Inside the Subduction Factory
Reproduced/modified by permission of American Geophysical Union.
John M. Eiler, Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, California
Geological models of subduction zones impact thinking about many of the central problems in the structure, dynamics, chemistry, and history of the solid earth. Should those models change, the effects will reach across the earth sciences. We are currently in the midst of such a change, brought on by several causes. First, the earth science community recently began an organized, multi-disciplinary study of subduction zones by way of the National Science Foundation's "Margins initiative." This has improved the quality of our descriptions of key focus areas and should continue to do so for the next several years. Less purposefully but of equal importance, several of the debates in subduction zone research have evolved in recent years, making us revisit both recent and old observations in a new light. This book consists of descriptions of these advances, written by the some of the leading participants. I describe the origin and organization of the book in the preface; here I review the context of previous thought regarding convergent margins, with particular focus on the link between tectonics and magmatism, and point out questions raised in the following chapters.
Section I: The Subducted Slab
Simon M. Peacock, Department of Geological Sciences, Arizona State University, Tempe, Arizona
Variations in subduction-zone seismicity, seismic velocity, and arc magmatism reflect differences in the thermal structure and metamorphic reactions occurring in subducting oceanic lithosphere. Current kinematic and dynamical models of subduction zones predict cool slab-mantle interface temperatures less than one-half of the initial mantle temperature. Weak rocks along the slab-mantle interface likely limit the rate of shear heating; surface heat flux measurements and other observations suggest interface shear stresses are 0 - 40 MPa, consistent with this expectation. Thermal models of the NE Japan, Izu-Bonin, and Aleutian subduction zones predict slabmantle interface temperatures of ~500 ˚C beneath the volcanic front. In such cool subduction zones, subducting oceanic crust transforms to eclogite at depths > 100 km and temperatures are too low to permit partial melting of subducted sediments or crust. In the Nankai subduction zone, where the incoming Philippine Sea Plate is unusually warm, predicted interface temperatures beneath sparse Holocene volcanoes are ~800 ˚C and eclogite transformation, slab dehydration reactions, and intermediate-depth seismicity occur at < 60 km depth. The geometry and vigor of mantle-wedge convection remains considerably uncertain; models incorporating strongly temperature-dependent mantle viscosity predict significantly higher slab-mantle interface temperatures.
Tim Elliott, epartment of Earth Sciences, University of Bristol, Bristol, UK
A new global compilation of high precision trace element and isotopic analyses from mafic island arc lavas is explored to highlight the key geochemical features of volcanic front lavas. Two distinct components from the slab can be identified in island arc lavas. One component dominates the budgets of most incompatible trace elements and another affects a smaller range of elements, most notably Ba, Pb and Sr. These contrasting slab components require different ultimate sources and mechanisms of transport to the mantle wedge. The first component is argued to be a melt of the down-going sediment, while the second is likely an aqueous fluid derived from the altered mafic oceanic crust. U-series nuclides constrain the timing of slab component additions. The sediment component is close to 238U-230Th equilibrium implying >350ky since the last major fractionation of U and Th, plausibly the time since sediment melting. Lavas dominated by the 'fluid' component can have extreme 226Ra-230Th disequilibrium together with elevated ratios of the stable Ba/Th analogue. In its simplest interpretation, this observation implies only a few thousand years elapse between 'fluid' release from the slab, and eruption of this component in arc lavas. In most cases the two subduction components appear to be added to a recently melt-depleted mantle. This suggests the mantle wedge is fed with material processed through a back-arc melting regime. These first order geochemical observations and inferences place important constraints on physical models of the subduction zone.
George Helffrich, Earth Sciences, University of Bristol, Bristol, United Kingdom
This chapter reviews the basic concepts involving waves that electromagnetic and seismic investigations rely upon, for intiutive understanding by non-specialists in either field. Following sections dealing with the general properties of waves, further specialized to each method, the review summarizes investigative results from key subduction zone studies that illustrate how the methods were used.
Section II: The Mantle Wedge
Douglas A. Wiens and Gideon P. Smith, Department of Earth and Planetary Sciences, Washington University, St. Louis, Missouri
The mantle wedge of a subduction zone is characterized by low seismic velocities and high attenuation, indicative of temperatures approaching the solidus and the possible presence of melt and volatiles. Tomographic images show a low velocity region above the slab extending from 150 km depth up to the volcanic front. The low velocities result at least partially from volatiles fluxed off the slab, which lower the solidus and thus raise the homologous temperature of the wedge material. Subduction zones with active back-arc spreading centers also show large low velocity and high attenuation seismic anomalies beneath the backarc basin, indicating that a broad zone of magma production feeds the backarc spreading center. The magnitude of the velocity anomaly is consistent with the presence of approximately 1% partial melt at depths of 30-90 km. The best-imaged arc-backarc system, the Tonga-Lau region, suggests that the zone of backarc magma production is separated from the island arc source region within the depth range of primary magma production. However, the anomalies merge at depths greater than 100 km, suggesting that small slab components of backarc magmas may originate through interactions at these depths. Slow velocities extend to 400 km depth beneath backarc basins, and these deep anomalies may result from the release of volatiles transported to the top of the transition zone by hydrous minerals in the slab. Observations of seismic anisotropy can provide direct evidence for the flow pattern in the mantle wedge. Slab parallel fast directions suggesting along-arc flow within the mantle wedge are found in most, but not all, subduction zones. The Tonga-Lau region shows a complex pattern of fast directions, with along-strike fast directions beneath the Tonga island arc and convergence-parallel fast directions to the west of the Lau backarc spreading center. The pattern of flow in the Lau backarc is consistent with southward mantle flow inferred from geochemical data. Geodynamic modeling suggests several possible mechanisms of flow within the mantle wedge, which may help explain the diverse observations. Viscous coupling between the backarc flow and the downgoing plate should produce induced flow within the backarc, with flow directions parallel to the convergence direction. In contrast, subduction zone roll-back may produce along-arc flow. This latter model matches the observations in most subduction zones suggesting that viscous coupling does not exert a strong control on the flow pattern in the mantle wedge.
Greg Hirth, Department of Geology and Geophysics, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts
In this manuscript we review experimental constraints for the viscosity of the upper mantle. We first analyze experimental data to provide a critical review of flow law parameters for olivine aggregates and single crystals deformed in the diffusion creep and dislocation creep regimes under both wet and dry conditions. Using reasonable values for the physical state of the upper mantle, the viscosities predicted by extrapolation of the experimental flow laws compare well with independent estimates for the viscosity of the oceanic mantle, which is approximately 1019 Pa s at a depth of ~100 km. The viscosity of the mantle wedge of subduction zones could be even lower if the flux of water through it can result in olivine water contents greater than those estimated for the oceanic asthenosphere and promote the onset of melting. Calculations of the partitioning of water between hydrous melt and mantle peridotite suggest that the water content of the residue of arc melting is similar to that estimated for the asthenosphere. Thus, transport of water from the slab into the mantle wedge can continually replenish the water content of the upper mantle and facilitate the existence of a low viscosity asthenosphere.
Glenn A. Gaetani, Department of Geology and Geophysics, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts
Experimental studies show that H2O affects most aspects of melt generation in the sub-arc mantle wedge. For example, dissolved H2O modifies the major element composition of peridotite partial melt by increasing the ratio of SiO2 to MgO + FeO, mimicking the effect of decreasing pressure during anhydrous partial melting. Comparison of the normalized (anhydrous) compositions of experimentally produced hydrous and anhydrous melts shows that SiO2 increases by ~1 wt% with addition of 3 to 6 wt% dissolved H2O, while FeO + MgO decreases by ~2 wt%. Furthermore, mobility of partial melt in mantle peridotite may increase due to the influence of H2O. Orthopyroxene-melt dihedral angles are ~70˚ under anhydrous conditions, trapping small amounts of melt at 4 grain junctions, but they decrease to ~52˚ under hydrous conditions, allowing connectivity down to very low melt fractions. Dissolved H2O also decreases melt density and viscosity that, combined with enhanced connectivity, allows hydrous melt to segregate very efficiently from residual peridotite. Less melt is produced by hydrous peridotite, relative to anhydrous peridotite, for a given temperature increase or pressure decrease, because of the monotonic decrease of dissolved H2O with increasing extent of melting. Primitive arc magmas with high pre-eruptive H2O contents may form when a peridotite partial melt that is initially near fluid saturation percolates upward through the mantle wedge, maintaining equilibrium with hotter, overlying peridotite by dissolving the surrounding rock (reactive porous flow). Adiabatic decompression melting may occur in regions where hot mantle flows from the back arc into the wedge corner, generating nearly anhydrous partial melt.
Shun-ichiro Karato, Yale University, Department of Geology and Geophysics, New Haven, Connecticut
Variations in water (hydrogen) content in Earth's (upper) mantle can be inferred from geophysical observations if the relationship between water content and relevant physical properties is known and if high-resolution geophysical measurements are available. This paper reviews the current status of mineral physics understanding of the effects of water on elastic and non-elastic deformation of minerals such as olivine and its influence on seismologically measurable properties. Important effects of water on seismic wave propagation are through indirect effects due to hydrogen-related defects in nominally anhydrous minerals as opposed to the direct effects caused by the formation of hydrous minerals. Two cases of indirect effects are reviewed: (i) effects through the enhancement of anelasticity and (ii) effects through the modifications of lattice preferred orientation. The former causes enhanced attenuation (low Q) and low velocities by the increase of water content and the latter modifies the nature of seismic anisotropy. Experimental data are reviewed to formulate ways to infer water content from seismological data and analytical equations are derived that relate velocity and attenuation anomalies to anomalies in temperature and/or water content. The results are applied to infer the distribution of water in Earth's upper mantle using seismological observations. In subduction zones, the regions of high water content in the shallow upper mantle (200 km) are inferred to be localized to the mantle beneath current or recent volcanoes although wider distribution is hinted in the deeper portions (200 km). In the upper mantle beneath hot spot volcanoes such as Hawaii and Iceland, both seismic wave attenuation and anisotropy measurements suggest the presence of a column of material with a high water content, indicating that Hawaii and Iceland are not only "hot" spots but also "wet" spots.
Section III: Focus Regions
M. J. Carr and M. D. Feigenson, Department of Geological Sciences, Rutgers University
Most Central American volcanoes occur in an impressive volcanic front that trends parallel to the strike of the subducting Cocos Plate. The volcanic front is a chain, made of right-stepping, linear segments, 100 to 300 Km in length. Volcanoes cluster into centers, whose spacing is random but averages about 27 Km. These closely spaced, easily accessible volcanic centers allow mapping of geochemical variations along the volcanic front. Abundant back-arc volcanoes in southeast Guatemala and central Honduras allow two cross-arc transects. Several element and isotope ratios (e.g. Ba/La, U/Th, B/La, 10Be/9Be, 87Sr/86Sr) that are thought to signal subducted marine sediments or altered MORB consistently define a chevron pattern along the arc, with its maximum in Nicaragua. Ba/La, a particularly sensitive signal, is 130 at the maximum in Nicaragua but decreases out on the limbs to 40 in Guatemala and 20 in Costa Rica, which is just above the nominal mantle value of 15. This high amplitude regional variation, roughly symmetrical about Nicaragua, contrasts with the near constancy, or small gradient, in several plate tectonic parameters such as convergence rate, age of the subducting Cocos Plate, and thickness and type of subducted sediment. The large geochemical changes over relatively short distances make Central America an important margin for seeking the tectonic causes of geochemical variations; the regional variation has both a high amplitude and structure, including flat areas and gradients. The geochemical database continues to improve and is already adequate to compare to tectonic models with length scales of 100 Km or longer.
Robert J. Stern, Geosciences Department, University of Texas at Dallas, Richardson, Texas
The Izu-Bonin-Mariana (IBM) arc system extends 2800km from near Tokyo, Japan to Guam and is an outstanding example of an intraoceanic convergent margin (IOCM). Inputs from sub-arc crust are minimized at IOCMs and output fluxes from the Subduction Factory can be more confidently assessed than for arcs built on continental crust. The history of the IBM IOCM since subduction began about 43 Ma may be better understood than for any other convergent margin. IBM subducts the oldest seafloor on the planet and is under strong extension. The stratigraphy of the western Pacific plate being subducted beneath IBM varies simply parallel to the arc, with abundant off-ridge volcanics and volcaniclastics in the south which diminish northward, and this seafloor is completely subducted. The Wadati-Benioff Zone varies simply along strike, from dipping gently and failing to penetrate the 660 km discontinuity in the north to plunging vertically into the deep mantle in the south. The northern IBM arc is about 22km thick, with a felsic middle crust; this middle crust is exposed in the collision zone at the northern end of the IBM IOCM. There are four Subduction Factory outputs across the IBM IOCM: (1) serpentinite mud volcanoes in the forearc, and as lavas erupted from along (2) the volcanic front of the arc and (3) back-arc basin and (4) from arc cross-chains. This contribution summarizes our present understanding of matter fed into and produced by the IBM Subduction Factory, with the intention of motivating scientific efforts to understand this outstanding example of one of earth's most dynamic, mysterious, and important geosystems.
Peter B. Kelemen, Dept. of Geology & Geophysics, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts
Based on a compilation of whole rock geochemistry for approximately 1100 lava samples and 200 plutonic rock samples from the Aleutian island arc, we characterize along-strike variation, including data for the western part of the arc which has recently become available. We concentrate on the observation that western Aleutian, high Mg# andesite compositions bracket the composition of the continental crust. Isotope data show that this is not due to recycling of terrigenous sediments. Thus, the western Aleutians can provide insight into genesis of juvenile continental crust. The composition of primitive magmas (molar Mg# > 0.6) varies systematically along the strike of the arc. Concentrations of SiO2, Na2O and perhaps K2O increase from east to west, while MgO, FeO, CaO decrease. Thus, primitive magmas in the central and eastern Aleutians (east of 174˚W) are mainly basalts, while those in the western Aleutians are mainly andesites. Along-strike variation in Aleutian magma compositions may be related to a westward decrease in sediment input, and/or to the westward decrease in down-dip subduction velocity. 206Pb/204Pb, 207Pb/204Pb, 208Pb/204Pb and 87Sr/86Sr all decrease from east to west, whereas 143Nd/144Nd increases from east to west. These data, together with analyses of sediment from DSDP Site 183, indicate that the proportion of recycled sediment in Aleutian magmas decreases from east to west. Some proposed trace element signatures of sediment recycling in arc magmas do not vary systematically along the strike of the Aleutians, and do not correlate with radiogenic isotope variations. Thus, for example, Th/Nb and fractionation-corrected K concentration in Aleutian lavas are not related to the flux of subducting sediment. Th/La is strongly correlated with Ba/La, rendering it doubtful that Ba/La is a proxy for an aqueous fluid component derived from subducted basalt. Ce/Pb > 4 is common in Aleutian lavas west of 174˚W, in lavas with MORB-like Pb, Sr and Nd isotope ratios, and is also found behind the main arc trend in the central Aleutians. Thus, Ce/Pb in Aleutian lavas with MORB-like isotope ratios is not always low, and may be affected by a component derived from partial melting of subducted basalt in eclogite facies. Enriched, primitive andesites, with high Sr/Y, steep REE patterns, and low Yb and Y, are an important lava type in the Aleutians west of 174˚W. High Sr/Y and Dy/Yb, indicative of abundant garnet in the source of melting, are correlated with major element systematics. Lavas with a "garnet signature" have high SiO2, Na2O and K2O. Enriched, primitive Aleutian andesites did not form via crystal fractionation from primitive basalt, melting of primitive basalt, mixing of primitive basalt and evolved dacite, or partial melting of metasomatized peridotite. Instead, as proposed by Kay , they formed by partial melting of subducted eclogite, followed by reaction with the mantle during ascent into the arc crust. In the eastern Aleutians, an eclogite-melt signature is less evident, but trace element systematics have led earlier workers to the hypothesis that partial melts of subducted sediment are an important component. Thus, partial melting of subducted sediment and/or basalt is occurring beneath most of the present-day Aleutian arc. This is consistent with the most recent thermal models for arcs. Enriched primitive andesites are observed mainly in the west because the mantle is relatively cold, whereas in the east, a hotter wedge gives rise to abundant, mantle-derived basalts which obscure the subduction zone melt component. Enriched primitive andesites, partial melts of eclogite, and products of small amounts of reaction between eclogite melts and mantle peridotite under conditions of decreasing magma mass, all have middle to heavy REE slopes that are steeper than those in typical Aleutian andesite and continental crust. Thus, direct partial melts of eclogite--without magma/mantle interaction--do not form an important component in the continental crust. Extensive reaction, with gradually increasing melt mass and melt/rock ratios ~ 0.1 to ~ 0.01, is required to increase heavy REE concentrations to the levels observed in most Aleutian andesites and in continental crust.
Section IV: Synthesis
Yoshiyuki Tatsumi, Institute for Frontier Research on Earth Evolution (IFREE), Japan Marine Science and Technology Center (JAMSTEC), Yokosuka, Japan
Improved understanding of the complex processes associated with subduction zone magmatism requires identifying and explaining tectonic, petrological, geochemical, and geophysical characteristics common to most subduction zones. These characteristics include: (1) the presence of dual volcanic chains within a single volcanic arc, (2) the constant depths to the surface of the subducting lithosphere beneath trench- and backarc-side volcanic chains, ~110 km and ~170 km, respectively, (3) greater volume of magma production beneath the trench-side volcanic chain (4) selective enrichment of particular incompatible elements, (5) systematic across-arc variations in incompatible element concentrations and isotopic/element ratios, (6) location of a high-temperature (>1350˚C) region within the mantle wedge, (7) occurrenceof localized high-temperatures (~1300˚C) immediately beneath the crust/mantle boundary, (8) location of very low-velocity regions within the mantle wedge. It is suggested here that dehydration processes and associated selective element transport, which take place in both the subducting lithosphere and the downdragged hydrated peridotite layer at the base of the mantle wedge, and secondary convection within the mantle wedge induced by plate subduction are largely responsible for the observed characteristics of subduction zone magmatism.
Peter B. Kelemen, Dept. of Geology & Geophysics, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts
We summarize petrological and seismic constraints on the temperature of arc lower crust and shallow mantle, and show that published thermal models are inconsistent with these constraints. We then present thermal models incorporating temperature dependent viscosity, using widely accepted values for activation energy and asthenospheric viscosity. These produce thin thermal boundary layers in the wedge corner, and an overall thermal structure that is consistent with other temperature constraints. Some of these models predict partial melting of subducted sediment and/or basalt, even though we did not incorporate the effect of shear heating We obtain these results for subduction of 50 Myr old oceanic crust at 60 km/Myr, and even for subduction of 80 Myr old crust at 80 km/Myr, suggesting that melting of subducted crust may not be not restricted to slow subduction of young oceanic crust.
Last updated Friday, July 27, 2007