1Mr Administrator,
Honourable Professors,
Dear colleagues and friends,

2It is a great honour to be able to present you today with some of the elements of the discipline that fascinates me. I would first and foremost like to thank the professors of the Collège de France for putting their trust in me, and to thank you all for being here.

3I would especially like to thank Édouard Bard, who presented you with the project of the Chair of Physics of the Earth’s Interior, as well as Xavier Le Pichon, who held the Chair of Geodynamics from 1986 to 2008, and whose encouragement played a decisive role in my career. At the end of the 1960s, Xavier Le Pichon both witnessed and helped engineer the plate tectonics revolution. He then dedicated the better part of his career to understanding the movements of these plates on the surface of the Earth and their deformation. The new Chair of Physics of the Earth’s Interior, as we shall see, will focus mainly on the current fundamental question of the profound origin of these tectonic motions.

4I would like to dedicate this lecture to the memory of my parents, Zofia and Kazimierz Romanowicz. Those who knew them will know how glad they would have been to be here.

Introduction

5Solid earth geophysics is a vast domain of applied mathematics, physics and chemistry (as well as biology), which encompasses a large number of disciplines. As it is impossible to cover everything in an hour, I had to make a selection. During this lecture, I would like to provide you not only with a necessarily highly simplified overview of our current knowledge on the structure and internal dynamics of the Earth, but also with insight into the most compelling questions driving research in this domain. After covering a few general points, I will briefly describe current progress and challenges in my own discipline, global seismology.

6I will start off by reminding you of the exceptional place the planet Earth holds in our solar system, which, by virtue of its consequences on our existence and our well-being, is a major driving force behind research. It benefits from the combination of a number of very particular conditions:

7(1) Earth is the only planet with a surface that meets temperature and pressure conditions close to those of the triple-point of water, that is, where water, so crucial to life, can coexist in its three forms: solid, liquid and gas.

8(2) It is also the only planet to currently have active plate tectonics.

9In fact, these two points are related. The presence of water plays an important if not fundamental role in plate tectonics: the rock hydration process influences rocks’ physical properties, making them more malleable and increasing their water retention capacity in a solid state. The presence of water also lowers the point of partial melting of rocks, an important phenomenon for the generation of magma at mid-ocean ridges and subduction zones. This is a very particular situation that is difficult to reproduce through modelling, where the lithospheric plates are essentially rigid, and deformation is almost entirely concentrated on their boundaries.

10It is estimated that the quantity of water enclosed in the rocks of the Earth’s mantle, in their hydrated form, could reach once, if not twice the combined volume of all oceans, which themselves cover two thirds of the surface of the Earth. This is however not what Jules Verne imagined in his Journey to the Centre of the Earth, an underground ocean with beaches and caves, which could not have resisted the enormous pressure exerted on the rocks; this is indeed a “hidden” ocean in solid rocks. One of the foci of current leading multidisciplinary research is to determine how the hydration of mantle rocks is distributed: is it limited to the first 100-200 km below the surface, or is it significant down to the upper mantle transition zone (400-700 km), or even deeper?

11(3) Finally, the Earth also has a sufficiently powerful internal magnetic field to create a real shield against cosmic rays in the ionosphere —yet another crucial element for life on Earth. This magnetic field is generated by movements of fluid in the liquid metallic core, mainly composed of iron, through a dynamo process. Note that the central part of the core, the inner core, is solid and composed of even purer iron crystals. It has played an important role in maintaining this dynamo over geological times: certain chemical elements lighter than iron are discharged into the liquid part of the core during the crystallisation of the inner core, which increases in size by a few millimetres every year. The rise of light elements towards the surface of the core generates convection in the outer core.

The role of plate tectonics and elastic rebound in the geographic distribution of earthquakes

12The price to pay for a planet with a dynamics so favourable to life is natural disasters, namely, of particular interest to us here, earthquakes, volcanic eruptions, and their consequences, such as tsunamis and landslides. The distressing images of the disaster triggered by the magnitude 9 earthquake in Japan on 11 March 2011, and the ensuing tsunami, are a vivid reminder of the considerable natural forces at play. Whether we think of the last earthquake in Japan, or those in Haiti in January 2010, in Chile in February 2010, those in New-Zealand in September 2010 and February 2011, or even the gigantic Sumatra earthquake from 26 December 2004 and its tsunami, already somewhat forgotten: all are the inevitable consequence of the movements of the Earth’s crust, associated with plate tectonics, the outlines of which I will now briefly recall.

13Plate tectonics, coupled with elastic rebound theory, provides a simple explanation, to first order, of the occurrence and location of the overwhelming majority of earthquakes and volcanic eruptions. It took over fifty years for plate tectonics, put forward as early as 1912 by Alfred Wegener in his paper on continental drift¹, to be definitively accepted. The observation of magnetic anomalies is what finally convinced the scientific community in the second half of the 1960s. When rocks solidify, forming the new crust at mid-ocean ridges, they acquire a magnetisation of identical polarity to that of the prevailing magnetic field at the time of solidification. Positioned symmetrically to the ridge, the magnetic anomalies constitute a record, in the rock, of the successive inversions of the magnetic field over time, as the rock moves away from the ridge. Xavier Le Pichon² was actually the first, in 1968, to measure with precision the speed of the plates’ movement, based on this type of magnetic data (fig. 1).

Figure 1. Magnetic anomalies and dating the seafloor

14Thus, the surface of our planet is covered by a dozen rigid plates, called lithospheric plates, of irregular shapes and sizes, and on average 80 to 100km thick. They move in relation to one another at relative speeds of a few centimetres every year —the speed at which our nails grow—, driven by currents of matter in the Earth’s mantle. Virtually all deformation is concentrated at the boundaries of the plates, which are primarily of three types. First, two plates move apart along submarine volcanic chains, called mid-ocean ridges, giving way to fresh magma which increases the surface of the plates as it solidifies, thereby forming a new crust. Second, the excess mass produced at one end is in turn absorbed by the subduction zones where one plate slides under another, in such a way that on average the total surface of the Earth remains more or less the same. It is also in these subduction zones, mainly in the Western Pacific and South America, that the most active volcanic chains are found, all situated at the front of the subducting plate, vertically from the place where this plate reaches a depth of about 100 km. Basically, these volcanoes are the consequence of processes linked to the partial melting and the dehydration of the oceanic crust, drawn down by the subducted plate when it reaches the temperature and pressure range favourable to these phenomena. Third, in some places, two plates slide past each other, as in California along the famous San Andreas fault.

15Most earthquakes occur along the edges of plates, thereby marking their boundaries. The devastating earthquakes of the last few years followed this pattern, highlighting plate boundaries. While they made a particularly strong impression, as they occurred in highly populated areas, every year an average of fifteen earthquakes with a magnitude greater than 7 are expected across the world.

16The idea of elastic rebound, formulated by Harry Fielding Reid in 1910, arose from the devastating earthquake of 1906 in San Francisco (magnitude 7.9)³. Using the geodesic triangulation measurements taken regularly in the region in the years preceding and following the earthquake, Harry Reid noticed that two points situated on either side of the fault, and at a certain distance from it, moved continually and gradually much faster than two points close to the fault, implying an intense deformation of the surrounding rock. As the rock’s maximum resistance to deformation was exceeded when the earthquake occurred, the rupture released the pressure within minutes by creating a clear displacement of its fault sides of 5 to 10 m, over a length of 450 km along the North Californian coast. Once the pressure is released, the cycle starts again until the next earthquake.

17In a few million years, the movement along an active fault can therefore reach tens of kilometres. This is clearly visible in the landscape in several places along the San Andreas fault in California. If one knows the average speed of the relative movement of the two plates and the seismic history of a specific point on a fault, one can deduce the displacement deficit along the fault. In certain locations, however, the plates gently slide past one another, without causing powerful earthquakes.

18The theories of plate tectonics and elastic rebound provide us with a general framework with which to anticipate the places where the most powerful earthquakes may occur. In order to identify these risk zones, seismologists have gathered data on past earthquakes and, where possible, the geometry of their rupture based on geological, geodesic and seismic data, to draw a world map of seismic potential. This map is focused mainly on the Pacific Ocean, where most of the subduction zones lie and where the most powerful earthquakes occur —those with a magnitude of 8 or more. In this case, the elastic rebound causes an upward displacement of the overriding plate. Since this generally takes place under the ocean, if the rupture reaches the surface of the seafloor, a considerable mass of water (several metres high over a length of hundreds of kilometres, or even a thousand, as in the case of Sumatra in 2004) is suddenly displaced, creating a situation likely to trigger a tsunami.

19By knowing the speeds of relative motion of the plates it is therefore possible to determine the approximate duration of the seismic cycle and anticipate the places where powerful earthquakes could occur, as well as their size.

20Nevertheless, the conditions leading to the generation of a powerful earthquake in a given place and time, and the precise way in which a seismic rupture is initiated and how it will evolve, currently remain fundamental questions to which we paradoxically have fewer answers now than before the Tōhoku (Japan) earthquake of 11 March 2011. In this specific case, Japanese seismologists had 1,300 years of history and precise data —thanks to GPS technology— on the speeds of convergence of the Pacific and North-American plates, to draw up a slip deficit map for the northern region of the island of Honshū and off the shore of the island of Hokkaidō. The 11 March 2011 earthquake did occur in a zone identified as about to “break”, but several factors took geophysicists by surprise: the largest part of the rupture occurred more to the East, closer to the Japan Trench, and therefore closer to the surface (a significant factor for the generation of tsunamis), and the magnitude of the earthquake was of 9.0, whereas no earthquake with a magnitude greater than 8.6 was expected in the region. There are other surprising aspects about the way the rupture of this earthquake occurred: in particular, displacements of over 50 metres of the two boundaries of the fault in relation to each other were recorded, something never seen before since the beginning of instrumental seismology. The 11 March earthquake also undermined a number of concepts that until then were considered as soundly established: the GPS network, though dense on Japan’s islands, did not detect the accumulation of deformation before the earthquake in terms of elastic rebound; how could the zone that ruptured during this earthquake have resisted 1,300 years of slip deficit accrued by 8 cm every year? Suffice it to say that the simple concept according to which certain parts of the boundaries of the plates are likely to break in very strong earthquakes and others not, needs reviewing. Nor can we rest assured any longer that we are able to produce precise estimates of the place and magnitude of the next earthquake, and that the only truly impossible parameter to predict is the exact date on which it will occur.

21The observation of phenomena at the boundaries of plates also has other surprises in store, such as the recent discovery of slow slip events (SSE)⁴: slip events without strong earthquakes observed in the North-West American subduction zone and in Japan by means of GPS measurements. These quasi-periodic events, recurring yearly in most cases, correspond to the backward motion of the plates, as though the subducting plate “retreated” a little before continuing to move forward. These SSE are accompanied by small earthquakes, so-called non-volcanic tremors, which illuminate the approximate zone affected by the slip. It is situated at the interface between the plates, a little below (deeper than) the blocked zone of the fault. Since they were discovered, SSEs and their non-volcanic tremors have been observed in many subduction zones, as well as on the San Andreas fault in central and Southern California. Owing to their quasi-periodicity, it has been possible to set up temporary seismic and geodesic monitoring networks to capture the successive events and compile a significant database on this newly discovered phenomenon. However the physical mechanism and its relation to the strong earthquakes expected in these regions are currently unknown.

22This subject deserves a lecture of its own. I will nevertheless set it aside to return to the more general question of the origin of the considerable forces governing plate tectonics. The absence of a precise answer to this question considerably delayed acceptance of Wegener’s continental drift. Yet no elements were missing —neither geographical (the shape of the coastlines) nor geological, nor data on the fauna and flora found on either side of the Atlantic— to demonstrate that in the geological past, over 200 million years ago, Africa and South America had once constituted a single continent. But why did they separate?

23Nowadays, we readily accept the fact that tectonic plate movements are due to convection currents in the Earth’s mantle, an effective mechanism for releasing our planet’s internal heat. A simple analogy would be that of a water container heated from below: the hottest and therefore least dense water rises from the bottom of the container towards the surface, where it cools down and becomes denser, before sinking back towards the bottom, thereby creating circulation organised into convection cells.

New tools to probe the Earth’s interior

24Studying the Earth’s interior is particularly challenging. Direct observation is impeded by technological difficulties, in particular the construction of drilling instruments capable of resisting the high temperatures (over 1,000 °C) found at relatively shallow depths: the deepest wells drilled until now do not reach further than 12 to 13 km which, to put it into context, barely represents two thousandths of the Earth’s radius (6,371 km). These boreholes have provided valuable information on the structure and composition of the shallow part of the Earth’s crust. But to find out more, indirect observation methods are needed to solve what is called an inverse problem: determining the structure inside a body by exclusively using data available on the surface surrounding it. Here you will recognise a familiar problem in the medical field, with all the non-invasive methods used to probe the human body (X-ray, MRI, ultrasound, etc.).

25Different types of observation on the surface of the Earth therefore need to be combined to probe its interior. This has given rise to several disciplines within the Earth sciences: ad hoc measurements and, more recently, satellite measurements (Magsat, Champ and Oersted) of the magnetic field and the gravity field, recording of the seismic waves generated by natural earthquakes, geochemical rock sampling, geodesic measurements and geological cartography. The interpretation of these observations needs to be combined with the theoretical and experimental modelling of currents of matter in the mantle and the core (i.e. geodynamics), and the modelling of the propagation of seismic waves in complex environments, as well as materials science and petrology, which provide information on the composition, the crystalline structure and the deformation properties of the rocks at the high pressures and temperatures of the Earth’s interior. To give you an idea, the pressure at the centre of the Earth is about 360 GPa, which corresponds to roughly 3.6 million times the atmospheric pressure, and the estimated temperature in the core is 5,000 +/– 1,500 degrees Celsius.

26Over the last few decades, each of these disciplines has made considerable progress in measurement techniques, which has led to a number of discoveries, in particular:

27GPS technology and radar interferometry (InSAR) currently allow us to detect relative movement across faults of the order of a millimetre per year, which has made it possible to determine tectonic movements on a global scale, to monitor the deformation of volcanoes so far as to be able to predict eruptions, and very recently to discover new phenomena, such as quasi-periodic tremors in subduction zones.

28With the construction of increasingly high performing presses, it is currently possible to reach pressures of over 150 GPa, in other words down to the Earth’s core, which, coupled with particle accelerators (cyclotrons), allows for an increasingly precise characterisation of the structure and physical properties of the rocks at the conditions found at the base of the mantle and down to the core.

29Considerable progress has also been made from a theoretical point of view, thanks to more and more powerful computers that allow for realistic conditions to be reached to model convection within the Earth’s mantle, in 3D, in a suitable spherical geometry, or for the precise simulation of seismic wave fields propagating in complex structures.

30The current challenge is to combine the information provided by these different disciplines, each of which contributes to an aspect of the whole but, on its own, is unable to resolve the fundamental problem of the internal dynamics of our planet. This problem consists in understanding how the convective engine of plate tectonics works —an engine which, over forty years after plate tectonics was universally accepted, is still the centre of lively debates.

31Since this is a dynamic system, it is not enough to know its current state. And to know its evolution over time, we must look at the past, from the time of the Earth’s formation about 4.56 billion years ago to now, knowing that the sampling of this past is essentially limited to the rocks situated close to the surface, and that only continents have preserved rocks which are over 200 million years old, precisely because of plate tectonics that endlessly recycles the oceanic plates.

Beyond the theories of plate tectonics and elastic rebound

32There are two main sources of heat inside the Earth: so-called residual heat, produced at the time of the Earth’s formation, which has not yet had the time to escape the core and the mantle; and heat released through the gradual disintegration of long-life radioactive elements (about 1 billion years or more), such as uranium (²³⁸U, ²³⁵U), thorium (²³²Th), and potassium (⁴⁰K), mainly concentrated in the granites of the continental crust.

33The contribution of crustal and mantle radioactivity to the heat released, and therefore to convection, is estimated at approximately half of the total energy released (~ 44-46 TW), but there is still a lot of uncertainty about the proportion of primordial heat transmitted from the Earth’s core towards the mantle: evaluations range from less than 10% to more than 20% of the total. The point of this debate, and more generally the question of the precise nature and form of mantle convection, is not purely academic: the heat flow at the boundary between the solid mantle and the liquid core is one of the main factors controlling the power of the dynamo that drives the internal magnetic field. A new technique currently under development may soon allow us to produce more precise estimates of the quantity of heat released by radioactive decay processes. This technique is based on the observation of geo-neutrinos, particles with no mass that are electrically neutral and are released through the beta decay of radioactive elements in terrestrial rocks. Special detectors have been built, mainly in Japan, and preliminary measurements seem to confirm that the heat released in the crust and mantle through radioactive decay corresponds to more or less half of the total heat released. This remains to be confirmed.

34More specifically, what are the physical and chemical mechanisms controlling plate movement? Are these mainly governed by the cooling at the surface, with the resulting gradual increase in plate density based on their age, which causes them to drop back down and sink into the mantle when they are sufficiently heavy, taking with them the displaced viscous rock? This is what we expect to see if the internal heating (i.e. the radioactivity) is prevailing. In that case, descending currents would be located where there are subducting plates; conversely, ascending currents would be more diffuse. Or does a hot lower boundary-layer play an active role by generating hot ascending currents that are more localised and relatively fixed in time: the mantle plumes? This boundary-layer is preferably situated at the core-mantle boundary or, alternatively, at the boundary between the upper and lower mantles, around 670 km deep. The mantle plumes concept was proposed by Jason Morgan in 1972⁵. It draws on the observations made through experiments (and numerical models) in the case of a fluid heated from below: the ascending currents are organised in narrow columns with larger heads, the plumes. This set of plumes is understood to be relatively fixed in relation to the mantle, inducing linear chains of progressively older volcanoes, as observed in several places. The manifestation of the plumes at the surface would then be the presence of hot spots: volcanoes, like Hawaii, in the middle of the Pacific Plate, which have no simple explanation in terms of plate tectonics, but could be explained if we posited that the plumes are fixed over time. For example, as the Pacific plate moves west over its plume, new volcanoes form, which could explain the progressively greater age towards the west of Hawaii’s volcanic chain, as well as other linear volcanic chains in the middle of the Pacific. This theory is however challenged by certain geophysicists who think that the hot spots are the manifestation of phenomena close to the Earth’s surface, and linked to the internal deformation of the lithospheric plates themselves. It is therefore important to try and detect any plumes there may be, along with the depth at which they originate in the mantle, in order to resolve this controversy and better determine the passive or active role of hot ascending currents in the global mantle circulation.

35Another fundamental question concerns the degree of coupling between mantle convection and surface tectonics: who controls what? The lithospheric plates move above a zone of the mantle called the asthenosphere (from the Greek asthenes, “weak, without resistance”), because of its low viscosity and poor resistance to long-term deformation, allowing the plates to slide more or less regularly at a pace of a few centimetres a year. But what is the relative strength of the forces pulling the plate in a subduction zone, of those pushing it when the new crust forms in a mid-ocean ridge, and finally of the resistance (or push?) at the limit between the lithosphere and the asthenosphere?

36Does convection affect the entire mantle, or is it limited to the upper mantle, which is demarcated from the lower mantle —more viscous as the pressure increases with depth—, through a jump in the seismic and mineral structure at a depth of 660 km?

37In addition to the proportion of the different types of internal heating, mantle circulation is complicated by the fact that the crystalline structure of the rocks and their rheology (that is, their deformation properties) change according to the pressure and therefore depth of the mantle. And finally, there is the added complexity of heterogeneous chemical composition, which consists primarily of two end-members. The first is inherited from the time of the Earth’s formation, and has not yet been sufficiently stirred to be homogeneously distributed throughout the mantle. The second is continually created at ocean ridges when new crust and lithosphere are formed, and, at a greater depth, in subduction zones, through partial melting processes that result in preferential separation of certain chemical elements into the products of melting, and others into the residual solid matrix.

38Drawing primarily on the study of the respective isotopic content of the volcanic lava sampled in the mid-ocean ridges’ volcanoes and in hot spot volcanoes, geochemists have shown that the mantle is far from being a uniform mix of rocks. They have revealed the presence of at least two “reservoirs” in the Earth’s mantle: the one presumably corresponds to the products of the partial melting that takes place during the formation of new crust at mid-ocean ridges, and the other to the “primordial” material, of which few samples have yet been taken, and which is situated somewhere at the bottom of the mantle, since this mantle has not yet had the time —or has not been able— to become homogeneous. Several extreme models of mantle circulation have therefore been developed, which need to be checked against the observations of seismic imaging mainly. These range from double-layer convection, where the upper mantle is almost completely isolated from the lower mantle, to single-layer convection, encompassing the full depth of the mantle, with a whole series of models in-between, most likely closer to reality.

39All these questions, as well as the question of the evolution over time of the Earth’s internal dynamics since its formation, currently motivate active research as well as intense debates in virtually all disciplines of the Earth sciences. In this respect, we know that the tectonics of the Archean Age —over two billion years ago— must have been somewhat different from modern day tectonics. A more dynamic convection, due to a more significant heat release, may have given rise to the formation of the oldest parts of the continents: these are lighter and float above the oceans; their particular structure and composition differ from those currently produced at mid-ocean ridges. But this is also open to debate.

What seismic imagery has contributed to our knowledge of the structure and dynamics of the Earth’s mantle

40I would now like to offer a more specific description of the progress made in the field of seismic imaging on a global scale, my specific research domain. In the interest of time, I am only going to talk about imaging of the mantle, and not that of the core or of the solid inner core. Seismic imaging is unquestionably the most effective method we have to “see” inside the Earth. It uses elastic waves emitted by earthquakes across the globe and recorded by highly sensitive wave sensors, or seismometers. These instruments can detect shifts to the order of a micron, corresponding to earthquakes of a magnitude greater than five, at distances as great as the antipode of the epicentre. Elastic waves illuminate the interior of the globe: on the one hand, like light, they are reflected or refracted on the obstacles encountered, and, additionally, their speed and amplitude are altered by the physical state of the matter that they encounter. By combining the recordings of many earthquakes in numerous stations throughout the world, we can use “tomographic” methods similar to those used in medicine to obtain 3D images of the Earth’s interior. However, unlike medical methods, in tomography we do not control the distribution of the sources of vibrations (situated primarily along the edges of the tectonic plates), and the distribution of stations is still very limited. Apart from logistic and political considerations, setting up high-quality seismic stations on the ocean floor remains a technical and financial challenge as two thirds of the Earth’s surface is covered by oceans and there are relatively few islands. Sophisticated techniques are therefore needed to extract as much information as possible from seismic recordings. This is currently one of the main contributions of cutting-edge research. Fortunately, seismic recordings are a rich source of information. First, there are two types of seismic waves: longitudinal waves (known as “P waves”) and transverse waves (known as “S waves”). As the latter do not travel in liquids, we have known with certainty, since the work of Oldham in 1906⁶ and those who followed in his footsteps in the first half of the 20th century, by observing seismic waves at different distances from the earthquake source, that the external part of the Earth’s core is liquid.

41More subtle variations in observed seismic waveforms have successively revealed the existence of a solid inner core⁷ and structural discontinuities present on a global scale in the “upper” mantle, at depths of 400 and 660 km. In particular, the comparison with materials physics has enabled scientists to establish that these discontinuities correspond to changes of the crystalline structure of minerals under the effect of pressure —called phase changes—, into increasingly compact structures. The discontinuity situated at 660 km, which marks the limit between the upper and lower mantle, is probably also a mechanical barrier that makes the transfer of matter between the two parts of the mantle difficult but not impossible.

42Seismic waves are also rich in frequency content. The waves with longer periods (the lowest vibration has a period of about 54 mn) travel in the form of surface waves that circle the Earth many times. Interference between them produces stationary waves, the Earth’s normal modes, the spectrum of which (i.e. the ensemble of their frequencies) is characteristic of the internal structure. Thanks to all these tools we have had a precise idea of the Earth’s spherically symmetrical structure for a long time (the first reliable models were produced in the late 1940s), that is, an onion-like structure consisting of successive layers: the crust and solid mantle, composed primarily of silicates, then the liquid core composed of an iron and nickel alloy and, finally, the solid inner core composed of even purer crystalline iron. In the 1940s, seismologist Keith Bullen⁸ divided the Earth into layers, to which he gave names corresponding to the first letters of the alphabet. Only one remains, the D″, which includes the deepest 200-300 km of the mantle, and the properties of which differ from those of the entire lower mantle above it. Note that seismological data inform us on the elastic structure, that is, on the variations of three parameters with depth: the respective speeds at which the transverse and longitudinal waves propagate, as well as their density. To convert these into fundamental physical parameters, namely temperature and chemical composition, they need to be compared with laboratory measurements of the elastic properties of rocks according to their composition, taken under the same temperature and pressure conditions. In particular, since it has been determined that the seismic discontinuities at a depth of 400 km and 660 km correspond to phase transitions in silicates, due to increasing pressure with depth, the average temperature at these depths is known fairly precisely.

43This research has yielded important knowledge on the radial structure of the Earth for over half a century. However surprises still await us. For instance, with access to increasingly high pressures, a new phase was recently discovered that could be stable just above the core-mantle boundary. This phase has been named post-perovskite, perovskite being the predominant mineral —composed of magnesium and iron silicate— in the lower mantle. This transition could be occurring in the D″ zone, marked in numerous places around the globe by a discontinuity in elastic velocities, as seismologists have observed.

44Another discovery, to which our French colleagues James Badro and Guillaume Fiquet contributed⁹, is that of an electronic transition of iron (high spin to low spin) at pressures and temperatures corresponding to those found in the middle of the lower mantle (at a depth of about 1500 km). This transition modifies the chemical and physical properties of the rocks, which could have consequences for the interpretation of seismic data at these depths, and for the dynamics of this part of the mantle. The seismic manifestation of this transition has however not yet been detected.

45The radial structure of the Earth provides us with a static image of the Earth’s interior. Our current goal is to map the expression of the internal dynamics, which manifests itself through so-called lateral variations in the seismic structure. For instance, by simplifying or ignoring the variations in composition, the hottest and therefore least dense regions, which correspond to the ascending currents, manifest themselves through slower seismic speeds than the cooler and denser regions, which correspond to descending currents. These lateral variations are imaged using seismic tomography, which follows a principle similar to the techniques used in medicine: waves (seismic waves in this case) generated at the surface of a body by sources of vibration (here, earthquakes) are sent through this same body, and are also recorded at the surface. Their travel time or amplitude can be measured and compared with the times or amplitudes predicted by a reference model, generally a spherically symmetric average Earth model, that is, a simplified model in which the physical parameters vary only with depth.

46Let us now look at what we have learnt from tomographic images. One could a priori expect such images to reflect a very simple convective system, where the ascending currents are found under mid-ocean ridges, and the descending currents are found at subduction zones, where the tectonic plates slide back into the mantle. In fact, seismic tomography, thanks to which we now have good constraints on long-wavelength characteristics (~ 2000 km) of the elastic structure of the Earth’s mantle, reveals a more complex situation that cannot yet be very clearly interpreted.

47In what follows, I will first present a series of cross-sections, each representing the lateral variations of the structure at a given depth. There will also be a few cross-sections in the vertical plane. The colours represent the zones of seismic speed, which are slower (in red) or faster (in blue) than average (fig. 2)¹⁰. The simplest interpretation to remember is that, to first order, the red zones correspond to higher temperatures (by several hundred degrees) and the blue zones to cooler temperatures.

Figure 2. Horizontal cross-sections in a recent global seismic tomography model at several depths in the mantle

48Thus a cross-section from the upper mantle, for example at a depth of 150km, confirms what plate tectonics predict: one can follow the system of mid-ocean ridges throughout the Pacific, Indian and Atlantic Oceans, which correspond to the hottest temperatures. In the Pacific Ocean, the temperature decreases as we move away from the East-Pacific Ridge, in line with the gradual cooling of the Pacific Plate. Also note the very cold regions situated under the oldest parts of the continents, known as cratons, in Canada, Scandinavia, Australia, Antarctica and Siberia, where the lithosphere —the rigid part of the plate— is twice as thick as the global average (~ 200 km). If it was only cooler, and therefore denser, this lithosphere would tend to sink into the mantle; however, since on the contrary it is stable, and floats without causing an anomalous signal in gravity measurements, we can deduce that its composition must be distinct from that of oceans at the same depth. It is very likely iron-depleted, which suggests a different mode of formation to that of the current oceanic lithosphere. At a greater depth, around 250-300 km, there is a noticeable change in the distribution of elastic speeds: we lose the signal of the ridges and old continents, as well as the strong correspondence with surface tectonics. At a depth of 300 km, the zones with a temperature higher than average have moved towards the centre of the Pacific and under Africa, lateral variations are much smaller (to the order of +/– 1-2% compared with +/– 8-10% at a depth of 150 km) and the continental signal has disappeared. At a depth of 600 km, just above the limit between the upper and lower mantles, the dominant signal is that of cold zones with an elongated shape, which correspond to the main subduction zones, at the western and south-eastern edges of the Pacific Ocean. These are more clearly visible in vertical cross-sections in regional models specifically built to image them, which take advantage of the abundant seismicity associated with subduction zones and of the numerous seismic stations at the edges of continents.

49We can see here that the answer to the question “double-layer or single-layer convection” is not straightforward: subducted plates behave differently around the world. Some, like in Central America, seem to extend continuously almost down to the base of the mantle, while others, as in Japan, seem to flatten out at the base of the upper mantle, where they meet strong resistance to penetration into the lower mantle. Here is one possible scenario: the plates first accumulate at the base of the upper mantle and, when they become heavy enough to overcome the viscous resistance, they suddenly penetrate into the lower mantle, in isolated “flushing” episodes (on a geological timescale) that perhaps also correspond to episodes of reorganisation of the plates on the surface.

50But let us get back to the horizontal cross-sections: for most of the lower mantle, current images do not reveal any grand structures, apart from those of the sliding plates in Asia and South America. On the other hand, as we get closer to the base of the mantle, there is a striking reorganisation of the structure and, starting from about 500 km above the core-mantle boundary, we begin to see two large equatorial and more or less antipodal zones of slow seismic velocity. These zones are respectively centred under the Pacific and Africa, and are surrounded by a ring of faster velocities, situated vertically from the zones where the subducted plates slide down (fig. 2). Although definite proof is still missing, these zones in blue probably correspond to the ultimate accumulation of subducted plates, which have “fallen” into the lower mantle —a graveyard of plates if you will. As for the two zones in red, they are often called superplumes, as they remind us of the presence of hot, less dense matter, which presumably rises towards the surface. In particular, they must correspond to the return flow of mantle convection.

51This however represents an oversimplified view. In reality, the debate is currently situated on two levels. First, the possibility of essentially double-layer convection is not excluded: one could conceive of a coupling between the convection in the lower mantle and the upper mantle at the level of these superplumes, as they may simply heat the base of the upper mantle by inducing hot currents above the hot currents of the lower mantle, without any significant transfer of matter. Second, we are now virtually certain that these superplumes are not only hotter, but also have a different chemical composition from that of the surrounding regions, and are probably denser. We derive this information from other seismological studies which unambiguously show that the external boundaries of the superplumes are very clear: we can see it in the way that the seismic wavefield, which interacts with the superplumes close to the core-mantle boundary, is noticeably deformed. We can therefore think of these two superplumes as two gigantic pillars, each 5,000 km in diameter and a few hundred kilometres tall, comprised of matter most likely containing a higher than average proportion of iron, and which rest on the core-mantle boundary and affect the form of mantle circulation. What is their true role in mantle convection? When did they come about? Are they the outcome of the sinking of subducted plates, which, when they reach the bottom of the mantle, sweep the denser layer of matter found there by forming these piles? Or do they stem from very ancient processes, perhaps dating back to the formation of the Earth? The idea that these superplumes may have existed for a very long time is corroborated by the remarkable observation that their position corresponds to an energetically stable configuration of the moments of inertia of the Earth as it rotates around its axis. To see this, we simply consider the part of the structure involved in the calculation of moments of inertia, that is, the longest wavelengths, what we call the degree 2 of the structure. As shown in this representation —all tomographic models agree very precisely on this point—, the axis of the structure of the two superplumes is in the plane of the equator, whereas the axis of rotation of the Earth, which is orthogonal, is inside the blue ring of faster seismic velocities separating them, in the plane of symmetry of this structure¹¹ (fig. 3). Any position of the axis of rotation in this plane corresponds to a stable configuration of the moments of inertia. Paleomagnetic reconstructions show that polar motion has been within this zone, at least for the last 250 or even 500 million years. Note that the position of the two superplumes corresponds very well to the geographic distribution of the main hot spots at the surface. The famous mantle plumes culminating in the hot spots could therefore have their roots in the superplumes. Other seismological studies have also shown a correlation between the position of the hot spots and the zones, on the edges of the superplumes, where very localised zones with extreme seismic properties are observed, suggesting the presence of partial melting, the so-called ultra low velocity zones. The problem is that the plumes have until now been very difficult to reliably detect with seismic imaging inside the mantle, due to their presumably small diameter —possibly less than 200 km. This lack of resolution is one of the major challenges of global seismic imaging: do the plumes rise from the base of the lower mantle, or from the thermal boundary-layer at the base of the upper mantle? Or, at the other extreme of the debate surrounding these plumes, are they simply shallow perturbations corresponding to the upwelling of partially melted matter from the base of the lithosphere, in the cracks caused by the strong tensions exerted on it?

Figure 3. Degree 2 of the structure in shear velocities in the deepest part of the mantle (at a depth of 2,800 km), seen in three recent tomographic models built by different research groups. S362ANI: Kustowski et al. (2008); SAW24B16: Mégnin and Romanowicz (2000); S20RTS: Ritsema et al. (1999).

Prospects for the future of global seismology

52The answer to these questions will have to await the development of finer tomographic images. This takes me to the last part of my presentation, where I briefly describe the prospects for the future of this discipline. Progress in global seismology over the last few years has gone hand in hand with improvements in the quality of seismic sensors, the possibility of digitally recording ever greater quantities of data, theoretical progress in the simulation of the seismic wavefield in complex 3D media, and the development of powerful computers for the forward calculations and the inversion of large matrices. Present challenges now revolve around characterising the physical and chemical nature of the transitions corresponding to the limits between the large structures in the deep mantle, and detecting and characterising small-scale structures. This involves being able to extract more information from seismic records than has been the case until now. Neither the remnants of subducted plates at the bottom of the mantle nor the plumes —if they exist— can be resolved with precision using traditional methods, which draw only on information contained in the “first” arriving waves, those that have propagated over the shortest paths, following the laws of ray theory. We now also need to take into account those waves that interact with low-amplitude diffracting objects of varying sizes. In practice, this requires working in a large frequency band, at short spatial wavelengths, and using the information contained in both the travel times and amplitudes of all the waves recorded, perhaps also applying signal amplification techniques.

53The limited distribution of earthquakes and seismic stations around the world is a significant challenge. Ideally, we would like to be able to uniformly sample the volume of the Earth. However, unlike other disciplines that use imaging methods, like medicine or oil exploration, we do not control the distribution of the sources, nor completely that of recording stations. Several promising pathways currently exist to overcome these difficulties.

54New horizons have recently come into view with progress in the numerical calculation of wavefields and massive data archiving. Powerful numerical schemes now exist to calculate the seismic wavefield in structures of any complexity, for example the spectral element method, well suited to the Earth’s spherical geometry. These methods can be used in various ways, either to directly model the observed records, or in the context of an inverse problem. Though still cumbersome, these calculations are promising for the development of the next generation of global tomographic models. Anisotropy and anelastic dissipation, which also influence wave propagation, can now be characterised better and provide complementary information about directions of flow, lateral temperature variations and the presence of partial melting. At the highest observable seismic frequencies, the development of dense networks of permanent seismic stations, such as the hi-net network in Japan, or of temporary ones, as with the USArray programme of Earthscope, are stimulating the development of techniques that are starting to reduce the differences between the tools of global seismology and those of applied geophysics.

55With the information contained in the direct and diffuse wave fields, very precise images of the subducted plates inside the mantle can be built. The results of seismic imaging under Japan can now be used to follow the evolution in the mantle of water driven into the subduction zones. The data from the temporary American network USArray, gradually covering the United States from west to east with 400 stations installed with an average spacing of 70 km, afford sufficient spatial sampling to study the fine structure of the lithosphere under the North American continent, as well as the details of the Pacific superplume at the base of the mantle. Finally, seismologists can offer precise values of the size and depth of heterogeneities and of the contrast in elastic parameters that they represent. These values can be compared with data from materials science to determine lateral variations in composition and temperature. For example, in the case of the recent discovery of post-perovskite, which appears to have the same domain of stability as the D″ layer of the mantle, materials scientists and geodynamists may converge with seismologists to confirm its presence in the deep mantle and to assess its consequences on mantle dynamics.

56The uneven distribution of sources of seismic energy presents a significant problem. One rapidly-evolving technique bypasses this constraint by exploiting the large quantity of wideband seismic data that have been continuously recorded for the last 30 years at many stations: the continuous background noise created by the ocean and the atmosphere can be used to build tomographic images using noise correlation methods. The power of this approach has been demonstrated for the study of the structure of the crust, which can draw on the significant presence of energy in the microseismic frequency band (1-15s). The extension of this method to longer periods offers interesting prospects for imaging of the upper mantle, at least down to depths corresponding to the base of the continental lithosphere.

57These techniques perform well in the continental regions, where there is a strong density of stations. This is not the case for the oceans, where the acquisition of seismic data is limited to a small number of islands. Yet there is no shortage of geodynamic problems to solve: for instance, we lack sufficiently precise knowledge of ocean basins’ deep structure and seismic anisotropy to identify several possible models of oceanic plate cooling. The deep mantle and core are still poorly sampled due to the lack of stations in the oceans. Despite efforts for over 25 years to instrument the seafloor, few long-term stations have as yet been set up on the seafloor. An internationally coordinated programme is needed to systematically cover the oceans with large-scale (1,000 km by 1,000 km) broadband seismic networks. Such networks should remain in place for at least one or two years to record a sufficient number of earthquakes and allow for the structure under the oceans, from the surface to the core, to gradually be illuminated.

58Finally, as the images produced by seismologists become more and more precise, it is important to make use of the complementary information provided by the different disciplines of the Earth sciences to solve the difficult inverse problem that the interior of the Earth and planets presents.

59Even if we agree that the images provided will one day offer extreme precision and even if, with anisotropic imaging, they inform us on past or present flow directions¹³, they produce only an instantaneous and therefore essentially static view of the Earth’s interior. This view needs to be completed using other methods, namely those of geodynamics, to access the dynamics and thermal evolution of the planet. Furthermore, the images obtained are of the distribution of the elastic or anelastic properties of the materials in the Earth’s interior, whereas we are interested in the distribution of temperatures and chemical compositions, which determine the distribution of density. These allow us to characterise the morphology of the convective cells, that is, the spatial distribution of the flow within the mantle. There is an ongoing effort to integrate the different disciplines contributing to knowledge of the Earth’s structure and dynamics, for example as part of the CIDER (Cooperative Institute for Dynamic Earth Research) programme.

60To conclude, I have only discussed here certain global aspects of the deep Earth, leaving others aside, particularly concerning core structure and dynamics. I hope, however, that I have convinced you that the multidisciplinary study of the Earth’s interior is a fascinating and highly active field, which is rapidly evolving using state-of-the-art technology, and concerns us all in so far as it fundamentally relates to the habitability of our planet.