Deep subduction of H 2 O and de ection of volcanic chain towards backarc near triple junction due to lower temperature

Transcription

1 Earth and Planetary Science Letters 181 (2000) 41^46 Deep subduction of H 2 O and de ection of volcanic chain towards backarc near triple junction due to lower temperature Hikaru Iwamori Department of Earth and Planetary Sciences, University of Tokyo, Tokyo 113, Japan Received 3 December 1999; received in revised form 5 June 2000; accepted 7 June 2000 Abstract In central Japan near the triple junction of the Pacific, Philippine Sea and North American (or Ohotsuku) plates, the volcanic chain deflects towards backarc compared to the adjacent northeast Japan and Izu^Bonin arcs, and lies V200^ 300 km above the Wadati^Benioff zone. Numerical modeling shows that thermal recovery of the subducting Pacific plate is slow, due to the overlapping Philippine sea plate, which shifts the dehydration reactions to greater depths along the Pacific plate and causes the magmatism above the deep Wadati^Benioff zone. The low geothermal gradient along the subducting Pacific plate also implies that a considerable amount of H 2 O is carried further down by phase A which is formed just above the subducting plate, without being released for magmatism. Central Japan can be regarded as an entrance for H 2 O into the deep mantle. ß 2000 Elsevier Science B.V. All rights reserved. Keywords: subduction; numerical models; water; dehydration; volcanic belts 1. Introduction The Paci c plate subducts beneath the Japan arc on the North American (or Ohotsuku) plate at the Japan trench and beneath the Izu arc on the Philippine sea plate at the Izu^Ogasawara trench. This geometry produces a long volcanic chain from Kamtchatka to the Marianas without large gaps, including the northeast Japan arc to the Izu arc (Fig. 1). In general, volcanic chains in subduction zones are parallel to the trenches, and 80^200 km above the Wadati^Benio zone [1,2]. This is the case with the northeast Japan and Izu arcs. Compared to the adjacent northeast Japan and Izu arcs, the volcanic chain in central Japan to the west of the Kanto region de ects towards the backarc region, and is 170^300 km above the Wadati^Benio zone (Fig. 1) [3,4]. In addition to the Paci c plate, the Philippine sea plate subducts beneath the Japan arc from the Sagami and Nankai troughs, with the triple junction near 34³N and 142³E (Fig. 1). This plate con guration is thought to have been established 15 Myr ago, with the formation of the Shikoku basin (a backarc basin of the Philippine sea plate) [5,6]. Consequently, beneath the Kanto region, the two plates overlap. Seismic studies with dense networks show that the subducted Philippine sea plate with a thickness of about 30 km exists in the mantle wedge above the subducting Paci c plate [7,8]. The leading edge of the Philippine sea plate reaches beneath the north Kanto region (V36.5³N, Fig. 1), where the depth of its upper surface is estimated to be at least 60 km [7^12]. This paper presents a two-dimensional model for the mantle ow, thermal structure, and uid X / 00 / $ ^ see front matter ß 2000 Elsevier Science B.V. All rights reserved. PII: S X(00)

2 42 H. Iwamori / Earth and Planetary Science Letters 181 (2000) 41^46 Fig. 1. Map showing the distribution of active volcanoes (solid triangles) and the plate con gurations with the depth contour of the upper surface of the Wadati^Benio zone of the subducting Paci c plate (thick broken lines) [3,4], and the northern limit of the Philippine sea plate beneath the Kanto and Tokai regions (thin broken lines) [7^12]. `I^O trench' indicates the Izu^Ogasawara trench. The plate velocities relative to the Japan arc (V10 cm yr 31 for the Paci c plate, V3 cm yr 31 for the Philippine sea plate) are shown by the arrows. distribution beneath central Japan where the two plates overlap. The main aim of this paper is to discuss the origin of de ection of the volcanic chain in the region, which would contribute to understanding the uid processes and magmatism in subduction zones in general. 2. Flow and thermal structures The plate con guration described in the previous section can be represented by the model cross-section shown in Fig. 2, along the extension of the arrow showing the motion of the Paci c plate (Fig. 1). In Fig. 2, the Paci c plate with a thickness of about 100 km is modeled as a subducting slab from the upper right corner with an angle of 30³ and velocity of 10 cm yr 31. Temperature boundary conditions appropriate for 130 Myr old plate are imposed on the right hand side of the box. The Philippine sea plate is represented by a xed slab in the mantle wedge above the Paci c plate, since the subduction velocity near the triple junction is small for the last 4 Myr (1.8 cm yr 31 ) [5] when compared to that of the Paci c plate. The viscous solid ow in the mantle wedge was calculated for this two plate con guration, together with the thermal structure, phase equilibria, uid generation^migration, and melting, using a nite di erence scheme after [13]. A cross sectional area of 520U300 km (lateral and vertical) has been discretized into 300U300 grid cells, with additional ner cells along the subducting slab. In the model of Fig. 2, the position of the Philippine sea plate is xed at (horizontal distance from the trench, depth) = (87,30)-(277,90)-(277,60)-(182,30) km, beneath the crust of 30 km thick, based on [7^12]. The depth of the Wadati^Benio zone beneath the points are 50, 160, 160 and 105 km, respectively. With this geometry, a drag ow in the mantle wedge of a constant viscosity, which is induced by subduction of the Paci c plate, is calculated: a biharmonic equation in terms of the stream function was solved numerically. If the Philippine sea plate does not exist, the solution coincides with an analytic corner ow solution in a simple triangular wedge [14]. On the backarc side boundary (left hand side wall of the box of Fig. 2), this analytic corner ow solution is assumed as a ow boundary condition. Then the energy equation which involves advection of heat along the ow, heat conduction, and heat for melting has been solved, based on the formulation by [13]. The thermal boundary conditions are as follows. The surface temperature of 0³C. An error function gradient for the plate age of 130 Myr old [15] and an adiabatic gradient underneath for the oceanic side boundary are assumed with the thermal expansion coe cient of 2.4U10 35 ³K 31, the heat capacity at constant pressure of 1.2U10 3 Jkg 31 K 31, the thermal diffusivity of 1.0U10 36 m 2 s 31, and the density of solid of 3.3U10 3 kg m 33. A linear gradient within the crust and within a thermal boundary layer of 18 km beneath the crust to produce the surface heat ux of 100 mw m 32, and an adiabatic gradient underneath for the arc side boundary are

3 H. Iwamori / Earth and Planetary Science Letters 181 (2000) 41^46 43 Fig. 2. The calculated distribution of (a) H 2 O (total amount contained in all the phases present, including the solid phase), (b) aqueous uid, and (c) melt. The oceanic crust with a thickness of 7 km, which is assumed to contain 6 wt% H 2 O initially, subducts from the upper right corner. The calculation procedure is after [13], and is explained in the text. The solid lines indicate the isothermal contours with a 200³C interval. The broken lines indicate the stream lines in the mantle wedge. PSP and PAP represent the Philippine Sea plate and the Paci c plate, respectively. assumed, which gives the potential temperature V1250³C. Zero heat ux at the bottom boundary. Melting is involved in the calculation as will be described later, and the latent heat of melting is set to be 3.5U10 2 Jkg 31 ³K 31. Heat generation by decay of radionuclides and viscous dissipation is not included. The calculated streamlines and the temperature contours are shown in Fig. 2 by the broken and solid lines, respectively. They show that the presence of the subducted Philippine sea plate signi cantly distorts the corner ow induced by the subduction of the Paci c plate, and inhibits hot mantle from owing into the corner region near the trench. Consequently, the thermal recovery of the subducting Paci c plate is slow. The temperature at the interface between the slab and the mantle

4 44 H. Iwamori / Earth and Planetary Science Letters 181 (2000) 41^46 wedge is V370³C at 100 km depth, and V550³C at 200 km depth in the model. By contrast, the calculated temperature for a model without the subducted Philippine sea plate is V490³C at 100 km depth, and V610³C at 200 km depth. The subducted Philippine sea plate in the mantle wedge also reduces the heat ux at the surface. The model predicts that a broad region of relatively low heat ux of 40^50 mw m 32, or less than half of that in the backarc region, should exist in the Kanto region, between the forearc region and the volcanic chain. This prediction is slightly higher than the observed heat ux in the region of V35 mw m 32 [3,16]. If the Philippine sea plate is not xed, heat can be absorbed to the Philippine sea plate more than the current model, which may account for the observed heat ux. 3. Fluid distribution and de ection of volcanic chain Fig. 3. The stability elds of major hydrous phases in the mantle wedge based on [13,17^19], and the P^T paths of the mantle hydrous layer (containing serpentine or phase A as a major hydrous phase) along the subducting Paci c plate (broken lines). The P^T paths are shown for a layer of V7 km thick overlying the plate upper surface. The low temperature gradient shifts the major dehydration reactions along the subducting plate to greater depths when compared to normal subduction zones. The distribution of H 2 O, uid and melt is calculated based on [13], and is shown in Fig. 2a^c, respectively. The calculation incorporates the phase relationships for basaltic and peridotitic systems at high pressures [17^19]. The parameterized phase diagrams are shown in [13] and Fig. 3. In the model, rst, the phase assemblage and the proportions of the phases in a local system is calculated at a given P, T and the H 2 O content, assuming chemical equilibrium in the local system. When the temperature of the local system exceeds the solidus temperature for the given P, T and the H 2 O content, the melt phase is preset. If the H 2 O content in the local system exceeds the maximum H 2 O content which can be contained in the solid (and melt) for the given P and T, the aqueous uid phase is present. After the calculation of the phase assemblage, each phase moves to advect the mass including H 2 O and the energy. The solid ow was described in the previous section. The movement of the aqueous uid phase is assumed to occur as a porous ow without compaction or expansion of the solid matrix after [13]. In general, the fraction of the aqueous uid phase is the order of V0.1 vol% or less (Fig. 2b), as the uid easily segregates. The melt phase is assumed to move with the solid (i.e., no melt segregation), which limits the melt distribution at the depth V80 km where the practical solidus of hydrous (H 2 O-undersaturated) peridotites has a minimum temperature [13]. Although the assumption of no melt segregation is unrealistic, the initiation of melting and the approximate distribution of the melt can be discussed with the current model. If melt segregation is involved in the model by porous or channel ow, the melt will distribute above the layer of V80 km. However, the melt will not be generated in the region below and above the subducted Philippine sea plate (Fig. 2c). The oceanic crust with a thickness of 7 km, which is assumed to contain 6 wt% H 2 O initially, subducts from the upper right corner of Fig. 2. Most of the water subducted with the oceanic crust is released by the time it reaches 100 km depth, where it forms a thin (V5 km) layer of

5 H. Iwamori / Earth and Planetary Science Letters 181 (2000) 41^46 45 serpentinite just above the oceanic crust. In normal subduction zones, breakdown of the serpentinite layer releases nearly all the water subducted at depths shallower than 200 km [13,20]. However, because of the low geothermal gradient beneath central Japan, breakdown of the serpentinite layer can occur at a depth greater than 200 km, and release H 2 O upwards to initiate melting in the far backarc side (Fig. 2c). Within the low temperature oceanic crust at 200^300 km depths, a small amount of H 2 O remains in lawsonite, and is gradually released and contributes to melting. This may produce the broad volcanic chain in central Japan. From the model and discussions above, it is shown that the essential role of the subducted Philippine sea plate on the magmatism is to shift the dehydration reactions to greater depths, rather than preventing melts from rising through. The P^T paths of the mantle hydrous layer along the subducting Paci c plate, between the upper surface of the plate and V7 km above the upper surface, are shown in Fig. 3 by dashed lines. The stability eld of major hydrous phases (serpentine, phase A, amphibole, chlorite and melt) in the peridotitic systems [18^20] are also shown. Even after breakdown of serpentine, phase A can contain V5 wt% H 2 O and can carry H 2 O further down to deep mantle. This hydrous layer containing phase A is visible as a yellowish layer along the deepest part of the slab in Fig. 2a. In this case, V20% of the water subducted is carried to the deep mantle (2.3U10 9 kg yr 31 per 1 km along the arc strike), and will a ect the rheology and phase relationships of the deep mantle. Is a similar process happening anywhere else on Earth? According to the analyses of the geometry of triple junctions [21], central Japan is probably the only place where one plate acts as an obstacle for corner ow induced by subduction of an other plate. However, similar e ects might be expected beneath parts of the Andes, if the crust is thick enough to prevent a hot mantle from owing into the corner region. There are other possibilities for deep subduction of H 2 O. If the lithospheric mantle of a cold subducting plate contains H 2 O, water can be brought down, since the thermal recovery within the plate is slow, and the hydrous minerals such as serpentine and phase A can be stable. Also a small amount of H 2 O can be carried down by nominally anhydrous phases [22], although it could be much smaller than that carried by serpentine and phase A. These possibilities should be tested in future studies. Acknowledgements The author thanks D. McKenzie, J. Maclennan, C. Richardson and D. Zhao for discussion and help, and M. Schmidt and the two anonymous reviewers for review. This work was supported by the Japan Society for the Promotion of Science.[FA] References [1] J.B. Gill, Orogenic Andesites and Plate Tectonics, Springer, Berlin, [2] Y. Tatsumi, S. Eggins, Subduction zone magmatism, Blackwell, Cambridge, [3] K. Yoshii, Compilation of the geophysical data around the Japan arcs (I) (in Japanese), Bull. Earthq. Res. Inst. Univ. Tokyo 54 (1979) 75^117. [4] D. Zhao, A. Hasegawa, H. Kanamori, Deep structure of Japan subduction zone as derived from local, regional and teleseismic events, J. Geophys. Res. 99 (1994) 22313^ [5] T. Seno, S. Maruyama, Paleogeographic reconstruction and origin of the Philippine sea, Tectonophysics 102 (1984) 53^84. [6] K. Okino, Y. Shimakawa, S. Nagaoka, Evolution of the Shikoku basin, J. Geomagn. Geoelectr. 46 (1994) 463. [7] M. Ishida, A.H. Hasemi, Three-dimensional ne velocity structure and hypocentral distribution of earthquakes beneath the Kanto^Tokai district, Japan, J. Geophys. Res. 93 (1988) 2076^2094. [8] M. Ishida, Geometry and relative motion of the Philippine sea plate and Paci c plate beneath the Kanto^Tokai district, Japan, J. Geophys. Res. 97 (1992) 489^513. [9] K. Shimazaki, K. Nakamura, T. Yoshii, Complicated pattern of the seismicity beneath metropolitan area of Japan: proposed explanation by the interactions among the super cial Eurasian plate and the subducted Philippine Sea and Paci c slabs, Mathematical Geophysics, Chateau de Bonas, France, 20^25 June 1982, Terra Cognita 2 (1982) 403. [10] M. Ishida, The spatial distribution of earthquake hypocenters and the three-dimensional velocity structure in the

6 46 H. Iwamori / Earth and Planetary Science Letters 181 (2000) 41^46 Kanto^Tokai district, Japan, J. Phys. Earth 32 (1984) 399^422. [11] K. Kasahara, Patterns of crustal activity associated with the convergence of three plates in the Kanto^Tokai area, central Japan (in Japanese), Rep. Natl. Res. Cent. Disaster Prev. 35 (1985) 33^137. [12] S. Noguchi, Con guration of the Philippine sea plate and seismic activities beneath Ibaraki Prefecture (in Japanese), Earth Mon. 7 (1985) 97^104. [13] H. Iwamori, Transportation of H 2 O and melting in subduction zones, Earth Planet. Sci. Lett. 160 (1998) 65^80. [14] G.K. Batchelor, An introduction to uid dynamics, Cambridge Univ. Press, Cambridge, 1967, 615 pp. [15] D.L. Turcotte, G. Schubert, Geodynamics, John Wiley and Sons, New York, [16] M. Yamano, Terrestrial heat ow and geothermal energy in Asia, M.L. Gupta, M. Yamano (Eds.), Oxford and IBH, New Delhi (1995) 173^201. [17] K. Bose, J. Ganguly, Experimental and theoretical studies of the stabilities of talc, antigorite and phase A at high pressures with applications to subduction processes, Earth Planet. Sci. Lett. 136 (1995) 109^121. [18] T. Kawamoto, J. Holloway, Melting temperature and partial melt chemistry of H 2 O-saturated mantle peridotite to 11 gigapascals, Science 276 (1997) 240^243. [19] M. Schmidt, S. Poli, Experimentally based water budgets for dehydrating slabs and consequences for arc magma generation, Earth Planet. Sci. Lett. 163 (1998) 361^379. [20] H. Iwamori, D. Zhao, Melting and seismic structure beneath the northeast Japan arc, Geophys. Res. Lett. 27 (2000) 425^428. [21] D.P. McKenzie, W.J. Morgan, Evolution of triple junctions, Nature 224 (1969) 125^133. [22] D.R. Bell, G.R. Rossman, Water in Earth's mantle ^ the role of nominally anhydrous minerals, Science 255 (1992) 1391^1397.