Partitioning of REE between calcite and carbonatitic melt containing P, S, Si at 650–900 °C and 100 MPa

Carbonatites host some unique ore deposits, especially REE, and fractional crystallization might be a potentially powerful mechanism for control enrichment of carbonatitic magmas by these metals to economically significant levels. At present, data on distribution coefficients of REE during fractional crystallization of carbonatitic melts at volcanic conditions are extremely scarce. Here we present an experimental study of REE partitioning between carbonatitic melts and calcite in the system CaCO3-Na2CO3 with varying amounts of P2O5, F, Cl, SiO2, SO3 at 650–900 °C and 100 MPa using cold-seal pressure vessels and LA-ICP-MS. The presence of phosphorus in the system generally increases the distribution coefficients but its effect decreases with increasing concentration. The temperature factor is high: at 770–900 °C DREE ≥ 1, while at lower temperatures DREE become below unity. Silicon also promotes the fractionation of REE into calcite, while sulfur contributes to retention of REE in the melt. Our results imply that calcite may impose significant control upon REE fractionation at the early stages of crystallization of carbonatitic magmas and might be a closest proxy for monitoring the REE content in initial melt.


Results
Carbonatitic melts in all the experiments quenched as aggregates of fine-grained dendritic crystals with abundant fluid bubbles probably dominated by CO 2 (Fig. 1, 2; Tables 2, 3). In phosphate-bearing samples, quenched melt is full of needle-shaped crystals of quench Ca-phosphate, presumably apatite (Figs. 1a, Fig. 2a-c). In experiments with SiO 2 fine crystals of wollastonite and Zr-Hf-silicate (zircon-hafnon) are also formed ( Fig. 2d; Table 2). Oxides of HFSE and REE in all samples form aggregates of fine crystals with complex composition (baddeleyite, perovskite-lueshite solid solutions and crystals of pyrochlore supergroup) (Fig. 2a-c). The crystals are few microns in size and are not suitable for representative analyzes by microprobe or LA-ICP-MS. Rounded, irregularly shaped, drop-like fluorite crystals are formed in samples with high content of F ( Fig. 2a; Table 2).
Calcite has various morphology: rhombohedral or prismatic (Fig. 1a, 2a), amorphous (Fig. 1b), needle-like (Fig. 1d,f) and tabular rounded crystals (laths) up to 250 µm in most experiments (Figs. 1c,e, 2c), which are typical for early generations of calcite in carbonatites 43 . The amount of calcite crystals in some samples is so high that they form a crystal mush at the bottom of the samples (Fig. 1c,e). Calcite major element composition is close to stoichiometric (Table 3).
Trace element composition of the run products presented in Table 4. Partition coefficients were calculated according to the Nernst formula as mass ratios of element content in a crystalline phase to its content in the melt: D Element = C Crystal /C Melt (Table 5). D Sr is > 1 for all samples (Fig. 3). In all experiments D Sr >> D REE , except NCPSi mixtures. Almost all D REE plots show a positive Eu and Y anomalies. In general, the slope of the D REE plots is positive, with the exception of NCPSi samples. D REE of calcites, published in our previous work 41 , are close to zero: 0.02-0.04 in NCF samples and 0.04-0.1 in sample with NCFP-6. In experiments with the NCFP mixtures, presented in this work, the averages of D REE are much higher: 0.1-0.2 and 0.52-1.39 (Fig. 3a). The averages of D REE in NCP samples vary in 0.55-1.4, but taking into account the analytical uncertainty for each individual element, it is almost the same (Fig. 3b). In experiments at highest temperatures the highest D REE are observed for both mixtures.

Discussion
Factors influencing the partitioning of REE into calcite in carbonatite systems. As seen from natural carbonatite samples, the early generations of calcites are characterized by increased content of cations, which size doesn't allow stoichiometrically replace Ca 2+ at crustal conditions: Mg, Sr, Ba are most abundant among them 43,44 . Sometimes it leads to formation of Ca-Ba-Sr "protocarbonates", which are unstable and decay into calcite and baritocalcite in the form of a solid solution. Consequently, elevated temperatures favor the entry of cations, whose size precludes their unlimited substitution for Ca 2+ . Along with these cations, REE may also enter the structure of calcite, which is also noted in natural samples: for example, the positive correlation between Ba, Pb, and LREE in early calcites of the Aley complex (Canada) 42,43 . The strong effect of temperature is confirmed by our experiments: in most experiments D REE > 1 at high temperatures (900-820 °C), while it is about unity or less at lower temperatures 41 .
As follows from a comparison with our previous experiments 41 , the presence of small amounts of phosphorus may be another positive and key factor for the fractionation of REE into calcite in carbonatite systems with high Na 2 O content. The presence SiO 2 in small amounts in the system may be another powerful driver for incorporating D REE into calcite. Sulfur, on the contrary, may promotes the dissolution (retention) of REE in the melt. However, to evaluate the total effect of these and other parameters (such as F, Cl) additional experiments are required.
Calcite as a monitor of REE. Our data show that, under the combination of some factors at the magmatic stage of evolution of carbonatite magma, calcite can retain a significant gross budget of REE and may be a closest proxy for estimation of initial REE content in the melt. For example, at high temperatures, in local areas of a carbonatite body, or when there is a lack of material for fractionation of early traditional mineral concentrators of a large amount of REE (such as apatite, monazite, etc.) even at low concentration of REE in the melt.
High susceptibility of calcite to postmagmatic changes, plasticity and recrystallization creates the possibility of releasing from calcite large bulk amounts of REE, Sr and Ba and their subsequent redeposition as a result of interaction with later melts or solutions or as a result of metamorphism. According to various estimates 42,43 , such processes can occur at shallow depths already at temperatures of 400-600 and even 730 °C, which can lead to    41 . About 45-55 mg starting mixtures were put in gold capsules and sealed by arc-welding in the flow of Ar. Experiments were performed at pressure of 100 MPa and temperature range of 650-900 °C in rapid-quench cold-seal pressure vessels at the German Research Centre for Geosciences (GFZ Potsdam). The temperature range is close to the parameters for nature cooling systems of carbonatitic melts. A detailed description of this type of the vessels described in 45 . The autoclaves at GFZ Potsdam are made of the Ni-Cr alloy Vakumelt ATS 290-G (ThyssenKrupp AG). Oxygen fugacity was not controlled, but believed to have been close to that of the Ni-NiO equilibrium, buffered by oxidation reactions of water, used as pressure medium, and the Ni-Cr alloy of the autoclave. Temperature measured by external Ni-CrNi thermocouple, calibrated against the melting temperature of gold. Temperature measurements are corrected for a temperature gradient inside the autoclaves, which was measured using a second inner thermocouple during the initial calibration of the vessels. The external thermocouple has an uncertainty of approximately ± 1 °C, and the total uncertainty of temperature measurements, including the uncertainties due to the temperature gradients, is estimated to be ± 5 °C. Pressure is measured by transducers, and results were checked against a pressure gauge. The transducers and gauge are factory calibrated and have an accuracy of better than ± 0.1 MPa. Run times varied from 20 to 141 h. In experiments with decreasing temperature, the temperature lowered gradually, with decrease of 5-7 °C in every 1-1.5 h. Samples were kept for at least 10 h at the final temperature before quenching. The quenching of the experimental samples in such apparatus is isobaric and less than a second.
After the experiments, samples were mounted in epoxy resin and polished with diamond polishing pastes without water to avoid dissolution of alkalis. Major components of the run products were analyzed using Cameca SX-50 and SX-100 electron microprobes in the GFZ Potsdam and energy-dispersive spectrometry (EDS) in combination with back-scattered electron imaging (BSE) using a MIRA 3 LMU SEM (TESCAN Ltd.) equipped with an INCA Energy 450 XMax 80 microanalysis system (Oxford Instruments Ltd.) in the Analytical Center for multi-elemental and isotope research SB RAS (Novosibirsk, Russia). Table 3. Major composition of run products (in wt.%). The numbers in the parentheses next to the phase name indicates the number of analyzes. The numbers in the parentheses next to the analyzes are the standard deviation and reported as the least unit cited. For example, 55.50 (64) should be read as 55.50 ± 0.64 wt.%. Fluorite (17) Table 4. Trace element composition of run products (in ppm). The numbers in the parentheses next to the phase name indicates the number of analyzes. The numbers in the parentheses next to the analyzes are the standard deviation. For example, 182 (48) should be read as 182 ± 48 ppm.   (1) 27 (1) 27 (1) 27 (1) 26 (1) 27 (1) 26 (1) 27 (1)   www.nature.com/scientificreports/ size of 24 µm, repetition rate of 8 Hz, and laser energy density of 5 J cm −2 . NIST SRM 610 was used as external standard and Ca as derived from EPMA as internal standard. BCR-2G was analyzed as the validation material known-unknown for accuracy control and its trace element concentrations are in agreement with published reference datavalues at < 10% 2σ RSD. Each analysis comprised 20 s background, 40 s ablation and 20 s wash-out. Data was processed using the trace elements IS data reduction scheme 46 in iolite 3.63 47 . Experimental experience gained with carbonatite systems under these conditions by us and by other researchers shows that the chosen duration of experiments is sufficient to achieve equilibrium in experiments with synthetic carbonatites due to the high rate of kinetic processes in them, especially with small experimental samples. Moreover, most of the duration of the experiments (from tens of hours to days) took place at the final temperature of the experiments to achieve equilibrium before quenching. During the subsequent analysis of the samples both by scanning microscopy and LA-ICP-MS, we did not find any zoning in the content of either the main components or rare elements in the center and on the periphery of the crystal phases.