¿ìè¶ÌÊÓƵ

Mountains move and ‘solid rock’ is not solid: for modern geologists
there is nothing particularly strange about these ideas. For the Earth’s
crust on which we live is merely a skin on the outside of a rocky mantle,
which in turn surrounds a metallic core. Both crust and mantle are composed
almost entirely of crystalline rock. On a human timescale it appears solid,
but over the longer periods that characterise geological change, the mantle
behaves more as a fluid. It is currents in the mantle that shift the oceanic
and continental crust around. This motion of the crust over the past few
hundred million years is well understood. But the underlying pattern of
flow in the mantle is not, and the question of how the mantle moves has
been the focus of hot debate for 20 years or more. Now geophysicists seem
at last to be moving closer to an understanding of convection processes
deep within the Earth – processes which underlay the making of the Alps
and the Pacific zone of earthquakes and volcanoes known as the Ring of Fire,
and which are still creating the Himalayas.

Geophysicists are in little doubt that the entire mantle is flowing:
only this mechanism can easily explain the detailed form of the Earth’s
surface and its gravity field. Plate tectonics is now used to explain much
of the geology seen at the surface: this model describes our planet’s surface
in terms of a dozen or so crustal plates that are moving horizontally at
several centimetres per year. Geologists have ample evidence that continents
and oceanic crust have shifted laterally by thousands of kilometres during
the past 200 million years or so – a long time by human standards, but less
than 5 per cent of Earth’s history.

This movement is just the outward sign of underlying convection as the
hot mantle slowly cools. But no one has been able to answer the basic question
of whether the mantle is a homogeneous mixture, or a set of separately convecting
systems. These two patterns of mantle flow are referred to as ‘whole-mantle’
and ‘layered-mantle’ convection.

Discovering the pattern of mantle flow is important, for a number of
reasons. First, it gives insights into the nature of the internal forces
that drive plate tectonics, and the earthquakes, volcanism and ore formation
which accompany it – indeed, of most geological processes in the crust.
Secondly, the flow pattern points to the ultimate source of the rocks found
at the surface. If convection is layered, the deepest mantle would not be
swept up towards the surface, so rocks from the lower mantle, which makes
up nearly two-thirds of the planet, would be virtually inaccessible. Finally,
because the upwelling of hot rock transports heat from the deep interior
towards the surface, the flow pattern holds clues to how efficiently the
Earth is cooling.

The whole-mantle pattern allows heat to be brought straight from the
molten core to the surface. In the layered pattern, heat is transferred
far less efficiently, as the upper mantle acts as a blanket, insulating
the deeper interior. The more heat is retained, the more slowly the planet
will cool and the longer vigorous geological activity will persist. So the
pattern of mantle convection essentially determines Earth’s geological
life cycle.

Thinking about the pattern of flow in the mantle dates back more than
40 years, well before plate tectonics was established as a credible theory.
The ideas of Alfred Wegener, the German meteorologist who championed the
concept of continental drift early this century, inspired several researchers
to consider the large-scale flow of the underlying mantle. By the early
1960s, the emerging picture was one of plain convection ‘cells’ – paired
upwellings and downwellings acting like rollers in a conveyor belt, where
the top of the belt represents the laterally moving crust. Such a flow pattern
is the simplest form known, that of Rayleigh-Benard convection. It is similar
to the motion of hot water as it cools in a pan.

Even at the time, this was considered a simplistic model for the Earth’s
inner workings. But the image of Rayleigh-Benard convection has stuck, and
with it the idea of the mantle as a single convection cell, stirring the
interior in a stately, homogeneous manner – the whole-mantle pattern, in
other words. This, after all, is the simplest pattern imaginable, and ‘Occam’s
razor’ – the principle that if there are several equally valid possible
explanations, the simplest is the one to go for – rules against more complex
schemes of mantle flow.

Unfortunately, the convective motion of real fluids turns out to be
a lot less straightforward, and tends towards complex, chaotic behaviour.
Mantle convection is all the more difficult to understand because it takes
place over timescales that are so far outside human experience. Geologists
can never be sure they fully understand how rocks and other planetary materials
behave over geological time. The only convincing way to determine the nature
of mantle convection is to look inside the Earth itself.

The most fruitful method of ‘viewing’ our planet’s interior is through
seismology. By examining acoustic waves generated from earthquakes or explosions,
and recorded by instruments placed around the Earth’s surface, seismologists
are able to deduce details of the structure of the interior. Seismologists
routinely study signals whose wavelengths span several orders of magnitude,
corresponding to frequencies over a range of millihertz to hertz, well below
the pitch of audible sound. In comparison with this thousand-fold range
of wavelengths, the human eye is sensitive to a pitifully narrow wavelength
range of about 400 to 700 nanometres. Moreover, there are two types of
seismic waves – compressional or longitudinal waves, and shearing or transverse
waves. Seismic waves are refracted and reflected by the internal structure
of the mantle, much as light is distorted by passing through a complex transparent
structure such as a glass sculpture.

From such seismological studies, geolo-gists have developed a picture
in which the mantle is subdivided into three shells: the upper mantle, which
extends down to 400 kilometres below the surface; the transition zone, from
400 to 650 kilometres down; and the lower mantle from 650 kilometres to
2900 kilometres down. Between the layers are boundaries, or interfaces,
at depths of about 400 to 410 kilometres and about 650 to 670 kilometres,
at which the density changes abruply. These boundaries are revealed by changes
in the velocity of seismic waves travelling through them.

Geologists ascribe these variations in density to a difference in the
mineral compositions of the layers. Their evidence is based on laboratory
studies which show that it is possible to transform upper mantle minerals
such as olivine, pyroxene and garnet, into spinel, and then, at higher pressures
still, into various perovskite forms by compressing them at the pressures
experienced at the interfaces – that is, at around 12 gigapascals (GPa)
at the 400-kilometre interface and 24 GPa at the lower interface – pressures
of around 2 to 4 million pounds per square inch. The properties of these
dense, high-pressure minerals match the densities and wave velocities seismologists
have observed for the transition zone and lower mantle. Most researchers
now believe that this settles the question of the basic minerals which exist
throughout the mantle.

3D earthquakes

But can seismology show how the mantle flows? Researchers have recently
taken to analysing thousands – even millions – of simultaneous earthquake-wave
readings at stations around the world. This method, known as seismic tomography,
was pioneered by Adam Dziewonski at Harvard University and Don Anderson
at Caltech, and is now used by researchers worldwide. Over the past decade,
this technique has yielded the first three-dimensional images of the Earth’s
internal structure. The images are built up from variations in the velocities
of seismic waves, which in turn reflect how the type, temperature and composition
of the rock varies with depth. So as well as providing detailed evidence
of what the mantle looks like, these images furnish valuable clues to the
pattern of mantle flow.

Unfortunately, however, these clues have so far not added up to a very
clear picture. The most recent seismic tomographic studies suggest that
there is little definitive evidence for intermixing in the upper and lower
mantle, which would support the layered mantle model. Last year, Yoshio
Fukao of Nagoya University completed one of the most complete worldwide
studies of the transition zone ever made. He and his colleagues discovered
that in some regions cold downwelling rock in the ‘subduction zones’ of
the upper mantle pushes as far as 200 to 400 km into the lower mantle. Yet
in many other places the subduction zones are deflected sideways at the
bottom of the transition zone. So it is not clear whether the upper mantle
rock can sink into, or remains buoyed up above the lower mantle. In other
words, the seismological findings do not show conclusively whether the entire
mantle is mixed or not.

Faced with these rather ambiguous results, mineral physicists have
began to pursue an alternative approach, albeit a less direct one. This
involves studying the properties of materials at the high pressures and
temperatures of the mantle: conditions that can, within limits, be simulated
in the laboratory. Researchers measure, under these conditions, the crystalline
structure, density and other properties of whatever mineral they suspect
exists in that portion of the mantle. These measurements are then compared
with seismological observations. If they match, this is taken as indicating
that the material and conditions of the deep mantle were probably deduced
correctly.

Under pressure

The same approach can be used to tackle the question of mantle flow.
Mineralogists measure some of the key seismic properties, such as density
and compressibility, that upper mantle rock would have under the conditions
that are found within the lower mantle. They do this to simulate what would
happen to upper-mantle rock if whole-mantle convection takes place. Comparisons
with seismological data should then reveal whether or not the lower mantle
consists of upper-mantle material that has simply been taken to greater
depth.

A few years ago, Elise Knittle – now at the University of California,
Santa Cruz – applied deep Earth conditions to upper-mantle minerals, such
as olivine and pyroxene, during her research in my laboratory at the University
of California, Berkeley. As she expected, high-pressure, ‘lower mantle’
perovskite was formed. More importantly, she also proved that the perovskite
would remain stable all the way to the bottom of the mantle. Knittle’s experiment
involved using X-ray diffraction, a crystallographic technique, to measure
the density and compressibility of the perovskite at pressures up to those
found in the deepest part of the mantle, which are higher than 100 GPa.
In these experiments, the rock is squeezed between the points of two gem-quality
diamonds, called ‘anvils’, and is heated by a powerful laser beam. One important
advantage of this setup is that the squeezed rock can be studied by passing
either visible light or X-rays through the diamonds.

At the time, however, we could only make high-pressure measurements
at room temperature. So the results were not directly comparable with the
seismic data from the lower mantle, which is at a temperature of about 2000
°C. We then conducted experiments at high temperature and atmospheric
pressure, and by combining these results with those from experiments at
high pressure and ambient temperature we were able to extrapolate to the
conditions of the lower mantle.

The results showed the properties of the rock at lower-mantle conditions
to be inconsistent with lower-mantle rock being derived from the same material
as the upper mantle. In particular, under the conditions prevailing in the
lower mantle, we found the density of upper-mantle material to be only 95
to 97 per cent of the observed density of the lower mantle. From this discrepancy,
we concluded that the upper and lower mantle had not intermixed significantly
over geological time.

At the time of Knittle’s work it was possible, though difficult, to
reproduce temperatures of between 4000 °C and 5000 °C at pressures
of 250 GPa – the problem was making measurements of samples at these extreme
temperatures and pressures. Since then, several research groups have succeeded.
For example, Reinhard Boehler, of the Max Planck Institute for Geochemistry
at Mainz, has just made high-temperature measurements on iron at pressures
of up to 200 GPa, and several other groups have recently made measurements
at pressures of up to 140 GPa, matching those found near the bottom of the
mantle.

Two groups in particular, each using slightly different experimental
methods, have taking our preliminary study an important step further. Yingwei
Fei, Russell Hemley, Ho-kwang Mao and Lars Stixrude of the Carnegie Institution
of Washington, and Nobumasa Funamori and Takehiko Yagi from the University
of Tokyo have used the high-intensity X-rays produced by synchrotron rings
– up to 10 000 times as intense as the X-rays we used – to probe the perovskite
at the high pressures and temperatures of the lower mantle. They were able
to extend, as well as check, our earlier measurements. The Carnegie group
used a diamond anvil cell to make measurements at 30 GPa and 700 °C,
while the Tokyo group, working with a larger cell, investigated a larger
sample at 1800 °C.

In line with our own findings, both groups’ results suggest that the
upper and lower mantle are composed of different materials, which differ
in density by a few per cent. The evidence is growing that intermixing between
the upper and the lower mantle is limited – in other words, the mantle appears
to convect in a layered pattern.

Computer simulations by Ulrich Christensen, then of the Max Planck Institute
for Geochemistry and David Yuen of the University of Minnesota show that
if the density difference between upper and lower mantle materials exceeds
2 to 3 per cent these regions convect separately, without intermixing, a
bit like oil on water. In these simulations, the interface between the
fluid layers becomes distorted to form a distinctive shape in which the
boundary is forced downwards beneath cold, sinking areas of the upper fluid.
If this model holds good for the mantle, the coldest downwelling regions
of the upper mantle – the subduction zones – should extend several hundred
kilometres down into the lower mantle. This may well explain the seismic
tomographers’ observations that subduction zones can push 200 to 400 kilometres
into the lower mantle before the upper-mantle rock is deflected sideways.

Leaky layers

The possibility that mantle flow is layered has attracted the interest
of several fluid dynamicists. Henri-Claude Nataf of the Ecole Normale Superieure
in Paris and Peter Olson of Johns Hopkins University in Baltimore, Ohio,
among others, have examined the movement of two separate convecting fluids,
one lying on top of the other, using both physical models and computer simulations.
They found that the interface between the fluid layers tends to become complex
over time, with small pockets containing a mixture of the two fluids developing
along it. It seems to me that this pattern of ‘leaky layering’ would be
a plausible model for the Earth’s mantle. According to this view, there
is a small amount of leakage between the upper and lower mantle.

Leaky layering is of special interest in considering mantle ‘plumes’.
These are unusually large, long-lived sources of magma that have led to
volcanic activity in places such as the Hawaiian Emperor Islands, the Deccan
Traps in India, and Yellowstone in the western US. Plumes are active for
many millions of years and are thought to emanate from the lower mantle
because they remain in place for a long time, ignoring plate tectonics.
The leaky layering idea would explain how this might happen and how subduction-zone
material might occasionally leak down into the lower mantle. The main constraint,
as far as mineral physics is concerned, is that this leakage must be small
enough not to have intermixed the upper and lower mantle completely over
Earth’s 4.5-billion-year history.

Earlier this year, fluid dynamicists stumbled on some complex flow patterns
associated with mineral transformations deep in the mantle. This discovery
is important because it suggests how the present-day flow patterns may
have come about. Until recently, fluid dynamical simulations often ignored
or simplified the mineral transformations that take place in the transition
zone. But powerful computers are now allowing the effects of the transition
zone to be modelled more realistically. They can also simulate complex,
time-dependent convection, such as that in the mantle, within the spherical
geometry that represents the Earth’s interior. Previously they had to work
in two dimensions.

Boundary changes

This year three separate teams of researchers from Japan, the US and
Canada discovered that the transition zone can have a significant effect
on the convection dynamics of the mantle. Their computer simulations of
changes in volume and heats of reaction of minerals being transformed at
the boundary between the upper and lower mantle suggest that upper mantle
material tends to stagnate in the transition zone. Even assuming whole-mantle
convection, stagnant upper-mantle rock builds up in the transition zone
and then periodically ‘flushes’ down into the lower mantle.

To be geologically realistic, these models will have to include more
details such as variations in the effective strength and fluidity of the
rock and the effects of partial melting. But calculations to date suggest
that even if the mantle was originally uniform in composition, mineral transformations
cause the flow to hesitate in passing through the transition zone, breaking
it up into a layered pattern.

Ultimate confirmation that the mantle is indeed layered awaits more
data from mineral physics, as well as seismology and other methods. One
possibility now being explored is the use of the neutrinos that flood in
from space as a probe for imaging the Earth’s interior. This would provide
a direct method for measuring the mantle’s density and composition. In the
complex picture of mantle convection now emerging, leaky layering, time-dependent
convection through transforming material, and remixing of surface rock into
the mantle all contribute to the richness of the process. It is this very
complexity that helps to keep our planet geologically vigorous, unlike the
Moon, Mercury and Mars which became tectonically ‘dead’ long ago.

Raymond Jeanloz is professor of geology and geophysics at the University
of California at Berkeley.