Most of the harzburgites from the Purang ophiolite are massive and show brownish colors (Fig. 2a). A few of them are serpentinized, whereby the alteration increases towards the chromitite inclusions. In general, they are coarse-grained and contain high amounts of olivine (80 vol %–90 vol %), about 10 vol %–20 vol % orthopyroxene, minor amounts of clinopyroxene (<5 vol %) and accessory amounts of Cr-spinel (Figs. 2, 3a). They show a characteristic porphyroblast texture with large olivine and orthopyroxene crystals (1–5 and 5–20 mm, respectively), which display strongly embayed grain boundaries filled with fine-grained olivine and pyroxene neoblasts (≤100–200 µm). Olivine and orthopyroxene porphyroclasts show ductile deformation, such as undulatory extinction, kink bands and tectonic twinning.

Most of the podiform chromitites in the Purang ophiolite are surrounded by dunite envelopes that vary from a few centimeters to several meters in thickness (Fig. 2b). The thickest and best-developed ones occur around disseminated chromitite. This type of dunite is moderately fresh, light yellowish-brown in color and massive. Typical samples consist of 98 vol % olivine and 2 % magnesiochromite (Fig. 3b); neither clinopyroxene nor orthopyroxene has been identified in these rocks, not even as inclusions in olivine or chromite. Most olivine grains are subhedral, form granular crystals, and show grain sizes between 2 and 6 mm. They rarely exhibit deformation structures such as kink banding or undulatory extinction. Spinel occurs as small euhedral to subhedral grains (Fig. 3b) and as larger (up to 5 mm), irregular, amoeboid crystals, some of which contain small inclusions of euhedral olivine.

This type of dunite lens is enclosed in harzburgite, forming very fresh and mainly brown rock inclusions (Fig. 2c). In general, the lenses are oriented northwest–southeast and show a width of 1–5 m and a length of 5–10 m. The rock is mainly composed of olivine (95–99 modal %, Fig. 3) and contains small amounts of Cr-rich spinel, orthopyroxene and clinopyroxene. The olivine is large in grain size, mainly 0.5–3 mm in length. This type of olivine does not contain any inclusions or exsolution textures. Locally, the rims of olivine are replaced by serpentine. Chromian spinel has a particle size of between 0.2 and 1 mm.

Type 3 dunite lenses and pods in the Purang peridotites are mostly 0.1–2 m in width and commonly aligned in a northwest–southeast direction, roughly parallel to the main foliation of the host harzburgites (Fig. 2d). The dunite is fresh and nearly unaltered (<2 vol % serpentine) with brownish-black to yellowish-brown colors. It consists mainly of olivine (95–99 modal %) and orthopyroxene (2 %–4 %) with minor amounts of clinopyroxene and Cr-poor spinel. Olivine grains are commonly 2.5–3 mm in diameter, but recrystallized grains are much smaller. Some of the largest grains show undulatory extinction and deformation twinning. Cr-poor spinel, 0.2–1 mm in size, is mostly randomly distributed in the rocks but can locally concentrate to form small, disseminated clusters.

Figure 7a and b compare Fo vs. NiO and MnO, respectively. Figure 7a shows that olivine compositions from harzburgite and the three types of dunite overlap to a large degree. However, forsterite contents in olivine type 1 dunite envelopes and dunite lenses with Cr-rich spinel reach somewhat higher values of Fo91.8 and Fo92.2, respectively, than harzburgite (Fo91.2) or dunite lenses with Cr-poor spinel (Fo90.7).

In Fig. 7b the olivine compositions of the different source rocks are presented in a MnO vs. Fo component diagram, again documenting similar compositions of olivines in harzburgites and type 3 dunite. Olivine of dunite that envelopes chromitite lenses shows a higher Fo component; the respective values of olivine from type 2 dunite lenses are even higher.

Lithium isotope compositions, the respective amounts of forsterite component of olivine, as well as selected chemical parameters of accompanying Cr-rich spinel of the harzburgite, and the three different types of dunite are listed in Table 2. Generally, the Li concentrations and δ7Li values of the olivines studied show large variations from 0.75 to 1.74 ppm and 6.19 ‰ to +14.72 ‰, respectively, and they are negatively correlated (Fig. 8). Note that the olivines in harzburgite exhibit the lowest amounts of Li (0.8 ppm on average; range of 0.75–0.89 ppm) and the highest concentrations of δ7Li (average of 13.8 ‰; range of 11.82–14.72 ppm), whereas those from type 1 dunite enveloping chromitite lenses have the highest amounts of Li (1.6 ppm on average; range of 1.48–1.74 ppm) and the lowest δ7Li concentrations (average of 7.1 ‰; range of 6.19 ‰–7.98 ‰). Li contents of olivines from type 2 and type 3 dunite lenses vary widely between 0.94 and 1.64 ppm, but some olivines of the type 2 dunite lenses have considerably higher amounts of Li (Fig. 8). The δ7Li values of olivine in the type 3 dunite lenses are in general higher (10.25 ‰–14.2 ‰) than those of the type 2 dunite lenses (6.77 ‰–10.76 ‰). In Fig. 9 a series of correlation diagrams display the amount of Li (in ppm) and the δ7Li values (in ‰) in olivine relative to the Fo component in olivine and the NiO and MnO content, as well as Cr# of the spinels in the different lithological types studied. These diagrams indicate that dunite formed in different environments and/or stages. Note, however, that olivine from harzburgite and type 3 dunite lenses with Cr-poor chromite exhibit quite similar chemical characteristics.

Spinel supergroup minerals (Bosi et al., 2019) are significant indicators that allow us to distinguish the different host rocks and gain insight into possible geotectonic provenance. In Fig. 10, discrimination diagrams were chosen to characterize the spinel compositions; the respective analyses are given in Table 3. Figure 10c, a Cr# [100×Cr/(Cr+Al)] vs. Mg# [100×Mg/(Mg+Fe2++Mn)] diagram, documents a compositional trend with increasing Cr# and Mg# from type 3 dunite, via harzburgite and type 1 dunite enveloping chromitite, to type 2 dunite. Magnesiochromite in type 2 dunite lenses, both in the matrix and as inclusions in olivine, is typically small and euhedral and shows Cr# between 66 and 67 and Mg# between 52 and 54, whereas spinel. in the type 3 dunite lenses yields Cr# between 19 and 20 and Mg# between 72 and 78. The diagram also documents that only the spinel supergroup minerals of the type 2 dunites are magnesiochromite, whereas those of the three other lithological types are spinel. The analyses of magnesiochromite from the type 2 dunite lenses fall into the compositional field of ophiolitic chromitites (Kamenetsky et al., 2001) and are located between the two fields of abyssal and forearc peridotites (Fig. 10c).

The Cr# vs. TiO2 diagram (Fig.10b) documents low amounts of TiO2 (<0.1 wt %) in spinel supergroup minerals of the type 1 dunite that envelopes chromitite and in the harzburgite; the respective Cr# is about 50 and 30. The type 3 and type 2 dunite lenses contain spinel supergroup minerals with TiO2 contents between 0.13 and 0.23. The correlation between the Cr# of spinel supergroup minerals and the Fo content of the adjacent olivine reveals a compositional trend within the olivine–spinel mantle array (OSMA; Arai, 1994; Fig. 10a).

Clinopyroxene is essentially diopsidic with compositions between 31.2 and 50.1 mol % En end-member (Table 4). Correlation diagrams for selected oxides are shown in Fig. 11.

In general, the clinopyroxenes in the harzburgite have the lowest Mg#, between 90.7 and 93.0. In the type 1 dunite that envelopes chromitite lenses, the Mg# is between 91.7 and 93.7. In the type 3 dunite lenses the Mg# is similar (95.3–96.0), but in the type 2 dunite lenses Mg# is considerably higher and varies between 91.9 and 93.0. The Al2O3 content of clinopyroxene is variable, lying in the range of 4.28 wt %–5.45 wt % in the harzburgites, 1.56 wt %–3.29 wt % in the type 1 dunite enveloping chromitite, 0.78 wt %–0.86 wt % in the type 2 dunite lenses and 3.48 wt %–5.24 wt % in the type 3 dunite lenses.

The contents of Cr2O3 in clinopyroxene of the peridotites range from 0.1 wt % to 1.32 wt %, and Na2O contents are below 1.0 wt %. The clinopyroxenes of harzburgites show the highest amounts of Cr2O3, Na2O and TiO2 (0.16 wt %–0.25 wt %) compared to the remaining peridotites. The amount of TiO2 in clinopyroxene is also relatively high in the type 3 dunite lenses (0.12–0.19), whereas it is significantly lower in the type 1 dunite enveloping chromitite and in the type 2 dunite.

The bulk rock characteristics and Li isotope ratios of the different types of dunite from the Purang ophiolite, as well as their corresponding mineral compositions, are quite different. The type 3 dunite lenses contain chromian spinel with the lowest Cr# observed, which may have experienced interaction with an aluminum-rich melt. In addition, exsolution textures were produced in type 3 dunite during uplift (Xiong et al., 2016; Fig. 10a). Note that the composition of chromian spinel and olivine from the type 3 dunite is similar to that in abyssal peridotite (Zhou et al., 2005; Arai and Miura, 2016). Note also that the type 2 dunite lenses and the type 1 dunite enveloping chromitite differ from the type 3 dunite. The type 1 dunites were possibly more intensely affected by an interaction with melt. A similar behavior was reported by Zhou et al. (2005), who suggested that the melt in a supra-subduction zone setting would be of a boninite composition, whereas the type 1 dunite envelope is considered to be formed by the reaction of a subduction-related melt with lherzolite or wehrlite (Zhou et al., 2005; Xiong et al., 2015; Zhang et al., 2019a, b). In an olivine vs. chromian spinel mantle-array diagram (Arai, 1994), the type 3 dunites and the harzburgites plot entirely within the area defined by abyssal peridotite, with approximately 8 % and 10 % partial melting (Fig. 10a), indicating MOR peridotite. In contrast, type 1 dunite envelopes and type 2 dunites, which can be estimated to have experienced about 20 % and 35 % partial melting, respectively, show compositional features similar to subduction zone (SSZ) peridotite (Fig. 10a), which may be the result of a multistage development.

The formation of high-Cr chromitites, dunites and refractory harzburgites is thought to result from varying degrees of reaction between a boninite-like melt and Cpx-bearing harzburgite (Zhou et al., 2005). In addition, most of the dunite envelopes are found only around disseminated chromitites, not the massive varieties (Fig. 2b). The prevailing view that podiform chromitites form only in SSZ settings by reaction between hydrous boninitic melts and mantle peridotites in the shallow mantle (Zhou et al., 1996, 2005; Uysal et al., 2005; Xiong et al., 2015) is based on the study of large chromitite bodies.

Kelemen et al. (1990) provided reaction experiments between a melt of olivine tholeiite and harzburgite at 5 kbar and 1050–1150 ∘C and demonstrated that the amount of SiO2 in the solid phase gradually increases with the consumption of the harzburgite. The Mg/(Mg+Fe) ratio in the solid phase also decreases until the harzburgite is finally consumed. Their study showed that a similar process might be expected during interaction with depleted mantle peridotites, e.g., when an arc of magma rises through the mantle lithosphere of the overriding plate. This process involves extensive reaction events between melt, olivine, orthopyroxene and spinel. Sakae and Jun-Ichi (2007) pointed out that through this boninitic melt–rock reaction a large number of different dunite lenses can be produced in a peridotite massif in New Zealand. The type 2 dunite lenses with Cr-rich spinel form intercalations within harzburgite, and the compositions of olivine, clinopyroxene and chromian spinel show gradual changes from harzburgite. Therefore, the type 2 dunite lenses may be the remnants of a mantle peridotite after high partial melting, and the type 3 dunite with Cr-poor spinel could result from an in situ reaction between an aluminum-enriched melt and peridotite. The type 1 dunite envelopes are probably a result of a melt–peridotite reaction.

Several authors have described clinopyroxene and magnetite intergrowth structures in olivine and interpreted them as exsolution features (e.g., Arai, 1978; Moseley, 1984; Xiong et al., 2017; Liang et al., 2014, 2018). Ashworth and Chambers (2000) and Ren et al. (2008), however, interpreted them as symplectites. The structures observed in the Purang dunites of the present study are here interpreted as exsolution lamellae. Exsolution of magnetite from olivine could presumably represent either a decomposition product of a solid solution between olivine and laihunite [Fe2+Fe23+(SiO4)2] or an oxidization product of Fe2+ in the olivine (Moseley, 1984).

Olivine accommodates extremely small quantities of Ca2+ and Na+ ions in its crystal lattice due to the larger ionic radii of these elements compared to those of Fe2+, Mg2+ and Mn2+. Olivine is thus not expected to exsolve clinopyroxene. Experimental studies on the olivine–melt system show that the higher the contents of CaO, Na2O and FeO in the melt, the higher the Ca content in the olivine (Libourel, 1999). In this study, olivine grains with exsolution microstructures point to a relatively high CaO content prior to exsolution (Fig. 4). The CaO content in the olivine cannot indicate fractional crystallization, nor can it be due to to the effects of oxygen fugacity, temperature, and pressure (Libourel, 1999). Thus, the variation in CaO content of olivine in the dunite from the Purang ophiolite (Fig. 4) may be related to later modification by melt or fluid. The clinopyroxene in the rod-like intergrowths of magnetite in the type 3 dunite olivine grains is aligned in the same direction as the needle-like intergrowths (Figs. 4, 5), indicating their simultaneous exsolution. It has been documented that clinopyroxene can crystallize from hydrothermal fluids at temperatures higher than 750 ∘C (Python et al., 2007; Akizawa et al., 2011; Akizawa and Arai, 2014). The association of clinopyroxene and magnetite and the presence of interstitial clinopyroxene in the Purang ophiolite suggest the involvement of hydrothermal fluids during the formation of the dunite. Hence, the reactant fluids could have originated from hydrothermal fluids. Re-equilibration between olivine and fluids at high temperatures could produce solid solutions with Ca2+ and Na+ in olivine. This feature has been well documented in dunites of the Luobusa ophiolite in the eastern part of the Yarlung–Zangbo suture zone (Ren et al., 2008). Because of the rapid uplift and cooling of the Purang peridotite (Liu et al., 2014), the exsolution lamellae in the olivine grains do not show plastic deformation; thus, their primary exsolution features are well preserved.

The isotopes of Li, 6Li and 7Li can be easily fractionated because of their large mass difference (∼15 %), resulting in large δ7Li variation in the terrestrial environment. For example, the δ7Li value of seawater is ∼+32 ‰ and that of N-MORB and MORB sources ranges from ∼+1.5 ‰ to +6.5 ‰, whereas altered MORB has a larger range from -2 ‰ to +14 ‰ (Chan et al., 1999; Decitre et al., 2002; Lundstrom et al., 2005; Vils et al., 2008). During subduction, Li in subducted slabs can be easily mobilized and carried into the overlying mantle wedge by fluids (Ishikawa et al., 2005; Savov et al., 2007). Li isotopic signatures developed in the crust and mantle can be used to study mantle heterogeneity. Abyssal peridotites have Li concentrations of <3 ppm, which are lower than those of rocks formed by a melt–peridotite reaction. Based on the dissolution of olivine during the process of partial melting, Li isotopic diversities of the different peridotites studied can be seen as the result of different amounts of melt extraction and high-temperature hydrothermal alteration (Su et al., 2016, 2017; Xiao et al., 2017) (Fig. 8).

The different types of olivine-rich rocks discovered in the Purang ophiolite formed due to different mechanisms. In olivine the Fo end-member content correlates with the different Li isotope characteristics in the various peridotites (Fig. 9). The Purang peridotites have heavy Li isotopes with high δ7Li values of 6.19–14.72, which are similar to arc lavas but also overlap with values found in metamorphic products such as eclogites and granulites, as well as altered MORBs and marine sediments (Fig. 8). Note that the type 1 dunite enveloping chromitite has higher amounts of Li than the harzburgite but the lowest values of δ7Li. Since Li is highly mobile, it is sensitive to low-temperature fluid–rock interaction (Chan et al., 2002). Lithium and (preferentially) 6Li move from olivine grains to the hydrothermal fluid during serpentinization, and as a result olivine becomes Li-depleted and isotopically heavier in Li (Decitre et al., 2002; Lundstrom et al., 2005; Wimpenny et al., 2010; Wunder et al., 2010; Su et al., 2016). In addition, crustal contamination may lead to an increase in the Li content in the magma and a change of its 7Li isotopic composition, especially during magma ascent and ophiolite emplacement (Teng et al., 2004). The Purang peridotite is high in MgO and rare-earth elements (REE) (Li et al., 2015; Xiong et al., 2018), and the Li content in olivine is very low (Table 2). This means that the process of crustal contamination is negligible. Teng et al. (2011), Xiao et al. (2016) and Bai et al. (2017) demonstrated the possibility of an intense exchange of chemical components between olivine and chromian spinel. It is widely accepted that Fe diffuses from olivine to chromite as the temperature decreases (Mg does the opposite), thus increasing the Fo component of olivine (Su et al., 2015; Xiong et al., 2015; Bai et al., 2017). Previous studies have suggested (e.g., Lundstrom et al., 2005; Tomascak et al., 2016) that because 6Li diffuses slightly faster than 7Li inside minerals, a negative anomaly of about 10 ‰–20 ‰ should be produced in olivine. This diffusion mechanism may explain the slightly negative correlation between Li and δ7Li observed in the Purang olivines of the different ultrabasic rocks (Fig. 8), but it does not explain the variability of the entire dataset of 10 ‰ of δ7Li difference and its correlation with magma differentiation indicators between the different dunites. The decreased δ7Li trend of olivine in dunite was explained as a result of Li diffusion from melt in a melt conduit to the host peridotite during melt–harzburgite interaction (Su et al., 2016).

Chemically or isotopically driven intragranular diffusion may affect the Li concentration of a mineral, depending on whether its growth occurred from a fluid or a melt (Tomascak et al., 2016). Even if the Li isotopes take more time to achieve compositional uniformity between melt and crystals than the Li abundance (Richter et al., 2014), the Li content in Purang peridotites is similar, whereas the δ7Li values have some unclear variation (Fig. 9). The δ7Li values of forsterite of the type 2 dunite lenses and the type 1 dunite enveloping chromitite are lower than those of the harzburgite and the type 3 dunite lenses. The same features are evident in NiO and Fo of olivine and Cr# of spinel, respectively, which show a negative correlation with the δ7Li values (Fig. 9).

Lithium is a moderately incompatible element during mantle melting processes and magma differentiation (Brenan et al., 1998). The concentration of Li in the magma increases as the degree of differentiation increases (e.g., Hamelin et al., 2009; Marks et al., 2007; Weyer and Seitz, 2012; Ackerman et al., 2015). The Purang olivine shows a tendency towards increased Li with a decrease in Fo component (Fig. 9). In addition, the Li content of olivine is negatively correlated with NiO and also negatively correlated with Cr# in spinel supergroup minerals(Fig. 9). These trends are consistent with trends that result from magma differentiation. Marks et al. (2007) found evidence that olivine fractionation drives the derived magma to move in the direction of high Li and that this elevated Li trend is thought to be related to the redox conditions of the magma. The diffusion process alone does not explain the preferred enrichment of δ7Li in the melt relative to olivine phenocrysts. The Purang peridotite acts as a quasi-stable aquifer system that can carry large amounts of fluids that were introduced during the SSZ subduction (Xiong et al., 2018). A large number of studies have demonstrated that Li, in general, is an element that can easily transport via a fluid phase (Brenan et al., 1998; Su et al., 2016, 2018). With respect to questions related to isotope fractionation processes, Chan et al. (1999) were able to show that 7Li becomes preferentially enriched in the fluid phase. Su et al. (2015, 2016) have demonstrated that during the subduction of the basaltic oceanic crust and overlying sediments, fluids may become released and percolate through the overlying mantle wedge; magma that is generated in such an environment can thus incorporate the Li isotope characteristics of these fluids.

The Purang ophiolite is one of the numerous ophiolite bodies located along the Yarlung–Zangbo suture zone in southern Tibet, which are considered to be either remnants of the Neo-Tethyan lithosphere emplaced during the collision between the Indian and Asian plates or as ancient subcontinental lithospheric mantle domains (Gong et al., 2016). It consists chiefly of mantle peridotites locally accompanied by gabbros, dikes and solidified lavas. The solidified lavas range from arc tholeiites to boninitic and OIB-like basalts. Most of the peridotites have “U-shaped”, chondrite-normalized REE patterns, which indicate the re-fertilization of depleted peridotites in a supra-subduction zone environment (Xiong et al., 2018).

However, the presence of both type 2 and type 3 dunite pods in the Purang peridotites and the presence of magnetite + clinopyroxene exsolution lamellae in olivine of the type 3 dunite provide important clues to the tectonic environment in which this ophiolite formed. There is strong evidence for a multistage origin of the Purang peridotites. After having reached relatively shallow levels in the upper mantle, the peridotites were trapped in a supra-subduction zone wedge, where they underwent interaction with SSZ melts and fluids (Xiong et al., 2019c). Initially, the hydrous melting of the peridotite produced Mg-rich magmas, which reacted with the peridotites to form dunite pods characterized by Mg-rich olivine and high-Cr spinel. In some ophiolites, such as Luobusa, the dunite pods are closely associated with podiform chromitites. A second stage of melting in the Purang peridotites produced boninitic magmas via lower degrees of partial melting, which were relatively rich in Ti4+, Al3+, Ca2+, Na1+ and Fe3+. During the reaction of these melts with the peridotites, small amounts of these elements may have become incorporated into the newly crystallized olivine. During subsequent cooling, they very likely became exsolved as lamellae of magnetite (containing Fe3+, Ti4+) and clinopyroxene (containing Al3+, Ca2+ and Na1+). Thus, our preferred explanation for the origin and evolution of the Purang ophiolite is that the mantle peridotites formed in a MOR environment and were then modified by later-stage melts and fluids in a SSZ setting during intra-oceanic subduction along the Indus–Yarlung–Zangbo suture (see Zhou et al., 1996, 2005; Malpas et al., 2003; Hébert et al., 2003; Dubois-Côté et al., 2005; Bédard et al., 2009; Liu et al., 2010; Xu et al., 2011; Xiong et al., 2013; Su et al., 2019). The discovery of a large number of water-bearing minerals in the Purang chromitite is a further indicator of fluid exchange processes (Xiong et al., 2013, 2018). The above-described characteristics and discussed processes for the Purang ophiolite are consistent with the geochemical behavior of Li isotopes in a subduction-zone setting, as described by Marschall et al. (2007), Wunder et al. (2006) and Yamaji et al. (2001), who found evidence for a progressive release of fluid from the subducted plate that migrated into the overlying mantle wedge and caused a lowering of δ7Li.