Carbonatite complexes are globally significant sources of rare earth elements (REEs); however, mechanisms governing REE deposition in various tectono-lithologic settings, encompassing host rocks, wall rocks, ore-controlling structures, and metasomatism, remain inadequately understood. The Zhengjialiangzi mining camp, situated within the extensive Muluozhai deposit (containing 0.45 million metric tons [Mt] at 4.0 wt % REE2O3) in the northern segment of the Mianning-Dechang belt, Sichuan (southwestern China), is characterized by a complex vein system that evolved within metamorphosed supracrustal rocks of the Yangxin and Mount Emei Formations. The mineralization is coeval with Oligocene intrusions of carbonatite and nordmarkite at ~27 Ma. The major gangue minerals include fluorite, barite (transitional to celestine), and calcite, with bastnäsite serving as the primary host for REEs in all analyzed orebodies. Several other accessory to minor minerals were identified in the ore veins, including some that had not previously been known to occur in the Muluozhai deposits (e.g., thorite and pyrochlore). The stable isotopic (C-O-Ca) and trace element compositions of calcite, along with whole-rock data, suggest that carbonate material was derived from the mantle and subsequently reequilibrated with the Yangxin marbles. The radiogenic isotope (Sr-Nd-Pb) compositions of vein material remained unaffected by wall-rock contamination and suggest a mantle source influenced by crustal recycling, consistent with other REE deposits hosted by carbonatite and nordmarkite in the region. The combined petrographic and geochemical evidence suggests derivation of Muluozhai mineralization from a carbonatitic source and interaction of carbonatite-derived fluids with wall rocks, xenoliths, and early-crystallizing mineral phases, particularly barite.

Globally, around 600 carbonatite localities have been identified (Humphreys-Williams and Zahirovic, 2021). However, only a limited number of these localities host economically viable rare earth element (REE) deposits (≥100 kt REO, where REO denotes total rare earth oxide content). China boasts a diverse range of REE deposits, playing a prominent role in the global REE market since the 1980s (Chakhmouradian and Wall, 2012; Kynický et al., 2012). Notable examples include Bayan Obo in Inner Mongolia (Yang, X.Y., et al., 2009, 2017; Yang, K.F., et al., 2011a, b; Ling et al., 2013; Smith et al., 2015), a cluster of intrusions in the Mianning-Dechang belt of Sichuan, China (Hou et al., 2009, 2015; Xie et al., 2009; Liu and Hou, 2017; Liu et al., 2019a, b), and Weishan in Shandong (Jia and Liu, 2019; Zeng et al., 2022). Some of these deposits are geologically complex, often linked with, or even hosted by, intrusive carbonatites. This deposit type represents the most economically important form of light REE (LREE = La-Eu) mineralization in the world, constituting approximately 65% of China’s reserves (Weng et al., 2015).

The genesis of these carbonatite-related REE deposits was not a simple one-stage process of ore deposition but involved a sequence of juxtaposed fortuitous tectonic, magmatic, and hydrothermal factors. These factors have been explored at length for Bayan Obo, China (Smith and Henderson, 2000; Smith et al., 2015; Yang et al., 2019; Wei et al., 2022), Mountain Pass, USA (Poletti et al., 2016; Watts et al., 2022), Maoniuping, China (Hou et al., 2006, 2015; Liu et al., 2019a, b; Zheng and Liu, 2019; Weng et al., 2021), Bear Lodge, USA (Moore et al., 2015; Andersen et al., 2016), and Amba Dongar, India (Hopp and Viladkar, 2018). Hydrothermal alteration and/or supergene weathering, as seen at Mount Weld, Australia (Jaireth et al., 2014) and South Africa (Venter et al., 2010; Harper et al., 2015), have significantly contributed to the REE concentration, resulting in a complex variety of rocks and mineralization styles (Kynický et al., 2012; Xie et al., 2019; Jia and Liu, 2020). Therefore, understanding the formation of carbonatite-related REE deposits is a complex endeavor, necessitating integrated studies of the primary host rocks and wall rocks and the microtextural and microchemical variations of constituent minerals within ore veins (Chakhmouradian and Zaitsev, 2012; Goodenough et al., 2021; Anenburg et al., 2022).

Until now, the behavior of REEs in carbonatite magma sources, their fractionation, and their hydrothermal dispersal or concentration have remained insufficiently constrained owing to a limited number of experimental (Veksler et al., 1998; Chebotarev et al., 2019, 2022; Anenburg et al., 2020a; Nabyl et al., 2020; Yuan et al., 2023), modeling (Hou et al., 2009; Xie et al., 2009; Walter et al., 2021; Zheng et al., 2023), and natural case studies (Bühn, 2008; Chakhmouradian et al., 2015, 2016a, b; Giebel et al., 2017, 2019a; Jyoti et al., 2019; Yang et al., 2019; Beard et al., 2022; Walter et al., 2022, 2023). Recently, it has been suggested that complexation with alkalis, rather than with anionic species such as halogens, is critical for significant REE transport in and around carbonatites as a precursor to economic-grade mineralization (Anenburg et al., 2020a, 2022; Yuan et al., 2023). Some carbonatites, however, exhibit initial magmatic deposition of REE phases (Wall and Mariano, 1996; Zaitsev et al., 2002; Weng et al., 2021; Zheng et al., 2023) or REE sequestration by calcium minerals (e.g., apatite; Bühn, 2001; Chakhmouradian et al., 2017). The compatibility of REEs in these minerals may be sensitive to independent compositional variables, such as silica activity (Pan and Fleet, 2002; Giebel et al., 2019a; Anenburg et al., 2020b). The identification of ore deposition process in each specific case may be further complicated by subsolidus processes, which may result in the erasure of primary petrographic characteristics (Chakhmouradian et al., 2016a, b).

The Zhengjialiangzi mining camp, within the Muluozhai deposit in the Cenozoic Mianning-Dechang belt (Fig. 1), presents a remarkable target for investigating the genesis of REE mineralization hosted by carbonatites. Its significance stems from its young age (~27 Ma; Liu et al., 2015a; Ling et al., 2016; Fu et al., 2019), exposure due to mining and exploration, and juxtaposition of igneous and hydrothermal mineral assemblages, which document a well-defined, traceable formation sequence. Furthermore, this camp holds significant, as yet unexplored, potential for REE resources. A complex carbonatite-nordmarkite intrusion was recently identified beneath the currently active underground operations, hosting mineralization within intricate vein systems.

Nevertheless, there is little published research concerning the genesis of ore veins in marbles within the Muluozhai deposit (Tian et al., 2006; Fu et al., 2019), including fluid inclusion studies (Zheng et al., 2019). Significantly, field relationships documented in the literature primarily rely on outdated geological survey reports, offering limited insights into the vein systems. Consequently, the origin and evolution of REE mineralization at Muluozhai remain elusive. Moreover, at Muluozhai, marble serves as the major host rock for mineralization, in contrast to most carbonatite-related REE deposits worldwide where silicate wall rocks predominate.

This study involves field, petrographic, mineralogical, and geochemical data for the mineralized vein system in the Zhengjialiangzi camp, which forms part of the Muluozhai deposit. In this contribution, mineralogical, trace element, and isotopic (Sr-Nd-Pb, C-O, and Ca) characteristics of the principal rock types are systematically described. These new data are used to constrain the source, distribution, structural controls, and evolution of REE mineralization, as well as provide further insights into REE metallogeny in orogenic systems.

The collision between India and Asia between 65 and 50 m.y. ago resulted in the formation of the Himalayan orogen (Yin and Harrison, 2000; Zhong et al., 2001; Mo et al., 2003) and had a significant impact on the western margin of the Yangtze craton (Yin and Harrison, 2000; Hou et al., 2003, 2006; Hou and Cook, 2009). This collision also led to the emergence of the N-S–oriented Jinpingshan orogenic belt (Luo et al., 1998). The Mianning-Dechang metallogenic belt formed between ~27 and ~11 Ma (Tian et al., 2008; Ling et al., 2016; Liu and Hou, 2017; Fu et al., 2019) and is situated approximately 100 to 200 km to the east of the Jinpingshan suture zone (Hou et al., 2009). In this region of the Yangtze craton, the basement consists primarily of high-grade Archean metamorphic rocks and Proterozoic metasedimentary rocks, covered by Phanerozoic clastic and carbonate sequences (Cong, 1988; Luo et al., 1998). This region was also influenced by the Permian Emeishan mantle plume, resulting in the formation of flood basalts and layered mafic intrusions (Shellnutt, 2014). However, the Mianning-Dechang belt originated within a collisional setting rather than a continental rift environment (Hou et al., 2009). Tectonically, the collision between India and Asia led to the development of a series of Cenozoic strike-slip faults, which acted as pathways for mantle-derived melts, including lamprophyres that formed between 40 and 24 Ma (Guo et al., 2005; Wang et al., 2012) and generally younger carbonatites (27–11 Ma; Liu and Hou, 2017).

The Mianning-Dechang intrusive complexes and their associated REE deposits occur as carbonatite dikes and stocks associated with silica-oversaturated, quartz-bearing syenites (nordmarkites) that intruded the Precambrian crystalline basement (Hou et al., 2006). The spatial distribution of the carbonatite complexes and related REE deposits was controlled by the N-S–trending Mianning-Dechang strike-slip fault system.

General geology

The Muluozhai REE deposit is located in the northern segment of the Mianning-Dechang belt and controlled by the strike-slip Yalongjiang fault, which connects with the Xianshuihe fault farther to the north. At Muluozhai, zircon in nordmarkite, bastnäsite, and fluorophlogopite in ore veins gave ages of 26.77 ± 0.32 (Liu et al., 2015a), 26.9 ± 0.2 (Ling et al., 2016), and 27.11 ± 5.3 Ma (Fu et al., 2019), respectively. The Muluozhai deposit contains 0.45 Mt of REOs at an average grade of 3.97 wt % (Hou et al., 2009) and can be subdivided structurally into the Diaolaoshan, Fangjiabao, and Zhengjialiangzi camps, which collectively consist of about 18 orebodies (Fig. 1). All these orebodies range 40 to 440 m in length and 1 to 30 m in width, and their grade is <5 wt % REOs. The ore consists predominantly of bastnäsite-barite-fluorite-calcite veins with an average grade of ~4.0 wt % REOs (i.e., ~5 modal % of bastnäsite). Field observations suggest that the deposit is situated within a monoclinal structure that tilts westward and, to some extent, is disrupted by later faulting. By reconstructing the three-dimensional structure of the deposit based on cross sections (Figs. 24), it becomes apparent that REE mineralization is accommodated by faults. In the Zhengjialiangzi camp, mineralized veins spatially follow the Chapuzi fault (F1) and three groups of related secondary faults that strike north-northeast to south-southwest, northeast to southwest, and northwest to southeast (Fig. 1). Geologic mapping in the Muluozhai area reveals the presence of Permian marine and subaerial supracrustal rocks located to the east of the F1 fault, near the deposit. Two Permian units are distinguished (Fig. 3): the Yangxin Formation and the overlying Mount Emei Formation. The Yangxin Formation consists of gray to white finely foliated marble in its upper part, blocky marble with black laminae in the middle section, and gray to white, coarse-grained granular dolomite marble at its base. Intrusions of nordmarkite and metadiabase crosscut the marble, and the mineralized veins are confined to fissures in the marbles, as well as their contact zone with metadiabase blocks. The younger Mount Emei Formation is exposed in multiple locations around the deposit. It consists of an upper section characterized by grayish-green metabasalts, chlorite schists, and extensive marble lenses with a thickness of 400 to 500 m. The lower part of this formation, with a thickness of 500 to 600 m, is composed of grayish-green massive metabasalt interspersed with amygdaloidal metabasalt. These rock types correspond to various eruptive cycles, each yielding a greenish variety at the base and a reddish, oxidized variety at the top (Fig. 3).

The study area contains four major lithological units (Fig. 3): (1) Cenozoic nordmarkite intrusions, which serve as the host rock for REE mineralization in some areas, are present in all three mining camps as discrete bodies <700 m in both length and width; these nordmarkite bodies intruded the Permian strata and Mesozoic granites; (2) Cenozoic granite porphyry, which lacks mineralization; (3) metadiabase intrusions, which were emplaced within the Permian basalts and marbles and were subsequently intruded by the nordmarkite (veins and veinlets containing REE minerals are locally present along the fractured contact zones of the metadiabase bodies); and (4) a granite pluton, emplaced to the southeast of the Zhengjialiangzi camp and adjacent to the unexposed nordmarkite intrusion.

Rare earth element mineralization is primarily located within fractured contact zones between the nordmarkite intrusions and metadiabases, as well as between metadiabases and marbles. Not all of the rock types reported by Ouyang et al. (2018) can be correlated with those identified in the present work. It is likely that their “alkaline granite” is a variety of nordmarkite, whereas their “monzogranite” is probably our granite porphyry. Nonetheless, the entire spectrum of granitic rocks analyzed by Ouyang et al. (2018) gave a tight cluster of emplacement ages, ranging from 26.98 ± 0.64 to 25.80 ± 0.18 Ma, with the youngest value corresponding to nordmarkite. It should be noted that the marbles and chlorite schists are older than 230 Ma (Ouyang et al., 2018).

Zhengjialiangzi vein system

In contrast to various ore types (breccias, stockworks, and disseminated ores), which occur elsewhere in the Mianning-Dechang belt and are hosted by nordmarkite intrusions, REE ores in the Zhengjialiangzi camp are hosted primarily by marbles and, to a lesser extent, schists and nordmarkite, as shown in cross sections A-Aʹ and B-Bʹ (Figs. 1, 2). Cross sections C-Cʹ and D-Dʹ (Figs. 1, 4) clarify the nature of the mineralized vein system on the basis of detailed observations within mine tunnels penetrating several orebodies. In the Zhengjialiangzi camp, REE mineralization is confined to a structurally complex system of anastomosing veins, veinlets, and stockworks that developed in tensional fissures, which are locally brecciated. The veins hosted by the marbles are numerous and show a variable thickness (1–30 cm), with some extending into the schist, nordmarkite, or their contact zones (Figs. 4, 5). The contacts between the veins and their host rocks are sharp, and many veins exhibit a banded, axially quasi-symmetric structure (Fig. 5), which suggests that they were deposited by the infilling of periodically opened tensional fissures.

Representative samples were selected from different vein types and several orebodies in the Zhengjialiangzi camp. Monomineralic fractions for grain mounts were prepared using conventional methods (crushing, sieving, heavy liquid separation, and mineral separation under a microscope), and polished thin sections were cut for microscopic examination and detailed petrographic analysis.

Cathodoluminescence (CL), backscattered electron (BSE) images, and electron microprobe analysis

Cathodoluminescence images were acquired using a Reliotron cold-cathode instrument at the University of Manitoba, Canada, operated in regulated mode at 7.5 kV and 400 to 500 mA, with an electronically cooled Nikon DS-Ri1 camera. BSE images (Figs. 68) and major element analyses of minerals (Tables 13) were acquired using a JEOL JXA-8230 electron probe microanalyzer (EPMA) at the Institute of Mineral Resources, Chinese Academy of Geological Sciences (CAGS), in Beijing, China, and a Cameca SX100 instrument at the University of Manitoba. Both instruments were operated at an accelerating voltage of 15 kV and beam currents of 10 nA (for apatite) and 20 nA (for other minerals), with beam sizes adjusted to avoid volatile loss and ranging from 1 to 10 μm. Calibration standards for the EPMA were carefully selected to minimize matrix effects and included the following natural and synthetic reference materials: jadeite (NaKα and AlKα on TAP, SiKα on PET), forsterite (MgKα, TAP), topaz (FKα, LED1), fluoroapatite (PKα, PET), K-feldspar (KKα, PET), wollastonite (CaKα and SiKα PET), hematite or fayalite (FeKα, LIF), MnO or spessartine (MnKα, LIF), rutile (TiKα, LIF), barite (SKα and BaLα, PET), SrTiO3 (SrLα PET), Na-Ba niobate (NbLα, PET), NaCl (ClKα, PET), thorite (ThMα, PET), and metallic Ta and U (TaMα and UMα PET). For bastnäsite and other REE-bearing phases, synthetic orthophosphate standards were employed to measure the abundances of individual REEs (Y, La, Ce, Pr, Nd, and Sm). Heavier lanthanides were sought but found not to be present at levels detectable by the EPMA in any of the minerals. Matrix corrections for fluorite and bastnäsite were carried out using the PAP correction procedure at the University of Manitoba with the rest of the minerals using ZAF correction procedure in the Institute of Mineral Resources, CAGS, in the microprobe operating software.

Laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS)

The same grains previously analyzed by EPMA were used for trace element analysis by LA-ICP-MS, with the assistance of sample maps to ensure precise beam placement; BSE, CL, and reflected-light images corresponding to areas analyzed by the EPMA were employed for that purpose. After considering potential spectral overlaps and molecular interferences, the following isotopes were chosen for analysis: 7Li, 9Be, 10B, 43Ca, 45Sc, 51V, 55Mn, 59Co, 63Cu, 66Zn, 85Rb, 88Sr, 89Y, 90Zr, 93Nb, 137Ba, 139La, 140Ce, 141Pr, 143Nd, 147Sm, 151Eu, 155Gd, 159Tb, 163Dy, 165Ho, 167Er, 169Tm, 172Yb, 175Lu, 208Pb, 232Th, and 238U.

Trace element analyses of calcite and barite (Table 4) were performed using a Teledyne Cetac Technologies Analyte Excite laser ablation system (Bozeman, Montana, USA), in conjunction with an Agilent Technologies 7700x quadrupole ICP-MS (Hachioji, Tokyo, Japan). Helium (370 mL/min) was used as a carrier gas to efficiently remove aerosols out of the ablation cell and was mixed with argon (~1.15 L/min) via a T connector before entering the ICP torch. Instrument configurations for Agilent Technologies 7700x included a 193-nm ArF excimer laser with a fluence of 4.5 J/cm2 and a 20-s background measurement (gas blank), followed by a 40-s analysis at a spot diameter of 33 µm and a repetition rate of 6 Hz. The U.S. Geological Survey basaltic glasses BIR-1G, BHVO-2G, and BCR-2G were employed as external calibration standards, and calibration and quality control utilized the synthetic glass standard NIST SRM 610. The oxide production rate was monitored during instrument calibration by measuring the ThO/Th ratio and keeping it below 0.2%. The raw data were subsequently processed offline using the ICPMSDataCal software, applying a 100% normalization strategy without an internal standard (Liu et al., 2008). The Chinese Geological Standard Glasses CGSG-1, CGSG-2, CGSG-4, and CGSG-5 (prepared by the National Research Center for Geoanalysis, Beijing, China) were employed for quality control (Hu et al., 2011).

The compositions of fluorite and bastnäsite were determined at the University of Manitoba using a 213-nm Nd-YAG Merchantek laser connected to a Thermo Finnigan Element 2 sector field-mass spectrometer. Data were collected using spot analyses with a 30- to 40-μm beam at a repetition rate of 5 to 10 Hz and power level of 80 to 85%. The incident pulse energy was 0.03 to 0.07 mJ, yielding a surface energy density of 4.0 to 5.6 J/cm2. Ablation was performed using Ar (plasma and auxiliary) and He (sample). The oxide production rate was monitored during instrument calibration by measuring the ThO/Th ratio and keeping it below 0.2%. Trace elements quantified included Sc, Mn, Zn, Sr, Y, Ba, lanthanides, Pb, Th, and U. Calibration and quality control utilized the synthetic glass standard NIST SRM 610. Analyses were conducted in a low-resolution mode (~300) using Pt skimmer and sample cones. Data reduction was carried out online using the GLITTER software (Van Achterbergh et al., 2001) and an in-house Excel-based program. Quality control was ensured by keeping fractionation at less than 10% and fractionation/error ratio at less than three.

Bulk geochemical analysis

Bulk geochemical analyses of typical ore samples, fresh marble and metasomatized exocontact marble (Table 5), nordmarkite, porphyritic granite, and greenschist (Table 6) were performed at the National Research Center for Geoanalysis, CAGS. Pulverized samples (0.7 g) were mixed with 5.3 g Li2B4O7, 0.4 g LiF, and 0.3 g NH4NO3 in a porcelain crucible. The powder mixture was transferred to a platinum alloy crucible, and 1 mL LiBr solution was added to it. The fluxed sample was then dried and melted in an automatic flame-fusion furnace and quenched to glass. The glass was analyzed using a PE8300 inductively coupled plasma-optical emission spectrometer for major elements; analytical errors were estimated to be <2% relative.

The following procedure was adopted for the analysis of REEs and other trace elements. Pulverized samples (50 mg) were dissolved in 1 mL HF and 0.5 mL HNO3 in 15-mL Savillex Teflon screw-cap capsules at 190°C for 1 day and dried, digested with 0.5 mL HNO3, and then dried again. The capsule content was digested with 0.5 mL HNO3 and dried again to ensure complete digestion. Then, the sample was digested with 5 mL HNO3 and held at 130°C in an oven for 3 h. After cooling, the solution was transferred to a plastic bottle and diluted to 50 mL before analysis. The sample solutions were analyzed by ICP-MS using a PE300D instrument. Standards GBW07120, GBW 07103, GBW 07105, and GBW 07187 were measured together with the samples; precision within 5% relative was achieved (Hu et al., 2011).

C-O isotope analysis

The C and O isotope compositions of carbonates in fresh and altered marble and ore samples (Table 7) were analyzed using a ThermoFinnigan GasBench II equipped with a CTC CombiPAL autosampler and linked to a Finnigan MAT253 mass spectrometer at the Institute of Mineral Resources, CAGS. Sample powders were loaded manually into 12-mL round-bottomed borosilicate vials and sealed using butyl rubber septa. The automatic injection tray holding 88 vials, including 18 aliquots of four national standards (GBW04405, GBW04406, GBW04416, and GBW04417), was employed in each run. The vials were automatically flushed with pure He (99.999%) for 10 min at a flow rate of 100 mL/min. Each sample was then reacted with 100% H3PO4 at 72°C for at least 4 h to liberate CO2. The CO2 was injected into a loop, separated via a chromatographic column (Poraplot Q, 25 m × 0.32 mm, Varian Ltd.) at a constant temperature (T) of 70°C, dried using a Nafion membrane, and analyzed with a MAT 253 isotope ratio-mass spectrometer (Thermo Scientific). The measurements of 18 reference carbonates indicate that the analytical precision was better than 0.1‰ for both δ13C and δ18O values. The analytical results are reported using the standard delta notation, in ‰ relative to the Vienna standard mean ocean water (V-SMOW) and Pee-Dee Belemnite (V-PDB).

Calcium isotope analysis

The Ca separation protocol used in this study was a modified version of the methodologies described by Chu et al. (2006), Owen et al. (2016), and Nan et al. (2015). The separation column comprised a 30-mL Teflon microcolumn (6.4 mm internal diam) filled with 2 mL of AG50W-X12 resin (200–400 mesh, Bio-Rad). The columns were precleaned with 20 mL 6 mol L–1 HNO3 and 3 mL 6 mol L–1 HCl and then conditioned with 3 mL 18.2 MΩ H2O prior to sample loading.

The ~1-mL sample solutions containing ~100 μg Ca were loaded and 19 mL 2 mol L–1 HCl was used to elute the matrix elements, including Na, Mg, K, Fe, and Mn. Calcium was then collected with 18 mL 2 mol L–1 HCl, which was followed by elution of trace Ca with 2 mL 3 mol L–1 HCl. The 2-mL aliquots were collected and measured both before and after the Ca extraction for their Ca contents to test whether the Ca elution curve drifted during the chromatographic process. After the separation, Ca collections were dried and dissolved in concentrated HNO3, evaporated to dryness three times and finally dissolved in 0.3 mol L–1 HNO3 for isotope analysis.

Calcium isotope analysis was undertaken using a Nu Instruments multicollector ICP-MS at the Laboratory of Isotope Geology, CAGS. The instrument was operated at low resolution to quantify 42Ca, 43Ca, and 44Ca isotopes. The standard sample bracketing of Belshaw et al. (2000) was used to correct for mass discrimination using NIST 915a or NIST 915b as a reference standard, with sample and standard solutions being matched to give 42Ca intensities with differences of <10%. All samples were analyzed three to five times during the same session. The external precision of Ca isotope ratios (44Ca/42Ca) was estimated to be better than 0.05‰ (2SD) based on the repeated measurements of NIST 915b. Calcium isotope values (Table 8) can be converted using the expression δ44/40Ca = 2.048 × δ44/42Ca (Heuser et al., 2016). In this study, the measured Ca isotope ratios are presented as δ44/40Ca relative to NIST SRM915a:

δ44/40Ca()=[(44Ca/40Ca)sample/(44Ca/40Ca)NIST915a1]×1,000.
(1)

Sr-Nd-Pb isotope measurements

The Sr-Nd-Pb isotope compositions were measured at the Laboratory of Isotope Geology, CAGS. For the Rb-Sr and Sm-Nd isotope analyses, selected rock samples were crushed, homogenized, and pulverized. The sample powders were spiked with mixed isotope tracers and then dissolved in Savillex with HF + HNO3. After dissolution, the samples were dried to remove HF and dissolved with HCl. The Sr and REE fractions were separated in solution using cationic ion exchangers AG 50 W (H+) (200–400 mesh) in columns. Then, Sr and Nd were collected in the proper sequence via leaching with 4 mol mL−1 HCl. The Sr fraction was purified using the above-described procedure for complete separation. Then, the solution was dried for isotopic measurement. The collected Nd fraction was evaporated and dissolved in 0.2 mol mL–1 HCl, and Nd was separated from the other REEs using cationic ion-exchange columns and Bis(2-ethylhexyl) phosphate (HDEHP) resins. The solution was then rinsed with 0.2 mol mL–1 HCl to remove Sm. Strontium isotope measurements were performed using a MAT 262 thermal ionization mass spectrometer. The mass fractionation corrections for the measured Sr isotope ratios were based on 88Sr/86Sr = 8.37521; reference standard NBS987 yielded 87Sr/86Sr = 0.710250 ± 10 (2σ). The Nd isotope measurements were done using a Nu Plasma II high-resolution multicollector ICP-MS. The raw data were corrected for mass fractionation based on 146Nd/144Nd = 0.7219. The JMC Nd standard yielded a 143Nd/144Nd ratio of 0.511125 ± 10 (2σ). The 87Rb/86Sr and 147Sm/144Nd ratios were calculated using the Rb, Sr, Sm, and Nd concentrations, and the initial (87Sr/86Sr)i and (143Nd/144Nd)i ratios and εNdt were calculated using concordant ages obtained earlier by sensitive high-resolution ion microprobe analysis of zircon (Table 9).

The relative abundances of Pb isotopes were quantified using a Nu Plasma II high-resolution multicollector ICP-MS using the method described by Belshaw et al. (1998). The samples were digested using a mixture of ultrapure HF and HNO3 at 800°C for 72 h, followed by purification using conventional ion-exchange chromatography (AG1X8, 200–400 resin). A portion of the sample (0.2 mL) was added and drip washed with 1 mL of 1 molL-1 HBr eluant five times and 0.5 mL of 2 molL-1 HBr eluant once. The Pb was collected by drip washing with 1 mL of 6 molL-1 HCl and 0.5 mL of 6 molL-1 HCl. The entire procedure blank was <0.1 ng. During the analytical session, repeat analyses of international standard NBS981 yielded the following values: 208Pb/206Pb = 2.1674 ± 0.0004; 207Pb/206Pb = 0.91478 ± 0.00018; 206Pb/204Pb = 16.9402 ± 0.0070; 207Pb/204Pb = 15.4966 ± 0.0030; and 208Pb/204Pb = 36.7155 ± 0.0120 (2σ).

Petrography and mineralogy

In the Zhengjialiangzi camp, extensive underground workings provide ample opportunity for studying the structure, compositional variability of REE ore veins (Fig. 5), microtextural characteristics, and interrelations of typical ore types and their host rocks, which are illustrated in Figures 6 through 8. The modal content of major constituents varies greatly, from carbonatites (~60–70 vol % calcite, ~20–25 vol % barite ± minor amphibole, aegirine-augite, mica, and fluorapatite) to high-grade barite-fluorite veins with a low proportion of calcite and wall-rock xenocrysts (Figs. 5, 6A, B). Banded ore veins (Fig. 5C, E) arise from the crudely parallel alignment of bastnäsite and barite crystals and, to some extent, also from the elongate shape of calcite grains in the endocontact zone. Coarser-grained lenses and brecciated blows (Fig. 5D, F) are massive and patchy in appearance owing to the association of bastnäsite, barite, fluorite, and calcite (± wall-rock fragments).

Barite forms large, commonly deformed crystals up to several centimeters in size (Figs. 5, 6A, B). In BSE images, we observed replacement of early, coarse-grained Sr-rich barite (approaching celestine with SrO content up to 25.4 wt %) by a cavernous aggregate of chalky or spongy fine-grained Sr-poor variety (Fig. 6C). Calcite occurs as an equigranular mosaic of variably deformed grains developed interstitially with respect to larger crystals of barite and bastnäsite, as well as filling fractures in these minerals (Fig. 6A, B) and wall-rock xenoliths (Fig. 6D, G). The mosaic-textured calcite is typically twinned and shows evidence of incipient ductile deformation (Fig. 6A). Purple fluorite is developed interstitially with respect to the other gangue minerals (Fig. 6A, B). Larger fluorite grains (>1 cm) are colorless in their interior and grade into a dark-purple variety in the rim.

Veins contain several types of xenoliths and a wide variety of accessory minerals (less than 5 vol %). The nordmarkite xenoliths comprise microcline, albite, quartz, biotite, clinopyroxene, titanite, fluorapatite, magnetite, and zircon. The clinopyroxene, ranging from diopside to aegirine-augite in composition, was partially replaced and overgrown by fluororichterite (Fig. 6D, E). In addition to the latter, metasomatic reworking (sodic fenitization) of nordmarkite fragments is expressed by the pervasive replacement of primary oscillatory zoned K-feldspar (blue CL) by microcline (yellowish-green CL) and albite (red CL, Fig. 6F) in association with calcite, Fe-rich dolomite, fluorophlogopite, and muscovite close to xenolith margins. By contrast, greenschist fragments underwent potassic fenitization, expressed by the replacement of precursor albite by K-feldspar and of biotite and chlorite by fluorophlogopite (Fig. 6G). The Yangx-in marbles are saccharoidal fine-grained rocks composed of zoned calcite and devoid of any REE minerals, barite or fluorite.

Silicate phases including fluorophlogopite, amphibole, and clinopyroxene are most common in the vicinity of nordmarkite and metabasalt xenoliths, where ore veins are enriched in xenocrystic quartz and feldspars. Fluorophlogopite occurs as platy zoned crystals enclosed in barite, fluorite, and calcite (Fig. 7A) or as metasomatic encrustations on xenoliths (Fig. 6G). Locally, fluorophlogopite is replaced by dolomite (Fig. 7B). Amphiboles form encrustations on quartz and feldspar xenocrysts in association with aegirine overgrowths (Fig. 7C) and strongly pleochroic acicular crystals up to 3 mm in length (Fig. 7D). Two distinct types of clinopyroxene were identified. Veins with nordmarkite xenoliths contain small (<100 μm) quartz xenocrysts overgrown by green acicular clinopyroxene (Fig. 7C, D). Samples hosting greenschist fragments contain stocky prismatic crystals of aegirine ranging from yellow to medium brown in color. These crystals are much larger than the green variety (up to 1.5 × 0.5 mm) and exhibit sector zoning in BSE images (Fig. 7E).

Quartz grains in ore veins are commonly xenocrysts based on their irregular morphology, intimate association with feldspars, and the presence of reaction-induced rims (e.g., Fig. 7C). However, clear euhedral crystals of quartz lacking any evidence of disequilibrium were also observed (Fig. 7B). Thorite forms stubby tetragonal prisms 5 to 20 μm in length (Fig. 7A, F), which are morphologically distinct from monoclinic huttonite (e.g., Förster et al., 2000).

Sulfides are represented predominantly by pyrite; much less common are galena and sphalerite, the former of which typically form equant crystals up to ~1 mm embedded in calcite (e.g., Fig. 6D). Molybdenite occurs as platy crystals up to 200 μm across, accompanied by fluorite, amphibole, fluorapatite, and pyrochlore-group minerals (Fig. 8A).

The principal REE host is bastnäsite-(Ce), which forms tabular crystals up to ~5 cm across and ~25 vol % in abundance (Figs. 5D-F, 6A, B). It was one of the earliest phases to crystallize and, as such, underwent fragmentation, ductile deformation, and fracturing. Bastnäsite crystals are typically fractured perpendicular to {0001}, and some appear bent, particularly in areas of deformed and dynamically recrystallized calcite (Fig. 6A). Fractures in bastnäsite are filled with calcite, barite, fluorite, or strontianite (Fig. 8B). Smaller crystals, their clusters, and their fragments are enclosed in large barite or fluorite grains (Fig. 6B).

Parisite-(Ce), monazite-(Ce), fluorapatite, and chevkinite-(Ce) rich in LREEs also occur in very small quantities. The only other REE fluorocarbonate is parisite-(Ce), which forms thin syntactic overgrowths on bastnäsite and minute (≤20 μm) crystals confined to late-stage parageneses developed in fractures in the latter mineral (Fig. 8C) and fluorite pockets. Monazite-(Ce) was observed only in veins containing greenschist xenoliths with apatite as scarce elongate crystals up to 80 μm in length included in calcite or aegirine adjacent to the xenoliths (Fig. 8D). Fluorapatite, found in several calciterich and bastnäsite-poor veins, invariably occurs in association with amphibole (Fig. 8E). Elongate fluorapatite crystals up to 0.7 mm in length occur as lenticular pods crudely aligned with the host-rock foliation (Fig. 8F). Ore veins containing nordmarkite fragments host small (<100 μm) poikilitic crystals of chevkinite-(Ce), attached to quartz xenocrysts (Fig. 7C) and minute pyrochlore grains (Fig. 8A).

The order of crystallization (Figs. 68) appears to be as follows: albite or microcline ± fluorophlogopite → bastnäsite (monazite or fluorapatite) + thorite → Na clinopyroxenes + chevkinite → Na amphiboles ± pyrochlore ± molybdenite → fluorophlogopite → sparry barite → calcite → chalky barite ± strontianite → fluorite + dolomite + parisite → pyrite (± galena) + muscovite + quartz.

Mineral chemistry

Calcite in ore veins contains elevated levels of Mg, Mn, and Fe (0.5–0.8, 0.5–0.6, and 0.3–0.9 wt % oxides, respectively), Zn, Sr, REEs, and Pb (9–15, 4,500–9,300, 1,000–3,000, and 58–116 ppm, respectively). In contrast, calcite in the marbles is relatively pure. Calcite in exocontact marbles shows intermediate values of these elements (Tables 1, 4). The chondrite-normalized REE patterns of calcite from the ore veins are shifted to higher values relative to their counterparts in the marbles, show nearly flat LREE patterns (LaN/NdN = 1–2), and lack Y anomalies (Y/Ho = 24–29). The metasedimentary calcite shows positively sloping chondrite-normalized profiles (LaN/NdN = 3–11; Table 4; Fig. 9A). The Y/Ho ratio of metasedimentary calcite is greater than ~45—i.e., exceeds the chondritic value (Table 4). Calcite from the exocontact zones in the marbles shows transitional levels of REE enrichment and Y-Ho decoupling (Y/Ho = 31–35). Barium-Sr sulfates have 43 to 50 mol % SrSO4 (celestine) in the primary barite and 4 to 27 mol % SrSO4 in the late-stage chalky variety. The chondrite-normalized patterns of barite exhibit a steep positive slope (LaN/NdN > 200) along with an inferred positive Y anomaly, although Ho was below detection limit (Fig. 9B). The Er values in barite are also mostly below the limits of detection, precluding estimation of the Y/Ho values by interpolation. Fluorite has high Sr and REE concentrations (1,800–1,900 and 300–800 ppm, respectively); its chondrite-normalized profiles are nearly flat (LaN/YbN = 3–4) and show a conspicuous positive Y anomaly (Y/Ho = 70–91; Fig. 9B).

Micas in the silicate xenoliths differ in composition from those in the ores (Table 2). Nordmarkite contains low-F, Al-Ti–rich biotite, which shows elevated levels of Mn and Ba (0.3–0.4 and 0.7–0.8 wt % oxide, respectively). Metabasalt clasts contain low-F, Al-Ti–rich phlogopite, which in metasomatized areas is replaced by fluorophlogopite (4.6–6.9 wt % F) with lower Al and Ti contents. The ores contain fluorophlogopite with variable Fe contents (Mg# = 0.63–0.86) but consistently higher F (6.1–8.5 wt %) and lower Al + Ti.

Amphiboles that replaced clinopyroxenes in nordmarkite (Figs. 6D, E, 7C) are close to fluororichterite in composition, whereas the acicular variety in carbonatites is predominantly potassic-fluoro-magnesio-arfvedsonite with occasional relict fluororichterite found in its core (Fig. 7D). Thus, the composition of amphiboles in the veins evolved by becoming enriched in F, Na, Mg, and Fe3+, but depleted in K, Ca, Al, Ti, Mn, and Fe2+ (Table 2).

Zoned clinopyroxene crystals and xenocrysts from nordmarkite xenoliths range in composition from diopside to aegirine-augite (6–65 mol % NaFeSi2O6, 8–75 mol % CaMgSi2O6, 12–47 mol % CaFeSi2O6; Table 2) and are characterized by appreciable Al and Mn (0.5–1.7 and 0.3–0.9 wt % oxide, respectively) and low Ti contents. The pale-green variety associated with fluororichterite in carbonatites (Fig. 7C) is aegirine-augite (70–74 mol % NaFeSi2O6, 7–14 mol % CaMgSi2O6, 11–18 mol % CaFeSi2O6) with low Al, Ti, and Mn abundances, whereas stocky brown crystals (Fig. 7E) are aegirine with elevated Al and V contents (1.1–3.0 and up to 0.9 wt % oxides, respectively). Thus, clinopyroxenes in the ore veins evolve toward aegirine-dominant and Al-rich compositions. Despite variation in luminescence (Fig. 6F), both K-feldspar and albite approach their ideal formulae.

Bastnäsite shows the preponderance of Ce over the rest of the REEs (Tables 3, 4) and the high content of Ba and Th (1,010–1,200 and 1,900–3,090 ppm, respectively). Chondritenormalized profiles (Fig. 9B) show a steep negative slope (LaN/YbN = 149,000–308,000; LaN/NdN = 8–11) and a negative Eu anomaly (Eu/Eu* ≈ 0.6) but no Y anomaly. The Th/U ratio is very high (190–406) and correlates positively with chondrite-normalized LaN/YbN values.

Other REE-bearing accessory minerals are LREE dominant (Table 3), and yttrium is present in detectable concentrations (a few thousand ppm) only in parisite and chevkinite (Figs. 7C, 8C). Fluorapatite (Fig. 8E, F) is enriched in LREEs, particularly in the intermediate zone showing plum CL (up to 5.7 wt % LREE2O3).

The pyrochlore (Fig. 8A) contains high levels of U (27.4–29.7 wt % UO2) and shows appreciable variation in the proportions of Na, Ca, Ti, Nb, and F (Table 3). Because some compositions have Nb > Ti whereas others have Ti > Nb, this mineral is best described as a series of compositions transitional between U-rich betafite and pyrochlore. The relatively higher Na and Ca contents in Ti-rich compositions imply that the principal substitution mechanisms involve the formation of vacancies in the (Na, Ca) site: Na+ + Ti4+ → □ + Nb5+ and Ca2+ + 2Ti4+ → □ + 2Nb5+. In thorite, the only substituent element of note is U (11.4–12.8 wt % UO2; Th/U = 5–6).

Whole-rock geochemistry

Yangxin marbles are characterized by low Mg, Mn, Fe (~0.28, 0.06, and 0.02 wt % oxides, respectively; Fig. 10A), and REE content (below ~25 ppm; Table 5); other trace elements are also low, with the exception of Pb in some samples (up to 22 ppm). Of note are very low V/Cr, Mo/U, and Th/U ratios (~0.3, 0.6, and 1.1; Table 5). Metasomatized marbles from exocontact zones of the mineralized veins exhibit intermediate levels of enrichment in most trace elements (Fig. 10A, B). One notable exception is Sr, which is decoupled from Ba, and in the exocontact zones approaches the same concentration levels as in the veins (Fig. 10C). The metasomatized rocks show detectable concentrations of Na and K and higher levels of Mg, Mn, and Fe (~0.38, 0.24, and 0.28 wt % oxide, respectively; Table 6). The exocontact marbles also differ from their precursor in REE budget, including a greater degree of fractionation between light and heavy REEs (LaN/YbN = 36–91 vs. 8–37), the absence of Ce anomaly (Ce/Ce* ≈ 0.96 and 0.72, respectively), and lower, near-chondritic Y/Ho ratios (on average, 29 and 46, respectively). Note that calcite samples from mineralized veins and marbles show the same Y/Ho variation trend as their host rocks, but REE-rich calcite from the veins does not show extreme enrichment in LREEs as shown by whole-rock analyses (cf. Tables 4, 5; Fig. 10A).

All nordmarkite samples are mildly peralkaline [molar (Na2O + K2O)/Al2O3 = 0.99–1.04] silica-oversaturated rocks (20–30% normative quartz) characterized by low MgO and CaO contents (≤0.7 and 2.8 wt %, respectively, after correction for the presence of accessory calcite; Table 6). All samples exhibit enrichment in REEs, Sr, Ba, Th, and U at low levels of Na, K, and some high field strength elements (HFSEs = Ti, P, Nb, and Ta). Their REE budget is characterized by strong enrichment in light lanthanides (LaN/YbN = 58–185). The chondrite-normalized trace element patterns of the nordmarkites are broadly similar to those of the mineralized veins but exhibit less enrichment in REEs, Sr, Ba, and Pb and lack Zr-Hf anomalies (Fig. 10D).

Stable isotopic compositions

Thirty-eight samples were selected for C-O isotope analysis (Table 7; Fig. 11A). Eight calcite samples from fresh marble have δ13 CV-PDB values ranging from 0.4 to 1.6‰ and δ18OV-SMOW values from 15.7 to 23.7‰. The marbles least affected by metasomatism show homogeneous compositions (δ13CV-PDB = 1.4 ± 0.2‰; δ18OV-SMOW = 23.4 ± 0.2‰) and plot in the range typical of carbonate rocks from the Yangxin Formation (e.g., Huang et al., 2016). Fifteen calcite samples from ore veins exhibit lower δ13CV-PDB and δ18OV-SMOW values (–6.0 to –1.1‰ and 10.0–12.8‰, respectively) that approach the compositional range of igneous carbonates (Taylor et al., 1967; Demény et al., 2004). Samples from the exocontact marble are characterized by intermediate isotopic ratios (δ13CV-PDB = –1.1 to 1.3‰ and δ18OV-SMOW = 11.9–17.3‰).

Calcite from ore veins is characterized by lower δ44/42Ca915a and δ43/42Ca915a ratios relative to the marble samples (Table 8). The data follow a mass-dependent fractionation trend, where the metacarbonates plot away from the origin (Fig. 11B). The absence of any appreciable excursion from the trend toward lighter isotopic compositions suggests nil contribution from radiogenic 40Ca (Schiller et al., 2016). For comparison, two calcite crystals from Guangtoushan at Maoniuping in the Mianning-Dechang belt (Liu et al., 2019a) were analyzed and showed less fractionation between the heavy and light Ca isotopes (i.e., slightly lower δ values; Fig. 11B). However, the C-Ca isotope signature from both deposits is consistent with igneous carbonates elsewhere (Sun et al., 2021).

Sr-Nd-Pb isotope compositions

The Sr-Nd-Pb isotope compositions of the nordmarkite, marble, and selected calcite, fluorite, and bastnäsite samples from ore veins are listed in Tables 9 and 10. Calcite contains high levels of Sr and Pb and negligible Rb, Th, and U (Table 4). Its measured Sr-Pb isotope ratios are essentially identical to the initial values calculated for ~27 Ma (Liu et al., 2015a). Fluorite contains sub-parts-per-million levels of Rb, Pb, Th, and U at several thousand parts per million Sr. Calculations show that its initial Sr and Pb isotope ratios are within the 2σ value of the measured ratios. Because both minerals contain appreciable levels of REEs, their initial Nd ratios are slightly lower than the measured values, but the difference does not exceed 0.01% (Table 9). Bastnäsite is strongly enriched in Th relative to U and Pb (Table 4). As a result, its (208Pb/204Pb)i values are slightly smaller (<0.1% relative) than the corresponding measured ratios, whereas the (206Pb/204Pb)i and (207Pb/204Pb)i values are virtually identical to the measured ratios (Table 10).

The Sr-Nd-Pb isotope compositions of calcite and fluorite from the ore veins are very similar and fall within a narrow range: (87Sr/86Sr)i = 0.70657 to 0.70673, εNd(t) = –4.4 to –4.2, (206Pb/204Pb)i = 18.29 to 18.33, (207Pb/204Pb)i = 15.630 to 15.636, and (208Pb/204Pb)i = 38.56 to 38.68. Bastnäsite has slightly less radiogenic Pb compositions: (206Pb/204Pb)i = 18.01 to 18.23, (207Pb/204Pb)i = 15.610 to 15.625, and (208Pb/204Pb)i = 36.98 to 37.93 (Tables 9, 10). It is noteworthy, however, that these ranges are small compared to the isotopic compositions of similar rocks elsewhere in China (Fig. 12). One of the two analyzed nordmarkite samples (P1L3-7-3) is isotopically indistinguishable from the ore vein material, whereas sample P3L1-1-0 has the same Pb isotope signature but is somewhat depleted in radiogenic Sr and enriched in radiogenic Nd [(87Sr/86Sr)i = 0.70579, εNd(t) = –3.7].

The marbles form two geochemically distinct groups: group 1 (fresh rocks) is characterized by higher Rb/Sr, Sm/Nd, 87Rb/86Sr, and 147Sm/144Nd ratios (0.0031 ± 0.0016, 0.21 ± 0.04, 0.41, and 0.12, respectively) relative to group 2 (meta-somatized material; 0.00036 ± 0.00024, 0.13 ± 0.01, 0.0020 ± 0.0015 and 0.08 ± 0.02, respectively). Group 1 yields higher (87Sr/86Sr)i values (0.70788–0.70789) than those of ore vein minerals (0.70657–0.70673) and is characterized by a much less radiogenic Nd and more radiogenic U-Pb signature [εNd(t) ≈ –10, (206Pb/204Pb)i = 18.60–18.61, (207Pb/204Pb)i = 15.651] than either the ore vein material or group 2. The group 2 marble is similar to the ore vein material in Nd and Pb isotope composition but yields slightly more radiogenic (87Sr/86Sr)i values (Tables 9, 10).

The origin of REE mineralization in the Zhengjialiangzi camp

The geochemical evidence indicates that fresh porphyritic granite and greenschist facies rocks in the Zhengjialiangzi camp have very low levels of REEs (289–333 and 118 ppm, respectively). Minerals containing essential REEs are not observed in these rocks, implying that these elements are distributed among accessory minerals such as zircon or apatite. Recently, it has been shown that wall-rock contamination of carbonatites could cause the formation of alkali and F-bearing silicates, and fluorite by reaction of K and Na in carbonatite with Si and Al in wall rocks, which would destabilize halogenated REE complexes and reduce REE solubility in melts and hydrothermal fluids, causing REE mineralization (Anenburg et al., 2020a; Zheng et al., 2023). However, the absence of REE mineralization in the fenitized wall rocks is inconsistent with the breakdown of alkali REE complexes. The marbles exhibit even lower concentrations of REEs (6–24 ppm) and show appreciable enrichment (1,190–2,000 ppm) only at the contact with the ore veins (Fig. 9A). The same observation applies to other nominally incompatible elements, which are depleted in fresh marbles compared to the ore veins (Fig. 10A-C). The available stable isotope data (Fig. 11) show that the exocontact marbles exhibit lower δ13CV-PDB and δ18OV-SMOW values than their precursors, trending toward the range typical of mantle-derived rocks (Fig. 11A). Regarding their radiogenic isotopic signature, these marbles (group 2) are very similar to, and in some cases indistinguishable from, calcite and fluorite in the ore veins (Fig. 12). All of this evidence points to the metasomatic overprinting of the Yangxin marbles, which entailed very limited fluid-assisted influx of REEs and rules out their role as a source of the REEs deposited within the vein system.

In other parts of the Mianning-Dechang belt, similar mineralized veins containing bastnäsite, barite, fluorite, and calcite, associated with carbonatite-nordmarkite complexes, commonly show REE concentrations up to several weight percent (Hou et al., 2006, 2015; Liu and Hou, 2017). The structural, petrographic, and geochemical evidence presented here strongly suggests a genetic link between alkalinesilicate and carbonatitic magmatism also at Muluozhai. In the Zhengjialiangzi camp, nordmarkite exclusively appears as vein-hosted xenoliths, undergoing partial metasomatism resulting the formation of albite, fluororichterite, aegirine, and minor fluorophlogopite (Fig. 6). This metasomatic overprint is characterized as sodic fenitization and indicates disequilibrium between the two rock types (Le Bas, 2008; Elliott et al., 2018). In other areas of the Mianning-Dechang belt, similar processes (albeit, on a larger scale) have been documented at Maoniuping, where fenitization was found to be contemporaneous with the emplacement of carbonatites (Liu and Hou, 2017; Liu et al., 2019a). The physical parameters of fenitization were constrained to temperatures of ~480°C and moderate salinity (10–18 wt % NaCl equiv; Zheng and Liu, 2019).

Further similarities include the spatial association of carbonatite and ore veins with nordmarkite, which is enriched in REEs (612–1,823 ppm), Sr (1,734–2,622 ppm), and Ba (3,922–9,280 ppm) but remarkably poor in HFSEs (16–38 ppm Nb, <1 ppm Ta). The presence of abundant bästnasite, fluorite, and Sr-rich barite in the Zhengjialiangzi camp ore strengthens the mineralogical parallels with the Maoniuping, Dalucao, Lizhuang (Liu and Hou, 2017), Weishan (Jia and Liu, 2019), and Bayan Obo deposits (Yang et al., 2019). The presence of voluminous nordmarkite bodies at Muluozhai (Fig. 1) and the resemblances to other ore systems within the Mianning-Dechang belt suggest a substantial uncharted potential for REE exploration. Calcite stable isotope compositions in the Zhengjialiangzi camp veins indicate a mantle source, whereas the observed variations in δ13CV-PDB and δ18OV-SMOW (Fig. 11A) signify isotopic reequilibration during or after emplacement.

Carbon and oxygen isotope values in fresh marble conform to the typical range found in the Yangxin group marbles (Huang et al., 2016), but drop to about –1 and 12‰, respectively, in recrystallized calcite adjacent to the marble-vein contacts (Table 7). A specific set of metasomatized marble samples displays a δ18OV-SMOW signature closely resembling that of the gangue calcite but higher δ13CV-PDB ratios, a phenomenon difficult to explain through isotopic reequilibration alone. Remarkably, the gangue calcite samples analyzed for Ca and C isotopes exhibit a greater δ13CV-PDB variation that extends toward exocontact marble values (–6.8 to –1.1‰). Thus, the apparently clustered data distribution on the δ13CV-PDB vs. δ18OV-SMOW plot may be an artifact arising from limited sampling. The observed isotopic variations can alternatively be modeled by marble decarbonation at high temperature. However, the Rayleigh decarbonation trend (no. 5 in Fig. 11A) does not explain the low-δ18OV-SMOW marbles and requires initial δ18OV-SMOW values that are too low in comparison with the actual fresh marble compositions. A similar departure of exocontact carbonate rocks from a decarbonation model to much lower δ18OV-SMOW values was reported in the Alta aureole (Utah) by Bowman et al. (1994). These authors interpreted it to result from the infiltration of 18O-poor fluids into the wall-rock marble. It is hence conceivable that the stable isotope variations in the marble-vein system at Muluozhai arose from a combination of processes (most likely, fluid infiltration into the wall-rock marble, fractionation of calcite and bastnäsite, and CO2 loss from the marble) rather than a single mechanism. Although the lowest isotopic values are in agreement with a mantle (i.e., carbonatitic) source (indicated by the PIC area in Fig. 11), it is likely that there was some contribution of C, O, and Ca to the ore system from the surrounding metasedimentary package.

The Sr-Nd-Pb isotope compositions of nordmarkite, marble, bastnäsite, and gangue minerals from the ore veins were used to constrain the source(s) of REE mineralization. Data in Tables 5 and 6 reveal significantly higher concentrations of Sr, REEs, and Pb in the ore veins and nordmarkites than in any of the wall rocks. Therefore, the measured radiogenic isotope characteristics closely represent the source and are unlikely to have been modified to any significant extent by wall-rock assimilation or hydrothermal overprinting. Initial isotopic values of ore vein material show little variation (Tables 9, 10) and yield a tight data cluster on conventional diagrams (Fig. 12). This cluster overlaps with one of the measured nordmarkite compositions in Sr, Nd, and Pb values, whereas the second nordmarkite sample shows a slightly less radiogenic Sr signature (Fig. 12A). Notably, however, all measured compositions follow the same trend as nordmarkites and carbonatites elsewhere in the Mianning-Dechang belt. These similarities imply that the Muluozhai intrusions originated from essentially the same source as those responsible for the formation of the giant Maoniuping and other smaller deposits within the same tectonic setting. Note that typical anorogenic alkaline complexes (e.g., those in the East African rift), whose origin is typically interpreted in the context of lithosphere-plume interaction (e.g., Bell and Tilton, 2001), differ appreciably from Maoniuping, Muluozhai, and other similar intrusions. The consistent isotopic signature found in the alkaline and carbonatitic rocks in the Mianning-Dechang belt, across the Tibetan Plateau, and in other carbonatite-related REE deposits such as Weishan (Wei et al., 2022) and Bayan Obo (Yang et al., 2019) is attributed to their origin from the subcontinental lithospheric mantle, which has been fertilized by CO2-rich fluids released from subducted marine sediments (Guo et al., 2005; Hou et al., 2015, 2023; Tian et al., 2015).

It remains debatable whether the carbonatites and spatially associated nordmarkites derive from the same parental magma or represent discrete magmas originating from the same mantle source (Fig. 13A). It is clear that the Muluozhai nordmarkite and vein material most closely approaching bona fide carbonatite (i.e., relatively enriched in silica, Mg, and alkalis) cannot represent conjugate immiscible liquids. First, the nordmarkite compositions are too high in silica and too low in alkalis to approach silicate melts on the low-pressure two-liquid solvus (Lee and Wyllie, 1998; Brooker and Kjarsgaard, 2011; Martin et al., 2013). Second, specific trace elements consistently partition into one of the immiscible liquids regardless of variations in alkalinity or water content. For instance, Sr and Ba favor carbonate melts, whereas Ga remains in the silicate phase (Martin et al., 2013). However, our data do not support this trend. Third, the available textural evidence indicates that the nordmarkite was fenitized at the contact with carbonatites (e.g., Fig. 6D, F), which effectively rules out derivation from the same parental magma by immiscible separation (Treiman and Essene, 1985). Thus, whether carbonatites and nordmarkites represent immiscible fractions, as suggested by their close spatial and temporal association and similar Sr-Nd-Pb isotope and trace element characteristics in the Mianning-Dechang belt (Hou et al., 2006), Weishan (Zeng et al., 2022), or Miaoya (Zhang et al., 2019) remains to be ascertained based on new evidence.

It is also unlikely that the examined calcite-bearing vein systems could derive from a hypothetical carbonate-rich trachytic or trachyandesitic melt by crystal fractionation of rock-forming silicates. The partitioning of trace elements between silicate melts, feldspars, and clinopyroxenes is well established through experimental data (Toplis and Corgne, 2002). Barium exhibits strong compatibility with K-feldspar (White et al., 2003), implying that magmas modified by voluminous feldspar fractionation will show relative depletion in Ba. Contrary to this expectation, the carbonatite and associated veins have similar or, in many cases, much higher Ba levels relative to either nordmarkite or alkali granite. Nickel, Co, Sc, and Cr are compatible in calcic clinopyroxenes, whereas the compatibility of V depends on redox conditions but is at least several times lower than that of Sc; Ni is more compatible than either Co or Sc (Norman et al., 2005; Fedele et al., 2009; Mollo et al., 2013). Therefore, clinopyroxene fractionation will lead to melts depleted in Cr, Co, and Ni with progressively lower Ni/Co and Sc/V but higher V/Cr ratios. However, we observed no such trend at Muluozhai: the carbonatite and other veins are enriched in Cr, Co, and Ni by at least one order of magnitude along with a lower V/Cr but higher Ni/Co ratio than the nordmarkite or alkali granite, whereas the Sc/V value is virtually the same in all three rock types (cf. Tables 5, 6). Hence, the presently available geochemical evidence is most consistent with the derivation of alkaline felsic rocks and carbonatites from discrete magma types tapping the same subcontinental mantle source. The petrogenetic relationships between the nordmarkite and spatially associated porphyritic granite (Fig. 1) are uncertain. However, the two rocks are sufficiently similar in trace element geochemistry (Fig. 10; Table 6) to imply a common lineage.

The higher SiO2 and lower Ba contents of the granite suggest that it could represent a more evolved product of trachytic magma fractionation (e.g., Nekvasil et al., 2000). A detailed study of mineral chemistry in both rock types (e.g., K-feldspar evolution) and isotopic analysis of the granite are required to explore this possibility. The Sr-Nd-Pb radiogenic isotope signature of exocontact marbles from the Zhengjialiangzi camp (group 2) is similar to that of the ore vein material. This is in line with the trace element evidence (Fig. 10) and supports the idea that the marbles underwent local metasomatic overprinting by Sr-REE-Pb–bearing fluids (Fig. 13A). Fresh marbles (group 1), however, are distinctly different in having more radiogenic Sr and U-Pb signatures but containing less radiogenic Nd (Fig. 12). Moreover, these rocks exhibit superchondritic Y/Ho and subchondritic Th/U ratios, negative Ce anomalies, and very low V/Cr and Mo/U ratios (Table 5). Their geochemistry is consistent with deposition in an oxidized, shallow marine environment (Guo et al., 2013; Wu and Zhang, 2019).

Evolution of the Zhengjialiangzi vein system

Deciphering the evolution of the Zhengjialiangzi vein system is not a trivial task, because it is developed in marble and calcite is the principal constituent of both the veins and their wall rocks. The isotopic evidence indicates that Ca, C, Sr, and REEs in the Zhengjialiangzi veins were initially derived from a carbonatitic source of mantle provenance, akin to those of the other Mianning-Dechang deposits (Figs. 11, 12). As expected, the Ca-C-O stable isotope signature of the veins was affected by their interaction with predominantly calcitic wall rocks, but the extent of this interaction is variable among the studied samples: cf. δ44/42Ca915a = 0.36 to 0.41‰ and δ13CV-PDB = –6.8 to –4.5‰ in the most primitive calcite samples vs. 0.44 to 0.51‰ and –4.2 to –1.1‰, respectively, in the contaminated vein material (Table 8). The whole-rock and LA-ICP-MS data (Tables 4, 5) provide further evidence for carbonatite-wall-rock interaction and, in particular, for the partial loss of Sr, REEs, and some other trace elements into the metasomatized exocontact zone. Notably, one element that does not show any significant enrichment in the metasomatized marbles is Ba, particularly in comparison with its kindred alkali earth element Sr. The Sr/Ba and Pb/Ba ratios in the metasomatized rocks increase dramatically relative to their precursor marbles (26 ± 19 and 0.7 ± 0.5 vs. 0.9 ± 0.7 and 0.09 ± 0.16, respectively) and beyond the range measured in the ore veins (Sr/Ba = 0.14 ± 0.08, Pb/Ba = 0.22 ± 0.19). Similar variations are observed in calcite from these three rock types. The lower mobility of Ba in comparison with Pb and, especially, Sr in the marbles is in agreement with experimental data that demonstrate much lower solubility of Ba2+ relative to Pb2+ and Sr2+ in sulfate-bearing fluids (Dove and Czank, 1995; Rollog et al., 2019). Of particular interest is also a sharp (one to two orders of magnitude) rise in Mo and Th in the exocontact marbles, which is not matched by their enrichment in U (Table 5). The relative immobility of U implies reducing conditions and low activity of sulfate and phosphate anions in the metasomatizing fluid (Krupka and Serne, 2002; Cumberland et al., 2016). This interpretation is consistent with the absence of sulfate and phosphate minerals in the metasomatized rocks and appreciable across-contact mobility of Mn and Ce: Mn/Mg and Ce/Ce* ratios increase from ~0.3 to 0.8 and from 0.7 to 1.0, respectively, from the protolith toward the contact.

Other elements that were locally affected by carbonatite-wall-rock interactions include Si, alkalis, Mg, and F. Similar element distribution patterns and associated REE mobilization during carbonatite-wall-rock interaction have been increasingly documented in carbonatites from the Mianning-Dechang belt (Zheng et al., 2023) and elsewhere, including Palabora, South Africa (Giebel et al., 2019b), Kieshöhe, Namibia (Walter et al., 2022), and Jacupiranga, Brazil (Chmyz et al., 2022). Of note, local enrichment of mineralized veins in silica is likely linked to the disaggregation and digestion of nordmarkite and metabasalt xenoliths. These processes are manifested by the presence of albite and quartz xenocrysts overgrown by aegirine and alkali amphiboles (Figs. 7C, 8A). As our petrographic data show, the REE silicate chevkinite and REE-rich fluorapatite are restricted to aegirine-amphibole–rich vein parageneses (Figs. 7C, 8F). Hence, enhanced silica activity in these xenolith-contaminated zones also triggered the crystallization of chevkinite and facilitated the incorporation of REEs in fluorapatite, according to the substitution Ca2+ + P5+ → REE3+ + Si4+. Although the diluting effect of assimilated silica on REE mineralization is in-significant in the Zhengjialiangzi veins because most of them are bordered by carbonate wall rocks, this effect will become significant in carbonatites intruding silicate rocks (e.g., Chakhmouradian et al., 2017). Alkalis and F were partially lost to fenitization of the silicate wall rocks and xenoliths. In nordmarkites, this involved the replacement of K-feldspar by albite and of clinopyroxene by fluororichterite; fenitization of metabasalts was potassic in nature and produced microcline and fluorophlogopite (Fig. 6). Our data demonstrate that sodic and potassic fenites do not always represent two discrete, temporally spaced phases of metasomatism (cf. Le Bas, 2008) and may arise instead from the reaction of the same carbonatitic magma with chemically distinct protoliths (e.g., K-rich syenitic rocks vs. K-poor albite-bearing metabasalts). Note also that, contrary to Le Bas’s (2008, p. 927) comment about the “near-absence of H2O (and F) in the K-feldspar fenites,” potassic fenitization of metabasalts at Muluozhai clearly involved an F-rich fluid, as indicated by the abundance of micas with 1.5 to 6.9 wt % F in the metasomatic assemblage (Figs. 6G, 8D). These observations may have significant implications for the development of fenitic aureoles at other carbonatite localities (Elliott et al., 2018) or the recently revived antiskarn model of carbonatite-wall-rock interaction (Anenburg et al., 2020a, b; Bouabdellah et al., 2022). We expect this diversity of metasomatic changes to reflect variations in the chemical characteristics of wall rocks and carbonatite-derived fluids.

Based on the current understanding of regional tectonics, we suggest that mantle-derived magmas were focused into lithospheric zones of weakness (Fig. 13A) localized along the system of sinistral strike-slip faults arranged sublongitudinally or splaying off the Anninghe fault. According to geochronological evidence, the northern part of the Mianning-Dechang belt, where the Muluozhai deposit is located, was tectonically active within a narrow time interval in the Late Oligocene (27–25 Ma). In the study area, faulted supracrustal rocks of the Paleozoic Yangxin and Mount Emei Formations were intruded by distinct rock assemblages that follow the local tectonics but do not show any obvious spatial correlation with one another (Fig. 1). This and the absence of any rocks or textures that could link the alkaline silicate rock types to carbonatites (e.g., syenites with interstitial calcite, or carbonate globules in a silicate mesostasis) further support the idea that their parental magmas do not share the same origin. The distribution of mineralized material in the Muluozhai deposit is distinct from that of those observed at the Maoniuping, Lizhuang, and Dalucao deposits, representing stockworks in fenitized nordmarkite, disseminated ores, and breccia pipes, respectively (Fig. 13B).

The available structural, petrographic, and geochemical data do not support a simple petrogenetic model involving the intrusion of carbonatitic magma into fractured marbles and its closed-system evolution. Numerous veins are associated with brecciated zones, which are particularly conspicuous within silicate wall rocks (Fig. 5A). Although injection of carbonatitic magma into dilated fractures was probably accompanied by intrusive brecciation (as indicated by fragmentation of such early-crystallized phases as bastnäsite and sparry barite; Fig. 6A, B), there are several lines of evidence for the involvement of hydrothermal fluids in the evolution of the Zhengjialiangzi vein system. First, metasomatic changes in the wall rocks are ubiquitous and involve the transport of trace elements into the marbles (Fig. 10) and of alkalis, Mg, and F into the nordmarkites and metabasalts (see also Fig. 6E, G):

4(Ca0.9Na0.1)(Mg0.7Fe2+0.2Fe3+0.1)Si2O6diopside (nordmarkite)+1.2Na++0.3K++0.3Mg2++0.7Fe3++1.5F+2H+=(Na1.6Ca0.5K0.3)(Mg3.1Fe3+1.1Fe2+0.8)fluororichterite (sodic fenite)Si8O22(F1.5OH0.5)+3.1Ca2++1.5OH
(2)

and

(Mg3Fe2+2Al)AlSi3O10chlorite(metabasalt)(OH)8+2K++Mg2++3F+3H4SiO40=2K(Mg2Fe2+)AlSi3O10(F1.5OH0.5)+fluorophlogopite (potassicfenite)9H2O+H+.
(3)

Note that the above reactions reflect the actual mineral compositions (Table 2) but ignore minor substitutions of Al and Ti (≤0.1 apfu) for Mg + Fe in some of the mafic silicates.

Our petrographic observations show that replacement textures are ubiquitous also within the ore veins. One of the most common and petrogenetically informative hydrothermal reactions identified in the Zhengjialiangzi samples is the replacement of the early, Sr-rich barite by the late-stage variety depleted in Sr (barite 1 and 2 in Fig. 6C). The net loss of material from the precursor resulted in the cavernous texture and overall chalky appearance of barite 2. Strontium released during this reaction was deposited as strontianite. On the basis of published experimental data, this texture can be interpreted to have formed by the reaction of primary barite-celestine compositions with a high-ionic-strength fluid at T < 370°C (Brower and Renault, 1971):

1.6(Ba0.5Sr0.5)SO4barite1+0.6CO2 + 0.6H2O=(Ba0.8Sr0.2)SO4barite2+0.6SrCo3+0.6(SO4)2strontianite+1.2H+.
(4)

Notably, sulfate minerals were not precipitated outside these partial pseudomorphs, while sulfide minerals (in particular, pyrite and galena) are very common. This implies that the (SO4)2– released by reaction (4) was reduced to HS at pH redox conditions where Sr-poor barite and sulfides could coexist. Reduction processes of this type are well known in biomediated diagenetic settings (e.g., Magnall et al., 2016) but can also occur at T > 200°C and low pH, particularly in the presence of graphite or short-chain hydrocarbons (Hanor, 2000). We did not observe any graphite in the fresh Yangxin marbles, but both CH4 and C2H6 were identified in CO2-rich fluid inclusions in barite, fluorite, and bästnasite by Zheng et al. (2021). These inclusions are interpreted to correspond to the final, low-salinity hydrothermal stage at T ≈ 200°C. Thus, we assign the replacement of Sr-rich barite and subsequent precipitation of sulfides to the same evolutionary stage, marked by CO2 buildup and hydrocarbon synthesis via the Fischer-Tropsch route (see further discussion in Zheng et al., 2021). Nonetheless, in contrast to their assertion that these processes “initiated extensive REE deposition at ~200°C” (p. 13), the accessible petrographic evidence indicates only limited REE deposition at the final hydrothermal stage, principally in the form of parisite overgrowths on bastnäsite (Fig. 8C). This contradicts the notion that REE mineralization in the Mianning-Dechang belt mainly owes to hydrothermal processes (Shu and Liu, 2019; Zheng and Liu, 2019; Zheng et al., 2021). Although sulfate-rich hydrothermal fluids can transport significant quantities of REEs (Wan et al., 2021, 2023), the early crystallization of voluminous Sr-rich barite at Muluozhai is more consistent with magmatic deposition preceding fluid release (manifested by fracturing and hydrothermal phases like chalky barite, muscovite, quartz and strontianite).

The present study revealed the mineralogical complexity of the Zhengjialiangzi vein system. Several of the identified minerals have not been previously reported in the Mianning-Dechang deposits. A comprehensive investigation of these minerals is beyond the scope of this study, but it would provide valuable insights into the evolution of carbonatitic igneous-hydrothermal systems. For example, thorite and molybdenite account for the bulk of the Th and Mo budgets (respectively) of the examined rocks. These elements are present in concentrations of up to 600 and 6,400 ppm, respectively, making them some of the highest values reported in carbonatites and related mineralized rocks. For example, calcite carbonatites from Bayan Obo were reported to contain up to 640 ppm Th (Le Bas et al., 1992), although 90% of published whole-rock analyses indicate concentrations below 150 ppm (authors’ unpublished data). Another example is Nolans Bore carbonatite in Australia, which exhibits exceptionally high thorium levels due to both thorite and thorianite (Anenburg et al., 2020b). Extreme enrichment in Mo (>100 ppm) is even less common and has so far been documented only at Dashigou in the Huanglongpu Mo deposit in the Qinling belt, central China (240–1,010 ppm; Xu et al., 2010). A detailed study of accessory minerals like thorite and molybdenite is clearly required to understand the highly inconsistent behavior of Th, Mo, U, and some other incompatible trace elements during carbonatite evolution. Moreover, even such rock-forming phases as members of the barite-celestine series warrant further work. These minerals are common in carbonatites and related REE deposits and occur in various paragenetic associations (Castor, 2008; Bolonin and Nikiforov, 2014; Moore et al., 2015; Broom-Fendley et al., 2017; this work).

Our data support the idea that carbonatite complexes hosting REE deposits originate from mantle sources influenced by crustal recycling and involving significant REE endowment (Hou et al., 2015, 2023) and recorded in syn- to postcollisional alkaline-carbonatite complexes worldwide (Hou et al., 2009, 2015; Goodenough et al., 2021; Beard et al., 2022). This sets them apart from typical anorogenic carbonatite-alkaline complexes (e.g., those in the East African rift), which arise from lithosphere-plume interactions (e.g., Bell and Tilton, 2001).

A comprehensive investigation of Ca-C-O-Sr-Nd-Pb isotopes variations provides constraints on the sources of carbonatite magmas and whether they have experienced metasomatic interaction with conduit and wall rocks during their evolution. Such studies are essential for metacarbonate-hosted intrusions, such as Muluozhai, Amba Dongar (Palmer and Williams-Jones, 1996; Williams-Jones and Palmer, 2002), and Wicheeda in Canada (Trofanenko et al., 2016); both latter occurrences host carbonatite-derived REE ore veins within metacarbonate wall rocks and share similarities with Muluozhai. The relatively less studied isotopic systems (Pb and Ca), in particular, prove very useful for deciphering the role of hydrothermal fluids in REE transport and deposition. The available experimental studies (Migdisov et al., 2009, 2016; Migdisov and Williams-Jones, 2014) should not be applied dogmatically and must always be placed in the actual geologic context, as has been recently done for a variety of REE deposits in East Asia, Malawi, and Canada (Trofanenko et al., 2016; Li and Zhou, 2018; Broom-Fendley et al., 2021; Zheng et al., 2021; Singh et al., 2022; Li et al., 2023).

Our study leads to the following main conclusions regarding REE mineralization in the carbonatite-nordmarkite complex at the Muluozhai deposit in the Mianning-Dechang REE belt.

  1. Fluorite, barite transitional to celestite, and calcite are the major gangue minerals, along with bastnäsite as the principal REE host in the orebodies studied here. Additionally, minor fluorapatite, chevkinite, and monazite enriched in LREEs occur in areas contaminated by silicate wall rocks. Also, several other accessory minerals, such as thorite, strontianite, molybdenite, and pyrochlore, have been recognized in the Muluozhai deposit for the first time.

  2. The combined petrographic and geochemical evidence suggests the derivation of bastnäsite mineralization from a carbonatitic source, characterized by extensive interaction between carbonatite-derived fluids, wall rock, xenoliths, and early-crystallizing mineral phases. The bulk of REE mineralization formed during this magmatic stage.

  3. The C-O-Ca stable isotope and trace element compositions of calcite, along with whole-rock data, indicate a mantle source for the carbonate material and its local reequilibration with the Yangxin marbles. The Sr-Nd-Pb isotope compositions of vein material show no appreciable contamination by wall rocks and indicate a mantle source affected by crustal recycling, in common with other carbonatite-nordmarkite-hosted deposits in the Mianning-Dechang belt.

We are grateful to Sam Broom-Fendley, one anonymous reviewer, and editors Lawrence D. Meinert and Paul G. Spry for their reviews, constructive comments, and suggestions that improved this manuscript significantly. This research was funded by the National Natural Science Foundation of China (92162216), the Second Tibetan Plateau Scientific Expedition and Research Project (2021QZKK0304), Sichuan Science and Technology Program (2023NSFSC0272), and China Geological Survey, Ministry of Natural Resources (DD 20221649). ARC acknowledges support from the Natural Sciences and Engineering Research Council of Canada, University of Manitoba, and Canada-China Scholars’ Exchange Program.

Yan Liu received his B.Sc. degree from the China University of Geosciences (Beijing) in 2002. He obtained a Ph.D. degree in 2010, also from China University of Geosciences (Beijing). Then he joined the Institute of Geology, Chinese Academy of Geological Sciences, as a postdoctoral researcher under the supervision of academician Hou Zengqian. From 2010 until now, his study has focused on genesis of carbonatites and related REE-Nb deposits, including natural case studies and high-temperature and high-pressure simulation experiments.

Gold Open Access: This paper is published under the terms of the CC-BY-NC license.