Search

• Sediment is the source of the earliest slab-derived fluids. • The serpentinizing fluids experienced extensive interaction with subducted basalts. • Breakdown of basalt plagioclase affected chemistry of the earliest fluids. • Breakdown of plagioclase resulted in calcium carbonate precipitation. • Lawsonite and rodingite appear as stable phases after plagioclase breakdown. The geochemistry of Mariana forearc serpentinites sheds light on how the subducting slab starts to devolatilize. International Ocean Discovery Program Expedition 366 targeted three serpentinite seamounts: Yinazao, Fantangisña and Asùt Tesoru, with corresponding slab depths of 13, 14 and 18 km, respectively. The fluid-mobile elements (FMEs) B, Rb, Cs, and Li become more enriched in the eruptive serpentinites with increasing slab depth. Yinazao serpentinites show strongly elevated Sr and Ba contents, which complements depletions in these elements seen in mafic clasts entrained in the serpentinites. Correlations between the FMEs and 87Sr/86Sr of the serpentinites reflect interactions among slab fluid, mantle and the subducting slab. With the enrichments of Sr-Ba in the Yinazao and B-Rb-Cs-Li in the Asùt Tesoru serpentinites, their 87Sr/86Sr converge toward values similar as those of the mafic clasts from values between sediments and the mantle. These observations indicate that sediment is the source of the earliest slab-derived fluids, which experienced extensive interaction first with mantle wedge peridotites to produce high pH fluids, and later with subducted seamount basalts. The decomposition of plagioclase in subducting basalts leads to precipitation of CaCO 3 from the high pH serpentinizing fluids and elevated Ca-Sr-Ba contents under low carbonate alkalinity conditions, as seen in the Yinazao serpentinites. The breakdown of clay minerals after lawsonite formation and the lizardite-antigorite transition release geochemically distinct fluids, with high B-Rb-Cs-Li and low Sr. This episode of fluid release results in the dissolution of CaCO 3 precipitated in the shallow subduction channel, and the generation of rodingite assemblages and epidote. Finally, these fluids evolve to have high carbonate alkalinity and low Ca-Sr-Ba contents, as seen in Asùt Tesoru serpentinites as well as those in South Chamorro and Conical Seamounts. This study reveals early plagioclase breakdown via high pH fluids in the subduction channel is important both for the generation of hydrous minerals and the precipitation of CaCO 3 in the subducting slab, subduction channel, and shallow mantle wedge. [ABSTRACT FROM AUTHOR]

• Fe and S speciation were measured in volcanic glasses from the Troodos Ophiolite. • Troodos glass ƒO 2 values are uncorrelated with volatile and fluid-mobile elements. • Mantle melts can be influenced by subduction processes without becoming oxidized. • The oxidizing capacities of slab-derived fluids and melts are variable. Oxygen fugacity (ƒO 2) varies systematically across tectonic environments. The typically high ƒO 2 recorded in arc settings relative to oceanic ridges is attributed to recycling of oxidized materials derived from subducting slabs into regions of the mantle that undergo partial melting. To evaluate further the relationship between ƒO 2 and mantle metasomatism, Fe and S X-ray Absorption Near-Edge Structure measurements were conducted on volcanic glass wafers sampled across the extrusive section of the Troodos Ophiolite, Cyprus, which is a type locality for studying the influence of subduction on mantle melting processes and the generation of oceanic crust. The glasses record ƒO 2 values relative to the quartz-fayalite-magnetite (FMQ) buffer of FMQ+0.13±0.16(1σ) to FMQ+0.74±0.23(1σ) upon quenching at the seafloor. At a given MgO content, the ƒO 2 values recorded by the Troodos glasses do not vary systematically based on type (i.e., boninitic versus tholeiitic) nor by sampling location, indicating that the lavas were derived from primary melts and mantle sources that were indistinguishable in their initial ƒO 2 levels. The glass ƒO 2 values do not vary with dissolved H 2 O contents, nor with trace element ratios (e.g. , Ba/La, Ce/Pb) and Pb and Sr isotopic compositions, all used to monitor interaction of slab-derived materials with the Troodos mantle source. The decoupling of ƒO 2 and H 2 O and fluid-mobile elements renders the Troodos an important endmember for interpreting global variations in the redox state of mantle melts formed in subduction-influenced settings. For example, the Troodos glasses are reduced relative to Izu-Bonin boninites and melt inclusions from Mariana and Cascades arc volcanoes, despite overlapping and elevated H 2 O contents, Ba/La, and S/Dy relative to mid-ocean ridge basalts. The lack of correlation between ƒO 2 and volatile contents and with proxies for subduction influence in the Troodos glasses is similar to basalts formed in back-arc environments and in near-trench settings in Izu-Bonin-Mariana. Such comparisons suggest that the slab-derived fluids that infiltrated the Troodos melting region were derived from a shallow slab and had low oxidizing capacity relative to fluids or melts which interacted with the mantle in the Izu-Bonin-Mariana arc and in the Cascades, which represent arc environments with comparable glass Fe XANES measurements. A simple framework accounts for local and global variability in ƒO 2 in fractionation-corrected glasses, where fluids of variable oxidizing capacity interact with depleted mantle to produce hydrous melts. The inheritance of volatile and fluid-mobile elements in regions of the mantle that undergo partial melting does not necessarily lead to higher ƒO 2 if the oxidizing capacity of the infiltrating hydrous fluids or melts is low. In back-arcs and near-trench environments, partial melts formed by interaction with fluids of low oxidizing capacity are similar to arc melts in their volatile contents, trace element, and isotopic signatures, but modestly oxidized relative to mid-ocean ridge basalts. [ABSTRACT FROM AUTHOR]

• The slab-unfolding links plates and slabs without assumed vertical slab sinking. • The Lesser Antilles slab was laterally transported up to ∼900 km after subduction. • We reconstruct up to 18 mm/yr lateral slab transport since 50 Ma. • A new pre-subduction ridge-transform system of the Proto-Caribbean is presented. The link between surface tectonic plates and mantle slabs is fundamental for paleo-tectonic reconstructions and for our understanding of mantle dynamics. Many seismic tomography-based studies have assumed vertical slab sinking and projected mantle features to the surface to reconstruct paleo-trench locations or explain tectonic features. Here, we used a slab-unfolding approach that does not require assumptions about sinking paths or rates to re-interpret the seismic structure of the Lesser Antilles slab underneath the Caribbean. A recent study invoked mainly vertical slab sinking and a highly folded and deformed slab to explain seismic Caribbean mantle structures. However, our results show that the upper-mantle Lesser Antilles slab structure can be better explained by limited intra-slab deformation and up to ∼900 km lateral slab transport towards the northwest after subduction. Our results indicate that such lateral slab transport can occur even with probable weaknesses in the slab that originate from a subducted fossil ridge-transform system. We ascribe the lateral slab transport in the mantle to a kinematic connection with the North American plate, which has migrated northwestward since the Eocene. [ABSTRACT FROM AUTHOR]

• Global primitive arc lavas (Mg# ≥60) display notable δ49/47Ti heterogeneity. • Residual rutile imposes high δ49/47Ti of 0.24 ± 0.06 ‰ on hydrous, silicic slab melts. • Primitive Aleutian rhyodacites have the same δ49/47Ti as predicted for slab melts. • A variably diluted signature of slab melts is found in all eight subduction zones. • A slab melt component is required to generate silicic primitive arc lavas. It is still a matter of intense debate to what extent partial melting of the subducting slab contributes to arc magmatism in modern subduction zones. In particular, it is difficult to differentiate between silicate melts formed by partial melting of the slab, and aqueous fluids released during subsolidus dehydration as the main medium for slab-to-mantle wedge mass transfer. Here we use δ49/47Ti (the deviation in 49Ti/47Ti of a sample to the OL-Ti reference material) as a robust geochemical tracer of slab melting. Hydrous partial melting of subducted oceanic crust and the superjacent sedimentary layer produces silicic melts in equilibrium with residual rutile. Modelling shows that such silicic slab melts have notably higher δ49/47Ti (+0.24 ± 0.06 ‰) than their protolith due to the strong preference of rutile for the lighter isotopes of Ti. In contrast, even highly saline fluids cannot carry Ti from the slab and hence hydrous peridotite partial melts have δ49/47Ti similar to mid-ocean ridge basalts (MORB; ca. 0 ‰). Primitive (Mg# ≥60) arc lavas from eight subduction zones that are unaffected by fractional crystallisation of Fe-Ti oxides show a more than tenfold larger variation in δ49/47Ti than found in MORB. In particular, primitive arc lavas display a striking correlation between SiO 2 content and δ49/47Ti that ranges from island arc basalts overlapping with MORB, to primitive rhyodacites with δ49/47Ti up to 0.26 ‰ erupted in the western Aleutian arc. The elevated δ49/47Ti of these primitive arc lavas provides conclusive evidence for partial melts of the slab as a key medium for mass transfer in subduction zones. The Aleutian rhyodacites represent a rare example of slab melts that have traversed the mantle wedge with minimal modification. More commonly, slab melts interact with the mantle wedge to form an array of primary arc magmas that are a blend of slab- and peridotite-derived melt. We identify primitive arc lavas with a clearly resolvable slab melt signature in all eight subduction zone localities, confirming that slab melting is prevalent in modern subduction zones. [ABSTRACT FROM AUTHOR]

• Indonesia contains arc volcanoes that are spatially and/or geochemically anomalous. • Giant porphyry deposits are spatio-temporally associated with anomalous volcanism. • We suggest these anomalous volcanoes and ore deposits are linked to slab tearing. • Slab tearing may also facilitate extremely explosive eruptions in island arcs. Most Holocene volcanoes in Indonesia display geochemical and geophysical properties characteristic of formation via mantle wedge metasomatism. However, some volcanoes (Toba, Ungaran, Muriah, Rinjani, Tambora, Sangeang Api, and Batu Tara) are anomalous, showing alkaline geochemical affinities, anomalously violent recent eruptions, and/or positioning above sections of slab generally considered too deep to induce mantle melting via slab dehydration. These anomalous volcanoes are spatially and temporally associated with giant Pliocene–recent porphyry Cu deposits (in thin continental or oceanic crust), contrary to global occurrences of porphyry Cu deposits that are typically restricted to thick (>45 km crustal thickness) continental arcs. Using geophysical data to model slab geometry and geochemical data to evaluate source compositions and melting mechanisms, we demonstrate that anomalous Indonesian volcanoes and giant porphyry Cu deposits are linked to slab segmentation due to slab tearing. In the slab beneath Toba, the mechanically weak Investigator Fracture Zone, which separates a more buoyant Cenozoic segment of the Indo-Australian plate from older, denser, and more steeply dipping Jurassic–Cretaceous lithosphere to the east, likely created a sub-vertical slab tear feeding volcanism via poloidal mantle flow. The alkaline volcanoes Muriah, Ungaran, Tambora, and Sangeang Api (as well as the sub-alkaline volcano Rinjani) are positioned along edges of slab segments defined by negative seismic velocity anomalies at depths of ∼185–400 km, which we interpret as elongate slab windows formed during localised choking of subduction by anomalously buoyant oceanic lithosphere. Although no negative velocity anomalies are associated with Batu Tara, we speculate that this alkaline volcano is located above a small-scale slab tear, potentially associated with the subducted Scott Plateau. We propose that poloidal asthenospheric flow along slab segment boundaries has triggered deep, low-degree partial melting with varying contributions of slab melt, sediment, and fluid. Products of this melting regime are sub-alkaline and alkaline lavas with trace element patterns enriched in incompatible elements relative to the main arc volcanoes. Based on spatial, temporal, and geochemical associations of giant Pliocene–recent porphyry Cu deposits with proposed slab tears (and an absence of evidence for slab melting in the formation of these deposits), we suggest that slab tears may have generated oxidised magmas by facilitating melting in the garnet stability field, emulating the petrogenetic effects of thick continental crust and slab melting. We further hypothesise that slab tearing locally enhanced magma flux, leading to extremely violent Quaternary eruptions at Toba, Rinjani, and Tambora. [ABSTRACT FROM AUTHOR]

• Magnitude and pattern of mantle convection are sensitive to slab sinking rates. • Bulk lower mantle flow is significantly slower than the average slab sinking rates. • Using new inferred kinematic observations can improve mantle convection models. • Unmixed heterogeneities can survive in the lower mantle for at least 1000 ma. Numerical models of Earth's mantle dynamics that aim to comply with a variety of surface observations and/or modern mantle structure, still predict a widely varying vigour of mantle flow which governs the long-term evolution of mantle structure and mixing. A yet unexplored source of intrinsic information on mantle flow characteristics are the geologically reconstructed average slab sinking rates. Here we evaluate from numerical experiments how average slab sinking rates relate to the vigour of mantle convection and mixing. We use a simplified mantle convection model and show that long-term mantle flow velocity and mixing is strongly sensitive to slab sinking rates. Models tuned to match lower mantle average sinking rates of 10–15 mm/a, yield lower mantle convection rates of only several mm/a. Furthermore, they reveal large unmixed regions in the mid-mantle which preserve 25% of 'primordial' lower mantle material after 1000 Ma, which may explain geochemical observations from hotspot volcanoes. [ABSTRACT FROM AUTHOR]