Mantle Hg isotopic heterogeneity and evidence of oceanic Hg recycling into the mantle

The geochemical cycle of mercury in Earth’s surface environment (atmosphere, hydrosphere, biosphere) has been extensively studied; however, the deep geological cycling of this element is less well known. Here we document distinct mass-independent mercury isotope fractionation (expressed as Δ199Hg) in island arc basalts and mid-ocean ridge basalts. Both rock groups show positive Δ199Hg values up to 0.34‰ and 0.22‰, respectively, which deviate from recent estimates of the primitive mantle (Δ199Hg: 0.00 ± 0.10‰, 2 SD)1. The positive Δ199Hg values indicate recycling of marine Hg into the asthenospheric mantle. Such a crustal Hg isotope signature was not observed in our samples of ocean island basalts and continental flood basalts, but has recently been identified in canonical end-member samples of the deep mantle1, therefore demonstrating that recycling of mercury can affect both the upper and lower mantle. Our study reveals large-scale translithospheric Hg recycling via plate tectonics. This work demonstrates distinct mass-independent mercury isotope fractionation in global basalts, which suggests mantle Hg isotope heterogeneity and reveals large-scale translithospheric Hg recycling via plate tectonics.

M ercury (Hg) is a unique heavy metal that has an active redox chemistry, high volatility, strong bioaccumulation, and extreme toxicity 2 . Both natural and anthropogenic activities emit large amounts of Hg to the atmosphere, mainly in the form of elemental Hg (Hg 0 ) 3 . Gaseous Hg 0 has a lifetime of~1 year in the atmosphere, allowing its global transport before deposition into terrestrial and oceanic ecosystems via both wet and dry deposition pathways 4 . Due to its toxicity to humans and wildlife, the geochemical cycle of Hg in Earth's surface environment (atmosphere, hydrosphere, and biosphere) has been extensively studied 2 . However, the cycling of Hg in Earth's interior reservoirs (e.g., crust and mantle) remains less studied 1,5 .
Mercury ore deposits are predominantly located in active continental margin settings, with two well-known mercuriferous belts-the Pacific rim metallogenic belt and the Alpine-Himalayan belt 4 , implying a causal link between Hg metallogenesis and plate subduction. Mercury in oceanic reservoirs (e.g., sediments, seawater) may be carried by subducting ocean slabs into the asthenospheric mantle, and can then be emitted during arc magmatism to form Hg-enriched hydrothermal ore deposits in active continental margin settings 6 . A recent study, based on Hg isotopes in ocean island basalts, suggests that crustal Hg is also recycled into the lower mantle 1 .
Mercury is the only metal that exhibits both significant isotopic mass-dependent fractionation (MDF, defined as δ 202 Hg) and mass-independent fractionation (MIF, defined as Δ 199 Hg, Δ 200 Hg, Δ 201 Hg, and Δ 204 Hg) 7,8 . Hg-MDF is produced during essentially all biological, chemical, and physical processes involving Hg 7 . MIF of 200 Hg and Δ 204 Hg is detected mainly in atmospherically derived samples, and values in rocks are generally too small to enable source tracing 8 . MIF of 199 Hg and 201 Hg mainly occurs during photochemical processes with little contribution from other reactions, and therefore provides clear source constraints 7 . On Earth's surface, photochemical processes result in negative Δ 199 Hg (−0.6 to 0‰) in terrestrial reservoirs (e.g., soil and vegetation) and positive Δ 199 Hg (0 to 0.4‰) in oceanic reservoirs (e.g., seawater and marine sediments) 7,8 . It has been estimated that the primitive mantle has near-zero Δ 199 Hg (Δ 199 Hg: 0.00 ± 0.10‰, 2 SD) and negative δ 202 Hg (−1.7 ± 1.2‰, 2 SD), based on Hg isotope analyses of 3 He-rich lavas 1 . A recent study also found significantly positive Δ 199 Hg (0 to 0.4‰) in hydrothermal ore deposits associated with arc magmatism, implying the recycling of Hg from marine sediments into arc lavas 6 .
Distinct Sr-Nd-Pb isotopic compositions of mid-ocean ridge basalts (MORBs), island arc basalts (IABs), ocean island basalts (OIBs), and continental flood basalts (CFBs) demonstrate a pronounced mantle isotope heterogeneity, which is generally accepted to result from recycling of crustal materials into the mantle via plate subduction 9,10 . MORBs are derived from partial melting of the upper mantle during the ascent of the asthenosphere beneath mid-ocean ridges. IABs result from partial melting of the mantle wedge at convergent margins due to the addition of subducted crustal materials (including volatile components and/or melts). OIBs and CFBs are genetically related to mantle plumes, which originate from the lower mantle or the core-mantle boundary 11 . A landmark study recently observed deviations in Δ 199 Hg in three well-characterized mantle endmembers (EM-1: −0.45 to 0.05‰; EM-2: 0.23‰; HIMU: −0.14‰), compared to the estimate of the primitive mantle (Δ 199 Hg: 0 ± 0.1‰, 2 SD) based on Hg isotope analyses of 3 Herich lavas, highlighting the great potential of using Hg isotopes to trace crustal Hg recycling into the mantle 1 .
To aid in understanding the deep Hg cycle and to probe possible mantle Hg isotope heterogeneity, we measured the Hg isotopic compositions of a variety of types of basalts at a global scale (Fig. 1). Samples include MORBs from the Mid-Atlantic Ridge (MAR, n = 5), Southwest Indian Ridge (SWIR, n = 3), and East Pacific Rise (EPR, n = 7); IABs (n = 9) from the southern part of the Mariana Island Arc; HIMU-like OIBs (n = 2) and EM1-like OIBs (n = 5) from the Pako guyot of the Magellan Seamount Chain in the West Pacific Seamount Province; and CFBs (n = 17) from the Siberian Trap, the world's largest flood basalt province. The geological background of these samples and analytical methods are given in the Materials and Methods section in detail. The sample information and analytical results are summarized in Supplementary Tables 1 and 2. We observed diverse Hg-MIF signals in these basalts, providing evidence of oceanic Hg recycling into the mantle.

Results and discussion
Hg concentrations and mass-dependent fractionation of Hg isotopes. In this study, total Hg (THg) concentrations vary from 0.65 to 2.35 ng/g for MORBs, from 0.54 to 1.28 ng/g for IABs, from 0.48 to 2.08 ng/g for OIBs, and from 0.91 to 4.29 ng/g for CFBs, with mean values of 1.48 ng/g, 2.25 ng/g, 1.63 ng/g, and 0.80 ng/g, respectively ( Fig. 2A). These concentrations at the low ng/g level are in the range of mafic-ultramafic rocks worldwide (i.e., 0.2-7.0 ng/g for basalts and peridotite) 12,13 .
The large variation of δ 202 Hg in the studied basalts could be caused by MDF during magmatic processes. Lighter Hg isotopes may be preferentially volatilized during Hg degassing 14 , and thus generate enrichment of heavy Hg isotopes in the residue phases. The lower δ 202 Hg of fumarolic gas (−1.74 ± 0.36‰, 2 SD) compared to condensed particles (−0.74 ± 0.18‰, 2 SD) in an active volcano (Vulcano Island, Italy) strongly supports this view 14 . In our study, the more negative values of δ 202 Hg in OIBs and CFBs (that are genetically related to the lower mantle) relative to those of MORBs and IABs (that originate from the upper mantle), as shown in Fig. 2B, could possibly be explained by more severe Hg degassing during the eruption of the latter compared to the former.
An alternative explanation for the variation of δ 202 Hg in the basalts studied is the mixing of Hg sources with distinct δ 202 Hg values. The continental crust, in particular sedimentary rocks and sediments, displays large variations of δ 202 Hg (−3 to 1‰), with a mean value of −0.7 ± 1.6‰ (2 SD) (reviewed by Blum et al. 7 ). The large variations of δ 202 Hg in MORBs and IABs in our study, coupled with their higher mean δ 202 Hg values compared to OIBs and CFBs, may be explained by the recycling of crustal rocks into the upper mantle. However, many different processes could trigger Hg-MDF 7 , which hinders the full understanding of Hg-MDF during magmatic processes. Thus, δ 202 Hg is not considered to be highly diagnostic of Hg sources. Instead, we focus on the Hg-MIF signal as a more reliable source tracer, as discussed below.
Crustal Hg recycling into the lower mantle. OIBs and CFBs are petrogenetically related to the lower mantle. As shown in  ARTICLE pool in the lower mantle, given the fact that OIBs and CFBs in this study as well as 16 previously studied OIBs (including 11 3 He-rich lavas) 1 show near-zero Δ 199 Hg values (Fig. 3A).
Recycling of oceanic Hg into the upper mantle. MORBs and IABs have mostly positive Δ 199 Hg values of 0.05 to 0.22‰ and -0.01 to 0.34‰ (Figs. 2C, 3A), with Δ 199 Hg/Δ 201 Hg of~1 (Fig. 3B). These values fall within the range of marine sediments and seawater, suggesting that a substantial amount of Hg in these rocks is of marine origin. Their positive Δ 199 Hg values may result from the fixation of seawater Hg during seawater-rock reactions. MORBs and IABs investigated here were collected from the seafloor and show variable weight loss on ignition (LOI) ranging from −0.9 to 0.4 wt% and 6.2 to 8.0 wt%, respectively (Supplementary Table 2). The very low LOI of the MORBs and their pristine chemical and mineralogical composition argue against seawater overprinting (Supplementary Table 2). Although IABs have high LOI, no correlations are observed between LOI and THg or Δ 199 Hg (Fig. 4), indicating that seawater-rock reactions are unlikely to be the mechanism responsible for the positive Δ 199 Hg in IABs. The OIB samples studied here were also collected from the seafloor. Although these OIBs have relatively higher LOI values (2.0 to 8.6 wt%, Supplementary Table 2) than MORBs, they have near-zero Δ 199 Hg values. Given the low Hg concentration in seawater (<0.02 to 2 ng/L) 25 , seawater-rock reactions can be further precluded as a reason for the positive Δ 199 Hg values in MORBs and IABs.
We attribute the positive Δ 199 Hg signals in the studied IABs to recycled marine sediments. At convergent margins, subduction of the oceanic slab carries a substantial amount of Hg from marine sediments into subduction zones. The downgoing slab undergoes dehydration and metamorphism and releases fluids and melts that in turn metasomatize the mantle wedge. Partial melting of such metasomatized mantle could generate IABs with positive Δ 199 Hg. Notably, given the low volatilization temperature of Hg, subducted Hg may be mostly released from the downgoing slab and returned to the Earth's surface through arc volcanism. This is supported by the positive Δ 199 Hg signal observed in volcanic arcrelated epithermal Hg-Au deposits in NE China and the Pacific rim (0 to 0.4‰, Fig. 3A, B, Deng et al. 6 and references therein). The most recent estimate of present-day global volcanic Hg flux (76 ± 30 t/year) 26 is slightly lower than the total mass of Hg in global subducted sediments (~87.4 t/year), calculated from the mass of global subducted sediment (1.4 × 10 15 g/year) 27 multiplied by the Hg concentration of average marine sediments (62.4 ng/g) 28 .
The positive Δ 199 Hg of MORBs could also be the result of the recycling of oceanic crust. In subduction zones, the subducted marine Hg may not be completely released via arc volcanism. The remaining Hg in the subducted slab could continue to descend into the asthenospheric mantle. This fraction of Hg is recycled into the asthenosphere and could be brought back to Earth's surface via basaltic magmatism at mid-ocean ridges. The upper mantle is heterogeneous in Hg isotopes, apparently due to the variable involvement of recycled crustal Hg. Such isotopic heterogeneities are consistent with large variations in Sr-Nd-Pb isotope compositions for MORBs from different mid-ocean ridges (e.g., MAR, SWIR, and EPR), which supports the model of recycling of crustal materials into the upper mantle [29][30][31] . The recycling of oceanic Hg into the upper mantle is also favored by the positive Δ 199 Hg values (0.07 to 0.27‰) observed in hotspring and sinter samples from the Yellowstone Plateau volcanic field 32 , which might be explained by upwelling asthenospheric mantle flow induced by the subducting Pacific plate 33 .
Geodynamic implications. Based on our current knowledge of Hg isotopic variation in mantle-derived rocks, the deep cycling of Hg is illustrated in Fig. 5. Photochemical reactions produce significant negative Δ 199 Hg values in terrestrial reservoirs (e.g., soil and vegetation) and significant positive Δ 199 Hg values in marine reservoirs (e.g., seawater and marine sediments). Mercury from terrestrial and marine reservoirs is subducted at convergent margins. Due to the low volatilization temperature of Hg, much of the subducted Hg is likely released back to the surface environment through arc volcanism, forming IABs and arc-related epithermal Hg deposits with significantly positive Δ 199 Hg values. The remaining Hg in the subducted slab is recycled into the asthenosphere and released at mid-ocean ridges by MORBs, which are characterized by significantly positive Δ 199 Hg values.
Some of the subducted Hg may also be recycled into the lower mantle, consistent with the recently observed significant Δ 199 Hg signals in some OIB samples 1 ; however, given that most OIBs and CFBs show near-zero Δ 199 Hg values, this recycled Hg likely represents only a small proportion of Hg in the lower mantle, and may be mostly located in the source of the mantle end-members. In the future, analysis of OIBs and CFBs from additional sites will be needed to better quantified the amount of recycled Hg in the lower mantle.
Overall, our study demonstrates mantle Hg isotopic heterogeneity and reveals large-scale translithospheric Hg recycling via plate tectonics. Hg tends to be depleted in volcanic rocks derived from the mantle (e.g., several ng/g in basalts) and enriched in crustal sedimentary rocks (e.g., 62.4 ng/g in average marine sediments 28 ). Given the significantly large Δ 199 Hg values and high Hg concentrations in crustal materials, we infer that the crustal signature cannot easily be erased by the primitive mantle signature (Δ 199 Hg~0). We conclude that Hg-MIF (Δ 199 Hg) can be a useful tool for studying crust-mantle interactions.

Methods
Samples and geological background. The sample locations are shown in Fig. 1 and summarized in Supplementary Table 1 CFBs (n = 17) were sampled from the Norilsk region at the northwestern margin of the Siberian platform, which is a part of the world's largest flood basalt province, the Siberian Traps (Fig. 1). The eruption of these CFBs occurred from the latest Permian to the Early Triassic (i.e., 248-252 Ma) and was associated with the break-up of Gondwanaland 39 . Trace element and Sr-Nd isotope data indicate that CFBs from the Norilsk region were formed by the interaction of plume-generated picritic magmas with the lithosphere 40 .
Chemical analyses. The samples were cut to expose fresh surfaces, washed by 18.2 MΩ cm water, air-dried, powdered, and homogenized, prior to chemical analysis at the Institute of Geochemistry, Chinese Academy of Sciences (IGCAS). Major elements were analyzed at the ALS Minerals-ALS Chemex, Guangzhou, China, using a PANalytical PW2424 X-ray fluorescence spectrometer. LOI values were measured by combustion loss of weight, with analytical uncertainties (RSD) of <5%. Total Hg (THg) concentration was determined by an RA-915+ Hg analyzer (Lumex, Russia), with a detection limit of 0.01 ng/g. Standard reference materials (GSR-2, andesite; BCR-2, basalt) were measured, yielding Hg recoveries of 94-106% and RSD of <9%. The samples were prepared for Hg isotope analysis using a double-stage tube furnace coupled with 40% anti aqua regia (HNO 3 /HCl = 2/1, v/v) trapping solutions 41 . Standard reference materials (GSR-2 and BCR-2) and method blanks were processed in the same way as the samples. The former yielded Hg recoveries of 90-105% and the latter showed Hg concentrations lower than the detection limit, precluding lab contamination. The preconcentrated solutions were diluted to 0.5 ng/mL with an acid concentration of 10-20% prior to Hg isotope analysis using a Neptune Plus multi-collector inductively coupled plasma mass spectrometry 42 . Hg isotope ratios were reported following the convention proposed by Blum and Bergquist 43 . MDF is expressed in δ 202 Hg notation in units of ‰ referenced to the NIST-3133 (analyzed before and after each sample): MIF is reported in Δ notation, which describes the difference between the measured δ xxx Hg and the theoretically predicted δ xxx Hg value, in units of ‰: 0.02 ± 0.06‰, 2 SD, n = 5) and BCR-2 (δ 202 Hg: -1.89 ± 0.11‰; Δ 199 Hg: 0.00 ± 0.07‰; Δ 200 Hg: 0.00 ± 0.06‰; Δ 201 Hg: 0.01 ± 0.06‰, 2 SD, n = 5) agree well with previous results [43][44][45] . The larger of the values of standard deviation (2 SD) for either NIST-3177, GSR-2, or BCR-2 are used to reflect maximum analytical uncertainties.

Data availability
All data is available in the main text or supplementary materials.