T23D-01 13:45h
Slab Dehydration: mechanical consequences on subduction zones dynamics
Subduction zones occur at the boundaries where tectonic plates converge and their dynamics is strongly coupled to the mantle wedge one. In this area, three main mechanical actors interact: the diving lithosphere, the overriding one, and the asthenosphere. To better understand the coupling phenomenon in the mantle wedge, we study the case of two converging oceanic plates, and we focus on the slab dehydration effect on the mantle wedge dynamics. Numerical experiments are performed using a thermomechanical code of convection. Water transfers are controlled by dehydration reactions within the slab and by hydration in the overlaying rocks. Dehydration and hydration reactions are both estimated according to accurate phase diagrams (Schmidt and Poli 1998, Bousquet et al. 1997). Rocks are assumed to be H$_2$O-saturated. Mantle rocks can be strongly weakened by the presence of water, that we model by adecreasing rock viscosity as a function of water content. Simulations show firstly that the amount of water released into the mantle wedge can hydrate the upper plate on about a 80 km thickness. Secondly, the hydrated rock softening in the mantle wedge enhances the corner flow. Furthermore, if the viscosity reduction coefficient, visco$_{dry}/$visco$_{wet}=f_{\nu}$, due to the presence of water is sufficiently large (greater than 50), secondary convection cells appear. As a consequence, the overriding lithosphere is delaminated by small blobs detachment and thins progressively. In these cases, the upper plate base is delaminated until the hydrated sublithospheric layer disappears. The delamination characteristic time seems to be proportional to $f_{\nu}^{-2/3}$. What brings about this upper plate erosion? Is it the enhanced corner flow, or the mechanical structure of the hydrated lithosphere? Simulations without subduction and localized hydration reactions in selected areas of the lithosphere and of the underlying asthenosphere are performed. For $f_{\nu} \ge 50$, a convective destabilization appears as in subduction experiment with similar characteristics. Therefore, for high hydrous strength reduction, the upper plate thinning during subduction is not controlled by the corner flow dynamics, but by the hydrated lithosphere strength. Thus we test the influence of the lithosphere bulk composition, using Gibbs free energy minimization calculations (de Capitani and Brown, 1987) to recalculate amounts of hydration. The water-contents we obtain in the sublithospheric layer are low because of amphiboles disappearance. This decrease of water contents strongly limits the hydrous softening in the upper plate. As a consequence, one can conclude from our preliminary results that the presence of amphiboles within the lithosphere favors local convection by reducing the viscosity.
T23D-02 14:00h
Geochemical Tracing of Mantle Flow above Subduction Zones
Geochemical tracing may be used to track mantle flow above and behind subduction zones and so provide an independent test of the applicability of seismic anisotropy measurements. The theory is that, if mantle flow is accompanied by decompression, then extraction of small degree melts from multi-component mantle leads to compositional gradients in the mantle in both isotope and trace element space. These gradients may be obtained by inverting geochemical data from the products of mantle melting. If mantle flow is accompanied by addition of a subduction fluid, then simultaneous melting and subduction component addition may also produce compositional gradients. Numerical experiments enable compositional gradients to be quantified in terms of the extent of melt extraction, mantle temperature and other variables. In addition, isotope and trace element systematics provide evidence for the provenance of the mantle entering the subduction system, with Hf isotopes and immobile trace elements providing a means of establishing provenance even from magmas generated directly above the dehydrating subducted plate. This work focuses on a series of geochemical maps which enable mantle flow to be traced for range of oceanic arc basin systems (Izu-Bonin-Mariana, Tonga-Vanuatu, Scotia, Manus) and, provisionally, some continental systems (Japan, Cascades). Using maps based on geochemical proxies for melt extraction (such as Ta/Yb), subduction-addition (such as Th/Ta) and mantle provenance (such as epsilon-Hf v epsilon Nd), it is possible to demonstrate the existence of a wide range of mantle flow regimes. Thus, the Izu and Japan systems appear to be characterised by simple trench-orthogonal flow, the Mariana system by dispersion away from several separate centers of mantle upwelling, the Tonga-Vanuatu system by unidirectional flow from beneath the Pacific plate in the north, and the Scotia system by bi-directional flow from both north and south. In a number of these cases, isotopic fingerprinting using immobile isotope ratios is critical for establishing the ultimate source of the mantle: for example, the mantle entering the Scotia system may be seen to originate from the Atlantic Bouvet domain rather than the Atlantic Tristan or the Pacific domains. These results are broadly in keeping with seismic anisotropy measurements to date while providing greater coverage but, of course, geochemical tracing is restricted to areas of magmatic activity. Integration of geophysical and geochemical methods may therefore be necessary to provide the maximum information on mantle flow.
T23D-03 INVITED 14:15h
Rheologic Controls on the Dynamic Evolution of Slabs in the Upper Mantle
Subduction of tectonic plates is characterized by long-lived subduction zones, asymmetric subduction and slab dip angles of 25--80$^\circ$ in the upper mantle. Several mechanisms proposed to explain the variation in observed dip include large-scale mantle flow, trench roll-back, and interaction of the slab with the transition zone. Previous dynamic models of subduction that include only Newtonian viscosity and moderately strong slabs generally fail to predict subduction angles less than 60--90$^\circ$ at shallow depths (100--300 km). We find that the observed characteristics of subduction are reproduced by viscous flow models, in which the rheologic structure is consistent with experimentally determined flow laws for Newtonian and non-Newtonian visco-plastic deformation of olivine. The properties of the models required to match the observed characteristics of slabs are: non-Newtonian viscosity in the mantle producing a weak mantle wedge ($10^{18}$--$10^{19}$~Pa s), a stiff slab interior ($10^{25}$~Pa s) limited by a plastic yield criterion and a weak plate boundary shear zone ($10^{20}$--$10^{21}$~Pa s). The shallow slab dip reaches a minimum of 25--30$^\circ$ for high convergence rates and a stiff slab, without trench roll-back or relative motion of the entire lithosphere with respect to the mantle, suggesting these other mechanisms are not the primary controls on slab geometry. The deep slab dip (350--650 km) decreases as the slab penetrates the stiffer (x10), Newtonian viscosity lower mantle, eventually stabilizing the upper mantle slab geometry.
T23D-04 INVITED 14:35h
Mantle wedge flow at the northern Cascadia subduction zone: Observational constraints and numerical models
We investigate mantle flow in the backarc of the northern Cascadia subduction zone (48-51\deg N). Surface heat flow, seismic velocity, effective elastic thickness, thermal isostasy, and xenolith studies indicate that the backarc is extremely hot, with estimated temperatures of 800-1000\deg C at the Moho (35 km depth) and a lithosphere thickness of only 50-60 km. The uniformly high temperatures are inferred across the entire 500 km width of the northern Cascadia backarc, despite the presence of the cool subducting Juan de Fuca plate on the west and the cool North America craton on the east. Local sources of heat are concluded to be negligible, and thus mantle flow is required to carry heat from depth into the backarc upper mantle. One constraint on backarc mantle flow directions comes from observations of seismic anisotropy, assuming that anisotropy is produced by the flow-induced lattice-preferred orientation anisotropic mantle minerals. Shear wave splitting from SKS arrivals at seismic stations in the Cascadia backarc shows significant anisotropy, with delay times of 1-1.5 s and fast directions parallel to the Juan de Fuca-North America convergence direction (~N70\deg E). These observations are consistent with the model of subduction-induced corner flow in the backarc mantle, generated by the entrainment of the mantle wedge material by the subducting plate. Numerical models are used to assess the thermal effects of corner flow. Using realistic model geometry and boundary conditions, we find that slab-driven corner flow at plate rates is too inefficient to transport the required amount of heat into the backarc to satisfy the thermal observations; an additional component of more vigourous flow is required. We propose that backarc mantle flow is dominated by vigourous small-scale thermal convection in a low viscosity backarc upper mantle. This is superimposed on a slower, regional corner flow type flow pattern, induced either by eastward subduction or westward motion of the over-riding plate. Although our study focusses on the Cascadia subduction zone, a compilation of thermal constraints and anisotropy observations suggests a similar backarc mantle flow regime at many subduction zones, including South America, Mexico, and NE Japan.
T23D-05 14:55h
Investigating the Link Between Mantle Flow and Seismic Anisotropy in Regions of Subduction.
Plate subduction is a fundamental feature of mantle convection and observations of seismic anisotropy hold insights into the mantle's response to subduction. In general, observations of SKS splitting from subduction regions worldwide show considerable variability in both orientation and magnitude. Splitting in local events that sample the backarc show relatively small amounts of splitting and usually trench parallel orientations. Splitting in deeper slab events exhibit a behaviour more like that observed for SKS phases. Cumulatively, these observations suggest significant sub-lithospheric and sub-slab anisotropy. These observations have motivated geodynamical numerical modelling of thermal convection to investigate sub-lithospheric flow in subduction environments. In order to gain insights into the connection between mantle flow and seismic anisotropy we present calculations developed for 3D Cartesian-geometry models of mantle convection with rigid plates. The calculations incorporate a fixed plate geometry with velocities that evolve dynamically with the convecting system. A depth-dependent Newtonian viscosity and internal heating are included in the calculations. Wrap-around boundary conditions are imposed on the model. Simple 3D models that mimic Earth-like subduction settings show that the mantle flow field can be very complex and qualitatively explain the diversity of splitting measurements worldwide. Slab morphology and plate age severely affect the flow in near-slab regions and can induce slab-parallel flow. Thermal flow models are converted to heterogeneous anisostropic elastic models and shear-waves are tracked through the resulting subduction models. Synthetic splitting parameters are then compared with those from subduction regions around the world. The model results help explain the highly variable splitting parameters observed worldwide. In the past, trench parallel flow has been explained via more ad-hoc models where the mantle is forced to flow around the slab. Our modelling shows that trench-parallel fast shear-wave polarisations can be more simply explained by thermal effects.
T23D-06 15:10h
Seismic Characterization of Mantle Flow in Subduction Systems: Can We Resolve a Hydrated Mantle Wedge?
The goal of this study is to provide new constraints on the resolvability of mantle flow in subduction zone settings as inferred by observations of seismic anisotropy. We are motivated by the multitude of seismic anisotropy observations in subduction systems that exhibit a broad range of shear wave splitting parameters. While fast polarization directions have typically been interpreted as a proxy for flow or maximum finite extension, experimental studies suggest that olivine slip systems change under hydrous conditions. The simple assumed relationship where fast directions are interpreted as parallel to mantle flow in anhydrous regions is not appropriate if the relative importance of slip systems varies, as would be expected in the mantle wedge of a subduction zone. For instance, fast polarization directions have been interpreted as orthogonal to mantle flow in some hydrated mantle regions, but the conditions that could produce seismic observations of flow-orthogonal fast directions due to hydrated mantle sources have not yet been examined. To this end, we predict shear wave splitting as a result of mantle silicate lattice-preferred orientation (LPO) development during mantle flow. We evaluate two-dimensional (2D) mantle flow models for LPO development using a theory that incorporates the combined effects of intracrystalline slip and dynamic recrystallization on textural development. We utilize the resulting textures to predict shear wave splitting for a population of seismic raypaths traversing the model by solving the Christoffel equation for each increment of a raypath and integrating the resulting predicted shear wave splitting recorded by each ray. For a model where LPO evolved to steady state under dry simple shear conditions, complete realignment of LPO after a transition to hydrated olivine rheology requires an additional $\sim$300% strain. However, we observe resolvable significant changes in fast directions ($>$10$\deg$) after only $\sim$125% strain. Using these results as a guide for regions of subduction systems that likely contain a transition from anhydrous to hydrated mantle, we expect that only very limited regions of the mantle may develop coherent LPO that reflects the hydrated system. We can test the hypothesis that trench-parallel fast directions indicate wet LPO textures by determining how far from the trench wet deformation conditions must hold, and whether this is consistent with estimates of where water release from the slab occurs. We are currently applying this analysis to a range of 2D subduction zone flow models to directly evaluate these effects and determine under what scenarios hydrated mantle may be imaged by shear wave splitting observations.
T23D-07 15:25h
Stress and Strain in the Mantle Wedge
In many subduction zones around the world the seismically fast vibration direction of split shear waves is parallel to the trench. Possible explanations include 3-D trench-parallel flow and water-induced olivine fabrics. The type-B olivine fabric, which may explain trench-parallel anisotropy, occurs above a critical stress under wet and low-temperature conditions. The formation of fabric is also dependent on the accumulation of sufficient finite strain in the dislocation creep regime. We investigate stress and finite-strain structures in the mantle wedge for high-resolution, time-dependent, finite element models with a kinematic-dynamic setup and water-dependent composite rheology. We use experimentally determined diffusion and dislocation creep parameters from Karato and Wu (1993) and Karato and Jung (2003). Our models show low stresses throughout most of the mantle wedge under dry conditions and that stress is even lower under wet conditions ($<$ 1 MPa). Relatively high stresses ($>$ 5 MPa) are confined mainly to boundary layers above the slab and below the overriding plate. These boundary layers are the best candidates for the water-induced type-B fabric that may be responsible for trench-parallel shear wave splitting in subduction zones.