Dehydration and melting experiments constrain the fate of subducted sediments

Abstract

Geochemical tracers demonstrate that elements are cycled from subducted sediments into the arc melting regime at subduction zones, although the transfer mechanism is poorly understood. Are key elements (Th, Be, Rb) lost during sediment dehydration or is sediment melting required? To investigate this question, we conducted phase equilibria and trace element partitioning experiments on a pelagic red clay for conditions appropriate to the slab beneath arc volcanoes (2–4 GPa, 600°–1000°C). Using both piston cylinders and multianvils, we determined the solidus, phase stabilities, and major element compositions of coexisting phases. The solidus (H 2 O + Cl fluid‐saturated) was located at 775 ± 25°C at 2 GPa, 810 ± 15°C at 3 GPa, and 1025 ± 25°C at 4 GPa with noevidence for complete miscibility between melt and fluid. This sediment composition produces a profusion of phases both above and below the solidus: garnet, jadeitic pyroxene, alkali‐rich amphibole, phengite, biotite, magnetite, coesite, kyanite, apatite, zircon, Cl‐rich fluids, and peraluminous to peralkaline granitic melts. At 2 GPa the phengite dehydration solidus is at 800°–825°C, while biotite breaks down between 850° and 900°C. To explore trace element partitioning across the solidus at 2 GPa, we used diamonds to trap fluids and melts. Both the bulk sediment residues and diamond traps were analyzed postexperiment by inductively coupled plasma–mass spectrometry (ICP‐MS) and inductively coupled plasma–atomic emission spectrometry (ICP‐AES) for 40 elements for which we calculated bulk partition coefficients (D = C solid/C fluid). Below the solidus, Rb, Sr, Ba, and Pb showed the greatest mobility (D ∼ 0.5–1.0), while at the solidus, Th and Be became notably partitioned into the melt (D values changing from >2.0 to <1.0). K and Rb D values fall below 1.0 when the micas breakdown. Only at the solidus do Th and Rb attain similar partition coefficients, a condition required by arc data. Taken together, the experimental results indicate that critical elements (Th and Be) require sediment melting to be efficiently transferred to the arc. This conclusion is at odds with most thermal models for subduction zones, which predict slab temperatures more than 100°C lower than sediment solidi. Thus the condition of sediment melting (with oceanic crust dehydration) may provide new constraints on the next generation of thermal/geodynamical models of subduction zones.