INTRODUCTION

Rare-earth elements (REE), including lanthanide group (Ln), are ascribed to strategic metals. They are widely applied in different high-tech fields and are termed as “industrial vitamins” (Сhen et al., 2022). The limited REE resources highlight the problems of exploration of their deposits and improvement of mining methods, as well as rational exploitation of host secondary raw material. The solution of these problems requires knowing of REE behavior in heterogeneous systems, their partitioning and solubility constants, characteristics of absorption, and speciation in minerals and chemical compounds.

Magnetite and hematite are frequently present in the REE-bearing iron ore deposits (Tallarico et al., 2005; Harlov et al., 2016), including the largest deposits, for instance, the world-class Bayan-Obo Fe–REE–Nb deposit (Inner Mongolia, China) (Huang et al., 2015). Magnetite and hematite are usually not considered as the main REE carriers (Kulik and Melgunov, 1992; Zamanian and Radmard, 2016), since REE are mainly hosted in calcium minerals–-fluorite, calcite, dolomite, and apatite. In this relation, the REE partition and co-crystallization coefficients in iron oxides practically have not been studied. This is a significant omission given the useful properties of magnetite as an indicator of composition of mineral-forming medium (Smagunov et al., 2021). This work reports the first results of study of multisystem, which includes practically all lanthanides and Fe oxides, magnetite and hematite, under hydrothermal conditions. The values of partition and co-crystallization coefficients were obtained and Ln-rich phases associated with these minerals were studied.

EXPERIMENTAL PROCEDURE AND ANALYTICAL METHODS

Experiments were performed using conventional technique of thermogradient hydrothermal synthesis at 450°С and 100 MPa (1 kbar) coupled with internal sampling of a high-temperature fluid for determination of its composition (Tauson et al., 2018; Smagunov et al., 2021). The experiments were carried out in stainless steel autoclaves ~200 cm3 in volume designed at the Institute of Crystallography of the Russian Academy of Sciences (Fig. 1). The autoclaves were equipped with titanium (VT-8) passivated containers (inserts) ~50 cm3 in volume, the upper part of which contains titanium fluid trap. For the first four days, isothermal conditions were maintained to homogenize the system. For subsequent twenty days, a temperature drop of 15°С was applied to the external walls of the autoclave, which corresponds to the actual temperature gradient T of ~0.1 degree/cm for this configuration of experiment.

Fig. 1.
figure 1

Scheme of thermogradient hydrothermal synthesis.

Full size image

The experiments were completed with autoclave quenching in cold flowing water with a temperature drop of ~5 degree/s. When inserts were recovered, solution was extracted from trap (solution рН was 8.3–8.4) and the trap was rinsed with aqua regia. Then, it was combined with rinsed waters and analyzed by atomic absorption spectrometry (for Fe) and inductively coupled plasma mass spectrometry (for Ln). However, the capture of fluid in trap is not controlled and to certain extent, is an occasional event. For this reason, the amount of captures solution is frequently insufficient and it is required to repeat the experiment.

Ammonium chloride solutions (10%) was used as a mineralizing solution for synthesis of magnetite (runs 1 and 2), while a batch contained all lanthanides, except for promethium. In run 1, 5 g Fe2O3 + FeO in 0.8 mol ratio was added with 0.88 wt % of each Ln in form of oxide III (IV for Се). In run 2, 5 g Fe2O3 + FeO in 0.5 molar ratio was added with 1.56 wt % of each Ln in the same form.

Hematite (runs 3 and 4) was synthesized in run 3 using 5% NH4Cl solution and batch consisting of 5 g Fe2O3 and 0.88 wt % of each Ln in form of oxide III (IV for Се). In run 4, we used 10% NH4Cl, while batch contained 5 g Fe2O3 and 1.56 wt % of each Ln in the same form.

The starting materials were prepared from Fe2O3 and FeO of analytical grade, NH4Cl of high purity grade, and lanthanide oxides III and CeO2 (99.0 to 99.93% of main components) manufactured by different companies.

The Ln contents in solutions were determined by ICP-MS on a high-resolution ELEMENT-2 mass-spectrometer (Finnigan MAT, Germany) at the Institute of Geochemistry of the Siberian Branch of the Russian Academy of Sciences. Operation conditions were as follows: generator power of 1200 W, reflected power <4 W, the gas flow (argon) rate of 0.8–0.95, 16 and 0.9–1.2 L/min for transporting, plasma-forming, and auxiliary flows, respectively, the solution supply rate of 0.9 mL/min, Meinhard concentric nebulizer, quartz nebulizer chamber with cooling up to 3°С; sample introduction time of 60 s; spectra acquisition time 100–120 s, rinsing (3% HNO3) time of 120–240 s, sensitivity of 106 imp/s per 1 ng/mL In, and the in-house Rh standard (2 ng/mL). The instrumental parameters were optimized using a tuning solution that contained 1 ppb of each analyzed element. Concentrations were calculated using certified SPEX CLMS-1-4 solutions (USA) with element concentrations of 0.1, 1.0, and 5.0 ng/mL and signal drift control using the 103Rh in-house standard. The matrix effect was eliminated by dilution of solutions of analyzed samples. The detection limits were estimated using 3σ criterion (Table 1). The results were controlled using standard multielement solutions (MERCK, Germany; NIST, USA).

Table 1.   Detection limits of elements in solutions (ICP-MS) and in crystals (LA-ICP-MS)
Full size table

The Ln contents in magnetite and hematite crystals were determined by LA-ICP-MS on an Agilent 7500ce (Agilent Tech., USA) using a New Wave Research UP-213 laser ablation system at the Limnological Institute of the Siberian Branch of the Russian Academy of Sciences. The experimental parameters were as follows: plasma power 1500 W, gas flow rates of 1.5 L/min, laser power 80%, frequency 10 Hz, beam size 55 μm, and acquisition time of 15 s. Scanning method was three spots per mass, with counting time of 0.1 s per one spot.

The calculations of Ln contents were based on the NIST 612 standard, which is suitable for obtaining reliable results when trace element concentrations in magnetite are of few ppm and lower (Nadoll et al., 2014; Smagunov et al., 2021). The calculated values of detection limits are presented in Table 1, and standard deviations, in Tables 2 and 3.

Table 2.   Lanthanide contents in magnetite (LA-ICP-MS), co-crystallization coefficients Ln/Fe and crystal/fluid partition coefficients
Full size table
Table 3.   Lanthanide contents in hematite (LA-ICP-MS), co-crystallization coefficients Ln/Fe and crystal/fluid partition coefficients
Full size table

The results of LA-ICP-MS analysis in separate spots on crystal faces were processed using technique allowing the identification of evenly distributed component of Ln admixture formally corresponding to the structurally bound form (Tauson and Lustenberg, 2008). No stable results were obtained for Tm using this approach, and data on this element are constrained by its contents in solutions.

Ln-rich phases formed simultaneously with magnetite and hematite were studied by methods of Scanning Electron Microscopy using EDS Microanalysis System AztecLive Advanced Ultim Max 40 with nitrogen-free detector (Oxford Instruments Analytical Ltd., England) (SEM-EDS) and Electron Probe Microanalysis (EPMA) applying MIRA 3 LMH (TESCAN, Czech Republic) and Superprobe JXA 8200 (JEOL Ltd., Japan) devices at the Center for Collective Use “For Isotope and Geochemical Studies” in the Institute of Geochemistry, Siberian Branch of the Russian Academy of Sciences.

RESULTS AND DISCUSSION

During experiment, crystals of magnetite (hematite) mainly of octahedral (Mt) and rhombohedral (Hem) habit up to 1 mm in size (Fig. 2) were synthesized in the growth zone in association with lanthanide-rich phases represented by light yellow lamellar and dark red acicular crystals (Fig. 3).

Fig. 2.
figure 2

Synthesized crystals of hematite (a) and magnetite (b): main habit forms.

Full size image
Fig. 3.
figure 3

Own Ln phases in the hematite crystallization field (run 4). Lamellar crystals are hydrous oxychloride of mainly light and medium Ln, acicular crystals are oxyhydroxide of Fe and medium–heavy Ln, and equant rhombic crystals are hematite.

Full size image

Compositions of analyzed solutions (quenched from insert and collected in trap) are shown in Fig. 4.

Fig. 4.
figure 4

The Ln contents in fluid from trap (1) and in quenched solution from insert (2) in runs on synthesis of hematite (a, run 3) and magnetite (b, run 2).

Full size image

The Ln contents measured by LA-ICP-MS and calculated partition and cocrystallization coefficients of Ln are presented in Tables 2 and 3 and in Fig. 5.

Fig. 5.
figure 5

Experimental results on the Ln distribution between magnetite (Mt), hematite (Hem), and hydrothermal solution at 450оС and pressure of 1 kbar. (a) Ln/Fe cocrystallization coefficient, (b) crystal–solution partition coefficient of Ln.

Full size image

It is seen in Fig. 4 that lanthanides practically are not retained in solution after cooling (quenching). After cooling of system with hematite (logfO2 > –21.6 bar at 450°С), ~1.3–7.5% Ln are preserved (except for Ce and Eu, which preserve the high-temperature contents at 17 and 33%, respectively). In the system with magnetite (at lower fO2), the Ln content in quenched solution is slightly higher (3.3–18% for most elements). Thereby, 33% Ce is preserved, while Eu is completely preserved. This suggests the higher stability of Ce and Eu complexes in the solution compared to other lanthanides. At the same time, the Eu co-crystallization coefficient does not increase and even shows obvious minimum (Fig. 5a), which is inconsistent with its inferred strong complexing in solution. According to (Migdisov et al., 2009), Eu chloride complexes at 150–250°С show no anomalies in the stability constants relative to the neighboring Ln. In the lanthanide series, these constants decrease only for heavy Ln—Tm, Yb, and Lu. However, judging from the formation of Ln oxychloride and oxyhydroxide (see below), the transfer in our experiments was provided by not only pure chloride complexes LnCl2+ and \(Ln{\text{Cl}}_{2}^{ + },\)but also by hydroxychloride Ln(OH)Cl+. Thus, the europium minimum in the fluid–mineral systems can be related not only with the absence of Eu mineral carriers, but also with its lower co-crystallization coefficient compared to other lanthanides (at least, for magnetite and hematite).

The DCe/Fe maximum in hematite in run 4 can be explained by more oxidizing conditions at increasing addition of Ln oxides in batch. This suggests the possible appearance of significant amount of Ce4+ that is more strongly complexed in solution than Ce3+. This is confirmed by the discovery of both these species on the hematite surface at similar conditions using X-ray photoelectron spectroscopy (Tauson et al., 2018). It should be however kept in mind that the Ce content in hematite in run 4 is close to the detection limit (Table 1) and even lower than detection limits in other cases.

Except for Eu, lanthanides are compatible in magnetite. Thereby, Er, Yb, and Lu are characterized by the relatively high degree of coherence (Table 2). Experiments with hematite yield slightly different results, but tendency to the growth of coefficient with increasing atomic number Ln is preserved and even intensified compared to magnetite (Table 3). Heavy lanthanides, beginning from Tb, are compatible in hematite. The assignment of element to compatible (coherent) and incompatible (incoherent) is based on its partition coefficient between solid phase and solution: Ds/aq > 1 and Ds/aq < 1, respectively.

At present, it is difficult to determine whether the obtained estimates correspond to structurally bound admixture of Ln in minerals. Although LA-ICP-MS analysis was performed for crystals with smooth pure faces, the limited area of the latter due to small crystal size could cause partial capture of surface component, which is more abundant than structurally bound form (Smagunov et al., 2021). This is partially supported by the high dispersion of data in run 1 (Table 2), where magnetite crystals were smaller and the number of analyzed spots was likely insufficient. The EPMA and SEM-EDS analyses revealed no the presence of REE-rich phases in the magnetite and hematite crystals. While processing LA-ICP-MS data, we attempted to recognize the structurally bound REE component using technique proposed in (Tauson and Lustenberg, 2008). Unfortunately, due to a few number and small size of crystals, we failed to apply the technique allowing complete omitting of the surface component of REE admixture. For this reason, the obtained data are qualified as maximum estimates of “true” partition and co-crystallization coefficients for structurally bound Ln admixtures. Nevertheless, the application of obtained data for the determination of Ln proportions in hydrothermal solutions in any case is more justified than the use of low-temperature absorption partition constants Kd (e.g., Alibert, 2016). For hematite in NaCl solution, Kd for light and medium Ln are over three orders of magnitude higher than partition coefficients obtained by us, whereas Kd for heavy Ln (Yb, Lu) seem to be comparable with data in Table 3 (Tao et al., 2004). It should be reminded that the REE behavior at elevated temperatures differs from theoretical predictions, in particular, Pearson rules (Williams-Jones et al., 2012). In addition, this difference can be explained by different mechanisms of admixture uptake under normal or nearly normal conditions (adsorption) and in hydrothermal thermogradient systems, when the crystal growth is provided by the surface non-autonomous phase (Tauson et al., 2019).

Note especially a clear tendency to increase in co-crystallization coefficients for heavy Ln, beginning from Gd–Tb (Fig. 5). To some extent, this can be caused by the decrease of solubility of their oxides (spinels) of the corresponding stoichiometry and structural type relative to the solubility of ferruginous minerals (Chernyshev, 1980). Indeed, experimental results show that an increase of DLn/Fe for heavy Ln is accompanied by the decrease of their contents relative to Fe in fluid and, to lesser extent, an increase of contents in the magnetite and hematite crystals. The latter can be explained by the effect of lanthanide compression: for instance, the Fe3+ substitution in the high-spin state in magnetite for Ln3+ in the octahedral site is controlled by the relative deviation of the effective ion radii by 37% for La and lower value (24%) for Lu.

The discovered effect could have important geochemical implication. It is believed that LREE have the higher geochemical mobility and are mainly transferred for greater distances from magma source compared to HREE (Williams-Jones et al., 2012). This would help to reveal the primary sources of ore elements. However, if fluid discharge will be accompanied by the formation of Fe oxide minerals, their subsequent analysis will show the HREE enrichment relative to LREE regardless of the distance to magmatic body due to the higher DLn/Fe for HREE. Therefore, the application of this typochemical criterion (proportions of heavy and light Ln) for identification of REE source requires knowing their speciation and coefficients of co-crystallization in minerals—their potential carriers. The observations of own lanthanide minerals in ore occurrences could be helpful in this context.

In our experiments (both in the magnetite and hematite field), own Ln phases were synthesized simultaneously (as indicated by their intergrowths, Fig. 3) but have contrasting compositions. These phases have no analogues among known hydrothermal REE minerals mentioned in overview (Migdisov et al., 2016). At present, their study has not been completed yet, but according to preliminary data, yellow transparent crystals are aqueous Ln oxychloride, which can be represented by formula LnClO∙H2O, where Ln are mainly light and medium lanthanides (La, Pr, Nd, Sm, Eu, Gd, and Tb) (Fig. 6а). Red acicular crystals are represented by Fe and Ln oxyhydroxide and, in contrast, are enriched in heavy Ln. Their idealized formula is (Fe2/3Ln1/3)OOH, where Ln are represented by heavy and medium lanthanides (Lu, Yb, Ho, Tb, Gd, Eu, Sm, Nd) (Fig. 6b).

Fig. 6.
figure 6

Energy-dispersive X-ray spectrum (SEM-EDX) of lamellar (a) and acicular (b) crystals associated with magnetite and hematite (Fig. 2).

Full size image

This example demonstrates that the coexistence of light and heavy REE within a single hydrothermal system is provided by the simultaneous crystallization of phases that selectively incorporate light and heavy lanthanides.

It is important for further experimental studies that an increase of Fe transfer in form of oxides in order to obtain larger crystals can be reached by eliminating the competitive Fe-bearing phase via transition to the more “acid” hydrothermal solution.

CONCLUSIONS

(1) The partition and co-crystallization coefficients of Ln in magnetite and hematite under hydrothermal conditions at 450°С and 1 kbar are determined, which are interpreted as maximum estimates of “true” coefficients for structurally bound admixtures.

(2) Since the majority of lanthanides are not retained in solution after quenching, the experiment on synthesis of iron oxide crystals should be accompanied by sampling of high-temperature fluid.

(3) Lanthanides (except for Eu) are compatible in hydrothermal magnetite. Thereby, Er, Yb, and Lu are characterized by the relatively high degree of coherence. Heavy Ln, beginning from Tb, are coherent elements in hematite.

(4) Tendency to increasing co-crystallization and partition coefficients for heavy Ln, beginning from Gd–Tb, is typical of both Fe oxides. This example shows that the application of such typochemical guide as proportions of heavy and light Ln for determining REE source requires knowing of their speciation and co-crystallization coefficients in minerals that are the potential REE carriers.

(5) Own Ln phases obtained in association with magnetite and hematite exemplifies the coexistence of light and heavy REE within a single hydrothermal system owing to the simultaneous crystallization of phases selectively incorporating light and heavy elements.

(6) An increase of acidity of mineralizing hydrothermal solution is likely required to provide the Fe transfer in favor of its oxides and, respectively, to obtain larger Ln-doped crystals.