GEOPHYSICAL RESEARCH LETTERS, VOL. 27, NO. 3, PAGES 425-428, FEBRUARY 1, 2000 Melting and seismic structure beneath the northeast Japan arc Hikaru Iwamori Geological Institute, University of Tokyo, Tokyo, Japan Dapeng Zhao Department of Earth Sciences, Ehime University, Matsuyama, Japan Abstract. Transportation of H20 associatedwith subduc- of the area [Zhao et al., 1992] (Fig.2). The NE Japan arc is tion of the Pacific plate beneath the northeast Japan arc one of the best test field of the model: (1) a denseseismic is modeled to predict distribution of aqueous solution and network with active seismicity have revealed the geometry melt, and the consequentP-wave velocity structure. The and velocity structures of the subducting Pacific plate, and observed velocity structure coincides well with the model the overlyingmantle wedge-crustsystem (e.g., [Zhao et al., for equilibrium transport of H20: most of H20 subductedis 1992, 1994]), (2) the geometricalrelationship between the brought down to 150-200 km depth along the slab, then the plate motion, trench and arc is simple, i.e., the Pacific plate aqueous fluid generated ascendsto initiate melting, which subducts nearly perpendicular to the Japan trench and the explainsthe low velocity regions(-•6% reduction) observed activevolcanic front (Fig.2), allowingus to apply the across- beneath the backarc, rather than beneath the volcanic front. arc 2-D modeling. Based on the comparison, we propose a If equilibrium transport of H20 occurs, initiation of melting model which accounts for the observations, including the beneath the backarcwith deep subductionof H20 is likely to regions with 6% reduction in P-wave velocity beneath the be the case also for other subduction zones with slabs older backarc, and discussthe fluid processesin subduction zones. than several tens of m.y., cold enoughto stabilize serpentine to a great depth. Model Predictions Introduction Transportation and distribution of H20 beneath the NE Transportation of H20 in subduction zones is key to Japan arc have been calculated based on 2-D numerical understanding the origin of arc magmatism. A series of modelingof Iwamori [1998] (Fig.la and b). In the model, fluid processesare thought to occur associated with sub- the subducting Pacific plate is assumed to drive the clock- duction of a H20-bearing plate: generation and migration wise corner flow of the solid in the mantle wedge by dragging of aqueousfluid, and its reaction with the convecting man- (dotted linesin Fig.la and b). A standardtemperature pro- tle wedgeabove the slab, involvingmelting (e.g., [Tatsumi, files for subduction zones with the potential temperature of 1989; Peacock,1990; Davies and Stevenson,1992; Iwamori, the mantle wedgeof 1250 øC (similar to those in [Davies 1998; Schmidtand Poli, 1998]). Theseprocesses need to be and Stevenson,1992; Furukawa, 1993b; Iwamori, 1997]) is consideredsimultaneously for understanding the behavior of set on the backarc side boundaries, which advects into the H20 and magmatism in subduction zones. region of interest by the corner flow. The thermal structure along the slab, and the style of The subducting oceanic crust of 7 km thick, which is fluid transportation have been shown to be the controlling assumed to contain initially 6 wt.% H20 in chlorite, law- factorsof the processesabove [Tatsumi et al., 1983;Peacock, sonite and amphibole, subducts from the upper-right corner 1990; Davies and Stevenson,1992; Peacocket al., 1994; Bose of Fig.1. It undergoes dehydration, and produces an aque- and Ganguly,1995; Iwamori, 1998;Schmidt and Poli, 1998]. ous solution of total 3 wt.% at 50 km and 5 wt.% at 100 km. Recent knowledge concerningthe phase relationships of hy- This significant amount of the aqueous solution is likely to drous peridotitic and basaltic systems(e.g., [Inoue, 1994; migrate upwards due to buoyancy as either porous or chan- Poll and Schmidt, 1995; Hirose, 1997; Kawamoto and Hol- nel flow through cracks. Intermediate-depth earthquakes, loway, 1997; Schmidt and Poli, 1998]) allowsus to model which occur at depths of greater than 50 km within the up- the fluid generation and transportation in subduction zones per part of the slab beneathNE Japan [Zhao et al., 1992], simultaneously[Iwamori, 1998], which predicts the distribu- are thought to be the consequenceof the dehydration and tion of fluids (both aqueoussolution and melt). migration of such a fluid which is supplied to the mantle In this paper, in order to test the predictionsand discuss wedge[Kirby, 1995; Davies,1999]. the fluid processes,we apply the model to estimate the fluid Two types of migration of the aqueous solution are as- distributionsbeneath the northeast (NE) Japan arc, and sumed in the model. One is porous flow with chemical equili- calculatethe consequentseismic velocity structures(Fig.l), bration during migration in Figs.la and c (hereafterreferred which are then compared with the 3-D tomographic images to as 'equilibriummodel'). The other is that, in Figs.lb and d, in addition to equilibrium porous flow, a part of the aque- Copyright2000 by the AmericanGeophysical Union. ous solution present is assumed to migrate as channel flow Papernumber 1999GL010917. such as through dykes, which leads to partial chemical dise- 0094-8276/00/1999GL010917505.00 quilibrium (hereafterreferred to as 'disequilibriummodel'). 425 426 IWAMORI AND ZHAO: MELTING AND SEISMIC STRUCTURE BENEATH NORTHEAST JAPAN Equi ! ibr ium model Disequi t ibrium model perature and fluid fraction on the seismicvelocity, based on the experimentsand thermodynamictheory [Murase and Kushiro,1977; Duffy and Anderson,1989; $ato et al., 1989]' o e.g., at 3 GPa and around 1000 øC, a temperature increase by 100 øC reducesthe velocityby •0.6 %, whereas,although 200 0.0 the large uncertainty exists due to scale and textural vari- ations of partial melts, the i % increase of melt fraction - 1 04 . reducesthe velocitysignificantly (by 2 to 3 % [$ato et al., 1989]). As has been discussed,the local melt fraction is thought to be less than a couple of per cent, due to effec- 0.96 200 C tive melt segregation. Here we assume the maximum local melt fraction of 2 vol.% above which the melt segregation Figure 1. Distributionof H20 (a andb) andthe correspond- takes place. The aqueoussolution generatedat depths of in- ing P-wave velocity (c and d) beneath the NE Japan arc for the terest contains a significant amount of silicate components equilibrium (a, c) and disequilibrium(b, d) models. Tempera- [Mibe et al., 1997],and its physicalproperties are expected ture and flows of the solid and the fluids have been calculated to be similar to the melt. Therefore, the same velocity-fluid after Iwamori [1998] in a cross-sectionalarea to 200 km depth fraction relationship with the melt is used for the aqueous with a fixed crust of 30 km thick and the subducting Pacific plate solution. In any casethe fraction of aqueoussolution is gen- (subductionangle 30 ø, velocity2 x 10-9 m s-1, age 130 m.y.). The dotted lines in the mantle wedge indicate the stream lines erally _•0.1% and has less effectson the velocity than the of solid, and the solid lines indicate the temperature contour of melt. An influenceof variable proportions of mineral phases 200øC interval. P-wave velocity structures have been calculated is also small: e.g., increaseof 5 vol.% of garnet increasesthe as is described in the text. The vertical line intersects the slab- P-wave velocity by 0.5%. For simplicity, in the model a con- mantle interface at 100 km depth, and nearly correspondsto the stant proportion of olivine: orthopyroxene: clinopyroxene: location of volcanic front. garnet (60:20:15:5)is assumedfor subsolidus. The results are shown in Figs.lc and d with P-wave ve- locity (Vp) normalizedto a standardvelocity at eachdepth When the aqueoussolution enters the mantle wedge, if observedbeneath the NE Japanarc [Zhao et al., 1992],the equilibrium porous flow occurs, a thin layer of serpentinite same normalization as in Fig.2. The cold Pacific plate shows (•7 kin) is formedto absorball H20 released,since it can the greatest Vp value, whereas the mantle wedge has lower contain •8 wt.% H20 [Schmidtand Poli, 1998]. Nearly Vp valuesdue to both the higher temperature and the pres- all the H20 is brought down to 150-200 km beneath the ence of fluids. The temperature variation within the mantle backarc,due to low temperatures along the slab. Chlorite is wedge can account for at most 4 % reduction of the P-wave also stable up to 150-200 km. This thin layer with abundant velocity at depths between 50 and 75km, whereas the pres- ence of fluids of 2 vol.% can account for 6 % reduction in H•O is visible in Fig.la as a 'red' layer along the subducting slab. Vp. From 150 to 200 km depths, serpentineand chlorite break The difference in the style of fluid transport leads to down to form a vertical column through which H•O is trans- a remarkable difference in the predicted seismic structure ported upwards by an aqueoussolution. When the aqueous (Figs.lc and d). In the equilibriummodel, low-velocityre- solution reaches a depth of •-80 km, the temperature of the system exceedsthe practical solidus temperature to cause o significantmelting. The P-T condition is closeto the cuspof the H20-undersaturatedsolidus around 2.5 GPa [Iwamori, 1998]. In this equilibriummodel, therefore,the arc mag- lOO matism is initiated in the backarc region, rather than di- rectly beneath the volcanic front as has been proposed in the previousstudies [Tatsumi, 1989; Davies and Stevenson, 200 1992; Schmidtand Poli, 1998]. The melt will be transported and focused towards the volcanic front either as a porous flow or through cracksdue to the stressgradient [Spiegel- 100 man and McKenzie, 1987; Davies and Stevenson,1992; Fu- rukawa, 1993a], which is likely to suppressthe maximum 200 melt fraction to be less than a couple of per cent.