Crystal Chemistry and Stability of Hydrated Rare-Earth Phosphates Formed at Room Temperature

Abstract

In order to understand the crystal chemical properties of hydrous rare-earth (RE) phosphates, REPO_4,hyd, that form at ambient temperature, we have synthesized REPO_4,hyd through the interaction of aqueous RE elements (REEs) with aqueous P at room temperature at pH < 6, where the precipitation of RE hydroxides does not occur, and performed rigorous solid characterization. The second experiment was designed identically except for using hydroxyapatite (HAP) crystals as the P source at pH constrained by the dissolved P. Hydrated RE phosphate that precipitated at pH 3 after 3 days was classified into three groups: LREPO_4,hyd (La → Gd) containing each REE from La-Gd, MREPO_4,hyd (Tb → Ho), and HREPO_4,hyd (Er → Lu). The latter two groups included increasing fractions of an amorphous component with increasing ionic radius, which was associated with non-coordinated water. RE_allPO_4,hyd that contains all lanthanides except Pm transformed to rhabdophane structure over 30 days of aging. In the experiments using HAP, light REEs were preferentially distributed into nano-crystals, which can potentially constrain initial RE distributions in aqueous phase. Consequently, the mineralogical properties of hydrous RE phosphates forming at ambient temperature depend on the aging, the pH of the solution, and the average ionic radii of REE, similarly to the well-crystalline RE phosphates.

1. Introduction

Rare-earth (RE) elements (REEs), such as the lanthanides (La-Lu), are important rare metals in geological resources, and commonly used for technological application owing to their fluorescent, catalytic, and magnetic properties [1]. REEs are also potential fission products and used as surrogates for trivalent actinides in various experimental studies in nuclear chemistry [2,3,4]. Among a variety of REE-bearing minerals, RE phosphate is one of the important REE hosts that have further applications to phosphors, though post-heat treatment is required [1].

Crystalline RE phosphates occur in four different structures: monazite (light REPO₄; P2₁/n), rhabdophane (light REPO₄·nH₂O; P6₂22 or C2), xenotime (heavy REPO₄; I4₁/amd), and churchite (heavy REPO₄·nH₂O; I2/a) [5,6,7,8]. In the crust (granite, metamorphic and/or metasomatic rocks, and pegmatite), monazite and xenotime are the stable phases for light REE and heavy REE, respectively, while rhabdophane and churchite occurrences are almost exclusively encountered in soil [9,10,11,12,13].

The structure depends on physico-chemical parameters such as pH, reaction time, and temperature [14,15,16]. Ionic radii also influence the crystal structure and morphology. Although the previous studies have reported on the mineralogical properties of all four types of RE phosphates and their application to accommodate foreign cations [17,18], these studies were mainly concerned with the well-crystalline phase formed at elevated temperatures at acidic pH ~1, e.g., [16,19]. The crystal chemical property of the hydrous RE phosphates that formed at ambient temperature has not been explored in detail. In addition, recent studies reported a novel synthesis of RE phosphate nanocrystals in P-free solution using microorganisms that release P from cell interiors at pH 3–5 at room temperature [20,21]; however, the electron diffraction pattern revealed vague patterns with diffused diffraction maxima owing to the presence of amorphous matter, which prevents the full identification of these nanoparticles. Hence, the first half of the present study reports a systematic characterization of hydrated RE phosphates, REPO_4,hyd, that formed at ambient temperature to understand their crystal-chemical properties. The formation of RE hydroxides formation was suppressed by using the pH condition of <6 in the present experiment.

In geological media, RE phosphates occur primarily as an accessary mineral such as monazite and xenotime, and the secondary precipitates of RE phosphates such as rhabdophane frequently play a key role as an indicative of various geochemical signatures such as for estimating paleo-environment [22,23] and migration of actinides such as Pu in the Oklo natural fission reactor [17,18,24,25,26,27,28,29]. The secondary RE phosphates commonly occur in weathered profiles in association with original phosphate minerals, such as apatite, which acts as a source of phosphorus (P) and provides unique reaction interfaces for the formation of secondary phosphates such as epitaxial growth and pseudomorphism [30]; however, the effects of the substrates has not been evaluated for the formation of hydrous RE phosphate. Thus, the second half of the present study demonstrates the time-course of the formation process of hydrated RE phosphates, REPO_4,hyd, over the HAP crystal at ambient temperatures at varying pHs constrained by dissolved phosphate to elucidate the evolution of secondary RE phosphate at the interfaces and the REE fractionation between solution and solid.

2. Materials and Methods

2.1. REPO_4,hyd Precipitation from Aqueous Trivalent REE and Phosphate in Solution

In the present study, RE_i refers to phases containing individual La-Lu elements but excludes Pm, and RE_all refers to phases containing all La-Lu elements except for Pm. To simplify the discussion, Sc and Y were not included in the present experiment. A series of hydrated RE phosphates, RE_iPO_4,hyd (where RE = La-Lu except for Pm) were synthesized from aqueous solutions. The reactants used as starting materials were La(NO₃)₃·6H₂O (99.9%), Ce(NO₃)₃·6H₂O (98%), Pr(NO₃)₃·xH₂O (99.5%), Nd(NO₃)₃·6H₂O (99.5%), Sm(NO₃)₃·6H₂O (99.5%), Gd(NO₃)₃·6H₂O (99.5%), Dy(NO₃)₃·6H₂O (99.5%), Ho(NO₃)₃·xH₂O (99.5%), Er(NO₃)₃·xH₂O (99.5%), and Yb(NO₃)₃·xH₂O (99.9%) supplied by Wako Pure Chemical (Osaka, Japan), Eu(NO₃)₃·6H₂O (99.9%) and Tb(NO₃)₃·6H₂O (99.9%) supplied by Strem Chemicals (Newburyport, MA, USA), Tm(NO₃)₃·xH₂O (99.9%) supplied by Alfa Aesar (Ward Hill, MA, USA), and Lu(NO₃)₃·4H₂O (99.95%) supplied by Kanto Chemical (Tokyo, Japan). A 0.1 mol·L⁻¹ RE³⁺ solution was prepared for each REE by dissolving the reactants in ultrapure water (Milli-Q^®, Merck Millipore, Billerica, MA, USA). The pH of each solution was adjusted to pH 3.0 ± 0.1 with NaOH and HNO₃ using a pH meter (Toko TXP-999i, Tokyo, Japan) with an electrode (Toko PCE108CW-SR). The margin of error in the pH measurements was within 0.1 pH units. In addition, 10 mL of a 0.1 mol·L⁻¹ NaH₂PO₄ (Wako Pure Chemical) aqueous solution adjusted to pH 9.0 ± 0.1 was added to 10 mL of a 0.1 mol·L⁻¹ RE solution with continuous stirring, and the pH was adjusted to 3.0 ± 0.1. The solutions were shaken by a rotary shaker for 3 days at room temperature. The suspensions were filtered with an Omnipore membrane filter with a pore size of 0.1 µm (Merck Millipore JVWP04700, Billerica, MA, USA) and rinsed with ultrapure water. The precipitates were dried in air at room temperature. RE_allPO_4,hyd was synthesized using the same procedure as RE_iPO_4,hyd and a solution with a total RE concentration of 0.1 mol·L⁻¹ (RE = La-Lu except Pm), in which the concentration of each REE was set to 7.1 mmol·L⁻¹.

To evaluate the effects of crystal aging, RE_iPO_4,hyd (RE = La, Tb, Dy, Ho, and Yb) and RE_allPO_4,hyd were reacted for an extended duration ranging from 3 to 30 days. The effect of pH was examined in the case of LaPO_4,hyd formation by varying the pH from 1 to 5. LaPO_4,hyd was synthesized at pH 1 by pouring 10 mL of 0.1 mol·L⁻¹ NaH₂PO₄ into 10 mL of 0.1 mol·L⁻¹ La(NO₃)₃. The pH of the NaH₂PO₄ and La(NO₃)₃ solutions were adjusted beforehand to 9.0 ± 0.1 and 1.1, respectively, to account for the number of protons that would be released to the solution during the reaction. For synthesis at pH 5, 10 mL of 0.1 mol·L⁻¹ NaH₂PO₄ was poured into 10 mL of 0.1 mol·L⁻¹ La(NO₃)₃, for which the pH values were adjusted beforehand to 12.2 and 5.3, respectively. After mixing the solutions, the pH was subsequently adjusted again to 5.0 ± 0.1. The same incubation and filtration procedures were performed as for the other RE_iPO_4,hyd phases.

Additional experiments were performed to investigate the thermal stability of RE_iPO_4,hyd formed at room temperature. RE_iPO_4,hyd (RE = Tb, Dy, Ho, and Yb) and RE_allPO_4,hyd synthesized from a solution at pH 3 for 3 days were annealed at 70, 120, 200, 300, 400, or 500 °C for 1 h in a muffle furnace (Yamato Scientific FM38, Tokyo, Japan).

2.2. REPO_4,hyd Formation on HAP Crystals

RE³⁺ solutions at a concentration of 2.0 mmol·L⁻¹ (RE = La, Tb, Yb) were prepared by dissolving RE(NO₃)₃·nH₂O in Milli-Q^® water, which was adjusted to a pH of 5 using HNO₃. 0.2 g of synthetic HAP in the form of Ca₅(PO₄)₃OH powder (Wako Pure Chemical) was added to 100 mL of the RE³⁺ solutions at room temperature. The HAP grains are rod-shaped, with approximate dimensions of 1.5 µm × 300 nm and a specific surface area of 6.31 m²·g⁻¹ [30]. The S/V ratio (where S is the apatite surface area and V is the volume of solution) was 1.3 × 10² cm⁻¹ in these experiments. The suspensions of HAP powders were gently agitated using a magnetic stirrer for 1, 3, 9, and 24 h and 3 and 10 days. The solution pH was measured for each duration. The precipitate was collected by centrifugation at 3000 rpm for 10 min at 25 °C, washed with ultrapure water, and dried in air at room temperature. The supernatants remaining after centrifugation were filtered under reduced pressure with an Omnipore membrane filter with a pore size of 0.025 µm (Merck Millipore VSWP04700), which was small enough to separate the precipitates from the supernatant. RE_allPO_4,hyd was also synthesized on HAP through the same procedure using a solution with a total RE concentration of 2.0 mmol·L⁻¹ (RE = La-Lu except Pm), in which the concentration of each REE was 0.14 mmol·L⁻¹.

2.3. Analytical Methods

The crystal structures of RE_iPO_4,hyd (RE = La-Lu) and RE_allPO_4,hyd prepared from solution at room temperature were determined using powder X-ray diffraction analysis (XRD, Rigaku MultiFlex, Tokyo, Japan or Rigaku SmartLab). The analysis was conducted using Cu Kα radiation (λ = 1.5418 Å) at 40 kV and 40 mA for the MultiFlex instrument, and at 40 kV and 30 mA for the SmartLab instrument in a scan range of 2θ = 10°–63° with a scan speed of 2° per min. A non-reflective silicon (111) holder was used to hold the specimens. The REPO_4,hyd that formed on HAP crystals was analyzed using the SmartLab XRD in a scan range of 10°–50° using a SiO₂ glass plate as a specimen holder. Each spectrum was smoothed with a modified Savitzky–Golay filter and the linear background was subtracted using JADE 7 software (Materials Data Incorporated, Livermore, CA, USA). Subsequently, peak separation and fitting were conducted using the same software to calculate the crystallite size from the Scherrer equation:

$$D=\frac{K\lambda }{\beta \cos \theta }$$

(1)

where D, K, λ, β, and θ refer to the crystallite size (Å), the Scherrer constant (in this study K = 0.9), the wavelength of X-rays (1.5418 Å), the full-width at half maximum of the peak (rad), and the Bragg angle at the peak (rad), respectively.

RE_iPO_4,hyd (RE = La-Lu) and RE_allPO_4,hyd prepared from solution at pH 3 for 3 days and 30 days were analyzed by Fourier transform infrared spectroscopy (FT-IR) using KBr disks (1 mg sample/100 mg KBr, 10 mm in diameter). Spectra were recorded at a resolution of 4 cm⁻¹ by summation of 32 scans.

Thermogravimetry–differential thermal analysis (TG–DTA) was performed on RE_iPO_4,hyd (RE = La-Lu) and RE_allPO_4,hyd prepared from solution at pH 3 for a duration of 3 days. Furthermore, RE_iPO_4,hyd (RE = La, Tb, Yb) and RE_allPO_4,hyd were reacted at pH 3 for 30 days and were analyzed to investigate the effect of aging. Each specimen was heated to 500 °C at a heating rate of 10 °C/min and then held for 10 min at 500 °C under a nitrogen gas flow of 250 mL/min using α-alumina as a reference material (Hitachi TG/DTA7300, Tokyo, Japan).

The supernatant after filtration was analyzed to obtain the concentrations of RE (RE = La-Lu), Ca, and P using inductively coupled plasma atomic emission spectroscopy (ICP-AES, PerkinElmer Optima 530 DV, Waltham, MA, USA).

High-resolution transmission electron microscopy (HRTEM), selected-area electron diffraction (SAED), high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM), and energy dispersive X-ray spectroscopy (EDS) were performed for all precipitate samples obtained in the syntheses using a JEOL JEM-ARM200F (Tokyo, Japan) with an accelerating voltage of 200 kV. The spherical aberration coefficient C_s is ~0 mm. The contrast in HAADF-STEM mode correlates with the mass and thickness of the target material [31]. TEM specimens were prepared by placing drops of a suspension of the synthesized precipitates onto a holey carbon mesh supported by a Cu grid.

3. Results

3.1. XRD Analysis

XRD patterns for RE_iPO_4,hyd (RE = La-Lu except Pm) and RE_allPO_4,hyd reacted at pH 3 for 3 days are shown in Figure 1a. The peak positions of RE_iPO_4,hyd (RE = La-Gd except Pm) correspond well, with a slight peak shift, to the structure of rhabdophane [8]. The patterns for RE_iPO_4,hyd (RE = Tb-Lu) and RE_allPO_4,hyd could not be identified due to the broadened peaks and lack of matching structures. Although a few of the RE_iPO_4,hyd (RE = Er-Lu) peaks appear to be similar to those of xenotime [5], the patterns lack a (200) peak and, thus, the structures were not determined convincingly. Furthermore, the patterns for RE_iPO_4,hyd (RE = Tb-Ho) display intermediate profiles between those of RE_iPO_4,hyd (RE = La-Gd except Pm) and RE_iPO_4,hyd (RE = Er-Lu). Based on the structural characteristics indicated by XRD analysis in this study, the structure of the RE_iPO_4,hyd precipitates are classified as LREPO_4,hyd for phases containing RE = La-Gd, MREPO_4,hyd containing RE = Tb-Ho, and HREPO_4,hyd containing RE = Er-Lu, respectively. In LREPO_4,hyd, which has a rhabdophane structure, the coordination numbers (CNs) of RE³⁺ ions are one third eight-fold and two-thirds nine-fold [8], resulting in an average CN of 8.7. RE³⁺ in MREPO_4,hyd and HREPO_4,hyd can be considered to have eight-fold coordination because the CN of xenotime and churchite is eight-fold. Based on the CNs of all possible RE_iPO_4,hyd phases, the average ionic radius of RE_allPO_4,hyd containing all RE (RE = La-Lu except Pm) is calculated to be 1.07 Å, which lies just between that of Gd (CN = 8.7, 1.09 Å) and of Tb (CN = 8, 1.04 Å) [32].

The XRD pattern of RE_allPO_4,hyd correlates to a combination of the GdPO_4,hyd and TbPO_4,hyd patterns. A linear combination of these two profiles indicates volume fractions of 35% GdPO_4,hyd structure and 65% TbPO_4,hyd structure (Figure 1b).

The effect of pH on LaPO_4,hyd precipitates are shown in Figure 1a,c. The full width at half maximum (FWHM) of each peak becomes smaller as the pH lowers.

In the aging experiments, there was little difference in the XRD profiles between 3 and 30 days for RE_iPO_4,hyd (RE = La, Dy, Ho, and Yb); however, the structures of RE_allPO_4,hyd and TbPO_4,hyd changed significantly to become rhabdophane-structured after 30 days of aging (Figure 1a,d). In the DyPO_4,hyd pattern, a small peak corresponding to the ($\overline{1}$11) plane of rhabdophane appeared after 30 days of reaction, although the other peaks remained unchanged. The crystallite sizes of LaPO_4,hyd calculated on the basis of the XRD patterns in Figure 1 using Scherrer equation are clearly a function of pH and the aging period (Figure 2a,b) and are summarized in

Table 1. Crystallite size of LaPO_4,hyd samples formed under four different conditions. The calculation was performed for five major peaks in the XRD pattern using the JADE software based on the Scherrer equation with a Scherrer constant K of 0.9.

Miller Index	LaPO4,hyd Crystallite Size (nm)
	pH 3, 3 days	pH 3, 30 days	pH 1, 3 days	pH 5, 3 days
(11)	2.6	3.5	2.9	2.0
(11)	5.0	5.7	5.4	3.9
(13)	3.4	4.0	3.7	2.7
(22)	3.7	4.2	4.2	3.2
(711)	6.5	7.4	6.5	4.6

.

The calculated lattice volume of LREPO_4,hyd reacted at a pH of 3 for 3 days are plotted as a function of the effective ionic radius in Figure 3a by fitting the ($\overline{1}$11), ($\overline{5}$11), ($\overline{3}$13), ($\overline{2}$22), and (711) peaks to the peaks of the rhabdophane structure using the JADE software. The calculated lattice parameters are summarized in

Table 2. Lattice parameters a, b, c and β, and lattice volume for RE_iPO_4,hyd (RE = La-Gd) calculated based on the peaks in the XRD patterns using the JADE software.

REE	Effective Ionic Radius (Å) 1	Lattice Parameter (Å)				Lattice Volume (Å3)
		a	b	c	β
La	1.197	29.17	7.19	12.29	116.3	2313
Ce	1.178	28.50	7.22	12.13	115.7	2249
Pr	1.161	28.61	7.06	12.14	115.8	2209
Nd	1.145	28.48	7.04	12.07	115.8	2179
Sm	1.114	28.34	6.99	12.01	115.8	2142
Eu	1.102	27.93	7.02	11.91	115.5	2108
Gd	1.089	27.89	7.06	11.85	115.5	2108

¹ The effective ionic radius r = 1/3 r_CN8 + 2/3 r_CN9.

. The calculated lattice volume displays a positive linear correlation with the effective ionic radius, indicating that the crystal lattice shrinks uniformly as the ionic radius decreases (Figure 3a). The lattice volumes calculated in the present experiment are larger than that of well-crystalline rhabdophane structure reported in [8], most likely owing to the presence of more water molecules in the structure of LREPO_4,hyd.

The crystallinity of RE_iPO_4,hyd (RE = Dy-Lu) was calculated based on the profile fitting of amorphous halos in the XRD patterns (Figure S2,

Table 3. The results of quantification of the crystallinity of RE_iPO_4,hyd (RE = Dy-Lu) calculated using the JADE software.

REE	Effective Ionic Radius (Å) 1	Crystallinity (%)
Dy	1.027	38.9
Ho	1.015	53.9
Er	1.004	51.8
Tm	0.994	57.9
Yb	0.985	66.2
Lu	0.977	70.0

¹ The effective ionic radius from [32].

) and is plotted in Figure 3b as a function of effective ionic radius. The results indicate that DyPO_4,hyd has the largest amorphous volume and that the amorphous fraction linearly decreases towards LuPO_4,hyd.

The XRD patterns of RE_iPO_4,hyd (RE = Tb, Ho, Dy, and Yb) and RE_allPO_4,hyd obtained from the annealing experiments are summarized in Figure 4. TbPO_4,hyd transformed to a monazite structure at 400 °C with small peaks occurring at a 2θ value around 30°, as indicated by the circle. DyPO_4,hyd appeared to transform to a xenotime structure at 400 °C, as indicated by the appearance of small peaks at a 2θ value around 25.8°. HoPO_4,hyd and YbPO_4,hyd also transformed to a xenotime structure at around 300 and 200 °C, respectively. RE_allPO_4,hyd appeared to retain its initial structure up to 200 °C and then transformed to a rhabdophane structure, which was retained up to 500 °C. In contrast, TbPO_4,hyd did not become rhabdophane-structured but directly transformed to a monazite structure when annealed in air. The thermal behavior of TbPO_4,hyd is different from that observed in the aging experiments, in which the structure simply changed to that of rhabdophane.

3.2. FT-IR Analysis

Infrared spectra for RE_iPO_4,hyd (RE = La-Lu except Pm) and RE_allPO_4,hyd reacted at pH 3 for 3 days are shown in Figure 5a. Two bands at 520–650 cm⁻¹ (labeled 1 and 2) correspond to the asymmetric P–O bending vibration, ν₃, while the bands at 1000–1100 cm⁻¹ (labeled 3 and 4) correspond to the antisymmetric P–O stretching vibration, ν₄ [33,34,35]. Bands 3 and 4 are not clearly separated for LREPO_4,hyd. The symmetric stretching mode ν₁, which typically appears as a small absorption band at ~966–985 cm⁻¹ [36] and the bending vibration ν₂ were not observed, possibly because those bands were superimposed in the ν₃ and ν₄ regions for the hydrated RE phosphates [35]. The wavenumbers of band 4 in the ν₃ region plotted as a function of the effective ionic radius (Figure 5b and

Table 4. List of the wavenumber of band 4 that corresponds to P–O antisymmetric stretching vibration, ν₄ in the FT-IR spectra obtained for RE_iPO₄ (RE = La-Lu) and RE_allPO_4,hyd from FT-IR spectra.

REE	Effective Ionic Radius (Å) 1	Wavenumber (cm−1)
La	1.197	1053.9
Ce	1.178	1057.8
Pr	1.161	1059.7
Nd	1.145	1062.6
Sm	1.114	1068.4
Eu	1.102	1071.3
Gd	1.089	1076.1
REall	1.074	1075.1
Tb	1.040	1075.1
Dy	1.027	1075.1
Ho	1.015	1078.0
Er	1.004	1078.0
Tm	0.994	1079.9
Yb	0.985	1082.8
Lu	0.977	1085.7

¹ Effective ionic radius: r_La-Gd = 1/3 r_CN8 + 2/3 r_CN9, r_Tb-Lu = r_CN8, and r_all is the average of r_La-Lu.

) exhibit a linear decrease with increasing ionic radius for LREPO_4,hyd and HREPO_4,hyd, while MREPO_4,hyd and RE_allPO_4,hyd did not exhibit such trends. The broad bands at ~1600 cm⁻¹ (band 5) and at ~3000–3500 cm⁻¹ (band 6) are derived from coordinated H₂O [34]. Nitrate impurities derived from the initial REE reagents produce the band at ~1380 cm⁻¹.

FT-IR spectrum of the aged TbPO_4,hyd and RE_allPO_4,hyd appeared to be similar to that of LaPO_4,hyd, especially at bands 3 and 4 (Figure 5c). These FT-IR results concur with the XRD analyses, i.e., they indicate a change in the structures of RE_allPO_4,hyd and TbPO_4,hyd to a rhabdophane structure after 30 days of aging in solution.

3.3. TG-DTA Analysis

Derivatives of the TG (DTG) and DTA spectra for RE_iPO_4,hyd (RE = La-Lu) and RE_allPO_4,hyd reacted at a pH of 3 for 3 days are shown in Figure 6. LREPO_4,hyd and HREPO_4,hyd display a two-step dehydration, which is indicated by the presence of a DTG upper peak with an endothermic DTA peak. Assuming that the material remaining after heating to 500 °C entirely comprise anhydrous RE phosphates and the weight loss is derived solely from the dissociation of H₂O in REPO_4,hyd, the amount of dissociated water was calculated for the total heating duration (Figure 7a) and for each dehydration step (Figure 7b,c), and compiled in

Table 5. Summary of dehydration or crystallization temperature and the amount of dissociated water n from the RE_iPO_4,hyd (RE = La-Lu) and RE_allPO_4,hyd during the TG-DTA analysis heated up to 500 °C. The value n corresponds to the number of H₂O molecule per unit formula.

REE	Effective Ionic Radius (Å) 1	Second Dehydration Temperature (°C)	Crystallization Temperature (°C)	Dissociated Water n
				Total	First	Second
La	1.197	165	-	1.95	1.30	0.66
Ce	1.178	147	-	1.71	1.03	0.68
Pr	1.161	143	-	1.86	1.14	0.72
Nd	1.145	137	-	1.76	1.00	0.76
Sm	1.114	127	-	1.54	0.80	0.73
Eu	1.102	117	-	1.58	0.75	0.83
Gd	1.089	133	-	1.96	1.13	0.83
REall	1.074	-	220	2.22	-	-
Tb	1.040	-	300	2.24	-	-
Dy	1.027	-	326	2.36	-	-
Ho	1.015	-	300	2.43	-	-
Er	1.004	178	-	2.64	1.77	0.86
Tm	0.994	169	-	2.63	1.66	0.97
Yb	0.985	150	-	2.56	1.54	1.02
Lu	0.977	160	-	2.39	1.40	0.99

¹ Effective ionic radius: r_La-Gd = 1/3 r_CN8 + 2/3 r_CN9, r_Tb-Lu = r_CN8, and r_all is the average of r_La-Lu.

. The starting temperature of the second dehydration step was determined by selecting the local minimum of the DTG spectrum or the peak of the DTA spectrum between the first and second dehydration steps. A diagram of the dehydration extent in the first step in HREPO_4,hyd indicates that the amount of water loss increases as a function of ionic radius (Figure 7b). The amount of water loss in the second dehydration step also appears to exhibit a linear decrease as the ionic radii decrease in LREPO_4,hyd and HREPO_4,hyd (Figure 7c). MREPO_4,hyd and RE_allPO_4,hyd each exhibited only one dehydration peak. For these phases, a DTG upper peak and a DTA exothermic peak are present, indicating that crystallization occurred during dehydration.

In addition, the DTG and DTA spectra of RE_iPO_4,hyd (RE = La, Tb, and Yb) and RE_allPO_4,hyd after 30 days of reaction time are shown in Figure 6 for comparison with samples measured after 3 days of reaction time. Although LaPO_4,hyd and YbPO_4,hyd did not appear significantly different from those samples that had reacted for 3 days, additional dehydration peaks appeared in the spectra of TbPO_4,hyd and RE_allPO_4,hyd reacted for 30 days.

3.4. ICP-AES Analysis

In experiments of HAP dissolution in a solution containing all REEs, the concentrations of RE (RE = La-Lu except Pm), Ca, and P were measured for each duration using ICP-AES. The results are summarized in

Table 6. The solution composition determined by ICP-AES at each reaction time during the formation of RE_allPO_4,hyd, in which HAP crystals were in contact with the solution containing all REEs.

Reaction Time	Concentration (mmol·dm−3)
	La	Ce	Pr	Nd	Sm	Eu	Gd	Tb	Dy	Ho	Er	Tm	Yb	Lu	Ca	P 1	pH
0 h	0.12	0.11	0.13	0.12	0.12	0.13	0.11	0.13	0.11	0.13	0.13	0.13	0.13	0.13	udl1	udl	5.06
1 h	0.10	0.08	0.09	0.09	0.08	0.08	0.08	0.09	0.08	0.10	0.10	0.10	0.09	0.10	0.70	udl	5.16
3 h	0.08	0.06	0.07	0.07	0.06	0.07	0.06	0.07	0.07	0.08	0.08	0.08	0.08	0.08	1.03	udl	5.53
9 h	0.07	0.05	0.06	0.06	0.05	0.06	0.06	0.06	0.06	0.07	0.07	0.07	0.06	0.07	1.23	udl	5.55
24 h	0.06	0.03	0.03	0.04	0.03	0.03	0.04	0.04	0.03	0.05	0.05	0.05	0.04	0.05	1.62	udl	5.63
3 days	0.04	0.01	0.02	0.02	0.02	0.02	0.03	0.02	0.02	0.03	0.04	0.03	0.03	0.04	1.86	udl	5.66
10 days	0.03	0.01	0.01	0.02	0.01	0.02	0.02	0.02	0.02	0.03	0.03	0.03	0.02	0.03	2.01	udl	- 2

¹ udl: Under detection limits (0.05 ppb for Ca and 15 ppb for P). ² The pH after 10 days was not available.

and the total REE concentrations are plotted in Figure 8a as a function of reaction time. The P concentration is under the detection limit for all durations. The Ca concentration increased linearly for periods from 3 to 24 h, whereas the total concentration of all REEs decreased inversely with the Ca concentration. The reaction did not reach equilibrium within the duration of this dissolution experiment (Figure 8a); however, more than 80% of the total initial REE concentration was removed from solution after 240 h.

Figure 8b shows the transition in the patterns of REEs in solution reacted with HAP. The four curves at La-Nd, Sm-Gd, Gd-Ho, and Er-Lu exhibit the lanthanide tetrad effect [37]. Furthermore, more heavy REEs remained in solution compared with light REEs, and this trend becomes more pronounced with reaction time. The possible concentrations of free RE³⁺, RECO₃⁺, and REOH²⁺ at these experimental conditions were calculated using Geochemist’s Workbench using their formation constants [38], and input data of P_CO2 = 10^−3.5 bar [39] and phosphate free solution. The calculation results clearly demonstrate that the concentration of free RE³⁺ decreases as the ionic radius decreases because the amount of other REE complexes such as RECO₃⁺ and REOH²⁺ increase with decreasing ionic radius (Figure 8c,d).

3.5. TEM Analysis

The grain shapes of REPO_4,hyd gradually change with ionic radius. HRTEM images show that RE_iPO_4,hyd (RE = La, Nd, Tb) and RE_allPO_4,hyd form rod-shaped particles of ~50 nm in length. TmPO_4,hyd and LuPO_4,hyd are spherically shaped with diameters of 10 nm (Figure 9). DyPO_4,hyd and HoPO_4,hyd precipitates do not exhibit any specific shape due to their amorphous characteristics. The SAED patterns of LREPO_4,hyd and RE_allPO_4,hyd display the same ring pattern as that of the rhabdophane structure. Although the SAED patterns of other REPO_4,hyd are indistinct due to the large amorphous fractions, the RE_iPO_4,hyd (RE = Tb-Lu) structure can be distinguished from the rhabdophane structure by the lack of an innermost diffraction ring. In EDS analyses, most analytical points from the RE_allPO₄ crystals contain uniform collective peaks at 4–8 keV corresponding to the L lines of REEs (Figure 10a). However, a few points display non-uniform spectrums, as shown in Figure 10b,c, in which enhanced peaks for Er and Nd occur, respectively.

RE_iPO_4,hyd (RE = La, Tb, and Yb) and RE_allPO_4,hyd formed on HAP crystals appear to differ slightly from those formed in mixed solutions. For example, LaPO₄ precipitated as fibrous crystals on HAP (Figure 11a–c), in contrast to the rod-shaped crystals precipitated from the mixed solutions. RE_allPO_4,hyd precipitates that formed on HAP exhibit shorter fibrous shaped crystals than the LaPO_4,hyd precipitates on HAP (Figure 11d–i). The crystal shapes were not significantly affected by the reaction time. YbPO_4,hyd precipitated at the rim of HAP crystals with a non-fibrous morphology (Figure 11j–k), reflecting the rapid precipitation of YbPO_4,hyd and consumption of phosphate upon the release of P from the surface of HAP crystals.

4. Discussion

4.1. Properties of REPO_4,hyd Formed at Room Temperature

The REPO_4,hyd that formed at room temperature in this study and its characteristics are summarized schematically in Figure 12. The LREPO_4,hyd that formed at room temperature exhibited a rhabdophane structure similar to [40]. In this study, monazite did not form an primary product from solution at room temperature, although Hikichi and Hukuo [41] and Hikichi et al. [14] reported that monazite formed from aqueous solutions at 50 °C and pH < 1 within 28 days and at pH 5 after 900 days. Previous studies also reported that monazite was formed at 90 °C from solutions at pH < 1 [40,42], proceeded by the rhabdophane structure losing hydrated waters as a function of the aging time [40]. The differences in pH and the elevated temperatures may account for the absence of monazite in the present experiments at room temperature. This study considers mildly-acidic pH ranges relevant to natural environments without forming RE hydroxides, and the results show that monazite is not likely to form at room-temperature conditions.

Hikichi et al. [14] reported that a churchite structure in DyPO₄, YPO₄, and YbPO₄ formed at a pH of 3 at 20 °C; however, Hikichi et al. [43] did not identify a churchite structure for YPO₄ at a pH of 3.7 at 50 °C. The present study also did not detect a churchite structure at a pH of 3 at room temperature. The HREPO_4,hyd structure did not match the structure of any RE phosphate minerals, and only a few peaks in the XRD pattern corresponding to the anhydrous tetragonal structure of the xenotime structure appeared in the HREPO_4,hyd patterns. Such characteristics in the XRD pattern were also recognized in a previous study by Lucas et al. [40] reporting the synthesis of YPO₄·nH₂O at 50 °C. As described in the results section, the structure of these phases cannot be identified conclusively. However, it is clear that the HREPO_4,hyd structure involves not only amorphous features but also some periodic features. In addition, the structural periodicity is very limited, most likely less than 10 nm as a first approximation based on the XRD peak broadness. Indeed, HRTEM images revealed aggregates of particles as small as ~10 nm. In some cases, although the structure could not be clearly seen initially, electron-beam irradiation resulted in the formation of xenotime structure (Figure 9). Such transformation prevented the acquisition of HRTEM images of HREPO_4,hyd particles with their original structure. The evidence implies that the original structure is closely related to very small, nanometer-sized particles with a tetragonal structure similar to that of xenotime.

The presence of a structured component in HREPO_4,hyd is also supported by the gradual decrease in wavenumber for band 4, corresponding to the antisymmetric stretching vibration, as a function of the REE ionic radius in the FT-IR spectrum. This feature indicates lattice contraction of a configured structure (Figure 5b). There is evidence that the lattices of monazite, rhabdophane, xenotime, and churchite contract and appear as a shift in the bands derived from the P–O stretching vibration mode [34,36,44,45]. The lattice shrinks as it incorporates smaller REEs, and the P–O distance becomes slightly shorter, resulting in greater wavenumbers for this vibration. It is noted that HREPO_4,hyd is partially composed of amorphous matter in addition to material with a configured structure, as shown in Figure 3b, and only the structured portion contributes to the shift in the wavenumber. The fact that there was no correlation between the effective ionic radius and the antisymmetric stretching vibration wavenumber in the MREPO_4,hyd sample may indicate that no configured structure is present in MREPO_4,hyd. Although Lucas et al. [40] detected a structural transformation to churchite after 1 h of reaction at 50 °C in solution, no such structural transition was observed in the present study in aging experiments for HREPO_4,hyd at room temperature. This indicates that the HREPO_4,hyd structure is stable at room temperature and that structural transformation is slow on the time-scale of the present study.

The XRD analysis revealed that the structure of RE_allPO_4,hyd comprises a combination of the GdPO_4,hyd and TbPO_4,hyd structures (Figure 1b); however, the both structures in RE_allPO_4,hyd are not composed of the end member compositions of GdPO_4,hyd and TbPO_4,hyd. Rather they are composed of all REE at almost same composition as that in the starting reagents as evidenced by the EDX analysis (Figure 10a).

In the LaPO_4,hyd precipitation experiments, the crystallite size increased with increasing reaction time from 3 days to 30 days (Figure 1a,d and Figure 2a). This ripening behavior in crystallites was also seen for other REEs, as evidenced by the smaller full-width at half-maximum of diffraction peaks after aging (Figure 1a,d). The crystallite size also increased as pH decreased (Figure 1a,c and Figure 2b). This can be explained by the saturation ratio Ω:

$$\Omega ={\left(\frac{{\mathit{IAP}}_0}{K_{\mathrm{s}0}}\right)}^{\frac{1}{\eta }}$$

(2)

where IAP₀, K_s0, and η refer to the ion activity product (mol²·dm⁻⁶), the solubility product (mol²·dm⁻⁶), and the number of ions per formula unit, respectively [46]. Using the parameters K_s0 = 10^−25.7 [47] and η = 2 for LaPO₄, the value of Ω at pH 1 was calculated in the present study to be four orders of magnitude smaller than that at pH 5 due to the smaller amount of PO₄³⁻ (

Table 7. Saturation ratio, Ω of the reacting solution with respect to LaPO_4,hyd. Activities of RE³⁺ and PO₄³⁻ were calculated using Geochemist’s Workbench with the thermodynamic database minteq. The input data are 0.05 mM RE³⁺, 0.05 mM PO₄³⁻, 0.1 M Na⁺, and 0.5 M NO₃⁻.

pH	Activity (mol·dm−3)		Ion Activity Product IAP0	Saturation Ratio Ω
	aRE3+	aPO43−
1	2.06 × 10−3	8.80 × 10−21	1.81 × 10−23	3.01 × 101
3	1.29 × 10−3	4.95 × 10−16	6.37 × 10−19	5.65 × 103
5	1.26 × 10−3	5.16 × 10−12	6.53 × 10−15	5.72 × 105

). Stumm [46] recorded that Ω values greater than 1 exponentially induce faster formation of smaller nuclei; hence, lower pH produces larger crystal sizes.

4.2. Thermal Stability of RE Phosphates Formed at Room Temperature

Because secondary RE phosphate minerals that form during rock weathering in surface environments are frequently subjected to subsequent thermal events, the thermal stability of REPO_4,hyd that formed at room temperature was evaluated. In the present study, the structures of REPO_4,hyd were classified into three categories: LREPO_4,hyd corresponds to the rhabdophane structure, whereas the structures of both MREPO_4,hyd and HREPO_4,hyd could not be conclusively identified. The thermal stability of rhabdophane has been thoroughly investigated in previous studies and the rhabdophane structure is known to be stable up to ~500 °C–800 °C [15,48,49], ~750 °C [50]. The present study focused on the thermal stability of the unidentified structures, MREPO_4,hyd and HREPO_4,hyd, which were stable in air up to 100–300 °C on the time-scale of the experiments. The thermal stability of these two types of phosphates was comparable to the stability of YPO₄·2H₂O, in which the churchite structure transforms to the xenotime structure by annealing at ~100 °C for 73 h and at ~200 °C within 1 h [50,51]. Non-coordinated water seems to be unrelated to the crystal structure, based on a comparison between the TG-DTA charts and the XRD patterns of heated samples (Figure 4 and Figure 6). Although Mesbah et al. [52] reported that two-step dehydration processes of well-crystalline rhabdophane in detail, the present study demonstrated that structural transitions occurred during the second dehydration step (after the local minimum in the DTG spectrum) for LREPO_4,hyd and HREPO_4,hyd, or during crystallization (at the DTA peak of exothermic) for MREPO_4,hyd and RE_allPO_4,hyd. As noted above, the initial structure can survive up to a certain temperature, at least 120 °C. The TbPO_4,hyd structure does not possess a perfect rhabdophane structure, but persists up to 300 °C under dry conditions, although rhabdophane forms after 30 days of aging in solution (Figure 1d). Thus, REPO_4,hyd formed at room temperature is more stable under dry conditions than wet conditions. Therefore, when the average ionic radii are larger than that of Tb (1.04 Å), the initial structure of REPO_4,hyd is not stable and will transform to the rhabdophane structure by aging under wet conditions at room temperature.

4.3. Formation of REPO_4,hyd on Apatite at Room Temperature

In REPO_4,hyd formation experiments using apatite as a source of P, the dissolution rate of HAP in a RE³⁺ solution was determined to be 7.3 × 10⁻⁹ mol·m⁻²·min⁻¹ based on the rate of Ca release from HAP for a period of 3–24 h, as shown in Figure 8a. The pH of the solution remained within 5.5–5.6 during the period, after an initial increase in pH from ~5.1 in the first 3 h. The dissolution rate calculated in the present study is comparable with the dissolution rate determined by a previous study in pure water; 10⁻⁷–10⁻⁹ mol·m⁻²·min⁻¹ at pHs ranging from 2–7 [53], indicating that the presence of soluble REEs in solution does not significantly affect the dissolution rate of HAP.

The solution pH adjacent to the HAP surface could be slightly higher than that of the bulk solution. However, RE hydroxide was not characterized in the present study, indicating that the local pH at the HAP surface did not exceed ~8.

RE phosphates precipitated at the surface of apatite crystals through dissolution–precipitation mechanisms without the involvement of further mechanisms, such as epitaxial growth and pseudomorphism, which were previously reported in experiments on apatite interaction with aqueous Pb [30]. In general, epitaxial growth and pseudomorphism only occur when the crystallographic characteristics, such as the lattice parameter of the secondary precipitate, are closely related to those of the substrate and when a difference in solubility between the original and secondary mineral acts as the driving force [54,55,56]. The crystallographic parameters of the RE phosphate precipitates that were characterized in the present study (LREPO_4,hyd (rhabdophane), MREPO_4,hyd, and HREPO_4,hyd) are not related to those of HAP, preventing the occurrence of those two specific mechanisms. Thus, dissolution–precipitation on apatite surfaces without the lattice relationship is the only mechanism expected for the initial formation of hydrous RE phosphates at ambient temperature in nature. Indeed, rhabdophane crystals associated with apatite were found to be crystallographically unrelated to apatite in natural environments [23], whereas the dissolution–precipitation mechanisms at high temperature >300 °C were reported to exhibit the topotaxial lattice relationship between apatite and monazite [57,58].

5. Implications

In the present study, only the rhabdophane structure was conclusively identified, which was constrained by the average ionic radius calculated according to the proportion of each constituent REE. This finding can be applied to predict the initial structure of hydrated RE phosphates formed in natural surface environments. However, natural systems involve more complicated species, such as other ions and organic matter. Organic ligands can produce smaller particles [59]. Organic matter, in particular, has the potential to control the stability of the rhabdophane structure, as some previous studies have reported that the rhabdophane structure was stable even for HREE (Yb, Lu) when organic matter was added to the starting reagent [60,61,62] or mechanical treatment was used to introduce it to the churchite structure [49,63]. A natural occurrence of rhabdophane-Y was reported by Takai and Uehara [64]; however, the mineral they discovered contains a wide variety of other LREE including ~20 wt % La₂O₃ and 15 wt % Nd₂O₃ in addition to ~15 wt % Y₂O₃. Thus, their report is consistent with the results obtained in the present experiments, which conclude that the structure is determined by the average ionic radius for phases that include multiple REEs.

Aqueous REEs frequently become immobilized in phosphate minerals by the consumption of apatite or they are simply bound to aqueous P during alteration processes [27,64]. Under typical environmental conditions, the hydrated RE phosphates observed in the present study are expected to initially form at nanoscale sizes. As natural water is not supersaturated with regard to RE phosphates [65] when compared with the composition of solutions used in this study, actual RE phosphates may slowly precipitate as larger crystals locally near apatite grains in natural aquifers than those observed in the present study.

The observed development of REE patterns during the formation and aging of RE_allPO_4,hyd suggests that REE fractionation during the formation of nanoscale RE_allPO_4,hyd is plausible. Therefore, the REE pattern for rhabdophane is not identical to that of the reacting solution. This is of particular importance when mineral REE patterns are used to estimate certain geochemical conditions of the ambient environment, such as oxidation states, compared with REE patterns of bedrock [22]. If specific phosphate minerals such as rhabdophane are dominant in a weathering profile, the chemical composition of the RE phosphate may constrain the bulk REE composition of the weathered profile, as demonstrated by a previous field study [66]. In the present study, only REE fractionation into the rhabdophane structure was examined. Further experiments are needed in order to evaluate REE fractionation into other structures such as MREPO_4,hyd and HREPO_4,hyd by varying the average ionic radius over all REEs.

Previous studies highlighted the importance of biomineralization of RE phosphates in constraining the mobility of REEs in the environment [20,21,67]. In these studies, biogenic Ce phosphate was characterized as having a needle-shaped rhabdophane structure [20], while Yb phosphate precipitated as amorphous nanoparticles ~50 nm in size [21]. The structure and morphology of these phases are very similar to those of LREPO_4,hyd and HREPO_4,hyd observed in the present study. Hydrous RE phosphate precipitation as a result of microbial activity was also reported in a weathering profile of granite [67]. Given that the majority of P in extracellular substances occurs as orthophosphate [68], the crystallographic nature of biogenic RE phosphates may be mainly controlled by the combined effects of the inorganic reaction processes elucidated in the present study and the effects of organically-bound P species.