A ‘hidden’ 18O-enriched reservoir in the sub-arc mantle

Plate subduction continuously transports crustal materials with high-δ18O values down to the mantle wedge, where mantle peridotites are expected to achieve the high-δ18O features. Elevated δ18O values relative to the upper mantle value have been reported for magmas from some subduction zones. However, peridotites with δ18O values significantly higher than the well-defined upper mantle values have never been observed from modern subduction zones. Here we present in-situ oxygen isotope data of olivine crystals in Sailipu mantle xenoliths from South Tibet, which have been subjected to a long history of Tethyan subduction before the India-Asia collision. Our data identify for the first time a metasomatized mantle that, interpreted as the sub-arc lithospheric mantle, shows anomalously enriched oxygen isotopes (δ18O = +8.03 ± 0.28 ‰). Such a high-δ18O mantle commonly does not contribute significantly to typical island arc basalts. However, partial melting or contamination of such a high-δ18O mantle is feasible to account for the high-δ18O signatures in arc basalts.

are trapped by the Sailipu ultrapotassic lavas in Lhasa Block from the southern Tibet Plateau (Fig. 1), which were erupted at ca 17 Ma (ref. 22). The Sailipu mantle xenoliths are small in size, with diameters commonly less than 2 cm. It has been suggested that they represent a relict mantle of the Asian lithosphere, which has been metasomatized by fluids/melts liberated from the subducted Tethys ocean plate and/ or the Indian continent 23 . In this study, olivines from nine Sailipu peridotite xenoliths have been analyzed for oxygen isotopes by secondary ion mass spectrometry (SIMS).

Results
The Sailipu mantle xenoliths mainly consist of olivine and orthopyroxene (Supplementary Text). All samples except SLP105 contain phlogopite, whereas clinopyroxene has been discovered in five samples. Spinel has only been found in three samples (e.g., SLP105, SLP127 and SLP154). Spinel in both SLP127 and SLP154 is interstitial among silicates, whereas it is included in orthopyroxene in SLP105. Phlogopite occasionally shows reaction textures with spinel.
Olivine grains from each studied xenolith have homogeneous oxygen isotope compositions (Supplementary Table S3), as indicated by small inter-and intra-grain variations that are always less than the precision of the method (less than 0.4%). Olivines from eight samples have d 18 O values varying from 15.22% to 15.41% (Fig. 3), with an average of 5.27 6 0.24% (n 5 221). Their d 18 O values plot within the range of the upper mantle values defined by both mantle-derived magmas and ultramafic rocks (5.18 6 0.28%; ref. 8,9). In contrast, olivines from sample SLP105 with the highest Fo content also have the remarkably elevated d 18 O values of 8.03 6 0.28% (n 5 72). To our knowledge, this is the highest value so far reported for mantle peridotites.

Discussion
Olivine in sample SLP105 could achieve the elevated d 18 O value from the host lava, as high d 18 O values up to 9.4% have been reported for the Sailipu ultrapotassic rocks 22 . Diffusion and reaction are two potential ways for host lavas to affect oxygen isotope compositions of the entrained mantle xenoliths. However, we suggest that neither process can account for the high d 18 O value of SLP105, as explained below. Mantle xenoliths with high temperatures are more likely to achieve oxygen isotopic equilibrium with the host lavas through diffusion; thus, high-temperature mantle xenoliths are expected to achieve the oxygen isotope signatures of the host lavas. However, the equilibrium temperature of sample SLP105 estimated using the Alin-orthopyroxene geothermometer 24 (see Supplementary Table 1) is 753uC (Fig. 4a), which is significantly lower than temperatures estimated for the other eight samples (i.e., 891-1072uC). Therefore, the oxygen isotope compositions of sample SLP105 should be least  We suggest that olivine in SLP105 obtained its elevated d 18 O value in the mantle through metasomatic processes by an 18 O-rich agent. Sailipu locates on the southern margin of the Lhasa block, which experienced a prolonged history of oceanic subduction (i.e., the Neo-Tethys Ocean) before the India-Asia collision at ca 55 Ma (ref. 20,21). Therefore, subduction is the most likely mechanism to bring such an 18 O-rich metasomatic agent to the mantle beneath Sailipu. Trace element compositions of clinopyroxenes from five samples support that they have been highly metasomatized (Fig. 2). They show enriched REE patterns (Fig. 2a) and are also strongly enriched in large ion lithophile elements (LILE), but depleted in high strength field elements (HFSE; Fig. 2b). In combination with the presence of phlogopite in Sailipu mantle xenoliths, such trace element characteristics of clinopyroxene have been explained as a result of metasomatism by slab-derived hydrous melts 23 . However, olivines in these five clinopyroxene-bearing xenoliths and other three clinopyroxene-free xenoliths have d 18 O values the same as the upper mantle. Therefore, a possible scenario is that slab-released hydrous melts, originally with high d 18 O values, were in oxygen isotopic equilibrium with the ambient mantle prior to the metasomatism 25 . Orthopyroxenes from these five clinopyroxene-bearing xenoliths (i.e., SLP113) also show variably LREE-enriched patterns, which are in stark contrast to the depleted pattern displayed by orthopyroxene in sample SLP105 (Fig. 2a). This implies that trace elements of  SLP105 have not been enriched by the metasomatic agent. However, the high d 18 O value in olivine requires that it has been metasomatized by a 18 O-rich agent. Such a decoupling between trace elements and oxygen isotopes could be reconciled if sample SLP105 was metasomatized by slab-released fluids rather than melts. It has been experimentally demonstrated 26 that slab-released fluids could contain trace elements three orders of magnitude less than the slab-derived melts at pressures less than 4 GPa. Therefore, we suggest that sample SLP105 had been metasomatized by slab-released fluids that were not in oxygen isotopic equilibrium with the ambient mantle before metasomatism. Sample SLP105 could achieve its high d 18 O value in the fore-arc mantle through exchange with hydrous 18 O-rich fluids liberated from the subducted slab at a high fluid/rock ratio 17 . A fore-arc mantle origin of SLP105 is consistent with its low temperature and the refractory composition. However, Sailipu locates at the back-arc areas of the Gangdese arc that was developed during subduction of the Neo-Tethys Ocean, which argues against that sample SLP105 was stemmed from the fore-arc mantle. Furthermore, absence of serpentine in SLP105 also does not favor its origin in a fore-arc mantle, because it has been suggested that the hydrated fore-arc mantle is mainly composed of serpentinites [27][28][29] . Peridotites in the mantle wedge just above the subducting slab could also be highly metasomatized by slab-released fluids/melts and thus have elevated d 18 O values, which has been proposed as the source for high-d 18 O phonolitic melts entrained in mantle olivines from the Tabar-Lihir-Tanga-Feni arc 14 . In this scenario, however, it is hard to explain how such metasomatized peridotites were incorporated into the continental mantle lithosphere beneath Sailipu. Finally, we suggest that sample SLP105 were derived from the sub-arc lithospheric mantle above the mantle wedge, which has been metasomatized by slab-derived fluids. Normally, slab-released fluids should percolate the asthenospheric wedge before they metasomatize the sub-arc lithospheric mantle. During transportation in the mantle wedge, they tend to achieve oxygen isotopic equilibrium with the ambient mantle, and thus, cannot preserve their high-d 18 O signatures. Nevertheless, the asthenospheric wedge disappears in the flat subduction setting 30 . Therefore, the slab-released fluids could directly permeate and also transfer the high-d 18 O signatures to the overriding lithospheric mantle. Flat subduction of the Neo-Tethys slab has been previously proposed on the basis of magmatism developed in the Gangdese arc 31,32 .

Lhasa block
Our results support that some portions of the sub-arc lithospheric mantle have d 18 O values higher than the upper mantle. Such mantle reservoirs have refractory compositions and commonly do not significantly contribute to typical island-arc basalts in normal subduction zones. This might be the reason that such an 18 O-enriched mantle reservoir is ''invisible'' from oxygen isotope compositions of typical island-arc basalts 13 . In anomalous tectonic settings like arc rifting, however, this non-convecting cold mantle could melt to produce arc magmas with high d 18 O values 17 . On the other hand, arc magmas could also achieve the 18 O-enriched signatures through interaction (e.g., contamination) with shallow high-d 18 O lithospheric mantle during ascending to surface 18 .

Methods
Major elements were measured on individual minerals using a JEOL JXA-8100 electron microscope with an accelerating potential of 15 kV and sample current of 10 nA at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS). Trace elements of clinopyroxene and orthopyroxene were analyzed using a laser ablation inductively coupled mass spectrometer (LA-ICP-MS) at IGGCAS. The LA-ICP-MS system consists of a 193 nm Geolas Pro laser coupled to an Agilent 7500a ICP-MS. Isotopes were measured in peak-hopping mode. A repetition rate of 8 Hz was used during analyses for all minerals. A spot size of 80 mm was used for the analyses of clinopyroxene, whereas a spot size of 120 mm was used for orthopyroxene. The NIST 610 glass was used as an external calibration standard and isotope 43 Ca was selected as an internal standard to quantify the analyses. The CaO content of NIST 610 used in the calculation is 11.45 wt.%. The data were reduced using the GLITTER 4.0 program.
Oxygen isotope compositions of olivine were analyzed in situ using CAMECA IMS-1280 ion microprobe at IGGCAS. Sample grains were prepared in epoxy adjacent to grains of a San Carlos olivine intralaboratory standard and then polished to a

Additional information
Supplementary information accompanies this paper at http://www.nature.com/ scientificreports Competing financial interests: The authors declare no competing financial interests.