Category: Geochemistry, mineralogy, petrology and volcanology

Following close on the heels of the hypothesis that iron in Precambrian banded iron formations was precipitated by bacteria (see BIFs and bacteria in February 2003 issue of Earth Pages News) is an account of the origin of silica that makes up roughly half the banding (Hamade, T. and 4 others 2003.  Using Ge/Si ratios to decouple iron and silica fluxes in Precambrian banded iron formation.  Geology, v. 31, p. 35-38).  The rare-earth elements and Nd isotopes in the iron-rich layers suggests that they probably originate from ocean-floor hydrothermal activity.  How their cherty layers formed has largely been overlooked.  Before the Cambrian Explosion there were no organisms that secreted silica in their skeletons.  Consequently, the dissolved silica content of Precambrian oceans was probably much higher than now.  Because silica becomes highly soluble only under very alkaline conditions, it may have been close to saturation in Precambrian seawater.  Quite small changes in seawater chemistry would result in its precipitation as fine-grained chert.  But the main issue is where the dissolved silica came from.

Hamade et al. examined the amount of germanium in the cherts, because it is in the same group of elements and acts as if it were a heavy isotope of silicon.  So it follows Si very closely in its distribution.  Alteration of mafic rocks by sea-floor hydrothermal activity dissolves Si, and so does weathering of continental materials; there is a dual source of seawater Si.  However, basalts have more than 10 times as much Ge as do granitic rocks, and the Ge/Si ratio is a good guide to the dominant source of Si.  Cherts in the BIFs from the famous Hamersley basin in Western Australia have Ge/Si ratios that increase with the amount of iron.  The most silica-rich BIFs seem to have formed from waters derived from continental areas, whereas the iron-rich varieties have a sea-floor hydrothermal signature.  The authors conclude that these BIFs formed on a continental shelf subject to regular, periodic upwellings of deep ocean water.

The steel in your car almost certainly contains iron mined from a banded iron formation or BIF.  These Precambrian sediments are the largest repository of high-grade iron ore on the planet, and nearly all of them formed before about 2 billion years ago, when Earth’s atmosphere and hydrosphere are reckoned by many to have contained very low amounts of free oxygen.  The enigma of BIFs is that, as well as vast amounts of iron, they contain equally large amounts of oxygen combined in hematite and magnetite.  However they formed, there must have been sufficient iron and oxygen in their environment to make these minerals in astounding quanties.  Iron is problematic, because in its Fe-3 form it is almost completely insoluble, and modern sea water contains very little because it is an oxidizing fluid now.  Nobody doubts that BIFs formed in a marine environment, and that would have had to contain plenty of soluble Fe-2.  So seawater before 2 Ga must have been a reducing fluid so that iron emanating from hydrothermal vents on the basaltic ocean floor could remain in solution and end up in near-surface water.  A popular explanation for the oxygen in BIFs is that it was released by the photosynthetic metabolism of blue-green bacteria, near to the basins where BIFs accumulated.  So BIFs mopped up any free oxygen that would otherwise have ended up in air or water and made both oxidising.  Eventually oxygen production outstripped that of soluble Fe-2 (perhaps by a gradual slowdown of sea-floor spreading) and thereby caused all hydrothermal iron to be precipitated near to ocean floor hydrothermal vents; the oceans became iron-poor after 2 Ga.

There is another plausible scenario for BIF formation, explored by a team from Canada, Britain, Australia and Denmark.  Some types of modern bacteria, chemolithoautotrophs and photosynthesisers that do not produce oxygen, are able to fix iron as Fe-3 hydroxides where there is very little oxygen or none at all.  The simple chemical equilibria that they exploit provide both energy and carbohydrate (Konhauser and 6 others 2002.  Could bacteria have formed the Precambrian banded iron formations?  Geology, v. 30, p. 1079-1082).  Evidence that such a process might have “grown” the massive BIFs comes from the famous Palaeoproterozoic Hamersley Group of Western Australia, the source of all the steel in cars produced in east Asia.  The Hamersley BIFs contain extraordinarily fine layers of iron oxides and silica, which may be annual or even daily records of biological cycles.  The key evidence lies in the relative concentrations of other elements in the deposit, phosphorus and trace metals (V, Mn, Co, Zn and Mo), which are close to the nutritional balance needed by the bacteria that Konhauser et al. suggest to have been involved.  Experiments with colonies modern bacteria of these kinds show that they are quite capable of depositing iron hydroxide at rates that would easily build vast thicknesses, given time.  Around 10²² individual cells could do the job at a rate that would have built the Hamersley BIFs – about 100 metres per million years.  That might seem to be an awful lot of bacteria, but it amounts to only about 40 thousand cells per cubic centimetre – far less than the number that build plaque on our teeth!

Phanerozoic marine strontium record throws spanners in the works

Jan Veizer of Ruhr University, Germany and the University of Ottawa is rightly known as “Dr Strontium”.  Almost single handedly he has created the record of strontium variation in seawater through geological time, by analysing carbonates that have extracted it along with calcium.  Input of strontium to the oceans is through continental weathering and hydrothermal solutions from the oceanic crust, and it has proved tempting to use variations in the Sr/Ca ratio of carbonates as a proxy for the rates of both processes, particularly using Sr isotopes.  It is not so simple however, as Thomas Steuber of Ruhr University and Veizer have shown (Steuber, T. & Veizer, J. 2002.  Phanerozoic record of plate tectonic control of seawater chemistry and carbonate sedimentation.  Geology, v. 30, p. 1123-1126).  As in many geochemical cycles, the other important process is burial of strontium in marine sediments, and that depends very much on the type of carbonate that carries it from solution.  Aragonite is between 8 and 4 times more efficient at mopping up dissolved strontium than the other common calcium carbonate, calcite.  So, if aragonite is the main carbonate that is buried, seawater strontium is likely to fall more rapidly than with calcite burial.  Which form dominates in sedimentation depends a great deal on the kind of animal that builds shells – most carbonate buried during the Phanerozoic has been of biogenic origin.  Corals and carbonate-secreting algae use aragonite, whereas molluscs, brachiopods, coccoliths and forams have calcite shells.

Other workers have suggested that there have been periods dominated by deposition of one or other form of calcium carbonate, mainly calcite until the mid-Carboniferous, then aragonite up to the mid-Jurassic, calcite through the Cretaceous and most of the Tertiary, and a current tendency for more aragonite.  Steuber and Veizer show how there is good correlation between changing ocean-crust formation and seawater Sr, and a negative correlation with the Mg/Ca ratio of seawater.  Clearly there are linkages between the three variables, as follows: hydrothermal alteration of new ocean crust exchanges Mg for Ca, so the rate of sea-floor spreading modulates the seawater Mg/Ca ratio; magnesium inhibits the formation of calcite, thereby encouraging aragonite formation; periods of slow spreading therefore favour a higher rate of strontium removal from seawater.  This has profound negative implications for the use of strontium isotopes in marine sediments to monitor the pace of continental weathering (the crux for some gross models of global climate change), and using the Mg/Ca ratio as a means of monitoring seawater temperature variations.

The cold, dense oceanic lithosphere that descends subduction zones is also rich in water.  These features result from the circulation of seawater through young basaltic crust, the exothermic hydration of originally anhydrous minerals in basalt and efficient convective cooling through hydrothermal processes.  Because of this, it might seem as though subduction is a means of re-introducing water into the mantle, thereby enhancing the ability of rising mantle plumes to melt.  The critical process that destines subducted lithosphere to sink inexorably is the conversion of oceanic crust to eclogite by high-pressure, low-temperature metamorphism in the subduction zone.  Eclogite consists mainly of garnet and the pyroxene omphacite, which confer its higher density than mantle peridotite, and the reactions which form them involve dehydration.  Rise of hydrous fluids from the descending slab is implicated in partial melting of the over-riding wedge of mantle to form the volatile-rich magmas that build volcanic arcs.  The higher gas content of arc magmas, compared with those at constructive margins and above mantle plumes, makes them explosive and able to build volcanoes high above sea level.  Most eclogites found at the Earth’s surface are accompanied by still hydrous metamorphic rocks of basaltic composition – blueschists – and others that clearly formed from the sedimentary veneer of the oceanic crust.  So, it might seem that blueschists and metasediments could carry a substantial amount of water into the mantle.  Eventually, its recycling through the mantle could influence later magmatic processes.

Testing this seemingly reasonable extension of the hydrological cycle depends on assessing the water content of newly erupted magmas.  This is virtually impossible for eruptions at the Earth’s surface, because low pressure results in water escape within the higher parts of the volcanic plumbing system, before lavas can be sampled.  However, eruptions onto the ocean floor deeper than a kilometre experience pressures high enough to keep gases in solution, which is why pillow lavas of true oceanic crust contain no signs of gas bubbles.  Crystallised oceanic basalts soon react with percolating water, and their volatile contents are meaningless.  Only the rapidly chilled margins are likely to retain their original composition, locked into quenched basaltic glass.  Even then, a direct measurement of water content can be misleading.  A cunning approach is to consider H₂O as if it behaved like a single element, based on its bulk distribution coefficient between melt and residual solid mantle.  That is close to the values for light rare-earth elements, such as cerium.  So a check for either degassing or contamination of basaltic glass with seawater is the glass’s H₂O/Ce ratio (decreased by the first and increased by the second process).  Jacqueline Dixon of the University of Miami, and co-workers from Harvard and the University of Rhode Island have used this method to assess the probable water content of the mantle source for mid-Atlantic Ridge basalts, whose lead and strontium isotopes suggest that their source was contaminated by older, recycled crust (Dixon, J.E. et al. 2002.  Recycled dehydrated lithosphere observed in plume-influence mid-ocean-ridge basalt.  Nature, v. 420, p. 385-389).  The surprising conclusion of their work is that oceanic basalts formed from mantle with a recycled component have considerably less water in them than those formed by melting of pristine mantle.  This suggests that subduction processes are extremely efficient (>92%) at removing volatiles from the subducted slab; lithosphere descending to depth is almost anhydrous.

Incidentally, the paper begins with an excellent explanation of the somewhat arcane distinctions between different mantle sources affected by lithosphere recycling and mixing.

See also: White, W.M. 2002.  Through the wringer.  Nature, v. 420, p. 366-367; and  Tectonics section below

Using carbonate sediments and fossil shells to assess how the composition of seawater has changed is a long-standing technique in sedimentary geochemistry.  Isotopes of strontium and oxygen have provided revolutionising windows on the pace of continental weathering and fluctuations in sea-surface temperature and continental ice cover for over 30 years.  The magnesium to calcium ratio in fossil shells has given insights into deep-water temperatures for the Cenozoic, more recently.  However, tracking changes in the bulk composition of seawater through time, through analyses of carbonates, is plagued by the continual chemical interaction between rocks and the waters with which they are in contact.  The Mg/Ca ratio of sea water is a potential proxy for the amount of hydrothermal activity on the sea floor, and thus the rate of sea-floor spreading.  This is not because oceanic basalts are magnesium rich compared with continental crust that provides much of the dissolved matter that enters the oceans, but because hydrothermal reactions tend to mop up dissolved magnesium and release calcium..  Unfortunately, magnesium also easily replaces calcium in carbonates during diagenetic processes, particularly dolomitisation.  There are two means of overcoming this hindrance, by analysing seawater trapped as fluid inclusions in evaporite minerals and the shells of echinoderms that still contain minute structures formed in life and are unlikely to have been altered (Dickson, J.A.D. 2002.  Fossil echinoderms as monitor of the Mg/Ca ratio of Phanerozoic oceans. Science, v. 298, p. 1222-1224).  Early results seem to match a prediction that while supercontinents existed, the length of mid-ocean ridges and therefore ocean floor hydrothermal activity were at a minimum.  Around the Precambrian boundary and during the Carboniferous to Jurassic periods, Mg/Ca was high at the time of the Vendian and Pangaea supercontinents.  During major bouts of continental break up – the Lower Palaeozoic and Mesozoic – the ratio is low.  Oddly, the ratio has risen to unprecedented high levels during the Cenozoic Era, when clearly there is high hydrothermal activity.

Despite the fact that the Mg-Ca record of the oceans is limited to just a few short time spans in the 545 Ma record of the Phanerozoic, plenty of geochemists and palaeobiologists are speculating about the possible consequences for evolution of changes in the bulk composition of seawater.  There have been major swings in the proportion of calcite to dolomite in carbonate sediments throughout geological time (see Bacteria and dolomites, January 2001 Earth Pages News).  Discussion now centres on the possible effect of changing Mg/Ca ratios on the waxing and waning of important carbonate secreting organisms, ranging from corals and molluscs that build reefs to the minute coccoliths that formed the Cretaceous Chalk.  Perhaps different groups responded differently to changing water composition, and maybe the Cambrian Explosion of shelly faunas was triggered somehow by a critical shift in the ratio.

See also: Kerr, R.A. 2002.  Inconstant ancient seas and life’s path.  Science, v. 298, p. 1165-1166

Deep carbon cycling, and gold mineralization

One of the more speculative aspects of the carbon cycle concerns the fate of carbonate sediments that descend subduction zones.  One popular hypothesis, with an acronym that is likely to amuse colloquially inclined, British readers (the BLAG model named after its three originators Berner, Lasaga and Garrels) avows that such carbonates contribute to CO₂ emissions from volcanoes above subduction zones by reacting with silica.  The presence in blueschists of abundant aragonite associated with silica suggests that if that does happen, not all carbonate is consumed and a great deal enters very long-term storage in the mantle.  Indeed, aragonite-magnesite associations are stable to pressures that are equivalent to depths of 240 km.  Rocks formed under exceptionally high-pressure conditions, which might shed further light on the deep part of the carbon cycle, are exceptionally rare.  One such occurrence is the Kokchetav massif of Kazakhstan, in which dolomitic marbles accompany eclogites.  Notable for the occurrence of metamorphic diamonds, Kokchetav rocks probably equilibrated deeper than 250 km, so the carbonates are particularly interesting.  Yongfeng Zhu and Yoshihide Ogasawara of Beijing University in China and Waseda University in Japan have found evidence for dissociation of dolomite in them (Zhu, Y. & Ogasawara, Y. 2002.  Carbon recycled into deep Earth: Evidence from dolomite dissociation in subduction-zone rocks.  Geology, v. 30, p. 947-950) during reactions that generate garnet and clinochlore.  The mineral textures reveal equilibria that involve the production of carbon and oxygen, rather than CO₂, so it is quite possible that reflux of CO₂ from subduction zones to the atmosphere may not be as significant as the “BLAGgers” suppose.

Interestingly, the same issue of Geology includes a paper on the geochemical conditions under which gold and copper enter subduction-zone magmas to source major ore deposits (Mungall, J.E. 2002.  Roasting the mantle: Slab melting and the genesis of major Au and Au-rich Cu deposits. Geology, v. 30, p.915-918).  Mungall focuses on the inability of chalcophile metals to enter magmas when sulphides are stable in the mantle.  Under those condition Au and Cu tend to enter sulphide melts whose density and immiscibility separate them from silicate melts.  Oxidation of sulphur is needed to overcome this tendency, and that requires high oxygen fugacity at the depths involved, suggested by him to accompany abundant iron-3 in the subducted materials.  That may be so, but release of molecular oxygen by high-pressure carbonate dissociation, as described by Zhu and Ogasawara, seems an even more likely means of freeing chalcophile metals to magmas.

Major-element chemistry of basalts provides proxies for key parameters involved in magmatism.  Sodium content, normalized to an MgO content of 8%, relates to the degree of mantle melting, and similarly normalized iron content helps assess the depth of melt production.  Such proxies help establish potential mantle temperatures – the temperature of magma that would erupt after rising adiabatically from different mantle depths.  Low Na_8.0 suggests high potential temperature in a magma’s source.

Vast repositories of basalt chemistry relate to every conceivable setting of magmatism, so Na_8.0 and Fe_8.0 numbers are useful in testing various hypotheses.  One of these is that slabs of continental lithosphere affect mantle convection, by forming insulating “lids” that control surface heat flow.  Eric Humler and Jean Besse, of the Université Denis Diderot in Paris, focus on the relationship between mantle potential temperature beneath ocean-ridge systems and their distance to passive continental margins (Humler, E. & Besse, J. 2002.  A correlation between mid-ocean ridge basalt chemistry and distance to continents.  Nature, v. 419, p, 607-609).  Leaving out the complicating factors of continental margins that involve subduction and ridges affected by hot spots, they found that recent ridge basalts show higher potential temperatures when the ridge is close to continental lithosphere than for more distant ridges.  This suggests that the mantle cools away from continents by between 0.05 to 0.1°C per kilometre.  This matches the well-known increase in depth to ridges as they become further from continents.  Rather than being inert passengers on modern plates, continents do play a role in the mantle’s thermal structure.

The scope for synopsis of geochemical data is boosted by wider availability of existing data.  How tedious it used to be, trawling paper journals for tables of analyses with which to compare ones own.  It is still quite a task, but there is light on the horizon, because geochemists at the University of Mainz in Germany have made their compilations for ocean-island volcanic rocks and those from large igneous provinces (flood basalts) available on the web as the initial input to the GEOROC (Geochemistry of Rocks of the Oceans and Continents) database (http://georoc.mpch-mainz.gwdg.de ).  A similar database for ocean-floor basalts is PETDB at Columbia University in the USA (http://petdb.ldeo.columbia.edu/petdb/).  Between them, the two web sites amass over 200 thousand analyses of major- and trace-elements, and isotopes, enough for even the most ardent user of  MS Excel!

Detrital platinum-group grains and “plum pudding” mantle heterogeneity

Evidence for the degree and longevity of geochemical heterogeneities in the mantle has largely stemmed from studies of basalts derived by mantle melting.  The great diversity of melting and fractionation processes involved in their genesis obviously complicates assessment of whether or not the mantle is a mixture of several chemical domains, even though it is suspected.  Indeed it is only to be expected as a result of 4.5 billion years of mantle melting events and recycling of surface materials that find their way into subduction zones, unless, that is, long-term convection is an efficient means of mixing.  A novel approach by a team from Stanford University, the University of Copenhagen and the US Geological Survey uses a combination of the rhenium-osmium radioactive decay scheme and the tendency for Re to enter melts, while Os is highly compatible to address this long-standing conundrum (Meibom, A. et al. 2002.  Re-Os isotopic evidence for long-lived heterogeneity and equilibration processes in the Earth’s upper mantle.  Nature, v. 419, p. 705-708).  The novelty lies in their use of detrital grains of platinoids in alluvium derived from the many ultramafic masses in the western USA, rather than individual basalts or peridotites themselves.

Measurements of ¹⁸⁷Os/¹⁸⁸Os in the grains span a wide range from extremely unradiogenic values to those signifying a high component of radiogenic ¹⁸⁷Os.  The data occupy a bell-shaped (Gaussian) frequency distribution.  While that probably reflects equilibration of old, unradiogenic material with radiogenic Os in melts derived from the mantle ultramafic rocks, and the destruction of any age information, it does point to mantle dotted with patches with different origins.

On a global scale, shifts in sea level recorded by stratigraphers and on seismic profiles stem from one of two main processes: changes in land-ice volume and the volume of the ocean basins.  The latter most often results from changing rates of sea-floor spreading, so that when it is rapid a greater volume of the lithosphere near spreading centres retains sufficient buoyancy to displace the oceans onto continental margins.  During slow spreading, cooling of the lithosphere and an increase in its density enlarges the deep abyssal plains, so that the oceans withdraw to low levels.  The mid-Cretaceous saw vast outpourings of plume-related lavas onto the floor of the West Pacific.  So large, that they reduced the volume of the Pacific basin enough to result in continental flooding that was unprecedented in the Phanerozoic Eon.

On a local scale, changes in sea level recorded by the stratigraphic record include those due to local processes, generally ascribed to tectonic events at continental margins, which involved rising continental lithosphere.  However, one of the greatest forces for local change in the continental freeboard is changing density of the lithosphere due to thermal effects.  Anywhere once affected by major igneous events should record relative falls in sea level during the acme of magmatism, and rises when activity waned.  The British Tertiary Igneous Province, a precursor to the eventual rifting of the North Atlantic under the influence of the Iceland plume is a good candidate for charting magma-sea level connections.  The central volcanic complexes of the Hebrides, and their enveloping flood basalt piles formed at the start of the Palaeocene (~60 Ma).  Around that time, much of the British Isles underwent several kilometres of vertical uplift and exhumation, whose effects remain today.  In the surrounding marine basins, this event is recorded by Palaeogene sandstone bodies, presumable derived by erosion of the uplifted crust.  Yet local Palaeogene sediments also record episodes of rising sea level.  John Maclennan and Brian Lovell of the French Institut de Physique du Globe and Cambridge University have modelled the likely effect on sea levels around the British Isles by crustal underplating of magmas formed during the BTIP magmatism (Maclennan, J. & Lovell, B. 2002.  Control of regional sea level by surface uplift and subsidence caused by magmatic underplating of the Earth’s crust.  Geology, v. 30, p. 675-678).

Up to 8 km of mafic igneous rocks seem to have ponded at the base of the British Isles’ crust while the BTIP was active.  This estimate stems from the fact that the lavas of the province evidence high-pressure fractional crystallization.  Calculations of the percentage of cumulates needed to generate the bulk chemistry of the BTIP lavas suggest that their volume far outweighs that of the volcanic part of the province.  Given estimates of the volume of underplated cumulates, modelling boils down to examining the consequences for lithospheric density of initial heating and its subsequent relaxation.  The Palaeogene sedimentary record provides good support for the model, with massive uplift from 60-56 Ma (the period when the BTIP was forming).  Sudden sea-level rise at the end of this period never reached the level prior to magmatism; in fact it amounts to one half the estimated uplift.  That is precisely in line with the underplating model.

Erupted at the time of the Palaeozoic-Mesozoic boundary, and coinciding with the largest mass extinction during the Phanerozoic, the Siberian Traps are by far the biggest example of flood-basalt volcanism known.  They blanket a huge area of the Siberian Platform.  To the east of their outcrops is a large extensional downwarp, known as the West Siberian Basin, where recent deep drilling has cut through up to 1 km of flood basalts.  Dating samples from 15 boreholes proves that these too are members of the Siberian Trap suite (Reichow, M.K. et al. 2002.  ⁴⁰Ar/³⁹Ar dates from the West Siberian Basin: Siberian flood basalt province doubled.  Science, v. 296, p. 1846-1849).  Combined, the two zones of Siberian Traps represent eruption of around 2.3 million km³ of plume-derived magma at around 250 Ma ago, possibly within 2 or 3 Ma.  Gas release from such a stupendous event is implicated in the Permian-Triassic mass extinction, either through climate change associated with CO₂ and SO₂, or toxic effects of hydrofluoric acid.  Unlike the end-Triassic and K-T extinctions, no clear evidence has emerged for coincident flood volcanism and major impact at the end of the Palaeozoic Era.  However, the use of tungsten isotopes as “fingerprints” for extraterrestrial debris in boundary sediments may help resolve the issue of whether an impact accompanied the Siberian Traps (see Tungsten and Archaean heavy bombardment, this issue)

Various indicators, such as the presence of detrital uranium oxide and iron sulphide grains in sediments older than about 2.3 Ga and the appearance of terrestrial sediments stained red by the presence of ferric (Fe-3) oxides thereafter, have long been used to suggest that atmospheric oxygen was a mere trace before that time.  Generation of oxygen through photosynthesis by simple organisms, principally blue-green bacteria, could have led to an oxygenated atmosphere when their productivity exceeded the tendency for oxygen to be consumed by reaction with reducing agents, such as abundant ferrous (Fe-2) iron in sea water, and by burial of carbon-rich dead organic matter.  That method is a central plank in the Gaia hypothesis.  However, geochemical considerations suggest another scenario for oxygenation (Catling, D.C., Zahnla, K.J. and McKay, C.P. 2001.  Biogenic methane, hydrogen escape, and the irreversible oxidation of early Earth.  Science, v. 293, p, 839-843).  Unless carbon burial exceeded the rate at which reductants supplied to the outer Earth (including exposure of buried carbonaceous sediments) by geological processes consumed oxygen, the atmosphere would remain low in oxygen.

Lacking in oxygen, the early atmosphere would have been able to support build-up of methane from biogenic processes – today methane is soon oxidized to carbon dioxide and water.  Carbon isotope evidence suggests that early life was dominated by methanogens, and such organisms alive today are genetically very primitive.  Consequently, methane is a good candidate for keeping average surface temperature above the freezing point of water at a time when the Sun’s output of energy was considerably lower than it is now.  All hydrogen-bearing compounds become dissociated high in the atmosphere, to release hydrogen atoms, and they readily escape the Earth’s gravitational pull.  Fortunately, this does not happen now because the only significant H-compound, water, cannot rise above the tropopause.  The decline in temperature upwards acts as a cold trap for water.  Were this boundary not in place, and it is largely due to the presence of ozone in the stratosphere which absorbs radiation to give higher-level warming, Earth would long ago have lost most of its water, as did Mars and Venus.  In the early atmosphere, methane would not have been “cold trapped”, and nor is it today.  So, during that period, hydrogen would steadily have leaked from the Earth.

The chemical outcome of such a simple process would have been a steady decline in the reducing capacity of the Earth as a whole, for hydrogen is a powerful reductant.  Because most of our planet’s hydrogen was locked in water from the time of its accretion, its escape must have resulted in a net gain of oxygen somewhere in the Earth system.  Increased methane productivity by methanogen bacteria during the Archaean and early Proterozoic would have enhanced this tendency for the whole Earth to become more oxidizing.  Catling et al. argue that the continental crust became more oxidized, so that any gases released from it by metamorphism would become less reducing.  That would have reduced the tendency for immediate consumption of oxygen produced by photosynthetic organisms, culminating in its eventual ability to exist in the atmosphere in balance with biological processes at around 2.3 Ga.

Zircons’ window on the Hadean

The oldest tangible rocks that are not completely changed by deep-crustal metamorphism are those of Isua in West Greenland.  Interleaved with gneisses that originated probably from calc-alkaline intrusions are rocks formed at the Earth’s surface around 3.8 Ga ago.  The general scene represented by this Akilia Association is in many respects familiar – the operation of plate tectonics, rapid generation of what was to become continental crust, abundant evidence for the action of liquid water and even the isotopic traces of living organisms.  That 750 Ma after the Earth’s accretion the last two were present is no surprise.  The oddity is that, despite decades of effort, there is still no sign of continents older than 4 Ga.  That crustal rocks which had undergone considerable evolution from their mantle source did exist in the missing half-billion years emerged from the discovery of detrital zircons as old as 4.4 Ga in much younger Australian sedimentary rocks.  Some of the rare, tiny grains show isotopic evidence that the magmas in which they formed had contact with liquid water at the surface.

As well as containing sufficient uranium to allow the dating of single grains by the U-Pb method, zircons also contain hafnium, which is chemically very similar to zirconium.  Measurable quantities of ¹⁷⁶Hf add to common ¹⁷⁷Hf by the decay of ¹⁷⁶Lu, giving a potential dating technique.  However, zircon contains only minute traces of lutetium, so that its ¹⁷⁶Hf/¹⁷⁷Hf ratio remains that of the ultimate source of its host rock.  Relative to hafnium, lutetium is more likely to remain in the residue left by partial melting of the mantle, or so theory suggests (geochemists can only deduce this from various lines of indirect evidence).  Consequently, mantle that has sourced continental crust builds up ¹⁷⁶Hf from the time such crust formed., whereas continental crust has significantly lower levels.  Studying hafnium isotopes in very old zircons is therefore a means of seeking periods when significant amounts of continental crust separated from the mantle.  Because such tiny amounts of the radiogenic hafnium are involved, an accurate decay constant for ¹⁷⁶Lu is vital (Scherer, E., Münker, C. and Mezger, K. 2001.  Calibration of the lutetium-hafnium clock.  Science, v. 293, p. 683-687).  Zircons from the oldest rocks in Greenland, Canada, Australia and South Africa fall into two, complementary groups; those with slight enrichment in ¹⁷⁶Hf and those with slight depletion.  Simple geochemical theory seems to indicate that indeed magmas similar to those that contributed to formation of the bulk of continental crust did form as early as 4.4 Ga ago.  However, zircons with younger Archaean ages show little sign of deviant hafnium, which suggests that a large proportion of the mantle was not involved in early sial formation.  Hadean continental material no doubt formed, but not much.  That is no surprise, for involvement of surface-derived water in mantle melting above zones where earlier lithosphere returns to the mantle, whatever their form, seems inevitable in a planet noted for its high water content.  That is the basic “recipe” for the formation of silica-rich magmas.

Two things stem from this work: the probable futility of seeking Hadean continents; the unlikelihood that the chemical heterogeneity of the mantle stemmed from Hadean continet formation on a massive scale.

See also:  Kramers, J. 2001.  The smile of the Cheshire Cat.  Science, v. 293, p. 619-620

The basalt floods draped over some great continental plateaux and considerable areas of the ocean floor, ocean islands far from plate boundaries and the volcanic provinces sitting at the ends of various oceanic island chains are with little doubt the product of plume-like masses rising from great depth in the mantle.  What is not so well agreed is just what bit of a plume underwent partial melting to make the magma, the depth at which that took place and the prevailing temperature.  There is some support for plumes that rise from a mantle transition zone about 700 km down, where there is an abrupt increase in temperature.  Such plumes form a hot head when they impact the lithosphere, and that should be the source for magma.  Plumes that rise from the core mantle boundary, should in theory have heads that are cooler than their tails, and which grow hugely by being stalled at the 700 km discontinuity.  The two combined might form little plumes that rise from a big head at 700 km that spreads laterally.  Nicholas Arndt gives a neat summary of these unseen ramifications in a recent issue of Nature (Arndt, N.  2000.  Hot heads and cold tails. Nature, 407, 458-461).

Arndt was moved to make his comments by evidence from Namibian flood basalts from the 128-138 Ma old Paraná-Etendeka large igneous province (Thompson, R.N. and Gibson, S.A. 2000.  Transient high temperatures in mantle plume heads inferred from magnesian olivines in Phanerozoic picrites. Nature, 407, 502-506).  Thompson and Gibson found highly magnesian olivine crystals, among more normal ones, in basaltic dykes that cut the Etendeka basalts.  The more Mg-rich an olivine is the more primitive (the more like the composition of the mantle) the magma from which it crystallized.  They calculate that these anomalous olivines equilibrated with a magma with 24% MgO (compared with the <10% of most basalts) – probably a komatiite.  But they are in much more evolved basalts, so they suggest that a primitive magma at the hot head of a plume that hit the lithosphere itself underwent fractional crystallization to produce plain basalt.  They draw from that the conclusion that the plume head was 300-400°C hotter than the surrounding mantle – as expected in the first plume model above.  Arndt is sceptical, partly because there are so many unknowns about the source region and partly because there are many other possible explanations.  He suggests more similar work and other kinds of geochemical research on large igneous provinces in general.  To that might be added looking for some of the possible mechanical consequences of hot or cool plume heads.

The Earth has a core made, probably, of alloyed iron, nickel and sulphur.  Much evidence points to the core having formed very early in our planet’s history, probably in its first 100 million years.  Core formation explains the depletion in iron of mantle rocks and magmas derived from them, compared with iron’s abundance in the cosmos.  Because some rarer elements have a 10 000 times greater tendency to partition into melts containing metallic iron than into silicates, such siderophile (‘iron-loving’) metals are also highly depleted in the outer Earth.  That is one of the reasons why gold and the platinum-group metals are so rare and highly prized at the Earth’s surface.  In fact, such noble metals are a lot more abundant than the presence of a metallic core could have allowed; they should be at vanishingly low abundances.

One solution to this paradox is that the ‘extra’ gold and PGEs arrived after core-formation had finished, the agency of delivery being continual bombardment by meteoritic debris in the first half billion years of the Solar System’s history.  The other is that somehow, the affinity of such metals for iron drops off at extremely high pressures.  German, Canadian and Australian geochemists  (Holzheid, A. et al., 2000.  Evidence for a late chondritic veneer in the Earth’s mantle from high-pressure partitioning of palladium and platinum.  Nature, v. 406, p. 396-399) have shown experimentally that such a decrease doesn’t occur, at least in the outermost 500 km of the Earth.  This points strongly to impacts having seeded the upper mantle with noble metals, and therefore, perhaps, with lots more besides.  This re-opens the old controversy between homo- and heterogeneous accretion of the Earth, tempered by the fact that more common siderophile metals, such as nickel and cobalt do not show mantle abundances that are in disequilibrium with core formation.  The distinction is not trivial, for much of Earth’s evolution has been driven by its internal composition, most especially its content of radioactive isotopes and water.

The Moon seems to have formed as a result of a gigantic impact of a Mars-sized body with the early Earth.  Since the Moon has neither a core nor its full cosmic complement of iron, such a catastrophic beginning (effectively ‘Year Zero’ for the geochemistry of both bodies) must have taken place after core formation in the Earth.  Because lunar rocks are so little changed by later events, its age is known with considerable accuracy – the Lunar Highlands are about 4450 million years old.  It would be interesting to compare gold and PGE abundances between Earth and its Moon, for that might reveal the period during which bombardment delivered siderophile elements.  Up to 3.8 billion years ago, both bodies received lots of visitors, culminating in a bout of huge impacts between 4.0 and 3.8 billion years ago that formed the huge lunar craters, that early astronomers termed maria or ‘seas’.