Melting in the mantle wedge

Compared with the relatively simple spectacle of MORB magmatism, the melting drama in arc tectonic settings is highly complicated. Partial melting, melt segregation and uplift in the mantle wedges are controlled by a number of different processes and parameters: the fluid and melt transfer from the subducting slab (in turn depending on slab composition, age, subduction velocity and angle); the composition of a given wedge and its dynamics and thermal structure; the thickness and composition of the overlying continental plate; the crystallization differentiation in crustal magma chambers. Because of these complexities a standard model of melt generation beneath a "standard arc" has not yet been put forward. The great variation in the data probably reflects the fact that processes are affected by many parameters that are independent of each other. Sets of observations and related models are reviewed by Ulmer (2001), Turner et al. (2003), Schmidt and Poli (2003) and Kelemen et al. (2003). Below we recall some principal observations (mainly already discussed in this chapter) and qualitatively outline major points of a scenario for the generation of arc magmas that take these observations into account. Radioactive isotope tracers are particularly important in this.

Three different processes are generally recorded by arc magmas: (1) the transfer of a mobile phase that originated in the slab under high-PT conditions; (2) the addition of "normal" fluids, derived from the slab at moderate PT, to an arc-melt source within the mantle wedge; (3) the partial melting of this metasomatized hybrid source. We now discuss these in turn.

(1) In arc magmas, Th and 10Be are the best tracers of high-temperature fluids from sediments (Plank and Langmuir, 1993). Also, Ce anomalies inherited from the sedimentary environment indicate the addition of slab melts (or supercritical fluids, Kessel et al., 2005) into the arc magma source (Section 25.6). In the course of slab dehydration and melting processes, the U/Th ratio is expected to be fractionated, causing a deviation of the (230Th/234U) activity ratios from secular equilibrium. However, in contrast with this expectation, the values of these activity ratios in arc volcanics are commonly ~ 0.95, deviating only slightly from equilibrium. The time scale for 230Th equilibration is relatively long, ~ 300 kyr (Elliott et al., 1997; Turner et al., 2003).

(2) The addition of "normal" slab-derived fluids to arc magma sources is clearly shown by e.g. the correlations seen in Figs. 25.3-25.5. The injection of a slab-derived fluid with a high 238U/232Th ratio into an arc magma source and the subsequent ingrowth of 230Th should generate an inclined array in the 230Th / 232Th versus 238U / 232Th isotope plot, with a slope < 1. This is indeed observed in several arcs. The chronological interpretation of such arrays gives ingrowth time scales of ~ 30 kyr or longer (Elliott et al., 1997; Turner et al., 2003). However, the departure of the (226Ra / 230Th) ratio from equilibrium (Fig. 25.5) dictates much shorter time intervals, less than 10 kyr, between the separation of a fluid (the carrier of hydrophilic 226Ra and 238U) from the downgoing slab and the eruption of arc lavas. A way to resolve this contradiction between robust data is to propose that the subduction-related modification of the mantle wedge, adding slab-derived elements such as Th, Sr and Be and various intermediate decay products of the U chains, happened in diachronous stages (Morris and Rayan, 2003).

(3) The value of (231Pa/235U) > 1 observed in arc volcanics also departs from equilibrium but cannot be explained by fluid addition to an arc magma source: Pa is not a hydrophilic element and the addition ofU would have yielded (231Pa/235U) < 1 instead. Protactinium is an extremely incompatible element, with D(solid/melt) ~10-5. Therefore an increase in the activity ratio could result from partial melting and melt-solid fractionation. This indicates a small melt fraction, ~ 0.001, during arc magma generation as U is also a highly incompatible element (Bourdon et al., 2003; Turner et al., 2003).

Traditionally, arc melting is considered to result chiefly from the lowering of the solidus of rocks in the mantle wedge by the introduction of hydrous fluids. This would not be readily compatible, however, with the extremely small partialmelt fraction reflected by (3) above. A scenario envisaging (i) slab melting and melt migration into a slowly upwelling hot peridotitic diapir, within the mantle wedge at some distance above the subducting slab, (ii) further metasomatic development of the diapir with slab-derived fluids and (iii) its decompression fractional melting in the upper region of the wedge has therefore been suggested (see Conder et al., 2002; Kohut et al., 2006, and references therein). As the rocks in a diapir approach the depth of slab melting (— 150 to 200 km), enriched hydrous melts or/and supercritical liquids from the slab metasomatize rock units in the wedge, thus increasing [Th] and fractionating the U/Th and (230Th/238U) ratios. As ascent proceeds, 230Th approaches secular equilibrium (Elliott et al., 1997). The rate of ascent estimated from the equilibration time scale is of order — 10 cm yr-1.

At somewhat lower depths, slab-derived fluids may interact with the diapir, yielding an enrichment in fluid-sensitive elements. The observed chemistry of arc rocks belonging to the calc-alkaline to tholeiitic compositions requires a small amount of water in the mantle source, of order 0.1%-0.5% (Ulmer, 2001). The arc lava O-isotope inventory also suggests a low bulk contribution of slab derivatives (melts, fluids), generally < 1% (Eiler et al., 2000). To reconcile the time scales dictated by 230Th and 226Ra disequilibrium, two or more subsequent fluid-injection events are postulated (Bourdon et al., 2003).

As ascending material reaches the solidus, the partial melting of metasomatized peridotite starts. Initially the melt fraction remains small (and increases slowly) because of the postulated slow uplift of the diapir and the higher viscosity of hybrid melts compared with MORBs; the small melt fractions lead to preferential partitioning of Pa into the melt compared with U, and the time scale for 231Pa equilibration is a few tens of kyr (Bourdon et al., 2003). Through mixing, the aggregate melt fraction could become quite large, approaching the value — 10% generally accepted from major- and moderately-incompatible-element data.

The data indicate a close temporal and spatial link between the final addition(s) of 226Ra-bearing fluid(s) and the segregation and removal of magmas with fast ascent velocities, —10 to 100 m yr-1 (Turner et al., 2000, 2003).

Generally melting and melt segregation proceeds at depth and temperature intervals within 120 to 30 km and 1450 to 1200 °C, respectively (Conder et al., 2002; Kohut et al., 2006). For example, in the case of the Japan arcs the seismological data indicate melt generation at 90 km depth (or somewhat below this depth if initially small melt fractions are postulated), efficient melt segregation and transfer through dykes and cracks at 65 km depth and melt accumulation below the base of the crust (Nakajima et al., 2005).

0 0

Post a comment