Intralanthanide Separation on Layered Titanium(IV) Organophosphate Materials via a Selective Transmetalation Process.

The lanthanides (Ln) are an essential part of many advanced technologies. Our societal transformation toward renewable energy drives their ever-growing demand. The similar chemical properties of the Ln pose fundamental difficulties in separating them from each other, yet high purity elements are crucial for specific applications. Here, we propose an intralanthanide separation method utilizing a group of titanium(IV) butyl phosphate coordination polymers as solid-phase extractants. These materials are characterized, and they contain layered structures directed by the hydrophobic interaction of the alkyl chains. The selective Ln uptake results from the transmetalation reaction (framework metal cation exchange), where the titanium(IV) serves as sacrificial coordination centers. The "tetrad effect" is observed from a dilute Ln3+ mixture. However, smaller Ln3+ ions are preferentially extracted in competitive binary separation models between adjacent Ln pairs. The intralanthanide ion-exchange selectivity arises synergistically from the coordination and steric strain preferences, both of which follow the reversed Ln contraction order. A one-step aqueous separation of neodymium (Nd) and dysprosium (Dy) is quantitatively achievable by simply controlling the solution pH in a batch mode, translating into a separation factor of greater than 2000 and 99.1% molar purity of Dy in the solid phase. Coordination polymers provide a versatile platform for further exploring selective Ln separation processes via the transmetalation process.

Introduction

The lanthanide (Ln) series, collectively made up of a group of 15 elements, has gained strategic importance in recent decades. The unique 4f electron structures entail distinctive and, in many cases, irreplaceable physical properties that make Ln essential components in advanced electronics, lasers, and permanent magnets.[1,2] Owing to the poor shielding of nuclear charge by 4f electrons, Ln3+ ions share extremely similar yet descending ionic radii across the series with an average difference between adjacent elements of only 1 pm (Ln contraction).[3] Consequently, Ln forms isostructural mineral compounds that are almost always found together in the geosphere.[4] Despite their considerably similar chemical characters, the magnetic and optical properties of Ln differ, and as a result, obtaining pure fractions of individual Ln is crucial for specific applications. The efficient separation of Ln remains a challenging task.

Historically, fraction crystallization was adopted to produce pure Ln.[5] This is proven to be tedious and time-consuming because hundreds of cycles are required. During the Manhattan project, ion-exchange chromatography utilizing organic resins was developed to separate Ln and actinides.[6] The difficulties in continuous operation and resin regeneration restricted its further scalability. Nowadays, continuous liquid–liquid extraction (solvent extraction) is the state-of-the-art technology for Ln separation on an industrial scale.[6] The separation is enabled by complex-forming extractants that selectively transfer the Ln ions from the aqueous to organic phases. Nevertheless, the inevitable liquid waste generation in the solvent extraction process has resulted in a search for more environmentally friendly separation methods.[7]

Recent advances in f-element coordination chemistry have reopened the window for selective crystallization, (bio)mineralization, and supramolecular self-assembly.[8−14] The small difference in ionic radii can be amplified during the crystallization with desirable multidentate ligands, resulting in metal–organic framework (MOF) materials with fine-tuned sizes, shapes, and crystal structures.[15] Often, these processes are energy-intensive because they are carried out hydrothermally. Biomineralization is active under milder conditions, yet the stability and reusability of artificial peptides are not satisfactory. Another approach—solid-phase extraction—completely eliminates the organic phase in solvent extraction. Organic–inorganic hybrid materials utilizing an array of inorganic porous supports and organic functional ligands have been developed for selective Ln sorption.[16−20] Recently, we have shown that titanium alkyl-phosphate functionalized mesoporous silica possesses solvating extraction capability.[21] The surface alkyl chains linked to titanium phosphate moieties mimic the structure of tri-n-butyl phosphate (TBP) and hence form complexes with Ln nitrates. However, these materials seem not to be perfect candidates for intralanthanide separation because of the similar complex formation constants across the Ln series.

Inorganic ion exchangers, namely, metal(IV) phosphate materials, are used for selective metal separation and decontamination.[17,22,23] Purely inorganic titanium[24] and zirconium phosphate[25] (TiP and ZrP) materials were studied for rare-earth separation via cation exchanges with the proton in the hydrogenphosphate groups. This reversible extraframework cation exchange reaction does not disrupt the building blocks of the material, and therefore the metal selectivity is primarily based on ion-sieve selection and phosphate coordination. Whereas inorganic phosphate offers limited Ln selectivity, simple organic phosphoric acids [e.g., di-n-butyl phosphoric acid (HDBP)[26] and di(2-ethylhexyl)-phosphoric acid (DEHPA)[27,28]] demonstrate more promising intralanthanide separation potentials. The ratio of the solubility product constants (Ksp) between Dy(DBP)3 and Nd(DBP)3 is reported to be more than 4000 (DBP is the deprotonated form of HDBP).[26] Titanium(IV) forms insoluble, layer-structured precipitates with DBP in the form of a coordination polymer.[29] The resultant material is composed of alternating bilayers of butyl chains and titanium(IV) phosphate. Although titanium(IV) and zirconium(IV) phosphonates were intensively studied,[30,31] there has been extremely limited information regarding the properties and application of titanium(IV) organophosphate materials in metal separation.

In the present work, we describe the synthesis and characterization of layered titanium(IV) butyl phosphate materials and their uptake behavior toward Ln3+ ions. Contrary to what we observed from the surface alkyl-phosphate grafted materials, the titanium(IV) butyl phosphate materials do not retain Ln3+ in a solvating extraction manner. Instead, the driving force for Ln3+ uptake originates from transmetalation reactions [framework Ti(IV) exchange]. Significant group separation was achieved for early and late Ln. This observation forms the basis of a Ln separation strategy that combines both the high selectivity from the crystallization method and the operational convenience from solid-phase extraction.

Experimental Section

Chemicals

Titanium(IV) tetrachloride (TiCl4, >97%) and HDBP (>97%) were purchased from Sigma-Aldrich. Lanthanide (including La, Ce, Pr, Nd, Eu, Gd, Tb, Dy, Yb, and Lu) solutions were prepared from their corresponding nitrate salts (>99.9%, Sigma-Aldrich). ortho-Phosphoric acid (>99.9%) and ethanol (>99.5%) were purchased from VWR Chemical. For metal (uptake) analysis, elemental standard solutions (1000 mg L–1, PrimAg-plus cert. ref. material) and nitric acid (SpA super purity, 67–69%) were obtained from ROMIL (Cambridge, UK). Ultrapure water (Milli-Q, Millipore) with a resistivity of 18.2 MΩ cm was used.

Synthesis of the TiP Materials

A precipitation method was employed for the synthesis of the TiP materials. By carefully dissolving TiCl4 in water under vigorously stirring (1000 rpm, ∼1 h), a completely transparent Ti(IV) precursor solution (1 M) was obtained. A series of phosphate precursor solutions (0.2 M total P) was prepared by diluting different amounts of 85% H3PO4 and HDBP into a 50/50 (v/v) water/ethanol mixture, with the final H3PO4-to-HDBP molar ratios at 1:0, 3:1, 1:1, 1:3, and 0:1. The precipitation reaction was carried out by the dropwise addition of 10 mL of the Ti(IV) precursor into 100 mL of the phosphate precursors while stirring (1000 rpm). Note that the P/Ti molar ratio in all synthesis liquors was fixed to 2. The suspension was aged at room temperature for 1 h under the same stirring condition. The resulting white slurry was then washed three times with the water/ethanol mixture through centrifugation (2 min, 5000g) and redispersion. Finally, the materials obtained were air-dried at 60 °C in an oven overnight. The materials synthesized from the phosphate precursors with different inorganic-to-organic phosphate ratios were denoted as TiP_x/y, where x/y = n(H3PO4)/n(DBP) = 1:0, 3:1, 1:1, 1:3, and 0:1 (Table ). All of them were ground and sieved to a particle size between 74 and 149 μm (100–200 mesh) before further study.

Table 1

Formulae and Molecular Mass for TiP Materials

		inorganic-to-organic P ratio
sample	compound formulae	in synthesis liquor	in obtained material	molecular mass (g mol–1)
TiP_1:0	TiO0.02(OH)2.88(HPO4)0.23(H2PO4)0.62·0.25H2O	∞	∞	183.9
TiP_3:1	TiO0.43(OH)1.84(HPO4)0.32(H2PO4)0.39(DBP)0.27·0.24H2O	3	2.63	215.4
TiP_1:1	TiO0.99(OH)0.55(HPO4)0.24(H2PO4)0.28(DBP)0.71·0.31H2O	1	0.73	277.4
TiP_1:3	TiO1.05(OH)0.13(HPO4)0.22(H2PO4)0.07(DBP)1.26·0.10H2O	0.33	0.23	360.2
TiP_0:1	TiO1.06(OH)0.01(H2PO4)0.24(DBP)1.63	0	0.15	429.3

X-ray Diffraction

The powder X-ray diffraction (XRD) patterns in the Bragg–Brentano geometry were recorded by a PANalytical X’Pert3 PW3710 MPD diffractometer control unit, a PW3020 vertical goniometer, and a monochromatic Cu Kα radiation source (λ = 1.54056 Å) operated at 40 kV and 40 mA.

Infrared Spectroscopy

Infrared (IR) spectra (650–4000 cm–1) were recorded on a PerkinElmer Spectrum One Fourier transform infrared (FTIR) spectrometer equipped with a Universal attenuated total reflection (ATR) sampling accessory.

Raman Spectroscopy

Raman spectra were recorded using a confocal Raman microscope (NTEGRA Spectra, NT-MDT) in the 100-to-1200 cm–1 region with a spectral resolution of 2 cm–1. The measurements were performed with a 532 nm Nd:YAG laser (output power 20 mW) and a 100× objective. The acquisition time was typically 1 min. We checked that the laser light did not damage the sample.

Surface Area

The Brunauer–Emmett–Teller (BET) surface area and porosity of the materials were determined at 77 K by the nitrogen (N2) adsorption–desorption method (TriStar II Plus, Micromeritics), and the samples were degassed at 150 °C for 5 h prior to the measurements (VacPrep 061 degassing unit).

Scanning Electron Microscopy

The surface morphology of the TiPs was observed using a Hitachi S-4800 FE-SEM (field-emission scanning electron microscopy) after they were sputter-coated with a 3 nm layer of Pd–Au alloy.

Compositional Analysis

The elemental contents (C and H) were analyzed using a Thermo Scientific Flash 2000 elemental analyzer. The contents of Ti and P were determined by a PerkinElmer Optima 8300 inductively coupled plasma optical emission spectrometer (ICP-OES) after total digestion in 65% HNO3 and 1% HF using a CEM MARS 5 microwave digestion system.

Thermogravimetry

The simultaneous thermogravimetry and differential scanning calorimetry–mass spectroscopy (TG/DSC–MS) analysis was performed using a simultaneous TG/DSC apparatus (STA 449F3 Jupiter, Netzsch) connected to a JAS-Agilent GC–MS (7890B GC/MSD5977A). In the dynamic TG measurements, the samples were heated from 30 to 800 °C at a heating rate of 20 °C min–1 under a helium gas flow of 40 mL min–1.

Solid-State Nuclear Magnetic Resonance

The solid-state 13C and 31P magic-angle spinning (MAS) nuclear magnetic resonance (NMR) spectra were collected on a Bruker AVANCE III 500 MHz spectrometer equipped with a 4 mm H/X/Y MAS probe. The samples were filled in a 4 mm zirconia rotor and measured at a MAS rate of 12 kHz. 13C spectra were recorded with 1H–13C cross-polarization (CP) pulse sequence, 2048 scans, a recycle delay of 5 s, spinal-64 1H decoupling, and a contact time 0.5 ms. The spectra were referenced with respect to adamantane at 38.5 ppm. 31P spectra were acquired with a 90° pulse (77 kHz rf), a 100 s recycle delay, and 64 scans. The 31P chemical shifts were referred to external 85% H3PO4 at 0 ppm.

Metal Concentration Determination

Determination of the metals was performed on an Agilent microwave plasma-atomic emission spectrometer 4200. Depending on suitability, one of Sc, La, or Fe was used as the internal standard for quality control. Samples were diluted to concentrations below 25 mg L–1, and a linear calibration curve was established by 0, 0.5, 1, 2, 5, 10, and 25 mg L–1 standard solutions.

Ln3+ Uptake Study

The Ln3+ ion-exchange extraction was studied in a batch mode. Typically, 50 ± 1 mg of TiP material was placed in a polyethylene vial with 10 mL of Ln3+-containing solution. Samples were equilibrated for 48 h (unless otherwise stated) by constant rotary mixing (50 rpm). The solid/liquid separation was then achieved by filtering through a 0.22 μm poly(vinylidene difluoride) syringe filter, and the clear solution was pipetted for concentration determination. Equilibrium pH was measured from the remaining filtrate.

The distribution coefficient (Kd, mL g–1) demonstrates the distribution of a Ln element (or a group of selected Ln elements) between the solution and the solid material. It was calculated by eq :where is the Ln3+ concentration in the solid (mg g–1) after the sorption, [Ln3+]i and [Ln3+]t are, respectively, the Ln3+ concentration of the solution before and after sorption (mg L–1), respectively, V is the volume of the solution (10 mL), and m is the mass of the material (0.05 g).

The separation factor (SF) between two Ln3+ ions or two groups of Ln3+ ions (Ln1 and Ln2) was subsequently calculated based on their Kd values (eq ):

Results and Discussion

Synthesis and Characterization of the TiP Materials

TiCl4 is prone to hydrolysis even in acid media, and the stable species of titanium in our synthesis liquor is TiO2+ (eq ), according to the Pourbaix diagram.[32]

The TiO2+ cations are then polymerized to form (TiO)2 chains. Two kinds of phosphate precursors were employed in the synthesis, inorganic ortho-phosphoric acid (H3PO4) and organic HDBP. Pure inorganic TiPs are a group of −H2PO4 and/or −HPO4 containing proton-type ion exchangers, which have been extensively studied.[33] The addition of HDBP to the synthesis liquor introduced the butyl phosphate functional groups onto the amorphous TiP assembly based on the inorganic backbone (Ti–O–P) by condensation reactions. In our work, among the TiP materials synthesized (Table ), TiP_1:0 is a purely inorganic material, whereas TiP_0:1 should supposedly contain only DBP groups. The others are categorized into organic–inorganic mixed TiPs.

Inorganic TiP_1:0 exhibited a typical amorphous phase with a broad low-intensity powder XRD peak (Figure a). The addition of HDBP resulted in the evolution of a peak at 2θ = 5.9–6.3°, which translates into an interlayer spacing of 1.41–1.51 nm. This observation corresponds with an earlier study and suggests that these materials contain self-assembled butyl group layers stacked via van der Waals interactions.[29] Despite the layered structure, the material lacks long-range periodic order. This is mainly due to the high reactivity of Ti(IV), leading to the formation of octahedral Ti-oxo-clusters and consequently amorphous coordination polymers. The high charge and polarizing power of Ti(IV) contribute to the chemical stability of the coordination polymer.[34] A similar situation applies to Zr(IV).[31]

Figure 1

Characterizations of the synthesized TiP materials. (a) Powder XRD patterns. (b) ATR–FTIR spectra. (c) Raman spectra with crystalline α-TiP as a comparison. (d) Nitrogen adsorption–desorption isotherms measured at 77 K. (e) SEM images (smaller magnification in the left column and higher magnification in the right column).

The ATR–FTIR peaks (Figure b) centered at 2959, 2934, 2906, and 2874 cm–1 are assigned to the stretching vibration of C–H in butyl groups. The 1465 cm–1 band represents the deformation vibration of the same groups.[29] The IR absorptions at 1023 and 733 cm–1 are typical vibrational modes for C–O–(P) in phase vibration and P–O3 symmetrical stretching, respectively.[35] The water content characterized by the O–H vibration at 1623 cm–1 becomes less evident with the increased organic content, as the hydrophobicity increases. The Raman spectra (Figure c) revealed additional information regarding the linkage between tetrahedral PO4 and octahedral TiO6 groups. The broad peaks at about 730 cm–1 in TiP_0:1, TiP_1:3, and TiP_1:1 materials are assigned to the vibration of Ti–O–Ti chains that contain nonbridging oxygen.[36−38] This peak moves toward higher Raman shifts and becomes less evident in TiP_3:1 and TiP_1:0 samples, and it is completely absent in crystalline α-TiP. We therefore infer that the heterocondensation between Ti–O–H and P–O–H groups competes with the homocondensation process for Ti–O–H during the synthesis. HDBP possesses only one P–O–H group, whereas H3PO4 harbors three. Because we fixed the overall P/Ti molar ratio at 2, the homocondensation reaction is expected to be more evident in the synthesis precursor with a higher amount of HDBP. In general, the Raman spectra of the TiPs exhibit an amorphous nature without any sharp peaks. The stretching vibrations of the isolated phosphate groups (in the 1000 cm–1 region) are only observable from TiPs with a higher inorganic phosphate content (TiP_1:0 and TiP_3:1).

N2 porosimetry analysis showed that the TiP samples have different textural characteristics, as summarized in Table . For the organic-containing TiP materials except TiP_3:1, the N2 adsorption–desorption isotherms (Figure d) are type II with very small hysteresis of type H3, suggesting monolayer–multilayer adsorption and aggregation of particles in these TiP materials. However, TiP_3:1 and purely inorganic TiP_1:0 sample gave a type IV(a) isotherm with a hysteresis in the p/p0 range of 0.45–1.0 and 0.7–1.0, respectively,[39] which is associated with capillary condensation taking place in their respective mesopores. In TiP_1:0, the hysteresis loop is H2(b), associated with pore blocking, with larger size distribution of the next widths, whereas in TiP_3:1, the change in hysteresis to type H3, due to aggregation of particles. The addition of small amount of organophosphate (in TiP_3:1) increased the surface area and decreased the average particle size (Table ). As a general trend, the BET surface area decreases drastically with the synthetic addition of organophosphate. The TiP_0:1 sample has a strikingly low surface area (1.2 m2 g–1), whereas the inorganic amorphous TiP_1:0 sample presents a contrasting value of 142 m2 g–1. The self-assembly of the butyl groups into a lamellar format resulted in lower pore volume, larger average pore diameter, and significant growth (aggregation) of the particles. This is consistent with the morphological observation by SEM images in Figure e. These TiP materials appear as irregular pellets that are connected to each other. It is worth noting that the inorganic TiP_1:0 sample exhibits irregularly shaped large lumps under microscopic observation. The addition of small amount of organophosphate (in TiP_3:1) increased the surface area and decreased the average particle size. The capillary condensation of N2 shifts to lower relative pressure from the TiP_3:1 to TiP_0:1 sample, reflecting a decrease in the pore volume due to the introduction of organophosphate. The structural and morphological design of the inorganic TiP materials would be beneficial when small amounts of organophosphate are added as structural modifiers. Nevertheless, this is not the target of the current work and would require further and detailed investigation.

Table 2

Textural Parameters of the Synthesized TiPs

sample	SBET (m2g–1)a	Vtotal (cm3g–1)b	Dpore (nm)c	Dparticle (nm)d
TiP_1:0	142	0.62	15.1	42.3
TiP_3:1	184	0.43	14.6	32.7
TiP_1:1	16.0	0.12	38.7	375.0
TiP_1:3	6.5	0.05	40.0	916.6
TiP_0:1	1.2	0.01	72.0	5168.2

SBET: BET surface area.

Vtotal: total pore volume estimated from the desorption branch by the Barrett–Joyner–Halenda (BJH) method.

Dpore: average pore diameter estimated from the desorption branch by the BJH method.

Dparticle: average particle size calculated by density functional theory.

In order to solve the formulae of the obtained TiP materials, an array of supplementary characterizations was made. The carbon (C), hydrogen (H), titanium (Ti), and phosphorus (P) contents were determined by elemental analysis (EA) and total digestion. The thermogravimetric analysis (TGA) (Figure a) weight loss at below 105 °C gave an estimate of the adsorbed water contents (0–3%, Table ). Solid-state 13C CP MAS NMR (Figure b) was run to check that all of the C was sourced from the organic DBP groups and there was no residual solvent. In addition, solid-state 31P MAS NMR with peak deconvolutions (Figure c) provided us the relative ratios between the −H2PO4 and −HPO4 groups. It is reported that peaks with a 31P chemical shift at −5 to −9 ppm result from −H2PO4 groups, whereas those at −15 to −22 ppm represent −HPO4 groups.[38] Peaks close to 1–2 ppm are assigned to DBP based on our liquid NMR results. The detailed peak positions and relative quantification of the inorganic phosphate groups are present in Table . Although the synthesis precursor for TiP_0:1 consisted only of HDBP, inorganic −H2PO4 groups were found in the final TiP_0:1. For this, a partial hydrolysis of HDBP in acid media would be a logical explanation.

Figure 2

Characterizations in relation to formulae calculations. (a) TG curves (solid line and left y-axis) and DSC curves (dashed line and right y-axis). (b) 13C CP MAS NMR spectra. (c) 31P MAS NMR spectra with peak deconvolutions. (d) Powder XRD patterns of the calcined TiP materials compared with ICDD reference patterns. (e) Raman spectra of the calcined TiP materials.

Table 3

Characterization Results Used for the Formulae Calculations

					31P NMR peak fitting				TGA
	EA		digestion		HPO4		H2PO4		DBP	mass loss %
material	C wt %	H wt %	P wt %	Ti wt %	ppm	%	ppm	%	ppm	105 °C	800 °C
TiP_1:0	0.1	2.8	15.1	27.5	–22.0, −21.0	27.6	–7.7	72.4		2.6	19.6
TiP_3:1	12.7	4.2	15.2	24.0	–22.0, −21.7	45.2	–5.3, −11.1	54.8	0.5, 1.9	2.2	30.1
TiP_1:1	24.5	5.4	13.8	17.3	–21.7, −18.7	46.4	–9.9	53.6	2.3	2.0	43.3
TiP_1:3	33.8	6.5	13.3	13.2	–21.5, −18.7	72.2	–3.8, −9.8	27.8	2.4	0.5	48.3
TiP_0:1	39.0	7.5	14.2	11.9			–9	100	2.3	0.0	56.6

To gain more insights into the composition of the TiP materials, they were calcined at 800 °C under atmospheric condition for 6 h. The resulting white powders were characterized by SEM (Figure S1 in the Supporting Information), XRD, and Raman spectroscopy. The calcined TiP materials are composed of titanium pyrophosphates (Figure d), including TiP2O7 and (TiO)2P2O7 (ICDD card numbers 00-038-1468 and 00-039-0207, respectively). The XRD peaks of these two pyrophosphates overlap to some extent; nevertheless, we were able to distinguish them using Raman spectroscopy. The peak at 720 cm–1 (Figure e) signifies the presence of Ti–O–Ti chains that contain nonbridging oxygen, as we discussed earlier. Therefore, only (TiO)2P2O7 can produce Raman scattering at 720 cm–1. A small amount of amorphous TiO2 is believed to be present in all calcined samples because of the oxygen-rich calcination environment. Amorphous TiO2 is a weak Raman scatterer and thus cannot be fully characterized with this technique.[40]

On the basis of all aforementioned characterizations, we were able to determine the composition of the samples. The calculated formulae are listed in Table . The detailed calculations and justifications are provided as the Supporting Information. For the TiP materials synthesized from mixed organic–inorganic precursors, the inorganic-to-organic P ratios in the obtained materials are lower than that in the synthesis liquor. Substantially higher Ti(IV) reactivity was observed for HDBP than for H3PO4. The obtained titanium(IV) butyl phosphate materials contain hydroxyl and oxy groups. The increased oxy anion content from the TiP_1:0 to TiP_0:1 series confirmed the results from Raman spectroscopy. Solvent extraction of Ti(IV) by DEHPA yields the extractable [TiO(DEHPA)2] complex, where deprotonated DEHPA acts as a bidentate ligand to occupy the vacant coordination sites of the (TiO)2 chains.[41,42] The TiP_0:1 material is believed to have a similar structure (Scheme ).

Scheme 1

Proposed Structure of the TiP_0:1 Material Showing the Layered Assembly and the Local Coordination of Ti(IV)

Preliminary Investigation of Ln Uptake Capability

The separation between early and late Ln3+ ions serves as a reasonable indicator of the intralanthanide separation capability of the TiP materials. Therefore, preliminary batch trials on competitive Nd–Dy uptake were conducted. The Nd–Dy pair was chosen also because of its industrial relevance with respect to the recycling of NdFeB magnets. In all cases, the total Ln3+ uptake increased with the elevated equilibrium pH (Figure S2a) and there were no significant differences in the total uptake amounts between the TiP_0:1 and TiP_1:0 samples. However, the SF(Dy/Nd) values for the TiP_1:0 sample stayed between 1 and 2 over the studied pH range (1–3), whereas the same values for TiP_0:1 increased with pH to more than 100 at approximately pH 2.8 (Figure S2b). In the hybrid TiP materials (TiP_1:3, TiP_1:1, and TiP_3:1), the SF(Dy/Nd) maxima were observed in the pH range of 2.2–2.6. The increase of Nd3+ uptake after this pH range was responsible for the subsequent decrease in SF(Dy/Nd). The results of the preliminary investigation indicated that TiP_0:1 exhibited the best intralanthanide separation potential. Using the purely inorganic TiP_1:0 as a comparison, the selective uptake of the heavier Ln3+ by TiP_0:1 should originate from the effect of the organophosphate functional groups. Therefore, the TiP_0:1 material, prepared from a purely organophosphate precursor, was selected for further separation and mechanism study in the current work. The hybrid TiPs would find applications when the selectivity requirements are not that high or when total Ln uptake is needed.

Proposed Ln3+ Uptake Mechanism and Kinetics

Ln3+ forms nitrate complexes in aqueous solutions. TBP is a solvating organophosphate extractant that coordinates to neutral Ln(NO)3 complexes (eq ):

Previously, we reported a group of titanium(IV) alkyl-phosphate grafted silica materials that retain Ln3+ in a solvating extraction manner.[21] The addition of nitrate salts significantly enhanced the uptake of Nd3+ and Dy3+ on these grafted materials, thereby suggesting that the uptake mechanism follows eq . However, the addition of 5 M NH4NO3 inhibited the Ln3+ uptake on the TiP_0:1 material (Figure S3). To elucidate the uptake mechanism, we closely monitored the Ln3+ uptake kinetics in two sorption systems.

The first proposed sorption system I consisted of 2 mM Lu3+ at an initial pH of 1.8. The uptake of Lu3+ appeared to be relatively slow, reaching less than 30% of the equilibrium uptake after 24 h of mixing (Figure a). Notably, significant amounts of Ti (soluble Ti(IV) ions irrespective of speciation) were dissolved into the solution during the first 24 h. The shape of the Ti dissolution amount curve in Figure a resembles that of the solution pH changes in Figure b. The Lu3+ uptake could, therefore, result from the transmetalation reaction: metal cation exchange with framework TiO2+, according to eq . Here, the barred ions refer to the solid phase.

Figure 3

Lu3+ uptake kinetics on the TiP_0:1 material (2 mM Lu3+). Effect of contact time on (a) Lu3+ uptake and Ti dissolution, (b) solution pH, and (c) n(Ti)/n(Lu3+) ratio.

The ratio of dissolved Ti to Lu3+ uptake [n(Ti)/n(Lu3+)] was calculated and is presented in Figure c. The n(Ti)/n(Lu3+) values are always below the theoretical value at 1.5 (according to eq ) because of the rapid precipitation reaction of TiO2+ (eq ), leading to the decrease in the amount of dissolved Ti, the solution pH, and the n(Ti)/n(Lu3+) value.

The second proposed sorption system II involved competitive Nd3+ and Dy3+ uptake at an initial pH of 3.6. Over the course of 20 d, the uptake of Nd3+ remained negligible (<1 mg g–1) while more than 97% of Dy3+ was removed from the solution (Figure a). Although the dissolved Ti concentrations were consistently lower than the detection limit (<10 μg L–1), the drastic drop in the solution pH through the initial 10 h of contact (Figure b) indicated the ion-exchange dissolution and rapid hydrolysis of TiO2+ (faster reaction rate for eq at higher pH). The calculated SFs [SF(Dy/Nd)] reached over 103 after 24 h.

Figure 4

Competitive Nd3+ and Dy3+ uptake kinetics on the TiP_0:1 material (1 mM equimolar mixture of Nd3+ and Dy3+). Effect of contact time on (a) Nd3+ and Dy3+ uptake, (b) solution pH, and (c) SF(Dy/Nd).

The sorption systems I and II were described by common kinetic models, and the detailed results are given in the Supporting Information (Figure S4). On the basis of the obtained high correlation coefficients (R2 close to 1) and the closeness of the modeled equilibrium uptakes compared with the experimental values, both of the kinetics in systems I and II seem to obey the pseudo-second-order model (Table S2). However, further model validation suggested that the errors are not normally distributed and certain points have an unreasonably higher level of influence with a Cook’s distance close to 0.5 (Figure S5–S6). The models especially failed to fit the uptake kinetics during the initial 24 h of contact (Figure S7). For system I, the initial accumulation of Ti4+ in the solution inhibited the equilibrium in eq from shifting to the right side. For system II, the initial rapid drop in solution pH influenced the reaction rate in eq and the speciation of Dy3+ in solution.

When the TiP_0:1 material was equilibrated in 0.1 M HNO3, a negligible amount of Ti was detected in the solution, thereby excluding the dissolution–precipitation mechanism. Ln3+ complexes with dialkylphosphoric acids were determined to be in a polymeric arrangement, in which each Ln3+ center is surrounded by six oxygen atoms from the ligand in a pseudo-octahedral environment.[43] Given the above observations, the Ln3+ uptake on the TiP_0:1 sample is illustrated in Scheme . The selectivity arises from the preferential transmetalation reaction (framework cation exchange). Tasaki-Handa has systematically studied the Ln3+/Ln3+ ion-exchange behavior in the coordination polymers based on DEHPA.[44,45] The ion-exchange affinity was ascribed to coordination preference and steric strain. On the one hand, because of the Ln contraction, the coordination preference follows the reverse atomic number sequence in the Ln series. Smaller ions coordinate stronger with the DBP groups. On the other hand, the steric strain effect amplifies the same selectivity order. The ionic radius of Ti4+ (74.5 pm) is significantly smaller than that of Ln3+ (86.1–103.2 pm, six-coordinated). The non-isostructural insertion of larger Ln3+ ions into the TiP framework requires more energy because of the higher degree of structural distortion. The level of steric strain caused by the TiO2+ ↔ Ln3+ exchange is expected to be even higher compared to that by the equivalent Ln3+ ↔ Ln3+ exchange. Thus, we were able to obtain higher SF(Dy/Nd) values compared with earlier studies using equivalent framework ion exchange[46] and fraction precipitation by organophosphoric acids.[10,26,27]

Scheme 2

Proposed Transmetalation Reaction on the TiP_0:1 Material Demonstrating the Exchange of Framework TiO2+ with Dy3+

To support the proposed mechanism, the Dy3+-loaded TiP_0:1 material (after equilibrating for 20 d in system II, denoted as TiP_0:1_Dy) was characterized and compared with pristine TiP_0:1 as well as the synthesized pure Dy(DBP)3 and/or Nd(DBP)3 coordination polymers. Powder XRD patterns (Figure a) reveal the shrinkage of interlayer distance after Dy3+ exchange and structural distortion. Two peaks appeared for the TiP_0:1_Dy sample corresponding to d spacings of 1.40 and 1.31 nm. Dy(DBP)3 crystallizes in a monoclinic fashion,[26] and the peak (d = 1.29 nm) here represents the distance between the chains. The ATR–FTIR spectra (Figure b) show a doublet at approximately 1170 and 1100 cm–1, supposedly because of the vibration of the POO– group.[44] This exact doublet shifted toward the value of Dy(DBP)3 in the TiP_0:1_Dy sample. The Raman and UV–vis diffuse reflectance spectra were also checked before and after Dy3+ loading, yet no evident changes were found. Hydrolysis induced small amounts of −H2PO4 groups, as they were found by the 31P NMR peaks with chemical shifts at −18.4 and −21.4 ppm in the TiP_0:1_Dy material (Figure c). Further studies in exploring the local coordination environment of the loaded Ln3+ are needed to fully justify our proposed mechanism.

Figure 5

Characterizations of Dy3+-loaded TiP_0:1 (TiP_0:1_Dy) compared with the original TiP_0:1 and/or synthesized Dy(DBP)3 and Nd(DBP)3. (a) Powder XRD patterns in the 2θ range of 5.0–7.5°. (b) ATR–FTIR spectra in the 900–1300 cm–1 range. (c) Solid-state 31P MAS NMR spectrum with peak deconvolutions.

Extraction from Dilute Ln3+ Mixtures

The ion-exchange uptake preference of Ln3+ on the TiP_0:1 material was tested under batch condition in a mixture consisting of 10 mg L–1 of all Ln3+ ions. Because of their chemical similarity, Sc3+ and Y3+ were also added to the mixture. The extraction appeared to be pH-dependent, and the increase of equilibrium pH increased the uptake (Figure a). The pH dependency can be explained by eqs and 6. Sc3+, having the smallest ionic radius, was completely retained by the material even at pHs as low as 2.0. The extraction efficiency and Kd value for Y3+ lay between those for Ho3+ and Er3+, and this is concurrent with their ionic radii for six-coordinated ions: 90.1 pm for Ho3+, 90.0 pm for Y3+, and 89.0 pm for Er3+.

Figure 6

Extraction rates (a) and distribution coefficients (Kd) values (b) for metal uptake on the TiP_0:1 material. The ionic radii of the tested elements with different coordination numbers are indicated in (c).[3] Initial solution composition: 10 mg L–1 mixture of all Ln3+ (except radioactive Pm3+) together with Sc3+ and Y3+. The maximum Kd value in the study is presented as 199 800 mL g–1, calculated from the detection limit (10 μg L–1).

Across the Ln series, the extraction rate did not follow Ln contraction completely. Instead, the Ln “tetrad effect” was observed.[47] The extraction rate showed discontinuities at the positions of 1/4 filling (between Nd and Pm) and half filling (Gd) of the 4f electron subshell. The discontinuity at 3/4 filling (between Ho and Er) was not visible because of high extraction rates. This effect is commonly observed in the distribution of Ln in seawater and minerals, and it originates from the variations of interelectronic repulsion for the ground states across the Ln series.[48] In terms of the distribution coefficients (Kd, Figure b), a clear dividing line can be drawn between Gd and Tb. The present study establishes that the group separation between early Ln (La–Gd) and late Ln (Tb–Lu) could be effectively carried out on the TiP materials. The group SFs [SF(Tb–Lu/La–Gd)] were calculated as 2.0 ± 0.3, 28.7 ± 0.4, 203.1 ± 3.4, and 112.5 ± 35.3 at, respectively, pH 2.0, 2.6, 3.2, and 4.2.

Binary Ln3+ Separation Models

Binary Ln3+ uptake experiments were carried out to investigate in detail the potential separation behavior. Seven combinations were chosen across the Ln series: La–Ce, Pr–Nd, Eu–Gd, Gd–Tb, Tb–Dy, Yb–Lu, and Nd–Dy. The uptake of Ln3+ by the TiP_0:1 material from the 1 mM equimolar binary mixture was monitored as a function of solution equilibrium pH. Among these combinations, the first six represent adjacent Ln pairs, whereas the last one is a typical example for early and late Ln separation. The adjacent Ln pair Gd–Tb was studied because their distribution coefficients signify a clear discontinuity in the earlier section.

Adjacent Ln pairs are known to share extremely similar chemical properties, and their separation is one of the most difficult tasks in inorganic separations. The similarities of the adjacent Ln pairs are observed based on the almost identical shape of their Kd versus pHeq curves (in Figures S8–S13, Supporting Information). The uptake preference (represented by the SFs, Figure a–f) in the current case strictly followed the reversed Ln contraction order (Lu3+ → La3+), whereby Ln3+ ions with smaller radii were preferentially retained. This deviates from the Ln tetrad effect. We suggest that the tetrad effect only becomes evident if the concentrations of Ln are relatively low. If there are large amounts of competing Ln3+ ions, the coordination and steric strain preferences favor the smaller Ln3+. The separation between Gd and Tb is significantly easier compared to that between Eu and Gd, although both pairs are adjacent Ln. These results confirm the group separation behavior that Ln series is divided into two subgroups: from La to Gd as early Ln and from Tb to Lu as late Ln in our current work. In addition, significant Nd–Dy separation was achieved (Figures g and S14), possibly because of their larger differences in ionic radii.

Figure 7

SFs between selected Ln3+ pairs [(a) La–Ce; (b) Pr–Nd; (c) Eu–Gd; (d) Gd–Tb; (e) Tb–Dy; (f) Yb–Lu; and (g) Nd–Dy] as a function of solution pH. Initial solution composition: 1 mM equimolar mixture of the selected Ln3+ pair. The corresponding Kd values are available in Figure S8–S14 in the Supporting Information.

The sole external parameter contributing to the separations is the equilibrium pH of the solution in our proposed system. This means that, through delicate control over the solution pH, optimal separations can be achieved either by leaving the majority of the larger Ln3+ in the solution or by exchanging the majority of the smaller Ln3+ onto the TiP material. The optimal molar fractions (best values obtained from the binary competitive batch systems) of the Ln in the liquid and solid phases are plotted in Figure a. Through only one uptake cycle, the heavier Ln in the adjacent Ln pairs was concentrated in the solid phase, representing some 60–80% of the total Ln fraction. Concurrently, the lighter Ln was concentrated in the liquid phase. The separation of Nd and Pr remained the most difficult one. Notably, almost complete Nd–Dy separation was achieved batchwise, as more than 99% of the loaded Ln was Dy.

Figure 8

(a) Stacked column plots showing the optimal Ln fractions in the solid and the liquid phases obtained from the binary uptake study at different equilibrium pH values. (b) Dependence of the best achieved SFs using a logarithmic scale on the ionic radii difference of the Ln3+ pairs.

Because the separation capability originates from the small differences in ionic radii, we plotted the best achieved SFs (on a logarithmic scale) versus the differences in ionic radii (Figure b). Ln3+ forms an Ln(DBP)3 complex; therefore, the ionic radii of six-coordinated ions were used. The linear regression line was drawn, and it appears that a larger difference in the ionic radii is positively correlated with the separation behavior. Overall, these results emphasize the potential technological significance of the TiP materials in Ln separation.

Transmetalation as a Metal Separation Process

Transmetalation is a type of organometallic reaction during which the organic ligands are transferred from one metal ion to another. It has been shown in recent years that MOFs and coordination polymers are applicable hosts for transmetalation. Partial, core–shell, complete, or even single-crystal-to-single-crystal metal exchanges were reported.[49] The metal exchange depends on ionic radii differences, preferential coordination geometries, kinetics, framework flexibility, and solvent effects.[50] It appears that transmetalation on water-stable MOFs and coordination polymers is worth exploring as a hydrometallurgical metal separation approach. The effects of codissolving metal ions (Fe3+, Al3+, Ca2+, etc.) in real samples need to be taken into consideration when further assessing the performance of the process. However, the intralanthanide separation is still preferably to be placed as a downstream process when all other metal ions, except for Ln3+, are eliminated.

Comparison with Literature Results

Table summarizes some intralanthanide separation approaches reported in recent years, with an emphasis on solvent extraction,[51] solvent leaching,[52] crystallization,[8,15,26] and ion-exchange methods.[25,46] Organic phosphorus-containing ligands are in general selective for Ln separation. Our method appears to be the best in terms of SF, though this is far from the sole judgmental factor for a successful separation process. Titanium(IV) serves as sacrificial ion-exchange sites, and the dissolved TiO2+ hydrolyzes into hydroxides. In reality, the complete regeneration of the material might not be possible because of the formation of titanate phases. Additionally, organophosphates can be attacked by H+, hydrolyzing into inorganic phosphate and alcohol. Finding a reasonable compromising point between selectivity and operational ease marks out our road ahead. The ultimate goal is a solvent-free, efficient, recyclable, and selective method that requires neither toxic chemicals nor intensive energy input.

Table 4

Intralanthanide Separation Capability Comparisons with Literature Reported Approaches

classification	method/reagent	target Ln pair	SF	references
solvent extraction	DEHPA	Nd–Dy	12.93	(51)
	Cyanex 923	Nd–Dy	12.79	(51)
solvent leaching	tripodal nitroxide ligand	Nd–Dy	359	(52)
crystallization	HDBP	Nd–Dy	>300	(26)
	borate	Nd–Dy	986	(8)
	camphorate	Tb–Dy	1.247	(15)
ion exchange	amorphous zirconium phosphate	Nd–Dy	1.9	(25)
	Ce(DEHP)3	La–Dy	1.6	(46)
	Fe(DEHP)3	La–Dy	18.4	(46)
	Al(DEHP)3	La–Dy	3.5	(46)
	titanium(IV)butyl phosphate	Nd–Dy	2065	this work
		Tb–Dy	11.5	this work

Conclusions

We have demonstrated the basis of a Ln separation strategy utilizing a transmetalation reaction on layered titanium(IV) organophosphate materials, as opposed to the extraframework ion-exchange materials. These coordination polymers were easily prepared by a precipitation reaction and showed potential in intralanthanide separation, especially for the separation between early and late Ln. The selective Ln uptake was controlled by solution pH and did not require any additional energy input (e.g., hydrothermal condition). The selectivity arose from the ligand coordination and steric strain preferences during the transmetalation reaction and structural distortion. The separation performance of the transmetalation reaction seems higher compared to the selective crystallization or solvent extraction process utilizing the same or a similar organophosphate ligand. That said, the stability of the material as well as the regeneration needs to be further studied to warrant the design of a true separation process. Through alterations of the framework metal (size and valence) and organophosphate ligand, a variety of new MOF or coordination polymer materials can be envisioned with enhanced and tunable metal selectivity in aqueous solutions.