The thermal state of the top of the subducting plate is strongly affected by the mantle wedge flow structure. We identify three main factors controlling the influence of the corner flow on the slab surface temperature: (1) the mantle rheology, (2) the interplate decoupling depth, and (3) the intensity of thermomechanical ablation of the overriding plate. The first two factors are discussed using results from published simulations. We perform 2D numerical simulations of mantle wedge flow to assess the role of the third factor. The non-Newtonian rheology depends on temperature, pressure, yield stress, and composition. The thermochemical convection code includes a water transfer model and the water weakening effect on mantle rocks. Slab dehydration triggers water saturation of the overlying mantle wedge and upper plate. When weakening of mantle rocks by hydration is included in the simulations, the interplate decoupling depth decreases and the upper plate thermal thinning increases. A relatively cold blanket develops on top of the slab below the interplate decoupling depth. Cold materials removed from the thinned upper plate are advected by the corner flow and are accreted to the viscous layer dragged along the slab surface. As a result, the thickness of the insulating layer covering the slab surface increases with the water weakening effect, together with the upper plate thermal thinning rate.Pressure–temperature (P−T) paths followed by crustal rocks during subduction quantify the influence of upper plate thinning processes on the slab surface thermal state. The slab surface temperature below 100 km can be lowered by as much as 130 °C due to increased thermal convection at the base of the upper plate. At shallower depths (< 100 km), this effect competes with the heating of the slab induced by a shallowing of the interplate decoupling depth. For a small water weakening effect, the enhanced corner flow mainly yields a shallow decoupling depth and warms the slab top from 50 to 100 km. For a large weakening effect, the slab surface is cooled by convective drips detaching from the overriding lithosphere that counter-balance the effect of a shallow decoupling depth. P−T paths in case of efficient upper plate thinning are thus inferred to be uniformly colder than predicted from the slab thermal parameter (age × velocity) and non-Newtonian wedge rheology.

The thermal state of the top of the subducting plate is strongly affected by the mantle wedge flow structure. We identify three main factors controlling the influence of the corner flow on the slab surface temperature: (1) the mantle rheology, (2) the interplate decoupling depth, and (3) the intensity of thermomechanical ablation of the overriding plate. The first two factors are discussed using results from published simulations. We perform 2D numerical simulations of mantle wedge flow to assess the role of the third factor. The non-Newtonian rheology depends on temperature, pressure, yield stress, and composition. The thermochemical convection code includes a water transfer model and the water weakening effect on mantle rocks. Slab dehydration triggers water saturation of the overlying mantle wedge and upper plate. When weakening of mantle rocks by hydration is included in the simulations, the interplate decoupling depth decreases and the upper plate thermal thinning increases. A relatively cold blanket develops on top of the slab below the interplate decoupling depth. Cold materials removed from the thinned upper plate are advected by the corner flow and are accreted to the viscous layer dragged along the slab surface. As a result, the thickness of the insulating layer covering the slab surface increases with the water weakening effect, together with the upper plate thermal thinning rate.Pressure–temperature (P−T) paths followed by crustal rocks during subduction quantify the influence of upper plate thinning processes on the slab surface thermal state. The slab surface temperature below 100 km can be lowered by as much as 130 °C due to increased thermal convection at the base of the upper plate. At shallower depths (< 100 km), this effect competes with the heating of the slab induced by a shallowing of the interplate decoupling depth. For a small water weakening effect, the enhanced corner flow mainly yields a shallow decoupling depth and warms the slab top from 50 to 100 km. For a large weakening effect, the slab surface is cooled by convective drips detaching from the overriding lithosphere that counter-balance the effect of a shallow decoupling depth. P−T paths in case of efficient upper plate thinning are thus inferred to be uniformly colder than predicted from the slab thermal parameter (age × velocity) and non-Newtonian wedge rheology.

At continental subduction initiation, the continental crust buoyancy may induce, first, a convergence slowdown, and second, a compressive stress increase that could lead to the forearc lithosphere rupture. Both processes could influence the slab surface P–T conditions, favoring on one side crust partial melting or on the opposite the formation of ultra-high pressure/low temperature (UHP-LT) mineral. We quantify these two effects by performing numerical simulations of subduction. Water transfers are computed as a function of slab dehydration/overlying mantle hydration reactions, and a strength decrease is imposed for hydrated mantle rocks. The model starts with an old oceanic plate ( 100 Ma) subducting for 145.5 Myr with a 5 cm/yr convergence rate. The arc lithosphere is thermally thinned between 100 km and 310 km away from the trench, due to small-scale convection occuring in the water-saturated mantle wedge. We test the influence of convergence slowdown by carrying on subduction with a decreased convergence rate (≤ 2 cm/yr). Surprisingly, the subduction slowdown yields not only a strong slab warming at great depth (> 80 km), but also a significant cooling of the forearc lithosphere at shallower depth. The convergence slowdown increases the subducted crust temperature at 90 km depth to 705 ± 62 °C, depending on the convergence rate reduction, and might thus favor the oceanic crust partial melting in presence of water. For subduction velocities ≤ 1 cm/yr, slab breakoff is triggered 20–32 Myr after slowdown onset, due to a drastic slab thermal weakening in the vicinity of the interplate plane base. At last, the rupture of the weakened forearc is simulated by imposing in the thinnest part of the overlying lithosphere a dipping weakness plane. For convergence with rates ≥ 1 cm/yr, the thinned forearc first shortens, then starts subducting along the slab surface. The forearc lithosphere subduction stops the slab surface warming by hot asthenosphere corner flow, and decreases in a first stage the slab surface temperature to 630 ± 20 °C at 80 km depth, in agreement with P–T range inferred from natural records of UHP-LT metamorphism. The subducted crust temperature is further reduced to 405 ± 10 °C for the crust directly buried below the subducting forearc. Such a cold thermal state at great depth has never been sampled in collision zones, suggesting that forearc subduction might not be always required to explain UHP-LT metamorphsim.

At continental subduction initiation, the continental crust buoyancy may induce, first, a convergence slowdown, and second, a compressive stress increase that could lead to the forearc lithosphere rupture. Both processes could influence the slab surface P–T conditions, favoring on one side crust partial melting or on the opposite the formation of ultra-high pressure/low temperature (UHP-LT) mineral. We quantify these two effects by performing numerical simulations of subduction. Water transfers are computed as a function of slab dehydration/overlying mantle hydration reactions, and a strength decrease is imposed for hydrated mantle rocks. The model starts with an old oceanic plate ( 100 Ma) subducting for 145.5 Myr with a 5 cm/yr convergence rate. The arc lithosphere is thermally thinned between 100 km and 310 km away from the trench, due to small-scale convection occuring in the water-saturated mantle wedge. We test the influence of convergence slowdown by carrying on subduction with a decreased convergence rate (≤ 2 cm/yr). Surprisingly, the subduction slowdown yields not only a strong slab warming at great depth (> 80 km), but also a significant cooling of the forearc lithosphere at shallower depth. The convergence slowdown increases the subducted crust temperature at 90 km depth to 705 ± 62 °C, depending on the convergence rate reduction, and might thus favor the oceanic crust partial melting in presence of water. For subduction velocities ≤ 1 cm/yr, slab breakoff is triggered 20–32 Myr after slowdown onset, due to a drastic slab thermal weakening in the vicinity of the interplate plane base. At last, the rupture of the weakened forearc is simulated by imposing in the thinnest part of the overlying lithosphere a dipping weakness plane. For convergence with rates ≥ 1 cm/yr, the thinned forearc first shortens, then starts subducting along the slab surface. The forearc lithosphere subduction stops the slab surface warming by hot asthenosphere corner flow, and decreases in a first stage the slab surface temperature to 630 ± 20 °C at 80 km depth, in agreement with P–T range inferred from natural records of UHP-LT metamorphism. The subducted crust temperature is further reduced to 405 ± 10 °C for the crust directly buried below the subducting forearc. Such a cold thermal state at great depth has never been sampled in collision zones, suggesting that forearc subduction might not be always required to explain UHP-LT metamorphsim.