This transcript was generated automatically and lightly edited for readability.

Oliver Strimpel: This is Geology Bites with Oliver Strimpel. In an active subduction zone, the downgoing slab of oceanic lithosphere is denser than the mantle that surrounds it. This generates the so-called slab pull force that is thought to be the predominant force that keeps plate tectonics going. But how does the plate start sinking in the first place? This question has puzzled researchers for a long time. David Bercovici is one of those puzzled researchers, but his work on the physics of mineral grains within an ocean plate now promises to provide some answers. He is professor of Earth and planetary science at Yale University. David Bercovici, welcome to Geology Bites.

David Bercovici: Thanks very much, Oliver. Thanks for having me.

Oliver Strimpel: Why is subduction initiation such a thorny problem?

David Bercovici: Well, it's one of those classic problems in trying to understand why Earth has plate tectonics at all. In general, as the lithosphere moves away from a mid-ocean ridge, and especially when we look at something like the Atlantic Ocean, which doesn't have any subduction zones yet, we're interested in how something like the Atlantic, an opening ocean basin, can form a subduction zone and eventually, maybe one day, look more like the Pacific or the Indian Ocean. So one of the general problems is that as that lithosphere thickens and cools away from a mid-ocean ridge, it makes the plate stronger. And eventually, in principle, that makes it heavier also, because it's getting colder and thicker, which would make it want to subduct. But the fact that it's getting stronger keeps it from subducting. And in fact, the effect of what we call thermal stiffening on the strength of the plate makes it such that it really shouldn't subduct at all. I mean, in principle Earth shouldn't have subduction zones, and we tend to see other planets, with Venus arguably being one, behaving that way, that don't have any distinct subduction zones. So Earth is, in a way, an enigma. So the big question is, how do we undo that thermal stiffening? How do we undo the effect of cooling as a plate thickens and gets stronger? And so that becomes a major problem: how to get a plate to subduct.

Oliver Strimpel: What theories have people come up with so far to explain how subduction starts?

David Bercovici: All of them, in the end, have to invoke some way of undoing the cooling and stiffening of the plate. One that has some field evidence is that an existing transform fault that's already in the oceanic lithosphere provides a weak zone around which a subduction zone can nucleate. And that seems to be something that's seen in the evolution of plates as they change direction.

Oliver Strimpel: But isn't that just kicking the can down the road? I mean, don't we then have the problem of how a transform fault forms in the first place?

David Bercovici: Yeah, I mean, absolutely true. You know, I guess in some sense, it's probably easier to make a transform fault, especially if it starts off as an offset at a mid-ocean ridge, where the lithosphere is thinner and warmer and weaker. But just getting an offset like this, where you get this weird curvature in a mid-ocean ridge, is in itself a big problem. I mean, it's another very interesting problem of why plate tectonics behaves the way it does, and it is not at all well understood.

Oliver Strimpel: So how have you approached the problem?

David Bercovici: Well, we're trying to understand the physics that would lead to how plates form subduction zones and transform faults and other features of plate tectonics. And it really comes down to understanding more about the physics of how solids flow. The mantle is a solid; it's not a liquid, and yet it flows like a fluid. You know, like ice flows; even glass flows as a fluid. But still, ice and rocks have a crystallographic structure, and they wind up deforming slowly over time through the migration of atoms under imposed stresses. So if you're squeezing something from, say, the left and the right, or pulling it from top and bottom, the atoms are going to effectively want to migrate, to diffuse; defects will want to move around to adjust the shape of that rock to the imposed stresses. So it would really want to elongate in the vertical direction if I'm pulling that way and squeezing from the sides. But that really comes down to the microscale of how rocks deform, and that gets into how atoms move around, bouncing out of little potential wells, to how grains, the mineral grains that make up rocks, wind up deforming, and how the defects inside the grains move around to accommodate deformation, to help release stress and to deform the rock under the imposed stresses.

Oliver Strimpel: Give me a flavor of what these scales actually are. You've talked about the microscale. What is that scale, and what are the other scales involved here?

David Bercovici: A good example is the mineral grains inside rocks. These mineral grains are typically of the order of, at their largest, hundreds of microns, down to a few microns, and a micron is comparable to, say, the width of a hair follicle. So the grains of rocks themselves can control how rocks respond to stress, how they deform. Because as an atom wants to diffuse, it's more convenient for it to travel along grain boundaries, the boundaries between grains, because those boundaries tend to be disorganized at the atomic scale, so atoms can basically travel along those like easy pathways, through these grain boundaries. So the more grain boundaries there are, the easier it is for the rock to deform, and that goes along with there being smaller grains. With a lot of smaller grains, there's of course a lot of grain boundary, which means there are a lot of channels through which atoms can flow as they're trying to accommodate this deformation. So that's why it's very important to understand how rocks deform, down to that micron scale and even below.

Oliver Strimpel: All right. So let's talk about how the microscale physics can help us with the subduction initiation problem, which, after all, is a large-scale, plate-scale phenomenon.

David Bercovici: Sure. One of the important things is to try to understand how we go from this microscale to the big scale, also referred to as the macroscale. When I say macroscale, I'm really thinking of the plate tectonic, or even the planetary, scale of about 100 kilometers. The typical thickness of a plate, especially as it's approaching a subduction zone, is about 100 kilometers, and it's over that scale that the temperatures will change, of course, from the cool surface temperature near the ocean, going towards hot mantle temperature as it reaches the lithosphere–mantle boundary. And then there is, of course, as I mentioned, the microscale, at the size of grains, which are 1 to 100 microns in size, the individual mineral grains. And then there's another important scale, which comes out of petrology, which we refer to as the mesoscale, sort of in between. These are petrologic units. I mean, if you want to think of it as the size of the rocks in which these mineral grains are embedded, these are of the scale of centimeters to 10 centimeters in size. And that scale also turns out to be quite important for much of what we're looking at.

Oliver Strimpel: Let's talk about the processes that you invoke to help us understand what might be going on when we get subduction.

David Bercovici: Right. So there are two processes that are worth breaking down. One is grain size reduction, which is really how the stresses that are imposed, for example at the tectonic scale, wind up influencing the evolution of those mineral grains: how they change the size of those mineral grains. And then the other is grain mixing, and this is something that happens as grains try to intersperse with each other and mix, from one mineral unit, like an olivine, into another unit, like a pyroxene, and that's at the scale of what we refer to as the petrologic scale, or the mesoscale. So first off, with regard to grain size evolution, or grain size reduction: if I do nothing to them, grains will grow. If I just let them sit there, they will tend to grow all on their own, the way bubbles will grow in foam. You probably know that when you're looking at different types of igneous rocks, you can use the grain size to infer whether they're intrusive or extrusive, because if they're intrusive, they're sitting in a hot environment, so at higher temperatures, being left alone, they'll grow faster. So intrusive rocks are typically characterized by having larger grain sizes, whereas when they're dumped on the surface, as extrusive rocks, they're immediately exposed to cold temperatures, and that cold temperature tends to impede grain growth, so they will tend to have much smaller grains. So that's something that you see frequently in the geological context. But if a grain sits there and grows on its own, then, because bigger grains make it harder for atoms to move around under the imposed stresses, larger grains will tend to make the rock stronger. So if the grains grow, they'll make the rock stronger, and if we reverse that and make the grains smaller, then the rocks get weaker, which would lead us into an important feedback mechanism.

Oliver Strimpel: Okay, so I can see then that grain size reduction would weaken a plate, or weaken a rock. So is that a process that happens progressively as the plate gets older, and something, then, that might counteract this natural hardening and stiffening that you referred to right at the beginning as being the reason why this is such a hard problem to begin with?

David Bercovici: Yeah, it's worth understanding what causes grain size to reduce, since, as I said, if I let a rock just sit there, the grains will grow. How do I get them to shrink? It's something that happens in a deforming environment, in something that's undergoing stress, for example in an ocean lithosphere that's opening, that's undergoing ridge push stress, which I'll get to in a second. So why does deformation cause grain size to reduce at all? This is something that's still a very active area of research. Deformation can reduce grain size in a process that I've sometimes referred to generically as damage, because it's an easy way to say it. But what it really is, is that defects inside the grains, at the subgrain scale, called dislocations, are generated by deformation. They're basically a way to accommodate deformation in the grain also. And as these dislocations are formed and move around, they can accumulate to form new grain boundaries, and that basically means that they spawn a new, smaller grain. So you basically spit out smaller grains from bigger grains, to make small grain sizes. This is more technically referred to as dynamic recrystallization, and it's something that's seen quite easily in laboratory experiments on all types of materials. But in the end, this means that deformation drives a reduction in grain size, which then weakens the rock. As I said, make the grains smaller and the rock gets weaker, and the deformation then focuses on those weak zones, which then allows more damage to the weak zones, which makes them weaker, and so on. And it's this positive feedback which is something that we're often looking for, to cause a strong lithosphere to undergo significant focused, or localized, weakening, which we say is the equivalent of developing a plate boundary, or developing a weakening of the lithosphere to allow subduction. So that's already an involved process, but it also seems that the process helps the most if the medium is a mixture of different minerals, at least two different minerals, for example olivine and pyroxene, as we think is the case in the upper mantle of the lithosphere. First of all, just a mixture of two minerals' grains with each other tends to impede growth. It stops the grains from getting bigger, and the rock from getting stronger, through a process called Zener pinning, and that's basically that the grains just get in each other's way. I mean, if a grain of olivine is trying to grow, and a pyroxene happens to be getting in its way, that olivine can't grow, because it would have to engulf the pyroxene, which is just not really energetically favorable. That process of grains blocking each other also seems to help the damage process, the grain size reduction. So this feedback between grain size reduction and deformation in mixed minerals is something that is often seen in mylonites, especially mantle peridotite mylonites and ultramylonites. Mylonites are these metamorphic rocks that occur in rapidly deforming belts, like plate boundaries, and they're also coincident with the appearance of very strong grain size reduction, and ultramylonites are called that because the grains get reduced down to a few microns, which is very, very tiny; it's like dust size. So the fact that we see these mylonites and ultramylonites occurring in rapid deformation belts gives some evidence to the idea that there is this feedback between deformation and grain size reduction that leads to this weakening. And so our idea of grain damage provides a physical theory for how that can proceed. But again, it seems to work best in something like a peridotite, which is a mixture of olivine and pyroxene.

Oliver Strimpel: Okay. You mentioned another process, grain mixing. Let's talk about that.

David Bercovici: So there are quite a lot of experiments trying to understand how grains would mix with each other. So for example, if I have a rock of olivine and a rock of pyroxene, each one of them made up of a population of grains, how do I get those two, when they're in contact with each other, to have their grains intersperse and mix with one another? And there are various lab experiments that indicate that this occurs under a deforming environment. It's not something that just happens easily, naturally, but once you start deforming the rocks, they start to undergo this mixing. And so my colleague Elvira Mulyukova, who's now at Northwestern University, and I have a model to try to explain this mixing at the grain scale, which happens in the direction of compression. When I'm deforming something, it seems that grains mix in the direction in which the rock is being compressed, which is also something that is seen in laboratory experiments.

Oliver Strimpel: What happens to the rock if the grain mixing happens preferentially in the direction of stress?

David Bercovici: So you have a rock, or a mixture of rocks, so olivine and pyroxene in their separate blobs, in this conglomerate, and each of these blobs is centimeters across. And if I put it under compression horizontally, perhaps the way it is in the oceanic lithosphere, then, if I'm squeezing it horizontally, the mixing is going to occur horizontally. That means that it's at the vertical boundaries between the olivine and the pyroxene that mixing is going to occur, because the mixing is always horizontal, so it has to pass through vertical boundaries between the olivine and the pyroxene. If it's a horizontal boundary, then the mixing is just parallel to the boundary; there's no exchange of mass between those two olivine and pyroxene blobs. So if I'm squeezing it from the sides, the mixing would occur across the vertical boundaries. Well, that's actually pertinent to what's going on at what's called a passive margin, of something like the Atlantic Ocean, which is opening up. It has no subduction zones, and the primary force in something like the Atlantic is ridge push. Ridge push is just from gravitational sliding from the ridge down slope: the weight of the ridge pushes outwards, and it causes the plate to be under horizontal compression. This is something that, in the end, we want to understand: how something like the Atlantic, which is opening up, might wind up getting a subduction zone on what's called its passive margin. The passive margin just simply means the margin between the ocean and the continent, say Europe or North America, or either of the Americas, and it's passive because there's no subduction zone; that's all passive means. And the question is, if that's where the oceanic plate is oldest and coldest and heaviest, wouldn't you typically see subduction zones form there? I mean, we often see subduction zones at the boundaries between ocean and continent, so it's always been a favorite idea that passive margins get heavy enough to collapse. So in our theory, the idea is that as something like the Atlantic, an ocean basin without subduction zones, is opening up, the ridge push force, which is the predominant force (there's no slab pull yet), is all in the horizontal direction. So in that case, the mixing that would occur between these olivine and pyroxene rocks, or blobs, sitting in the lithosphere would occur primarily across the vertical boundaries between those rocks, at that 1-centimeter-to-10-centimeter scale.

Oliver Strimpel: So how does this directional mixing relate to the problem of subduction initiation?

David Bercovici: So in that case, I've got these vertical boundaries, these vertical zones that are 1 to 10 centimeters tall, but maybe millimeters or centimeters thick, depending on how much mixing occurs, in those mixing zones. That's where the olivine and the pyroxene are interspersed with each other, where the grains are interspersed with each other as they're mixing across the zone. And that is where this grain damage is most effective. As we said before, the process of grain size reduction works best where the different types of mineral grains are interspersed with each other. And so these vertically aligned mixing zones, right at the boundary between the olivine and pyroxene rocks, would be zones of weakening, so you would wind up getting these vertical weak bands. They're not kilometers long; they're centimeters tall and millimeters wide, and they basically wind up leading to a fabric, a weak fabric, in the lithosphere that is vertically aligned. In other words, it's like a deck of cards that is standing on its end.

Oliver Strimpel: Oh, I see. Okay. A deck of cards is very weak to being slid over itself, like being shuffled.

David Bercovici: Yeah, exactly. So if I have the deck of cards standing on its end, it's weak to being slid vertically up and down, but it's very strong to being bent horizontally. It's easy for the deck to slide downward, in a sense. The vertical weak bands provide that preferential weakness in that direction.

Oliver Strimpel: So let me see if I understand this. You're saying grain size reduction causes weakness, since smaller grains are less strong than bigger grains, and because we've got this compressive stress in a plate before it subducts, which is coming from ridge push, the grain size reduction happens preferentially along vertical boundaries between grains, because that's the direction in which grains are mixing, and the mixing is a very efficient way of reducing grain size. So we get these vertically aligned regions of weakness on what you called the mesoscale, like a few centimeters.

David Bercovici: That's exactly right. The only thing I would say is that the vertical mixing occurs not between the grains, but at the vertical boundary between the rocks, the collections of grains of olivine on one side and pyroxene on the other side. There's a vertical boundary between them, and the mixing occurs between those. That's an important distinction, because it says something about how the petrological structure, what the lithosphere looks like at the scale of the petrological signature that's pushed into it from the mantle below, matters quite a lot in terms of setting up those vertical bands.

Oliver Strimpel: Let's suppose we actually have subduction starting. Then we go from a compressive, ridge-push type of force to an extensional, tensile, slab-pull force. Yet we know that if slab pull is driving subduction, the plates must retain at least some tensile strength in that direction. So what happens in your model once a plate actually starts to subduct?

David Bercovici: Our model is not going through the whole process of seeing the subduction zone initiate. In a way, what we're doing is setting up the plate to be weak enough to subduct. And again, the vertically aligned weak zones are important because if that plate wants to sink downwards, it's weak in that direction; it's weak to that downward sinking and bending. So it sets it up nicely for initiating subduction. And also, to point out, as the plate thickens and grows, that whole region of weak fabric would be predicted to grow to be pretty thick also, you know, a big fraction of the lithosphere. But as you said, once we get subduction going, at some point, what does that do? Because the plate, as you said, is under slab pull, which would presumably put it under tension. It's not always observed; stress in the lithosphere is not always a very clean signature. But in principle, slab pull would put it under tension, and in that case, because the tension is in the direction of the plate and compression is vertical, the mixing would be predominantly in the vertical direction. And that would mean that, instead of the vertical boundaries between these rocks, these petrologic units, undergoing mixing, it's the horizontal boundaries that would undergo mixing, because it's in tension in the horizontal direction. And this would, by the same train of logic, lead to horizontal weak bands. It's not like once you've got subduction going they would all just flip around, because it takes time for mixing to occur and for things to adjust. But in the long term, you know, after maybe tens of millions of years, you might start to see some of that vertical weak band structure erased, and replaced with a horizontal weak band structure. And what that says is that if I want to continue to pull the plate, those horizontal weak bands are not going to make the plate weaker. In other words, I can pull a deck of cards. Even though the deck of cards is weak to sliding, I can still pull the deck of cards, and it's quite strong if I'm pulling it end to end. So that still allows the slab pull force, as the slab is subducting, to continue to pull the plate behind it.

Oliver Strimpel: Is there any way to test this theory?

David Bercovici: First of all, with regard to just grain size reduction and mixing, again, we see a lot of evidence of that in mylonites, when you're looking at that scale. You can see evidence of this in laboratory experiments, deforming rocks, deforming mixtures of rocks. There are a lot of very recent, interesting experiments that have seen this type of grain size reduction coincident with mixing, which is very interesting and something that we're working on quite a bit. Now, with regard to seeing it at the big scale, the plate tectonic scale, our main source of observations is seismology, and it's not always easy to see. So I guess the best thing you can say is sort of a negative result. In that regard, you can use seismological measurements to infer something about the fabric of plates. When I say fabric of plates, that has nothing to do with what I'm talking about in terms of the weak zones; it's mostly to do with crystallographic orientation. So for example, if I have the Pacific plate, which is moving over the mantle, the direction of deformation, the shearing of the mantle, especially in the lithosphere as it's being pulled along, tends to align crystals in the direction of plate motion, in a phenomenon called crystallographic preferred orientation. And it really has to do with how the olivine lattice structure aligns itself in the direction in which the crystals are being strained. And seismology can see a signature of that, which implies that that crystallographic orientation is horizontal and in the direction of plate motion, and that's a very distinct measurement in the Pacific. But the Atlantic is interesting, because the Atlantic has no subduction zones, and measurements of that same crystallographic orientation in the Atlantic show that it's much less; it's probably something like 50% less than in the Pacific. And our model would help predict why the Atlantic has less of that signature from crystal preferred orientation. It does two things. One is that a lot of the mixed zones that are weak don't allow this crystallographic orientation. It has to do with the fact that they have small grain sizes, and the way they deform, by diffusion creep, precludes crystals from being crystallographically oriented that way, which normally happens in dislocation creep. But the other thing is that our model would also say that there is a vertical fabric from these weak bands. Again, it's a different phenomenon, having to do with these weak bands being aligned, but that would, in a sense, muddle the signature of anisotropy. What seismology sees in anisotropy is a thing called shear wave splitting. It's not measuring something directly in the minerals; it's looking at seismic waves passing through the rocks and the lithosphere. In shear wave splitting, you have two different types of S waves, or shear waves. It's like waves on a string: there are ones that basically waggle up and down vertically, and ones that go side to side, and they're called SV, vertical S waves, and SH, horizontal S waves. And depending on the direction those waves are going, if they have totally the same speed, it's what's called an isotropic rock. So that would be a rock that has no fabric to it. But if a rock has some fabric to it, then those two waves can have different speeds, depending upon the direction in which the wave is moving. And so from this shear wave splitting, the fact that these two shear waves have different speeds, you can infer that the lithosphere has this fabric to it, and it's very strongly seen in the Pacific, which has this fabric aligned in the direction of plate motion. And again, this is much weaker in the Atlantic, and so we would contend that our mechanism allows for that 50% weakening in the Atlantic. So it's not a direct observation of our mechanism. It's basically a way of saying that our mechanism helps explain why you see less of it in the Atlantic.

Oliver Strimpel: That's interesting. So this vertical banding of weak zones counteracts the crystallographic orientation that would otherwise tend to make crystals go horizontal, and we actually see that in the seismic wave anisotropy, which is less in the Atlantic than in the Pacific. So that's at least consistent with your theory, if not a confirmation.

David Bercovici: Right. Yeah, well, it's a long shot to get any of these theories confirmed. That's the nature of our field.

Oliver Strimpel: Does this theory help us understand why subduction, or more generally plate tectonics, appears to be unique to the Earth in the solar system?

David Bercovici: Yeah, as far as we know. I mean, our dataset is pretty small, obviously. So with regard to just the grain damage part of it, in other words grain size reduction: the classic explanation for why the Earth has plate tectonics is that water weakens everything. We would take a different tack: that it's not so much that water weakens things, because the Earth's lithosphere is not necessarily any wetter than that of any other planet. What we would contend is that water, or the presence of liquid water, helps keep the surface of the Earth cooler. And it's a bit of a chicken-or-egg question, because plate tectonics also is an integral part of that problem. It's a combined rock cycle, from plate tectonics, and the hydrological cycle, from liquid water existing, that drives a geological carbon cycle that scrubs carbon dioxide out of the atmosphere and keeps our planet's surface at a pretty stable temperature, relatively speaking, for hundreds of millions of years. It's a sort of well-known feedback, how the geologically driven carbon cycle keeps the climate stable. So we would contend that this cooler, long-term cooler, surface temperature of Earth allows for a cooler lithosphere. And while the cooler lithosphere does make the lithosphere thermally stiffen (as it gets cooler, it wants to get stronger), it is also very good at inhibiting grain growth and also at enhancing grain damage. And so it's a good condition for the mechanism we're proposing, in terms of enhancing grain size reduction. Whereas if you're on another planet that is much hotter, maybe grain growth is faster and grain damage is weaker. Something like Mars, which is, of course, colder, is much smaller, and so it has smaller stresses and much weaker convection, because it's lost a lot of its heat. So the theory does use that idea of what water, in effect, does to the surface, and how that winds up controlling the tectonic environment. So surface conditions play a big role, and the size of the planet as well plays a big role, in whether a planet might have plate tectonics, at least according to our model. With regard to the mixing component of it, that's a very new and interesting question. The mixing also is, of course, going to be thermally controlled, but it also seems to be something that's controlled by the fact that you need to have a mixture of two different rocks, and the nature of the rocks themselves may control how easily the grains can mix with each other. So in a way, it's saying that the composition of the planet and its petrology may make a big difference to whether or not a planet is likely to have these types of mechanisms that would allow subduction zones, or more plate boundaries, to form. But this is very new stuff that we're talking about, and hopefully it opens up a new direction in how petrology informs a lot about how a planet's plates, or the conditions leading to plate tectonics, might evolve, from early on, say post-magma ocean, to the present, and how the composition of a planet in a given solar system may play an important role in whether it has plate tectonics or not.

Oliver Strimpel: David Bercovici, thank you very much.

David Bercovici: Well, thank you for having me. It was enjoyable.

Oliver Strimpel: For more about Geology Bites, as well as pictures and illustrations that support this podcast, you can go to geologybites.com.