poirier 1994 pepi

8/17/2019 Poirier 1994 PEPI

1/19

PHYSICS

OFTHE EARTH

ANDP LAN ETA RY_________ INTERIORSELSEVIER Physics of the Earth and Planetary Interiors 8 5 (1994) 319—337

Light elements in the Earth’s outer core: A critical review

Jean-Paul Poirier

Département des Géomatériaux (URA CNRS 734), Institut de Physique du Globe de Paris, 4 Place Jussieu,

75252 Paris-cedex 05, France

Received 8 October 1993; revision accepted 8 February 1994

Abstract

There is little doubt that densities for the Earth’s outer core, inferred from seismology, require that it isconstituted of an alloy of liquid iron and light elements. However, the nature of the light alloying elements is stilluncertain as it depends in a large measure on the conditions of accretion of the Earth and formation of the core.

The arguments brought forward for or against silicon, oxygen, sulphur, hydrogen and carbon are critically reviewed.There is no reason to consider that only one element is present in the outer core. Experimentally determinedand/or calculated ternary and quaternary phase diagrams are needed to provide constraints on the nature of thelight elements.

“There is no reason to believe that the core is a particularly clean system” — D.J. Stevenson (1981)

“All discussions of the nature of the light element suffer from too few data and too many extrapolations” — R. Brett

(1984)

1. Introduction shock-wave experiments, was later calculated tobe 10 ±2% (Jeanloz, 1979) and confinned by

More than 40 years ago, Birch (1952), from static determination of the equation of state of

seismic data, interpreted the outer core as ‘liquid s-Fe up to 3 Mbar (Mao et al., 1990) (Fig. 1); theiron, perhaps alloyed with a small fraction of uncertainty arises from the seismic density profile

lighter elements’. He suggested carbon and sili- and from the imperfect knowledge of the temper-con. The fact that the outer core is mostly iron ature in the core and the thermal expansionwas later established beyond reasonable doubt by coefficient of iron at core conditions. Birch (1964)

Birch (1961, 1964), who confirmed that the den- assigned the density deficit to lighter elements insity of the core was about 10% lower than the solution — silicon, sulphur or oxygen. Over thedensity of iron at the core pressures and tempera- years, a number of elements lighter than iron —

tures, and that the seismic parameter (1 = K/p, silicon, sulphur, oxygen, hydrogen and carbon —where K is the bulk modulus and p the density) were considered by various workers, singly or

of the core was higher than that of iron. The more rarely in combination (Fig. 2).difference in density between the core and the The nature of the light elements in the outerhigh-pressure phase of iron, on the basis of core is a standing problem of prime importance;

0031-9201/94/$07.00 ~ 1994 Elsevier Science BY. All rights reserved

SSDI 0031-9201(94)02948-B


2/19

320 J.-P. Poirier/Physicsof the Earth and PlanetaryInteriors 85 (1994) 319—337

PREM (5000 K)

150 200 250 300 350

Pressure (CPa)

Fig. 1. Comparison of the density of solid pure a-Fe at high pressures and 5000 K with the core density (Preliminary Reference

Earth Model; PREM). The fusion volume of iron (about 3%) is not taken into account. After Badding et al. (1992).

it conditions in particular the existence and value What is the composition of the inner core

of a freezing point depression at the inner core (Jephcoat and Olson, 1987)? Are chalcophile ele-boundary (ICB), on which the answers to major ments depleted in the mantle by volatilization orquestions in geophysics and geochemistry hang: by being sequestered in the core? What is thee.g. to what extent are light elements released at significance of the U—Pb age of the Earth (Over-the ICB, thus inducing compositional convection sby and Ringwood, 1971)?thought to power the geodynamo (Braginsky, It can be taken for granted that any light1964; Loper, 1978)? By how much is the tempera- solute element in suitable proportions in liquidture at the ICB lower than the melting tempera- iron will decrease its density to make it compati-

ture of pure iron at 3.3 Mbar (Stevenson, 1981)? ble with seismological Earth models (Birch, 1952);

however, the number of potential alloying ele-ments is restricted by the following cosmochemi-

30 ________________________________ cal and metallurgical constraints (Stevenson,

1981):

(1) The light elements must be sufficiently

20 abundant in the accreting Earth, thought to be of I chondritic composition. However, they must not

Si be volatilized by the heat of accretion and escapeentirely.

~ io H (2) It is currently accepted that most of the

core formed early during accretion (Oversby andC Ringwood, 1971; Allègre et al., 1982), hence be-

1950 1 960 1970 1990 2000 fore the Earth reached its present size. The lightYear elements should then be able to partition into

Fig. 2. Cumulative number of papers on light elements in the liquid iron at relatively low pressures (it may becore, as a function of publication date, useful to remember that the central pressure of


3/19

J.-P. Poirier/Physicsof the Earth and Planetary Interiors 85 (1994) 319—337 321

the Moon is about 50 kbar and that of Mars Ringwood, 1977). In what follows, without anyabout 450 kbar) and preferably help form a preconceived idea nor theory of my own about

metallic liquid melt (e.g. eutectic) at tempera- the nature of the light elements in the core, I willtures lower than the melting point of pure iron. systematically review the literature from Birch

(3) The outer core is presumably homoge- (1952) to the present day, and try and sort out the

neous, hence the light elements should remain factual information and the speculation but-soluble in iron at the current pressures of the tressed by ad-hoc models. In an admittedlycore. pedestrian way, I will successively examine the

(4) At least some of the light elements must be case for or against silicon, sulphur, oxygen, hydro-

released in the melt during crystallization of the gen and carbon, or mixtures thereof, in an orderinner core and the freezing point depressed, if following more or less the evolution of fashion

compositional convection is to be effective, during the last 40 years, as reflected by the cumu-The nature of the light elements is tightly lated number of papers on one or the other of

linked to the mode of formation of the core and the light elements (Fig. 2). Although sometimesit has been deplored that ‘models of core forma- listed among the possible light elements, nitrogen

tion are further complicated by our lack of knowl- was never the object of any attention. Alder (1966)

edge of the chemical composition of the core’ claimed that magnesium could be present in the(Jones and Drake, 1986) or pointed out that ‘the core, but this conclusion resulted from calcula-nature of the most abundant light element in the tions of the solubility of MgO based on unavail-core is important for determining whether core able physical quantities. Ringwood and Hibber-formation was dominated by high pressure or low son (1991) later showed experimentally that MgO

pressure processes’ (Newsom and Sims, 1991). I is one of the least soluble oxides in molten iron,

take, of course, the opposite view and regret that even at high pressure.the present knowledge of core formation does not Is the outer core in equilibrium or not with theprovide more constraints on the nature of the mantle? This problem is obviously related to thelight elements. composition and formation of the core. The con-

The most valuable information on the possible troversy is centred on the measured abundances

compositions of the outer core comes from phase of siderophile elements in the upper mantle, but

diagrams of binary, ternary and quaternary Fe— it is always possible to reproduce them in a some-light elements systems at pressures up to the ICB what ad-hoc fashion by accreting cocktails of dif -pressure. Some binary and ternary diagrams have ferent chondrites in various proportions at van-been experimentally determined to pressures of ous times. I do not believe that the answer to this

about 100 kbar, and calculated to higher pres- problem may contribute much to selecting orsures when equations of state for the end-mem- eliminating an element as the major light element

bers are known. It may be noted here that about in the core. However, in view of the importance4% nickel is thought to be present in the core of the topic in the literature, I will succinctly deal(Brett, 1971) and although it does not appreciably with it.

change the density of liquid iron, its presence

should not be forgotten, as phase diagrams of

systems Fe—Ni—light elements may be signifi- 2. Siliconcantly different from those of systems without

nickel (Urakawa et al., 1987). Birch (1952) first remarked that: ‘any of theMore or less succinct reviews of the literature most abundant elements will reduce the density

on light elements in the outer core have been of iron. The effect of carbon and silicon arewritten, as part of papers or books on the core perhaps the most familiar, a reduction of density

(Brett, 1976; Stevenson, 1981; Jacobs, 1987, 1992; by 10% requiring only small percentages of theseJeanloz, 1990), or as an introduction to papers elements.’ MacDonald and Knopoff (1958) thensupporting the presence of a given element (e.g. pointed out that an Earth with an iron—nickel


4/19

322 f-P. Poirier/Physics of the Earth and Planetary Interiors 85 (1994) 319—337

core and a penidotitic or eclogitic mantle would gases. The fact that some unreduced FeO ishave an (Fe + Mg + Ni)/Si ratio higher than that present in the mantle is attributed to lack of of the chondritic meteorites, in contradiction with equilibrium. Support for this theory is found in

the accepted tenet that the average composition the mineralogy of meteorites, similar to blastof the Earth is chondnitic; they concluded that furnace assemblages (Ringwood, 1959), and in

there must be silicon in the cone and that the the fact that there is 2—6 at.% Si in the metallichigher the silicon content assumed in the mantle, phase of enstatite chondrites (Ringwood, 1961).the less silicon there must be in the core. Mac- Ringwood (1959) also noticed, as had MacDonald

Donald and Knopoff (1958) provided another in- and Knopoff (1958), that there is a higher propor-dependent argument for the presence of silicon tion of Si0

2 in the silicate phase of chondnites

(or indeed any light element) in the core: they than in the Earth’s mantle and therefore, if thestarted from the observation that Bullen’s density compositions of Earth and chondrites are to bedistribution and an interpolated Thomas— similar, the silicon missing in the mantle must beFermi—Dirac equation of state yield a weighted in the core; a satisfactory Earth model requiresmean atomic number Z = 22 for the outer core the presence of about 20 wt.% Si in the core.

(Knopoff and Uffen, 1954), obviously too low for Urey (1960) took exception to the model pro-a core of iron (Z = 26) and nickel (Z = 28). The posed by Ringwood (1959), on the grounds thatvalue of Z for the outer core can be brought reduction of both iron and silicon would entail

down to 22 by the introduction in the core of the evolution of very large quantities of CO andabout 20 wt.% Si. They did not rule out the that no satisfactory mechanism for the escape of presence of sulphur, although they saw difficul- gases of molar weight 28 from a planet even the

ties in the fact that an Fe—S mixture with appro- size of Mars is known. He proposed that somepriate Z for the core would require an unrealistic residual carbon and some sulphur dissolved inhigh initial abundance of sulphur or some process iron would give a satisfactory explanation of the

by which silicon is lost to the Earth with respect density deficit of the core.to the more volatile sulphur. Knopoff and Mac- Balchan and Cowan (1966) performed shock-Donald (1960), using equations of state deter- wave experiments on Fe—Si alloys (4 and 19.8

mined from shock-wave data for iron and other wt.% Si) up to 2.7 Mbar and determined pres-metals, later found that a material containing sure—density and sound speed—density curves,between 20 and 30 wt.% Si is consistent with which they compared with the curves for the

seismic data. Earth. They found that their results were consis-

Ringwood (1959) simultaneously proposed a tent with a core containing 14—20 wt.% Si, inmodel of accretion of the Earth and formation of agreement with the suggestions of MacDonaldthe core resulting in the incorporation of silicon and Knopoff (1958) and Ringwood (1959).in the core. He assumed that the Earth accreted From shock-wave theory, Stewart (1973) calcu-at low temperatures from oxidized dust and gas lated all possible Hugoniots compatible with the

of cosmic composition, trapping carbonaceous seismic properties of the outer core. He foundcompounds. As temperature rises during accre- that his results were compatible with 8—20 wt.%

tion, melting and convection occur and the Si in the core, but did not rule out other light

trapped carbon reduces the oxides, much as in a elements.blast furnace, according to the reactions Brett (1971), although agreeing that silicon is

likely to be a major element in the core, dis-FeO+C—~Fe+CO .

agreed with Ringwood (1959, 1961, 1966) on theFeO + CO —~ Fe + CO2 question of disequilibrium between core and

SiO + 2C —~ Si + 2CO mantle and mustered arguments for equilibrium,2 answered by Ringwood (1971). The problem of

A metallic phase containing iron, nickel and equilibrium vs. disequilibrium will be discussed insilicon then segregates; CO and CO2 escape as Section 7. In a later review, Brett (1976) came to


5/19

J.-P. Poirier/Physicsof the Earth ana Planetary Interiors 85 (1994) 319—337 323

the conclusion that ‘we totally lack solid evidence Wänke’s model, raising doubts about the possibil-

on what the light element might be’. ity of a nebula of solar composition ever achiev-

Ringwood (1977) acknowledged that his accre- ing sufficiently reducing conditions for 10—20% Sition model for incorporating silicon in the core in the core. Based on their study of the solubilitywas ‘very specific’ and met with difficulties, of mantle oxides in molten iron at high pressures,

pointed out by Brett (1971, 1976), essentially be- and on Knittle and Jeanloz (1989) experiments incause it led to gross disequilibrium between man- a diamond-anvil cell, they suggested that the onlytie and core, and he thereupon suggested that way to incorporate Si in the core would be byoxygen (as FeO dissolved in iron), and not silicon, dissolution of Si0

2 from mantle silicates at high

was the light element in the core. It must be pressure, but even then they saw difficulties ow-

noted that Ringwood did not give very cogent ing to the fact that Ti02, which is more solublereasons (other than disequilibrium) for rather than SiO2, should be depleted in the mantle if abruptly abandoning the idea of silicon. Later on, SiO2 was dissolved in the core; as it is not, they

he eliminated altogether the need for silicon in concluded that there must be very little Si02 inthe core by suggesting that it is not the Earth the core. It must be noted that the argumentmantle that is depleted in silicon with respect to against Si, based on Ti02, holds only if onethe cosmic primitive composition, but the chon- assumes that Si in the core comes from dissolu-drites that are enriched (Ringwood, 1989). tion of the mantle at high pressure.

In fact, the main objection to Ringwood’s early Allègre et al. (1994) calculated the ratio Si/Femodel for incorporating silicon in the core was in the core without assuming a priori the pres-made by proponents of sulphur as the major light ence of silicon in the core (and without devising

element (see Section 3): reduction of silicon from an accretion model to acount for it). The ratiosilicates requires a high temperature during ac- (Si/Fe)core is calculated from the known ratio

cretion, which would have volatilized elements (Si/Fe)mantie and from the ratio (Si/Fe)Earth,

more volatile than Si (e.g. 5) that, however, are which must be estimated:

still present; it also produces vast quantities of (Si/Fe)CO and CO2 that have to be blown off (Rama coreMurthy and Hail, 1970, 1972). Wänke (1981) and [(S1/Fe)Earth “fm~(Si/Fe)mantie] /f~ Wänke and Dreibus (1988) circumvented this dif- where fm and ft are the mass fractions of iron inficulty by proposing an accretion in two stages: the mantle and core, respectively.

first, a highly reduced, devolatilized material con- The ratio (Si/Fe)Earth for the bulk Earth is

taming Si in metallic form was accreted (thus calculated from the Si/Al vs. Fe/Al trend estab-removing the problem of high-temperature re- lished for meteorites (and assumed valid for theduction to the solar nebula or planetesimals Earth), and the value (Fe/Al)Earth is calculated

stage), then, when the Earth had reached about from the corresponding ratio for the mantle, tak-two-thirds of its present mass, and after the met- ing into account the fact that aluminum does notals and sulphides had segregated to form the enter the core:

core, a more oxidized, volatile-rich material wasadded (accounting for the chondritic proportion (Al/Fe) Earth = fm~ (Al/Fe) mantle

of siderophiles in the upper mantle). Wänke could Allègre et al. found about 11 wt.% Si in thetherefore build an Earth with Cl chondrite abun- core. This would give a density deficit in the coredances and a core containing 12.5% Si. He fur- of only 6—7%. Other elements must therefore bether suggested (Wänke, 1981) that the presence present (a few per cent S and 0 would give theof silicon in the core might result in the precipita- correct density).tion of the intermetallic compound Ni2Si to form From the metallurgical viewpoint, there is no

the inner core, as first proposed (without the difficulty in having silicon dissolved in iron duringleast justification) by Herndon (1979). accretion at low pressures: the Fe—Si phase dia-

Ringwood and Hibberson (1991) objected to gram at ambient pressure exhibits a eutectic point


6/19

324 J.-P. Poirier/Physicsof the Earth and Planetary Interiors 85 (1994) 319—337

wt % Si 355 10 20 30

I I ____

1500 2S~j~1~ç ;)

141 /~f~~$W4. f\ 1400 C~

*20 ~

1300 195 23 5 ~

~‘ 1200 1212° 1203 15 ~ ~

I ~‘;::~ ~ ~ ______________________________900 1500

825 160d t~,. ~800T F. “ ~ ,~‘ .~ . ~‘ .~‘ .~‘

-.5 wt%Si

700 “ Fig. 4. Fe—S—Si miscibility gap at 1 atm (after Raghavan,

1988).600

500 I ‘i”T’ calculations were based on the ambient pressure

Fe io 20 30 40 50 densities of iron—nickel and troilite!); sulphur

At % Si would then ‘obviate the necessity’ of having Si in

Fig. 3. Fe—Si phase diagram at 1 atm (after Raghavan, 1988). the core, with the apparently unpleasant (why?)

consequence of core—mantle disequiiibrium.

It was, however, Rama Murthy and Hall (1970,at 1200°Cand 20.5 wt.% Si (Fig. 3), and there is 1972) who really started an interest in sulphur.no reason to believe that Si would be less soluble They noticed that sulphur was depleted in theat high pressure. Silicon also lowers the melting crust and mantle relative to the other volatilepoint of iron at atmospheric pressure. However, elements by several orders of magnitude, whereas

the ternary diagram Fe—Si—S at 1 atm exhibits a halogens, water and rare gases were present in

vast miscibility gap in the liquid state, which about their chondritic abundances (Fig. 5). Thewidens as temperature increases (Raghavan, depletion could not be due to preterrestrial frac-1988); this, of course, would limit the solubility of sulphur in Fe—Si (Fig. 4). The gap possibly shrinksat high pressure, but we do not know that it does

/ /~

Mason (1966) was the first to suggest definitely ~ 10that sulphur might be the major element in the ~

core: assuming that the Earth had a composition io-3~- ,,,of enstatite chondrite, he calculated the propor- C N H~ ~F ci B, , ____________tion of troilite FeS in the core and found that it Fig. 5. Abundance of light elements in the Earth, referred to

would give the correct density deficit (but his chondritic abundance (after Rama Murthy and Hall, 1970).


7/19

J.-P. Poirier/Physics of the Earth and PlanetaryInteriors 85 (1994) 319—337 325

soc I I of the Fe—Ni—S system from 30 to 100 kbar and,

o using equations of state for Fe and for non-

1400 ~ stoichiometric Fe09S (King and Ahrens, 1973), he

- ~ I. extrapolated the densities of the compositions on

1188. the liquidus to the pressure of the ICB; compar-

• i ing these values with the seismically determined

density, he found an average composition of the915 outer core between 8 and 11 wt.% S.

Urakawa et al. (1987) studied the solubility of 800 41’2 sulphur and oxygen in the system Fe—S—O up to

Fe~ - ,~ 150 kbar. They found that the immiscibility gap in

Boo the liquid region narrows with increasing pres-• ~ sure, and suggested that it might disappear above

400 - 36.5 250 kbar. Addition of nickel reduces the immisci-

bility of the liquids. Urakawa et al. (1987) also

200 found that alloying with S and 0 reduces the38.8 interfacial tension between the metallic melt and

I I I silicates and oxides: at high pressure the metallicFe 10 20 30 40 50 melt wets the grain boundaries and forms a net-

wt % s work of liquid. Goarant et al. (1992) also foundFig. 6. Fe—S phase diagram at 1 atm (after Raghavan, 1988). complete wetting of grain boundaries of magne-

siowüstite and perovskite by Fe—O—S melt atpressures between 700 and 1300 kbar.

tionation, as meteorites exhibit no anomaly, nor The melting temperatures of sulphides andcould it be due to volatilization during accretion, Fe—lO wt.% S mixtures were measured in dia-

as the other volatiles would have been lost too (in mond anvil cells by Williams and Jeanloz (1990)

particular, i29Xe produced by the decay of short- up to about 1 Mbar and by Boehler (1992) up to

lived 1291) They thought it unlikely that sulphur about 500 kbar. Although the temperatures found

could be hidden in the lower mantle, as any by Williams and Jeanloz are at least 500°Chigherprocess that would have removed metallic Fe into than those of Boehler, there is agreement on the

the core would also have removed sulphur. The fact that the eutectic behaviour in the Fe—S sys-only remaining possibility was that sulphur was tern persists up to very high pressures and that

sequestered in the core during core formation, the Fe—10% S alloy melts several hundreds of thanks to the existence of a low melting point degrees below the melting point of pure iron orFe—S eutectic (Brett and Bell, 1969) (Fig. 6). of iron sulphide.

They found that a mixture of 40% Cl chondrites, Using solid-state and liquid-state physics mod-

50% ordinary chondrites and 10% iron mete- els and results of shock-wave and melting experi-orites provides a satisfactory composition for the ments, Svendsen et al. (1989) performed some-Earth, with a core containing 15 wt.% S (but a what involved calculations of the liquidus at high

mantle richer in FeO than pyrolite). pressure in the Fe—FeS system, assuming idealityThe idea that sulphur might segregate with of the liquid and complete immiscibility of Fe and

iron in the proto-core was quickly accepted (e.g. FeS in the solid state. They found that the corn-Lewis, 1971) and a great number of experimental position of a S-rich core is more likely to lie ondata on the melting of the Fe—FeS system under the Fe-rich side of the eutectic (which implies apressure, as well as on the equation of state of welcome depression of the freezing point of pure

sulphides, was obtained in the following years. iron). However, in a follow-up paper the sameUsselman (1975) investigated the pressure de- workers (Anderson et al., 1989) concluded thatpendence of the liquidus in the iron-rich portion solid solution between S and Fe is possible at


8/19

326 f-P. Poirier/Physicsof the Earth and Planetary Interiors 85 (1994) 319—337

high pressures and that the Fe—S system exhibits ~ % ocontinuous solid and liquid solutions (still with a 20000; 05 22 23 24

freezing point depression, as the melting point of LI / \

FeS is lower than that of Fe). The possibility of 1800L / \ L

11

high-pressure solid solution between S and Fe 1600 1523°

was confirmed by electronic structure calculations ° ~II\ by Boness and Brown (1990) and Sherman (1991). •‘~ 1371° 50.92 51.26

Shock-wave equations of state were deter- 1200 wustite

mined for pyrrhotite FE7S8 by Ahrens (1979) and 1000 FeO

Brown et al. (1984) and for pyrite FeS2 by Ahrens 800

and Jeanloz (1987); they were used, together with

equations of state for iron, to estimate the mixing 60 560° _____________ratio of sulphur and iron matching densities in 400 51.42the liquid outer core. The inferred sulphur con-tents (assuming sulphur is the only light element

200Fe 1 2 3 50 51 5 2 53

in the core) are in good agreement: 9—12 wt.% Al % 0(Ahrens, 1979), 10 ±4 wt.% (Brown et al., 1984) Fig. 7. Fe—O phase diagram at 1 atm (after Raghavan, 1988).and 11 ±2 wt.% (Ahrens and Jeanloz, 1987).

Brown and McQueen (1982), using the pyrrhotite

data from Ahrens (1979) and their own shock- density of the mixture, estimated from approxi-wave equation of state for c-Fe, found a smaller mate equations of state, was consistent with that

percentage of sulphur — 5—10 wt.% — essen- of the core. Bullen (1973), with a similar ratio-tially because they took into account the differ- nale, proposed that the outer core consisted of

ence in volume between solid and liquid iron. Fe2O (12.5 wt.% 0).

Although sulphur is a most convenient light However, oxygen as a major light element in

element to have in the core as it forms a eutectic the core did not really become fashionable untilwith iron at low pressures, is still soluble in iron Ringwood (1977) revived the idea of Dubrovskiyat high pressures and lowers its freezing point, and Pan’kov and argued that, instead of silicon

there has never been a very satisfactory answer to that he had favoured earlier, or sulphur (thean embarrassing problem (e.g. Ringwood, 1977): current contender at the time), FeO ought to beif sulphur, a very volatile element, were hidden in considered. The phase diagram of the Fe—FeO

the core in sufficient quantity to account for the system at atmospheric pressure exhibits a largedensity deficit, it would be less depleted in the liquid miscibility gap and the solubility of FeO inbulk Earth than several less volatile, non- molten iron is very small near the liquidus (Fig.siderophile elements, such as Na, K or Cl. 7). However, the solubility increases rapidly with

temperature, and Ringwood, following Dubrov-

skiy and Pan’kov (1972), suggested that FeO be-

4. Oxygen comes metallic at high pressure, and that conse-quently the miscibility gap between ionic and

Some 20 years ago, Dubrovskiy and Pan’kov metallic liquids should disappear at a pressure

(1972), as an alternative to the theory according estimated at about 300 kbar. Using a rough esti-to which the core was not iron but a high-pres- mate of the density of metallic FeO at coresure metallic phase of the mantle silicate pressures, Ringwood concluded that, to fit the(Ramsey, 1949), suggested that the iron outer seismic data, the core should contain about 44

core of the Earth might contain more than 50% wt.% FeO, equivalent to 10 wt.% 0.in mass of iron oxide FeO, metallized under Shock-wave experiments (Jeanloz and Ahrens,

pressure, that would have gravitationally segre- 1980) showed that Fe0 940 does indeed undergo a

gated from partially melted lower mantle; the phase transition, with density increase of 4% at


9/19

J.-P. Poirier/Physics of the Earth and Planetary Interiors 85 (1994) 319—337 327

about 700 kbar, and that the Hugoniot data were

consistent with the outer core containing about 2000

10 wt.% 0. However, Yagi et al. (1985) found no

large discontinuous density increase at 700 kbar “~in FeO statically compressed at room tempera- ~ 1900 Lm I

1875°

\ / L,,+Li

\ ~ture up to 1.2 Mbar. Knittle and Jeanloz (1986, L.

1991a) later showed that above 700 kbar the ~. i~oo

electrical conductivity of FeO becomes almost \ /1760’ V Eequal to that of iron (Fig. 8), and concluded that ~ Lm+ F.Octhe transition observed by Jeanloz and Ahrens 1700

Lm 1670(1980) is due to the ‘metallization’ of FeO and Fe5. Fe05

that there is complete miscibility of the Fe—FeO ______________________________liquids above 700 kbar, in reasonable agreement 0 20 40 60 80 100with the prediction of McCammon et al. (1983) Fe wt % FeObased on extrapolation of the Fe—FeO phase Fig. 9. Phase diagram for the system Fe—FeO at 160 kbardiagram at high pressures. (after Ringwood and Hibberson, 1990). Lm: metallic Fe—O

Sherman (1989) remarked that the ‘metalliza- melt, Li: ionic FeO—Fe melt, Fec: crystalline iron, FeOc:crystalline wiistite.

tion’ of FeO is not necessarily related to a struc-

tural change owing to a change of the characterof the Fe—0 bonding from ionic to metallic; it ismore likely that the sudden increase in electricalconductivity is due to a Mott transition (insula- see anything at low temperature, where delocal-tor—metal transition) consisting in a delocaliza- ization probably does not occur.tion of the Fe(3d) electrons over the metallic The solubility of FeO in liquid iron at highsublattice, while the Fe—O bonding remains ionic; pressures was investigated in a series of experi-this could explain why Yagi et al. (1985) did not ments at the Australian National University, first

in a piston cylinder apparatus up to 40 kbar

(Ohtani et al., 1984), then in a multi-anvil appara-tus at 160 kbar (Kato and Ringwood, 1989; Ring-wood and Hibberson, 1990, 1991); these experi-ments confirmed the increase of solubility of FeO

at high pressures (Fig. 9) and allowed an estimateWUstite (DAC) of the eutectic composition and temperature on

the Fe-rich side of the phase diagram (10 wt.%0, 1670°C), as well as of the depression of the

freezing point (about 28°Cper 1 wt.% 0 at 160

kbar).

Knittle and Jeanloz (1991b) measured the van-~ 10ation of the melting point of FeO up to 800 kbar

WUstite (Shock-wave)in a laser-heated diamond-anvil cell, and found

Fe that the slope of the melting curve increasesabove 700 kbar and that FeO melts at a highertemperature than Fe. At 830 kbar, the alloy of

IC’ ~3~’~ composition Fe069O031 is found to melt at aPressure (GPo) temperature intermediate between the melting

Fig. 8. Electrical resistivity measurements on Fe0 ~O under point of Fe and that of FeO. Knittle and Jeanlozpressure. Above 700 kbar, the resistivity becomes comparable concluded that at these high pressures, the eutec-with that of iron (after Knittle and Jeanloz, 1986). tic disappears to be replaced by a two-phase


10/19

328 f-P. Poirier/Physics of the Earth and Planetary Interiors 85 (1994) 319—337

spindle between solid and liquid solution, andthat, consequently, oxygen, far from depressing 4000

the freezing point of the alloy, would increase it. —FeO

present in the core to counteract this effect, but

Another element (e.g. sulphur) should then be (BoenIer,1992)\~~

// Fe:10~.%Fe0the overall depression should not be more than a 3000 ~Rin~00da~Hibirson, 1990few hundreds of degrees. However, Anderson etal. (1989) remarked that the fact that the melting ~point of an Fe—FeO alloy is intermediate be- 2000tween those of Fe and FeO does not necessarily

Fe:8wt%FeOrule out the existence of a eutectic (Fig. 10), such 7 8

as the one they calculated. In any case, the com-1000

position of an Fe—FeO alloy consistent with outer + Fe: 3Owt. % FeCcore density should fall on the FeO-nich side of

the phase diagram, leading to an enrichment of 0 0.2 0.4 0.6 0.8 1.0 1.2 1 .4 1.6 1.8 2 .0the solid phase (inner core) in oxygen during P ( Mbar)freezing.

On the basis of electronic structure calcula- Fig. ii. Melting curves of Fe, FeO and Fe—FeO alloys (afterBoehler, 1993).

tions, the existence of a solid solution between Feand 0 at high pressures is also thought to be

unlikely (Boness and Brown, 1990; Sherman, seems therefore to be no disagreement as to the1991). fact that, at core pressures, oxygen does not lower

Boehler (1992, 1993) investigated melting of the freezing point of iron.FeO and Fe—FeO alloys at pressure up to 580 Knittle and Jeanloz (1991b) and Goarant et al.

kbar, in a laser-heated diamond-anvil cell. Al- (1992) investigated the reaction between lower-

though he found values of the melting point of Fe mantle material (silicate perovskite and magne-and FeO much lower than those of Knittle and siowüstite) and molten iron above 700 kbar. They

Jeanloz, he did find, like them, that FeO melts at found that liquid iron infiltrated between the

a higher temperature than Fe, and that there was grains of the solid phase and dissolved FeO from

no significant difference between the melting it. Analytical transmission electron microscopy of

points at high pressure of iron and Fe—8 wt.% the samples after reaction in the laser-heatedFeO and Fe—30 wt.% FeO alloys (Fig. 11). There diamond-anvil cell (Goarant et al., 1992) showed

depletion of the magnesiowüstite in FeO andcomplete wetting of the grain boundaries, in

agreement with the results at lower pressure of

Urakawa et al. (1987).There is therefore little doubt that molten iron

of the core and oxides of the lower mantle react

at high pressure, thus enriching the core metal in~id_Solid oxygen. Knittle and Jeanloz (1991b) even claimed

that ‘it may be that the entire budget of light

alloying component in the outer core has come

________________ ________________ from chemical reactions with the mantle’. This,Fe FeO Fe reQ

however, seems unlikely, for the core—mantle re-Fig. 10. Schema showing that a value intermediate between action is a very high pressure process and therethe melting points of Fe and FeO for the melting point of an

must surely have been some light elements intro-alloy Fe—FeO (dot) can be compatible with a eutectic (right)as well as with a solid solution (left) (after Anderson et al., duced at low pressure during the early stages of 1989). core formation.


11/19

f-P. Poirier/Physics of the Earth and PlanetaryInteriors 85 (1994) 319—337 329

5. Hydrogen

6000

Hydrogen, although a possible candidate for P 100GPa

lowering the density of the core, was long ne-

glected for two reasons: it was thought that it 5000

would escape during accretion (e.g. Jeanloz, L1990), and Ringwood (1966) had argued that, as 4000hydrogen enters into interstitial sites in iron, itwould not significantly decrease the density of 6 .~L L+LH

2

the core. As to the last point, Stevenson (1977) 3000mustered evidence to the contrary: interstitial E —hydrogen expands the host metal; he also calcu- 2000 ‘~ C + Li2

lated that hydrogen might form an FeH hydride / at the pressure of only a few kilobars. About 1 1000 I wt.% of hydrogen would be enough to account / ~ 4’SH2for the density deficit of the core and this quan -______________ _____________tity would be provided if only 10% of the accret-ing material was a low-temperature condensate Fe Fe H H

containing water, and if the water reacted with Fig. 12. Proposed phase diagram of the Fe—H system at 1Mbar (after Fukai, 1992).Fe at a pressure such that H could enter insolution and be retained. Stevenson concluded

that hydrogen was most probably one of the ele- Fukai and Suzuki (1986), using two different

ments contributing to the density deficit of the models of compressibility of the light elements tocore (Stevenson, 1977), but later eliminated it on fit the observed density deficit. They proposedthe basis that it was too insoluble (Stevenson, two core compositions, corresponding respec-

1981). tively to 10% and 15% of the low-temperature

All the work on hydrogen in Fe was done condensate in the accreting material: FeH041recently in Japan and at the Geophysical Labora- C0050013S003 and FeH063C007O023S005. It

tory of the Carnegie Institution of Washington. should be noted that Si was omitted withoutSuzuki et al. (1984) studied the reaction between justification.enstatite, iron and hydrous minerals (simulating a Badding et al. (1991, 1992) investigated the

mixture of enstatite and Cl chondrites) at 50 kbar reaction between iron and hydrogen in situ in aand temperatures between 1000 and 1200°C. diamond-anvil cell up to 620 kbar, at room tern-

From the presence of olivine (containing FeO) perature, using X-ray diffraction with synchrotronand iron in the reacted product, they indirectly radiation. They noticed a sudden increase in vol-inferred the presence of hydrogen in the iron, on ume of iron at 35 kbar, corresponding to thethe basis of the reaction formation of the FeH hydride, which they found

Fe + water —~FeH + FeO stable up to the maximum pressure investigated,X and they determined its structure and its equa-

The metal had melted 500°Cbelow the melting tion of state. They also performed thermody-

point of pure iron. namic calculations relative to the reactionFukai and Suzuki (1986), using calculated or

(2+x)Fe+H20-*2FeH+FeOestimated values of the atomic volumes of various xlight elements (H,C,O,S) at high pressures, calcu- and found that it was favoured at high tempera-lated their solubility in iron and the resulting tune and pressure. From the experimentally de-

decrease in density. Fukai (1992) proposed a termined equation of state, they found that thehigh-pressure phase diagram for the Fe—H sys- core density deficit can be accounted for by more

tern (Fig. 12) and re-examined the results of than 40 mol % hydrogen.


12/19

330 f.-P. Poirier/Physicsof the Earth and Planetary Interiors 85 (1994) 319—337

6. Carbon retained in appreciable quantities during accre-tion. He allowed 1 wt.% C at most in the core.

Although Birch (1952) mentioned it, carbon The only serious investigation of carbon in thewas only occasionally given lip service. Carbon is core is very recent. Wood (1993) first addressedavailable in the accretional material and i t is the question of carbon volatility, and showed that

soluble in iron at low pressure (4.3 wt.% at the 1 the volatility of carbon could be greatly reducedatm eutectic point). However, Ringwood (1966) by even the modest pressures (less than 50 kbar)eliminated it for the same reason as hydrogen, at which the differentiation of the Earth’s core isbecause it formed interstitial solid solutions with thought to occur. He calculated the solubility of iron. Brett (1976) even thought that it could in- carbon in liquid Fe in equilibrium with the gas

crease the density of iron. This reasoning does produced from Cl carbonaceous chondrites and

not take into account the expansion that can found that, between 1 and 10 kbar, carbon would

occur around an interstitial, and indeed carbon, enter liquid iron at concentrations between 2 and

albeit modestly, lowers the density of liquid iron 4 wt.%. He then proceeded to calculate the phasenear its melting point (Ogino et al., 1984). Ring- diagram of the Fe—C system at high pressures,wood (197) eventually recognized that carbon using thermodynamic properties of y-Fe, C, Fe,

could decrease the density of iron at high pres- Fe—C liquids and Fe3C (the equation of state of

sure, but pointed out that it was too volatile to be Fe3C, for which there are no experimental data,

300c

2400~ 2600 /. —-.

Grapt~iIe.

2200 1~” 2000 Liquid Liquid

• ~oo _______ Fe3C,

1400 , F.3C. , Fe, Fe3C

Fe Fe3C C ash Ic

looc 1200

(a) S 2 w.i:ht % C:,bon 8 tO (b) 0 2 Weight % CirbOti 8 tO

2600 , /

5200 Liquid (L)2400 Ca,bon

____ 00~—

(c) Weight % Carbon (d) Weight % Carbon

Fig. 13. Phase diagram of the system Fe—C at 1 bar (a), 50 kbar (b), 150 kbar (c) and 1.36 Mbar (d) (after Wood, 1993).


13/19

f-P. Poirier/Physics of the Earth and Planetary Interiors 85 (1994) 319—337 331

was estimated), as well as constraints from exper- grounds’. Brett’s calculations of reaction kinetics

iments at 30 and 50 kbar. It was found that, and partition coefficients with new data (and

between atmospheric pressure and 150 kbar (Fig. assumed values of certain thermodynamic quanti-13), the stability field of the carbide Fe

3C in- ties) tended to show that a disequilibrium statecreases dramatically, and the eutectic composi- would be unlikely during core formation, but they

lion shifts to even lower carbon concentrations were contested by Ringwood (1971). Urakawa(2.1 wt.% C at 150 kbar against 4.3% at 1 bar); (1991) experimentally investigated the partitionextrapolation to the pressure of the core—mantle coefficient of nickel between magnesiowüstite andboundary shows an enhancement of these fea- Fe-rich metal up to 170 kbar and 2200°C, and

tunes. It therefore appears that carbon cannot used the results to estimate the partition of nickelcontribute more than half of the budget of light between metal and bulk mantle silicate. He foundelements in the core, if melt segregation occurs at that at high pressure, Ni transfers into the silicatelow to moderate pressures. Wood then investi- from the iron alloy; but even though he advo-

gated the Fe—C—S system and found that the cated nickel equilibrium partitioning, he still

miscibility gap gradually closes with increasing needed a specific model of mantle differentiationtemperature and pressure: at core temperatures, to account for the nickel concentration in themost of the Fe—S—C liquids with a composition present upper mantle.consistent with the density deficit of the core (2) To account for the Fe

3 ~/Fe2 + ratio of

would form one stable liquid. Furthermore, the fresh basaltic glasses and penidotites, a pyrolite

extrapolations yield the result that, for even very mantle should have a ratio Fe3~/Fe2~~0.05—0.1,low values of the C/S ratio, the first phase to whereas it is expected to be at least an order of crystallize would be the carbide Fe

3C, thus lead- magnitude lower for metal silicate equilibrium

ing Wood to the conclusion that the inner core (Ringwood, 1966). Brett (1971) suggested that theprobably consists of Fe3C. rocks of the upper mantle had had time to re-

equilibrate at higher crustal oxygen fugacities, a

conclusion again disputed by Ringwood (1971).7. Core—mantle equilibrium or disequilibrium? McCammon (1993), using Mössbauen spec-

troscopy, experimentally determined the Fe3 + /

The controversy about core—mantle equilib- EFe ratio in magnesiowüstite in equilibrium withnium or absence thereof essentially hinges on iron up to 180 kbar, and found that it decreasesthree points: the abundances of siderophile ele- with increasing pressure. Extrapolation of herments in the upper mantle, the state of oxidation results to conditions at the top of the lower

of the mantle and the segregation process of the mantle suggest values of the ratio Fe3 7EFecore liquid, lower than 0.05 for equilibrium, too low to be(1) The abundance of siderophile elements consistent with electrical conductivity measure-

(Ni, Co, Au, Pt, etc.) in the mantle, as deter- ments on lower-mantle material, which point tomined from analyses of upper-mantle penidotites, conduction by charge transfer between Fe3 + and

is much higher than the concentrations expected Fe2~(see e.g. Shankland et al., 1993); McCam-from equilibrium partitioning between mantle and mon concluded that it is unlikely that the present

core, assuming solar composition for the bulk lower mantle is in equilibrium with the core.Earth (Ringwood, 1966) and assuming the upper The problems of the discrepancy of themantle is representative of the whole mantle. For siderophile abundances and the oxidation degree

instance, nickel concentration in the mantle ma- of the lower mantle have usually been avoidedterial is about 2000 ppm, whereas 10—100 ppm rather than resolved by devising specific modelswould be expected for equilibrium partitioning of core formation and differentiation (Jones and

(Urakawa, 1991). The conclusion that core and Drake, 1986; Newsom and Sims, 1991) or of inho-mantle were in gross disequilibrium appeared to mogeneous accretion of various proportions of Brett (1971) ‘to be highly unlikely on intuitive reduced and oxidized condensates (Ringwood,


14/19

332 1.-P. Poirier/Physicsof the Earth and Planetary Interiors 85 (1994) 319—337

1977; Wänke, 1981; Wänke and Dreibus, 1988). to interfacial energy, is also reduced in about theAs pointed out by Newsom and Sims (1991), same proportions by oxygen (Fig. 14) and by

there is a correlation between the degree of suc- sulphur: 1 wt.% S reduces the surface energy by a

cess of a model and the number of adjustable factor of three (lida and Guthrie, 1988). Carbon,parameters. however, does not seem to have any significant

(3) A somewhat different question, of a less effect. There are no data on silicon in iron, butgeochemical nature, is also relevant to the equi- the value of the surface energy of liquid silicon (a

librium vs. disequilibrium controversy: did the metal) is 0.865 J m2, lower than that of pure

segregating liquid metal gather into large blobs liquid iron: 1.872 J m2 (lida and Guthrie, 1988).that sank rapidly through the slag without having As it is a rule of thumb that solutes with a lowthe time to equilibrate (e.g. Stevenson, 1981), or surface energy tend to lower the surface energydid the cone fluid wet the grain boundaries of the of the solution, it is reasonable to expect thatsilicate and equilibrate while percolating down silicon will lower the surface energy of liquid

(e.g. Arculus et al., 1990)? The answer depends iron.

on the nature of the solute light element and its At high pressures, we have direct observa-

influence on the interfacial energy between ox- tional evidence that liquid iron, containing oxy-ides and liquid iron. gen in greater concentrations than at 1 atm, as

There are very few measurements of interfa- well as sulphur, completely wets grain boundaries

cial energy between solid oxides and liquid met- of oxides (Urakawa et al., 1987; Goarant et al.,

als, but they indicate that alloying of iron with 1992).light elements may lower the interfacial energy: I therefore think that it is relatively safe to

only 1 at.% (less than 0.3 wt.%) of oxygen in assume that, if the cone liquid contains severalliquid iron lowers the interfacial energy with alu- light elements (a reasonable supposition as wemina from 2.4 to 0.6 J ~ 2 (Chaklader et al., have seen), some of these will lower the intenfa-1981). Surface energy of liquid iron, a good guide cial energy with solid oxides enough for the liquid

to wet the grain boundaries and percolate downas thin intergnanular films, at low as well as athigh pressures. It follows that it is possible that

local equilibrium is achieved between the segre-gating metal and the solid slag. This may not be

the case if the silicate is molten: drops of metal1.8 might then be able to sink rapidly enough to

—0 0— .

Fe-C prevent local equilibrium.1.6 ‘•°••\ Local equilibrium, however, even if it is

1 4 \ achieved, in no way implies that the cone and themantle are, on ever were, in equilibrium. The

1 2 \‘\ liquid percolating down samples a whole range of temperatures, pressures and oxygen fugacities (as

Fe-O already pointed out by Ringwood (1959)), and theresulting segregated cone is certainly not in equi-

~ 0.81- Fe-S librium with any part of the mantle, let alone theupper mantle.

0.6 - Although abundances of elements and oxida-tion state measured on upper-mantle rocks must

0.4 0001 0 01 1 be taken into account in models of Earth accre-0.0005 - Weight % addition lion and differentiation, I do not believe that

Fig. 14. Effect of C, S and 0 on the surface tension of liquid these data can provide useful constraints on theFe (after lida and Guthrie, 1988). nature of the light elements in the outer core.


15/19

J.-P. Poirier/Physicsof the Earth and Planetary Interiors 85 (1994) 319—337 333

8. Conclusion (3) The conclusion that there are several lightelements in the outer core seems inescapable. It

(1) The density deficit of the core can be is not even obvious that one element should be

accounted for by most of the light elements con- particularly dominant.sidered — Si,S, 0, H, C. However, each element The best constraints on the proportions of

is subject to metallurgical constraints and is corn- different elements compatible with the geophysi-patible with only a limited class of Earth accne- cal data (and on the composition of the innertion or core formation models: cone) are, in my opinion, of a metallurgical na-

(a) Silicon is available and is soluble in iron at ture: phase diagrams of iron-rich (and iron—low (and very probably high) pressures; it lowers nickel-rich) ternary and quaternary systems

the melting point of iron. It is compatible with should be experimentally determined up to highearly core formation, but it demands very neduc- pressures and/on calculated, using thermody-ing conditions during accretion, or accretion of namic data and equations of state.

an already reduced material.

(b) Sulphur is available and is soluble in ironat low and high pressures; it lowers the melting Acknowledgements

point of iron. It is compatible with early cone

formation. However, geochemical questions con- I gratefully acknowledge fruitful discussions

cerning the abundance of sulphur are still pend- with Francois Guyot. I thank Claude Allègre,ing. Francois Guyot, Jean-Louis Le Mouël and David

(c) Oxygen is available, but it is soluble in iron Price for reading the manuscript and providing

in reasonable quantities only at high pressures. It useful comments. This work was partly supportedis incompatible with early core formation, and by CNRS (URA 734). This is IPG Contribution

can be introduced into the core by reaction with 1312.the mantle only when the Earth has reached alarge size. Also, oxygen does not significantly

lower the melting point of iron and would not be Appendix: Calculation of the density deficit (withreleased during crystallization of the inner core. respect to pure Fe) of a solution of light elements

(d) Hydrogen is available and might be re- in iron

tamed at low or moderate pressures by formingan iron hydride. It probably would lower the Most workers announce that the density deficitmelting point of iron, of the core is accounted for by a given mass

(e) Carbon is available, but it is not soluble fraction of light element, but do not generallyenough in iron (even at high pressures) to ac- explain how this result was obtained. Although

count for the whole density deficit of the core. It the calculation is simple in principle, I thought itlowers the melting point of iron. may be useful to give it here, if only to show that

(2) Most of the light elements considered lower it is not independent of the equation of state of

the interfacial energy between liquid iron and the light element. Instead of the light element, it

solid slag. The segregating core fluid probably may be more convenient to consider a light-ele-percolates down, achieving local equilibrium with ment-rich end-member (e.g. FeO or FeS) whose

the solid mantle at temperature and pressures equation of state is known.varying with depth. If the silicate is molten, even Let us consider a solution of a light element Xlocal equilibrium might not be reached. Hence, it in Fe. Let PFe be the density of pure iron at coreis very likely that the core is not in equilibrium pressures and temperatures and p~the density of

with the upper (or lower) mantle. However, es- light element in the same conditions (from equa-

tablishing whether the core is or is not in equilib- tions of state). Let mFe, VFe and m~,v~be therium with the mantle affords no constraint on the molar masses and partial molar volumes of Fe

nature of the light elements. and light element in the core, respectively (con-


16/19

334 f-P. Poirier/Physicsof the Earth and Planetary Interiors 85 (1994) 319—337

sidering partial molar volumes takes into account and Eq. (A6) becomes

the excess volume, which may not be negligible;however, in view of the fact that the available — 1) = 0.11 (A8)data usually are relative to an end-member and

not to the element in solution, the excess volume As an example, let us consider successively the

is neglected and v,.~is the molar volume of the cases where the light elements are S, 0 and Si.light end-member). The mass fraction of the light Using the equations of state for s-Fe (Brown and

element X is McQueen, 1982), FeO (Jeanloz and Ahrens,

1980), FeS (Brown et al., 1984) and Fe—20 wt.%Ic = mFe + m~ (Al) Si (Balchan and Cowan, 1966), we obtain from

(A6) approximate values of the mass fraction of

The density of the core fluid is light element compatible with a core densitydeficit of 10%:

mFe +m~ (A2) PFe

VFe + 1.3 hence fFeO 40 wt.%,P FeO

We haveand f

0~9wt.%

—=——(l—f~)+—f~=——+f~(—— PFe 137 hencefFes 3Owt%,1

1 1 1 /l

1)P PFe Px PFe \P~ P~e PFeS

(A3) and f~~llwt.%

and PFe 1.12 hence fFe-20S 40 wt.%,

=f~~(—— (A4) and f~1~l8wt.%/ 1 1 ) PFe—2OSj

PFe \Px PFe

The density deficit for various mass fractionsCarrying p from (A3) into (A4), we obtain of each element alone is shown in Fig. Al. The

fact that the curve for silicon is below the curves1 PFe

— ~ ) for oxygen and sulphur (instead of between them— 1

(AS) as might be surmised, as the atomic weight of SiPFe f~(~— 1) + 1 is intermediate between those of 0 and 5) is

Px

Taking ‘~P/P~e = 0.01, we obtain the simple 0.25formula

0.11 0.2________ C)

f~= ~ (A6) .~—l

>. 0.15Px

(0C

sity p~(at core conditions) and mass fraction f~ ~ -If there are several light elements X~,of den- ~ 0 1VEq. (AS) becomes 0 0.05

C.)

0~i~

‘~P— PX / (A7) 0 0.05 0.1 0.15 0.2 0.25Mass fraction of light element

PFe 1) + 1 Fig. Al. Core density deficit as a function of the mass fraction px of oxygen, sulphur and silicon.


17/19

1.-P. Poirier/Physics of the Earth and Planetary Interiors 85 (1994) 319—337 335

________________________________________ sure chemistry of hydrogen in metals: in-situ study of iron0.2 I hydride. Science, 253: 421—424.

Badding, J.V., Mao, H.K. and Hemley, R.J., 1992. High-pres-‘ö5 sure crystal structure and equation of state of iron hy-.._ 0.150

dride: implications for the Earth’s core. In: Y. Syono andM.H. Manghnani (Editors), High-pressure Research: Ap-C

0

plication to Earth and Planetary Science. Terrapub, Tokyo,

~43J0.1 pp. 363—371.Balchan, A.S. and Cowan, G.R., 1966. Shock compression of en

two iron—silicon alloys to 2.7 Megabars. J. Geophys. Res.,Ct) 0.05(0 71: 3577—3588.

Birch, F., 1952. Elasticity and constitution of the Earth’s

_______________________________________ interior. J. Geophys. Res., 57: 227—286.0

Birch, F., 1961. Composition of the Earth’s mantle. Geophys.0 0.05 0.1 0.15

Mass fraction of s J.R. Astron. Soc., 4: 295—311.Birch, F., 1964. Density and composition of mantle and core.

Fig. A2. Mass fractions of Si, S and 0 giving a core density J. Geophys. Res., 69: 4377—4388.

deficit of 10%.Boehler, R., 1992. Melting of the Fe—FeO and the Fe—FeS

systems at high pressure: constraints on core temperature.

Earth Planet. Sci. Lett., 111: 217—227.Boehler, R., 1993. Temperatures in the Earth’s core from

probably due to the large negative excess volume melting point measurements of iron at high static pres-

of silicon in solution in liquid iron: —

36% for the sures. Nature, 363: 534—536.composition FeSi (Wilson, 1965), which causes Boness, D.A. and Brown, J.M., 1990. The electronic band

the density of the solution to be greater than that structure of iron, sulfur and oxygen at high pressures andthe Earth’s core. J. Geophys. Res., 95: 21721—21730.

ofan ideal solution. If silicon, sulphur and oxygen Braginsky, SI., 1964. Magnetohydrodynamics of the Earth’s

were simultaneously present, the composition of core. Geomagn. Aeron., 4: 698—712.

the alloys giving a core density deficit of 10% Brett, R., 1971. The Earth’s core: speculations on its chemicalcould be found from (A8) (Fig. A2). equilibrium with the mantle. Geochim. Cosmochim. Acta,

35: 203—221.

Brett, R., 1976. The current status of speculations on the

composition of the core of the Earth. Rev. Geophys. Space

References Phys., 14: 375—383.Brett, R., 1984. Chemical equilibrium of the Earth’s core and

Ahrens, TJ., 1979. Equations of state of iron sulfide and upper mantle. Geochim. Cosmochim. Ada, 48: 1183—1188.constraints on the sulfur content of the Earth. J. Geophys. Brett, R. and Bell, P.M., 1969. Melting relations in the Fe-richRes., 84: 985—998. portion of the system Fe—FeS at 30 kbar pressure. Earth

Ahrens, T.J. and Jeanloz, R., 1987. Pyrite shock compr ession, Planet. Sci. Lett., 6: 479—482.isentropic release and composition of the Earth’s core. J. Brown, J.M. and McQueen, R.G., 1982. The equation of stateGeophys. Res., 92: 10363—10375. for iron and the Earth’s core. In: S. Akimoto and M.H.

Alder, B.J., 1966. Is the mantle soluble in the core? J. Geo- Manghnani (Editors), High-Pressure Research in Geo-phys. Res., 71: 4973—4979. physics. D. Reidel, Dordrecht, pp. 611—623.

Allègre, C.J., Dupré, B. and Brévart, 0., 1982. Chemical Brown, J.M., Ahrens, T.J. and Shampine, D.L., 1984. Hugo-aspects of the formation of the core. Philos. Trans. R. Soc. niot data for pyrrhotite and the Earth’s core. J. Geophys.London, Ser. A, 306: 49—59. Res., 89: 6041—6048.

Allègre, CJ., Poirier, J.P., Humler, E. and Hofmann, A., Bullen, K.E., 1973. Cores of terrestrial planets. Nature, 243:

1994. The chemical composition of the Earth in prepara- 68—70.

tion. Chaklader, A.C.D., Gill, W.W. and Mehrotra, S.P., 1981.

Anderson, W.W., Svendsen, B. and Ahrens, T.J., 1989. Phase Predictive model for interfacial phenomena betweenrelations in iron-rich systems and implications for the molten metals and sapphire in varying oxygen partial pres-

Earth’s core. Phys. Earth Planet. Inter., 55: 208—220. sures. In: J. Pask and A. Evans (Editors), Surfaces and

Arculus, R.J., Holmes, R.D., Powell, R. and Righter, K., i990. Interfaces in Ceramics and Ceramic—Metal Systems,

Metal—silicate equilibria and core formation. In: H.E. Plenum, New York, pp. 421—431.

Newsom and J.H. Jones (Editors), Origi n of the Earth, Dubrovskiy, V.A. and Pan’kov, V.L., 1972. On the composi-Oxford University Press, Oxford, pp. 251—271. tion of the Earth’s core. Acad. Sci. USSR, Phys. Solid

Badding, J.V., Hemley, R.J. and Mao, H.K., 1991. High-pres- Earth (transl.), 7: 452—455.


18/19

336 1.-P. Poirier/Physicsof the Earth and Planetary Interiors 85 (1994) 319—337

Fukai, Y., 1992. Some properties of the Fe—H system at high Lewis , J.S., 1971. Consequences of the presence of sulfur in

pressures and temperatures and their implications for the the core of the Earth. Earth Planet. Sci. Lett., 11: 130—134.

Earth’s core. In: Y. Syono and M.H. Manghnani (Editors), Loper, D.E., 1978. The gravitationally powered dynamo. Geo-

High-Pressure Research: Application to Earth and Plane- phys. J. R. Astron. Soc., 54: 389—404.tary Science. Terrapub, Tokyo, pp. 373—385. MacDonald, G.J.F. and Knopoff, L., 1958. On the chemical

Fukai, Y. and Suzuki, T., 1986. Iron—water reaction under composition of the outer core. Geophys. J.R. Astron. Soc.,

high pressures and its implication in the evolution of the 1: 284—297.Earth. J. Geophys. Res., 91: 9222—9230. Mao, H.K., Wu, Y., Chen, L.C., Shu, J.F. and Jephcoat, A.P.,

Goarant, F., Guyot, F., Peyronneau, J. and Poirier, J.P., 1992. 1990. Static compression of iron to 300 GPa and Fe0 8Ni02

High-pressure and high-temperature reactions between sil- alloy to 260 GPa: implications for composition of the core.

icates and liquid iron alloys, in the diamond anvil cell, J. Geophys. Res., 95: 21737—21742.

studied by analytical electron microscopy. J. Geophys. Mason, B., 1966. Composition of the Earth. Nature, 211:Res., 97: 4477—4487. 616—618.

Herndon, J.M., 1979. The nickel silicide inner core of the McCammon, C., 1993. Effect of pressure on the composition

Earth. Proc. R. Soc. London, Ser. A, 368: 495—500. of the lower mantle end member Fe,O. Science, 259:

lida, T. and Guthrie, R.I.L., 1988. The Physical Properties of 66—68.

the Liquid Metals. Clarendon Press, Oxford, 287 pp. McCamrnon, C.A., Ringwood, A.E. and Jackson, I., 1983.Jacobs, J.A., 1987. The Earth’s core 2nd edn. Academic Press, Thermodynamics of the system Fe—FeO—MgO at high

London. pressure and temperature and a model for formation of

Jacobs, J.A., 1992. Deep interior of the Earth. Chapman and the Earth’s core. Geophys. J.R. Astron. Soc., 72: 577—595.

Hall, London. Newsom, H.E. and Sims, K.W.W., 1991. Core formation dur-

Jeanloz, R., 1979. Properties of iron at high pressures and the ing early accretion of the Earth. Science, 252: 926—933.

state of the core. J. Geophys. Res., 84: 6059—6069. Ogino, K., Nishiwaki, A. and Hosotani, Y., 1984. Density of Jeanloz, R., 1990. The nature of the Earth’s core. Annu. Rev, molten Fe—C alloys. Nippon Kinzoku Gakkaishi, 48:

Earth Planet. Sci., 18: 357—386. 1004—1010.Jeanloz, R. and Aiirens, T.J., 1980. Equations of state of FeO Ohtani, E., Ringwood, A.E. and Hibberson, W. , 1984. Com-

and CaO Geophys. J.R. Astron. Soc., 62: 505—528. position of the core: II. Effect of high pressures on theJepheoat, A. and Olson, P., 1987. Is the inner core of the solubility of FeO in molten iron. Earth Planet. Sci. Lett.,

Earth pure iron? Nature, 325: 332—335. 71: 94—103.

Jones, J.H. and Drake, M.J., 1986. Geochemical constraints Oversby, V.M. and Ringwood, A.E., 1971. Time of formation

on core formation in the Earth. Nature, 322: 221—228. of the Earth’s core. Nature, 234: 463—465.Kato, T. and Ringwood, A.E., 1989. Melting relationships in Raghavan, V., 1988. Phase diagrams of ternary iron alloys.

the system Fe—FeO at high pressures: implications for the Part 2: Ternary systems containing iron and sulphur. In-

composition and formation of the Earth’s core. Phys. dian Institute of Metals, Calcutta.Chem. Minerals, 16: 524—538. Rama Murthy, V. and Hall, H.T., 1970. The chemical compo-

King, D. and Abrens, T.J., 1973. Shock compression of iron sition of the Earth’s core: possibility of sulfur in the core.

sulfide and the possible sulfur content of the Earth’s core. Phys. Earth Planet. Inter., 2: 276—282.

Nature Phys. Sci., 243: 82—84. Rama Murthy, V. and Hall, H.T., 1972. The origin and

Knittle, E. and Jeanloz, R., 1986. High pressure metallization chemical composition of the Earth’s core. Phys. Earthof FeO and implications for the Earth’s core. Geophys. Planet. Inter., 6: 123—130.Res. Lett., 13: 1541—1544. Ramsey, W.H., 1949. On the nature of the Earth’s core. Mm.

Knittle, E. and Jeanloz, R., 1989. Simulating the core—mantle Nat. R. Astron. Soc. Geophys. Suppl., 5: 409—426.

boundary: an experimental study of high-pressure reac- Ringwood, A.E., 1959. On the chemical evolution and density

tions between silicates and liquid iron. Geophys. Res. of planets. Geochim. Cosmochim. Acta, 15: 257—283.

Lett., 16: 609—612. Ringwood, A.E., 1961. Silicon in the metal phase of enstatiteKnittle, E. and Jeanloz, R., 1991a. The high-pressure phase chondrites and some geochemical implications. Geochim.

diagram of Fe0 940: a possible constituent of Earth’s core. Cosmochim. Acta, 25: 1—13.

J. Geophys. Res., 96: 16169—16180. Ringwood, A.E., 1966. Chemical evolution of terrestrial plan-

Knittle, E. and Jeanloz, R., 1991b. Earth’s core—mantle ets. Geochim. Cosmochim. Acta, 30: 41—104.boundary: results of experiments at h

poirier 1994 pepi

Documents