Multiple growth of zirconolite in marble (Mogok metamorphic belt, Myanmar): evidence for episodes of fluid metasomatism and Zr–Ti–U mineralization in metacarbonate systems

Fluid infiltration into (meta-)carbonate rocks is an important petrologic process that induces metamorphic decarbonation and potential mineralization of metals or nonmetals. The determination of the infiltration time and the compositional features of reactive fluids is essential to understand the mechanism and process of fluid–rock interactions. Zirconolite (ideal formula: CaZrTi₂O₇) is an important U-bearing accessory mineral that can develop in metasomatized metacarbonate rocks. In this study, we investigate the occurrence, texture, composition, and chronology of various types of zirconolite from fluid-infiltrated reaction zones in dolomite marbles from the Mogok metamorphic belt, Myanmar. Three types of zirconolite are recognized: (1) the first type (Zrl-I) coexists with metasomatic silicate and oxide minerals (forsterite, spinel, phlogopite) and has a homogeneous composition with high contents of UO₂ (21.37 wt %–22.82 wt %) and ThO₂ (0.84 wt %–1.99 wt %). (2) The second type (Zrl-II) has textural characteristics similar to those of Zrl-I. However, Zrl-II shows a core–rim zonation with a slightly higher UO₂ content in the rims (average of 23.5 ± 0.4 wt % (n=8)) than the cores (average of 22.1 ± 0.3 wt % (n=8)). (3) The third type (Zrl-III) typically occurs as coronas around baddeleyite and coexists with polycrystalline quartz. Zrl-III has obviously lower contents of UO₂ (0.88 wt %–5.3 wt %) than those of Zrl-I and Zrl-II. All types of zirconolite have relatively low rare earth element (REE) contents (< 480 µg g⁻¹ for ΣREE). Microtextures and compositions of the three zirconolite types, in combination with in situ zirconolite U–Pb dating using laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS), reveal episodic fluid infiltration and element mobilization in the dolomite marbles. The first-stage infiltration occurred at ∼ 35 Ma, leading to the formation of Mg-rich silicates and oxides and accessory minerals (Zrl-I, baddeleyite, and geikielite). The reactive fluid was characterized by high contents of Zr, Ti, U, and Th. After that, some Zrl-I grains underwent a local fluid-assisted dissolution–precipitation process, which produced a core–rim zonation (i.e., the Zrl-II type). The final stage of fluid infiltration, recorded by the growth of Zrl-III after baddeleyite, took place at ∼ 19 Ma. The infiltrating fluid of this stage had relatively lower U contents and higher SiO₂ activities than the first-stage infiltrating fluid.

This study illustrates that zirconolite is a powerful mineral that can record repeated episodes (ranging from 35 to 19 Ma) of fluid influx, metasomatic reactions, and Zr–Ti–U mineralization in (meta-)carbonates. This mineral not only provides key information about the timing of fluid flow but also documents the chemical variation in reactive fluids. Thus, zirconolite is expected to play a more important role in characterizing the fluid–carbonate interaction, orogenic CO₂ release, and the transfer and deposition of rare metals.

1 Introduction

Understanding the infiltration of fluids and melts into (meta-)carbonate rocks is of particular importance because these processes can cause significant CO₂ release through fluid/melt-rock interactions during orogenesis (e.g., Evans, 2011; Ferry, 2016; Carter and Dasgupta, 2018; Stewart et al., 2019; Guo et al., 2022) and generate many skarn-type deposits of critical metals (e.g., Kerrick, 1977; Meinert, 1992; Meinert et al., 2005; Deng and Wang, 2016; Xie et al., 2021) and colored gemstones (e.g., Themelis, 2008; Searle et al., 2020; Guo et al., 2021). Precisely dating the fluid infiltration events and constraining the compositions of reactive fluids are critical for understanding the duration and episodes of fluid flow, fluid–carbonate interaction progress, sources of fluids, and ore-forming processes (Yuan et al., 2008; Deng et al., 2014; Zhang et al., 2022). The influx of reactive fluids and melts in magmatic–hydrothermal–metasomatic systems might be episodic and was likely to occur at various stages and under various pressure–temperature conditions (e.g., Barker et al., 2006; Brice et al., 2019; Guo et al., 2021). Clarifying the fluid infiltration history and characterizing the reactive fluid behavior in the metasomatic metacarbonate system are thus highly challenging.

The occurrence of Zr-bearing or Ti-bearing phases provides a good opportunity to uncover the complex infiltration history in carbonate rocks. Zirconolite (ideal formula: CaZrTi₂O₇) occurs as an accessory mineral in a variety of lithologies, not only in metasomatized carbonate rocks (Gieré, 1986; Gieré, 1990; Gieré and Williams, 1992; Zaccarini et al., 2004) but also in many SiO₂-poor magmatic rocks, such as carbonatites, kimberlites, alkaline rocks, and lunar basalts (Williams and Gieré, 1996; Giére et al., 1998). In metasomatized marbles, zirconolite develops in various reaction zones and typically shows complex chemical zonation (e.g., Gieré and Williams, 1992). This mineral coexists with other metasomatic minerals (e.g., baddeleyite, geikielite, thorianite, uraninite, titanite, and spinel) and grew by the crystallization from the reactive fluids or the replacement of other Zr–Ti phases (Gieré and Williams, 1992; Zaccarini et al., 2004; Proyer et al., 2014). Importantly, zirconolite typically has high U and Th contents (up to 24 wt % and 22 wt %, respectively), low common Pb contents, and a high U–Pb isotopic system closure temperature of ∼ 800–900 ^∘C (Rasmussen and Fletcher, 2004; Wu et al., 2010), which make it an ideal geochronometer of fluid flow and metasomatism.

In addition to Th and U, several groups of trace elements can be incorporated into the structure of zirconolite. For example, the Ca site can be substituted by rare earth elements (REEs); the Ti site can be substituted by some high-field-strength elements (Nb and Ta), Fe, Cr, Al, Zn, and Mg; and the Zr site can be substituted by Hf (Gieré et al., 1998). The enrichment or depletion of different elements in zirconolite is closely associated with the lithology of the host rock, formation setting, and growth stage (Gieré et al., 1998), which allows the compositional signatures of the reactive fluid/melt to be constrained (e.g., Gieré and Williams, 1992; Zaccarini et al., 2004). Therefore, the investigation of zirconolite potentially provides key information on the time-resolved fluid infiltration history and compositional variation of reactive fluids in (meta-)carbonate rocks.

In this study, we investigate the occurrence, texture, mineral compositions, and chronology of different types of zirconolite in fluid-infiltrated marbles from the Mogok metamorphic belt (MMB, central Myanmar). Our results reveal multistage growth of zirconolite, which was caused by episodic fluid infiltration and Zr–Ti–U mineralization events at ages ranging from ∼ 35 to ∼ 19 Ma. Moreover, we compare and summarize the compositional signatures of zirconolites from various lithologies.

2 Geological setting and sample petrography

The MMB is located in central Myanmar (Fig. 1a). This narrow sickle-shaped belt extends approximately 1500 km from the Andaman Sea north to the eastern Himalayan syntaxis and is distributed along the northwestern margin of the Shan Plateau and southward between the Shan scarp and Sagaing fault (e.g., Barley et al., 2003; Gardiner et al., 2015; Mitchell et al., 2007; Searle et al., 2007, 2017). The formation and evolution of the MMB are closely related to the closure of the Neo-Tethys ocean, India–Asia collision, and post-collision extension and uplift (Searle et al., 2017). The MMB is mainly composed of various types of amphibolite-to-granulite-facies metasedimentary and metaigneous rocks (e.g., Mitchell et al., 2012; Searle et al., 2017) and considered to represent the exhumed lower and middle crustal metamorphic rocks of the Sibumasu (Asia) plate. This belt produces some of the world's best examples of colored gemstones, such as spinel, ruby, and sapphire (e.g., Searle et al., 2007, 2017; Guo et al., 2021; Zhang et al., 2021).

Figure 1 (a) Geological map (modified from Themelis, 2008) showing the distribution of the Mogok metamorphic belt (MMB). (b) A simplified geological map (modified from Guo et al., 2021) of the Nayaung Oke area showing the lithologic distribution and sample locality.

The main lithologies of the MMB include various types of gneisses, migmatites, marbles, schists, and calc–silicate rocks (Mitchell et al., 2007; Searle et al., 2017). In addition, syenites, granitoid rocks, dikes, leucosomes, pegmatites, and hydrothermal veins were widely observed in the MMB (Mitchell et al., 2007; Gardiner et al., 2015; Searle et al., 2017, 2020). In contacts between the magmatic rocks and marbles, metasomatic rocks develop by the influx of fluids or melts, leading to the formation of gemstone-bearing reaction zones (e.g., Themelis, 2008; Guo et al., 2016, 2021; Searle et al., 2020). The peak metamorphic pressures and temperatures of the MMB rocks range from 5 to 12 kbar and from 625 to > 950 ^∘C (Searle et al., 2007; Yonemura et al., 2013; Win et al., 2016; Thu and Enami, 2018; Chen et al., 2021; Lamont et al., 2021). Ultrahigh-temperature metamorphism (> 900 ^∘C) has been documented in the Thabeikkiyin and Bernardmyo areas (Chen et al., 2021; Lamont et al., 2021).

The metamorphic ages of the MMB range from ∼ 60 to 20 Ma (Searle et al., 2017; Lamont et al., 2021). Many studies revealed multiple metamorphic ages for single MMB rocks, indicating a multistage metamorphic evolution of these rocks (e.g., Win et al., 2016; Lamont et al., 2021). Ultrahigh-temperature metamorphism was determined to have occurred at 43–32 Ma (Lamont et al., 2021) and ∼ 25 Ma (Chen et al., 2021). New U–Pb dating for zircon or titanite indicates that metamorphism related to ruby formation occurred at 22–25 Ma (Zhang et al., 2021; Phyo et al., 2020). Intrusions of diorites, granite charnockites, and syenites in the MMB have large variations in age ranging from ∼ 170 to 13 Ma (e.g., Mitchell et al., 2012; Searle et al., 2020; Shi et al., 2021). The age of metasomatism (contact metamorphism) varies with the time of infiltrating fluids or melts from these intrusions (Searle et al., 2020; Guo et al., 2021).

The study region is located in the Nayaung Oke area (Fig. 1). This area comprises dolomite marbles, calc–silicate rocks, gneisses, and schists. In addition, small amounts of low-grade metamorphic rocks, sedimentary rocks, and intrusions are present. The Sedawgyi gneisses have a Cambrian U–Pb protolith age of 491 ± 4 Ma, and a biotite granite dike in these gneisses shows a zircon U–Pb age of 17.0 ± 0.3 Ma (Mitchell et al., 2012). Syntectonic deformed hornblende syenites have emplacement ages of 33.11 ± 0.93 to 30.90 ± 0.64 Ma (Barley et al., 2003). In the marbles, polycrystalline mineral reaction zones, which are mainly composed of metasomatic minerals (spinel, forsterite, and phlogopite) and carbonates, have been found (Guo et al., 2021). These reaction zones represent the flow channels of reactive fluids and the products of fluid–marble interactions. Three episodes of fluid influx have been revealed in the reaction zones based on mineral zonations, replacement textures, and baddeleyite–zircon dating (Guo et al., 2021). The first (35–36 Ma) and second (23–24 Ma) infiltration episodes are related to the syenite magmatic events, and the third episode (∼ 17 Ma) is related to the Si-rich fluids derived from the gneisses.

In this study, two samples (13MDL76-A and 13MDL76-B; Guo et al., 2021) of reaction zones in the marbles (Fig. 2a) were investigated. Both samples are composed of calcite (36 vol %–47 vol %), forsterite (18 vol %–36 vol %), spinel (7 vol %–18 vol %), phlogopite (3 vol %–17 vol %), and small amounts of dolomite and accessory minerals (baddeleyite, geikielite, rutile, zirconolite, and uraninite). The two samples have TiO₂ contents of 0.22 wt %–0.35 wt %, Zr contents of 71.1–128 µg g⁻¹, U contents of 65.1–118 µg g⁻¹, and Th contents of 5.17–9.55 µg g⁻¹ (Guo et al., 2021). Mineral abbreviations are according to Whitney and Evans (2010), and zirconolite and uraninite are abbreviated as Zrl and Urn, respectively.

Figure 2 Specimen photograph, photomicrographs (in plane-polarized light), and BSE images of Zrl-I and associated minerals. (a) Reaction zones in the marbles. (b–e) Zrl-I occurring as irregularly shaped crystals (0.04–10 mm in size) in the matrix of the reaction zones. Coexistence of spinel, forsterite, phlogopite, calcite, and Zrl-I containing calcite inclusions from samples 13MDL76-A (b–c) and 13MDL76-B (d–e). (f–g) Zrl-I from samples 13MDL76-A (f) and 13MDL76-B (g) in the resin mount. There is no obvious compositional zonation in Zrl-I.

3 Analytical methods

3.1 Microtexture and X-ray mapping

The petrographic observations were conducted on thin sections using an optical microscope and a Zeiss Gemini 450 field-emission scanning electron microscope (FE-SEM) equipped with an X-ray energy dispersive spectrometer and a pneumatically retractable backscattered electron (BSE) system at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS). The observations, BSE image acquisition, and compositional maps were operated at accelerating voltages of 15–20 kV, a beam current of 5 nA, and a working distance of 8.5 mm. The X-ray mapping was performed by energy dispersive X-ray spectroscopy (EDS) with a dwell time of 2000 µs per point.

3.2 Mineral compositions by microprobe analyses

The compositions of zirconolite, baddeleyite, geikielite, and rutile were analyzed using a Cameca SX Five field emission electron probe micro-analyzer (EPMA) at IGGCAS. Analyses were operated in wavelength-dispersion mode (WDS) with an acceleration voltage of 20 kV, a beam current of 20 nA, and a beam diameter of 1 µm. The crystals used for element analyses were as follows: one thallium acid phthalate (TAP) crystal for Si, Mg, and Al analyses; one large pentaerythritol (LPET) crystal for Zr, Nb, Ca, K, Sc, Th, U, and Y analyses; one large lithium fluoride (LLIF) crystal for Ti, La, Nd, Mn, Sm, Gd, Er, and Hf analyses; and another LLIF crystal for Fe, Ce, Pr, Eu, Dy, Yb, and W analyses. The peak counting time was 10 s for Si, Mg, Zr, Nb, Ti, Fe, La, Ce, and Mn, and the background counting time was 5 s at the high- and low-energy background positions. For the other elements, the peak counting time was 20 s and the background counting time was 10 s at each site. The following excitation lines and calibrant materials have been used: U – Mα, synthetic U glass; Zr – Lα, zircon; Hf – Lα, synthetic Hf metal; Ti – Kα, rutile; Ca – Kα, rhodonite; Nb – Lα, synthetic Nb₂O₅; Sc – Kα, synthetic Sc₂O₃; W – Lβ, scheelite; Mg – Kα, synthetic MgO; Si – Kα, rhodonite; Mn – Kα, rhodonite; Th – Mα, synthetic Th glass; Al – Kα, K-feldspar; K – Kα, K-feldspar; Fe– Kα, hematite; Y – Lα, synthetic Y glass; La – Lα, synthetic La glass; Ce – Lα, synthetic Ce glass; Pr – Lβ, synthetic Pr glass; Nd – Lα, synthetic Nd glass; Sm– Lβ, synthetic Sm glass; Eu – Lα, synthetic Eu glass; Gd – Lα, synthetic Gd glass; Dy – Lα, synthetic Dy glass; Er – Lβ, synthetic Er glass; and Yb – Lα, synthetic Yb glass. The standard materials used for both U and Th (from P and H Developments Ltd.; https://www.pandhdevelopments.com/, last access: 14 December 2023) were synthesized glasses. The glass standard used for Th analyses contains 7.40 wt % Al, 27.3 wt % Si, 15.8 wt % Ca, 5.18 wt % Th, and 44.6 wt % O; and the glass standard used for U analyses contains 7.42 wt % Al, 27.5 wt % Si, 16.1 wt % Ca, 3.85 wt % U, and 44.9 wt % O. The synthetic standard used for REE analyses was also Si–Al–Ca–O glasses with a single 10 % REE added. Averaged detection limits (3σ) were as follows (concentrations in µg g⁻¹): U – 658, Zr – 507, Hf – 444, Ti – 248, Ca – 73, Nb – 392, Sc – 72, W – 1434, Mg – 110, Si – 145, Mn – 220, Th – 676, Al – 77, K – 73, Fe – 219, Y – 344, La – 651, Ce – 220, Pr – 220, Nd – 220, Sm – 1114, Eu – 397, Gd – 438, Dy – 373, Er – 495, and Yb – 436.

The major and minor element compositions of silicates and oxides (forsterite, spinel, and phlogopite) were measured by a JOEL-8100 EPMA at IGGCAS. The measurements were collected in WDS mode with an acceleration voltage of 15 kV, a beam current of 10 nA, and a beam diameter of 1 µm. Natural and synthetic oxides were used as standards. The precision is ∼ 1.5 % for Ca, Ti, U, and Zr elements (> 5 wt %) but is ∼ 5 %–20 % for other elements due to their low concentrations.

Figure 3 Photomicrographs (in plane-polarized light) and BSE images of Zrl-II and associated minerals. (a–b) Zrl-II occurring as irregularly shaped crystals (0.04–10 mm in size) in the matrix of the reaction zones. Coexistence of forsterite, phlogopite, calcite, and Zrl-II containing calcite inclusions from sample 13MDL76-A. (c–d) Core–rim zoning structure of Zrl-II from samples 13MDL76-A (c) and 13MDL76-B (d).

Figure 4 BSE images of Zrl-III (in the rein mount) occurring as small, rounded to platy polycrystalline grains around baddeleyite and forming continuous or discontinuous coronas from samples 13MDL76-B (a–c) and 13MDL76-A (d). (a–b) Zrl-III coronas that extend into the interior of baddeleyite along some fine fractures. (c–d) Zrl-III coexisting or intergrown with quartz. The boundaries between the baddeleyite and Zrl-III show either a zigzag shape (b–d) or are smooth (a).

3.3 U–Pb ages of zirconolite

Zirconolite U–Pb dating was carried out by laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) employing an Agilent 7500a Q-ICP-MS instrument (Agilent Technologies, USA) coupled to a 193 nm ArF excimer laser system (Geolas HD, Lambda Physik, Göttingen, Germany) at IGGCAS. The method and procedure are similar to those described in Xie et al. (2008). Helium was employed as the carrying gas to improve the transporting efficiency of ablated aerosols. The laser beam diameter was 16 µm at a repetition rate of 3 Hz, and the density of energy was ∼ 4.0 J cm⁻². The Phalaborwa zirconolite reference (SIMS ²⁰⁷Pb $/$ ²⁰⁶Pb age = 2067 ± 9 Ma, 1s, n = 16; Wu et al., 2010) was measured every six sample spot analyses to calibrate Pb $/$ U ratios and U contents. The resulting data were reduced based on the GLITTER program (Griffin et al., 2008). Weighted ²⁰⁶Pb $/$ ²³⁸U mean dates were calculated using the ²⁰⁷Pb correction of common Pb, assuming a common Pb composition corresponding to the two-stage crustal Pb model of Stacey and Kramers (1975). The age calculations and plotting of Tera-Wasserburg diagrams were made using IsoplotR (Vermeesch, 2018).

3.4 Trace element analysis of zirconolite by LA-ICP-MS analyses

Trace element contents of zirconolite were determined by the same instrument used for the measurement of U–Pb isotopes at IGGCAS. The method is similar to those outlined in Wu et al. (2018). The density of energy at the ablation spots was ∼ 4.0 J cm⁻². The laser beam diameter was 32 µm at a repetition rate of 5 Hz. The spot locations of trace element analyses are close to those of U–Pb isotope analyses. ⁴³Ca was used as the internal standard. The NIST 610 glass standard (Pearce et al., 1997) was used as the calibration material, and the glass standards of ARM-1 (Wu et al., 2019, 2021) and BCR-2G (Rocholl, 1998) were analyzed as unknown samples to monitor the data quality. The resulting data were also reduced based on the GLITTER program (Griffin et al., 2008). For most trace elements (> 0.10 µg g⁻¹), the accuracy is better than ± 10 % with an analytical precision (1 RSD, relative standard deviation) of ± 10 %.

Figure 5 Ca (a), U (b), Ti (c), Zr (d), Si (e), and Hf (f) X-ray maps showing the replacement of baddeleyite by Zrl-III in Fig.4d. Zrl-III coexists or is intergrown with quartz (e).

Figure 6 (a) Uraninite inclusions in forsterite from sample 13MDL76-B. (b–c) Intergrowth of geikielite and rutile in the matrix from samples 13MDL76-A (b) and 13MDL76-B (c). (d) Rutile coexisting with spinel, phlogopite, and calcite from sample 13MDL76-A.

4 Results

4.1 Occurrence of zirconolite and associated minerals

Several thin sections were prepared from samples 13MDL76-A and 13MDL76-B to examine the occurrences and textures of zirconolite. Moreover, zirconolite and baddeleyite grains were separated from each sample and mounted in resins for further observation. Three types of zirconolite were recognized based on their occurrence, texture, and coexisting mineral assemblage.

The first type of zirconolite (Zrl-I) typically occurs as irregularly shaped crystals (0.04–10 mm in size) in the matrix of the reaction zones. Zrl-I coexists with forsterite, spinel, calcite, phlogopite, and dolomite (Fig. 2b–e) and contains calcite inclusions (Fig. 2e). No compositional zonation can be observed in Zrl-I (Fig. 2c, e–g). The second type of zirconolite (Zrl-II) also occurs as an isolated mineral in the matrix. The shape, grain size, coexisting mineral assemblage, and inclusions of Zrl-II are similar to those of Zrl-I (Fig. 3). However, Zrl-II exhibits core–rim zonation. The rims of Zrl-II are brighter than the cores in the BSE images (Fig. 3c and d).

The third type of zirconolite (Zrl-III) exclusively occurs as small, rounded to platy polycrystalline grains around the external edges of baddeleyite and forms continuous or discontinuous coronas that are ≤ 25 µm thick (Fig. 4). We prepared a resin mount of separated baddeleyite to observe Zrl-III (Fig. 4) because baddeleyite can rarely be observed in thin sections. The width of zirconolite around a single baddeleyite grain is highly variable at different locations. The boundaries between the baddeleyite and zirconolite are either zigzag-shaped (Fig. 4b–d) or smooth (Fig. 4a). The original shapes of baddeleyite grains and the reaction interface can be found. The growth of Zrl-III also occurs in the interior of baddeleyite along some fine fractures that extend to the outside of the baddeleyite grains (Fig. 4a). The BSE images (Fig. 4c and d) and element maps (Fig. 5) show that Zrl-III typically coexists with or is intergrown with fragmental, porous, polycrystalline quartz.

Moreover, in both samples, some Ti-bearing and U-bearing accessory minerals, such as geikielite (MgTiO₃), rutile (TiO₂), and uraninite (UO₂), can be observed in the matrix or as inclusions in silicate minerals (Fig. 6).

Table 1 Representative major and minor element compositions (wt %) and formulas of different types of zirconolite from the MMB.

bdl: below detection limits. C: core. R: rim.

Figure 7 Compositional variations in zirconolite from the MMB. (a) Plots of CaO vs. UO₂ (wt %) discriminating between the different types of zirconolite. (b–d) Plots of Mg${}^{\mathrm{2}+}+$ Fe²⁺ vs. Th${}^{\mathrm{4}+}+$ U⁴⁺ (apfu) (b), Nb${}^{\mathrm{5}+}+$ Al³⁺ vs. Ti⁴⁺ (apfu) (c), and Hf vs. Zr⁴⁺ (µg g⁻¹) (d) showing negative or positive correlations between them.

4.2 Mineral compositions

4.2.1 Zirconolite

The compositions of zirconolite are presented in Tables 1 and 2. Zrl-I has 28 wt %–30 wt % ZrO₂, 31 wt %–33 wt % TiO₂, and 8.5 wt %–9.6 wt % CaO (Fig. 7a; Table 1). The UO₂ content of Zrl-I ranges from 21 wt %–23 wt %, and the ThO₂ content ranges from 0.84 wt %–1.99 wt %. Zrl-I also contains small amounts of MgO (2.6 wt %–3.2 wt %), FeO (1.2 wt %–1.3 wt %), Al₂O₃ (0.46 wt %–0.74 wt %), HfO₂ (0.58 wt %–0.72 wt %), Nb₂O₅ (0.37 wt %–0.52 wt %), and Sc₂O₃ (≤ 0.03 wt %) (Fig. 8c; Fig. S1; Table 1). The contents of REEs are very low (generally lower than the limit of detection by EPMA analyses). LA-ICP-MS analyses indicate that the ΣREE of Zr-I ranges from 123 to 174 µg g⁻¹ (Fig. 9; Table 2). Zrl-I exhibits a light rare earth element (LREE)-depleted and flat heavy rare earth element (HREE) pattern in the chondrite-normalized diagram with negative Eu anomalies (Fig. 9). The Cs, Rb, and Ba contents of Zrl-I are lower than the limits of detection (Table 2).

Zrl-II has ZrO₂ (28 wt %–30 wt %), TiO₂ (30 wt %–32 wt %), CaO (8.6 wt %–9.5 wt %), UO₂ (21 wt %–24 wt %), and ThO₂ (0.71 wt %–2.0 wt %) contents similar to those of Zrl-I (Fig. 7a; Table 1). The brighter rims in the BSE images have slightly higher UO₂ and lower CaO contents than the darker cores (Fig. 7a; Table 1). The contents of MgO, FeO, Al₂O₃, Sc₂O₃, and HfO₂ in both cores and rims are similar (Fig. S1). Both the cores and rims have low REE contents (125 to 158 µg g⁻¹ for ΣREE) and show a LREE-depleted pattern similar to that of Zrl-I (Fig. 9).

Figure 8 (a) Ca–Ti–Zr compositional variations (apfu) of zirconolite. (b) Ca–(Th + U)–(REE + Y) compositional variations (apfu) of zirconolite. (c) Plots of Nb₂O₅ vs. HfO₂ (wt %) of zirconolite. Zirconolite from nepheline syenite or syenite (Platt et al.,1987; Ventura et al., 2000; Bellatreccia et al., 2002; Haifler et al., 2017; Melluso et al., 2017), ultra-mafic to mafic rocks (Williams, 1978; Lorand and Cottin, 1987; Stucki et al., 2001; Rajesh et al., 2006; Azzone et al., 2009), carbonatite (Williams and Giere, 1996), metasomatic rocks (Purtscheller and Tessadri., 1985; Gieré, 1986; Gieré and Williams, 1992; Zaccarini et al., 2004; Pascal et al., 2009; Proyer et al., 2014), and lunar basalt (Busche et al., 1972; Li et al., 2021; Wang et al., 2021).

Figure 9 Chondrite-normalized REE patterns (Sun and McDonough, 1989) of different types of zirconolite from the MMB.

Compared to Zrl-I and Zrl-II, Zrl-III is found to have higher contents of ZrO₂ (32 wt %–38 wt %), TiO₂ (36 wt %–44 wt %), CaO (12 wt %–15 wt %), Nb₂O₅ (0.74 wt %–1.3 wt %), and HfO₂ (0.93 wt %–1.04 wt %); much lower contents of UO₂ (0.88 wt %–5.3 wt %); and slightly lower contents of MgO (0.24 wt %–1.3 wt %) and FeO (0.47 wt %–1.2 wt %) (Table 1; Fig. S1). The ThO₂ contents range from 0.60 wt %–6.8 wt %. The Al₂O₃ contents of Zrl-III (0.52 wt %–0.77 wt %) are similar to those of Zrl-I (0.46 wt %–0.74 wt %) and Zrl-II (0.49 wt %–0.72 wt %). The LA-ICP-MS analyses indicate that the contents of REEs in Zrl-III (366 to 477 µg g⁻¹ for ΣREE) are higher than those of Zrl-I and Zrl-II (Fig. 9; Table 2). In the chondrite-normalized diagram, Zrl-III shows a nearly flat REE pattern (except La and Ce) with a negative Eu anomaly (Fig. 9).

Table 2 Representative trace element compositions of zirconolite (concentrations in µg g⁻¹).

bdl: below detection limits. C: core. R: rim.

4.2.2 Baddeleyite, rutile, and geikielite

The compositions of baddeleyite, rutile, and geikielite measured by EPMA are presented in Table S1. Baddeleyite is mainly composed of ZrO₂ (94 wt %–98 wt %) and HfO₂ (2.3 wt %–5.2 wt %). This mineral also contains minor amounts of TiO₂ (0.18 wt %–0.52 wt %) and Nb₂O₅ (0.13 wt %–0.22 wt %). Rutile is composed of TiO₂ (> 97 wt %) and contains minor amounts of ZrO₂ (0.03 wt %–0.59 wt %), FeO (0.02 wt %–0.09 wt %), CaO (0.07 wt %–1.1 wt %), and Nb₂O₅ (≤ 0.54 wt %). Geikielite is mainly composed of TiO₂ (63 wt %–66 wt %), MgO (27 wt %–30 wt %), and FeO (5.2 wt %–7.0 wt %). The contents of UO₂ and ThO₂ in the three minerals above are lower than the limits of detection by EPMA analyses.

Figure 10 (a–b) Tera-Wasserburg U–Pb concordia diagrams for Zrl-I from 13MDL76-A (a) and 13MDL76-B (b). (c–d) Tera-Wasserburg U–Pb concordia diagrams for Zrl-II from 13MDL76-A (c) and 13MDL76-B (d). The red and black ellipses represent the core and rim of Zrl-II, respectively. (e) Tera-Wasserburg U–Pb concordia diagrams for Zrl-III from 13MDL76-B. The discordia lines in the Terra-Wasserburg diagrams are forced through a ²⁰⁷Pb $/$ ²⁰⁶Pb value of 0.84 ± 0.05. The ²⁰⁷Pb $/$ ²⁰⁶Pb values were estimated using the two-stage crustal Pb model of Stacey and Kramers (1975) for all samples. The gray areas in Tera-Wasserburg U–Pb concordia diagrams show that the variation in the estimated ²⁰⁷Pb $/$ ²⁰⁶Pb value has little effect on the calculated ages for low common Pb contents for Zrl-I and Zrl-II (f₂₀₆ < 2 %, Table S2; Li et al., 2012). The data-point error bars are 1σ.

Figure 11 Schematic model (not-to-scale) of multiple episodes of fluid infiltration in the marbles (a, c, e) and the growth of Zrl-I (b), Zrl-II (d), and Zrl-III (f).

4.2.3 Silicate minerals and spinel

The compositions of silicate and oxide minerals (forsterite, spinel, and phlogopite) coexisting with zirconolite were analyzed, and the results are presented in Table S1. Forsterite has near-endmember compositions with X_Mg [= Mg $/$ (Mg + Fe²⁺)] = 0.99. Spinel has high Al contents of 68 wt %–70 wt % and X_Mg of 0.98–0.90. The Cr₂O₃ contents of spinel range from 0.80 wt %–3.3 wt %. Phlogopite has near-endmember compositions (X_Mg=0.99–1) and high TiO₂ contents of 0.57 wt %–1.5 wt %. The K (0.63–0.87 apfu, atomic proportion per formula unit) and Na (0.11–0.38 apfu) contents of phlogopite are high, and the Ca (≤ 0.002 apfu) contents are low. In addition, phlogopite contains considerable amounts of F (1.6 wt %–1.9 wt %).

4.3 U–Pb ages of zirconolite

The U–Th–Pb isotope contents of zirconolite are presented in Table S2 and shown in Fig. 10. The f₂₀₆ values of zirconolite in the MMB range from 1 %–7 %, indicating low proportions of common Pb. The $\mathrm{Th}/\mathrm{U}$ ratios of Zrl-I (0.03–0.08) and Zrl-II (0.04–0.08) are lower than those of Zrl-III (0.40–2.9). A total of 18 analyses on Zrl-I from sample 13MDL76-A and 16 analyses on Zrl-I from sample 13MDL76-B give lower intercept ages of 34.8 ± 0.5 Ma (2σ, mean square weighted deviation, MSWD = 1.7; Fig. 10a) and 35.2 ± 0.5 Ma (2σ, MSWD = 2; Fig. 10b), respectively, in Tera-Wasserburg concordia diagrams. Their weighted mean ²⁰⁶Pb $/$ ²³⁸U ages after ²⁰⁷Pb correction are 34.9 ± 0.5 Ma (2σ, MSWD = 1.6) and 35.3 ± 0.5 Ma (2σ, MSWD = 1.8), respectively. Four analyses on the 13MDL76-A Zrl-II and seven analyses on the 13MDL76-B Zrl-II yield lower intercept ages of 34.4 ± 1.0 Ma (2σ, MSWD = 0.17; Fig. 10c) and 34.9 ± 0.8 Ma (2σ, MSWD = 1.5; Fig. 10d), respectively. Their weighted mean ²⁰⁶Pb $/$ ²³⁸U ages after ²⁰⁷Pb correction are 34.4 ± 1.0 Ma (2σ, MSWD = 0.14) and 34.9 ± 0.8 Ma (2σ, MSWD = 1.1), respectively. There is no difference in the ages between the Zrl-II cores and rims. Six analyses of Zrl-III from 13MDL76-B define a regression line with a lower intercept age of 18.6 ± 0.9 Ma (2σ, MSWD = 1.1; Fig. 10e). The weighted mean ²⁰⁶Pb $/$ ²³⁸U age after ²⁰⁷Pb correction is 18.6 ± 0.9 Ma (2σ, MSWD = 0.92).

5 Discussion

5.1 Multiple growth of zirconolite in the reaction zones of marbles

The studied reaction zones in marbles were formed by episodic infiltrations of external reactive fluids (Guo et al., 2021). Thus, the growth of zirconolite in these reaction zones is associated with the different stages of fluid influx and mineral reactions.

Zrl-I occurs in the matrix and typically coexists with carbonates (calcite and dolomite), forsterite, spinel, and phlogopite, and it also contains calcite inclusions (Fig. 2a–f). The boundaries between Zrl-I and coexisting minerals are generally sharp, suggesting the simultaneous formation of these minerals. Therefore, Zrl-I developed during the first infiltration process that led to the metasomatic formation of forsterite, spinel, and phlogopite by the replacement of dolomite at 704–750 ^∘C and ∼ 8 kbar (Guo et al., 2021). Zrl-I has very high ZrO₂ (28 wt %–30 wt %), TiO₂ (31 wt %–33 wt %), UO₂ (21 wt %–23 wt %), and ThO₂ (0.84 wt %–1.9 wt %) contents, and contain considerable amounts of Nb, Ta, and Hf (Table 1; Figs. 7c, 7d, 8c), indicating high contents of these elements in the K–Al–Si-bearing infiltrating fluid. This interpretation is in accordance with the petrographic observations that abundant rutile, uraninite, and baddeleyite occur in the reaction zones (Fig. 6), indicating high Zr–Ti–U contents in the reactive fluid. The metasomatic reactions involved in the formation of Zrl-I at the expense of dolomite can be summarized as follows:

$$\begin{array}{}\text{(R1)}& {\displaystyle }\begin{array}{rl}\mathrm{2}\mathrm{Dol}& +\mathrm{4}{\mathrm{TiO}}_{\mathrm{2}}\left(\text{aq}\right)+{\mathrm{ZrO}}_{\mathrm{2}}\left(\text{aq}\right)\\ & =\mathrm{Zrl}+\mathrm{Cal}+\mathrm{2}\mathrm{Gk}+\mathrm{3}{\mathrm{CO}}_{\mathrm{2}},\end{array}\text{(R2)}& {\displaystyle }\begin{array}{rl}\mathrm{2}\mathrm{Dol}& +\mathrm{2}{\mathrm{TiO}}_{\mathrm{2}}\left(\text{aq}\right)+{\mathrm{ZrO}}_{\mathrm{2}}\left(\text{aq}\right)+{\mathrm{SiO}}_{\mathrm{2}}\left(\text{aq}\right)\\ & =\mathrm{Zrl}+\mathrm{Cal}+\mathrm{Fo}+\mathrm{3}{\mathrm{CO}}_{\mathrm{2}},\end{array}\text{(R3)}& {\displaystyle }\begin{array}{rl}\mathrm{2}\mathrm{Dol}& +\mathrm{2}{\mathrm{TiO}}_{\mathrm{2}}\left(\text{aq}\right)+{\mathrm{ZrO}}_{\mathrm{2}}\left(\text{aq}\right)+\mathrm{2}{\mathrm{Al}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}\\ & =\mathrm{Zrl}+\mathrm{Cal}+\mathrm{2}\mathrm{Spl}+\mathrm{3}{\mathrm{CO}}_{\mathrm{2}},\end{array}\text{(R4)}& {\displaystyle }\begin{array}{rl}\mathrm{6}\mathrm{Dol}& +\mathrm{2}{\mathrm{TiO}}_{\mathrm{2}}\left(\text{aq}\right)+{\mathrm{ZrO}}_{\mathrm{2}}\left(\text{aq}\right)+{\mathrm{Al}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}+\mathrm{6}{\mathrm{SiO}}_{\mathrm{2}}\\ & +{\mathrm{K}}_{\mathrm{2}}\mathrm{O}+\mathrm{2}{\mathrm{H}}_{\mathrm{2}}\mathrm{O}=\mathrm{Zrl}+\mathrm{5}\mathrm{Cal}+\mathrm{2}\mathrm{Phl}+\mathrm{7}{\mathrm{CO}}_{\mathrm{2}},\end{array}\text{(R5)}& {\displaystyle }\begin{array}{rl}\mathrm{Dol}& +\mathrm{2}{\mathrm{TiO}}_{\mathrm{2}}\left(\text{aq}\right)+{\mathrm{UO}}_{\mathrm{2}}\left(\text{aq}\right)\\ & =\text{U-Zrl}\left({\mathrm{UZrTi}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{7}}\right)+\mathrm{Cal}+{\mathrm{CO}}_{\mathrm{2}},\end{array}\text{(R6)}& {\displaystyle }\begin{array}{rl}\mathrm{Dol}& +\mathrm{2}{\mathrm{TiO}}_{\mathrm{2}}\left(\text{aq}\right)+{\mathrm{ThO}}_{\mathrm{2}}\left(\text{aq}\right)\\ & =\text{Th-Zrl}\left({\mathrm{ThZrTi}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{7}}\right)+\mathrm{Cal}+{\mathrm{CO}}_{\mathrm{2}}.\end{array}\end{array}$$

Zrl-II has a grain shape, size, and inclusions similar to those of Zrl-I. Both types of zirconolite have the same coexisting mineral assemblages. Moreover, Zrl-II cores have the same compositional variations and REE patterns as those of Zrl-I (Fig. 3d). Thus, we consider that the growth mechanism of Zrl-II cores is similar to that of Zrl-I. The formation of irregular Zrl-II rims is related to a later alteration process because the sharp core–rim contact of Zrl-II is consistent with fluid-assisted replacement reactions (e.g., Putnis and Austrheim, 2010).

The corona textures of Zrl-III around baddeleyite and the jagged grain boundaries between the two minerals (Fig. 4) indicate the growth of Zrl-III by the replacement of baddeleyite. Similar textures of zirconolite rimming baddeleyite have been previously observed in metacarbonate rocks (e.g., Purtscheller and Tessadri, 1985; Tropper et al., 2007), mafic–ultramafic magmatic rocks (e.g., Rasmussen and Fletcher, 2004; Azzone et al., 2009; Hurai et al., 2018), and lunar basalt (e.g., Rasmussen et al., 2008; Li et al., 2021). The formation of the Zrl-III corona is closely related to fluid metasomatism via a mechanism of interface-coupled dissolution–precipitation (ICDP) (e.g., Putnis and Austrheim, 2010; Guo et al., 2017). Zrl-III does not form on all the boundaries of baddeleyite (Fig. 4), suggesting that Zrl-III formed only on the sites where the infiltrating fluid was able to pass. Calcium in Zrl-III is locally derived from calcite, and Ti may be derived from the infiltrating fluid. The large compositional variation for the Zrl-III corona after a single baddeleyite grain (Fig. 4a, Table 1) may be caused by different degrees of fluid–mineral interaction and element exchange at different sites of the baddeleyite grain boundary. Based on the reactants and products, the following reaction can be inferred:

$$\begin{array}{}\text{(R7)}& \mathrm{Bdy}+\mathrm{2}{\mathrm{TiO}}_{\mathrm{2}}\left(\text{aq}\right)+{\mathrm{CaCO}}_{\mathrm{3}}=\text{Zrl-III}+{\mathrm{CO}}_{\mathrm{2}}.\end{array}$$

Zrl-III has variable and much higher UO₂ (0.88 wt %–5.3 wt %) and ThO₂ (0.60 wt %–6.8 wt %) than baddeleyite (UO₂ ≤ 0.23 wt %, ThO₂ ≤ 0.06 wt %; Table S1), indicating that the infiltrating fluids responsible for Zrl-III also contain certain amounts of U and Th. However, the UO₂ contents of Zrl-III are obviously lower than those of Zrl-I and Zrl-II (Fig. 7a; Table 1), implying that the later-stage infiltrating fluid has lower UO₂ contents than the first-stage infiltrating fluid. Interestingly, Zrl-III is found to be intergrown with polycrystalline quartz (Figs. 4c, d and 5e). This phenomenon has rarely been reported in natural rocks (Rasmussen and Fletcher, 2004; Rasmussen et al., 2008, 2009). The coexistence of Zrl-III and quartz in the MMB samples indicates that this stage of the reactive fluid was locally SiO₂-saturated. Guo et al. (2021) revealed that a gneiss-derived later-stage fluid with high SiO₂ activity infiltrated the marbles at ∼ 680 ^∘C and 5 kbar. We therefore suggest that the development of Zrl-III is also associated with this stage of high-${\mathit{\alpha }}_{{\mathrm{SiO}}_{\mathrm{2}}}$ fluid influx.

In summary, the above results indicate that the three types of zirconolite record multiple episodes of fluid infiltration in the marbles. The first stage of infiltration led to the formation of Zrl-I and Mg-rich silicates and oxides, as well as other accessory minerals (baddeleyite and geikielite) (Fig. 11a and b). The reactive K–Al–Si fluid was also enriched in Zr, Ti, U, and Th. After that, a portion of Zrl-I grains experienced a local fluid-assisted dissolution–precipitation process, producing a core–rim zonation (i.e., the Zrl-II type) (Fig. 11c and d). The final stage of infiltration formed Zrl-III at the expense of baddeleyite. This final stage of fluid was characterized by relatively low U contents and high SiO₂ activities (Fig. 11e and f).

5.2 Constraints on episodic infiltration by zirconolite U–Pb dating

The high U and Pb contents and low common Pb contents in zirconolite provide a good opportunity to elucidate the time-resolved infiltration history and episodic fluid–rock interactions in the marbles using in situ dating technology. The analyses of Zrl-I grains from both samples yield similar lower intercept ages of 34.8 ± 0.5 Ma (2σ, MSWD = 1.7, n=18; Fig. 10a) and 35.2 ± 0.5 Ma (2σ, MSWD = 2, n=16; Fig. 10b), consistent with the U–Pb ages of baddeleyite cores (35.8 ± 0.8 Ma; Guo et al., 2021) in the same samples. This result further supports the simultaneous formation of Zrl-I and baddeleyite, and therefore the ages represent the time of the earliest infiltration event in these marbles. Because the Zrl-I ages are similar to the emplacement ages of hornblende syenites near the marbles (from 33.1 ± 0.9 to 30.9 ± 0.6 Ma; Barley et al., 2003), it is reasonable to infer that the Zr–Ti–U–Th-rich metasomatic fluid is derived from the syenitic magma (Searle et al., 2020).

Zrl-II from the two samples gives lower intercept ages of 34.4 ± 1.0 Ma (2σ, MSWD = 0.17, n=4; Fig. 10c) and 34.9 ± 0.8 Ma (2σ, MSWD = 1.5, n=7; Fig. 10d), which are in accordance with those of Zrl-I within the uncertainties. The ages of the Zrl-II cores and rims are indistinguishable within the uncertainties. Thus, Zrl-II most likely represents Zrl-I grains that have experienced a local dissolution–precipitation process, which occurred slightly later than the formation of Zrl-I (i.e., Zrl-II cores).

The analyses of the Zrl-III corona around baddeleyite yield a lower intercept age of 18.6 ± 0.9 Ma (2σ, MSWD = 1.1, n=6; Fig. 10e), indicating an infiltration event postdating the formation of Zrl-I and Zrl-II. The ages are consistent with the time of the final stage of infiltration in the marbles constrained by zircon U–Pb dating (17.05 ± 0.27 Ma) in the same samples (Guo et al., 2021). The ages are also close to those of a biotite granite dike in the nearby Sedawgyi gneisses (17.0 ± 0.3 Ma; Mitchell et al., 2012). We thus suggest that the last stage of SiO₂-rich infiltrating fluids was derived from the partial melts of the gneisses during the Miocene uplift cooling process of the MMB (Bertrand et al., 1999, 2001; Garnier et al., 2006; Mitchell et al., 2012).

Therefore, the multiple formation stages of zirconolite record episodes of infiltration in the marbles ranging from 35 to 19 Ma during the India–Asia collision and uplift of the MMB (Fig. 11a–f).

5.3 Comparison with other types of zirconolite

Zirconolite occurs in a variety of rock types, including metasomatized metacarbonates, mafic–ultramafic magmatic rocks, carbonatites, syenite, syenite pegmatite, placer deposits, and extraterrestrial samples (see the review of Williams and Gieré, 1996). This mineral has several cation-acceptor sites, which allows a series of cation substitutions ranging in ionic radii from 0.40 to 1.14 A^∘ and charges from 2⁺ to 6⁺ (Giereì et al., 1998). The compositions of natural zirconolite vary extensively, and the predominant substitutions involve (1) U, Th, Pb, and REEs for Ca; (2) Hf for Zr; and (3) Al, Nb, Ta, Fe, Mg, Mn, Zn, and W for Ti (Gieré et al., 1998). Clarifying the correlation between zirconolite compositions and types of host lithologies is essential to understand the mineral growth mechanism and element substitution, as well as to establish a potential framework to trace the origin of zirconolite.

The MMB zirconolite, which formed by the metasomatism of dolomite marbles, is characterized by high UO₂ (0.88 wt %–24 wt %) and ThO₂ (0.6 wt %–6.8 wt %) contents and low REE contents (below the detection limits by EPMA). The U${}^{\mathrm{4}+}+$ Th⁴⁺ content correlates well with the Mg${}^{\mathrm{2}+}+$ Fe²⁺ content (Fig. 7b), indicating that the following coupled substitution causes the U⁴⁺ and Th⁴⁺ enrichments in zirconolite (Giereì and Williams, 1992): (U, Th)${}^{\mathrm{4}+}+$ (Mg, Fe)${}^{\mathrm{2}+}=$ Ca${}^{\mathrm{2}+}+$ Ti⁴⁺. The positive correlation between Nb${}^{\mathrm{5}+}+$ Al³⁺ and Ti⁴⁺ (Fig. 7c) supports the fact that the enrichments of Nb (maybe also Ta) are mainly governed by the following coupled substitution: (Nb, Ta)⁵⁺ + Al${}^{\mathrm{3}+}=$ 2Ti⁴⁺ (Giereì and Williams, 1992). The incorporation of Hf in zirconolite is controlled by the substitution of Hf${}^{\mathrm{4}+}=$ Zr⁴⁺, as indicated by the positive correlation between the two elements (Fig. 7d).

Figure 8a–c compare the compositions of zirconolite from the MMB samples with those from a variety of lithologies reported in previous studies, including mafic–ultramafic magmatic rocks (kimberlites and lunar basalts), carbonatites, syenites, and metasomatized carbonate rocks. MMB zirconolite has similar contents of Ca, Ti, and Zr to other types of zirconolite (Fig. 8a). The U + Th contents of zirconolite are highly varied (from 0.002 to 0.48 apfu) for all types of lithologies except the lunar samples, the zirconolite of which typically has relatively low U + Th contents (< 0.013 apfu) (Fig. 8b). In contrast, different types of zirconolite have distinct REE + Y concentrations (Fig. 8b). Generally, zirconolite in syenites, lunar basalts, and mafic–ultramafic magmatic rocks have high total REE + Y contents (up to 0.65 apfu). Zirconolite formed during metasomatic processes of carbonate rocks (skarn) and marbles, including the MMB zirconolite, typically exhibits the lowest contents of REE + Y (from 0 to 0.20 apfu). The REEs contents of zirconolite from carbonatites and most mafic–ultramafic rocks are in between those of the syenites and metasomatic rocks. The Nb₂O₅ and HfO₂ contents for different types of zirconolites are also highly variable (Fig. 8c). Zirconolite from the metasomatic rocks and lunar samples generally have lower Nb₂O₅ contents than those from carbonatites and syenites.

5.4 Implication for multiple stages of infiltration in metacarbonates and Zr–Ti–Th–U mineralization

Fluid/melt infiltration into metacarbonates and fluid/melt-rock interactions occur widely in subduction–collision orogenic belts, which may induce significant CO₂ release and mineralization of critical metals (e.g., Stewart et al., 2018; Guo et al., 2021, 2022; Kerrick, 1977; Meinert et al., 2005; Deng and Wang, 2016; Xie et al., 2021). A growing number of studies have indicated that the infiltration–metasomatic processes in metacarbonates are complex and involve multiple episodes of reactions (e.g., Jamtveit et al., 1993; Satish-Kumar et al., 2010; Brice et al., 2019; Tan et al., 2022; Guo et al., 2019, 2021; Zeng and Liu, 2022). However, it is difficult to precisely constrain the time and fluid chemical characteristics of each infiltration episode. Studies on zirconolite provide a new solution to address this problem. Our new results on the MMB reaction zones indicate that zirconolite can grow at various stages in response to the episodic fluid infiltrations and fluid-assisted replacement reactions. On the one hand, high U–Th–Pb contents and advances in the in situ isotopic analytical technique enable the zirconolite to yield the time of fluid influx (Sect. 5.2). On the other hand, compared to other accessory geochronometers (e.g., baddeleyite, zircon, and rutile) in metacarbonate rocks, zirconolite can accommodate more groups of trace elements in its crystal lattice (Sect. 5.3). Therefore, this mineral also provides more information on the compositions of introduced reactive fluids. Here, a comprehensive study on the MMB samples presents an example to elucidate the fluid influx history and reactive fluid compositions in the marbles using zirconolite. More importantly, the growth of zirconolite in metacarbonates is typically accompanied by the consumption of carbonates and the release of CO₂ (Reactions R1–R7). Thus, zirconolite may be of particular importance for characterizing the decarbonation process and the flow of carbonic fluids.

Our results indicate an efficient mobilization and transfer of Zr (Hf), Ti (Nb and Ta), and U (Th) during the fluid–rock interactions, although high-field-strength elements (HFSEs) were typically regarded as sparingly soluble components in aqueous fluids (e.g., Keppler and Wyllie, 1990, 1991; Audétat and Keppler, 2005; Tropper and Manning, 2005). Experimental and geological studies indicate that the enrichment of F and other halogens can significantly elevate the solubility of HFSEs by complexing with HFSEs in aqueous fluid (Keppler and Wyllie, 1990, 1991; Peiffer et al., 1996; Antignano and Manning, 2008; Cuney, 2009; Rapp et al., 2010; Bali et al., 2011; Migdisov et al., 2011; Tanis et al., 2016; Karmakar, 2021). The high F contents in phlogopite (1.6 wt %–1.9 wt %) from the reaction zones indicate that the reactive HFSE-bearing fluids also contain F. We thus suggest that F played an important role in transporting HFSEs in the fluids (Gieré, 1986; Salvi and WilliamsJones, 1996; Zhao et al., 2016; Guo et al., 2016; Duan and Li, 2017; Zeng and Liu, 2022), and the growth of phlogopite during the fluid–marble interaction decomposed F–HFSE complexes to deposit Zrl and other HFSE-rich minerals. In addition, we reveal that the mineralization of Zr–Ti–U-rich phases in (meta-)carbonate rocks could be episodic. This provides a new insight into the transport and precipitation of HFSEs during crustal fluid–rock interactions.

6 Conclusions

Three textural types of zirconolite (Zrl-I, Zrl-II, and Zrl-III), which formed by multiple episodes of fluid–marble interaction during the India–Asia continental collision, have been recognized in the reaction zones of MMB dolomite marbles. These types of zirconolite record important information about the timing of episodic infiltration and the chemical compositions of the reactive fluids. Zrl-I records the first episode of fluid infiltration at ∼ 35 Ma with high U and Th contents in the reactive fluid. Zrl-II, which has a core–rim compositional zonation, reflects a local dissolution–precipitation process slightly later than the first episode of fluid infiltration. Zrl-III, which formed by the replacement of baddeleyite, documents a final stage of SiO₂-rich fluid infiltration at ∼ 19 Ma. This late reactive fluid has lower U contents. All the infiltrating fluids have low REE contents. The fluids responsible for the Zrl-I and Zrl-II formation were derived from the syenite intrusions, whereas the fluid responsible for the formation of Zrl-III was associated with the granite dike in nearby gneisses. This study illustrates that the combination of detailed petrographic study, mineral chemistry, and in situ U–Pb dating on multistage-formed zirconolite can be used to constrain the time-resolved fluid infiltration history in metacarbonates and reactive fluid compositions.