This transcript was generated automatically and lightly edited for readability.

Oliver Strimpel: This is Geology Bites with Oliver Strimpel. The surface of the Earth is capped by a thin rocky layer called the lithosphere. The lithosphere is composed of discrete plates which float on the asthenosphere, the ductile upper part of the mantle. Since the asthenosphere is far too deep down for us to access it by drilling, our knowledge of it comes from mantle rock that finds its way to the surface, such as in ophiolites or kimberlites, or via geophysical phenomena that probe the mantle remotely, such as seismic waves, gravity and magnetic fields. But many of the properties of the mantle that are thought to determine the mechanisms of plate movements, subduction and magmatism are poorly understood, as they depend on the detailed physical and chemical behavior of the mantle under the conditions that prevail at depth.

David Kohlstedt has sought to replicate these conditions in the lab so as to observe these critical mechanisms directly. He is professor emeritus at the School of Earth and Environmental Sciences at the University of Minnesota. David Kohlstedt, welcome to Geology Bites.

David Kohlstedt: Thank you. I am delighted that you have invited me.

Oliver Strimpel: What exactly are the properties of the asthenospheric mantle that you try and learn about in the lab?

David Kohlstedt: Our experiments are designed to study the viscosity of the asthenosphere. The question that motivates us is, why is the asthenosphere so weak? Is it simply that the temperature has reached a critical point and the asthenosphere begins to flow, much like warming molasses in January? Or is it the presence of melt and water that makes the asthenosphere weak? To answer these questions, we have concentrated our research on quantifying the dependence of viscosity on the melt fraction and on water content.

Oliver Strimpel: And though those two quantities affect each other, they're two separate things, and I understand that in your experiments you study the effect of each of these separately. Let's start with the melt fraction. Do you experiment on samples of mantle rock that we can find on the surface?

David Kohlstedt: No, we don't use rocks provided by Mother Nature. Those rocks that have come to the surface in kimberlites or ophiolites have been on the surface long enough that they begin to weather and introduce products that we'd rather not have in our samples during the experiments. So instead we take gem-quality olivine, which is the dominant mineral in Earth's upper mantle, and we combine it with basalt. We take the gem-quality olivine and grind it into a powder, and then grind some of the basalt into powder and mix a specific amount of basalt in with the olivine. The basalt at high temperature will melt, but the amount of melt that you get from the basalt is independent of temperature, so that we can really isolate the effect of melt fraction on the deformation behavior of the material.

We also sometimes use lherzolite for our samples. In lherzolite, the melt is formed by melting the olivine plus orthopyroxene plus clinopyroxene, which react together to form the melt. But in that case the amount of melt is a function of temperature. So we use two different approaches, two different types of samples, and then the challenge is to demonstrate that you get essentially the same results taking either approach.

Oliver Strimpel: That's interesting. So you actually control the amount of melt in your samples through really the composition, the mixture of materials that you put into your sample to begin with.

David Kohlstedt: Yes, that's exactly right. And it's a way in which we can really design the experiments to probe, for example, what is the strength of material that has 1% melt versus having 10% melt?

Oliver Strimpel: And presumably you keep those as dry as possible, so as to take water out of the equation completely for those experiments.

David Kohlstedt: Yes, it's important for us to try to design experiments where we can vary one parameter at a time. It makes understanding the physics of what's going on in the problem much more straightforward, and there aren't complications arising from having three or four different variables changing, where every time you change one variable, you change two or three others at the same time. So that's really an important aspect of the way we try to design our experiments.

Oliver Strimpel: Let's talk a bit about water, then. How do you control for the amount of water in your sample?

David Kohlstedt: We work primarily with the mineral olivine, and we fabricate samples, again, by powdering gem-quality single crystals, hot-pressing them and making them back into a solid cylinder of material. In this case, we put an olivine sample into a nickel capsule. Nickel helps keep the olivine in its stability field, so that it doesn't oxidize or reduce. To that capsule we add a couple of drops of water, and then we seal the capsule; we weld it shut. Then we vary the water content in the sample by heating the assembly to high temperature and high pressure, and the water concentration, or water fugacity, in the capsule is a function of temperature and pressure. Once at high temperature and pressure, the water dissociates and produces hydrogen ions, which diffuse fairly rapidly into the olivine.

Oliver Strimpel: So you introduce a fixed amount of water into these experiments at the beginning, and then you control for the amount of water by varying the temperature and pressure. And somehow you know what the effective amount of water is inside the sample as a function of temperature and pressure.

David Kohlstedt: We do. We've done calibration experiments prior to doing the deformation experiments, in which we simply go through this routine, take the sample to a certain temperature and a certain pressure and then quench it, and then use infrared spectroscopy to measure the amount of water in the sample. So when we do a deformation experiment and we know what the temperature and pressure conditions are, we know what the water content in the sample will be.

Oliver Strimpel: What is the size of the samples that you use in your experiments?

David Kohlstedt: Typically a sample is about 1 centimeter in diameter and 2 centimeters in length.

Oliver Strimpel: So the conditions in the mantle are extreme: enormous pressures and high temperatures. What conditions do you aim to achieve in your samples during the experiments?

David Kohlstedt: We typically go to a pressure of 300 megapascals and a temperature of about 1,200 to 1,300 degrees Celsius. Now, that corresponds only to a relatively shallow depth in the Earth, of 10 kilometers, so we certainly need to be able to determine the flow laws that allow us to extrapolate to higher pressures and higher temperatures.

Oliver Strimpel: Do you not try and simulate those greater depths and higher temperatures because it's just technically too difficult?

David Kohlstedt: It is technically difficult, that's for sure. Our limitation is that we're doing experiments in a steel pressure vessel which has a hole drilled down the axis of the vessel, which limits our ability; maybe 1 gigapascal would be an upper limit of what would be possible. But there are other researchers who have been very clever and designed apparatuses that can go to significantly higher pressures, up to 10 gigapascals, which will get you down to roughly 300 kilometers depth in the Earth. These pressure vessels are taken to a synchrotron, a synchrotron source for X-rays, and the high-intensity beam of X-rays is shone on the sample during the deformation experiment. That allows you to measure the length of the sample, and therefore monitor the rate of deformation of the sample, and, from X-ray diffraction, to measure the stress that's being applied at the same time. So if you have stress and strain rate, you have viscosity.

Oliver Strimpel: But in those experiments that get to these much higher pressures and temperatures, presumably they don't do it with a sample size of a few centimeters, like you're doing.

David Kohlstedt: That's a really good point. The sample size there typically is a millimeter by a millimeter, so the samples get much smaller as the pressure goes up. Sample sizes tend to get progressively smaller.

Oliver Strimpel: OK, so you have this centimeter-scale sample. How do you actually sustain these high temperatures and also these incredibly high pressures?

David Kohlstedt: In our lab, we focus on using gas as a confining medium for the sample, and the pressure vessel that we use is about a foot across and about 3 feet long. Down the center of that vessel, there's a hole drilled that's about 3 inches in diameter, so it's a relatively small hole in a relatively large vessel. And then the trick is that you put end caps on the vessel that seal it. So you can pump gas into the vessel and increase the pressure through a series of pumps, each stage of which increases the pressure of the gas. The gas comes basically out of a welder's argon tank at maybe 10 megapascals, and you can increase the pressure with the series of intensifiers until you get up to 300 megapascals.

Inside that bore slides a furnace. The furnace design in some ways is relatively simple. There's a lot of insulation that keeps the heat from the center of the furnace from going out and overheating the steel vessel, and the furnace is basically simply a wire-wound aluminum oxide tube. It works much like your toaster works at home. The whole thing is cooled on the outside to keep the steel basically at room temperature while the center of the furnace gets up to high temperature. You can run a piston through one of the plugs at the end of the pressure vessel, and it's moved by a mechanical motor on the outside of the vessel. And all the instrumentation that allows us to measure the stress, and the displacement gauges, are inside the vessel, and that design allows us fairly high-resolution experiments. If you put all the instrumentation outside, you could do similar types of measurements, but you then have to deal with the fact that you're also deforming the piston at the same time, at least elastically, if not plastically. And you have other distortions that affect the stress; there's friction on the piston, so you have to account for the friction. With the instrumentation inside, you get around those complications.

Oliver Strimpel: How do you make a sensitive instrument survive under those conditions inside your apparatus?

David Kohlstedt: This is a really good question, and it takes clever engineering to make it happen. If you just buy a commercial strain gauge off the shelf, they're typically attached to a sample via an epoxy. Put epoxy inside a pressure vessel, and when the gas gets up to a pressure of 300 megapascals, it permeates the epoxy really quickly, and that's not a problem at that point. But when you lower the pressure back down, all that argon gas trapped inside the epoxy wants to get out, and it tries to diffuse out, and when it does, it forms bubbles. And when you now have a high-pressure bubble and you take it down to one atmosphere, the bubbles will explode and your strain gauge is destroyed. So it's complicated, but people have worked on this problem to try to design displacement transducers that will work under very high-pressure conditions.

Oliver Strimpel: So how do you actually go about determining the viscosity of your sample under these conditions?

David Kohlstedt: Well, our focus has been on two types of experiments. Until about 2000, or the turn of the century, we did most of our experiments in uniaxial compression. We would put a sample inside the pressure vessel, confine it with high-pressure gas, and then use a piston coming through the lower seal in the vessel to push on the sample and compress it and make it grow shorter as a function of time. The other approach we've taken is torsion experiments, where you can twist the sample to fairly high strain. If you do one full revolution of a sample in torsion, you can get strains on the order of 10.

Oliver Strimpel: Just to explain the units: strain is defined as the ratio of a change in a dimension to the original dimension, and strain rate is the strain change per second.

David Kohlstedt: And that's kind of fun to do, because the Earth does deform rocks to fairly high strains, and you can begin to reproduce that. Again, what we measure in these experiments typically is the length of the sample as a function of time, that is, how long the stress has been on the sample, and you can watch the sample get shorter. Or, if you're twisting, you can watch the deformation of the sample rotate through. We also measure the stress on the sample at the same time. So we typically apply a constant rate of deformation and measure the stress that's required to keep that rate of deformation going. We've really focused on the steady-state part of the problem, where you twist the sample at a constant rate and eventually the stress will level off. When the rate of deformation is fixed by the stress level, you've reached a steady-state condition, and that's where we've largely focused our effort.

Oliver Strimpel: So what are the time scales of your experiments? I mean, I know on Earth these things happen over millions of years. How long do you take to rotate your sample, say through 360 degrees or two revolutions, and achieve that steady state that you're then trying to measure?

David Kohlstedt: For torsion experiments, we typically work at strain rates of about 10⁻⁴, maybe 5 × 10⁻⁵ per second. But one of the points about gas-medium deformation apparatuses is that they really do take some level of monitoring. So we typically start the experiment early in the morning and run it for maybe a 12-hour period, and that's what really dictates the strain rate. We could deform more slowly, and in compression experiments we do. We might deform at 10⁻⁶ per second, but the Earth is still out there at 10⁻¹⁴. So the difference between 10⁻⁴ and 10⁻⁶ doesn't get you hugely closer to what's going on in the Earth. And so, again, the element that's important is to try to design experiments where you really take advantage of what you can do well, and then, as we've talked about a little, have equations, or flow laws, that allow you to extrapolate that with some level of confidence to the Earth.

Oliver Strimpel: How long does it take, roughly, then, to get up to that steady state?

David Kohlstedt: Within an hour or two, you can get from ambient conditions to steady-state deformation. And the nice part about that is that if you want to determine, for example, how the viscosity of the rock depends on stress, you can look at several different stresses during the experiment and then go back, analyze the data and determine the constitutive relation that describes the dependence of viscosity, in this case on stress. Or you could do temperature-stepping experiments and maybe look at three or four different temperatures on the same sample. So we try to extract more than just getting to steady state. We try to kind of tickle the sample a little bit once we have it at conditions, and get it to tell us exactly how it wants to behave.

Oliver Strimpel: Let's talk about your results, then. How does the viscosity at various temperatures and pressures depend on the melt fraction and the water content?

David Kohlstedt: The viscosity of these olivine-rich rocks at a given pressure and temperature decreases approximately linearly with increasing water content in the sample. For melt fraction, the viscosity decreases approximately exponentially with increasing melt fraction in the sample.

Oliver Strimpel: So it's very, very sensitive to melt fraction.

David Kohlstedt: It is very sensitive to melt fraction. If you add 1% melt, you get a decrease in viscosity of about a factor of 2. If you go to 10% melt, you decrease the viscosity by well over a factor of 10. So if you can store much melt in the sample, it will be very effective at reducing its viscosity. Of course, you have to keep in mind that melt has a lower density than the solid rock around it, so it does want to rise buoyantly out of the mantle. So there is some competition between how much melt you can maintain and its effect on the viscosity of the rock.

Oliver Strimpel: But then the melt fraction, as you just said, is more mobile. So does that actually escape from your sample? And do you find it all kind of clustering around the edge of your capsule?

David Kohlstedt: Fortunately, gravity doesn't play a very big role, because our sample sizes are so small, so in fact surface tension is more than adequate to keep the melt distributed essentially homogeneously throughout the sample.

Oliver Strimpel: OK, so the melt stays within the sample, but how is it structured? Does it form in particular regions, and does that give us any insight into what happens to melt when it occurs naturally 100 kilometers down?

David Kohlstedt: It's a really interesting question. The first thing that happens if you start twisting a partially molten sample is that the melt pockets, the melt which sits in triple junctions between grains, begin to get themselves oriented. They know about the stress field and about pressure gradients in the sample. If you start to twist to higher strains, typically a strain of one or higher, the melt actually begins to segregate. So at first the melt pockets at the grain scale orient, and then those melt pockets feel pressure gradients developing throughout the sample, and the melt will segregate and form melt-rich regions, which we've called melt-rich bands, or sheets of melt that are separated by regions that are basically melt-free. So you produce bands that are spaced maybe 100 microns apart, and the amount of melt in those bands might reach 25%, while the melt fraction in the regions between the bands actually drops way down. So you produce these really melt-rich regions that can act as shear zones, because these high-melt-fraction regions have low viscosity, so you can shear on them easily, and they're also high-permeability paths. Those paths may be important as one way in which you can remove melt from the melt source region and get it to volcanic structures sitting near the surface of the Earth.

Oliver Strimpel: In the real asthenosphere, we think there is both melt and water present. Can we use your results to predict what happens in those circumstances?

David Kohlstedt: We can certainly make some inferences from our experimental results. I think ultimately these results go to people who are doing numerical models of convection and subduction, but we can understand where these results might be important. One area that we've talked a little bit about is what happens beneath the mid-ocean ridge as upwelling mantle begins to melt as it depressurizes, and that melt moves upward. As it does, the water that's in the mantle partitions into that melt, and the melt, as it goes up buoyantly to the surface to form oceanic crust, has to strip the mantle in that region of its water, and therefore you should be strengthening that part of the mantle. So one way in which you could think about forming a lithospheric plate is by having the melt move through the rock, suck out all the water and leave behind dry rocks that are stronger than the wet rocks. So that's one way in which you can think about the evolution of a lithosphere.

And there's been a group in Japan who've looked at what happens in the oceanic lithosphere, seismically looking at the transition between the lithosphere and the asthenosphere. That interface is often called the lithosphere-asthenosphere boundary, and that transition is relatively sharp. It occurs over a distance of about 10 kilometers, and it's really hard to see how you can do that just by changing the temperature, which changes relatively gradually as you go down to greater depths. They've argued that if you had even 1% melt in the asthenosphere and you allowed it to segregate into these melt-rich sheets that we talked about, the net viscosity of that part of the mantle, the asthenospheric part of the mantle, would in fact be significantly lower than in the lithosphere above it. Again, not with a large amount of melt, but with a small amount of melt, where that melt is localized, or focused, into these melt-rich bands.

Oliver Strimpel: Well, that's really interesting. So it suggests that really the mechanical contrast between the lithosphere and the asthenosphere is fundamentally down to the melt fraction.

David Kohlstedt: Certainly you can make the case for that. The experimental results from the lab would support that point of view. Again, it requires, in that particular model, that the melt segregate into melt-rich sheets, which provides some areas that are really quite weak sitting beneath the lithosphere. I think there's still quite a bit of debate about exactly how this works. There have been a number of papers written more recently that take a more detailed look at the roles of water and the role of melt.

Oliver Strimpel: We already alluded to the enormously different time scales, and then of course there are also the space scales of your experiments as compared to those that prevail in the tectonic settings that we're trying to understand. How do we deal with that? And are you able to develop scaling laws that somehow or other enable us to get from the 24-hour, centimeter scale of your experiments to the vastly slower, millions-of-years and kilometer scale of the actual Earth?

David Kohlstedt: Well, fortunately, we're not the first people to work on the strength of materials. There are whole fields of materials science and engineering, for example, that have been very concerned with the strength of materials, or the viscosity of materials. They also have engineering problems where you need to extrapolate in time scale. Not as extreme as we're interested in, but they might need to extrapolate from testing materials over a day or two, much like we do, to applications that might need to last maybe 100 years. That might be the case in buildings or other kinds of structures. Or you're testing jet engines that need to last for 30 or 40 years. One really wants to do testing where you have the scaling laws that allow you to go out to at least 30 or 40 years. So that helps a lot.

And there's also a lot known about the details of the physics of ductile deformation of materials. We know that deformation occurs primarily by the movement of defects. Those defects can be point defects, like atoms and ions moving around, or they can be dislocations, which are line defects moving through the material. So we at least have the starting skeleton of what we're trying to work out in terms of constitutive relations, or flow laws, for extrapolation. And once we've done the extrapolation, once we've taken our lab results and extrapolated them to a depth of 300 kilometers and a temperature of 1,500 degrees Celsius, we can look in the Earth and find places where there are geophysical measurements that give us some constraints. People have used analysis of the geoid, for example, and analysis of the rate of postglacial rebound as ways of constraining the strain rate and calculating the stresses that should be applicable under those deep Earth conditions. And it turns out that the lab results do a pretty good job of predicting what, in fact, the Earth is telling us it's doing.

Oliver Strimpel: Is it possible to say that we know what the viscosity is on geological time scales at, say, 300 kilometers depth and 1,500 degrees Celsius to within 10%, or a factor of 2, or just the right order of magnitude?

David Kohlstedt: The right order of magnitude might be roughly where we're at right now. I think these things can be tightened up with additional experimental work, but there are a lot of complications that haven't even been put into the experiments at this point. I've been talking about olivine, but the Earth is made up of more than olivine. What happens if you add a little bit of pyroxene to that sample? Or what happens if you change the grain size? So it's a big parameter space, and I would argue there's a real need for additional experiments, and additional experimentalists to do those experiments, in order to make the flow laws and the extrapolation of those flow laws more robust than they are right now.

Oliver Strimpel: I did an earlier session with Allen McNamara, who models the deep mantle, does computer simulations, and no doubt uses what he can from experiments such as yours as inputs to his models. And I asked him what the biggest uncertainty was in his models, and he said that he really thought that we didn't know the viscosity at depth to within several orders of magnitude, because it's extremely nonlinear in the way it depends on these parameters. So getting down to one order of magnitude is something we should be extremely proud of.

David Kohlstedt: Well, I think it's quite amazing, actually, that we can do that, if you think about all the complications in the Earth in terms of knowing what the temperature is. The pressure we can calculate quite well, but there's a lot involved in trying to constrain what the physical parameters are in terms of rock composition and water content and melt fraction. There are a lot of variables in the system, and what's down there is a long way away.

Oliver Strimpel: Let's talk a bit about some of the implications of your experimental results for our understanding of plate tectonics. For example, does your work help us understand whether the asthenosphere might have the convection cells that have been theorized, with lithospheric plates being carried along above them?

David Kohlstedt: The viscosity measurements from the lab do help constrain what's going on in the convection cells. For example, when I first started thinking about this problem, people were modeling convection with a linear, or Newtonian, viscosity, where the rate of deformation was linearly proportional to the stress. That has really changed due to the laboratory experiments, which demonstrate quite clearly that the viscosity is indeed a function of the stress, not a Newtonian or linear viscosity. I think that has entered into people's calculations of convection cells and has had some effect on how people think about modeling convection, locally near the surface but even at greater depths in the Earth.

Oliver Strimpel: In an earlier episode, David Bercovici talked about how stressed mineral grains in oceanic lithosphere can radically weaken oceanic plates, which in turn could help explain how such plates can sag down into the mantle and start subducting. Does your work support this idea?

David Kohlstedt: We've recently undertaken experiments in which we combined two minerals, still not all the complexity of the mantle, but taking two minerals, olivine and pyroxene, mixing the two together, and then deforming them to very large strain and watching the evolution of the microstructure and the strength of the rock as the strain increases. And much as was predicted, the two phases, as they mix together, often form shear zones that are weak, and these are the shear zones that, extrapolated back to the subduction problem, allow the plate to bend and deform and to soften as it goes down deep into the mantle. So yes, there's a really direct connection between David Bercovici's work and what we've been doing. And our hope is that we're able to provide him with, for example, constitutive equations that describe the strength of the rock, or the viscosity of the rock, as a function of, for example, how much pyroxene is present. What is the olivine-to-pyroxene ratio? How quickly does the strength evolve? As you go to higher strains, these rocks tend to get weaker and weaker and weaker.

Oliver Strimpel: Do your results on the melt fraction help us understand how the mantle melts and causes volcanism?

David Kohlstedt: Once the melting has occurred, our results really focus on telling you something about how strong the rock is and how rapidly you can transport melt from the source region to the surface. Again, if you form these melt-rich bands, they really are high-permeability bands and allow melt transport to occur much more quickly than it would if the melt had to move through the olivine grains. Like water moving through sand, it would move fairly slowly. But if you can concentrate the melt into certain regions, the permeability and rate of melt transport are much higher through those melt-rich regions than in the water-through-sand kind of analog.

Oliver Strimpel: Do your experimental results help us understand the conditions under which rocks give way suddenly under applied stress and trigger an earthquake?

David Kohlstedt: If you think about it in terms of depth in the Earth, we've mapped out the lower part of the deformation behavior, and there have been studies that have focused totally on brittle behavior that have mapped the upper part, near the surface, as a function of depth. If you take, as a crude but first-order approximation, the two rheological laws, the one for brittle deformation up above and the one for ductile deformation down below, and look where they intersect, that begins to get you into the right regime of where you have to have interaction between brittle and ductile processes. If you look at the location where these two rheological laws intersect each other, you're getting yourself into the earthquake-forming region, and then it's a matter of mapping out in detail: now I've got to have both brittle processes and ductile processes happening, and how do they interact? Does one stimulate the other? Does one enhance the other? Does one suppress the other? You begin to know where the earthquakes are occurring, and the complication really is trying to understand those physical details at the microstructural level: how, in fact, does a crack forming cause dislocations to be activated, combining the brittle behavior of cracks and the ductile behavior produced by dislocations?

Oliver Strimpel: David Kohlstedt, thank you very much.

David Kohlstedt: It has been a pleasure chatting with you, Oliver.

Oliver Strimpel: To see pictures and illustrations that support this podcast, go to geologybites.com, where you'll also find transcripts and a subject matter index of all the episodes. There you can also give me feedback, which I welcome, as well as sign up to get my emails about new episodes.