Physics of Subduction | Thermal Structure | Arc Magmas | Applications to Convergent Margins

Subduction Zones are the largest ‘geosystem’ on the planet, extending for thousands of kilometers horizontally and hundreds or thousands of kilometers vertically. In spite of the importance, many introductory textbooks present the subduction process poorly. A simplified version of the "textbook paradigm" follows:

The ‘textbook paradigm’ is illustrated in Fig. II-A. This model is attractive to instructors and students because it is plausible and simple, and has been clearly and consistently presented in introductory textbooks for the past two decades. Nevertheless, the textbook paradigm does not reflect with what we know about the forces driving the plates, the strength of the asthenosphere, coupling between subducted plate and the mantle beneath the arc, how the subducted plate is heated, and how arc magmas are generated. The following discussion is intended to help rectify this situation by providing a concise overview of our present understanding of how subduction zones operate.

Physics of Subduction

Davies and Stevenson (1992) carried out numerical experiments that provide the most recent and comprehensive treatment of subduction zone processes, especially interactions between moving plates, mantle, fluids and melts. Fig. II-1 is adapted from their results, and it shows the most important aspects of subduction zone processes. In order to grasp subduction fundamentals, it is necessary to understand the difference between lithosphere (strong and cool; the plate of plate tectonics) and asthenosphere (weak and warm; the substrate over which the plates move) on the one hand, and between crust and mantle lithosphere on the other. This is because the driving force for subduction (and for plate motion in general) resides mostly in the excess density of mantle lithosphere, but the chemical elements that will be processed in the Subduction Factory reside mostly in the crust. Another way to put this is that: to understand subduction's physics, we must know the mantle lithosphere; but, to understand its chemistry, we must know the crust and sediments.

Subduction is driven by the excess density of the lithosphere relative to the underlying asthenosphere, and because the lithosphere thickens as it cools, this excess density increases with age. Mantle lithosphere cools by conduction and thus thickens with age:

This expression indicates that the 'subductability' of oceanic lithosphere varies directly and simply with its age. The thickness of the oceanic crust is always about 6 km and sediment thicknesses may vary from a few hundred to a few thousand meters. The great bulk of the subducted plate consists of mantle lithosphere. Using equation (1), it can be shown that the approximately 100 km thick lithosphere shown in Fig. II-1A is appropriate for 100 million year (Ma) old seafloor. It is essential to note that subduction result from the relatively greater density of the subducted plate relative to the underlying asthenosphere, and the greater the age of the subducted lithosphere, the greater its bulk density. This can be shown by using reasonable densities for the crust, mantle lithosphere, and asthenosphere of rc = 3.0 grams per cubic centimeter (g/cc), rl = 3.3 g/cc, and ra = 3.25 g/cc. 100 Ma old seafloor (crust = 6km, mantle lithosphere = 100km) will have a bulk density of 3.28 g/cc. This is about 1% greater than that of asthenospheric mantle, and we can expect such old plates to sink readily into the mantle if given the opportunity. This is best seen along the convergent margins of the Western Pacific, where seafloor of Jurassic and Cretaceous age (170 to 65 Ma) is subducted. Evidence that this sinking is gravitationally favored - that is, denser lithosphere is exchanging places with more buoyant asthenosphere - can be found in the observation that such arcs have fewer earthquakes, steep subduction zones, and back-arc extension. The Western Pacific arcs subduct old lithosphere and thus are Mariana-type subduction zones.

It should be appreciated that the motion of the plate in a subduction zone is not strictly down dip, that is, parallel to the seismic zone. Hamilton (1988, p. 1507) warns about "...the false assumption that a subducting plate rolls over a hinge and slides down a slot that is fixed in the mantle...". There is also a significant vertical vector of sinking, so that subducting slabs sink more steeply than the dip of the Wadati-Benioff zone, and trenches retreat, or 'roll back'. Perhaps the best proof of this is the fact that the Pacific Ocean basin is becoming smaller — it has to in order to accommodate the widening of the Atlantic Ocean. The Pacific Ocean could not shrink if trench ‘rollback’ did not occur.

In contrast to the situation for old seafloor, young seafloor - because of its lower density - resists subduction. Three (3) Ma old seafloor will have a lithospheric thickness of 23 km (6 km crust and 17 km mantle), and will have a bulk density of 3.22 g/cc, less dense than the underlying asthenosphere. Examples of this can be found in the Eastern Pacific, particularly next to South and Central America, where evidence that young lithosphere is resisting being returned to the mantle includes violent earthquakes, flat subduction zones, and back-arc compression. Figure 1-X shows the relationship between the age of subducted lithosphere and the style of the subduction zone, with "Chilean type" subduction zones involving young, buoyant lithosphere and "Mariana type" subduction zones involving old, dense lithosphere. It is important to point out that whereas the "Chilean type" subduction zone is reminiscent of the textbook paradigm, this is not a stable - that is, normal - subduction zone. Subduction of young lithosphere soon ends by intersection of spreading ridge and trench, as happened in during mid-Tertiary times along the western U.S. This often results in development of a transform margin. In contrast, sliding and sinking of old lithosphere beneath "Mariana type" subduction zones is energetically stable and likely to continue indefinitely.

One important way to define lithosphere and asthenosphere is that lithosphere is the outer portion of the earth that is heated (and cooled) by conduction, whereas heat is transported through the asthenosphere by convection. This definition holds not only while the lithosphere cools and thickens beneath the seafloor but also after it has been subducted. Conduction is a very inefficient way to transmit heat, so that subducted materials are heated to the temperature of the surrounding mantle only after 10's to 100's of millions of years. Figs I-XX and I-XXX show beautifully how the subducted lithosphere remains cool (and seismically fast) relative to the ambient mantle, for tens and even hundreds of millions of years. In the situation shown in Fig. II-1A, the subducted plate has a velocity of 7cm/year - typical for subduction zones - and will reach a depth directly beneath the line of arc volcanoes after only about 2 million years (note:1mm/year = 1km/million years). This is far too little time for the subducted lithosphere to heat to the temperature of the ambient mantle, and thermal models indicate that subducted lithosphere generally does not melt. Melting of subducted crust may occur when very young - and therefore hot - seafloor is subducted (Defant and Drummond, 1990), but this is unusual. The unfortunate tendency of many introductory geology texts to show the subducted crust being melted beneath the line of arc volcanoes is a lamentable misrepresentation of our science's understanding of this fundamental and distinctive earth process.

While the cold, subducted plate is slowly warmed by the ambient mantle, the ambient mantle is also cooled. The result of all this is that subduction zones involving an old descending plate are by far the coldest parts of earth's interior. Support for this statement is found in the great depth of seismicity in subduction zones. In all other parts of our planet, earthquakes are restricted to the outer, cold, and therefore brittle layers. In most tectonic settings, earthquakes are found to depths of no more than 10 km, but in subduction zones earthquakes to depths of almost 700km. This is somewhat puzzling, because subduction zones are associated with volcanoes — the subduction zones around the Pacific are parallel to the ‘Pacific Ring of Fire’. The key to understanding this paradox is found in Fig. II-1. The only thing that is set in motion in this model is the subducting plate, but this drags the asthenosphere below the overriding plate with it. The asthenosphere beneath the overriding plate immediately adjacent to the descending plate is dragged with the plate deeper into the earth. Sinking material must be replaced by asthenospheric mantle from elsewhere. This results in induced convection beneath the arc, so named because it is caused by motion of the subducting plate. As a consequence of induced convection, the asthenosphere beneath the overriding plate is continuously replenished. Thus the temperature of the asthenosphere beneath the arc volcanoes is not affected much by the presence of the cold subducted lithosphere, but remains at a temperature found for asthenosphere at simlar depths elsewhere inside the earth. The continuous replenishment of normal asthenospheric mantle beneath the arc is an essential part of understanding why the coldest parts of our planet are associated with lots of volcanoes.

Generation of Arc Magmas

If the subducted plate does not generally melt to form arc lavas, what does? The key can be gleaned from a simplified phase diagram for peridotite (Fig. II-4). The mantle, both asthenospheric and lithospheric, is composed of peridotite. As shown in Fig. II-1B, the maximum temperature directly beneath arc volcanoes lies at a depth of about 80 km, or about 25 kbar. At this pressure, wet peridotite begins to melt at a much lower temperature - as much as 400°C lower - than dry peridotite. We know that arc magmas are much richer in water than magmas from any other tectonic setting, and that this water is ultimately derived from the subducted plate. If water can be delivered to the region of maximum temperature beneath arc volcanoes (region labelled 'Likely Region of Melt Generation' in Fig. II-1B), then melting is likely to occur. However, it is not obvious how water can be delivered from the subducting plate to this region. Inspection of Fig. II-1A shows that the 1200°C isotherm is at its closest about 30 km from the subducted plate. Note that both the subducted plate and the adjacent mantle are moving downward. The question of how arc magmas are generated depends on understanding how water gets from the subducted plate to the likely region of melt generation.

This part of the story is complex, as revealed by inspection of Fig. II-5, which shows detail in the mantle wedge (area X-Y-Z of Fig. II-1A). Subducted sediments and oceanic crust may carry 5% and 1.5% water, respectively, into the subduction zone. This water is continuously released. This release is initially due to squeezing shut of cracks and pore spaces, but as subduction proceeds mineral reactions become more important in controlling water release. Amphiboles are thought to be especially important in carrying water to great depths, as these are stable to depths of about 100km (Fig. II-4). At this depth, amphiboles in the crust breakdown and the oceanic crust is reconstituted into much denser eclogite, composed of pyroxene and garnet.

To reach the zone of melt generation, fluids derived from the subducted slab have to ‘tack’ across the downwelling mantle, like a sailboat sailing against the wind. Fig. II-5 shows water being released continuously from the subducted crust, and this rises up into the overlying mantle where it may form amphibole peridotite. This mantle is dragged downwards with the subducted slab. It seems that amphibolite - metamorphosed basalt - is stable to only about 80 km deep. At this depth (B on Fig. II-5) all of the amphibolite breaks down to form eclogite, releasing a large amount of water up into the mantle and forming abundant amphibole peridotite (C). This is carried down to the maximum depth of stability for amphibole peridotite, about 100 km, where it breaks down to anhydrous peridotite and water (D). The water released rises vertically; note that this carries the water away from the subducted slab and towards the likely region of melting. At some point the rising water will react with the mantle to form amphibole peridotite again (E). This will be carried down with the descending mantle until the amphibole breaks down again (F). Amphibole peridotite forms again (G) and breaks down again (H). In this zig-zag fashion the water is carried away form the subducted slab and into warmer parts of the mantle wedge.

The pink area inside the solid line in Fig. II-5 is the area where the mantle will partially melt if sufficient water is provided. Above point H, the mantle is sufficiently hot that water added to it leads to melting (I). The mantle is still moving downward, so melt will not be able to rise until enough of it accumulates to form diapirs (K). Because these contain perhaps 10% melt, they are much less dense than the surrounding mantle and can rise through it, at a rate of perhaps 1m/year (L). If they rise sufficiently fast, diapirs will change temperature only slightly, and the decrease in pressure they experience during ascent may lead to further melting, so the diapir may be 30% melted when it arrives at the base of the lithosphere beneath the arc. At this point the magma separates from the unmelted part of the diapir and an arc volcano.

It is very useful to apply the outline of how subduction zones operate presented above to a real subduction zone. Recall that there are two subduction zone end-members which are defined by the age of the subducting lithosphere (Fig. I-X). The Chilean-type is inherently unstable, not only because subduction of progressively lithosphere becomes increasingly difficult, but because at some point the ridge and trench will met and the convergent margin will become a transform margin, as was the case for western North America during mid-Tertiary time. In contrast, Mariana-type subduction zones are stable because denser lithosphere is sinking beneath less dense asthenosphere, and in this sense are the more 'normal' type of subduction zone.

Figure II-6 presents the Mariana convergent margin in true scale, with all of the important elements of the subduction zone model presented here. This is a tectonic cartoon in many respects. For example, while we know the thickness of the crust, we know very little about the thickness of the lithosphere. The very thick lithosphere beneath the forearc is consistent with observed low heat flow, and the nearly vertical boundary between asthenosphere and lithosphere beneath the arc is consistent with thermal models as well as the requirement that melting occurs near the interface between water from the subducted crust and convecting asthenosphere. Similarly, we know that seismicity defines a subduction zone geometry like that shown here, and that seismicity ends at about 700km. The different behavior of the crust and mantle lithosphere on the subducting plate reflects the fact that the crust is a chemical entity while the mantle lithosphere reflects thermal considerations. As these crust and mantle lithosphere heat up as subduction proceeds, the latter will thin whereas the former will not. Not also that there is a back-arc basin spreading center behind the line of arc volcanoes. Note that the magmatic systems associated with the back-arc basin spreading center and arc volcanoes are associated with different mantle flow regimes. Arc diapirs ascend through down-welling mantle, while back-arc basin diapirs ascend with upwelling mantle. The scientific thrust of the Melville cruise concerns understanding whether or not melts are generated twice in this part of the Subduction Factory — once beneath the back-arc basin and once beneath the arc.

References

Click here