Melting of subducted sediments reconciles geophysical images of subduction zones

Subduction zones subduct a package of sediments that are compositionally distinct from the underlying oceanic crust and mantle. While sediments dehydrate and contribute to serpentinite formation at shallow depths (<10 km, T <700 °C), the remaining sediments can begin to melt at higher pressures and temperatures typical of the fore-arc (approximately 2.5–3.2 GPa; ~675–950 °C). This melting can occur trenchward of the region where typical basaltic–andesitic arc melts are generated. The authors pose the question of how the fate of subducted sediments influences geophysical observations—particularly the commonly observed strong fore-arc electrical conductors in magnetotelluric (MT) models—and the geochemical and seismic characteristics of subduction zones. They propose that silicic, volatile-rich sediment melts infiltrate and react with the overlying mantle wedge at temperatures too low to generate arc melts, forming a potassium-rich phlogopite–pyroxenite metasome and releasing saline fluids, thereby creating the observed conductive anomalies and affecting seismic behavior.

Prior work has established the dehydration of subducted sediments and serpentinization of the fore-arc mantle, influencing earthquake depths and seismic properties. Global thermal models (e.g., Syracuse et al., 2010) and constraints from exhumed rocks (Penniston-Dorland et al., 2015) frame the range of slab top temperatures, with some exhumed records indicating hotter conditions than models. MT studies commonly reveal strong conductors in fore-arcs, previously attributed to accumulations of dehydration fluids, reactions at the basalt–eclogite transition, or complex melt pathways. Experimental studies show sediments can produce hydrous silicic melts at sub-arc depths and that phlogopite can be anomalously conductive, especially when F-rich. These lines of evidence motivate a unified model wherein sediment melting and subsequent mantle metasomatism create the observed MT conductors and influence geochemical signatures.

Experimental reaction setup: Sediment–peridotite reaction experiments were used to characterize products of sediment melt interaction with depleted mantle. A two-layer capsule with 50/50 marine siliciclastic sediment (IODP ODP 161-976 B 18×3 105–106.5; <10% carbonate; 5–10 wt% H2O) and depleted peridotite (dunite ZD11-53: >97% olivine, ~2% spinel, <1% clinopyroxene) simulated sediment subduction. P–T conditions were 2–3 GPa and 750–900 °C (fore-arc setting), run in a piston-cylinder apparatus (Univ. of Mainz). Post-run, capsules were sectioned, polished, and analyzed by EPMA (JEOL JXA-8200). Experiments at ~3 GPa and 800–900 °C produced a reaction zone (~500 µm thick, increasing with T, melt fraction, and duration) of phlogopite–pyroxenite between sediment and dunite. Potassium partitions into phlogopite; the zone is enriched in Ca and Si (Cpx and Opx). Given sediment melts can hold 5–10% volatiles, and considering additional fluids from underlying serpentinites, reactions produce a fluid phase alongside metasome formation.

MT forward and inverse modeling: Subduction geotherms and slab top temperatures were taken from published models (including van Keken et al. for Cascadia). The subsurface was simplified into domains: anhydrous harzburgite (80% olivine, 20% pyroxene), phlogopite–pyroxenite metasome (20% phlogopite, 80% pyroxene), and regions containing 1% sediment melt, 1% arc melt, and 1% saline fluid (5% NaCl) as appropriate. Assumptions: sediment melting initiates where slab top T ≥675 °C; overlying mantle comprises metasome with water-rich sediment melt (12% H2O) until T drops below 700 °C, where melt crystallizes to metasome plus saline fluid; arc melting begins at ~800 °C and extends ~50 km laterally, with water-rich arc melt (12% H2O) extending through the overlying mantle and crust; melt/fluid phases extend upward to ~10 km depth (brittle–ductile transition). A 500 m, 500 Ωm surface layer simulated a sedimentary basin. Electrical conductivities: Components sourced from lab data—olivine (Gardès et al.), orthopyroxene (Dai & Karato), melts with 12% H2O (Sifré et al.), phlogopite (Li et al., including F effects), and saline fluid with 5% NaCl (Sinmyo & Keppler). Conductivities were geometrically combined using a modified Archie’s law assuming good connectivity at 1% melt/fluid. Sensitivity tests used 0.5% and 5% melt for Cascadia. Forward responses were generated with MARE2DEM at 25 synthetic MT stations at 4 km spacing over 17 periods (10–10^5 s), with 5% noise, then inverted from a 100 Ωm half-space. Cascadia inversions achieved rms ~1.5; Kyushu rms ~2.3. Synthetic models were compared with published MT inversions (Cascadia CAFE line; Kyushu across Kirishima), respecting differences such as inclusion/exclusion of a sharp slab boundary.

• A new model posits that hydrous sediment melts form in the fore-arc at slab top temperatures of ~675–950 °C and pressures ~2.5–3.2 GPa, below the peridotite solidus. These silicic, volatile-rich melts react with the overlying mantle to form a K-enriched phlogopite–pyroxenite metasome and release saline fluids.
• Reaction experiments at 3 GPa and 800–900 °C yield a ~500 µm-thick phlogopite–pyroxenite reaction zone enriched in K (phlogopite) and Ca–Si (pyroxenes), consistent with metasome formation in fore-arcs; sediment melts contain 5–10% volatiles and produce a coexisting fluid phase.
• Synthetic MT modeling for Cascadia reproduces a strong fore-arc conductor located ~20–30 km trenchward of the volcanic front (Mount Rainier), with the most conductive regions just above the slab (regions A/A′) and at ~10–30 km depth trenchward (region C) due to saline fluids from metasome crystallization. The arc column shows weaker conductivity because arc melts are less conductive than saline fluids at crustal temperatures. Inversion rms error ~1.5.
• Synthetic MT modeling for southern Kyushu reproduces a shallow conductor rising from the slab slightly trenchward of Kirishima. Agreement is strong above ~50 km depth; below this, resolution is poor in published data, yielding discrepancies. Inversion rms error ~2.3.
• The model explains why MT conductors peak in fore-arcs rather than under volcanic arcs: F-rich phlogopite and saline fluids in the metasome dominate conductivity. It also reconciles differences with seismic models: serpentinite yields low seismic velocities but weak MT response; phlogopite is highly conductive yet produces limited seismic velocity anomalies.
• Geochemical implication: the metasome is a likely source of K-rich (high-K to shoshonitic) magmas during slab rollback or fore-arc melting events; observations in Kyushu (Aso vs Kirishima) and Izu–Bonin–Mariana ‘alkalic volcano province’ are consistent with this.
• Seismicity implication: low frictional strength of phlogopite and presence of melts/fluids reduce likelihood of large thrust earthquakes in the metasomatized fore-arc, while elevated fluid pressures may promote small magnitude earthquakes and episodic tremor and slip; small M<3 events align with modeled ascent of sediment melt in Cascadia.

The findings address the long-standing question of the origin of strong fore-arc conductors in MT images by proposing a physically and chemically consistent mechanism: fore-arc melting of subducted sediments that react with and metasomatize the mantle wedge, forming conductive phlogopite–pyroxenite and releasing saline fluids. Synthetic resistivity structures based on experimentally constrained conductivities, geotherms, and plausible melt/fluid fractions reproduce key features of MT inversions in Cascadia and Kyushu, including the trenchward offset of the conductor from the volcanic front. This model unifies geophysical and geochemical observations by explaining fore-arc conductivity, potential sources of K-rich magmas during tectonic reconfiguration (e.g., slab rollback), and the distribution of seismicity. It reconciles MT–seismic disparities: saline fluids and F-rich phlogopite yield high MT conductivity without necessarily producing strong seismic velocity anomalies, while serpentinite affects seismic velocities but contributes little to conductivity. Overall, the results highlight the importance of sediment-derived melts in controlling fore-arc physical properties, arc magma sources, and seismic hazard patterns.

The study proposes and tests a model in which hydrous sediment melts form in the fore-arc, react with overlying mantle to create a conductive phlogopite–pyroxenite metasome, and exsolve saline fluids. Synthetic MT modeling constrained by experiments and geotherms reproduces observed fore-arc conductors in Cascadia and Kyushu, explaining why the strongest conductors are trenchward of arc volcanoes. The model provides a genetic link between subducted sediments, fore-arc metasomatism, fore-arc conductivity, K-rich volcanism during slab rollback, and seismicity patterns. Future work should image regions predicted to lack a conductive metasome (sediment-starved subductions) and test areas like the Izu–Bonin–Mariana ‘alkalic volcano province’ with MT. Improved MT resolution at >50 km depth, joint MT–seismic inversions, and further experiments quantifying volatile budgets and phlogopite conductivity under varying F–H2O conditions would refine and validate the model.

• MT inversions are non-unique; synthetic inversions show that forward-modelled broad low-resistivity zones can invert to thinner, deeper features while still fitting the data.
• Resolution at depths >50 km is poor in the Kyushu dataset, leading to discrepancies between synthetic and published models; inversion schemes impose differing conductive/resistive structures under limited resolution.
• The modelling simplifies subduction zone complexity (e.g., compositions, structures, fluid/melt pathways) and uses assumed thresholds (e.g., sediment melting at 675 °C; arc melting at ~800 °C) and fixed melt/fluid fractions (typically 1%).
• High-pressure experiments may overestimate volatile contents by not accounting for prior fluid loss; natural systems also receive additional fluids from serpentinites not fully captured in the experimental setup.
• Conductivity mixing uses modified Archie’s law with assumed good connectivity; real connectivity may vary with topology and strain.
• Upper crustal heterogeneity and structural controls on fluid migration are simplified (e.g., uniform brittle–ductile transition at 10 km; generic surface basin layer).