Reconsideration of Neo-Tethys evolution constrained from the nature of the Dazhuqu ophiolitic mantle, southern Tibet

Abstract

The nature (i.e., sub-oceanic, sub-arc or sub-continental) of ophiolitic mantle peridotites from the eastern Neo-Tethyan domain in southern Tibet has been hotly debated. This uncertainty limits our understanding of the history and evolution of the eastern Neo-Tethys Ocean. Here we present petrological, geochemical and Re–Os isotopic data for the mantle peridotites from the Dazhuqu ophiolite in the central segment of the Yarlung Zangbo suture zone, southern Tibet. Samples collected include both spinel lherzolites and spinel harzburgites. The lherzolites have spinel Cr# [Cr/(Cr + Al), ~ 0.3–0.4] comparable to those of typical abyssal peridotites. In contrast, the harzburgites have spinel Cr# (~ 0.3–0.7) overlapping with the ranges of both abyssal and fore-arc peridotites; two samples have spinel Cr# higher than 0.6, which is probably ascribed to intense melt–rock interactions. Clinopyroxene trace element modeling indicates that the Dazhuqu mantle peridotites have experienced 0–6% garnet-facies melting followed by 10–18% melting in the spinel stability field. This is similar to the degree of garnet-facies melting inferred for many abyssal peridotites and implies deep initial melting (> 85 km), which distinguishes the Dazhuqu mantle peridotites from fore-arc peridotites (commonly < 80 km in origin). The Dazhuqu peridotites have unradiogenic ¹⁸⁷Os/¹⁸⁸Os of 0.11836–0.12922, which are commonly lower than the recommended value of primitive upper mantle (PUM). All but one samples yield relatively younger Re depletion ages (T_RD = 0.06–0.81 Ga) with respect to the only one sample having an older T_RD age of 1.66 Ga. Re–Os isotopes and highly siderophile element (HSE) compositions of the Dazhuqu peridotites are similar to those of abyssal peridotites and the Oman southern massifs but are distinct from non-cratonic sub-continental lithospheric mantle (SCLM) xenoliths and sub-arc mantle. We emphasize the similarity between the Dazhuqu and Oman ophiolites, both representing Neo-Tethyan oceanic lithosphere and implying ridge–trench collision.

Introduction

A Wilson cycle describes the tectonic evolution of ocean basins from continental rifting, through ocean opening, subduction to closure (Wilson 1968). Consequently, much of the history of a disappeared ocean is “hidden” in ophiolites, which are regarded as fragments of oceanic lithosphere emplaced onto continents (Gass 1968; Moores and Vine 1971). In the past four decades, however, numerous studies have demonstrated, based on geochemical fingerprinting, that a significant proportion of ophiolites worldwide were generated at subduction zones rather than at mid-ocean ridges (Alabaster et al. 1982; Miyashiro 1973; Pearce et al. 1984; Pearce and Robinson 2010; Whattam and Stern 2011). The former has been classified as supra-subduction zone (SSZ) ophiolites, whereas the latter has been termed mid-ocean ridge (MOR) ophiolites (Alabaster et al. 1982). A SSZ ophiolite is newly born as a consequence of oceanic crustal accretion mostly in a fore-arc basin and may mark major subduction initiation processes (Dilek and Furnes 2011; Whattam and Stern 2011). Based on this hypothesis, peridotites within certain ophiolites are often interpreted as juvenile fore-arc mantle (sub-arc mantle), rather than as abyssal peridotites beneath MOR (sub-oceanic mantle). On the other hand, ancient (> 1.0 Ga) osmium isotope signatures indicate in one aspect a sub-continental lithospheric mantle (SCLM) affinity for some ophiolitic peridotites, e.g., the Alpine–Apennine ophiolites [ages were recalculated from the osmium data reported by Alard et al. (2005), Snow et al. (2000) and van Acken et al. (2008)]. An extensively accepted interpretation is that such SCLM-type peridotites were exhumed and emplaced close to the rifted continental margins and ocean–continent transition (OCT) settings (e.g., McCarthy and Müntener 2015; Müntener and Manatschal 2006; Rampone et al. 1998; Tribuzio et al. 2004). These observations diminish the direct linkage between ophiolites and disappeared oceanic lithospheres.

The Neo-Tethys Ocean is an ocean that existed between the continents of Gondwana and Eurasia during much of the Mesozoic (Gehrels et al. 2011; Metcalfe 2006; Sengor and Natalin 1996). The eastern part of Neo-Tethys Ocean closed along the Yarlung Zangbo suture zone (YZSZ) in southern Tibet (Dewey et al. 1988; Pan et al. 2012; Yin and Harrison 2000; Zhu et al. 2013). The best geologic record of this ancient ocean observed to date is the E–W trending Yarlung Zangbo ophiolites, which have been regarded as peculiar fragments of oceanic lithosphere emplaced onto the Indian plate (Girardeau et al. 1985a, b; Nicolas et al. 1981). Recent studies, however, have argued that the Yarlung Zangbo ophiolites formed in a fore-arc (Dai et al. 2011, 2013; Xia et al. 2003; Xiong et al. 2016, 2017) or back-arc basin (Hébert et al. 2012 and references therein) and could not represent the “true” Neo-Tethyan oceanic lithosphere. A bone of contention is the nature of the mantle peridotites within these ophiolites. Geochemical observations revealed that some mantle massifs along the YZSZ experienced high degrees of partial melting and displayed U-shaped whole-rock rare earth element (REE) patterns (Dai et al. 2011, 2013; Xia et al. 2003; Xiong et al. 2017; Zhou et al. 2005). The two features were ascribed to an interaction between depleted mantle peridotites and boninitic melts (Dai et al. 2011, 2013; Zhou et al. 2005) and hence support a fore-arc tectonic setting. However, these arguments are confusing because abyssal peridotites at modern ocean ridges also sometimes show U-shaped or spoon-like whole-rock REE patterns (e.g., Day et al. 2017). On the other hand, Gong et al. (2016) and Griffin et al. (2016) argued that ancient mantle signatures in the Yarlung Zangbo ophiolites, along with the relatively high Na but low Nd contents in clinopyroxenes, provide evidence for a SCLM affinity, as is the case for the Alpine–Apennine ophiolites (Alard et al. 2005; Müntener et al. 2010; Snow et al. 2000; van Acken et al. 2008). The mantle massif examined in previous studies (i.e., the Purang mantle peridotites from the western YZSZ; Gong et al. 2016), however, displays Os isotopic ratios that are consistent with those of certain abyssal peridotites (e.g., Day et al. 2017; Harvey et al. 2006; Lassiter et al. 2014; Liu et al. 2008). These complex results underscore the need for a systematic Re–Os study of the Yarlung Zangbo ophiolitic mantle. Definition of the nature (sub-oceanic, sub-continental or sub-arc) of these mantle rocks, based on systematic observations of their petrological and chemical features, is absolutely critical to decipher the origin of the Yarlung Zangbo ophiolites and hence to reconstruct the history of the Neo-Tethys Ocean.

In this study, we present mineral, petrological and geochemical compositions of mantle peridotites collected from the Dazhuqu ophiolite in the central segment of the YZSZ. These results suggest that the Dazhuqu mantle peridotites are of sub-oceanic affinity, rather than representing SCLM or sub-arc mantle. They are suitable for establishing the link between the Yarlung Zangbo ophiolites and the Neo-Tethyan oceanic lithosphere, and thus advancing our understanding of the evolutionary history of the Neo-Tethys Ocean.

Geological framework and sample description

Geological framework

The Tibetan plateau is a collage of several tectonostratigraphic terranes (Fig. 1a), which are separated by different E–W trending sutures (Dewey et al. 1988; Pan et al. 2012; Sengor and Natalin 1996; Yin and Harrison 2000; Zhu et al. 2013). The YZSZ is the southernmost suture (Fig. 1a, b) and has been regarded as the site where the Neo-Tethys Ocean closed (Yin and Harrison 2000; Zhu et al. 2013). Ophiolites in the YZSZ extend discontinuously about 2000 km from the Nanga Parbat syntaxis in the west to the Namche Barwa syntaxis in the east (Hébert et al. 2012). Traditionally, the Yarlung Zangbo ophiolites have been divided into three segments from west to east. Ophiolites in the central segment are also termed the Xigaze ophiolite, including ophiolitic massifs in Ngamring, Jiding, Luqu and Dazhuqu (Bao et al. 2013). They were sandwiched between the Xigaze fore-arc basin (XFB) sedimentary rocks to the north and the Triassic–Cretaceous accretionary mélanges to the south (Fig. 1b). The base of the XFB sediments is the Chongdui Formation, which is a tuffaceous chert section with interbedded tuff layers of 119–113 Ma (Wang et al. 2017).

Fig. 1

a Simplified geological map of the Tibetan Plateau showing the major terranes and sutures; modified after Wu et al. (2008). KSZ Kunlun suture zone, JSZ Jinsha suture zone, BNSZ Bangong–Nujiang suture zone, YZSZ Yarlung Zangbo suture zone. b Geological map of southern Tibet illustrating the distribution of the Yarlung Zangbo ophiolites; modified after Dai et al. (2011). c Simplified geological map of the Dazhuqu ophiolite, modified after Wang et al. (1987)

Overall, the Xigaze ophiolite has lithological assemblages that are quite distinct from the ideal “Penrose-type” ophiolite sequence (Nicolas et al. 1981). Mantle peridotites from the Dazhuqu (also termed Lianxiang or West Dazhuka) massif in the eastern part of the Xigaze ophiolite have a maximum thickness of about 4.5 km and are overlain by ~ 3 km thick crustal rocks (Fig. 1c), including gabbros, diabases and basalts (Girardeau et al. 1985b). Geochronological studies of the crustal rocks indicate that the Dazhuqu ophiolite was generated during 124–128 Ma (Dai et al. 2013; Liu et al. 2016; Malpas et al. 2003), which is coeval with the age of radiolarian cherts overlying the ophiolite (Upper Barremian; Ziabrev et al. 2003). Gabbros from the Dazhuqu ophiolite occur as layered intrusions of hundreds of meters in thickness or as meter-scale veins intruding into the mantle section. Both uppermost mantle and gabbro sections are cross-cut ubiquitously by diabase dikes or sills. Diabases proportionally increase upwards and transfer gradationally into a sheeted sill complex, with a strike parallel to the main foliation of the peridotites and to the major lithostratigraphic unit boundaries (Girardeau et al. 1985a, b). The sheeted sill complex is overlain by pillow and massive basaltic lavas, which in turn are commonly in thrust contact with the XFB sedimentary rocks. These geological features have been used to suggest that the Dazhuqu ophiolite was generated at a slow- to ultraslow-spreading MOR setting (Girardeau et al. 1985a, b; Nicolas et al. 1981; Liu et al. 2016; Wang et al. 1987; Wu et al. 2014).

Sample description

Mantle peridotites from the Dazhuqu ophiolite has commonly been subjected to intense alteration, but remarkably fresh peridotites occur as domes within the serpentinized peridotites (Fig. 2a, b). They mainly consist of harzburgites with volumetrically subordinate lherzolites (Fig. 2c). The field relationship between harzburgites and lherzolites is, however, obscure. Occasionally, irregular dunite patches (1–10 m) occur within the harzburgites (Fig. 2d). Some of the fresh peridotites commonly display clear primary high-temperature foliations. Occasionally, the mantle peridotites are cross-cut by different pyroxenite veins, including clinopyroxenites, orthopyroxenites and websterites (Fig. 2e). Northwards, the uppermost part of the mantle section is commonly intruded by diabase, rodingite and gabbroic dikes or sills (Fig. 2f).

Fig. 2

Field photos of the Dazhuqu mantle peridotites. a Fresh mantle peridotites having a dome-like shape included by serpentinized peridotites. b Outcrops of fresh Dazhuqu mantle peridotites. c Dazhuqu lherzolites. d Dunites. e Fresh Dazhuqu mantle peridotites cross-cut by pyroxenite veins. f Intensely serpentinized peridotites from Dazhuqu intruded by rodingite dikes

Thirty-three peridotites from the Dazhuqu massif have been studied (GPS: N 29°16.41′, E 89°31.66′). Among them, 21 samples are spinel harzburgites, which have clinopyroxene modal compositions varying from 0.7 to 4.9%, and the rest are spinel lherzolites with clinopyroxene modal compositions of 5.1–7.3% (Table 1; modal composition estimates are based on systematic petrographic observations, and on combined calculations using whole-rock and mineral major element compositions). Spinel harzburgites are composed of fresh olivine, orthopyroxene, clinopyroxene and spinel. Three out of the 21 harzburgites show a protogranular texture, in which olivine and orthopyroxene display curvilinear and polygonal shapes and are also locally recrystallized and show triple junctions (Fig. 3a). 18 of the Dazhuqu harzburgite samples display a porphyroclastic texture, in which both olivine and orthopyroxene porphyroclasts are surrounded by fine-grained neoblasts (Fig. 3b). Thin (< 10 µm) exsolution lamellae of clinopyroxenes are widely developed in orthopyroxene porphyroclasts (Fig. 3b), indicating Ca saturation over the major cooling path. This is consistent with a closure temperature of 867–1139 °C based on the calculation using the Ca-in-orthopyroxene geothermometer proposed by Brey and Köhler (1990) (at a given pressure of 15 kbar).

Table 1 Modal compositions of the Dazhuqu mantle peridotites

Fig. 3

Photomicrographs and BSE images of the Dazhuqu mantle peridotites. a The Dazhuqu harzburgites show a protogranular texture, in which olivine and orthopyroxene display a curvilinear and polygonal shape. Some olivine and orthopyroxene are locally recrystallized and show triple junctions. b The Dazhuqu harzubrgites display a porphyroclastic texture, in which orthopyroxene porphyroclasts are surrounded by fine-grained neoblasts. Exsolution lamellae of clinopyroxenes are widely developed in orthopyroxene porphyroclasts. c Clinopyroxenes from the Dazhuqu lherzolties display a porphyroclastic texture. d Brownish spinels of the Dazhuqu harzburgites commonly have a holly-leaf shape. e The Dazhuqu lherzolites typically contain a few amounts of subhedral sulfides (mainly pentlandite). f Sulfides are rare in the Dazhuqu harzburgites, which only contain a few grains of Ni-rich alloys included in olivines or orthopyroxenes. Abbreviations are as follows: Ol olivine, Opx orthopyroxene, Cpx clinopyroxene, Sp spinel, Pn pentlandite

Clinopyroxenes from the Dazhuqu harzburgites commonly occur as small grains (0.1–0.2 mm), while those from lherzolite samples developed into porphyroclasts with a grain size of ~ 2 mm (Fig. 3c). Brownish spinels are ubiquitous in both the Dazhuqu harzburgites and lherzolites and commonly show a holly-leaf shape (Fig. 3d). They occasionally occur as inclusions within orthopyroxenes and olivines. The lherzolites and harzburgites show different sulfide petrographic features, with a few (< 1%) subhedral sulfides (mainly pentlandite) only occur in the former that are rare in the latter (Fig. 3e, f).

Analytical methods

Fresh Dazhuqu peridotite samples were cut to make thin-sections for petrographic observations and mineral major and trace element analyses. The residual samples were then crushed to powders with sizes of ~ 200 mesh for whole-rock major element, highly siderophile element (HSE) and Re–Os isotope measurements. All analyses conducted in this study were done at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS) in Beijing, China.

Whole-rock and mineral major element measurements

0.5 g sample powders were precisely weighted and then mixed with 5 g Li₂B₄O₇ to make fused glass discs for whole-rock major element analysis using an AXIOS Minerals X-ray fluorescence (XRF) spectrometer. Loss on ignition (LOI) was determined by the measurement of the relative weight loss of 0.5 g sample powder after heating in an oven at ~ 1000 °C for 1.5 h. The 95% confidence limits for XRF analysis are ± 0.04% to ± 0.05% for SiO₂ and Al₂O₃, ± 0.01% to ± 0.02% for Fe₂O₃, MgO, Na₂O and K₂O and < ± 0.01% for TiO₂, MnO, CaO and P₂O₅, respectively, according to triplicate measurements of the USGS standard GSR-1 in the lab.

Major element compositions of minerals were determined by electron probe micro-analysis (EPMA) using a JEOL JXA8100 instrument. During analysis, quantitative measurement was carried out using wavelength-dispersive spectrometers (WDS) with an accelerating voltage of 15 kV and a beam current of 20 nA. The peak and background counting time were 20 s and 10 s, respectively, for each spot. Natural minerals and synthetic oxides were used as standards, including albite for Na, diopside for Si, Ca and Mg, haematite for Fe, orthoclase for K, and synthetic Cr₂O₃, TiO₂, Al₂O₃, MnO and NiO for Cr, Ti, Al, Mn and Ni, respectively. All raw data have been processed an on-line correction using a JEOL-modified ZAF (atomic number, absorption, fluorescence) procedure.

Clinopyroxene trace element determination

Clinopyroxene trace element concentrations were measured by laser ablation on an Agilent 7500a inductively coupled plasma–mass spectrometer (ICP–MS) coupled with a 193 nm Coherent COMPex Pro ArF Excimer laser. Isotopes were measured in a peak-hopping mode, using a laser diameter of ca. 120 µm and a repetition rate of 8 Hz. The NIST 610 glass was used as an external standard for calibration and the NIST 612 standard was used for monitoring signal drift. The results of NIST 610 are listed in Online Resource Table S2 and are consistent with the recommended values of Pearce et al. (1997) for all elements analyzed. The relative error for most elements between the measured and recommended values of NIST 610 is very low, with a highest value of 0.74% for U. Calcium (⁴³Ca) was used as an internal standard and the raw data were reduced using the GLITTER 4.0 program (Griffin et al. 2008).

HSE and Re–Os isotope analyses

The HSE contents and Re–Os isotopes were determined by isotope dilution (ID). The detailed procedures have been described by Chu et al. (2009). About 2 g sample powder and appropriate amounts of a ¹⁸⁷Re–¹⁹⁰Os mixed spike and a ¹⁹¹Ir–⁹⁹Ru–¹⁹⁴Pt–¹⁰⁵Pd mixed spike were digested with 3 ml purified concentrated HCl and 6 ml purified concentrated 16 N HNO₃ in a Pyrex^® borosilicate glass Carius Tube. After digestion at ~ 220 °C for about 72 h, Os was extracted from the aqua regia solution into CCl₄ and then back-extracted into HBr followed by purification via micro-distillation (Birck et al. 1997). Rhenium, Ir, Ru, Pt and Pd were separated and purified by exchange chromatography using a 2 ml resin (AG 1 × 8, 100–200 mesh) (Walker et al. 2008) for further elution. The Re–Ru, Ir–Pt and Pd fractions were eluted with 6 N HNO₃, 13.5 N HNO₃ and 10 N HCl, respectively. The Re fraction was then purified using a small-size anion exchange column with ~ 0.1 ml resin. Each fraction of the Re, Ir, Ru, Pt and Pd was heated to dryness and re-dissolved in 1 ml 0.8N HNO₃ for analysis by ICP–MS.

Osmium isotopic compositions were measured at the thermal ionization mass spectrometer (TIMS) Laboratory using a GV Isoprobe-T mass spectrometer. During measurement, Os was loaded onto platinum filaments and Ba(OH)₂ was used as an ion emitter. Os isotopic ratios were corrected for oxygen isotope fractionation using the standard oxygen isotope corrections of Nier and online mass fractionation using ¹⁹²Os/¹⁸⁸Os = 3.08271 and for off-line spike corrections. During analysis, a Johnson-Matthey standard (UMCP) yielded the ¹⁸⁷Os/¹⁸⁸Os ratio of 0.11378 ± 2 (2σ; σ = relative standard deviation; n = 4), which is consistent with the result (¹⁸⁷Os/¹⁸⁸Os = 0.113791 ± 15) analyzed by Shirey and Walker (1998). An international serpentinite standard UB-N was used as an external standard and had a measured Os concentration of 3.90 pg and ¹⁸⁷Os/¹⁸⁸Os ratio of 0.12690 ± 3 (2SE). The in-run precisions for Os isotopic ratios of all samples were better than ± 0.2% (2σ). The total procedural blanks were about 0.93–0.97 pg for Os (n = 2). Rhenium and other HSE (Ir, Ru, Pt and Pd) were measured at the MC–ICPMS Laboratory using a Thermo Fisher Neptune Plus mass spectrometry. The measured results of UB-N were 3.48 ng for Ir, 7.07 ng for Ru, 7.30 ng for Pt, 5.84 ng for Pd and 0.209 ng for Re, respectively. These values are consistent with the recommended values of UB-N from Meisel and Moser (2004) with Ir = 3.38 ng, Ru = 6.30 ng, Pt = 7.42 ng, Pd = 6.11 ng and Re = 0.206 ng. The total procedural blanks for Ir, Ru, Pt, Pd and Re were 4–10 pg, 18–28 pg, 24–29 pg, 33–37 pg and 2–4 pg, respectively. The in-run precisions for ¹⁸⁵Re/¹⁸⁷Re, ¹⁹¹Ir/¹⁹³Ir, ¹⁹⁴Pt/¹⁹⁶Pt, ¹⁰⁵Pd/¹⁰⁶Pd and ⁹⁹Ru/¹⁰¹Ru were commonly 0.1–0.3% (2RSD). Replicate analyses were conducted on sample 14DZQ12 to check the reproducibility, yielding relative differences of ~ 1% for ¹⁸⁷Os/¹⁸⁸Os ratio, of ~ 0.5% for Pd but of ~ 5–18% for other HSE, which are probably slightly influenced by chemical separation processes.

Results

Whole-rock major element compositions

Whole-rock major element compositions of the Dazhuqu mantle peridotites are listed in Table 2. Most of the studied samples are quite fresh with the LOI less than 1.1 wt.%, but two samples (14DZQ74 and 14DZQ76) have relatively higher LOI values of 1.8–3.2 wt.%. On the whole, these samples have homogeneous Mg# [atomic Mg/ (Mg + Fe)] values of ~ 0.92 (Table 2). The Dazhuqu harzburgites contain 41.8–45.5 wt.% MgO, 0.4–1.3 wt.% Al₂O₃, 0.5–1.3 wt.% CaO (Fig. 4) and ~ 0.01 wt.% TiO₂, but negligible Na₂O contents, which are comparable to the most depleted abyssal peridotites [data compilation by Niu (2004)] and the least refractory peridotites from the Izu–Bonin–Mariana (IBM) fore-arc basin (Parkinson and Pearce 1998). Compared to the harzburgites, the Dazhuqu lherzolites have lower MgO (41.8–44.0 wt.%) but higher Al₂O₃ (1.1–1.6 wt.%), CaO (1.4–1.9 wt.%) and TiO₂ (0.01–0.02 wt.%) contents. It is clear that the Dazhuqu lherzolites are less refractory than both the Dazhuqu harzburgites and the IBM fore-arc peridotites (Fig. 4). The Dazhuqu harzburgites and lherzolites both plot parallel to the modeled partial melting trends from Niu (1997). However, most of these samples show lower Al₂O₃ and CaO contents at a given MgO, compared to the compositions assumed by both polybaric near-fractional melting (25–8 kbar) and isobaric batch melting (20 kbar) (Niu 1997; Fig. 4).

Table 2 Whole-rock major element compositions (wt.%) of the Dazhuqu mantle peridotites

Fig. 4

Plots of whole-rock MgO against Al₂O₃ and CaO contents of the Dazhuqu mantle peridotites. The data plotted are recalculated to an anhydrous composition. Fields of the Izu–Bonin–Mariana (IBM) fore-arc refractory peridotites are after Parkinson and Pearce (1998). Data of abyssal peridotites are from the compilation by Niu (2004). The polybaric near-fractional (1% melt porosity) melting (with initial pressures from 25 kbar) and isobaric batch melting (at pressure = 20 kbar) curves are after Niu (1997)

Mineral chemistry

Mineral major element average compositions are listed in Online Resource Table S1. Olivines from the Dazhuqu harzburgites have forsterite contents [Fo; 100 × Mg/ (Mg + Fe)] of 90.3–91.3. Coexisting spinels have Cr# [Cr/(Cr + Al)] of ~ 0.3–0.7 and Mg# of ~ 0.5–0.7; overall, they contain TiO₂ contents less than 0.1 wt.%. The Dazhuqu lherzolites contain olivines and spinels with compositions overlapping with some of the harzburgites, but the olivine Fo contents (89.7–90.6) and spinel Cr# values (~ 0.3–0.4) of the lherzolites are both in relatively lower ranges (Fig. 5). In addition, the spinel Cr# of both the Dazhuqu harzburgites and lherzolites display a positive correlation with the olivine Fo contents but a negative correlation with the spinel Mg# (Fig. 5). These spinels share chemical characteristics overlapping with the ranges of both abyssal [see the review by Warren (2016)] and fore-arc peridotites (Parkinson and Pearce 1998; Pearce et al. 2000), with roughly equal numbers of samples plotting out of both fields (Fig. 5). In comparison, spinels from the harzburgites of the Luqu ophiolite (Zhang et al. 2017), which is located about 70 km west to Dazhuqu, have relatively higher Cr# values and plot all within the range of fore-arc peridotites, though most of them are within the range of abyssal peridotites (Fig. 5).

Fig. 5

Spinel Cr# versus olivine Fo (a) and spinel Mg# (b) of the Dazhuqu mantle peridotites. Fields of residual abyssal peridotite, abyssal dunite and veined peridotite are after the review in Warren (2016) and fields of fore-arc peridotite are from Parkinson and Pearce (1998) and Pearce et al. (2000). Data of the Luqu harzburgite and reactive dunite are from Zhang et al. (2017). The samples of 14DZQ78 and 14DZQ80 are the only two Dazhuqu samples in this study with spinel Cr# above 0.6

Clinopyroxenes from the Dazhuqu mantle peridotites have relatively low Na₂O contents of 0.04–0.47 wt.% (Fig. 6), which are in the lower end of the abyssal peridotite field (Warren 2016) but are not consistent with the range for continental peridotites (Hellebrand and Snow 2003). It should be noted that the correlation of the Na₂O contents in clinopyroxenes and Cr# values in spinels from a few Dazhuqu harzburgites clearly deviates from the modeled fractional melting trend from a Na-rich source (Fig. 6).

Fig. 6

Clinopyroxene Na₂O versus spinel Cr# of the Dazhuqu mantle peridotites. The fractional melting trend (with 1% melting increments) and the field of continental peridotite are according to the review in Hellebrand and Snow (2003); the field of abyssal peridotite is after the data compilation in Warren (2016)

Trace element concentrations of clinopyroxenes from the Dazhuqu mantle peridotites measured by in situ LA–ICPMS are listed in Online Resource Table S2. Overall, clinopyroxenes from these samples show chondritic to sub-chondritic trace element abundances (Fig. 7a). The REE abundances are much lower than those of depleted mid-ocean ridge basalt mantle (DMM) source and the average composition of abyssal peridotites (Warren 2016). Most of the clinopyroxenes from both the harzburgites and lherzolites display decreasing chondrite-normalized abundances from the heavy REE (HREE) through the middle REE (MREE) and the light REE (LREE), though the latter group in some of the harzburgites shows considerable variation (Fig. 7a). Furthermore, samples with relatively high LREE concentrations also display elevated Sr and Zr contents (Fig. 7b).

Fig. 7

Chondrite-normalized REE (a) and primitive mantle-normalized trace element (b) patterns of clinopyroxenes from the Dazhuqu mantle peridotites. The data are measured by in situ LA–ICPMS. REE of clinopyroxene after 5%, 10% and 15% fractional melting from a depleted MORB mantle (DMM) source (Workman and Hart 2005) is shown, using the modeled results described in Hellebrand et al. (2002). The field of whole-rock compositions of the Dazhuqu mantle peridotites is after Xia et al. (2003). The average composition of abyssal peridotites (Aver. AP) from Warren (2016) is shown for comparison. Chondrite and primitive mantle values are proposed by Sun and McDonough (1989)

HSE and Re–Os isotopic compositions

Whole-rock HSE and Re–Os isotopic compositions of 21 samples of the Dazhuqu mantle peridotites are shown in Table 3. All Dazhuqu samples have consistent and relatively flat IPGE (Ir-group platinum group elements; i.e., Os, Ir and Ru) patterns (Fig. 8) when normalized to CI chondrite (Horan et al. 2003). Their measured Os/Ir and Ru/Ir ratios (non-normalized) range from 0.7 to 1.6 and 0.9–1.6, respectively. Most Dazhuqu peridotites also have relatively flat patterns in PPGE (Pd-group PGE; i.e., Pt and Pd) and Re, but seven harzburgite samples show variable PPGE patterns and also moderate to strong depletion in Re (Fig. 8). In addition, two out of seven harzburgites have elevated Pt and particularly Pd abundances relative to the recommended values of primitive upper mantle (PUM; Meisel et al. 2001). As a whole, the measured Pt/Ir, Pd/Ir and Re/Ir ratios (non-normalized) of the Dazhuqu mantle peridotites range from 0.5 to 2.4, 0.5–3.6 and 0.04–1.0, respectively.

Table 3 Re–Os isotopes and HSE compositions of the Dazhuqu mantle peridotites

Fig. 8

CI chondrite-normalized highly siderophile element (HSE) patterns of the Dazhuqu mantle peridotites. Data of primitive upper mantle (PUM) and chondrite are from Meisel et al. (2001) and Horan et al. (2003), respectively. Data of the mantle peridotites from Oman southern massifs are from Hanghøj et al. (2010) and Lorand et al. (2009), and data of abyssal peridotites are after the compilation by Day et al. (2017) with three highly fractionated samples excluded. The field of HSE data of the Yarlung Zangbo ophiolites in the literature is after Dai et al. (2011) and references therein

The Dazhuqu mantle peridotites contain 2.55–5.64 ppb Os and 0.01–0.37 ppb Re (Table 3), yielding ¹⁸⁷Re/¹⁸⁸Os ratios of 0.01–0.35. They have ¹⁸⁷Os/¹⁸⁸Os ratios of 0.11836–0.12922 (Fig. 9a) that are similar to the values reported for mantle peridotites from the Purang and Dongbo ophiolites in the western segment of the YZSZ (Gong et al. 2016; Liu et al. 2012; Niu et al. 2015), but are lower than the recommended value (0.1296 ± 0.0008) of PUM (Meisel et al. 2001). In addition, a roughly positive covariation exists between ¹⁸⁷Re/¹⁸⁸Os and ¹⁸⁷Os/¹⁸⁸Os ratios for most Dazhuqu peridotites, but some harzburgite samples deviate, either above or below, from the correlation (Fig. 9a). The ¹⁸⁷Os/¹⁸⁸Os ratios do not show a correlation with whole-rock Al₂O₃ contents (Fig. 9b). Re depletion ages (T_RD) of the Dazhuqu peridotites calculated relative to PUM range from 0.06 to 1.66 Ga (Table 3).

Fig. 9

Re-Os isotopes of the Dazhuqu mantle peridotites. Data of primitive upper mantle (PUM) are from Meisel et al. (2001). Data of the Purang and Dongbo ophiolitic peridotites are from Gong et al. (2016), Liu et al. (2012) and Niu et al. (2015). Data of abyssal peridotites and of non-cratonic SCLM xenoliths are after the compilations by Day et al. (2017) and Luguet and Reisberg (2016), respectively. Dashed lines in Fig. 9b represent two rough positive correlations between ¹⁸⁷Os/¹⁸⁸Os ratio and whole-rock Al₂O₃ content of the non-cratonic SCLM xenoliths

Discussion

Garnet-facies initial melting and melt–rock interaction

The Dazhuqu mantle peridotites show variably refractory compositions that plot parallel to the modelled partial melting curves (Fig. 4), indicating that they must have experienced variable degrees of melting. However, the Dazhuqu samples in this study mostly contain lower Al₂O₃ and CaO at a given MgO and could not be perfectly modeled neither by polybaric fractional melting nor by isobaric batch melting (Niu 1997). Petrographic observations showed that orthopyroxene porphyroclasts from the Dazhuqu peridotites display extremely irregular outlines and are commonly surrounded by fine-grained olivine neoblasts (Fig. 3b), indicating orthopyroxene resorption and replacement by olivine. This feature has been widely interpreted as resulting from porous flow migration and interaction of peridotites with pyroxene-undersaturated melts (Rampone et al. 2008; Seyler et al. 2007). Therefore, the deviation from the partial melting trends could be ascribed to melt–rock interaction, resulting in the addition of olivine (Niu 1997).

Using the empirical equation [F = 10 × ln (Cr#_spinel) + 24, F here means the extent of melting in percent] established by Hellebrand et al. (2001), it can be estimated that the Dazhuqu peridotites have been subjected to 10–18% of fractional melting of a depleted mantle source. Similarly, in the diagram of spinel Cr# versus clinopyroxene Yb (Fig. 10a), the majority of the Dazhuqu samples plot close to the pure fractional melting trend (Liu et al. 2008) and yield 10–20% of partial melting. However, these samples are characterized by steeper MREE to HREE patterns (Fig. 7a) relative to the average composition of abyssal peridotites (Warren 2016) and the modelled fractional melting trend (Hellebrand et al. 2002). This observation is of particular importance for evaluating the potential role of residual garnet during initial melting, which has been proposed in both abyssal and ophiolitic peridotites (Brunelli et al. 2006; Gong et al. 2016; Hellebrand et al. 2002; Hellebrand and Snow 2003; Zhang et al. 2017). When garnet breaks down as peridotites rise into the spinel stability field, the HREE that had been hosted in the garnet will be redistributed between the minerals of a spinel peridotite, yielding an equilibrium clinopyroxene with higher HREE concentrations and lower Sm/Yb ratio (Hellebrand et al. 2002). In the (Sm/Yb)_N versus Yb_N diagram [Fig. 11; N: chondrite-normalization by Anders and Grevesse (1989)], comparison with the peridotite melting model of Hellebrand et al. (2002) suggests that the majority of Dazhuqu peridotites have experienced 0–6% garnet-facies melting followed by 10–18% spinel-facies melting. Garnet + spinel facies melting has also been identified in mantle peridotites from the Luqu ophiolite (Fig. 11), which, in comparison, have been subjected to relatively higher degrees of melting than the Dazhuqu peridotites (Zhang et al. 2017). It should be noted that some Dazhuqu samples plot above the modelled melting trend of Hellebrand et al. (2002) (Fig. 11), indicating modification by late-stage refertilization processes, i.e., melt–rock interaction (Brunelli et al. 2006; Hellebrand et al. 2002).

Fig. 10

a Clinopyroxene Yb_N versus spinel Cr# of the Dazhuqu mantle peridotites. The fractional melting trend is from Liu et al. (2008). b Clinopyroxene (Ce/Yb)_N versus Yb_N of the Dazhuqu mantle peridotites. The fractional melting trends (solid lines) for both spinel and garnet facies are after Brunelli et al. (2006) and Hellebrand et al. (2002). A quantitative calculation for refertilization has been carried out, comparing the refertilization model described by Brunelli et al. (2006). The modelled compositional paths (dashed lines) are followed by a variably depleted residual clinopyroxene refertilized with quasi-instantaneous melts extracted after 5, 10 and 15% source partial melting. Detailed parameters for melting and refertilization modeling are listed in Online Resource Appendix I. Data of abyssal peridotites are after the review in Warren (2016)

Fig. 11

Clinopyroxene (Sm/Yb)_N versus (Yb)_N of the Dazhuqu mantle peridotites. The trajectories of melting and fields of abyssal peridotites are from Hellebrand et al. (2002). The modeled melting curves describe a three-step process: (1) melting in the garnet stability field, (2) garnet breakdown, (3) melting in the spinel stability field. The labels on the melting curves represent percent melting in the garnet or spinel stability fields. Parameters for melting modeling are after Hellebrand et al. (2002) (Appendix I). Data of the Luqu peridotites are from Zhang et al. (2017), and IBM fore-arc mantle from Parkinson and Pearce (1998) and Chen and Zeng (2007). The refertilization trend (the red dashed arrow) refers to the modeling result from Brunelli et al. (2006) and Hellebrand et al. (2002). Data of chondrite are from Anders and Grevesse (1989). Sp spinel, gt garnet, cpx clinopyroxene

Melt–rock interaction experienced by mantle peridotites can be modeled both qualitatively and quantitatively. Clinopyroxenes in several Dazhuqu peridotites, particularly harzburgites, show enrichment in some incompatible elements (such as LREE, Sr, Zr, Hf and Ti), though most of them display LREE-depleted patterns (Fig. 7). In the diagram of (Ce/Yb)_N versus Yb_N of clinopyroxene, the Dazhuqu peridotites do not follow the fractional melting trend for both spinel- and garnet-facies (Brunelli et al. 2006), but deviate to high Ce/Yb ratios (Fig. 10b). This indicates that the Dazhuqu samples have experienced late-stage enrichment after melt depletion, which resulted in the elevated Ce contents of clinopyroxenes. Clinopyroxene (Ce/Yb)_N/Yb_N is consistent with variably depleted residual clinopyroxene refertilized by a quasi-instantaneous melt extracted after ~ 12% source partial melting (Fig. 10b; Brunelli et al. 2006).

The effects of melting and melt–rock interaction on HSE abundances

It is well known that HSE concentrations of the residual mantle are predominantly controlled by sulfide stability because of the extremely high metal/silicate partition coefficient (e.g., Brenan et al. 2016; Luguet and Reisberg 2016). Due to the stability of sulfides (or alloys or oxides) at low degrees of partial melting, the concentrations of HSE in the residual mantle would increase but their patterns do not change significantly. At larger degrees (> 15–20%) of partial melting, PPGE and Re behave incompatibly whereas IPGE still behave compatibly (Büchl et al. 2002; Keays 1995; Lorand et al. 1999; Morgan 1986), resulting in fractionated HSE patterns. These hypotheses are consistent with the sulfide petrography and HSE patterns of the Dazhuqu mantle peridotites. The Dazhuqu lherzolites typically contain a few (< 1%) subhedral sulfides (mainly pentlandite; Fig. 3e), which is consistent with the sulfide stability at a low extent of melting (i.e., ~ 10–14%). In contrast, sulfides have been rarely discovered in the Dazhuqu harzburgites, which only contain a few grains of Ni-rich alloys (Fig. 3f). This indicates the almost complete consumption of sulfides from the harzburgites at high degrees of melting up to 20%. The sulfide petrography is in a good agreement with the flat HSE patterns of the Dazhuqu lherzolites and the PPGE- and Re-depleted patterns of a few harzburgites (Fig. 8). All but three Dazhuqu samples have consistent Ru/Ir ratios (Fig. 12a); they can be explained by batch melting from PUM, as modeled by van Acken et al. (2010), assuming monosulfide solid solution (mss)–sulfide melt equilibrium. In contrast, the Re/Ir ratios of the Dazhuqu samples decrease sharply at a high extent (> ~ 20%) of melting (Fig. 12b) and deviate from the model of van Acken et al. (2010). This suggests the efficient removal of relatively incompatible Re during melt depletion. As for the Pd/Ir ratio versus whole-rock Al₂O₃ diagram (Fig. 12c), all Dazhuqu samples deviate clearly from the mss–sulfide liquid melting model, a feature that is probably ascribed to the non-consideration of phases (such as alloys and oxides) other than sulfides (van Acken et al. 2010). It should be noted that two Dazhuqu harzburgites (14DZQ75 and 14DZQ78) have elevated Pd/Ir ratio with respect to the other samples (Fig. 12c). The Pd enrichment may be a common feature of the Yarlung Zangbo ophiolitic mantle [Fig. 8; data from this study and Dai et al. (2011) and references therein]. This feature has been also reported for the Oman and Ligurian ophiolitic peridotites and for some abyssal (MARK area) and continental peridotites (Hanghøj et al. 2010; Lorand et al. 2009; Luguet et al. 2003, 2004; Pearson et al. 2004). A possible mechanism is the refertilization of the Pd-depleted residual mantle by Pd-rich, S-saturated Cu–Ni sulfide melts (Lorand et al. 2009; Luguet and Reisberg 2016). This implies a ubiquitous melt–rock interaction between the Yarlung Zangbo upper mantle and Pd-rich sulfide melts, though the interactions are typically in a small volume (Luguet and Reisberg 2016).

Fig. 12

Measured a Ru/Ir, b Re/Ir, c Pd/Ir and d Pt/Pd (non-normalized) versus whole-rock Al₂O₃ correlation diagrams of the Dazhuqu mantle peridotites. Partial melting modeling from van Acken et al. (2010) are shown in a–c for comparison, where solid lines represent residual peridotite compositions after partial batch melting assuming monosulfide solid solution (mss)–sulfide melt equilibrium (fractional melting trend is similar and not shown), and dashed line in b assuming moderately incompatible bulk partitioning of Re. The PUM composition used by the modeling of van Acken et al. (2010) was after Becker et al. (2006). Ranges of chondritic Ru/Ir, Re/Ir and Pd/Ir ratios are after Horan et al. (2003), while the abyssal peridotite compositions are after the compilation by Day et al. (2017). Data of mantle wedge xenoliths from the Kamchatka arc and the trend of those from the Tabar–Lihir–Tanga–Feni (TLTF) arc are after Kepezhinskas and Defant (2001) and McInnes et al. (1999), respectively

Ancient Os isotope signals: SCLM or mantle heterogeneity?

Re–Os isotope systematics provide critical information for constraining the nature and time of melt depletion processes in mantle peridotites, particularly for their ancient depletion signals, and are robust against late-stage metasomatism and contamination processes with respect to lithophile isotope systems (e.g., Sm–Nd and Rb–Sr) (Rudnick and Walker 2009). All studied Dazhuqu mantle peridotites have unradiogenic ¹⁸⁷Os/¹⁸⁸Os ratios (0.11836–0.12922) relative to PUM (Meisel et al. 2001), yielding Re depletion ages, i.e., minimum ages of melt extraction, of 0.06–1.66 Ga (Table 3). All but one Dazhuqu samples plot within the range previously reported for mantle peridotites from other ophiolitic massifs along the YZSZ (Fig. 9a), which are also within the field defined by abyssal peridotites (Fig. 9b). Sample 14DZQ71 has the lowest ¹⁸⁷Os/¹⁸⁸Os ratio (i.e., 0.11836) among all reported whole-rock data for mantle peridotites from the YZSZ, which yields a very old T_RD age of 1.66 Ga. This sample stands out from the other samples, which all have ¹⁸⁷Os/¹⁸⁸Os ratios clustered between 0.1241 and 0.1292 and have T_RD ages younger than 810 Ma (Table 3; Fig. 9a). The highly depleted Os isotopic composition of the sample 14DZQ71 is somewhat like ancient Os isotope signals preserved in a few abyssal peridotites (Alard et al. 2005; Brandon et al. 2000; Day et al. 2017; Harvey et al. 2006; Lassiter et al. 2014; Liu et al. 2008; Snow and Reisberg 1995) and in ophiolites (Alard et al. 2005; Büchl et al. 2002; Gong et al. 2016; Liu et al. 2012; Snow et al. 2000; van Acken et al. 2008). For example, highly depleted ¹⁸⁷Os/¹⁸⁸Os ratios with T_RD ages up to 2.66 Ga have been observed in the sulfides from the Ligurian ophiolitic peridotites (Alard et al. 2005). Old depleted components recorded by abyssal and ophiolitic peridotites are typically explained by two processes (see Rampone and Hofmann (2012) for a review). One hypothesis attributes ancient signatures to preservation of refractory domains in the heterogeneous asthenospheric mantle that have avoided erasure by mantle convection (i.e., mantle heterogeneity; Liu et al. 2008). Another hypothesis suggests that depleted domains result from SCLM slices “stranded” in the asthenosphere during continental rifting and break-up (e.g., Boillot et al. 1989; Müntener and Manatschal 2006; O’Reilly et al. 2009). The second model is best manifested by the St Peter-Paul and Sal Island peridotites (Atlantic Ocean) (Bonatti 1990; Coltorti et al. 2010), the peridotite bodies from the Zabargad Island (Red Sea) (Snow and Schmidt 1999) and the Alpine–Apennine ophiolitic peridotites (Snow et al. 2000; van Acken et al. 2008). For example, the St Peter-Paul peridotites systematically contain lower spinel Cr# and Mg# values and orthopyroxene Al₂O₃ and clinopyroxene Na₂O contents relative to the abyssal peridotites from the Mid-Atlantic Ridge, but comparable to those of the Zabargad peridotite bodies. Therefore, the St Peter-Paul peridotites are thought to be relics of sub-continental mantle left behind during the opening of the equatorial Atlantic (Bonatti 1990), as is the case for the northern and central part of the Zabargad peridotite bodies, which were interpreted as ancient continental lithospheric mantle left by the Red Sea Rift (Snow and Schmidt 1999). Similarly, some old domains within the Alpine–Apennine ophiolites are widely considered as orogenic peridotites trapped in the Jurassic Tethys Ocean (Müntener et al. 2010; Rampone et al. 1995).

The following lines of evidence, however, argue against that this old Dazhuqu mantle peridotite represents an isolated slice of SCLM. First, sub-continental mantle is characterized by the existence of aluminium- and iron-rich spinels, as is the case for the St Peter-Paul (Bonatti 1990) and Zabargad (Bonatti et al. 1986) peridotites. As shown is Fig. 5b, the St Peter-Paul and Zabargad mantle peridotites contain both relatively lower spinel Cr# and Mg# [SPP-Zabargad trend; according to the data compilation by Bonatti (1990)] than residual abyssal peridotites. In this respect, the Al–Fe enrichment of the SPP-Zabargad trend is quite different from the mineral chemical features of the Dazhuqu peridotites, which display chromium and magnesium enrichments in spinels that are comparable to residual abyssal peridotites (Fig. 5b). On the other hand, the majority of abyssal peridotites contain a relatively lower sodium content with respect to continental peridotites at a given spinel Cr# (Hellebrand and Snow 2003), though there is a large overlap between the two, as is shown in Fig. 6 of this study. Continental peridotites and some abyssal peridotites clearly deviate from the fractional melting curve modeled from a Na-rich source by Hellebrand and Snow (2003) (Fig. 6), which could be interpreted as sodium metasomatism even at high pressure in the spinel stability field (Müntener et al. 2010). The Dazhuqu peridotites from this study have extremely low clinopyroxene Na₂O and plot along the fractional melting curve, though a few harzburgites contain relatively higher clinopyroxene Na₂O due to melt refertilization (Fig. 6). It indicates that the Dazhuqu samples are residuals of low-pressure melt extraction and are comparable to most abyssal peridotites (Hellebrand and Snow 2003; Müntener et al. 2010).

Second, we here point out an Os-related indicator for distinguishing the Dazhuqu peridotites from typical non-cratonic SCLM. Numerous studies have focused on potential correlations between chemical fertility indexes (e.g., whole-rock Al₂O₃ content, clinopyroxene Yb abundance and spinel Cr# value) and isotopic (Os and Hf) ratios (Bizimis et al. 2007; Day et al. 2017; Rampone and Hofmann 2012; Reisberg and Lorand 1995; Stracke et al. 2011; Warren et al. 2009). In a prior study, Reisberg and Lorand (1995) present two examples of striking correlations between ¹⁸⁷Os/¹⁸⁸Os ratio and whole-rock Al₂O₃ content in orogenic peridotites (i.e., the Ronda and Eastern Pyrenean ultramafic massifs). These correlations are most simply explained as isochron analogues with Al₂O₃ proportional to Re/Os (Reisberg and Lorand 1995). Data compilation of non-cratonic SCLM xenoliths from Luguet and Reisberg (2016) has indeed yielded two rough positive correlations between ¹⁸⁷Os/¹⁸⁸Os ratio and whole-rock Al₂O₃ content (Fig. 9b). For non-cratonic SCLM xenoliths, the slope of correlations between ¹⁸⁷Os/¹⁸⁸Os and whole-rock Al₂O₃ probably depends on age, that is, young SCLM would be expected to show an approximately flat trend (Luguet and Reisberg 2016). Interestingly, such correlations have also been observed in present-day abyssal peridotites (Day et al. 2017; Warren et al. 2009). In these cases, the presence of old, refractory mantle domains in the MORB source was interpreted as evidence for a SCLM affinity (Rampone and Hofmann 2012; Stracke et al. 2011). Systematic examination of the Dazhuqu mantle peridotites in this study, however, indicates that no clear correlations between ¹⁸⁷Os/¹⁸⁸Os and whole-rock Al₂O₃ (Fig. 9b) and clinopyroxene Yb (not shown here) exist in these samples. In addition, it is evident that the sample (14DZQ71) with the lowest ¹⁸⁷Os/¹⁸⁸Os ratio does not correspond to the sample with the lowest whole-rock Al₂O₃ and clinopyroxene Yb contents. It is clear that non-cratonic SCLM xenoliths with Os characteristics similar to the Dazhuqu samples can indeed be found, but such SCLM samples are rare. In contrast, the Dazhuqu samples studied all plot within the range of abyssal peridotites (data compilation from Day et al. (2017)) (Fig. 9b). These results argue that an abyssal peridotite affinity for the Dazhuqu samples is much more likely, though their Os isotopic compositions could in theory be consistent with both continental and oceanic affinities. Hence, we conclude, based on all lines of evidence, that the Dazhuqu peridotites are less likely to be samples of isolated sub-continental slices stranded in the oceanic asthenosphere. They represent, in the strict sense, residues of heterogeneous oceanic mantle with preservation of small amounts of old depleted domains, which have experienced ancient melting. These old mantle domains escaped homogenization by asthenospheric convective mixing, as was previously suggested to explain the case of ancient and highly heterogeneous abyssal peridotites beneath the Gakkel Ridge (Liu et al. 2008).

Neo-Tethyan sub-oceanic mantle

Some studies have identified the Xigaze ophiolitic peridotites (including the Dazhuqu massif) as fore-arc peridotites (Dai et al. 2013; Dubois-Côté et al. 2005; Dupuis et al. 2005a; Xia et al. 2003; Xiong et al. 2017). This hypothesis was based mainly on (1) the refractory nature and high degree of melting and (2) the U-shaped whole-rock REE patterns of these mantle peridotites. The Dazhuqu peridotite samples comprise ~ 1/3 lherzolite, though harzburgite is the major rock type (Table 1). One interpretation of this could be that the extent of melting of the Dazhuqu peridotites may be less than typical fore-arc peridotites, which consist mainly of harzburgite (Parkinson and Pearce 1998; Pearce et al. 2000). Whole-rock major element compositions show that the Dazhuqu peridotites contain relatively higher Al₂O₃ (0.4–1.6 wt.%), CaO (0.50–1.9 wt.%) and lower MgO (41.8–45.5 wt.%) contents than the refractory IBM fore-arc mantle (these values are 0.12–1.04 wt.%, 0.30–2.36 wt.% and 42.4–50.0 wt.%, respectively) (Parkinson and Pearce 1998) (Table 2; Fig. 4a, b). These features confirm that the Dazhuqu samples are less refractory (i.e., lower degree of melting) than typical fore-arc peridotites.

The most significant chemical indicator for the high degree of melting of oceanic peridotites is high spinel Cr# value (Arai 1994; Dick and Bullen 1984; Hellebrand et al. 2001). Spinels with Cr# above 0.6 have been reported for modern fore-arc peridotites (Parkinson and Pearce 1998; Pearce et al. 2000) but very rarely for abyssal peridotites (Warren 2016). Therefore, Cr# values above 0.6 in spinels are widely considered as a chemical index for fore-arc peridotites. However, there are some cases where abyssal peridotites at modern MOR indeed contain spinels with Cr# > 0.6, such as the spinel peridotites from the Newfoundland margin drilled during ODP Leg 210 at Site 1277, North Atlantic Ocean (Müntener and Manatschal 2006) and some dunites and gabbro- and pyroxenite-veined peridotites from modern MOR (Warren 2016, as shown in Fig. 5b in this study). The spinels with Cr# above 0.6 of the Newfoundland case was interpreted as inheritance from ancient sub-arc mantle, which is consistent with the rifting processes forming the North Atlantic Ocean (Müntener and Manatschal 2006). However, the lack of extensional allochthons in the Yarlung Zangbo suture and regional areas (Gong et al. 2016) makes it a different case from the rifting-controlled OCT such as the Newfoundland–Iberia region. In contrast, the high-Cr# (> 0.6) spinels from the dunites and veined peridotites (Warren 2016) could be ascribed to a reactive origin, i.e., intense melt–rock interactions. It is well known that dunites and pyroxenite veins are formed by focused melt flow in the upper mantle (Büchl et al. 2002; Dick et al. 1984, 2010; Dygert et al. 2016; Kelemen and Dick 1995; Kelemen et al. 1995). Meanwhile, harzburgites can also be formed by melt percolation at a relatively small melt/rock ratio (Kelemen et al. 1992). As for the Dazhuqu mantle rocks in this study, the lherzolites have spinel Cr# that are less than 0.4, consistent with the compositions of residual abyssal peridotites (Fig. 5). It indicates that these lherzolites have experienced relatively low extents of melting typical of MOR environments. In contrast, the Dazhuqu harzburgites have spinel Cr# values overlapping with the ranges of both fore-arc and residual abyssal peridotites, with 2 (14DZQ78 and 14DZQ80) out of the 21 harzburgites containing spinel Cr# > 0.6 (Fig. 5). These two harzburgites are characterized by compositions indicative of intense melt–rock interactions, e.g., extremely elevated clinopyroxene LREE abundance (Fig. 7a), high clinopyroxene Ce and Sr contents (Fig. 7b) and (Ce/Yb)_N (Fig. 10b) and (Sm/Yb)_N (Fig. 11) ratios. They contain lowest clinopyroxene (0.7% and 1.5%, respectively) and highest olivine + orthopyroxene modal abundances among the Dazhuqu samples (Table 1), a composition that could be interpreted by the reactive scenario proposed by Kelemen et al. (1992) of dissolving clinopyroxene and producing olivine and orthopyroxene. In addition, the two samples also have spinel Cr# comparable to those of the dunites from the Luqu ophiolite (Fig. 5), which has been interpreted as the result of melt–rock interactions at high melt/rock ratios (Zhang et al. 2017). We argue that the high-Cr# signature of some Dazhuqu harzburgites is probably derived from porous melt flow and melt–rock interaction, with lower melt/rock ratio with respect to the dunites and veined peridotites from modern MOR (Warren 2016). Thus, the extent of melting experienced by the Dazhuqu samples is comparable to that of the most depleted abyssal peridotites at modern MOR.

On the other hand, fluid fluxing increases the degree of melting in the mantle wedge, leaving peridotites that display strong depletion in their Ti concentrations (Bizimis et al. 2000). Hence, the extent and rate of depletion in clinopyroxene Ti content have been regarded as indicators for distinguishing hydrous melting in an arc setting from dry melting beneath MOR (Bizimis et al. 2000). Comparing the dry and hydrous melting results from Bizimis et al. (2000) (Fig. 13), the Dazhuqu peridotites have moderate clinopyroxene Ti contents that are comparable to the most depleted end of abyssal peridotites (Warren 2016). Most of the Dazhuqu samples deviate from the modeled dry melting trend (Bizimis et al. 2000) but show enrichments in Zr content of clinopyroxenes (Fig. 13) that could be ascribed to intense melt–rock interaction, as has been proposed for abyssal peridotites (Brunelli et al. 2006). These trace element features diverge significantly from the modeled refertilization–hydrous melting trend, which characterizes SSZ peridotites (Fig. 13; Bizimis et al. 2000).

Fig. 13

Clinopyroxene Ti versus Zr of the Dazhuqu mantle peridotites. Data of abyssal peridotite are after the review in Warren (2016). Field of SSZ peridotite and melting trend are from Bizimis et al. (2000). Detailed parameters for melting modeling used by Bizimis et al. (2000) are listed in Appendix I. The boundaries of refertilization–hydrous melting model (Bizimis et al. 2000) are the limits of clinopyroxene compositions using the minimum (min) and maximum (max) concentrations in the fluid component, respectively. The minimum fluid components of Ti and Zr are 0 ppm and 16 ppm respectively, while the maximum fluid components of these elements are 436 ppm and 365 ppm, respectively. The melt–rock interaction trend (dashed line) are based on trajectories of the Dazhuqu harzburgite samples from this study and refertilization modeling results from Brunelli et al. (2006) (Appendix I)

The Dazhuqu peridotites show U-shaped whole-rock REE patterns (Fig. 7a), which were interpreted in previous studies (e.g., Xia et al. 2003) as a result of interaction between peridotites and boninitic melts. Similar arguments have also been proposed in studies of other ophiolite massifs along the YZSZ (e.g., Dai et al. 2011, 2013; Dubois-Côté et al. 2005; Xiong et al. 2017; Zhou et al. 2005). Boninite is commonly characterized by LREE enrichment relative to MREE and HREE and is traditionally taken as an indicator for a fore-arc setting (Crawford et al. 1989 and references therein). However, boninite is not the only candidate for the U-shaped REE patterns in the Yarlung Zangbo ophiolitic peridotites, because most (non-MORB) melts with which peridotites could interact are likely to have elevated LREE/HREE, such as OIB (ocean island basalt). In addition, abyssal peridotites at present-day MOR may also display LREE-enriched patterns in their whole-rock compositions (e.g., Day et al. 2017), demonstrating that the participation of boninitic magmas is not always required to produce such patterns. In this respect, it is misleading to attribute the U-shaped whole-rock REE patterns of the Dazhuqu peridotites to interaction with boninitic melts in a fore-arc setting.

Garnet-facies melting experienced by the Dazhuqu samples provides potential information to distinguish the Dazhuqu peridotites from fore-arc peridotites. About half of abyssal peridotite samples [data from Hellebrand et al. (2002)] show relatively low (Sm/Yb)_N ratios at a given Yb_N in clinopyroxene, indicating garnet-facies melting (Fig. 11). In contrast, fore-arc peridotites, exemplified by the IBM mantle peridotites (Parkinson and Pearce 1998), commonly have much higher (Sm/Yb)_N ratios at a given Yb_N in clinopyroxenes (Fig. 11). Only a few samples of the IBM system display weak depletion (< 3% partial melting) in the garnet stability field; most of the IBM samples have experienced solely spinel-facies partial melting. This distinction of the Dazhuqu and abyssal peridotites relative to the IBM samples might be associated with the depth where initial melting occurs. The Dazhuqu peridotites must have been derived from a depth that is not shallower than the spinel–garnet transition zone (about 85 km in depth; Robinson and Wood 1998). In contrast, fore-arc mantle is typically derived from a shallower depth (commonly < 80 km; Chen et al. 2018), inconsistent with the data from Dazhuqu mantle peridotites.

Most Dazhuqu lherzolites and harzburgites display relatively flat HSE patterns, though some harzburgite samples show variable PPGE and Re depletions because of high degrees of partial melting (Fig. 8). The HSE patterns and abundances of the Dazhuqu mantle peridotites are consistent with those of the most depleted end of abyssal peridotites. In this light, the Dazhuqu mantle rocks, particularly harzburgites, are equivalent in composition to the southern massifs (including Semail and Wadi Tayin) of the Oman ophiolite (Fig. 8), which are interpreted as residues of partial melting as depleted as the most depleted abyssal peridotites beneath a paleo-MOR (Hanghøj et al. 2010). This is exemplified by their relatively refractory whole-rock compositions (Fig. 4), high spinel Cr# values (Fig. 5) and extremely low clinopyroxene Na₂O and Yb contents (Figs. 6, 10a). It should be noted that the mantle peridotites from the YZSZ (including Dazhuqu), along with the southern Oman massifs and most abyssal peridotites, show unfractionated Pt–Pd patterns, though a few of them are modified by Pd-rich sulfide melt–rock interactions (Fig. 8). This feature is fundamentally different from the fractionated Pt–Pd of some sub-arc mantle (e.g., mantle wedge xenoliths). For example, some mantle wedge xenoliths from the Kamchatka and Tabar–Lihir–Tanga–Feni (TLTF) arcs have extremely high Pt/Pd ratios (up to 126; Fig. 12d), which have been ascribed to multi-stage hydrous melting followed by the formation of Pt–Fe alloys typical of subduction zones (Kepezhinskas and Defant 2001; McInnes et al. 1999). In this light, the Dazhuqu mantle peridotites are different than the sub-arc mantle. A sub-arc mantle could experience an extremely high extent of melting because of the addition of subducted slab-derived H₂O-bearing fluids. In contrast, at a “dry” MOR environment, mantle rocks have commonly experienced a relatively lower extent of melting and hence do not show highly elevated Pt/Pd ratios. Similarly, the extent of melting experienced by the Dazhuqu mantle peridotites in this study must be lower than that of the Kamchatka and TLTF mantle wedge xenoliths with high Pt/Pd ratios. Therefore, we argue that the Dazhuqu samples are more similar to abyssal peridotites in their HSE pattern and abundance, other than typical sub-arc mantle, though the Pt–Pd indicator must be critically used.

Current knowledge has shown that it is hard to identify unequivocal chemical indicators distinguishing sub-oceanic mantle from its sub-continental and sub-arc counterparts (Rampone and Hofmann 2012). As for the Dazhuqu mantle peridotites from this study, their spinel compositions overlap with both fore-arc and abyssal peridotites; in addition, their Os isotopic ratios are also within the range of both abyssal peridotites and non-cratonic SCLM. Nonetheless, garnet-facies initial melting, abyssal peridotite-like HSE pattern and abundance and the lack of a correlation between Os isotopic ratio and some chemical fertility indexes positively suggest a sub-oceanic affinity for these rocks, rather than SCLM or sub-arc mantle. The Dazhuqu lherzolites and harzburgites show compositions that are similar to typical and most depleted abyssal peridotites, respectively. It implies their variable degrees of partial melting, which in turn could be as high as the extent of melting experienced by fore-arc mantle.

Broader implications for tectonic evolution of the Neo-Tethys Ocean

Ophiolites were widely interpreted, in early studies, as fossil oceanic lithosphere formed beneath MOR that were later emplaced onto continents (Gass 1968; Moores and Vine 1971; Nicolas et al. 1981). However, based on the arc-like geochemistry of associated crustal rocks, it has been argued that many ophiolites may have been generated in a SSZ setting (Alabaster et al. 1982; Miyashiro 1973; Pearce et al. 1984; Pearce and Robinson 2010; Whattam and Stern 2011). A SSZ origin has also been extensively proposed for the Yarlung Zangbo ophiolites (Dai et al. 2013; Dubois-Côté et al. 2005; Dupuis et al. 2005a; Hébert et al. 2012; Malpas et al. 2003; Xia et al. 2003; Xiong et al. 2016, 2017). Dai et al. (2013) and Xiong et al. (2016) argued that the Yarlung Zangbo ophiolites formed in a fore-arc basin above the subducted Neo-Tethys Ocean. They also proposed that the Cretaceous (120–130 Ma) crustal rocks in the ophiolites were accreted during fore-arc spreading and that the immense Neo-Tethyan oceanic lithosphere was subducted beneath the Asian continent (Dai et al. 2013). Hence, the only relicts of the Neo-Tethyan oceanic lithosphere are mafic rocks with OIB-type geochemistry found as blocks in the mélange to south of the ophiolites (Aitchison et al. 2000; An et al. 2017; Cai et al. 2012; Dupuis et al. 2005b, 2006; Searle et al. 1987; Ziabrev et al. 2004). Our study, however, suggests that the Dazhuqu mantle peridotites most probably formed as sub-oceanic mantle beneath the Neo-Tethyan MOR. They are unlikely to represent juvenile fore-arc mantle formed following subduction initiation (Whattam and Stern 2011). On the contrary, they were trapped, as an exotic block, at the southern margin of the Lhasa terrane (the Asian continent) in the Lower Cretaceous (~ 124 Ma; Liu et al. 2016) and became the basement for the fore-arc basin sedimentary rocks deposited soon afterwards (~ 119 Ma; Wang et al. 2017). Therefore, the Yarlung Zangbo ophiolites, particularly some massifs in the middle part (e.g., Dazhuqu), comprise an important component of the Neo-Tethyan oceanic lithosphere. The preservation of Neo-Tethyan sub-oceanic mantle along suture zones is not uncommon, as has been also proposed for the Oman ophiolite. However, its mechanism is still poorly constrained. Nicolas and Boudier (2017) proposed ridge–trench collision for the emplacement of the Oman ophiolite, which is probably plausible for the preservation of paleo-MOR oceanic lithospheres on land. This in turn underscores the need for a universal test of this mechanism along the Neo-Tethyan ophiolites in the near future.

Concluding remarks

We highlighted in this study some chemical indicators for the nature of the Dazhuqu mantle rocks in Tibet. (1) The derivation depth of initial melting and CI chondrite-normalized Pt–Pd (i.e., PPGE) pattern were used to distinguish sub-oceanic mantle from sub-arc mantle, because the former could be derived from the garnet stability field and has commonly experienced relatively lower extent of melting with respect to the latter. (2) The Na₂O of clinopyroxene and Al–Fe of spinel and the association between Os isotopic composition and some chemical fertility indexes were used to distinguish sub-oceanic mantle from non-cratonic SCLM.

The Dazhuqu mantle peridotites in this study are more similar to sub-oceanic mantle other than sub-arc or sub-continental mantle. We argue that the Dazhuqu mantle rocks most likely represent “fossil” abyssal peridotites formed beneath the Neo-Tethyan mid-ocean ridge. This directly implies an efficient preservation of the Neo-Tethyan sub-oceanic mantle along the Yarlung Zangbo suture zone, which in this light has a similar scenario to the Oman ophiolite.