The studied samples consist of picrites (n = 4), LT (n = 14), and HT (n = 47) basalts. The picrites were collected from Jianshui (southern Yunnan) and Dali (western Yunnan) (Fig. 1b). The LT basalts are from Jianshui, Dali, Wenshan (southern Yunnan), Lijiang (northern Yunnan), and Miyi (southern Sichuan) (Fig. 1b). The HT basalts are sampled from Wenshan, Jianshui, Kunming (central Yunnan), Dali, Lijiang, Miyi, Xuanwei (northeastern Yunnan), Leibo (southern Sichuan) and Weining (northwestern Guizhou), covering the most part of the ELIP (Fig. 1b). The samples show a porphyritic texture, with olivine, clinopyroxene and plagioclase as phenocrysts (Supplementary Figs. 2, 3). Only samples without mineral alteration were selected for further elemental and isotopic analysis (Table S1 in Supplementary Data 1).

Boron elemental and isotopic compositions of the ELIP picrites and basalts

The ELIP samples generally have a wide range of B concentrations from 1.0 μg/g to 53.2 μg/g and δ11B from −13.1‰ to +27.6‰.The ELIP picrites have boron concentrations ranging from 5.8 to 13.2 μg/g (avg. 10.1 ± 7.7 μg/g, 2sd) and δ11B from +17.1‰ to +27.6‰ (avg. +21.2 ± 11.2‰, 2sd). The HT basalts have variable boron concentrations (1.0–53.2 μg/g; avg. 13.1 ± 23.2 μg/g, 2sd) and isotopes (δ11B = −12.8‰ to +18.4‰; avg. +2.6 ± 18.5‰, 2sd). The LT basalts have boron concentrations of 2.2–19.4 μg/g (avg. 7.7 ± 10.6 μg/g, 2sd) and δ11B of −13.1‰ to +7.9‰ (avg. −4.5 ± 14.4‰, 2sd). Overall, the majority of the ELIP samples have higher B concentrations and δ11B values than those of MORBs and OIBs, but overlap with those of mafic arc magmas25,31 (Fig. 2). For a given MgO, the studied samples have variable B concentrations and δ11B values and trace element ratios, such as B/Ce (Supplementary Fig. 4). The ELIP samples show a wide range in boron isotopic compositions, while their radiogenic isotopes exhibit a relatively narrow range (Fig. 3). Nonetheless, δ11B values display a clear positive correlation with boron concentrations and B/Ce ratios (Figs. 2 and 4).

Fig. 2: The δ11B and B (ppm) composition of the ELIP picrites and basalts. a Plots of δ11B vs. B diagrams. b Kernel density function plot of δ11B for data in (a). Mid-ocean ridge basalt (MORB)25, Ocean island basalt (OIB)31, and mafic arc magmas24 data are shown for comparison. Full size image

Fig. 3: Boron and radiogenic isotopes systematics of the ELIP rock samples. Plots of δ11B vs. a 87Sr/86Sr, b 143Nd/144Nd, and c 207Pb/206Pb for the ELIP picrites and basalts. All radiogenic isotopes are initial ratios that were age-corrected to 260 Ma. Serpentinite fluid is assumed to be derived from seawater altered slab peridotite at ~3% serpentinization. Mixing curves between different end-member components (ROC, sediments, FOZO-like mantle, and serpentinite fluid) are calculated per their chemical compositions, summarized in Supplementary Table 1. The contents and isotopic compositions of B, Sr, Nd of seawater are obtained from the literature27,96,97,98. Data for the FOZO mantle and slab peridotite component are from the literature25,61,66,99. The contents and isotopic compositions of B, Sr, Nd, and Pb in recycled oceanic crust are compiled from the literature10,11,61. Data for subducted ancient sediments are sourced from the literature100,101. Full size image

Fig. 4: Boron isotopes and B/Ce ratios of the ELIP rock samples. The enriched mantle 1 (EM1) component is assumed as mixture of ~90% recycled oceanic crust (ROC) and ~10% ancient sediments in Fig. 3. Boron contents and isotopic compositions of the ROC and sediments are from the literature10,11. Elemental and isotopic compositions of the FOZO mantle component are from the literature25,99. The δ11B and B/Ce of slab serpentinites (light blue rectangle) are from the literature11,62,71. MORB and OIB (pink rectangle) and mafic arc magmas data sources are the same as in Fig. 2. Full size image

Possible sources for the heavy B isotopes in the Emeishan mantle plume

The observed heavy B isotopes may be derived from either a mantle source or shallow level processes, such as post-magmatic alteration, crustal contamination, subcontinental lithospheric mantle (SCLM) overprint, and magmatic differentiation. The ELIP samples generally have variable loss on ignition (LOI; 0.4–4.5 wt.%), indicating different degrees of post-magmatic alteration. Therefore, it is crucial to evaluate the effects of alteration on the elemental and isotopic compositions of the studied samples. Seawater alteration would produce elevated δ11B values as observed in a number of MORB glass samples25. However, the Emeishan mantle plume impinged on the South China continental lithosphere34. A rapid, kilometer-scale crustal doming prior to the eruption of the Emeishan flood basalts has been recorded by the thinned carbonates in the Middle Permian Maokou Formation that underlies the flood basalts39. The crustal uplift induced the complete regression of seawater from South China at the end of the Maokou stage39. On the other hand, the post-erupted Emeishan flood basalts were covered by the Late Permian or Triassic terrestrial clastic facies in the studied area39. Consequently, the Emeishan basalts may have a rare chance to be altered by seawater. Detailed microscopic observations on samples that had experienced different degrees of alteration (Supplementary Discussion) suggest limited influences of alteration on the boron elemental and isotopic compositions. Furthermore, no correlation has been observed between LOI and fluid mobile elements (e.g., B) and related isotopes (e.g., δ11B, 87Sr/86Sr i , and 206Pb/204Pb i ) (Supplementary Fig. 5a–d). The ELIP samples have low chemical index of alteration (CIA) values of 29–43, which are below the threshold (~50) for altered basalts50. The δ11B values also do not correlate with chemical weathering indices, such as CIA and Sr/Sc ratios (Supplementary Fig. 5e, f). Notably, the heavy boron isotope signatures are observed regardless of the values of LOI or chemical weathering indices (Supplementary Fig. 5b, e, f). Aqueous fluid tends to preferentially leach 11B during the post-magmatic hydrothermal alteration of basalts51, therefore, this process cannot account for the observed heavy boron isotopes in the ELIP samples. Most importantly, the B/Ce, B/Nb and δ11B, as well as the radiogenic isotopes, show good correlations with the indices of partial melting degree and source heterogeneity (e.g., La/Yb, La/Sm, Sm/Yb and Nb/Yb; Fig. 5 and Supplementary Fig. 6), demonstrating that the elemental and isotopic variations of the ELIP samples are mainly controlled by source components rather than post-magmatic alteration.

Fig. 5: Boron and radiogenic isotopes and trace element systematics of the ELIP rock samples. Plots of a δ11B, b 143Nd/144Nd, c 87Sr/86Sr, and d B/Ce vs. La/Yb for the ELIP picrites and basalts. The B/Ce ratios of the olivine-hosted melt inclusions in the picrite sample are plotted for comparison. “High-F melt, refractory lithologies” refer to a high fraction of melts from more refractory and depleted lithologies in the mantle source. “Low-F melt, fusible lithologies” refers to a low fraction of melts from more fusible and enriched lithologies in the mantle source. The literature data for the ELIP picrites and basalts for comparison in b are compiled from the database GEOROC. Full size image

Melt inclusions enclosed in early-formed olivine crystals are shielded from post-magmatic alteration, and are generally regarded as representing nearly unevolved, primary melt compositions compared with the whole-rock compositions5,52. Therefore, a time-efficient and effective approach to evaluating the influence of post-magmatic alteration on whole rock B elemental composition is to examine whether the B/Ce and B/Nb ratios in melt inclusions are consistent with those of the bulk rock. Olivine-hosted melt inclusions in picrite sample (23WS-77A-1) are oval, with sizes ranging from 40 to 60 μm in diameter (Supplementary Fig. 7). Prior to analysis, the melt inclusions were homogenized to glass by reheating and quenching them in a 1 atm furnace35. The elemental concentrations of melt inclusions are listed in Supplementary Data 1 (Table S2). The olivine-hosted melt inclusions in picrite have B contents ranging from 5 ppm to 28 ppm (avg. 11 ppm). They have B/Ce ratios of 0.23–2.58, B/Nb ratios of 0.51–8.91, and La/Yb ratios of 1.06–9.38. These ranges are very similar to those of bulk picrites (Fig. 5e, f). These results further support the idea that the boron elemental and isotopic compositions of the bulk rocks were inherited from the source and have not been altered by post-magmatic alteration. Considering the possible interactions between mantle-derived magmas and the continental lithosphere during their ascent to the surface, it is necessary to evaluate the influence of crustal contamination and SCLM overprint on the elemental and isotopic compositions of studied ELIP samples. The South China continental crustal materials are characterized by relatively high SiO 2 and Th/La, and low Nb/La and non-radiogenic Nd isotopes53 (Supplementary Fig. 8). Asthenospheric mantle-derived melts that have undergone marked crustal contamination would show negative correlations between SiO 2 and Nb/La, and εNd(t), and between Th/La and εNd(t), which were not observed in our studied samples (Supplementary Fig. 8). Moreover, the relatively light boron isotopic composition of the continental crust (δ11B = −9 ± 2.4‰)25 makes crustal contamination an improbable explanation for the heavy boron isotopes observed in the ELIP samples (Supplementary Fig. 8). Sedimentary carbonates, such as those found in the Maokou Formation, are widely observed to underlie the Emeishan CFBs. Since marine carbonates typically exhibit elevated B isotope ratios (δ11B = +3.4‰ to +26.2‰, avg. 16.5‰)54,55,56, it is necessary to consider the possibility that the high B isotopic compositions of the Emeishan magmatic rocks are inherited from the carbonate-rich strata. We have performed binary mixing calculations. Our modeling results suggest that mixing of the plume component with carbonates cannot adequately reproduce the observed Sr-B isotopic compositions in the ELIP samples (see Supplementary Figs. 8 and 9).

The SCLM underlying the western South China may have been modified by multiple paleo-subduction events such as the Neoproterozoic and Paleotethyan subduction systems57,58. The lower Ce/Pb ratios (16 ± 7, 1sd) observed in the studied samples, compared to those of OIB and MORB (25 ± 5, 1sd)59, could be attributed to the involvement of subduction-metasomatized lithospheric mantle components60. However, we argue that a continental lithospheric mantle source model fails to explain the B isotopic signatures of the studied samples for the following reasons: (1) Previous studies have suggested that the elemental and isotopic compositions of the olivine-hosted melt inclusions from Emeishan picrites were inherited from the mantle source, and were unaffected by lithospheric mantle35. Their Ce/Pb ratios (15 ± 4, 1sd)61 are consistent with those of our samples (16 ± 7, 1sd), indicating that the low Ce/Pb ratios in the ELIP samples are also inherited from the mantle source, rather than being a consequence of lithospheric mantle contamination62. (2) The Late Permian South China lithospheric mantle was characterized by relatively unradiogenic Nd isotopes (εNd 1.59, B/Nb > 4.4, and δ11B > +27.62‰. These values are outside the range of known mantle composition (Fig. 4). The light boron isotopic signature of subducted sediments makes it an unlikely source for a heavy-B fluid endmember21. Large boron isotope shifts are more plausibly explained by low-temperature fractionation processes operating near the Earth’s surface. Seawater-induced alteration of oceanic crust and serpentinization of lithospheric mantle within the subducting slabs are the most common low-temperature processes with the potential to influence the elemental and isotopic characteristics of boron in mantle rocks13,68. While the AOC is characterized by heavy δ11B (avg. +3.4‰)68, its B, particularly 11B, is mostly lost during dehydration reaction at shallow depths ( + 7‰, reaching up to +40‰)71,72. Additionally, they also exhibit low Ce/Pb ratios (8.7–9.2)62 and represent a potential end-member component contributing to the low Ce/Pb ratios in the ELIP magmas. Unlike the AOC, slab serpentinites form the deepest and coldest core of the slab, allowing for boron transport into the deep mantle before complete dehydration13,18,73. There are two potential pathways through which serpentinite-derived fluids can reach the deep mantle. Firstly, in the subduction channel, serpentinized oceanic lithospheric mantle undergoes dehydration at sub-arc depths along a hot geotherm, releasing fluids that metasomatize the overlying mantle wedge11,74. The resulting metasomatized peridotite primarily experiences partial melting, giving rise to island arc magmatism, rather than being carried downward with the subducting slab into the deeper mantle11,74. Alternatively, slab serpentinites have the potential to be subducted to the mantle transition zone. In this case, as the cold subducting slab descends beyond the sub-arc mantle depth, serpentines would transform into metamorphic olivine or DHMS17,18, which could retain and transport water and FMEs to deeper mantle regions, such as the MTZ19. The heavy B isotopic compositions observed in the Emeishan picrites and basalts are consistent with the second scenario. Additionally, previous studies on B isotope fractionation during the transformation from serpentine to metamorphic olivine at high pressure have found that the secondary olivine mostly inherits the δ11B signatures of the antigorite precursor (Δ11B Ol-Atg of −0.7 ± 3.4 ‰)18, which implies that there is little B isotopic fractionation during serpentinite dehydration75,76. To date, evidence of B-isotope fractionation due to B loss during phase transitions among serpentine polymorphs and during olivine-in reaction at high pressures has not yet been documented in the natural rock records13,77.

Therefore, our new B elemental and isotopic data strongly suggest a contribution from recycled subdcuted slab serpentinites in the source of the Emeishan mantle plume. We integrated our new B isotope data into the previously proposed source components (i.e., FOZO-like DM and EM1)35,61 for the Emeishan mantle plume. The B-Sr-Nd-Pb isotopes of the ELIP samples can be modeled as a three-component mixing array (Fig. 4). In addition to the previously proposed source components for the Emeishan mantle plume, including FOZO-like DM and EM1 components, the addition of 0–5 wt.% serpentinite-derived fluids is essential (Figs. 3, 4). The ELIP magmatic rocks exhibit a bimodal distribution of B isotopic compositions, with only a limited number of samples (n = 5 over a total of 57) having values of δ11B between −2‰ and +5‰ (Fig. 2). This may reflect incomplete mixing between low- and high-B mantle end-members. Our model posits that the fluids were mainly sourced from slab peridotite that had undergone approximately 3% serpentinization. The degree of slab serpentinization can be high, but when it reaches approximately 20%, the serpentinized peridotite becomes increasingly buoyant and is more frequently exhumed, possibly as a result of buoyancy processes73,78. As a result, slab lithospheric mantle peridotites that experience deep subduction generally exhibit low degrees (<20%) of serpentinization73. Most importantly, modeling results that incorporate a ~ 3% serpentinized mantle as an endmember align well with observed trends in the isotopic and elemental (e.g., B/Ce ratios) compositions (Figs. 3, 4). Our new mass balance modeling highlights the contributions of serpentinites. The elevated δ11B values, coupled with increasing B/Ce and B/Nb ratios, indicate the involvement of boron-enriched fluid in the source (Fig. 3). This fluid component is similar to the serpentinites in that they have high δ11B, B/Ce, and B/Nb13,62. In particular, the B isotopes display an independent behavior relative to those of Sr, Nd, and Pb (Fig. 4), probably due to the high fluid mobility of B. This further suggests that aqueous fluids may have played a critical role in the formation of the ELIP melts.

Deep water cycling via subducting serpentinized lithosphere

While thermal modeling and experimental results suggest that it is possible16,17, the geochemical pathway for subducted serpentinite transporting surface water into the deep Earth has not been definitively tracked. Most OIBs show no detectable heavy B isotope signals from surface fluids31. Only limited evidence is available from the boron isotopes: i) those in blue (Type IIb) diamonds, which formed at the transition zone to lower mantle depths, and ii) those in the Icelandic OIBs32,79,80. The slightly elevated B isotope in blue diamonds (δ11B = −9.0‰ to −0.5‰) and Icelandic basalts (δ11B = − 11.6‰ to −1.0‰) compared to typical OIBs is interpreted to reflect minor fluid contributions from subducted oceanic slabs in the deep mantle32,79,80. The distinctive boron isotopic signatures between modern OIBs and ELIP samples may indicate their different geodynamic inheritance. Modern OIBs typically originate from mantle reservoirs that are distant from, and therefore minimally influenced by, subduction-related metasomatism30. In contrast, the high δ11B, B/Ce and B/Nb signatures observed in Emeishan CFBs and picrites likely reflect the involvement of slab serpentinite-derived fluids (which can be up to 5 wt.%) in the source (Figs. 3, 4), and thus indicate a distinctive Paleozoic tectonic history of the South China Block. Previous studies proposed that the upper and lower mantle are relatively anhydrous due to the low solubility of water in upper and lower mantle minerals81. In contrast, the MTZ between them is considered to be the primary water reservoir in the solid Earth due to the capacity of the higher-pressure polymorphs of olivine, such as wadsleyite and ringwoodite, to incorporate notable amounts of water4. Slab subduction could lead to hydration of the MTZ at least locally beneath subduction zones3,82. The extent of this hydration largely depends on the age and velocity of the subducting slab8. While water in the young and hot slab can be expelled into the upper mantle, rapid subduction of an old and cold slab is more effective in transporting surface water deeper into the MTZ83, and this process can trigger intraplate volcanism showing some geochemical signatures similar to arc lavas, but located far away from the trench3,82. The MTZ is therefore the most likely source of water for the Emeishan mantle plume. During the Paleozoic, masses of continental blocks detached from Gondwana and then drifted north towards Eurasia84. This process led to the opening of the Paleo-Tethys ocean during the Early Ordovician and to a long period of rapid and cold oceanic subduction, that lasted from the Late Ordovician to the Triassic85,86(Fig. 6a). This would allow the hydrated slab to transport large amounts of water into the MTZ4. Therefore, we envisage that the MTZ, at least locally beneath the South China Block, is likely to contain a considerable amount of surface water transported by subducted slab serpentinites during the Middle/Late Permian (Fig. 6b).

Fig. 6: Schematic diagrams showing the plate configuration and deep Earth water recycling. a The Late Permian global plate configuration with the paleogeographic position of the South China Block and the Emeishan mantle plume. The plate configuration is based on Scotese et al.102. b Geodynamic model showing the deep water recycling via subduction of slab serpentinites. The plume (red column) passes through the transition zone, which contains subducted slab serpentinite-derived H 2 O and B in DHMS and/or metamorphic olivine. The plume may entrain hydours materials (light blue circles) from the transition zone5, thereby introducing H 2 O and B into the mantle plume. Further ascent of the plume generates more melt during decompression (large ellipse). The blue shapes within the ellipse represent the fluid component in the mantle source, while the grey and dark shapes reflect the fusible and depleted mantle components of the source, respectively. Full size image

Our study highlights the role of slab serpentinites in deep water recycling. Subduction of serpentinized slab lithospheric mantle provides an efficient pathway for transporting surface volatiles, including water and FMEs, into the deep mantle. On the other hand, a hot and dry mantle plume can trap water as it rises through the hydrous MTZ, and returns it to the Earth’s surface5 (Fig. 6b). The contrast in water solubility between the upper mantle and the MTZ leads to dehydration melting when a mantle plume enters the upper mantle87. The ELIP samples with more enriched Sr-Nd-Pb isotopic compositions tend to show lower δ11B, B/Ce, and B/Nb ratios. Notably, the B-Sr-Nd-Pb isotope, as well as the B/Ce and B/Nb ratios of the ELIP samples all exhibit systematic correlations with indices of partial melting degree and source heterogeneity (e.g., La/Yb, La/Sm, Sm/Yb, and Nb/Yb; Fig. 5 and Supplementary Fig. 6). These observations suggest that the serpentinites-derived fluids in the mantle plume vary on small spatial and temporal scales. The hydrated mantle region will undergo a higher degree of partial melting, leading to enhanced melting of more refractory depleted components. The resulting melts are characterized by higher B/Ce, B/Nb, δ11B, 143Nd/144Nd, and 206Pb/204Pb, and lower La/Yb, Ti/Y and 87Sr/86Sr (Figs. 5, 6b and Supplementary Fig. 10). In contrast, melting in dry mantle regions will result in a low fraction of melting, with the fusible EM1-like enriched components dominating the resulting melts. These melts have lower B/Ce, B/Nb, δ11B, 143Nd/144Nd, and 206Pb/204Pb, and higher La/Yb, Ti/Y and 87Sr/86Sr (Figs. 5, 6b and Supplementary Fig. 10). Our results further demonstrate that the ELIP LT and HT magmas were most likely originated from a similar source given their quite similar B-Sr-Nd-Pb isotope ranges. Therefore, no correlations are observed between isotopic compositions and HT/LT basalts.

Identifying modern analogues is challenging, because they must satisfy two critical criteria: a prolonged cold slab subduction in the region and subsequent ascent of a mantle plume behind the trench. One potential modern analogue may be the Columbia River LIP, where the Miocene CFBs are thought to be intimately linked to the Yellowstone mantle plume88,89,90. Prior to the Columbia River LIP eruption, the Farallon oceanic slab underwent prolonged subduction beneath the North American plate, possibly extending into the mantle transition zone91,92. Therefore, similar heavy boron isotope signatures may be expected to occur in the Columbia River CFBs.