodichlorophosphazene [Cl-(PCl2=N)n-PCl3] + [PCl6] − , where n = 4-7, as a substrate for the preparation of PAPNA analogue, with the same chain length according to the literature data [41]. The proposed reagents were considered to be prospective candidates for EIPAs preparation due to the possibility of well-known phosphazene-phosphazane rearrangement (PPR) in their units. The driving force of such an arrangement is the supposed high susceptibility of hydrogen or α-carbon atoms in phosphate-like groups, which are attacked by the most basic nitrogen atom of the main chain [42].
The occurrence of PPR and molecular structures of resulting isomers was determined by NMR analysis. MALDI-TOF mass spectrometry was additionally used for confirmation. PPR is observed already at 25 °C for various phosphazenes with hydroxyl groups at the phosphorus atoms. For monosubstituted hydroxy or alkoxy derivatives of TCDP, PPR leads to a significant change in the character of the signals in their 31 P NMR spectra. Magnetically nonequivalent phosphorus atoms of hydroxyphosphazo (alkoxyphosphazo) and phosphonyl groups, initially characterized by two doublet signals, become magnetically equivalent with only one singlet in spectra after the rearrangement due to the formation of the second phosphonyl group.
In the products, containing 2-ethylhexyl substituents, the rearrangement proceeds very slowly at 25 °C. Moreover, as it can be seen from the 31 P NMR spectrum ( Figure S1) of the product obtained via TCDP and an excess of 2-ethylhexanol reaction at room temperature, the substitution of chlorine atoms itself is insignificant. In accordance with NMR data the main product is a mono-substituted non-rearranged derivative of TCDP (1), characterized with two doublet signals: at −0.2 ppm (phosphazenyl atom) and −11.2 ppm (phosphonyl atom). The temperature increase leads to the formation of products with a higher degree of chlorine atoms substitution as well as to PPR.
At elevated temperatures, PPR probability (70 °C and higher) approaches 100% even for the mono-substituted phosphazene and decreases slightly when all chlorine atoms in chlorophosphazene have been replaced by 2-ethylhexyl substituents. Thus, the resulting product of the reaction between 1 mol of TCDP and 0.75 mol of 2-ethylhexanol at 70 °C in dioxane is actually a mixture of the original TCDP and its mono-substituted derivative, tetrachlorodiphosphonyl-N-(2ethylhexyl)-amine (2), which has a singlet signal at -5.2 ppm in the 31 P NMR spectrum ( Figure S2) due to the magnetic identity of the phosphorus nuclei. The formation of 2, in this case, occurs according to Scheme 3. It should be noted that along with the PPR, short linear oligophosphazenes are susceptible to an adverse reaction of an attacking agent with α-carbon atoms of organic substituents, mainly those located at terminal units, resulting in the elimination of the side ether R-O-R and the formation of -O − Na + fragments [43]. The subsequent replacement of the sodium atom by hydrogen during purification leads to the formation of -N=(RO)2P-OH groups, which proceeds further with PPR to yield the aminophosphonyl form -NH-(RO)2P=O.
To show the equivalence of the structures obtained by nucleophilic substitution of chlorine atoms in the TCDP molecule and linear oligophosphazene, the reaction of the model shortest monophosphazene [Cl3P=N-PCl3] + [PCl6] − was carried out and subsequent interaction of the latter with water and sodium alkoxide was studied. The composition of the reaction mixture was characterized by 31 P NMR spectroscopy.
Partial hydrolysis of the phosphazene salt [Cl3P=N-PCl3] + [PCl6] − was carried out in dichloromethane media at room temperature. The molar ratio of phosphazene salt to H2O was 1:2 in order to replace one chlorine atom in the cation and to neutralize the counterion (Scheme 4). Hydrolysis leads to the ion pair destruction with the formation of the phosphorus oxychloride molecule. The system of doublets in the regions δP = −3.1 and −11.6 ppm in the 31 P NMR spectrum explicitly characterizes TCDP. The absence of signals of TCDP hydrolysis by-products in the spectrum indicates lower reactivity of phosphazophosphonyl compared with the monophosphazene salt.
The same products (TCDP, POCl3) and trifluoroethyl chloride were yielded from the reaction of 1 mole of [Cl3P=N-PCl3] + [PCl6] − and 2 moles of sodium trifluoroethylate. In this case, the formation of the terminal phosphonyl group accompanied by trifluoroethyl chloride elimination (Scheme 5). Thus, the nucleophilic substitution of chlorine atoms in [Cl(PCl2=N)n-PCl3] + [PCl6] − oligomers with alkoxy fragments can be considered to be a method for the direct preparation of phosphazophosphonyls of a given length.
Syntheses of EIPAs were carried out at 100 °C in dioxane (Schemes 6 and 7). A small, broad signal at 0 ppm in the 31 P NMR spectrum of EIPA-2 showed a significant number of overlapping signals of the oligomers with different chain lengths. The significant broadening of the signals in the NMR spectra of PAPNA (Figure 3c) in comparison with EIPA-2 is obviously due to the wider molecular weight distribution of the first.
The MALDI-TOF mass spectrum of EIPA-1 ( Figure 5) is in a good agreement with the chemical structures of the products described above, determined by 31 P NMR spectroscopy.
Structures in Table 1 are given in accordance with the NMR spectroscopy data. Fragments of the phosphazene form -P(OR)2=Nare not excluded. MALDI-TOF analysis of the industrial extractant PAPNA shows the presence of shorter-chain oligomers ( Table 2). The extraction capacity of EIPA-1, EIPA-2, and PAPNA as a reference sample, was estimated by change in REE ions concentration in the aqueous phase during extraction. The decrease in the concentration of REE ions in the aqueous phase is the result of their transition to the organic phase (RED-3 diluent was used). REE were extracted from nitric acid aqueous solutions (Cinitial = 0.25-0.3 mol/L, CHNO3 = 0.1 mol/L) using an organic solution of the corresponding extractant (Cext = 0.3 mol/L).
The extractant concentration in the organic phase corresponds to the concentration of H + ions, which can be replaced by REE ions. A volume ratio of organic and aqueous phases was 1:1. It is clearly shown (Figure 6) that EIPA-1 and EIPA-2 compounds are characterized by higher extraction ability and lower selectivity than PAPNA during REE extraction.       To confirm the participation of P-OH and NH groups in the complex formation and determine the role of P=O groups, the extraction ability of model triphenoxyphosphazodiphenoxyphosphonyl (PhO)3P=N-(PhO)2P=O containing only one P=O group was studied. It turns out that this compound extracts acids from aqueous solutions well, but does not extract REE in any quantities. Thus, the sole presence of functional fragments of the P=O type does not determine the extraction capacity towards REE in acidic conditions. The calculation of the extractant and REE concentrations ratios studied in the organic phase after extraction (Table 3) showed a higher selectivity of EIPA-1: heavier elements are more easily transferred to the organic phase. The reason for this is the higher steric availability of large-diameter lanthanide atoms (Figure 7a, shows a complex with a tetra-substituted TCDP derivative). The proposed structure is generally recognized for the complexes of this type of compounds (with two phosphonyl groups) with a light group of REE [5,23]. The aforementioned extractant to REE concentrations ratio in the organic phase very accurately confirm the formation of complexes of precisely 3 EIPA molecules and 1 metal ion. A similar structure with phenyl substituents at phosphorus atoms was described by Efimova et al. [44]. The small size of the lanthanide atom presumably can lead to the formation of complexes containing up to 6 fragments of extractant molecules (Figure 7b), in which the formation of hydrogen bonds between P=O and H-N occurs. This is confirmed by the higher ratios of CEIPA-1: Corg with values of 5.03 for La and 5.08 for Ce (Table 3). Chen et al. previously described the analogous structure of the REE complex with 2ethylhexylphosphonic acid mono-2-ethylhexyl ester [28].
EIPA-2 compounds form complexes with metal salts through the -P(O)(OR)-NH-functional groups, the average number of which is 2-3 per 1 molecule (Table 1). Ratios CEIPA-2: Corg for different REE ions slightly differ and vary from 3 to 3.5 with the exception of Ho (ratio CEIPA-2: CHo = 4.00).  Taking into account that CEIPA-2 corresponds to the concentration of H + ions, 3-4 EIPA-2 molecules can be combined with 1 REE ion depending on the number of phosphorus atoms in their chain. On average, only one functional group per molecule can participate in complex formation, while other functional groups tend to coordinate with each other in the organic phase.
The extraction of REE by EIPA-2 is a typical cation exchange process, expressed (with the exception of La and Ho) as follows: Ln 3+ + 3HA = [Ln(A)3] + 3H + . The remaining functional -P(O)(OR)-NH-fragments not directly interacting with the REE ion and non-functional -P(O)(OR)-NRfragments of these molecules are located in the outer sphere of the metal-containing complex. Thus, the properties of the complex will depend to a greater extent on fragments located in the outer sphere of the complex and to a lesser extent will be defined by the nature of the REE ion itself. As a result, the extraction selectivity of EIPA-2 is very weakly expressed with respect to certain REE ions. The organic phase is equally saturated with all metals. Metal concentration Corg after extraction is close to 0.1.

Conclusions
EIPAs can be prepared by reacting linear phosphazenes with alcohols of various structures, in particular, 2-ethylhexanol. The number of phosphorus-containing units in the chain does not change during the substitution of chlorine atoms in linear oligophosphazenes with 2-ethylhexyl groups. Therefore, the chain length of EIPAs can be set using the starting chlorophosphazene with the selected chain length. The chain length of the initial linear chlorophosphazenes can be easily adjusted due to the living cationic polymerization of Cl3P=NSiMe3.
Analysis of the chemical structure of compounds presented in the samples of polyalkylphosphonitrilic acid (PAPNA), EIPA-1 (Scheme 6) and EIPA-2 (Scheme 7) shows that the selectivity of PAPNA towards REE ions primarily depends on the presence of functional dihydroxyphosphonylamines with structures 5 and 6 ( Table 2).
A significant content of non-functional fragments -P(O)(OR)-NR-along with functional -P(O)(OR)-OH or -P(O)(OR)-NH-groups in EIPAs contributes to a high collective extraction ability to REE. These fragments are observed mainly in EIPA-2 but also present in the PAPNA.
Oligomeric 2-ethylhexyl esters based on linear phosphazenes, considered to be PAPNA analogues, show a clearer ability to collective REE extraction. An increase in the chain length of the ester leads to a better extraction capacity towards light lanthanides, but a decrease in their selectivity.
Supplementary Materials: The following are available online at www.mdpi.com. Figure S1: The 31 P NMR spectrum of the products prepared by the reaction of TCDP and 2-ethylhexanol (20% excess) at 22 °C in dioxane. Figure

Conflicts of Interest:
The authors declare no conflict of interest.