Does subduction quit every now and then?

Callan Bentley at NOVA Geoblog has a good discussion of an article published in Science by Paul Silver and Mark Behn about the possibility that plate tectonics has been intermittent, rather than a continuous feature of Earth’s geology. I was intrigued by the press releases, but the paper itself seems to have problems, and I’m curious whether I’m simply being dense, or whether these problems really exist.

The basic premise of the Silver and Behn paper is this: when continents collide, a subduction zone goes out of existence. Cartoons of pre-Mesozoic* collisions (such as the Appalachian collisions that Callan shows on his blog) usually show subduction initiating someplace in a nearby ocean, to balance sea-floor spreading somewhere else on the planet. But what if that doesn’t happen? And what if there aren’t major subduction zones on the opposite sides of the colliding continents to take up the motion? Is it possible that world-wide plate motions could simply stop, until enough heat built up in the mantle to force continents to rift (through initiation of mantle plumes, if they really exist)?

It’s an intriguing idea. It’s true that subduction zones don’t seem to regularly start in places where oceanic crust is particularly old, and that collisions don’t always result in the formation of new subduction zones. There are only two dinky little subduction zones in the Atlantic (the Caribbean and Scotia arcs), and there is no new subduction zone south of India. On the other hand, there is a new subduction zone forming on the north side of Papua New Guinea, after the northern continental shelf of Australia clogged the subduction zone. And the idea that supercontinents could insulate the mantle beneath them, leading to increased heat flow, could explain one of the mysterious features of the Precambrian basement in my neck of the woods: the ~1.4 billion-year-old “anorogenic” granites, and the pervasive high temperature/low pressure metamorphism that can be found even far from the contact of any exposed granite.

But the proposal implies a particular sequence of events... and the data used to support the model suggests that things actually happened in a different order.

The order ought to be this:

1) A Pacific-type ocean basin (in which the ocean basin existed before breakup of the previous supercontinent) closes, creating a supercontinent.

2) The majority of worldwide subduction zones are caught in the middle of these collisions, and disappear.

3) The lack of subduction causes mid-ocean ridges to slow, as well, and plate motion decreases (or, possibly, stops).

4) Heat builds up in the mantle until plate tectonics resumes (presumably driven by the rising of hot material and some kind of rifting; subduction would have to begin at the same time to keep the Earth the same size).

Silver and Behn argue that Pangea shouldn’t fit this model, because when the Caledonian/Appalachian/Variscan/Ural collisions occurred to form Pangea, the oceans that closed were young. (And there were many active subduction zones around the edges of Pangea, off the west coasts of present-day North and South America, among other places.) Instead, they focus on two older supercontinents: Rodinia, which formed as a result of the Grenville orogeny around a billion years ago, and Pannotia/Gondwana, which formed slightly later.

Their primary way of testing the model is to look at the geochemical evolution of the mantle as a way of measuring the amount of subduction that occurs. They use two different measures of mantle chemistry: the Nb/Th ratio (which should increase in mantle-derived rocks through time, because Th is preferentially incorporated into subduction zone melts), and the ratio between two isotopes of He.**

The test is summarized in their figure 1, which I can’t show because it’s behind a paywall. So I’ll draw some sketches to show why I’m confused. The approximate shape of their estimated subduction flux looks like this:


There are two peaks, one in the late Archean (around 2.5 billion years ago) and one in the Phanerozoic. And there’s one minimum, around 1 billion years ago.

If you add the times during which Rodinia and Pannotia/Gondwana formed by closure of Pacific-type basins, the diagram looks approximately like this:


Do you see why I’m confused? Yes, there appears that global subduction decreased (if the proxies are correct) between 2.5 and 1.0 billion years ago. But the minimum subduction rate occurred just before Rodinia formed – exactly the opposite order than one should expect if the formation of Rodinia caused subduction zones to shut down. In fact, the formation of Rodinia occurred just before the worldwide subduction rate began to increase, as if the formation of supercontinents actually caused new subduction zones to form in other parts of the world.

There is another, even older supercontinent (Nuna) proposed, but it isn’t clear exactly what kind of ocean basin closed to form it. And it formed during the middle of the decline in subduction rates.


In fact, the only supercontinent whose formation correlates to a decrease in the proxies for subduction rate is Pangea – and we know that subduction occurred before, during, and after Pangea existed. Pangea was the one supercontinent they excluded from consideration.

So... interesting idea. But their measures of subduction flux don’t seem to support it.

I have no trouble with the idea that plate tectonic rates may have varied through time (though I would love to see, say, a compilation of paleomag data as an independent test of the geochemical proxies). But their explanation – that the formation of supercontinents causes subduction to halt – doesn’t fit the data.

If this study is correct, it would have interesting implications for things like the driving mechanism of plate tectonics. But the data doesn’t seem to fit the conclusions. Am I missing something critical about their reasoning – some reason why the decline in subduction should precede the collisions?

I’m not teaching plate tectonics again until 2010, so there’s time for comments, replies, and discussions before I would discuss this in class.

Reference: Silver, Paul G., and Behn, Mark D., 2008, Intermittent plate tectonics?: Science, v. 319, p. 85-88, doi: 10.1126/science.1148397.

*Plate movements from the Jurassic to the present are much better constrained than pre-Mesozoic motions, because magnetic lineations in the ocean floor provide a quite detailed record. So the changes in plate motions after the Himalayan collision can be examined worldwide.

**I don’t know how quickly mantle chemistry should respond to the amount of subduction – that gets into questions about how quickly the mantle mixes, and I can’t even say how readily the mantle mixes vertically, let alone how easy it is for an aesthenospheric wedge above a long-lived subduction zone to mix with the wider mantle. But we’re talking about lots of time in the Precambrian, so maybe that isn’t really an issue.