ÐÓ°ÉÔ­´´

Lava erupting from volcanoes is an awesome sight: incandescent fountains at
1000 degree C or more feed red and yellow rivers of rock that turn dull and
grey as they cool and solidify. From this volcanic rock, geologists hope to
find out about the processes that occur deep within the Earth.

The world’s oldest volcanic rocks are a group known as komatiites. Formed
from what were once the hottest lavas that ever erupted, komatiites have
three times as much magnesium as other volcanic rocks, and curious textures
too. These geological oddities are testing current thinking on how magmas
cool and crystallise. They also provide a direct method of measuring the
temperature of the Earth’s mantle – the layer beneath the crust – in
Archean times, between 2.5 and 4 billion years ago . And they are
giving geologists new clues about the way valuable mineral deposits of
platinum, chromium and nickel were formed.

Komatiites are relatively recent additions to the geological canon. In 1969,
the geologists Morris and Richard Viljoen (they were not only brothers, but
also identical twins) were exploring the region around the Komati river in
the Barberton Mountains of South Africa when they came across some unusual
rich red-brown rocks. Their eyes were drawn to what looked like lavas with
pillow-like structures, and some other, larger flows. A few of these rocks
had interlocking masses of large, flat, crystal shapes. The Viljoens named
these rocks komatiites, after the area in which they found them.

When they analysed their samples, the Viljoens found them to be composed
mainly of silicate minerals, as all igneous rocks are. What was unusual
about them, apart from the texture, was that they contained far more
magnesium than most igneous rocks – three times as much as Hawaiian basalt
lavas and the huge eruptions on the mid-ocean ridges, for example. This high
concentration of magnesium changes the physical properties of the lava, and
would have caused it to erupt at unusually high temperatures – as high as
1560 degree C. The lavas would also have had very low viscosities, because
the viscosity of silicate melts increases with the length of silicate
chains, and magnesium in the melt inhibits the formation of such chains.

The discovery of komatiites came as something of a shock. In the first half
of this century, high-magnesium lavas had been the subject of a long-running
dispute between two of the founding fathers of igneous petrology. Norman
Bowen and Harry Hess, working in the US, disagreed over whether or not the
very high temperatures needed to produce magmas rich in magnesium could
exist in nature. Hess suggested that water, probably released from minerals
decomposing in the mantle, might reduce the temperatures at which the magmas
melt. But Bowen showed experimentally that high temperatures were indeed
necessary. By 1960, the dust had settled, and it was generally accepted that
volcanic rocks with high levels of magnesium could not exist. The Viljoens’
discovery of komatiites stirred up the argument all over again.

PUZZLING ROCKS

The komatiites they had found were part of the Barberton greenstone belt, a
band of Archean volcanic and sedimentary rocks that have survived to the
present day without too much deformation or metamorphism. Soon after their
discovery, it was found that greenstone belts in Canada, Australia and
Finland also harboured komatiites. The unusual texture that caught the
Viljoens’ attention was also noticed by mining geologists near valuable
nickel sulphide deposits, particularly at a mine at Kambalda in Western
Australia. The pattern of interwoven blade-shaped crystals reminded the
Australian geologists of the local desert grass called spinifex – so they
called it spinifex texture.

Spinifex texture is one of several distinctive features of komatiite lava
flows all over the world. It dominates the upper part of a flow, the
‘spinifex zone’, while well-shaped, solid crystals in a matrix of
fine-grained rock dominate the lower ‘cumulate zone’. Between these two
layers there is often a third, much thinner layer, in which the crystals lie
roughly horizontally
(see Figure 1).FIG-mg18834201.GIF
Under a microscope, the crystals in
komatiites are seen to have an extraordinary variety of odd shapes. Though
the original minerals have mostly been replaced by clumps of secondary
crystals such as serpentine and chlorite, their shapes remain clear. These
characteristic shapes, along with the high magnesium content of the rocks,
indicate that the original mineral was olivine, an iron magnesium silicate.

STRANGE SHAPES

Some of the olivine crystals have the typical rounded shape familiar to
petrologists from samples of the mineral in basalts and coarse-grained
igneous rocks, for example. Others have strange and unusual forms (see
Figure 2). There are crystals that grew as outlines or skeletons of familiar
crystal forms and others that have repeatedly branched to form dendritic
crystals, named for their resemblance to the branches of trees. Hopper
crystals – stubby or elongated grains with perfect crystal outlines but
hollow cores – make up most of the upper part of the spinifex zone. Here
they are randomly oriented, but lower down there are olivine crystals shaped
like plates or blades, stacked parallel or almost parallel to each other
like a pack of cards. This is the part of the flow that looks like spinifex
grass. In the cumulate zone below, the olivine crystals are rounded and
solid, and form the bulk of the rock.

This sort of variation in crystal shape holds valuable clues to how a rock
has crystallised. In the early 1970s, Colin Donaldson, a crystallographer at
the University of St Andrews, became intrigued by similarities in the
crystal structures of komatiites, lunar rocks and some samples from Harris
Bay on the Isle of Rum, off the west coast of Scotland. How were such
curious shapes formed? In a series of experiments he found that the rate of
cooling as crystals formed made a big difference to their shapes. If the
rocks cooled slowly, then ordinary olivine crystals resulted. But if they
cooled faster, or if the liquid rock had no nuclei or ‘seeds’ on which
crystals could grow, or if the liquid were very high in magnesium, dendritic
or skeletal crystals grew.

At that time, most geological theories on the formation of different types
of igneous rock concentrated on explaining the chemical changes, without too
much concern for the physics. But even while Donaldson was working, ideas
developed in the field of fluid dynamics were starting to revolutionise the
way geologists thought about igneous rocks. So Donaldson’s results set
geologists thinking about the sort of physical mechanisms that might have
affected the rate of cooling of crystals.

One possibility was thermal convection, a process which geologists have for
a long time believed to be important in the cooling of the large
underground bodies of molten rock known as magma chambers (see ‘A stirring
tale of crystals and currents’, New ÐÓ°ÉÔ­´´, 25 November 1989). Any hot
body, whether it is a cup of tea, a lava flow, a magma body, or the Earth’s
hot, solid mantle can cool by either or both of two mechanisms: conduction
or convection. In conductive cooling, the hot material is stagnant and the
cooling rate is relatively slow. In convection, density differences
resulting from cooling at the margins of the body cause the material to
move, albeit rather sedately in the case of the mantle.

A QUESTION OF CONVECTION

Fluid dynamic studies of convection have shown that one of the most
important factors for predicting whether or not convection will occur is
the thickness of the body. Others are the difference in temperature between
the body and its surroundings, and the viscosity of the material of which
the body is made. Big temperature differences favour convection, while high
viscosities inhibit it. The thicker the body, the more likely it is to
convect. So the Earth’s hot mantle, some 5000 kilometres thick, convects
even though it is solid rock, while typical lava flows, which are only a
few metres thick, are unlikely to convect.

The possibility that convection also played an important role in the
formation of komatiite structure was supported by experiments done by
Stuart Turner in the early 1980s at the Australian National University in
Canberra. He looked at how convection in simple mixtures of salts and water
could give rise to spinifex texture. Observing crystals growing with
spinifex texture, he found that the less concentrated solution of the salts
left behind form a layer of liquid around them. The crystals do not grow
quickly in this layer, because the salts they need are in short supply
there. But the tips of the crystals poke out, into the convecting fluid
beyond, and here they do grow, forming blade-like spinifex texture.
Convection, it seemed, could be important.

By the mid-1980s, new physics-based theories about the evolution of igneous
rocks were dominating geologists’ thoughts. Convection occurring in magma
chambers was being invoked to explain chromium and platinum deposits found
in the US and Zimbabwe and interesting layers in coarse-grained rocks the
world over. But Donaldson’s and Turner’s experiments had not been conclusive
for komatiites – fast cooling might not be the only answer. Unusual chemical
circumstances such as very high concentrations of a particular element or
the absence of crystallisation nuclei could also have played a role.
Furthermore, these experimental results did not seem to provide an
explanation for komatiites, because all lava flows are generally too thin
and too viscous to convect.

Then, in 1986, Turner, along with Herbert Huppert and Steve Sparks of the
University of Cambridge, took a closer look at the fluid dynamical theory.
Taking the composition of komatiites as their starting point, they
calculated whether the lavas that formed these rocks would be fluid enough
for convection to be likely. The calculations suggested that they would,
making komatiites potentially useful to geologists as examples of rocks
formed by convection – frozen remains that would tell them indirectly but
conveniently how magma chambers behaved.

A better understanding of what goes on in magma chambers would assist
geologists trying to predict volcanic eruptions, because active magma
chambers are the reservoirs which feed volcanoes like Kilauea in Hawaii. It
could be invaluable to mining geologists too, because frozen magma chambers
can be a source of valuable mineral deposits . But active
chambers can only be studied by remote sensing, and the large size and long,
complex history of frozen chambers make it hard to unravel the way they
worked when molten.

Lava flows have much simpler histories, which makes them easier to study.
The Huppert and Sparks theory for komatiites was part of their larger theory
of how magmas cool. Groups in the US, France, Australia and Japan are
working on similar theories, which all need to take account of the profound
physical changes that occur as hot magma becomes solid, crystalline rock.
Cooling and crystallisation change both the viscosity and density of the
magma, and the crystals themselves create additional complications.

To improve understanding of the way magmas cool, a detailed chemical and
physical history of how a particular komatiite flow crystallised has to be
defined, and from this a detailed model can be constructed. The model can
then be used to make predictions of what rocks will form, which can be
matched against the rocks actually found in nature. Geologists ought to have
been able to put most of this picture together from studies on fresh,
unaltered rocks. But in komatiites there is a major problem: fresh rocks are
hard to find, and most of the olivine is missing, replaced by secondary
minerals.

MISSING CLUES

The crystal shapes within the rock, together with its overall chemical
composition, leave no doubt that olivine was originally dominant. But with
the olivine almost entirely replaced, and because the water content of the
rock is so high – sometimes amounting to as much as 10 per cent by weight –
it is almost impossible to work out a detailed history. With only unaltered
komatiites to work on, there seemed no way forward. Geologists could still
not answer questions like: which layer of crystals formed first? Did all
the layers come from the same lava? Did the solid crystals in the lower
layer form, like the skeletal ones, as the lava slowed and cooled, or were
they picked up from somewhere else? Faced with so much uncertainty,
geologists could not assemble the detailed information needed to test the
theory.

But meanwhile, some very fresh komatiites had been discovered by Tony Martin
of the University of Zimbabwe. The Reliance formation of the Belingwe
greenstone belt in southern Zimbabwe – one of the best studied belts,
thanks to work by Mike Bickle and Euan Nisbet – was soon afterwards dated by
Nick Arndt and Catherine Chauvel at the Max Planck Institute for Chemistry
in Mainz to be 2.7 billion years old, and therefore to date from the
Archean
(see Figure).FIG-mg18834202.GIF
More interesting still, some of the olivine crystals in the
cumulate layer of this rock had survived unchanged. There was even some
glass – representing the most recent composition of the melt – within some
of the hollow crystals. In 1987, Nisbet, now at Royal Holloway, University
of London persuaded the Mainz institute, the Natural Science and Engineering
Research Council of Canada and the University of Saskatchewan to join forces
and sponsor the SASKMAR project to drill a core in these outcrops. Martin
coordinated the project, which produced continuous samples of 150 metres of
the komatiites.

The SASKMAR cores were even fresher – less altered – than the outcrops. The
drill cut through many typical komatiite flows, each about 10 metres thick,
which showed signs of especially fast cooling at the top and bottom. Each
flow had the characteristic layers – spinifex texture at the top,
flat-lying crystals below, with the lower two-thirds made of closely packed,
rounded crystals.

The komatiites of the Reliance formation had rather less magnesium than many
other flows but the freshness of the flows has allowed researchers led by
Arndt, by Mike Bickle at the University of Cambridge and by Mike Cheadle at
the University of Liverpool to compile a history of when and how the
different parts of the rock crystallised. To do so they combined chemical
analysis of individual minerals and of the rock as a whole, and
measurements of grain sizes, with modelling of the cooling and
crystallisation of the flows.

Too simple

The results have reopened the debate on how the komatiites formed. There is
a clear discrepancy between the detailed crystallisation history of the
fresh rocks in these lavas and the predictions of theoretical models based
on convection. Far from confirming the fluid dynamical predictions of
Huppert and Sparks about convection, this history shows that the komatiites
cooled either by conduction alone or by conduction with some weak convection
early on.

It now seems that the Huppert and Sparks model was too simple, and did not
take into account two different effects: first, that crystals in the middle
of the flow greatly increased the bulk viscosity of the crystal-liquid
mixture; and secondly, that the mush of spinifex crystals growing from the
top of the solidifying lava flow inhibits convection.

These results have important consequences for the convection theory. They
imply that as the spinifex crystals grew downwards, they formed a mushy
mixture that was too rigid to convect. This insulated the flow, narrowing
the temperature difference that helps to drive convection. Within the flow,
the crystals that were to form the cumulate layer increased the viscosity of
the mixture of crystals and liquid to the point where convective flow was
slowed down or even stopped. So conduction, rather than convection, must
have dominated the cooling of these flows.

In a quest to account for the unexpected effect of the crystal mush,
geologists and fluid dynamicists are again setting to work in the field and
the laboratory, and at the computer. Research groups in Australia, the US
and France are pursuing a whole new set of experiments and theories. Their
ultimate aim is to understand the physical processes that dictate how magmas
cool. If they succeed, they will have the key to the Earth’s astonishing
diversity of igneous rocks.

Rebecca Renner is a geologist and writer. Steve Barnes and Rob Hill are
geologists with CSIRO Division of Exploration and Mining in Perth,
Australia.

* * *

1: Clues about the mantle

One of the best sources of information about the evolution of the early
Earth are the komatiites, because of what they have to tell about the
composition and temperature of the Archean mantle, in which they were
formed. The mantle, which is composed mainly of olivine with smaller amounts
of other iron and magnesium-bearing silicate minerals, melts in a way
characteristic of all igneous rocks.

The first melts to form have a different composition from that of the mantle
as a whole. By contrast komatiites, with their high magnesium content, come
from magmas produced when a high proportion of the mantle melts, and
there-fore give a better picture of its composition.

When hot komatiite magmas travelled from the mantle through the crust on
their way to the surface, they melted some of the crust along the way, and
carried some of this molten material along with them. As a result, many
komatiite lavas are contaminated with crustal rocks. But because geologists
know the chemical signature of this contamination, they are able to allow
for it and study the tiny amounts of trace elements and long-lived isotopes
in the komatiite itself. From this they obtain chemical clues to how the
Earth evolved. The isotopes and trace elements show whether the mantle
material from which the komatiite formed had ever melted before and, if not,
what material it had lost in the process.

For instance, Catherine Chauvel, now at the University of Rennes, has shown
that the mantle source for the komatiites of the Reliance formation in
Zimbabwe must have endured a major melting episode millions of years before
the flows erupted. In fact, the episode that she identifies has also left
its signature on other komatiites and leads geologists to believe that its
effect was widespread in the Archean mantle. What this partial melting
represented is not yet known, but it could relate to the early formation of
continents.

Another, more glamorous geological oddity provides some insight into this
subject. Diamonds form only under high pressures and temperatures, the sort
of conditions to be expected at the bottom of the continental lithosphere
beneath some 100 to 150 kilometres of rock. Radioactive dating of inclusions
in diamonds has shown that many are very old, often over 3 billion years.

For the Archean continental lithosphere to have been this thick it must
have had a long time to cool, possibly 500 million years or more. Evidence
from komatiites, however, suggests that the mantle was hot – about 200
degree C hotter than today, according to current estimates.

In 1980 two geologists in the US, David Walker and Ed Stolper, put forward
the theory that under certain conditions olivine crystals might float, not
sink, in komatiite melt. Two years later, in conjunction with Euan Nisbet
now of Royal Holloway, University of London, Walker developed this idea
further, controversially suggesting that komatiites were formed in a large
magma ocean deep within the mantle. So the Archean may not only have had a
surprisingly cold crust and hot mantle, but huge magma oceans too.

* * *

2: Minerals from lava

The value of komatiites is more than academic. About 10 per cent of the
world’s nickel comes from nickel sulphide deposits in komatiite lava flows
in Western Australia’s Yilgarn block. These deposits were formed when liquid
consisting mainly of iron, sulphur and oxygen, and up to 15 per cent
nickel, separated out from komatiite lava – in a process very similar to
smelting – and accumulated at the bottoms of lava flows.

The way in which this liquid formed and became concentrated into an ore body
is still being investigated. The amount of sulphur that can dissolve in
komatiite lava is limited. Adding more to a lava which already includes as
much sulphur as it can hold causes an immiscible iron-sulphur-oxygen-nickel
melt to separate. Nickel, copper, gold and platinum rapidly became
concentrated in this melt. Alternatively, sulphide melt droplets can form
due to cooling and crystallisation of olivine from a komatiite lava which is
already saturated with sulphur on eruption. This second model was the one
which most nickel geologists favoured until the development over the past
ten years of new ideas on the origin of the Kambalda deposits in Western
Australia.

The Kambalda nickel deposits consist of nearly pure iron-nickel sulphide,
which solidified from long, ribbon-like ponds of sulphide magma, called ore
shoots, at the base of extensive, thick lava flows. The first step in
understanding these flows came in 1983 when Mike Lesher, at the University
of Western Australia, realised that the ribbons developed within large lava
channels, which regularly overflowed their banks and solidified to form
extensive, sheet-like flows. This theory accounts for the shape of the ore
bodies, but does not explain why only some lava channels contain ore. Where
did the extra sulphur come from in those that do?

The second major advance at Kambalda arose from observations made in the
early 1970s. The sheet flows flanking the channels often had thin layers of
sulphide-bearing sediment underneath and between them. There was no sign of
this sediment either in the channels or, more importantly, under the ore
shoots. Also, Lesher’s mapping had shown that the ore-bearing channels often
seemed to be in linear troughs in the underlying basalt. This led Herbert
Huppert and Steve Sparks of the University of Cambridge to suggest in 1984
that the komatiite lava had produced these channels by melting and eroding
its own floor. Such ‘thermal erosion’ of underlying sediment would provide
the excess sulphur needed to produce an immiscible sulphide liquid, which
then accumulated at the floor of the channel.

Several pieces of evidence support this theory. For one, thermal erosion had
already been seen on a small scale in recent basalt lava flows, and
geologists think it probably explains the ‘sinuous rilles’ seen on the Moon.
For another, the lava was so hot that it would have been no more viscous
than water, and so could have flowed turbulently over great distances. It
would easily melt and erode soft sediments with much lower melting points.

There is also evidence from elsewhere in the Yilgarn block for the existence
of large thermal erosion channels made by komatiite lavas. In the 1970s,
mining geologists looking for nickel came across ‘pods’ of dunite, a rock
composed entirely of coarse, tightly packed olivine crystals. The pods, up
to a kilometre thick and up to 2 kilometres wide, formed the thickest parts
of a continuous sheet of olivine-rich rocks more than 100 kilometres long.
Until the mid-1980s these rocks were thought to be part of a dyke.

Detailed field mapping by our team at the Australian national research
organisation, CSIRO, has shown that the dunite pods are linked by komatiite
lavas. We discovered at least one area where a dunite pod’s base was a
thermal erosion channel which cuts through layered sediments and volcanic
ash beds beneath. The dunite pods are now believed by most geologists to be
cross sections through vast komatiite lava rivers, hundreds of metres wide
and at least tens of kilometres long, formed in a similar way to the
Kambalda channels, but on a much larger scale.

Several of the pods contain sulphide, finely dispersed between olivine
grains. Its distribution throughout the dunite may be due to the scale of
flow. At Kambalda, molten sulphide was dragged along the base of the channel
beneath the komatiite lava. In the much larger lava rivers that formed the
dunite pods, all the sulphur derived from thermal erosion was dissolved in
the lava, and then segregated out again later as dispersed droplets, at the
same time as olivine was crystallising.

Rob Hill and Steve Barnes