Introduction

In the framework of plate tectonics, subduction introduces crustal material deep into Earth’s interior1. Due to the non-equilibrium transformation of pyroxene to majoritic garnet, slab viscosity, and/or the presence of a reduced viscosity layer at the phase change boundary, some subducting slabs may stop sinking in the mantle and instead stagnate in the mantle transition zone1,2,3,4 (MTZ, the depth interval with sharp seismic discontinuities around 410 and 660 km)5. Seismic observations and mineral physics models6,7,8,9,10,11 show that segregation of oceanic crust from ancient or adjacent subducted slabs can result in accumulation and formation of enriched basaltic reservoirs in the MTZ over large lateral distances far from the imaged stagnant slabs. Geochemical studies of inclusions within ultradeep diamonds and intraplate volcanism12,13,14,15,16,17,18,19,20 suggest that the Enriched Mantle end-member 1 (EM1, characterized by notably low 206Pb/204Pb and 143Nd/144Nd ratios, coupled with intermediate 87Sr/86Sr compositions)21,22 or HIMU (high μ = 238U/204Pb) type reservoirs could reside in the MTZ, leading to a causal link between subducted material and geochemical heterogeneity. In addition to the enrichment of basaltic crust and sediments, the MTZ might also act as an enormous deep reservoir for volatiles19,23,24,25,26 (e.g., hydrogen, nitrogen, and halogens) given the high storage capacity of transition zone mineral phases19,27,28,29,30. However, the genetic connection between subducted material, volatiles, and enriched mantle endmembers stored in the MTZ remains poorly understood.

Mercury (Hg) is the only volatile metal that exhibits both isotopic mass-dependent fractionations (MDF) and mass-independent fractionations (MIF) in natural samples31. The MDF of Hg occurs in a variety of physical, chemical, and biological processes, while the MIF-Hg of odd isotopes (odd-MIF) are primarily caused by magnetic isotope effects during photochemical reactions at or near the Earth’s surface31,32. In general, the primitive mantle or the mantle least impacted by crustal recycling33 shows no MIF-Hg (i.e., ∆199Hg value of 0 ± 0.1‰), whereas MIF-Hg signatures of oceanic (e.g., seawater, pelagic sediment, and altered oceanic crust) and terrestrial (e.g., soil, coal, plant, and terrigenous sediment) environments produced by photoreduction of aqueous Hg2+ are characterized with positive and negative ∆199Hg values, respectively32. Given the lack of MIF-Hg during subduction metamorphism, mantle melting, and crystal fractionation33,34,35,36,37,38, the recognition of MIF-Hg in mantle-derived materials is therefore a very powerful tracer for surface-derived material recycled into the mantle33,35,39,40,41,42,43,44,45. Although a substantial quantity of Hg may be lost from the subducted slabs at forearc and sub-arc depths and returned to the surface via arc volcanism35,38,40, recent Hg isotope studies of ocean island basalts (OIB) suggest that some surface-derived Hg may survive subduction devolatilization and enter the deep mantle33,35. However, it remains uncertain whether recycled Hg isotope signatures can be isolated in the MTZ over extended periods.

Geophysical observations46,47 and magnetotelluric data48,49,50 (Fig. 1 and Supplementary Fig. 1), together with petrological and geochemical studies and numerical modeling17,18,49,50,51,52,53, strongly suggest that Cenozoic intraplate volcanism in northeast Asia is a direct manifestation of volatile-rich upwelling in the big mantle wedge above the stagnant slab in the MTZ54, rather than a direct derivation from the metasomatized sub-continental lithospheric mantle (see the “Methods” section for more details).

Fig. 1: Schematic map, electrical structure, and seismic tomography of the studied volcanic fields in northeast China49,97.
Fig. 1: Schematic map, electrical structure, and seismic tomography of the studied volcanic fields in northeast China49,97.
Full size image

a Map49 showing sampling locations (red triangles) and other intraplate volcanos (blue triangles). The grey triangles show the distribution of Holocene arc volcanoes. The white dashed lines show the depth contours of the Wadati–Benioff deep seismic zone in the subducting Pacific slab. The white solid line with small triangles is the location of the current Japan trench, and the large white arrow represents the subduction direction of the Pacific plate. b Horizontal slice of a 3D low-resistivity anomaly49, with the contours labeled with numbers are estimated water content (in ppm) in the asthenosphere based on existing resistivity and geotherm, and c vertical cross-section of S-wave velocity tomography97. The locations of the horizontal slice and the vertical cross-section are shown in (a) by the black dashed lines A-A’ and B-B’, respectively. WDLC Wudalianchi, EKS Erkeshan, KL Keluo, NM Nuomin, XGLH Xiaogulihe, XK Xunke, RFE Russian Far East, SL Shuangliao, LG Longgang, CBS Changbaishan.

Here we report Hg isotope data for the Cenozoic EM1-type continental intraplate potassic basalts (Supplementary Fig. 2) from five representative volcanoes in NE China (Fig. 1 and Supplementary Fig. 1). These basalts exhibit some of the least radiogenic Pb and Nd isotope signatures documented globally13, and are therefore commonly regarded as the continental analogs of typical EM1-type oceanic basalts exemplified by those from Pitcairn Island55,56,57. Although several geochemical studies have been conducted on these continental intraplate basalts, the origin of their EM1 signatures remains contentious. Recent evidence increasingly indicates that the EM1 signatures are unlikely to originate from delaminated lower continental crust, and instead necessitate contributions from recycled sediments13,52,58,59,60,61,62. Nonetheless, whether these sedimentary inputs are primarily pelagic or terrigenous remains unresolved. Our Hg isotope data show that surficial Hg from recycled terrigenous sediments is recognized in these EM1-type continental lavas, thereby providing key evidence that surface materials can survive subduction and be delivered to the MTZ for over a billion years and are responsible for the chemical heterogeneities therein.

Results and discussion

The total Hg (THg) concentrations in the studied samples range from 0.1 to 5.5 ppb, with an average of 1.5 ppb, comparable to the range previously reported for basaltic rocks worldwide35. The δ202Hg, ∆199Hg, and ∆201Hg values vary from −1.26 to 1.05‰, −0.23 to 0.00‰, and −0.09 to −0.03‰, respectively, all exceeding the corresponding 2 SD analytical uncertainties of 0.10‰, 0.06‰, and 0.08‰.

MIF-Hg signatures (negative ∆199Hg) derived from recycled terrigenous sediments

The lack of correlation between the loss on ignition and ∆199Hg or δ202Hg in intraplate potassic basalts in NE China (Supplementary Fig. 3) indicates that alteration had a negligible effect on the Hg isotope compositions of the studied samples. Crustal assimilation cannot explain the negative ∆199Hg and positive δ202Hg anomalies in these samples because the continental crust63 has ∆199Hg (0.03 ± 0.15‰) indistinguishable from the mantle values (0.00 ± 0.10‰) and has δ202Hg (−1.15 ± 1.01‰) lower than most of our samples. Also, the high La/Yb ratios of our samples (up to 120; Supplementary Fig. 4) preclude significant crustal assimilation because the continental crust64 has a low La/Yb of 10.5. We therefore suggest that the MIF-Hg signatures in our basalts derive from a deep mantle reservoir.

Figure 2 shows the negative ∆199Hg signatures of our samples differ from the values (0.00 ± 0.10‰) of the “primitive mantle” or mantle least impacted by crustal recycling33. They are also significantly different from mid-ocean ridge basalts (MORB) and arc basalts, which exhibit positive ∆199Hg signatures35. Odd-MIF of Hg isotopes is primarily generated by photochemical reduction of aqueous Hg2+ to gaseous Hg0 at Earth’s near-surface environment31,32. During this reduction process, a negative ∆199Hg signal is generated in the product Hg0 that enters the atmosphere, while a complementary positive ∆199Hg is preserved in the residual aqueous phase32. Due to its transient residence time, elemental Hg0 with negative ∆199Hg in the atmosphere will be transported globally and deposited eventually into the surficial terrestrial reservoirs such as soil, coal, plants, and terrigenous sediments. By comparison, marine reservoirs feature mainly positive ∆199Hg because of wet deposition of aqueous Hg2+ in open oceanic sediments31,65. The contrasting and complementary ∆199Hg values in marine and terrestrial sediments are direct manifestation of photoreduction fractionation in generation of these distinct Hg reservoirs31,32. Because of the magnetic isotope effect31, photochemical fractionation of aqueous Hg2+ also produces a slope of ~1 in the ∆199Hg−∆201Hg space. All the samples studied here show negative ∆199Hg values with a ∆199Hg/∆201Hg ratio of 1.08 ± 0.07 and fall well in the range of terrestrial sediments (Fig. 2). These observations suggest our basalts sample an EM1-type mantle source modified by recycled terrigenous sediments, which inherited the odd-MIF Hg signatures from photoreduction at Earth’s surface31. Negative ∆199Hg values down to −0.45 ± 0.01‰ have been previously documented in the Pitcairn EM1 OIBs33, and these complement the negative ∆199Hg data reported in the Chinese EM1-type basalts (Fig. 2). These observations provide unambiguous evidence for the involvement of recycled terrestrial sediments in the EM1 mantle source.

Fig. 2: Mercury isotope compositions of the studied intraplate basalts.
Fig. 2: Mercury isotope compositions of the studied intraplate basalts.
Full size image

a ∆199Hg versus δ202Hg. b ∆199Hg versus ∆201Hg. The red rectangle in (a) represents the range of primitive mantle or the mantle least impacted by crustal recycling. The gray line in (b) has a slope of 1.08 ± 0.07 (2σ), consistent with the magnetic isotope effect associated with photoreduction of aqueous Hg2+ at Earth’s surface32. Also shown for comparison are compiled data33,35,41,42 of marine sediments, terrestrial sediments, seawater, igneous rocks, MORBs, arc basalts, Tubuai HIMU basalt, CFBs, and basalts from Magellan seamount, Afar, Iceland, Samoa, and Pitcairn islands. Note that MIF-Hg isotopes can be used to effectively distinguish between pelagic (marine) sediments and terrigenous (continental) sediments, with the former displaying positive ∆199Hg values (blue triangles) and the latter negative ∆199Hg values (solid gray circles). Therefore, the negative MIF-Hg values observed in our basalt samples point to a contribution of terrigenous sediments.

MDF-Hg composition (positive δ202Hg) of the EM1 endmember

In addition to MIF-Hg, our samples also exhibit a large range of MDF-Hg (δ202Hg = −1.26‰ to +1.05‰). One of the most striking features of our Hg isotope data is that the measured δ202Hg values of basalts correlate with indicators of mantle heterogeneity, including Sr–Pb isotopes and incompatible trace element ratios, such as U/Pb (Fig. 3). Several processes, such as abiotic chemical reactions (e.g., chemical or photochemical reduction/oxidation), physical adsorption/dissolution/evaporation, metamorphic devolatilization, and magmatic degassing, can trigger MDF-Hg31. In particular, degassing of magmas during eruption is known to generate considerable δ202Hg variability, with the gas and residue enriched in light and heavy Hg isotopes, respectively34. However, the observed correlations between δ202Hg and 87Sr/86Sr, 206Pb/204Pb, and 208Pb/204Pb in our samples cannot be explained by degassing during magma fractionation or by any other igneous processes, albeit these processes may indeed be responsible for the scatter of the basalt data in these plots (Fig. 3). Instead, our data indicate that the range of δ202Hg is an intrinsic property of the mantle source and requires mixing between two (at minimum) distinct sources (Fig. 3).

Fig. 3: Variations of Hg-MDF with radiogenic isotopes and trace element ratios.
Fig. 3: Variations of Hg-MDF with radiogenic isotopes and trace element ratios.
Full size image

Plots of δ202Hg versus a 87Sr/86Sr, b 206Pb/204Pb, c 208Pb/204Pb, and d U/Pb for the studied intraplate basalts. The low and high δ202Hg endmembers are consistent with the involvement of the lithospheric mantle (LM) and the EM1 component, respectively. The average 87Sr/86Sr (0.70457 ± 0.00031) of LM13 is plotted in (a) for comparison. The best-fit regressions with R2 and p values and the 95% confidence intervals are shown.

Recent studies have shown that melt-lithospheric mantle interactions played an important role in modifying the composition of EM1-type continental basaltic magmas in NE Asia13,60,62,66,67, and a binary mixing process between a depleted endmember (lithospheric mantle) and an EM1-type component provides a plausible explanation for the correlations observed in the Chinese continental EM1 basalts13,62. The low δ202Hg endmember, with low Sr and high Pb isotope ratios (Fig. 3), is therefore attributed to the lithospheric mantle beneath NE China13. This suggestion is also supported by the subcontinental lithospheric mantle (SCLM)-derived13 high Nd and Hf isotopes, low Ba/Th and K/U ratios, and Hf anomalies in these EM1-type basalts (Supplementary Fig. 4). In turn, the high δ202Hg rocks with highly unradiogenic 206Pb/204Pb, 208Pb/204Pb, and enriched 87Sr/86Sr are consistent with a greater contribution from an EM1-type source (Fig. 3). These radiogenic isotope characteristics require long-term isolation of the EM1 component in the MTZ13,17, with low time-integrated 238U/204Pb and 232Th/204Pb, and high time-integrated 87Rb/86Sr ratios, consistent with the measured low U/Pb and high Rb/Sr ratios of this endmember (Supplementary Fig. 4). Therefore, these observations suggest that the EM1 endmember is characterized by an inherent high δ202Hg signature that is similar to the δ202Hg values (up to −0.01‰) of the Pitcairn EM1-type OIBs33.

The negative ∆199Hg values and the ∆199Hg−∆201Hg slope of ~1, which matches those of the atmospheric photo-redox cycle in both NE China samples and Pitcairn OIBs, suggest that recycling of ancient terrigenous sediments resulted in an isotopically heavy Hg into the EM1 endmember source. A Monte Carlo mixing simulation between the depleted lithospheric mantle and ancient terrigenous sediments can reproduce the broad trend observed in our samples in the plots of 202Hg versus 206Pb/204Pb and 87Sr/86Sr (Supplementary Fig. 5), further supporting the above conclusion. Note, however, that it is currently unclear whether the high δ202Hg signature required for the terrigenous sediments is an inherent sediment property or a derivative obtained through subduction-modification. Although a small fraction of modern terrigenous sediments has heavy Hg isotope values (δ202Hg can be as high as 1.5‰, Fig. 2), most of them still fall within the range of unmodified mantle (δ202Hg = −1.7 ± 1.2‰). Experiments have demonstrated preferential loss of light Hg isotopes from sediments during metamorphism, implying that loss of light Hg isotopes to the overlying mantle may occur, leaving isotopically heavy Hg to be subsequently recycled into the mantle36. This suggests that the most likely mechanism for recycled terrigenous sediments in the EM1 source to possess the isotopically heavy Hg is through metamorphic devolatilization at subduction zones, despite the Mariana lavas having slightly negative δ202Hg35.

Linking ancient terrigenous sediments and EM1 reservoir in the MTZ

Recent hypotheses suggest that the EM1 signature may be generated by long-term storage of subducted pelagic sediments in the MTZ13,17. However, as shown in Fig. 2, pelagic (marine) sediments predominantly exhibit positive ∆199Hg and ∆201Hg values, whereas our samples consistently yield negative values, indicating a closer affinity with terrigenous sediments. Our MIF-Hg data, therefore, clearly distinguish terrigenous sediments from pelagic sediments (Fig. 2) and unambiguously demonstrate that recycled terrigenous sediments were responsible for the negative ∆199Hg and elevated δ202Hg values observed in the EM1 endmember. During subduction, terrigenous sediments derived from the upper continental crust may undergo melting or relamination and return rapidly back to the sub-arc magmas or accreted/stored in the shallow lithosphere68,69. Nevertheless, our Hg isotope data demonstrate that terrigenous sediments can also survive the sub-arc processes and be subducted and accumulated in the MTZ, forming the EM1 reservoir17 therein.

Although our Hg isotope data provide unambiguous evidence for the presence of recycled terrigenous sediments in the EM1 mantle source, they do not by themselves give age constraints on these sediments. The EM1-type unradiogenic Pb isotope compositions require that terrigenous sediments be isolated in the MTZ without being fully equilibrated with the ambient mantle over long timescales13,17. The negative correlation between MDF-Hg isotopes and 206Pb/204Pb (Fig. 3), together with the supra-geochron 207Pb/206Pb signature of the EM1 component (Supplementary Fig. 6), indicates that the high-δ202Hg endmember is ancient (>109 years), possibly as old as 2.2 billion years13. This suggests that the terrestrial sediment signature observed in the NE China EM1 lavas is not related to recent sediments associated with the Pacific plate. Therefore, Cenozoic subduction of the Pacific slab perturbed the MTZ18, leading to liberation of ancient subducted materials with EM1 signatures stored therein and triggering the formation of these continental EM1-type basaltic magmas (Fig. 4).

Fig. 4: Conceptual cartoon illustrating the Earth’s deep mercury cycle.
Fig. 4: Conceptual cartoon illustrating the Earth’s deep mercury cycle.
Full size image

The Earth’s primitive mantle or the mantle least impacted by crustal recycling has ∆199Hg of 0.00 ± 0.10‰ and δ202Hg of −1.7 ± 1.2‰, respectively33. During subduction, the oceanic crust and marine sediments deliver oceanic Hg with positive MIF signature into the source of arc magmas35,40 whereas the remaining marine Hg is recycled back into the upper mantle as suggested by the positive ∆199Hg values in MORBs33,35. Some oceanic crust with positive ∆199Hg features can even be transported into the MTZ to form the HIMU reservoir12,15,16,20. Recycling of ancient terrigenous sediments with negative ∆199Hg signatures into the MTZ forms the EM1 reservoir that feeds continental EM1-type volcanism. Devolatilization of Hg from subducted oceanic crust and marine and terrigenous sediments will release light MDF-Hg to the shallow mantle36 and then carry the complementary high-δ202Hg signatures into the EM1 or HIMU components stored in the MTZ. Although a very limited number of EM1 and HIMU samples do show signals of MIF-Hg33, the near-zero ∆199Hg values in most mantle plume-derived OIBs and continental flood basalts indicate that negligible surficial Hg can be transported into the lower mantle33,35. However, the EM1 endmember preserved in the MTZ can be entrained into ascending mantle plumes forming the EM1-type OIBs. ULVZ ultra-low velocity zones, LLSVP large low-shear-velocity provinces, and DMM depleted MORB mantle. The figure is modified from ref. 104. The vertical direction of the figure is not to scale.

Implications for deep mercury cycle

From a global perspective, most of the subducted Hg should be returned back to the surface through slab dehydration, forming arc magmas and arc-associated epithermal Hg deposits with positive ∆199Hg signatures35,40, while the remaining Hg in the residual slab can be recycled into the upper35, or even the lower mantle33,35 (Fig. 4). Although a limited number of EM1 and HIMU samples show strong signals of MIF-Hg33, the near-zero ∆199Hg values in most of the OIBs and continental flood basalts sourced from deep-seated mantle plumes indicate that, overall, little surficial Hg is transported into the lower mantle33,35 (Fig. 4). As the MTZ is an important host for the EM1 and HIMU reservoirs12,15,16,17,20,52, we propose an alternative scenario where some recycled sediment-related MIF-Hg signatures in EM1 and HIMU OIBs likely resulted from entrainment of MTZ material into the mantle plumes, in addition to originating directly from the deep lower mantle (Fig. 4). Combined with the fact that subducted slabs along normal-to-warm geotherms would undergo partial melting of carbonated metabasalt or metasediment in the MTZ and loose most of their volatile cargo before penetration into the lower mantle23, we predict the MTZ may act as a deep Hg barrier that controls recycling of Hg in the mantle and Hg is therefore rarely carried beyond this depth. However, secular cooling of Earth’s ambient mantle can potentially allow a large fraction of subducting slabs enriched in heavy Hg to be transferred into the MTZ and even the lower mantle in the recent geological history. The delivery of Hg, likely within mercury-poor and heavy-Hg-enriched oceanic crust and sediment, to the MTZ is consistent with the capability of cold subducted slabs to carry other recycled volatile elements such as hydrogen, carbon, nitrogen, halogens, and helium to these depths, as is confirmed by the anomalous volatile signals of surface origin enveloped in MTZ-seated diamonds14,19,28. Future studies of Hg isotope composition of inclusions in ultradeep diamonds are therefore critical to understanding the deep Earth’s Hg cycling. The long-term preservation of the recycled Hg isotope signatures introduced into the MTZ over a billion years ago via ancient subduction, and their afterwards resurfacing through EM1 or HIMU magmatism, suggest that, except for the survival of volatile heterogeneities from early Earth’s accretion and subsequent late veneer34, volatile anomalies introduced by plate tectonics can be preserved in the MTZ for over a billion years.

Methods

Sample descriptions

Cenozoic continental intraplate volcanos are widely distributed in NE Asia (Fig. 1). The studied samples were collected from the Xiaogulihe, Wudalianchi, Erkeshan, Keluo, and Nuominhe volcanic fields, located on the northern margin of the Songliao basin and covered an area of over 3000 km2 (Fig. 1). All the new geochemical data are provided in Supplementary Data 1–4. Based on K-Ar dating, these intraplate basalts erupted from ~9.6 Ma in the Miocene to 1721 CE70,71,72. All samples used for Hg isotope analysis are fresh, and no identifiable secondary minerals can be observed. They exhibit porphyritic to aphyric texture and massive to vesicular structure. Olivines ±  clinopyroxenes are the most ubiquitous phenocryst phases, with plagioclase, leucite occurred as minor phases. The groundmass consists mostly of quenched glass + olivine + clinopyroxenes + feldspar + Fe-Ti oxides13,52,58,59,66. Geochemically, they are alkaline, potassic to ultrapotassic rocks and show OIB-like trace element patterns52,58,59,61,62,66,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86. Isotopically, they are featured with low radiogenic Nd and Pb and moderate enriched Sr isotope ratios, which are even more extreme than those of the Pitcairn Island with a typical EM1 signature55,56,57,87, therefore they are commonly considered as continental equivalents of the EM1-type oceanic basalts13,60,61,62. Although the origin of the EM1 signature in OIBs remains debated, existing models suggest that it could be associated with delaminated/eroded lower continental crust88,89, delaminated SCLM90,91,92, or the recycled subducted sediments55,56,57. In recent years, increasing attention has been devoted to understanding the location and nature of the EM1 signature beneath continental regions (continental EM1), such as that expressed by the intraplate basalts in NE Asia13,58,59,61,62,66,67,84,85. Based on mineralogy, trace element and isotope geochemistry, previous studies consistently suggest that this continental EM1-like component is ancient and has been stored either in the metasomatized lithospheric mantle or within the MTZ52,58,59,61,62,66,67,71–86. However, a SCLM origin is not well supported for the following reasons: (1) the estimated equilibrium temperature of the primary magma exceeds by ~110 °C the maximum temperature expected for SCLM-derived magmas at the same pressure52. (2) Regional lithospheric mantle xenoliths typically exhibit depleted Sr–Nd–Hf isotopic signatures, which contrasts with the relatively enriched Sr–Nd–Hf isotopic compositions of the studied EMI-type basalts13,66,93. (3) An increase in lithospheric thickness corresponds to lower Sr but higher Nd and Pb isotopic ratios in these basalts, a pattern that runs completely counter to the expected trend if the EM1 component were sourced from the basal lithospheric mantle13,66,67. (4) Compared with mantle peridotite xenoliths that display MORB-like Mo isotopic compositions, these basalts possess distinctly lighter Mo isotopic signatures60. (5) More importantly, these basalts are characterized by anomalously high K/U and Ba/Th ratios and positive Hf* anomalies in the primitive mantle-normalized trace element spider diagram13. Such features cannot be readily accounted for by shallow melting of subducted sediments or metasomes within the mantle lithosphere. Instead, they are best explained by deep melting with the contribution of K-hollandite (KAlSi3O8), a high-pressure polymorph of K-feldspar phase that is only stable under conditions exceeding ~10 GPa, formed during phase transitions of subducted sediments at mantle transition zone depths13,94,95. Therefore, the EM1 component is unlikely to have been derived from the SCLM. Considering the combined geophysical46,47,96,97, magnetotelluric48,49,50, petrological52, geochemical13,17,49,62, and geodynamic constraints18,85, a MTZ origin is thus strongly favored, which appears to be the most plausible source for this continental EM1 component.

Major element analyses

Whole-rock major elements were determined at the Wuhan Sample Solution Analytical Technology Co., Ltd., Wuhan, China. Major element concentrations were determined by a Rigaku-3080 X-ray fluorescence (XRF) employing a Rh-anode X-ray tube with a voltage of 40 kV and a current of 70 mA after samples were fused in a high-frequency melting furnace. The analytical precision, relative standard deviation (RSD) and accuracy relative error (RE) are both better than 2% for the major element concentrations determined during this study.

Trace element analyses

Trace-element concentrations of whole-rock samples were determined by ICP-MS (Agilent 7500a with shielded torch) after acid digestion of the samples in high-pressure Teflon bombs at the State Key Laboratory of Geological Processes and Mineral Resources, China University of Geosciences, Wuhan. Sample powder (200 mesh) was weighed into a Teflon bomb and moistened with a few drops of Milli-Q ultrapure water. Then, 1.5 ml of HNO3 and 1.5 ml of HF were added to the Teflon bomb, which was sealed in a steel jacket and heated in an oven at 190 °C for 48 h to completely dissolve the sample. After opening the bomb and evaporating the solution on a hotplate at ~115 °C to dryness, 1 ml of HNO3 was added to the Teflon bomb and evaporated to a second round of dryness. The resultant salt was redissolved by adding ~3 ml of 30% HNO3, resealed in a steel jacket and heated in an oven at 190 °C for 12–24 h. The final solution was diluted to ~100 g with a mixture of 2% HNO3 for ICP-MS analysis. Analyses of international rock standards (AGV-2, BHVO-1, BHVO-2, GSR-1, GSR-3 and BCR-2) indicate that the precision and accuracy are better than 5% for most elements and ~10% for some transitional elements (Supplementary Data 5). The detailed sample-digestion procedure for ICP-MS analyses and the analytical precision and accuracy for trace elements have been described previously98.

Whole rock Sr–Pb isotope analyses

Whole-rock Sr–Pb isotopic compositions were determined using a Neptune Plus MC-ICP-MS (Thermo Fisher Scientific, Dreieich, Germany) at the Wuhan Sample Solution Analytical Technology Co., Ltd., Wuhan, China. The Neptune Plus, a double-focusing MC-ICP-MS, was equipped with seven fixed electron multiplier ICs and nine Faraday cups fitted with 1011 Ω resistors. Prior to analysis, ~100 mg of sample material was dissolved in HF + HNO3 acid in Teflon bombs at ~195 °C for 2 days. 87Rb/86Sr ratios were calculated using measured whole-rock Rb and Sr concentrations determined by ICP–MS. The resulting measured 87Sr/86Sr ratios were normalized to 86Sr/88Sr = 0.1194. The interference elements Ca, Rb, Er, and Yb have been completely separated by the exchange resin process99. The remaining interferences of 83Kr+, 85Rb+, 167Er++, and 173Yb++ were corrected based on the method described previously100. One international NIST 987 standard was measured for every seven samples analyzed. Comparison between our results of Sr isotope standard and recommended values show that they are identical within uncertainty to the published ranges, with an analytical precision better than 0.002%. Full details of the Rb–Sr procedures used in this study are given in ref. 99.

For Pb isotope determination, the Faraday collector configuration of the mass system was composed of an array to monitor 204(Pb + Hg), 206Pb, 207Pb, 208Pb, 203Tl, 205Tl, and 202Hg. Pb single-element solution from Alfa (Alfa Aesar, Karlsruhe, Germany) was used to optimize instrument operating parameters. An aliquot of the international standard solution of 100 μg L−1 NIST 981 was used regularly to evaluate the reproducibility and accuracy of the instrument. Typically, the signal intensities of 208Pb+ in NIST 981 were higher than ~6.0 V. The Pb isotopic data were acquired in the static mode at low resolution. The routine data acquisition consisted of ten blocks of 10 cycles (4.194 s integration time per cycle). The total time of one measurement lasted about 7 min. The exponential law was used to assess the instrumental mass discrimination in this study. Mass discrimination correction was carried out via normalization to a 205Tl/203Tl ratio of 2.38714 (the certified value of NIST SRM 997). Because of the difference of the mass bias behaviors between Pb and Tl, all measured 20xPb/204Pb ratios of the samples were normalized to the well-accepted NIST 981 values of 208Pb/204Pb = 36.7262 ± 31, 207Pb/204Pb = 15.5000 ± 13, 206Pb/204Pb = 16.9416 ± 13. One NIST 981 standard was measured for every ten samples analyzed. Analyses of the NIST 981 standard yielded external precisions of 0.03% (2RSD) for 20xPb/204Pb ratios. In addition, the USGS reference BCR-2 yielded results of 208Pb/204Pb = 38.736 ± 17, 207Pb/204Pb = 15.628 ± 3, 206Pb/204Pb = 18.756 ± 10 (2 SD, n = 22), respectively, which is identical within error of 0.03% to their published values101.

Hg isotope analyses

THg contents of the studied basalts were measured using a DMA-80 Hg analyzer with Hg detection limit of 0.01 ng/g. Analyses of the standard reference materials (GSR-2 and BCR-2) suggested THg recoveries of 90–110% with uncertainties for triplicate analyses of <10%. The samples were then processed following the two-stage thermal combustion and pre-concentration protocol before mercury isotope analysis102. Rock standards (GSR-2 and BCR-2) and method blanks were processed similarly to the samples and yielded THg recoveries of 95–100% for standards. The blanks have lower Hg concentrations than the detection limit, suggesting negligible lab contamination. The pre-enriched solutions were diluted to 1 ng/mL Hg and measured by a Neptune Plus MC-ICP-MS, following the method published previously103. Hg-MDF is expressed in δ202Hg (‰) which is the per mil difference between the 202Hg/198Hg ratio of samples and NIST-3133 mercury standard:

$${{{{\rm{\delta }}}}}{\scriptstyle{{202}}\atop} \!{{{\rm{Hg}}}}(\textperthousand )=\left[{\left({\scriptstyle{{202}}\atop} \!{{{\rm{H}}}}{{{\rm{g}}}}/{\scriptstyle{{198}}\atop} \!{{{\rm{H}}}}{{{\rm{g}}}}\right)}_{{{{\rm{sample}}}}}/{\left({\scriptstyle{{202}}\atop} \!{{{\rm{H}}}}{{{\rm{g}}}}/{\scriptstyle{{198}}\atop}{{{\rm{Hg}}}}\right)}_{{{{\rm{standard}}}}}-1\right]\times 1000$$
(1)

MIF-Hg compositions are expressed in Δ notations, which are the difference between the measured δxHg and the theoretically expected δxHg values:

$$\Delta {\scriptstyle{{{{{\rm{x}}}}}}\atop} \!{{{\rm{H}}}}{{{\rm{g}}}}={{{\rm{\delta }}}}{}{\scriptstyle{{{{{\rm{x}}}}}}\atop} \!{{{\rm{H}}}}{{{\rm{g}}}}-{{{\rm{\delta }}}}{\scriptstyle{{202}}\atop} \!{{{\rm{H}}}}{{{\rm{g}}}}\times \beta$$
(2)

where x is 199, 200, 201 and 202; and β is the mass-dependent scaling factor and is 0.2520 for 199Hg, 0.5024 for 200Hg, and 0.7520 for 201Hg from previous publication32. Multiple measurements of the NIST3177, GSR-2 and BCR-2 standard solutions showed that the analytical uncertainties of δ202Hg, Δ199Hg, Δ200Hg, and Δ201Hg are 0.08, 0.03, 0.03, and 0.05‰ for NIST3177, 0.08, 0.03, 0.03, and 0.05‰ for GSR-2, and 0.08, 0.03, 0.03, and 0.05‰ for BCR-2, respectively, all fall within published ranges. The 2 SD of standard NIST-3177 represents the analytical precision of our samples, with δ202Hg, Δ199Hg, and Δ201Hg of 0.10‰, 0.06‰ and 0.08‰, respectively.