resent-day Earth's mantle structure is dominated by a degree-2 spherical harmonic featuring two equatorial and antipodal mantle domains bisected by a subduction girdle surrounding the Pacific Ocean. Each of the two antipodal African and Pacific mantle domains features a large low shear-wave velocity province (LLSVP) in the lower mantle. Although most mantle plumes are believed to have originated from the edges of the LLSVPs in the two mantle domains since at least ca. 200 million years ago (Ma) 1 , we still know little about the nature and evolutionary histories of the two LLSVPs.Opposing models exist regarding how and when the two LLSVPs formed. According to one model, they are quasi-stationary, long-lived through Earth history (4.0-2.0 billion years ago) and uninfluenced by plate tectonics 1,2 . By contrast, another model claims the LLSVPs are dynamic in their formation, evolution (including demise) and geographic locations and are linked to the assembly and breakup of supercontinents 3,4 . Palaeomagnetic data, mantle plume records during the last two supercontinent cycles and dynamic modelling results have been used to suggest that the antipodal LLSVPs are related to whole-mantle convection driven by plate motion 5,6 , particularly by circum-supercontinent subduction 7 . As such, the shape and location of LLSVPs both in time and space are dynamically linked to the formation of supercontinents. It has been further speculated 3,7,8 that when the antipodal LLSVPs, whose locations are controlled by the subduction girdle surrounding the supercontinent, are positioned off the Equator, the centrifugal force of Earth's spin would bring them to the Equator through true polar wander, that is, wholesale rotation of the entire silicate Earth relative to the spin axis 9 . Such a strong coupling between the outer layer of the planet (tectonic plates) and deeper mantle domains is consistent with both continental and oceanic plume records over much of Earth history 10,11 . However, to test contrasting geodynamic models, it is fundamental to characterize and compare the geochemical compositions and evolutionary paths of the two mantle domains.Radiogenic isotopes (lead (Pb), strontium (Sr) and neodymium (Nd)) and noble gas isotopes (for example, helium (He)) of basaltic lava flows originating from hotspots in the present-day oceans appear to be the best tools for evaluating any systematic compositional difference between the two mantle domains 12 . A first-order isotopic distinction can be made between mid-ocean-ridge basalts (MORBs) and hotspot basalts (or ocean island basalts (OIBs)), which are thought to tap reservoirs in the depleted upper mantle 13 and the more primitive lower mantle 14 , respectively. While most MORBs are isotopically relatively uniform, OIBs show substantial diversity due to deep recycling of various subducted lithospheric components such as oceanic and continental lithospheric materials, including sediments 13 .Geographically, isotopic signatures from oceanic basalts have a purported hemispheric distinc...

Mantle melts provide a window on processes related to global plate tectonics. The composition of chromian spinel (Cr-spinel) from mafic-ultramafic rocks has been widely used for tracing the geotectonic environments, the degree of mantle melting and the rate of mid-ocean ridge spreading. The assumption is that Cr-spinel’s core composition (Cr# = Cr/(Cr + Al)) is homogenous, insensitive to post-formation modification and therefore a robust petrogenetic indicator. However, we demonstrate that the composition of Cr-spinel can be modified by fluid/melt-rock interactions in both sub-arc and sub-mid oceanic mantle. Metasomatism can produce Al-Cr heterogeneity in Cr-spinel that lowers the Cr/Al ratio, and therefore modifies the Cr#, making Cr# ineffective as a geotectonic and mantle melting indicator. Our analysis also demonstrates that Cr-spinel is a potential sink for fluid-mobile elements, especially in subduction zone environments. The heterogeneity of Cr# in Cr-spinel can, therefore, be used as an excellent tracer for metasomatic processes.

The most dominant features in the present-day lower mantle are the two antipodal African and Pacific large low-shear-velocity provinces (LLSVPs). How and when these two structures formed, and whether they are fixed and long lived through Earth history or dynamic and linked to the supercontinent cycles, remain first-order geodynamic questions. Hotspots and large igneous provinces (LIPs) are mostly generated above LLSVPs, and it is widely accepted that the African LLSVP existed by at least ca. 200 Ma beneath the supercontinent Pangea. Whereas the continental LIP record has been used to decipher the spatial and temporal variations of plume activity under the continents, plume records of the oceanic realm before ca. 170 Ma are mostly missing due to oceanic subduction. Here, we present the first compilation of an Oceanic Large Igneous Provinces database (O-LIPdb), which represents the preserved oceanic LIP and oceanic island basalt occurrences preserved in ophiolites. Using this database, we are able to reconstruct and compare the record of mantle plume activity in both the continental and oceanic realms for the past 2 b.y., spanning three supercontinent cycles. Time-series analysis reveals hints of similar cyclicity of the plume activity in the continent and oceanic realms, both exhibiting a periodicity of ∼500 m.y. that is comparable to the supercontinent cycle, albeit with a slight phase delay. Our results argue for dynamic LLSVPs where the supercontinent cycle and global subduction geometry control the formation and locations of the plumes.