Mountain roots and the survival of cratons

Abstract

What controls the deformation of the continents, the survival of ancient cratons and the roots of mountains? James Jackson explains in his Harold Jeffreys Lecture, 12 November 2004.

This is a detective story in the Earth sciences. From the earliest days of plate tectonics it was known that the continents do not deform in the same way as the oceans. This is evident from a map of earthquake epicentres (figure 1), showing the narrow bands of earthquakes in the Indian ocean, contrasting with the epicentres distributed over the broad mountainous regions of the Middle East and Central Asia. Earthquakes occur when faults move, so a map of earthquakes is a map of active deformation. Plate boundaries in the oceans are essentially single faults, defined by joining the epicentre dots on a map, whereas the very concept of plate boundaries on the continents is often unhelpful: it is meaningless to ask what plates Lhasa or Athens are on, since Tibet and Greece are both parts of wide deforming regions, and not parts of any rigid plate.

Figure 1 : . Earthquake epicentres (red) in Asia and the Middle East are dispersed over very wide mountainous regions, and contrast with the narrow bands of earthquakes offshore in the central Indian and Atlantic oceans. This contrast between the deformation of the continents and oceans is fundamental. Earthquakes in central Asia result from the collision of India with the Eurasian continent, but note that India, an ancient continental craton, is relatively undeformed.

The ocean–continent contrast is not unexpected. The plates make up the lithosphere, the outer strong layer of the Earth, typically about 100 km thick and consisting of the crust and part of the underlying mantle. In the oceans the crustal part is a fairly uniform 7 km thick, whereas on the continents it is typically 30–80 km. The crust is less dense than the mantle, so a plate with thick continental crust is more buoyant, preventing it from sliding beneath another plate at collision zones. Instead it crumples to make mountains. Attempts to understand this process require a knowledge of the mechanical properties of continental lithosphere, and it is recent developments in this area that are the subject of this article.

Earthquakes and collisions

As so often occurs in Earth sciences, focus on one question inadvertently illuminates another. The earthquakes in figure 1 result from the ongoing collision between India and Asia, whose geological boundary lies in the Himalaya. Yet, whereas Asia has crumpled up as far north as Mongolia, ∼3000 km from the geological contact, India is virtually undeformed. India is part of the Jurassic Gondwanaland supercontinent that fragmented to make Africa, South America, Australia and Antarctica. The interiors of all these continents are ancient (usually older than 3 billion years), flat and have remained undeformed for a very long time: they are called shields or cratons. They have a history of colliding with other cratons to form mountain belts between them, which later split apart again along the same sutures, leaving the cratons themselves intact and undeformed. The ancient cratons have an ability to survive that has long puzzled geologists.

For many years people have used the depth distribution of earthquakes and the gravity anomalies associated with topographic loads as indicators of lithosphere strength, by which is meant the ability to sustain elastic stresses over geological timescales with negligible flow.

In the case of earthquakes, a temperature-dependent change from shallow, friction-dominated slip on faults to deeper, aseismic creep processes is expected, and in most continental regions, earthquakes are indeed restricted to the upper half of the crust. But variations are seen, and in particular, in some of the old cratons, such as north India and parts of East Africa, earthquakes occur throughout the thickness of the crust (figure 2). There is little evidence for earthquakes in the mantle beneath the continents, in spite of there being many earthquakes in the oceanic mantle, beneath the much thinner oceanic crust. (I am not referring here to the very deep earthquakes in cold oceanic lithosphere that is transported back into the mantle at oceanic trenches, but to earthquakes at depths up to ∼40 km within the stable ocean basins.)

Figure 2 : . A north–south cross-section through the Himalaya at 90°E. Red dots are earthquakes within the crystalline crust of the Indian shield (blue), and occur down to the crust–mantle interface (the “Moho”: thick blue line). White dots are shallower earthquakes in the Tibetan crust (orange) or the upper crust of India (green), which has been deformed, thickened and scraped off the lower Indian crust (black arrow) in the Himalaya. The positions of the boundaries between the green, blue and orange areas are conjectural at depth, but the Indus suture is the geological boundary between Tibetan and Indian rocks at the surface. The vertical line at 920 km distance is the northern limit of the Indian mantle lithosphere at depth, from seismic anisotropy. (Adapted from Jackson et al. 2004)

The wavelength on which the lithosphere bends, and creates gravity anomalies to support topographic loads, can be used to estimate the thickness of the strong elastic layer providing the support, in the same way that a thick plank bends on a longer wavelength than a thin one. Recent reviews of this rather technical and controversial subject (McKenzie and Fairhead 1997, Maggi et al. 2000, McKenzie 2003) found that the effective elastic thickness (T_e) determined from gravity and topography tracked the thickness of the layer in which earthquakes occur (T_s). Thus larger values of elastic thickness were found where earthquakes occur throughout the crust. Although T_e is not always well resolved by the data, they found, in general, that T_e < T_s and they found nowhere where the data required that T_e > T_s. As they pointed out, the simplest interpretation of these results is that the long-term strength of the continental lithosphere resides in the layer that generates the earthquakes; which is either the upper crust or the whole crust, but does not include the mantle.

These results immediately contrasted the continents with the oceans, where the mantle both generates earthquakes and contributes to long-term elastic strength, but where the crust is much thinner. They also focused attention on the ancient cratons, and in particular north India, where the values of both T_e and T_s suggested that the Indian shield was unusually strong (figure 2).

Crustal thickness

As patterns in the earthquake depths and gravity anomalies became clearer, so too did our knowledge of crustal thickness variations. Modern, broadband digital seismometers can be used to determine the crustal thickness by detecting converted longitudinal-to-transverse waves generated by the crust–mantle interface. These waves arrive a little later than direct longitudinal waves from distant earthquakes and the delay depends on the crustal thickness. Widespread use of this technique on the continents produced two important results for this story.

Firstly, it was found that the thickness of the crustal root beneath southern Tibet reaches 80–90 km (figure 2): significantly thicker than we previously thought. It was always known that the crust was thick beneath continental mountains: the crust floats on the mantle like ice on water (though the mantle is solid, not liquid, and deforms by creep), and high elevations are supported by deep, buoyant roots. But a thickness of 80–90 km poses other problems: at those depths the minerals in typical continental rocks should transform to a different, much denser, mineral assemblage called eclogite (figure 3). They cannot have done so in the Himalayan root, or its greater density would mean both that Tibet would be at much lower elevation and that its higher velocity would make the crust–mantle interface there undetectable. The crust must somehow have remained in the less-dense mineral assemblage called granulite; but why? An extra surprise was that some unusually deep earthquakes, 80–90 km beneath southern Tibet, were now seen to be so close to the crust–mantle boundary that we could not really distinguish which side they were on.

Figure 3 : . This shows the expected mineral assemblages (white boxes) as a function of depth and temperature in the crust. The deepest part of the Himalayan root zone (50–90 km) should be in “eclogite” conditions since it will have followed a downward path similar to the blue line (“subduction zone geotherm”). But if this were true, the rocks would be very dense (∼3.5 Mgm m⁻³, compared with 3.2 Mgm m⁻³ in the mantle), and southern Tibet would not be at 5 km; the seismic velocities at depth would also be too high to see the Moho (crust–mantle interface). The crust must have remained metastable in the much less dense (∼2.8 Mgm m⁻³) “granulite” assemblage, characteristic of the ancient Indian shield where it is exposed at the surface. The yellow circle is the approximate temperature–pressure position recorded by the host granulite rocks in Norway (figure 5), even though the mineralogy in the “fossil earthquake” melts shows they reached the conditions of the green circle. The host granulite was again metastable, at pressures and temperatures beyond its nominal stability limit. (See Jackson et al. 2004)

Secondly, there were more signs that the cratons were odd. For example, in Finland, part of the ancient Scandinavian craton, the crust can reach 65 km thickness, the same as that in the Alps, even though the country is essentially at sea level. Younger continental regions at sea level, such as the United Kingdom, usually have a crustal thickness around 30 km. Something must be unusual about the mantle part of the lithosphere beneath the crust in the cratons.

Lithosphere thickness

At this point, bearing in mind the puzzle that gravity and earthquake evidence appeared to show the continental mantle was weak, whereas the oceanic mantle appeared strong, we looked for independent evidence of lithosphere thickness and temperature. Such evidence exists in two forms: the geochemistry of mantle nodules brought to the surface in volcanic eruptions, and the imaging of velocity structure in the Earth by seismic tomography.

Figure 4a shows an estimate of the temperature profile beneath the NW part of the ancient Canadian craton. It uses pressure and temperature estimates from the chemical compositions of mantle nodules, to which are fitted a steady-state geotherm that has to connect with the convecting interior beneath the plates, whose temperature is known. An important feature of this profile is that the temperature gradient is much steeper in the crust, where radioactive isotopes of K, U and Th are concentrated, than in the mantle lithosphere, where there is no significant internal heat production. The other important effect, incorporated here for the first time, is that conductivity in the mantle is a strong function of temperature, changing by a factor of about two over the thickness of the lithosphere, so the mantle gradient is concave upwards.

Figure 4 (a) right : . A temperature profile through the lithosphere (plate) based on pressure–temperature estimates from the geochemistry of mantle nodules at the Jericho mine, in the northern Canadian craton (green circles). The lithosphere consists of the crust (lightest green) and part of the mantle, which in turn includes a rigid mechanical boundary layer (MBL) and a lower thermal boundary layer (TBL) that connects with a convecting interior below the plate. (Figure from Dan McKenzie)

(b) above: Seismic velocity perturbation relative to a standard Earth reference model at a depth of 125 km beneath North America. The high velocity of the Canadian shield, whose geological boundary at the surface is marked by a yellow line, is evident in the blue colours and is related to its lower temperature. The Jericho mine, in figure 4a, is the yellow circle.

(c) above: A vertical seismic velocity section along the black line in 4b through the Jericho mine. The high velocity lid of the Canadian shield has an abrupt base at about 220 km (red line), in good agreement with the base of the lithosphere estimated in 4(a) (roughly halfway through the TBL). (Figures 4b and 4c from Keith Priestley)

The profile in figure 4a predicts the heat flow at the surface, and is compatible with observations. It also suggests a lithosphere thickness of 220 km, which can be compared with the velocity structure from seismic tomography. The tomographic image in figure 4b shows the high velocity lid of the cold Canadian craton extending to precisely the depth suggested by the mantle nodule analysis. Thus, for the first time, the seismology, geochemistry and heat-flow measurements are all consistent and indicate that the lithosphere thickness is about 220 km: roughly twice that of young continental regions and the ocean basins. Both the nodules and the tomography show that this is a general result: the cratons have significantly thicker lithosphere, sometimes with an unusually thick crust as well.

But the ability to construct accurate profiles like that in figure 4a also tells us something else: the temperature at the crust–mantle interface beneath the cratons. There are no earthquakes beneath this part of Canada, but where earthquakes occur in other cratons, such as Siberia and east Africa, it appears that the temperature at the interface is typically ∼600 °C. This turns out also to be the temperature at which we estimate the mantle becomes aseismic in the oceans (allowing for the temperature-dependent conductivity). Thus the conundrum as to why earthquakes occur in the oceanic mantle but not the continental may have a simple solution: the mantle generates earthquakes only when it is colder than ∼600 °C, which is common in the oceans but very rare on the continents. The neglect of temperature-dependent conductivity had previously led us to overestimate the temperatures in the oceanic mantle and underestimate them in the continents.

Fossil earthquakes in Norway

A remaining puzzle is why the cratons, such as north India, are both unusually strong and produce earthquakes in the lower crust. There are two obvious possibilities: that they are unusually cold, or that they are dehydrated (even very small amounts of water reduce creep strength dramatically). A clue to this puzzle came from an unexpected source: evidence of earthquakes about 400 million years old in Norway.

In the early 1990s, Håkon Austrheim and colleagues from Oslo described some friction-generated melts from the exposed ancient root zone of the Norwegian Caledonian mountains (the same mountain system that formed the Appalachians and Scottish Highlands). These melts formed in earthquakes as the result of heating on fault-slip surfaces at high confining pressures (figure 5), and were immediately quenched to form glass. What is unusual about these melts is that their mineralogy, which in electron micrographs shows the original dendritic and skeletal forms characteristic of rapid growth in quenching, is that of the eclogite assemblage (figure 3). These were fossil earthquakes at depths of at least 60–70 km, yet the host (unmelted) rock is still in the granulite assemblage. This indicates that the granulite was metastable and could only form the stable mineral assemblage when it melted.

Figure 5 (a) : . Fossil earthquake in Norway. The black band with flame-like injections coming off it is frictional melt (quenched to glass) from an earthquake about 400 Ma ago. The mineralogy of the quenched melt is that of the eclogite assemblage (figure 3) indicating a depth of at least 50 km. The host rock is granulite.

(b): Granulite–ecologite mechanical contrasts in Norway. Scale bar is 10 cm. The banded granulite (top) has behaved essentially rigidly (the banding itself is inherited from earlier deformation), while the eclogite (bottom) has flowed in a ductile shear zone. The transformation from granulite to eclogite requires water as a catalyst; without water, the granulite remains metastable. (See Jackson et al. 2004. Figures from Håkon Austrheim [Austrheim and Boundy 1994, Bjornerud et al. 2002])

Austrheim's group was able to show that the key to this process is water: very small amounts of hydrous mineral phases are seen in the eclogite melts, but the host granulite is completely anhydrous. Infiltration of water along cracks led to more pervasive eclogite formation, accompanied by a dramatic loss of strength: the eclogite is deformed by ductile flow, whereas the granulite is essentially undeformed except where offset by brittle slip on melt-generating surfaces.

Two things stand out from Austrheim's remarkable work. The first is the all-important effect of water, whose catalytic effect is that of an on–off switch: with no water the granulite–eclogite transformation simply does not occur, and the granulite remains metastable. This is not that surprising, as the transformation involves a wholesale reorganization of both chemistry and mineral structure and is known to be very difficult to achieve in the solid state unaided by a catalyst. The second is the association between metastability and mechanical strength. The metastable granulite retains its mechanical integrity and (inherited) internal structure until it transforms to eclogite and becomes weak.

Survival of cratons and roots

Thus we return to India. Whereas being either cold or anhydrous would account for the strength of Indian craton and its lower crust, penetrating beneath southern Tibet to form the massive Himalayan crustal root, the message from Norway is clear: only being dry will allow the Tibetan root to survive as metastable granulite. Thus it is likely that the unusual strength of the cratons derives from their dry lower crust: granulite is an anhydrous assemblage left behind after melting in the past extracted granite (and with it all the water) from the crust.

In Norway it seems that the transformation of granulite to eclogite along fractures is initiated by water, introduced during earthquake rupture; it may even be that water induces the fracture itself. At depths of 60–80 km the only reasonable source of water, given that the granulite is dry, is likely to be the mantle itself, possibly from the pressure-sensitive breakdown of hydrous mineral phases within it. Those enigmatic deep earthquakes beneath Tibet, so close to the crust–mantle boundary, may be modern analogues of the fossil earthquakes in Norway.

Thus we are beginning to see what is special about cratons: they achieve their extra strength through their dry lower crust. They also have unusually thick lithosphere, which is stabilized by chemically depleting the mantle beneath their crust. This too was achieved through earlier melting events, which are known to have removed garnet and reduced the iron content, thereby considerably reducing the density. The result is a mantle component of the cratonic lithosphere that is relatively buoyant compared to younger mantle lithosphere but, because it is much thicker than normal, reduces the overall buoyancy of the plate, thus allowing Finland, with its thick crust, to be at sea level. Cratons assembled of such material are virtually indestructible: much stronger than anything they are likely to run into, and too buoyant to be pushed back into the Earth's interior. They can ram into weaker continental crust, as India has into Tibet, but eventually the upper part of the Himalayan crustal root will be removed by erosion or tectonic processes leaving the lower crust of India to survive.

A typical story?

It is common in the Earth sciences for apparently simple questions, such as “what is special about the continental cratons?”, to lead us in unexpected directions. This enquiry involved a range of clues, from earthquakes, gravity, mantle geochemistry, seismology, rock mechanics and even the metastability of minerals and ancient fossil earthquakes. Along the way other puzzles are illuminated, such as why the mantle beneath the oceans produces earthquakes but not that beneath the continents. It is very likely, for instance, that the same effects are present on other planets. Venus has mountains that rise 10 km above the mean surface elevation. Because Venus contains much less water than Earth, such mountains are also likely to be supported by roots of metastable granulite. In many ways the story recounted here is a typical experience in the Earth sciences: important clues are in unexpected places; to recognize those clues requires a broad general knowledge of how the planet works; and focusing narrowly on what is apparently the core of the problem is often not the way to solve it. These lessons are all familiar to practising Earth scientists, but less so to scientific administrators and managers of directed science programmes.

The Harold Jeffreys Lecture

The Royal Astronomical Society awards this lecture each year to a scientist distinguished for their work in geophysics, in memory of Harold Jeffreys, RAS President and Gold Medallist.