Abstract

The preservation of ultra-high-pressure and super-reducing phases in the Neotethyan Luobusa ophiolite in Tibet suggests their deep origin near the mantle transition zone. Dunite and harzburgite core samples from the Luobusa Scientific Drilling Project show supra-subduction zone geochemical signatures and equilibration temperatures of c. 950–1080°C. Olivine shows A-, B-, C- and E-type fabrics, and combinations of A- and E-type or B- and E-type fabrics. Transmission electron microscopy observations show straight dislocations and the activation of multiple slip systems [100](010), [001](010), [001](100) and [100](001) in olivine. The mean water content in olivine, orthopyroxene (Opx) and clinopyroxene (Cpx) from 24 peridotite samples was 16 ± 5, 90 ± 21 and 492 ± 64 ppm, respectively, which is different from the water content of hydrated peridotites above the mantle wedge. The trace element compositions of Cpx exclude significant metasomatism after melt extraction. The high hydrogen partition coefficient between Cpx and Opx (DHCpx/Opx = 5.56 ± 0.96) implies equilibrium at high pressures and rapid exhumation. Based on deformation experiments, the B- and C-type fabrics could be formed in a subduction zone at depths >200 km, whereas the A- and E-type fabrics were produced in the shallow mantle. In a process triggered by slab rollback, the Luobusa peridotites may have been rapidly exhumed within a subduction channel and mixed with the lithospheric mantle of the forearc.

Supplementary material: Major oxide contents in Opx, Cpx and spinel, trace element concentration in Cpx, micrographs and TEM images of peridotite are available at https://doi.org/10.6084/m9.figshare.c.4307828

The fabrics of ophiolitic peridotites record thermodynamic processes from the mid-ocean ridge to the subduction zone, followed by their final emplacement at a convergent boundary (Dilek & Robinson 2003 and references cited therein). The lattice-preferred orientations (LPOs) of olivine and orthopyroxene (Opx) from both a fast-spreading ridge, such as the Oman ophiolite (Boudier & Nicolas 1995), and a slow-spreading ridge, such as the Lanzo massif in the Alps (Boudier 1978), indicate that dislocation creep is the dominant deformation mechanism in the oceanic lithosphere. The LPO of olivine in naturally deformed peridotites is generally characterized by the [100] axis parallel to the lineation and the (010) plane parallel to the foliation, i.e. the A-type fabric (Ben Ismaı̈l & Mainprice 1998; Tommasi & Vauchez 2015; Michibayashi et al. 2016; Tommasi et al. 2016), which can be attributed to the dominant activation of the slip system [100](010) at high temperatures and low strain rates under upper mantle conditions (e.g. Carter & Avé Lallemant 1970; Zhang et al. 2000; Hirth & Kohlstedt 2003). Because olivine is the most abundant mineral in the upper mantle, the A-type olivine fabric is used to infer the direction of mantle flow from seismic anisotropy (e.g. Blackman & Kendall 2002; Park & Levin 2002 ).

Deformation experiments investigating the effects of water, stress, pressure and temperature have found a number of different fabric types in olivine (e.g. Carter & Avé Lallemant 1970; Jung & Karato 2001; Jung et al. 2006; Ohuchi et al. 2011; Ohuchi & Irifune 2013). In addition to the A-type fabric, B- and E-type olivine fabrics, produced by the predominance of the [001](010) and [100](001) slip systems, respectively, have been observed in peridotites from supra-subduction zones (SSZ) – for example, the B-type fabric was seen in peridotites from the Sanbagawa metamorphic belt of SW Japan (Mizukami et al. 2004; Tasaka et al. 2008), the E-type fabric was seen in peridotites from the accreted Talkeetna arc in south-central Alaska (Mehl et al. 2003) and the Hidaka metamorphic belt in northern Japan (Sawaguchi 2004). It is still not clear whether the trench-parallel seismic anisotropy observed in forearc regions is caused by the B-type olivine fabric in the water-rich mantle wedge or the A-type olivine fabric resulting from trench-parallel flow (Long & Becker 2010).

Based on the water-induced fabric transition in olivine in deformation experiments (Jung & Karato 2001), the C-type fabric in garnet peridotites from ultra-high-pressure (UHP) metamorphic terranes is often attributed to ductile deformation in the hydrous deep mantle, e.g. the Cima Di Gagnone of the Central Alps (Frese et al. 2003; Skemer et al. 2006), the Norwegian Caledonides (Katayama et al. 2005) and the North Qaidam UHP belt (Jung et al. 2013). However, Xu et al. (2006) reported that the C-type fabric from the Zhimafang garnet peridotites in the Sulu UHP terrane was formed under 4–7 GPa, 750–950°C and water-poor conditions in a continental subduction zone. This interpretation was confirmed by the C-type fabric in water-poor olivine from the Xugou peridotites in the Sulu UHP terrane (Wang et al. 2013b). In addition, water-poor olivine from the Western Gneiss Region (Norway) has developed the B-type fabric in strongly sheared peridotites and the C-type fabric in garnet peridotites that had undergone UHP metamorphism at P > 6 GPa and 850–950°C, implying a water-independent fabric transition in a cold and dry continental subduction zone (Wang et al. 2013a).

The Luobusa ophiolite, located in the eastern part of the Yarlung–Zangbo Suture Zone (YZSZ) in southern Tibet, represents a remnant of the Neotethyan oceanic lithosphere in an SSZ environment (Zhou et al. 1996, 2005, 2014; Xu et al. 2015a; Dilek & Yang 2018). Different models, such as a mantle plume (Yang et al. 2014; Xiong et al. 2015), channelized mantle upwelling from the transition zone (Griffin et al. 2016) or slab rollback-induced channel flow from the transition zone (Dilek & Yang 2018) have been proposed to interpret the occurrence of UHP and super-reducing phases (e.g. in situ diamond, moissanite and native Fe) in the Luobusa peridotites and chromitites. Only the A-type olivine fabric has so far been observed in the Luobusa peridotites at outcrop and without measurements of the water content (Xu & Jin 2010; Sun et al. 2016), which casts doubt on the implications of the peridotite microstructure for the interpretation of deep mantle processes.

Here we present an integrated study of harzburgite and dunite samples from two boreholes of the Luobusa Scientific Drilling Project. The A-, B-, C- and E-type olivine fabrics, as well as different slip systems in olivine, were recognized in the core samples. Combined with the chemical composition and water content of olivine, Opx and clinopyroxene (Cpx), we propose that the Luobusha ophiolite presents a mixture of peridotites from different depths in a subduction channel. Channel flow from the transition zone to shallow depths induced by slab rollback provides an important pathway for the incorporation of subducted material into the oceanic lithosphere and a connection with plate tectonic processes.

Geological setting

As the tectonic boundary between the Eurasian and Indian plates, the nearly east–west-striking YZSZ extends for >2000 km and separates the Lhasa Terrane in the north from the Himalayan Orogen in the south (Fig. 1a). The onset of India–Asia continental collision has been constrained to 59 ± 1 Ma by stratigraphic dating (Hu et al. 2016). The Gangdese arc along the southern margin of the Lhasa Terrane is an Andean-type continental margin formed by northward subduction of the Neotethyan oceanic lithosphere from the Late Triassic to the Paleocene and has been overprinted by post-collisional magmatism in the thickened crust. The Gangdese batholith is mainly composed of gabbros, diorites, granodiorites, granites and leucogranites, with four discrete stages of magmatism at 205–152, 109–80, 65–41 and 33–13 Ma (e.g. Mo et al. 2007; Wen et al. 2008; Chung et al. 2009; Ji et al. 2009; Zhu et al. 2011; Zhang et al. 2014).

The ophiolite massifs and ophiolitic mélanges along the YZSZ preserve two age peaks of generation: 180–150 and 130–110 Ma (Hébert et al. 2012; Xu et al. 2015b). All the ophiolites show geochemical features of an SSZ environment, particularly arc and back-arc settings (Hébert et al. 2012). The Luobusa ophiolite in the eastern YZSZ is the largest chromite mine in China and extends c. 42 km along-strike and covers an area of c. 70 km2. It mainly consists of harzburgites, dunites, sparse podiform chromitites and mafic cumulates (Zhou et al. 1996; Xiong et al. 2015). The Luobusa ophiolite was formed at a mid-ocean ridge in the Mid-Jurassic, as constrained by a whole-rock Sm–Nd isochron age of 177 ± 31 Ma for gabbro dykes (Zhou et al. 2002) and a sensitive high-resolution ion microprobe U–Pb age of 163 ± 3 Ma for mid-ocean ridge basalt (MORB)-type diabase (Zhong et al. 2006). It was then re-fertilized by boninitic melts in a mantle wedge above the subduction zone at c. 126 Ma (Malpas et al. 2003; Zhou et al. 2005, 2014).

Bounded by two south-dipping thrust faults, the Luobusa ophiolite was thrust over the upper Oligocene–lower Miocene conglomerate of the Luobusa Formation to the north and is underlain by Upper Triassic flysch to the south (Fig. 1b) (Yamamoto et al. 2007; Liang et al 2011; Xu et al. 2015a). Although all the ultramafic rocks in the Luobusa ophiolite appear to have equilibrated at relatively shallow depths as spinel- or plagioclase-bearing peridotites (Hébert et al. 2003; Xiong et al. 2015), the occurrence of UHP (diamond, TiO2 II, coesite, stishovite pseudomorph, phase BMJ) and super-reducing phases (native elements, alloys, carbides, nitrides and moissanite) in the Luobusa peridotites and chromitites shows that they originated from the deep upper mantle or the mantle transition zone under conditions of very low oxygen fugacity (Bai et al. 1993; Robinson et al. 2004; Yang et al. 2007, 2014; Dobrzhinetskaya et al. 2009; Yamamoto et al. 2009; Xu et al. 2015b; Griffin et al. 2016; R.Y. Zhang et al. 2016). The abundant clinoenstatite lamellae in enstatite from lherzolite and harzburgite core samples of the Luobusa Scientific Drilling Project suggest that the Luobusa peridotites originated at P > 7 GPa (R.Y. Zhang et al. 2016). In addition, crustal minerals (e.g. zircon, quartz, corundum, K-feldspar, plagioclase, apatite and amphibole) have been found as mineral inclusions in the Luobusa peridotites and podiform chromitites (Robinson et al. 2015). These findings have initiated a heated debate about the formation mechanisms of the Luobusa ophiolite, as well as the recycling of the subducted continental and oceanic materials.

Sample description

To constrain the petrological and structural variations of the Luobusa ophiolite, the Luobusa Scientific Drilling Project drilled two boreholes (29° 13′ 24.19″ N, 92° 11′ 36.33″ E) at an altitude of 4378 m: the 1478.8 m deep LBSD-1 borehole and the 1853.8 m deep LBSD-2 borehole (Fig. 1b). Serpentinization is common in the Luobusa peridotites and some of the core samples are strongly serpentinized and fragile, which makes it difficult to identify the lineation and foliation of the peridotites.

We selected 17 relatively fresh harzburgite and seven relatively fresh dunite samples from the LBSD-1 and LBSD-2 boreholes (Fig. 1c; Table 1). Most of the samples show a degree of serpentinization <5%, although samples B186, B199, B281, B329, B137, B155, B226, B368 and B536 have a degree of serpentinization of 5–10%. The serpentine minerals are lizardite and/or chrysotile aggregates and appear along grain boundaries or fractures. Theses samples are coarse grained and show granoblastic or porphyroclastic textures.

The harzburgite samples contain olivine (65–85 vol.%), Opx (10–28 vol.%) and Cpx (1–5 vol.%), with minor spinel and serpentine (Fig. S1). Cpx occurs as either small interstitial grains or exsolution lamellae in Opx grains. The dunite samples consist of 85–95 vol.% olivine, minor Opx, and interstitial Cpx and spinel. The grain size of olivine is usually 1–3 mm, although it sometimes reaches 8 mm. Both olivine and Opx show plastic deformation features, such as irregular grain boundaries (Fig. 2a), kink bands in olivine (Fig. 2b), the undulose extinction of coarse-grained olivine (Fig. 2c–d) and the dynamic recrystallization of fine-grained olivine (0.1–0.5 mm grain size) (Fig. 2e–f). We also observed abundant exsolution lamellae of Cpx in Opx grains (Fig. 2g–h), which has previously been reported in harzburgite core samples from the LBSD-1 borehole and attributed to the transformation of clinoenstatite from high-temperature Ca-bearing orthoenstatite at pressures >7 GPa (Zhang et al. 2017). Some harzburgite samples show coarsely vermicular symplectites of Opx + Cr–Al spinel ± Cpx (Fig. 3), which have been interpreted as the breakdown products of majoritic garnet, with estimated minimum pressures >13 GPa (Griffin et al. 2016).

Methods

Mineral composition analyses

The major element compositions of olivine, Opx, Cpx and spinel from 23 peridotite samples were determined using a Shimadzu EPMA1600 electron microprobe at the Chinese Academy of Sciences Key Laboratory of Crust–Mantle Materials and Environments at the University of Science and Technology of China and a JEOL JXA-8100 electron microprobe at the State Key Laboratory for Mineral Deposits Research at Nanjing University. The analytical conditions were a 20 nA beam current, a 15 kV accelerating voltage and a 1 µm spot diameter. For each sample, the major element concentrations in each mineral were the mean value of five to eight grains. The trace element concentrations in Cpx from 11 harzburgite samples was acquired using an Agilent 7700× laser ablation inductively coupled plasma mass spectrometer with an energy density of 10 J cm−2, a spot diameter of 44 µm, a frequency of 4 Hz and ablation time of 40 s. For each sample, four to ten Cpx grains were analysed to give a mean value.

Microstructural analyses

The LPOs of olivine and Opx from nine harzburgite samples were measured using the electron back-scattered diffraction (EBSD) technique. We used a JEOL JSM-6490 scanning electron microscope equipped with an Oxford Nordlys-S EBSD detector and Channel 5+ software at the State Key Laboratory for Mineral Deposits Research. The highly polished thin sections were tilted by 70° and we used an accelerating voltage of 20 kV, 200× magnification and a working distance of 14–23 mm. Samples B106 and B180 were measured by automatic mapping, whereas the other seven samples were measured by manual indexing. The grain size of Opx in harzburgite B143 was so large that we could not find enough grains for plotting. Excluding sample B143, at least 100 grains of olivine and 50 grains of Opx were analysed for each sample to ensure the reliability of the statistical analyses.

For samples without evident lineation and foliation, non-oriented thin sections were prepared for microstructural analyses. Despite variations in composition and water content, Opx in peridotites develops a stable LPO characterized by the [001] axis parallel to the lineation and the (100) plane parallel to the foliation (e.g. Manthilake et al. 2013; Bystricky et al. 2016; Soustelle & Manthilake 2017; Jung 2017). Therefore the [001](100) fabric of Opx was used as a coordinate reference to rotate the EBSD data, i.e. the point maxima of the [001] axis parallel to the X direction and the point maxima of the [100] axis parallel to the Z direction. The pole figures of olivine and Opx were plotted using the Petrophysics program of Mainprice (1990).

To examine the slip systems of olivine, we carried out transmission electron microscopy (TEM) observations of olivine at the Institute of Geology and Geophysics, Chinese Academy of Sciences. The samples for the TEM imaging were prepared by focused ion beam milling using a Carl Zeiss Auriga Compact system. The acceleration voltage was 0.1–30 kV and the beam current was 1 pA–50 nA. The resolution ratio was 5 nm (30 kV, 1 pA). Imaging and electron diffraction were performed on a JEOL JEM-2100 transmission electron microscope using an LaB6 electron shooter and an accelerating voltage of 200 kV.

Water content measurements

Fourier transform infrared spectrometry (FTIR) was used to determine the water content of olivine, Opx and Cpx from 17 harzburgites and seven dunites. A Vertex 70 spectrometer coupled with a Hyperion 2000 microscope at the State Key Laboratory for Mineral Deposits Research and a Nicolet 5700 spectrometer coupled with a Continuμm microscope at the School of Earth and Space Sciences, University of Science and Technology of China, were used for comparison. Both instruments have a KBr beam splitter and a mercury cadmium telluride detector cooled by liquid nitrogen. The thickness of the double-polished thin sections used for the water content measurements varied between 150 and 300 µm. Prior to analysis, the thin sections were placed in an oven at 120°C for at least 8 h to remove any free water from the surface and fractures. The unpolarized FTIR spectra were collected in the range 4000–2800 cm−1 at room temperature, with 128 or 256 scan times and a resolution of 8 cm−1. Depending on the grain size, apertures of 50 × 50 or 25 × 25 µm were used for optically clean, crack- and inclusion-free regions. For each sample, 4–27 grains of each mineral were measured to obtain the mean water content.

A modified form of the Beer–Lambert law was used to calculate the hydrogen concentration:  
CH2O=A/(I×t)
(1)
where CH2O is the content of hydrogen species in ppm H2O, A is the integrated area (cm−1) of the absorption bands in the region of interest, I is the integral specific absorption coefficient (ppm−1·cm−2) and t is the sample thickness (cm). The infrared spectra were integrated from 3325 to 3650 cm−1, the region dominated by the stretching vibrations of OH bonds. The integral specific coefficients were 14.84 ppm−1 cm−2 for Opx and 7.09 ppm−1·cm−2 for Cpx (Bell et al. 1995). For olivine, 1/0.188 ppm−1·cm−2 was adopted as the integral specific coefficient (Bell et al. 2003), which could be 3.5 times the concentration derived from the Paterson (1982) calibration for unpolarized samples.

Results

Mineral chemistry and equilibration temperatures

Table 1 lists the mean major element compositions of olivine from the Luobusa peridotite samples. The Mg# values, defined as the Mg/(Mg + Fe) atomic ratio, vary between 89.6 and 90.6 for olivine in the harzburgites and between 87.7 and 92.5 for olivine in the dunites. Except for dunite B520, with an Mg# for olivine as low as 87.7, the mean Mg# of olivine in the dunites is 91.0, higher than the mean value of 90.1 for the harzburgites (Fig. 4a). The Opx and Cpx grains fall into the composition ranges of enstatite and diopside, respectively (Table S1). The Mg# values of Opx are 89.8–90.8 in the harzburgite samples and 91.4–92.2 in the dunite samples, whereas those of Cpx are in the range 92.4–94.1 in the harzburgite samples and 94.2–95.0 in the dunite samples (Table S1). The Mg# values of Opx and Cpx in the harzburgites and dunites show a negative correlation with the Al2O3 content (Fig. 4b, c). The Cr# values in spinel, defined as the Cr/(Cr + Al) atomic ratio, are 20.4–37.1 for the harzburgite samples and 54.9–78.1 for the dunite samples (Table S1 and Fig. 4d). The spinel grains in the harzburgites are characterized by a higher Al2O3 content (35.56–48.65 wt%), a lower Cr2O3 content (18.57–31.29 wt%) and a higher MgO content (13.84–17.31 wt%) than those in the dunites (Table S2). The Mg# values in spinel are 25.3–46.9 for the harzburgite samples and 58.0–68.9 for the dunite samples.

The relationships between the Mg# in olivine and the Cr# in spinel (Fig. 5a) and between the Mg# and Cr# in spinel (Fig. 5b) indicate that the Luobusa harzburgites are abyssal peridotites subjected to 10–20% partial melting, whereas the more depleted dunites are SSZ peridotites that experienced a high degree of partial melting and were then modified by melt–rock interactions in a mantle wedge, as suggested by previous studies (Zhou et al. 1996, 2005; Xiong et al. 2015). The negative correlation between the Al2O3 content in Opx and the Cr# in spinel (Fig. 5c) confirms the different origins and different degrees of partial melting of the Luobusa harzburgites and dunites.

Assuming a pressure of 1.5 GPa, which corresponds to equilibration in the spinel stability field, the equilibration temperatures of our samples were calculated using the Ca-in-Opx thermometer (Brey & Köhler 1990), the Al-in-Opx thermometer (Witt-Eickschen & Seck 1991) and the two-pyroxene thermometer (Taylor 1998) (Table 2). There is no data for dunites B269, B512, B520 and B536 due to their lack of Opx porphyroblasts in thin section. Orthopyroxene porphyroblasts in the harzburgites and dunites are very uniform in their CaO content (0.44–1.11 wt%), but variable in their Al2O3 content (2.96–4.24 wt% for the harzburgites and 0.68–1.83 wt% for the dunites). For dunites B155 and B256, the Al-in-Opx thermometer yields equilibration temperatures of 776–896°C, c. 100–250°C lower than the values from the Ca-in-Opx and two-pyroxene thermometers (Table 2). The low-Al Opx in the Luobasa dunites has been attributed to the reaction of boninitic melts with the surrounding peridotites (Zhou et al. 2005). Given the better performance of the Taylor (1998) two-pyroxene thermometer for peridotites (Nimis & Grütter 2010), the equilibration temperatures of the studied harzburgites and dunites are c. 950–1080°C.

It is noteworthy that some Opx grains in the harzburgites have a very high CaO content, e.g. sample B120. Such samples show abundant exsolution lamellae of Cpx in Opx grains (Fig. 2h), which has been interpreted as the transformation of clinoenstatite from orthoenstatite at pressures >7 GPa (Zhang et al. 2017). The deep origin of sample B120 is consistent with the coarse vermicular symplectites of Opx + Cr–Al spinel ± Cpx (Fig. 3), which has been attributed to the breakdown of majoritic garnets in Luobusa harzburgites with estimated minimum pressures >13 GPa (Griffin et al. 2016). Therefore the calculated temperature, assuming P = 1.5 GPa, only yields constraints on the temperature of their last equilibrium stage at shallow mantle depths.

Trace element composition of Cpx

The Cpx from the 11 Luobusa harzburgites show consistent rare earth element (REE) patterns, with a slow decrease in heavy to medium REEs from Yb to Eu and a rapid decrease in light REEs from Sm to Ce (Fig. 6a). Cpx from the harzburgites contains extremely low concentrations of the more incompatible lithophile elements (Table S3) and has negative anomalies in Sr, Zr and Ce relative to neighbouring REEs in the trace element patterns (Fig. 6b). These features are similar to the Cpx found in abyssal peridotites (Johnson et al. 1990) and agree with previous studies of the Luobusa harzburgites (Zhou et al. 1996, 2005). Therefore the Luobusa harzburgites did not experience pervasive metasomatism after melt extraction.

Fabrics of olivine and pyroxene

After rotation of the EBSD data, Opx from eight harzburgite samples and one dunite sample show the concentration of grains with the [001] axis sub-parallel to the lineation and the (100) plane sub-parallel to the foliation (Fig. 7). This fabric pattern agrees with the typical LPO of Opx in peridotites (e.g. Tommasi et al. 2016; Jung 2017). Therefore the structural coordinate system is established with X parallel to the lineation, Z normal to the foliation and Y normal to the lineation in the foliation plane. The pole figures of olivine can be plotted using the rotated EBSD data.

The olivine in our samples shows complex fabrics (Fig. 8; Table 3). Harzburgites B143 and B169 have developed a typical A-type fabric with the maximum concentration of the [100], [010] and [001] axes sub-parallel to the X, Z and Y directions, respectively, implying dominant activation of the [100](010) slip system. By contrast, harzburgite B106 shows a B-type fabric characterized by the maximum concentration of the [001] axis parallel to the lineation and the [010] axis perpendicular to the foliation, suggesting the predominance of the [001](010) slip system. Harzburgite B153 has developed an E-type fabric due to the activation of the [100](001) slip system. Harzburgite B245 show a C-type fabric with the point maxima of the [001] axis parallel to the lineation and those of the [100] axis perpendicular to the foliation, reflecting the dominant activation of the [001](100) slip system.

It is noteworthy that olivine from other samples shows maximum concentrations of the [100] axis and the [010] axis sub-parallel to the lineation and perpendicular to the foliation, respectively, but the [001] axis does not concentrate along the Y direction as in the typical A-type fabric. For dunite B256, the [001] axis of olivine concentrates sub-perpendicular to the foliation, suggesting the activation of both the [100](010) and [100](001) slip systems, i.e. a combination of A- and E-type fabrics. For harzburgites B98, B137 and B180, the [001] axis of olivine forms two-point maxima sub-parallel to the lineation and sub-perpendicular to the foliation, respectively. Given the orthorhombic crystallography of olivine, the asymmetrical pole figures allow us to distinguish the corresponding slip direction and the slip plane of olivine. For example, the two-point maxima of the olivine [001] axis from sample B98 are sub-parallel to those of the [100] and [010] axes, respectively, suggesting the activation of both the [001](010) and [100](001) slip systems. Therefore the three harzburgite samples record a combination of B- and E-type fabrics.

The bright-field TEM images of olivine grains were used to infer the dominant slip systems in olivine and compared with the EBSD-derived LPO of olivine (Fig. 9). The Burgers vectors were determined using the g · b = 0 and g · (b Ʌu) = 0 criteria, where b is the Burgers vector, g is the diffraction vector and u is a unit vector along the dislocation line. The dislocation direction was obtained through the corresponding selected-area electron diffraction (SAED) pattern. The index of the point in the SAED pattern represents the plane of the corresponding index in reciprocal space. Figure S2 shows the bright-field images of olivine and the corresponding SAED patterns of the studied samples. The observed slip systems in olivine grains from the Luobusa peridotites are summarized in Table 3.

Almost all the dislocations in the bright-field images of olivine from the Luobusa peridotites are straight (Fig. 9a–c), suggesting that the dominant deformation mechanism is dislocation creep at relatively low temperatures (either low temperature and low pressure, or high temperature and UHP). Dislocation loops (Fig. 9a), which are formed at high temperature or through dislocation cross-slipping (Raterron et al. 2007, 2009), are rare. The occasionally observed square-like figures (Fig. 9c) and dislocation tangles (Fig. 9d) are characteristic of low-temperature deformation at high pressure (Raterron et al. 2004). Dislocations are aligned along a particular direction and form sub-grain boundaries. The preservation of straight dislocations and sub-grains in olivine suggests that the Luobusa peridotites experienced limited annealing.

Water content

The representative FTIR spectra of olivine, Opx and Cpx from 17 harzburgite samples and seven dunite samples are given in Figure 10. The absorption bands of structural hydroxyl appear in the range 3400–3570 cm−1 for olivine (Fig. 10a). The infrared spectra of Opx are characterized by three hydroxyl absorption bands in the ranges 3410–3420, 3515–3525 and 3560–3570 cm−1, whereas those of Cpx are in the ranges 3450–3465, 3530–3535 and 3645–3648 cm−1 (Fig. 10b). The broad absorption band near 3400 cm−1 has been attributed to the inclusion of molecular water, whereas the absorption peaks near 3690 and 3717 cm−1 are caused by the OH stretching bands of serpentine. The absorption band at 3594 cm−1 matches the OH stretching bands of hydrous minerals, such as vermiculite group minerals (e.g. Khisina et al. 2001; Johnson et al. 2002; Beran & Libowitzky 2006; Aubaud et al. 2007). The very high absorptions between 3720 and 3750 cm−1 occur in some olivine grains (Fig. 10a) and may be caused by non-hydrogen-bonded OH, such as inclusions of a sheet silicate (Bell et al. 2003). The contribution of these extrinsic hydrogens was excluded in the baseline correction used to obtain the water content of olivine and pyroxene.

Table 3 shows that the water content of olivine is very low (6–25 ppm) with a mean of 16 ± 5 ppm. By contrast, both Opx and Cpx have high water contents. The water content of Opx is 56–141 ppm with a mean of 90 ± 21 ppm, whereas the water content of Cpx varies from 373 to 1273 ppm. If we exclude the extremely water-rich Cpx with 1273 ppm in harzburgite sample B153, the mean water content of Cpx is 492 ± 64 ppm.

Discussion

Implications of water content in Luobusa peridotites

Consistent with previous studies on the Luobusa ophiolite (Zhou et al. 1996, 2005; Dilek & Furnes 2011, 2014; Xiong et al. 2015), the studied harzburgites are abyssal peridotites subjected to 10–20% partial melting, whereas the more depleted dunites are SSZ peridotites that have been modified by melt–rock interactions in a mantle wedge (Fig. 5). The subducted oceanic lithosphere has lost most of its water at depths >200 km due to the continuous breakdown of hydrous minerals such as serpentine, talc and chlorite (Hacker et al. 2003). The mantle wedge above the subducted oceanic lithosphere generally has a high water content as a result of metasomatism and fluid infiltration, as evidenced by arc magmatism. However, olivine in 24 peridotite core samples from the Luobusa ophiolite is very poor in water, whereas Cpx is extremely rich in water relative to olivine and Opx (Table 3). The REE patterns of Cpx indicate that the Luobusa peridotites did not experience pervasive metasomatism after partial melting (Fig. 6).

Figure 11a shows that the diffusion of hydrogen in pyroxene is about 500 times slower than in olivine (Mackwell & Kohlstedt 1990; Ingrin et al. 1995; Stalder & Skogby 2003). Given the temperature dependence of hydrogen diffusion, the very low water content in olivine shows that widespread serpentinization of the Luobusa peridotites occurred at shallow depths. The combined δ13C and nitrogen data of microdiamonds from the Luobusa peridotites indicate that these microdiamonds were formed over a narrow and cold temperature range (<950°C) and were incorporated into the chromitites and peridotites near the mantle transition zone during a short residence time (i.e. within several million years) at high temperature in the deep mantle (Xu et al. 2018). Assuming 5 myr for the exhumation of microdiamonds and their host rocks from the transition zone to the shallow mantle depths, this implies a very fast exhumation rate of 6–8 cm a−1. Based on the hydrogen diffusion coefficients and fast exhumation rates, Cpx and Opx in the Luobusa peridotites have been subjected to limited hydrogen loss during exhumation, whereas olivine may have been subjected to significant hydrogen loss when temperatures exceeded 1000°C (Fig. 11b). This implies that olivine was strongly dehydrated when the peridotites were subducted to depths >200 km (i.e. the geothermal gradient is >5°C km−1) or when they entered the mantle transition zone.

When the slab reaches the mantle transition zone, the melt–rock interactions could significantly dehydrate olivine given the partition coefficient of hydrogen between the melt and olivine (Hirschmann et al. 2009). The partition coefficient of hydrogen between Cpx and Opx (DHCpx/Opx) is 5.56 ± 0.96 for the Luobusa peridotites. This value is much higher than DHCpx/Opx of 1.2–3.5 from previous measurements on natural peridotites (Bell & Rossman 1992; Bell & Ihinger 2000; Peslier et al. 2002, 2007, 2008, 2010, 2012; Peslier & Luhr 2006; Grant et al. 2007; Li et al. 2008; Yang et al. 2008) (Fig. 12). The experiments on hydrogen partitioning in mantle minerals show a trend of increasing DHCpx/Opx with pressures up to 4 GPa (Aubaud et al. 2004; Hauri et al. 2006; Tenner et al. 2009; Kovács et al. 2012; Rosenthal et al. 2015; Demouchy et al. 2017). The abundant exsolution lamellae of Cpx in Opx grains (Fig. 2g, h) from the Luobusa peridotite core samples suggest an origin at P > 7 GPa (Zhang et al. 2017). This is consistent with the estimation of P > 13 GPa from the coarse vermicular symplectites of Opx + Cr–Al spinel ± Cpx as breakdown products of majoritic garnet in harzburgite samples (Fig. 3) (Griffin et al. 2016). Therefore the extremely high DHCpx/Opx in the Luobusa peridotites probably benefits from a higher DHCpx/Opx at pressures >7 GPa and less hydrogen loss in Cpx than in Opx during rapid exhumation. It is worth noting the high water content of high-pressure phases of olivine, i.e. wadsleyite and ringwoodite, from volatile-rich kimberlite xenoliths (Pearson et al. 2014). Given the localized non-equilibrium phases in the Luobusa peridotites and chromitites, the transition zone could be heterogeneous in both composition and water content (Dilek & Yang 2018). The high water content of Cpx and Opx from the Luobusa peridotites show that the subducted oceanic lithosphere can transport a lot of water into the mantle transition zone via pyroxene.

Exhumation of Luobusa peridotites in a subduction channel

It is still unclear how the UHP and super-reducing phases were formed and preserved in the Luobusa peridotites and chromitites, which are generally regarded as relicts of the Neotethyan oceanic lithosphere and are characterized by the geochemical signatures of an SSZ environment (Griffin et al. 2016; Dilek & Yang 2018 and references cited therein). Although the mantle plume model (Yang et al. 2007) and the rapid channelized mantle upwelling model (Griffin et al. 2016) can explain the coexistence of shallow pre-metamorphic chromitites and peridotites and their metamorphism in the mantle transition zone at pressures >13 GPa, it is difficult to explain the widespread straight dislocations and sub-grains in olivine and the complex olivine fabrics in the Luobusa peridotites, which are easily modified at high temperature of mantle upwelling. In addition, the preservation of crustal minerals (e.g. zircon, quartz, corundum and K-feldspar) in ophiolitic peridotites and chromitites suggests the mixing of subducted continental crustal materials with oceanic lithosphere in a subduction channel (Yang et al. 2014; Robinson et al. 2015; Dilek & Yang 2018).

All the studied peridotites are coarse grained with granoblastic or porphyroclastic textures, which excludes stress-induced B-type fabrics in mylonitic peridotites (Wang et al. 2013a) or E-type fabrics in a hydrated ductile shear zone at the Moho transition zone (Michibayashi & Oohara 2013). Deformation experiments on olivine aggregates show different fabric types with increasing temperature and pressure. Carter & Avé Lallemant (1970) deformed dunites and peridotites under conditions of 0.5–3 GPa and 300–1400°C in the presence or absence of water. They found that the predominant slip systems in olivine changed from [001]{110} at low temperatures to [100]{0kl}(D-type fabric) and then [100](010) (A-type fabric) with increasing temperatures. Their observations were confirmed by the A-type fabric of olivine aggregates in simple shear deformation experiments at 300 MPa and 1300°C (Zhang et al. 2000). However, deformation experiments on single crystals of forsterite using a multi-anvil apparatus showed that UHP will favour [001] slip in olivine at pressures >7 GPa temperatures of 1200–1400°C (Couvy et al. 2004; Raterron et al. 2007, 2009). More recent deformation experiments have found that the A-type fabric in dry olivine aggregates changes to a B-/C-type fabric at 7.6 GPa and 1400°C (Ohuchi et al. 2011). However, at 7.2–11.1 GPa and 1127–1500°C, olivine will develop A- and B-type fabrics under water-rich and moderately wet conditions, respectively, whereas the C-type fabric is dominant under water-poor conditions at pressure >9.6 GPa and 1400°C (Ohuchi & Irifune 2013). Although we cannot exclude the possibility that different water contents in olivine affected the development of fabrics during exhumation, the preservation of straight dislocations (Fig. 9) and the general consistency between olivine fabrics and slip systems (Table 3) suggest the deformation of olivine under water-poor and relatively low-temperature conditions because a high water content and high temperature will facilitate the modification of dislocations. Compared with deformation experiments (Couvy et al. 2004; Raterron et al. 2007, 2009; Ohuchi et al. 2011; Ohuchi & Irifune 2013), the B- and C-type olivine fabrics in the Luobusa peridotites could be formed at pressures >7.6 and >9.6 GPa, respectively, whereas the A- and E-type olivine fabrics could be formed by dynamic recrystallization in the shallow mantle.

The ophiolites along the YZSZ were first obducted onto the continental margin of proto-India or onto a series of intra-oceanic island arcs (Aitchison et al. 2007; Hébert et al. 2012). The Luobusa ophiolite was formed at a mid-ocean ridge in the Mid-Jurassic (Zhou et al. 2002; Zhong et al. 2006) and then re-fertilized by boninitic melts in a mantle wedge above the subduction zone at c. 126 Ma (Malpas et al. 2003; Zhou et al. 2005, 2014). The 40Ar/39Ar dating of amphiboles from the Luobusa amphibolites yields amphibolite facies metamorphism at 90–80 Ma, which was interpreted as the initial tectonic displacement of the Luobusa ophiolite (Malpas et al. 2003). By contrast, zircon U-Pb ages of gabbro and amphibolite from the Luobusa ophiolite yield 128–131 Ma (C. Zhang et al. 2016), comparable with the 40Ar/39Ar cooling ages of amphiboles at 127–124 Ma for strongly foliated amphibolites in the Xigaze ophiolite in the central YZSZ (Guilmette et al. 2009). This implies multiple emplacement events before the closure of the Neotethyan Ocean along this suture zone. Slab rollback and southwards trench migration at c. 130–120 Ma has been proposed to explain the accretion of the Neotethyan forearc lithosphere in Tibetan ophiolites (Xiong et al. 2016).

Based on these geochronological constraints and the multiple magmatic events of the Gangdese arc, we propose a model in which channel flow induced by slab rollback can be used to explain the mixture of peridotites with different olivine fabrics and UHP minerals in the Luobusa ophiolite. As shown in Figure 13a, the Luobusa ophiolite probably represents the remnants of forearc oceanic lithosphere above a north-dipping subduction zone. The Neotethyan oceanic lithosphere was subducted to the transition zone in the Early Cretaceous. The slab rollback at 130–120 Ma formed a new subduction system to the south of the original system and triggered very rapid exhumation of the Luobusa peridotites and chromitites along the subduction channel between the subducting oceanic crust and the big mantle wedge. The return flow in the oceanic subduction channel resulted in a mixture of peridotites with different origins and fabrics, including peridotites and chromitites derived from the transition zone, UHP phases bearing SSZ peridotites and chromitites from the big mantle wedge, and SSZ peridotites that had been trapped in the subduction channel at shallow depths (Fig. 13b). This scenario is comparable with the dual origin of the Dabie–Sulu peridotites (Liou et al. 2000) and the exhumation of UHP rocks in continental subduction zones (Zheng 2012), but the return depth of the subducted materials is deeper than expected.

Compared with the mantle plume model, the exhumation of peridotites and chromitites in an oceanic subduction channel occurred at relatively low temperatures and a high exhumation rate. This model allows the addition of crustal minerals to peridotites and chromitites, the preservation of fabrics and dislocations in olivine, and the preservation of the δ13C and N signatures of microdiamonds. The subduction channel provides an important pathway for the transportation of subducted oceanic and continental materials from the mantle transition zone to shallow depths. In addition, due to the very low water content in olivine and the lack of hydrous minerals at depths >200 km, the water released from the subduction channel to the mantle wedge was limited during exhumation of the UHP rocks, which explains the interruption of magmatism in the Gangdese arc between 130 and 120 Ma. Therefore our model supports the exhumation of diamonds and UHP minerals by channel flow driven by slab rollback in forearc settings (Dilek & Yang 2018).

In a similar manner to the different fabrics seen in water-poor olivine from peridotites in the Western Gneiss Region (Norway) (Wang et al. 2013a), the Luobusa peridotites provide evidence for the water-independent development of fabrics in a subduction channel. Given the recent discovery of microdiamonds and UHP minerals in other ophiolites (Yang et al. 2014; Dilek & Yang 2018), it appears that the oceanic lithosphere has a more complex deformation history than that predicted based on plate tectonics theory. Further investigations of microstructure of ophiolites from different tectonic settings is required to investigate the heterogeneity of the oceanic lithosphere and the recycling of material through subduction channels.

Conclusions

We studied the composition, microstructure and water content of dunite and harzburgite core samples from the Luobusa Scientific Drilling Project to explore the complex deformation history of the diamond-bearing Luobusa ophiolite. All the peridotite samples show an SSZ signature and equilibration temperatures at c. 950–1080°C. The development of A-, B-, C- and E-type fabrics of olivine and the preservation of straight dislocations in olivine suggest their deformation at relatively low temperatures. Olivine is water-poor, with a mean water content of 16 ± 5 ppm due to the loss of hydrogen, whereas Opx and Cpx preserve high water contents and high hydrogen partition coefficients (DHCpx/Opx = 5.56 ± 0.96). The REE patterns of Cpx exclude the possible hydration of the Luobusa peridotites by mantle metasomatism. The subducted oceanic lithosphere can therefore transport a lot of water into the mantle transition zone via pyroxene. The B- and C-type fabrics in water-poor olivine represent a fossil fabric formed at great depths, whereas the A- and E-type fabrics were formed at shallow depths during exhumation. Our results, combined with previous studies, we propose channel flow driven by slab rollback to explain the mixture of different olivine fabrics, the extremely high DHCpx/Opx, and the occurrence of UHP and super-reducing mineral inclusions in the Luobusa peridotites and chromitites.

Acknowledgements

We are grateful to Y. Dilek for his invitation and encouragement to submit this paper, and for discussions with Y. Dilek, R. Wirth, W.L. Griffin and S.Y.O'Reilly. Constructive comments from T. Tsujimori and an anonymous reviewer were very helpful in improving this paper.

Scientific editing by Yildirim Dilek

This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License (http://creativecommons.org/licenses/by/4.0/)