Subduction: an overview
As we have discussed in the article on sea floor spreading, rock is generated at mid-ocean rifts; when it is produced in this way it is hot and therefore buoyant. As rock spreads out from the rifts, it cools down, and also thickens as material from the upper mantle accumulates on the underside of the cooling crust, both of which processes make it denser. Eventually, as a result of this, it becomes gravitationally unstable and liable to plunge into the athenosphere.
It cannot simply sink straight down like a foundering ship, for two reasons. First, it is still attached to the more buoyant portion of the plate. Second, in order to do so it would have to displace the material of the athenosphere in a way that is physically implausible. To grasp this point, consider a sheet of paper floating on the surface of a bowl of water. The paper is more dense than the water and should sink, but it cannot sink straight down in its horizontal attitude without pushing up the water surrounding it. However, if one edge of the paper is forced beneath the water, then the sheet of paper can descend down and sideways, sliding edge-on through the water.
And this is what happens to oceanic crust when it subducts; forced beneath another more buoyant plate, it slides edge-first into the athenosphere. The portion of the plate being thrust down in this way is called the slab. A typical angle of descent is 45 degrees, but the slab can descend at any angle from near-horizontal to near-vertical.
The diagram to the right shows a cross-section of the process. (The reader should note that the diagram is somewhat schematic in nature and not entirely to scale.)
In the diagram, oceanic crust (on the left) is being subducted under continental crust (on the right). The triangular gap formed where it bends to go under the continental crust is known as a trench. For reasons that we shall discuss below, the subducted slab undergoes partial melting at a depth of about 120 km, producing volcanoes above the melting.
In the diagram I have chosen to show oceanic crust being subducted beneath continental crust. An oceanic plate can equally well subduct beneath another oceanic plate with much the same results, only forming a chain of volcanic islands and seamounts instead of a volcanic mountain range.
In the diagram I have shown the trench filled up with sediment, forming what is known as an accretionary wedge or accretionary prism. However, not all trenches have such a feature: this depends on the rate at which sediment is transported into the trench and the rate at which it is subducted along with the plate. As a general rule, an accretionary wedge will be present when the trench is close to a continent, as this ensures a greater supply of sediment than can be found further out to sea.
Evidence for subduction
In this section, we shall look at the evidence for subduction.
The existence of sea-floor spreading
This raises the obvious question of where it goes to. The earth's surface cannot simply inflate like a balloon; and looking at the sea floor, we can see that neither does it ruck up like a carpet too big for the room it's laid in. The only possible explanation is that crust is being destroyed, and the only reasonable way that this can happen is for the crust to be recycled back into the mantle. So the evidence for sea floor spreading is on the face of it evidence that subduction must happen: or at the very least that something must be destroying oceanic crust.
Age of the sea floor
One consequence of the production of the sea floor at rifts and destruction of sea-floor at subduction zones is that the sea floor should consist of rocks younger than the Earth itself. And this is what we find: the oldest rocks on the ocean floor date to about 200 million years old, as compared to 4 billion or so years for the oldest continental rocks.
As with the previous point, this is good evidence that the oceanic crust is being destroyed by some mechanism. We can now turn to the evidence that the mechanism is indeed subduction.
If a slab of cold, brittle, elastic rock is really being subducted into the athenosphere, then we should expect it to undergo earthquakes as large bodies of rock in this state always do when placed under stress.
Because it is possible to measure the depth of an earthquake (see here for details) as well as its latitude and longitude, observations of earthquakes allow us to form a good picture of what happens when a slab descends into the athenosphere. The diagram to the right, for example, shows the locations of earthquakes of magnitude 4.5 or greater occurring between 1973 and 2004 where the Indo-Australian plate is subducted under the Lesser Sunda Islands. The profile of the slab is clearly traced out by the incidence of earthquakes. (The US Geological Survey's Earthquake Hazards Program has many such maps of subduction zones.)
We should note that without subduction there shouldn't be earthquakes below the lithosphere at all. In rocks below the lithosphere the temperature and pressure are such that rocks under stress at that depth should bend, stretch and flow, producing no earthquakes. The only plausible explanation for quakes at these depths is that colder more brittle rock from the surface has been thrust down into the athenosphere. (In considering this, the reader should bear in mind that the slab is several kilometers thick and that heat flows very slowly through rocks, so there is no difficulty in a subducted slab remaining brittle and relatively cold for millions of years after it enters the athenosphere.)
Volcanoes and subduction melting
Water lowers the melting point of rock in a manner similar to how salt lowers the melting point of ice; the result is that when the wet rocks of the crust reach the right zone of temperature and pressure, about 120 km below the surface, the rocks of the crust undergo partial melting. (The fact that this should occur at 120 km depths can be verified in the laboratory by simulating the appropriate temperature and pressure.)
Naturally the hot and relatively less dense molten rock rises through the athenosphere and escapes to the surface as lava, forming volcanic mountain chains if the slab is subducting under continental crust, or chains of volcanic islands and seamounts if the slab is being subducted under a oceanic crust. It is estimated that up to 20% of subducted rocks are returned to the crust by this mechanism.
Because we can use seismic readings to measure the angle of the slab, and because we know that partial melting occurs at a depth of about 120 km, we can use simple geometry to calculate where, relative to the subduction zone, the volcanic arcs should occur. The fact that they always do occur in the right place supports the idea that geologists really do know what's going on in the athenosphere.
The existence of trenches were once a mystery to geologists. But the mechanism of subduction makes everything clear: there must be a trench, roughly triangular in cross-section, between a descending slab and the plate under which it is subducted.
If this is the true explanation, then we should expect to find these trenches (and under the right circumstances accretionary wedges) in geographical association with the earthquakes and volcanism which we also attribute to subduction; and this is in fact what we find.
The Earth looks as it should if subduction really is taking place; what is more, there are phenomena such as earthquakes in the athenosphere, or the presence of chains of volcanoes at just the right distance from subduction zones, which no-one can explain on any other hypothesis. It is therefore reasonable to conclude that subduction really is taking place.