当前位置:首页 >> 天文/地理 >>

Iron-Water Reaction under High Pressure and Its Implication in the Evolution of the Earth


Iron-Water Reactionunder High Pressure and Its Implication
in the Evolution of the Earth
Yuh Fukai

Department Physics, of ChuoUniversity, Bunkyo-ku, Tokyo
Toshihiro Suzuki

Institute SolidState for Physics, University Tokyo,Minato-ku of

Wehave establishedexperiments thermodynamical by and calculations a large that amount hydroof

gen bedissolved under hydrogen can iniron high pressure, causing appreciable reductiondensity of and melting point. More recent experiments shown thereaction Fewith have that of hydrous silicates under high pressure temperaturesGPa,<_ and (_< 5 1200?C, min) 20 yields hydride iron (Felix), olivine pyroxand ene, thatcoagulation particlesgreatly and ofiron is facilitatedthepartially in molten ofthesilicates. state
Onthebasis these of experimental findings Fe- H and - H20systems, propose theFe- H20 on Fe we that reaction should played crucial in theevolution theearth, have a role of including particular corein the mantle separation dissolutionhydrogen theearth's and of in core. tentative A calculation themass on flowin theevolution process made assuming was by a 10:90 mixture low-temperature andhighof (LT)

temperature components primordial (HT) forthe material Fe/(Fe Mg)=0.11forthe and + whole mantle. Thecomposition core deduced, of the thus together compressibilityat high with data pressures, isfound to give reasonable valuesfor the densitydeficit of the outer core. If, conversely, we adoptthe density reduction input,the composition the outercoreis estimated be as of to
FeHo.25Co.o5Oo.13So.o3. Finally, some implications thepresent of picture other on terrestrial planets are
examined briefly.



Recently, discovered thesolubility hydrogen iron we that of in is enhanced drastically under high pressures suggested and on this basisthat a large amount of hydrogenmay have been accumulated in the earth's core in its evolution. In this paper, we

reexamine this hypothesis the presence large-scale in of melting of silicates growingterrestrialplanets,an important on recognition hasbeenderived that from lunarexploration, and therebywishto demonstrate a consistent that pictureof planetary evolutioncan be naturally constructed. Among the threemajor elements terrestrial of planets,Si, Mg, and Fe, the behaviorof Fe in the evolution process beis lievedto be rather uniqueand it thereforedeserves special consideration.The uniqueness arisesfrom the fact that its redox statecanbe easilychanged with environment alsofrom its and ability to collectsiderophile elements the core formation in

Fe and H20) wouldhaveevaporated duringpassage through theatmosphere, asa consequencethereaction + H20 and of Fe -? FeO+ H2, practically of the waterwouldhavebeenlost all into space the form of H2 gas.This is highlyimprobable, in however, because is generally it agreed that the average of size accreting planetesimals --- 10km asa result gravitationwas of
al instabilities smaller-sized for ones[Safronov, 1972;Goldreich and Ward, 1973;Adachi et al., 1976;Greenberg al., 1978]. et In this situation,the evaporated masscan be estimated be to

negligibly small,however thicktheatmosphere be. Lange may andAhrens [1982a] haveperformed impact-dehydration experiments some on hydrous minerals showed particular and in that impactgenerated wateratmosphere become may comparable
in weightto the present-dayoceanicwater [Langeand Ahrens,

1982b,1984].Abe and Matsui [1985]havegonea stepfurther
and showed that as a result of the blanketing effect of the

process. discussed As below,these two features Fe should of
have been of crucial importancein many phasesof planetary evolutionthat ultimately led to the core-mantledifferentiated

impact-generated water atmosphere, surfacetemperature the of the growing earthwasraised above solidus silicates, the of andmelting silicates of should haveoccurred largescale on for theentireaccretion period.Theirconclusion large-scale of melting of silicates of particular concernto our discussion is because

It is believedthat a greaterproportion of Fe was accreted the solubilityof water in silicates increased is abruptly onceparin metallic form, and a primary causefor the changeof its redox state was its reaction with water. Thus the redox state of

tial melting starts. It follows then that almost all the accreting water can be retained in solution in molten silicates. We may Fe is determinedessentially the amount of accretedwater. by expected therefore that the Fe- H20 reaction shouldhave ocOnly 1?7o waterin the primordialmaterialof the earth can of curred rather extensivelyin molten silicates and affected the be shownto produceFeO comparable the FeO contentof to evolutionary processin a profound manner. the mantle. In regard to dissolutionof siderophileelementsin metallic There have beenconflictingviewson the degreeof retention Fe, it may be recalled that the density of the earth's core is of water in the accretionprocess.Ringwood [1978, 1979] sugsmallerby -- 10?7o than pure Fe, and this densitydeficithasbeen gested a greater that partof theprimordial material (including ascribedto the presenceof some light elementsin molten FeNi alloy. As a candidate for the light elements,S was consiCopyright 1986by the American Geophysical Union. dered to be most plausible. The Fe-S systemwas thoroughly investigated Usselman[1975a,b], and the solubility of S in by Paper number 5B5549. molten Fe is known to be large at all pressures.More recent 0148-0227 / 86/ 005 B- 5549505.00



experiments the Fe-O system Ohtaniand Ringwood on by [1984],Ohtaniet al. [1984]and by M. Kato (unpublished
manuscript, 1986) have shown thesolubility O in molten that of

Feisgreatly increased under high pressures, strongly suggesting that O shouldalsobe present the outer core. in

Thepossible presence H in thecore of through reaction the Fe+ H20 -? FeO+ Felixwas noted Stevenson by [1977] has but not beenpursued seriously because solubility H in Fe the of was thought betoolowandthevolatility H20 tOO to of high
to cause significant any effects [Stevenson, 1981].His ideawas revived, however, when enhanced the solubilityof in FeunH derpressure discovered was [Antonov al., 1980g et Fukaietal., 1982].Based thisandotherhighpressure on experiments on the Fe-H20 and Fe-hydrous minerals reaction [Fukaiand

1(? 1

9 8

? 162


Akimoto, 1983; Fukai,1984a; Suzuki al., 1984], suggestet we edthathydrogen be a majorcause thedensity can of deficit of theouter core. Nowthatwehave learned water that may have existed abundantly molten in silicates, proposition our has
become even more realistic.



O (?4


Theconstruction thispaper asfollows. startby of is We
describing properties theFe-Hsystem theFe- H20 some of and reaction under pressure high (sections 3)in order pro2 and to videa material basis subsequent for discussions.section In 4,
characterization the primordialmaterialthat carriesFe and of

H20 ismade, itsaccretion and process discussed. disis Here,
cussion focusedon how H20 wasretainedin the interior of is the growing earth. The Fe- H20 reaction that occurred the in proto-earth and the core-mantle separation ensued disthat are cussed qualitatively section Some in 5. quantitative considera-


30 35

ld 6





tion of these processes madein section including are 6, of in of estimation FeOcontent theearth's of of mantle possible Fig. 1. Solubility hydrogen ironasa function temperature and
composition thecore,followed some of by discussions other on
relevant problems in section 7.


10'4 )

andpressure. Experimental for 0.1 MPa (1 atm)of hydrogen data gas

(liquid[Schenck Wiinsch, and 1961] solid and phase Silvaand [da McLellan, 1976]) shown open are by circles. Solid curves represent the
results thermodynamical of calculations using these solubility data,the equationof stateof fluid hydrogen,and hydrogen-induced volume

to beextremely small. shown Figure it amounts only not beenincluded[Fukai and Sugimoto,1983]. As in 1, to 0.06at ?7o themelting at pointof Fein 1 atmof H2 gas[Schenck
and Wiinsch, 1961;da SilvaandMcLellan, 1976].The dissolu-

?2I? 2.9 ?3/H atom.The maximum = concentrationassumed be is to interaction between hydrogen atomshas The solubility H in Fe undernormalpressure known H/Fe = 1.0, and the elastic of is

tionprocess endothermic a large is with positive of solu- hydrogen-induced heat volume is nearly the same in all metaltion (A Hs>0), andthe solubility varies rathersmoothly over hydrogen systems, fiI4=2.9+0.2,? [Peisl, 3 1978;Fukai, 1984b],
differentphases: fi(bcc), c?, ?(fcc), andliquid. The mechanism and disregarded pressure the dependence fill. (This may be of of howH atoms dissolved transition are in metals nowfairly justified is fromthediscussionssection in thepressure in 6 range wellunderstood [Gelattet al., 1978; Fukai, 1985].H atoms oc- P= 10 GPa.) The maximum concentration was assumedto be cupyinterstitialsites,and form bondingstates with surround- H/Fe = 1 by analogywith othermetal-hydrogen systems. The ingmetal atoms. theenergetics bonding-state As of formation, results the calculation H in ? -Fe are shown Figure of for in whichessentially determines heat of solution,shouldbe 1. The rapid increase solubilitywith increasing the in pressure largely determined metal-atom by species, insensitivity originates the of fromthefactthatthetotalvolume change thedisin the solubility latticestructure naturallyunderstandable. solution to is process V) is negative, (A and consequently, pA V the High-pressure experiments performed the Fe-H system term in the solutionenthalpytendsto offset the positiveheat on

have shown, however, the that solubility isenhanced enormously solution at high pressures. of
underhighpressure. Antonovet al. [1980]found that the iron

In fact, the solubility the solidphase affected longin is by
range elastic interaction between H atoms. As the elastic ener-

hydrideof composition Fell0.8can be formedat 6.7 GPa, 250?C,and Fukai et al. [1982]obtained FeH0.?3at 3 GPa,

gy that costsin incorporatingone H atom in the lattice is lo-

weredby latticeexpansion, whichin turn is caused other by

Subsequently, Fukai and Sugimoto[1983]performedther- H atoms in solution, the net effect of the elastic interaction is modynamical calculations the solubilityof H in a number to decrease solution of the enthalpy proportion hydrogen in to conof transitionmetalsusingthe solubilitydata for 1 atm of H2 centration. Calculations including effecthavealsobeen this pergas theequation state fluidhydrogen hightemper- formed. illustrated Figure thesolid and of of at As in 2, solubility increased is atures pressures. and Underhighpressures, enthalpy so- by this effect especially high-pressures, the resultsobthe of at and lution determined low pressures at mustbe modified to include tained are in good agreementwith the observations cited above. the pA V term. In our calculation, we utilized the fact that This assures reliability of the calculation. the




had taken placeat 900?C and 6.4 GPa. The Felix was, however, not observed because decomposes it immediately whenthe pressure is released.


- .......... "--.

6.7 GPa

Subsequently, order to simulatemore closelythe reaction in that may have occurred in the proto-earth, we have extended

? 1(31 '??
? 152


our high-pressure experiments iron-enstatite-water to system
[Suzuki et al., 1984]. Starting material was a mixture of iron

powder, enstatite (MgSiO3), brucite (Mg(OH)2), talc (Mg3Si401o(OH)2), silicicacid(SiO2.xH20, x = 0.636) havand ing different compositions. It was discovered that the run
products of the reaction at 5 GPa and 1000? --- 1200?C con-



sistedof metallic iron, olivine, and pyroxene.(Recentexperimentshave shownthat by addition of AI(OH)3 in the starting material,Fe-containing garnetcanalsobe produced alongwith olivine and pyroxene.)The presence Fe-containingolivine of
implies that the reaction

FeO+ MgSiO3 (Mg,Fe)2SiO4 -?






Fig. 2. Comparisonbetweencalculationand experimentson the solubility of hydrogenin iron under high pressure.Solid and dashedcurves represent,respectively, resultof thermodynamicalcalculations the with and without including the effect of elastic interaction between hydrogen atoms [Fukai and Sugimoto, 1983]. The solid circle is the datum point of Antonov et al. [1980], and open circlesare those of Fukai et al. [1982].

had taken place. The removal of oxide from the surface of Fe particles appears to facilitate the progressof the reaction. In someof the 1200?Cruns, where needlelikequenchcrys-

talsof silicates wereobserved the recovered in specimens, large balls of iron of 0.1 -- 0.3 mm in diameterwere formed during the reactiontime of 20 min. (The averagesize of Fe particles in the startingmaterial was <_ 20/zm.) An exampleof suchrun productsis shownin Figure 3. The coagulationprocess Fe of
particles was, in fact, not easily controllable. In some cases, all the iron was found to be in one large block at the bottom of the reaction cell. It seemsthat once the coagulationof Fe particlesstarted,it proceeded rather rapidly. We alsonotedthat in order for balls of iron be formed, the startingmaterial must contain sufficientamount of water to causepartial meltingof
the silicates.

Unfortunately, the solubilityin molten Fe, which is more relevant to our present problem, is hard to obtain theoretically. However, considering that the elastic interaction is absent in liquids and the manner of bonding-state formation should be nearly independent of lattice structures, we may expect that the solubility in liquid can be approximated by calculations for a solid phase without including the elastic interaction, such as shown in Figure 1. Thus we have reasonably good estimatesof the solubility of H in molten Fe. It is as large as H/Fe= 0.2-- 0.4 near the melting point at a pressure severalgigapascals. of This rapid increase of solubility with pressurehas important geophysical implications, as discussedin the following sections.

The fact that the coagulationtook place at 1200?C implies that the melting point of Fe was lowered by at least -? 500K by dissolutionof hydrogen.(The meltingpoint of Fe at 5 GPa is -? 1700?C[Stronget al., 1973].) asthe concentration estiis mated at H/Fe = 0.3 -? 0.5 under these experimental conditions, the rate of decrease of the melting point with H concentration of the sameorder of magnitudeas that of other is
interstitial solute atoms such as C and N.


THE Fe-H20


The Fe-H20 reaction is known to produce Fe304 and H2 gas under normal conditions. When the reaction is made to proceed in an enclosed container, the atmosphere becomes more reductive, and the reaction takes the form

Fe + H20 -? FeO + H2
As the pressure of H2 gas is increased, a part of it should be dissolved in Fe,
....: ...........

?e (?)?-? ?e? +
and in consequence, Fe- H20 reaction should produce FeO the
and Felix.

Sucha high-pressure experimenthas beenperformed by Fukai and Akimoto [1983] using Fe and AI(OH)3 as starting materials. The presenceof FeO in the run product was taken to be
the evidence that the reaction

Fe+ AI(OH)3-->FeO+ FeAl?O4+ FeH?

Fig. 3. Photomicrographof recoveredspecimen high-pressure of experiment on Fe - hydrous silicate system. A mixture of iron, talc and AI(OH)3 was held at 5.4 GPa and 1200?C for 20 min and quenched. Large "balls of iron" were surroundedby olivine, pyroxene, and garnet. Range of view is 1.5 mm in width.



The mechanism of how the coagulation of Fe particles is facilitatedin theseexperiments not beenclarifiedyet. It canhas not simply be the result of lowering of the melting point of Fe becausethe presenceof S in Fe also lowers the melting point but is known to hinderthe coagulation process [e.g.,Takahashi, 1983]. We can only say that the surface of Fe particles in contact with partially molten silicates containingwater is maintained in the condition for easycoagulation. These observationshave profound implication in considering processof core-mantle the separation.

There have beenincreasing piecesof evidencethat the primordial material of terrestrial planets was a mixture of nebula condensates which had been formed under very different P, T conditions. Here we adopt a view that the primordial material of the terrestrial planets was a mixture of two components; a low-temperature (LT) component similar in compositionto carbonaceouschondrite (C1) and a high-temperature (HT) component which is an aggregate of highly reduced, metal-rich devolatilized material similar to E-chondrite, in the ratio of
10:90. The estimate of this ratio from the Cs/U ratio is 11:89








Fig. 4. Temporal variation of the fraction of water retained in the accretion process.The water retention factor, defined as the ratio of water content of surfacesilicates that of (average)primordial material, to is shown as a function of radius of the growing earth. Only in a rather short period around R/Ro = 1/4, the water is nearly completelydevolatilized to form the atmosphere. After this, silicatesnear the surface become partially molten and allow dissolution of all the accreted water. The general feature is quite insensitiveto the choice of parameters. For this calculation, the values adopted are the water content of the primordial material = 0.01, efficiency of impact dehydration = 0.2, accretion

[Anders et al., 1971] and about 10:90 from the four volatile elements (Zn, C1, In, Cs) [Ringwood, 1977]. As additional support for the two-component model, we quote the observation by infrared spectrometry that asteroids in the main belt can be classifiedinto two distinct types similar to C- and E-chondrites [Chapman et al., 1975; Gaffey and McCord, 1977]. The fractionation trend of oxygen isotopes is also consistent with this model; among variousplanetary materials, only C1- and E-chondrites fall on the terrestrial line of fractionation [Clayton et al., 1977; Clayton and Mayeda, 1984]. Analysisof the condensation sequence the solar nebula sugin gests that these two components accreted as separate planetesimalsof --- 10km in size in the earlier stage of accretion, becoming larger in the later stages[Safronov, 1972; Goldreich and Ward, 1973; Adachi et al., 1976; Greenberg et al., 1978]. What is important for our purpose is that a large amount of water should have been contained in the LT component. If it is identified as C l-chondrite, the average water content amounts to --- 20 wt %, in the form of both free water and structural water in hydrous minerals (such as serpentine, chlorite, brucite, epsomite,etc.). On the other hand, Fe shouldhave existedin oxidized from in LT component and in metallic form in HT component. Let us examinehow the water in the primordial material was retained in the interior of a planet in the course of its accretion. A temporal variation of the fraction of water retained in the growing earth is shown schematicallyin Figure 4 [A be and Matsui, 1985]. During the initial period of accretion (R/Ro- 0.2), sincethe impact dehydrationis negligible, all the water can be retained inside [Lange and Ahrens, 1982b]. By the time the proto-earth becomesheavy enough to causeimpact dehydration,it is alsoheavyenoughto keepthe water vapor in its gravitational field, and allows the water atmosphereto

time= 5 x 107years(courtesy Abe and Matsui[1985]). of

silicatesincreasesabruptly and the equilibrium can be established in time between the water dissolved in the molten sili-

catesand the atmosphere. Therefore the accretion proceedsin steadystate, and all the water accretedis dissolvedin the molten silicates. In the last stage of accretion, as a flux of infalling planetesimals subsides, the surface temperature is lowered gradually, resulting eventually in the solidification of the silicates. Then the water in the silicatesin the surface region goes into the atmosphere, as well as the water accreted afterward. Finally, the atmosphere condensesto form oceanic water. Overall, nearly 100 % of the accreted water was incorporated in the interior of the earth, leaving only a very small proportion in the form of atmosphere.A very similar picture is believed
to hold for CO2.

This scenarioof evolution provides a natural explanation for the extensive melting of silicates on the terrestrial planets, as well as for the origin of the atmosphere. The conclusionfrom

Ar-isotope studiesthat "catastrophicdegassing"occurredat

least within5 x 108 years fromtheaccretion [Ozima,1975] is
also consistent with this view. It must be recognized,however, that degassing characteristicsof water should be rather different from that of rare gasesbecauseof its large solubility in
molten silicates and its reaction with metallic iron. This latter

aspect will be discussedin the next section.


build up. The water atmosphere thus generatedtendsto raise the surfacetemperature by blanketing effect and to accelerate the dehydration process.During this period, all the water accretedshouldbe evaporatedto form the atmosphere.This continues until the surface temperature reachesthe solidus of of constituent silicates, whereupon the solubility of water in the

The product of the Fe-H20 reaction which takes place in molten silicates varieswith pressure and thereforewith depth from the surfaceof the growing planet. Near the surface, the reactionproduct consists FeO and H2 gas,which evolves of out of the molten silicatesand is lost eventuallyinto space.As we go deeperfrom the surface,the largerproportion of hydro-




Cold Primordial Material



Dehydrated Material


Liquid Metallic Phase

(Fe-Ni alloy+ H,S,O,C,etc.)
(Partially) Molten Silicate
Impact Generated Atmosphere

Fig. 5. Schematic picture thescenario core of of formation. Theproto-planet asa cold (a) starts body primordial of material. Asit grows size, impact (b) in the velocity infalling of planetesimals isincreased, theprimordial and material isdehydrated impact formtheatmosphere.Since atmosphere theaccretional by to (c) this traps energy, surface the temperatureis raised, silicates the becomes partially molten a layerof molten containing lightelements, as and Fe some such H and builds At a certain S, up. instant, spontaneous (d) asymmetry and cold occurs the primordial dehydrated (plus material)
is crushed,and (e) the proto core of the planetis formed.

The overall processof planetary evolution can be envisaged gen gets dissolvedin Fe. As a crude measure,the solubility at 1500?C amounts to -- 0.1 and 0.2 at 2 and 4 GPa. respectively as follows, slightly modifying the picture of Stevenson [1977] (Figure 1). These pressurevaluescorrespond,respectively,to (seeFigure 5). As planetary growth proceeds,the partial meltthe depth of about 50 and 100 km at the final stage of accre- ing of silicatesstarts from the surface, leaving the inner solid part (cold primordial) intact. A layer of molten Fe containing tion of the earth, and nearly 4 times as deep at the initiation light elementsgradually builds up between the cold primordial of melting of silicates. and partially molten silicatesand, at a certain instant, underIt must be recognizedhere that the amount of hydrogen that can be retained in the interior of the growing planet is deter- goesdrastic overturn with the cold primordial as a result of the mined by kineticsrather than equilibrium properties.If we were Rayleigh-Taylor instability, forming the protocore of the earth. to assume that accreted material stayed in the near-surface After this event, the steady state growth continues until the regionlong enoughto attain thermal equilibrium,we would have accretion comes to the end. This picture differs qualitatively from Stevenson'sin that the to acceptthat all the hydrogenshouldhave beenlost into space as a result of Fe- H20 reaction. In reality, however, this could surfacetemperature was higher than the solidusof silicatesdue hardly be the case. Immediately after impact fracturing of an to blanketing effect of water atmosphere and that hydrogen reacaccretedplanetesimal,the reaction of iron in fractured pieces should be dissolved in the core as a result of Fe-H20 in its flight in the water atmosphere may certainly occur but tion. Moreover, the overturn occurred in a much earlier stage undoubtedly to a very limited extent. (This is, in fact, neces- than was conceivedof by Stevenson, and the cold primordial smaller. According to Figure 4, it amounts sary for maintaining the water atmosphere around the grow- was correspondingly of ing planet.) In the molten silicates, iron particles, typically to lessthan 2 ?7o the total mass of the Earth, and is therefore submillimetersin size, will rapidly coagulateinto droplets and of minor importance for the whole history of evolution. Some passthroughthe near-surface region,decomposing smallfrac- quantitative aspectsof the present scenario will be discussed a tion of water on the way. It is only after thesestages that suffi- in the next section.
cient time is available for the Fe-H20 reaction to reach its

completion. Droplets of Felix, thus produced, will rapidly grow in size into large blobsand will act as scavenger other light elements, of suchas C, O, and S, as they sink into the deep interior. It may be noted here that the dissolution of other light elements (especially O) into Fe should also be controlled by kinetics rather than equilibrium properties.By the time molten Fe blobs reach the depth wherethe solubilityof O is large, the sizeof the blobs may be too large for equilibrium partition of O betweeniron
an silicates to be established.



Adopting the picture of planetary evolution describedabove, we now attempt to make some quantitative calculations on the stateof the presentearth, includingthe compositionof the core, in particular. We state in the beginning that the following calculations are primarily intended to test and demonstrate the overall consistencyof the present picture. Thus the reader is warned against acceptingthe numerical values too literally. Calculations must start with more concrete specification of the primordial material. We represent the composition of the



TABLE 1. Compositionof Primordial Materials Adopted.
Low-Temperature Component*
Fe Ni 0 0

High-Temperature Component**
34.5 2.0





1.18 22.86 1.19 5.65 2.99

1.7 0 0 0 0

with the model of chemically stratified mantle but not with the model of homogeneous pyrolite mantle. Knowing the total amount of MgO (23.9 ?7o) and the Fe/Mg ratio in the mantle, we can estimate the amount of FeO in the mantle to be 5.3 ?7o. The FeO supplied from the LT component (2.3%) amounts to only 40?7oof the FeO in the mantle; the rest has to be supplied by the Fe-H20 reaction. The total amount of FeO produced by the Fe- H20 reaction can be estimated from the amount of H20. Assuming that all the water in the LT component was accreted and reacted with Fe, we obtain 7.66 ?7oof FeO. Thus 7.66-3.0=4.66 ?7oof FeO can be




* Composition of Cl-chondrite (Orgueil). ** Derivedfrom the average 11 E-chondrites reducing amount of by the of SiO2 to fit the solar systemabundanceof Mg/Si (= 1.075) and
removing all volatiles.

regarded to be dissolved in the core, together with a greater proportion of hydrogen produced by the reaction (0.21 ?7o). All the C and S are regarded to go into the core. The sum of the materials that enter the core amounts to 32.6 070,which agrees excellently with the value of PREM, 32.4 ?7o. The composition
of the outer core thus formed becomes FeHo.4?Co.o5Oo.?3So.o3.

The density reduction of the outer core due to dissolution
of solute atoms can be calculated if the ratio of atomic volumes

of solute (ils) and molten iron (flo) is known for the relevant LT component that of Orgueil,one of the type I carbonaby P, T conditions. For a solute concentration x (in atomic ratio), ceous chondrite 1). Regarding HT component, adopt the relative density reduction is given by (C the we

the composition that we havederivedfrom the average 11 of E-chondrites[Mason, 1966] by adjustingthe amount of Si to g?o Ap x ns_Mo the solar system abundance ratio, Mg/Si = 1.075 [Andersand Ebihara, 1982],removing volatiles all and otherminor elements, Po 1+X?o and renormalizing.The composition thus derivedis given in where Ms/Mo is a ratio of the atomic weight of the solute to Table 1, together with that of the LT component (Orgueil).As
describedin section 4, the primordial material is assumedto be a mixtureof the LT and HT components the ratio of 10:90. in The next stepis to follow the materialsbudgetin the evolution process,by taking account of Fe- H20 reaction and dis-

solutionof light elements the core.The mass-flow in diagram is shownschematically Fig. 6. For simplicityof calculation, in the Ni content in the HT componentis included in Fe. SiO2 (and otherrefractoryoxides,MgO, A1203,etc.) in both LT and HT components omitted from the figure; they all go into are the mantle to constitutea major part of it. Before going to estimatethe core composition,let us examinethecomposition themantle.Ringwood[1975]hassugof gested,on the basisof compositions mantle xenoliths, that of the wholemantlemay be regarded be of pyrolitecomposito tion M?.57SiO3.57, whereM represents mixtureof Fe and Mg a in the atomic ratio Fe/(Fe+ Mg) = 0.11. It must be noted, however,that as long as we acceptthe large-scale melting of silicates the evolutionprocess, is logicallymore consistent in it to accept chemically a stratifiedmantleratherthan a homogeneousmantle as conceived Ringwood.A chemicallystratified by mantle is formed as a result of gravitational differentiation in moltensilicates; lower mantlecontains the almostexclusively ferromagnesiansilicates of perovskite structure having a stoichiometry pyroxene,and the upper mantle is approxiof mately of the pyrolite composition [Anderson, 1984; E. Ohtahi, unpublished manuscript, 1985]. If we assume that Fe/(Fe + Mg) - 0.11 appliesto both upper and lower mantles,
the average mantle composition can be obtained as

that of Fe and ?o is a densityof pure Fe (?o = Mo/9o). We shall here estimate the density reduction at the core-mantle boundary (CMB), where the pressureis --- 135 GPa and the temperature is believed to lie between 3000 and 4000 K. For hydrogen, a theoretical calculation has been performed by K. Terakura (personalcommunication, 1983) on the pressure-volume (P-V) relation of iron hydride Fell (NaC1 structure) up to the pressure of 380 GPa. The pressure dependence of 9s/9o obtained is shown in Figure 7. It becomes0.19 at 135 GPa. His calculation, basedon the electronicenergy-bandtheory, has given good agreementwith the shock wave data on vanadium hydride [Syono et al., 1984] and is therefore believed to be fairly reliable. For oxygen, from a recent analysis of the effect of high pressure on the solubility of FeO in molten Fe, the partial molar volume of molten FeO is obtained as 10.S1 cm3 at 4 GPa and 1600?C [Ohtani et al., 1984]. As 90 of molten Fe under the cor-

responding T condition estimated --- 7.2 cm 9s/9ois P, is at 3,
obtained as --- 0.50. The same value is assumed for the P, T condition of CMB. For carbon, lacking experimental data, 9s/9o --- 0.50 is also assumed. For sulfur, 9s is estimated from

'1--12 , 0
LT comp.


(Mgo.89Feo. ll)?.13SiO3.13. (The resultof recentseismic analysis (PREM) that the massof the upper and lower mantlesis 17.7?7o and 49.3 070, respectively, the total massof the earth [Dzieof wonskiand Anderson, 1981] has also been used.)

FeO r


comp. Fe?


Fe Hx

The average composition of the silicate that can be formed Fig. 6. Mass-flow diagram of the evolutionaryprocessof the earth. from the primordial material adopted here is found to be For the sakeof clarity, SiO2 and other nonvolatile oxides, all of which (Mgo.89Feo. ll)l.2oSiO3.2o. resultis in reasonable The agreement go to the mantle, have been omitted from the figure.



metallic iron and water was accretedthroughout the entire peri0.3
od of accretion.

Naturally, the detailsof the process dependenton the comare position of the primordial material assumed. The total mass ?oo CMB of the earth's core is obtained correctly here becausewe have 1 adopted the HT component similar in composition to Echondrites;they are known to contain much larger amounts of Fe than other groups of chondrites. It is by this choice of the HT component that we have dispensed with the need for preferential accumulationof metallic Fe in the initial accretional 0.1 H in ?,-Fe stage, as suggestedby a number of investigators. The overall redox state of Fe is determined by the amount of H20, which Calc. by Terakura in the present argument, has been constrainedindirectly by the amount of other volatiles.The amount of C and S in the primordial material cannot be better constrained by other evidence. , , , 0 100 200 300 4 ?) 0 So that the quantitative calculation performed in section 6 p, GPa should be admitted as a rather tentative one. However, we mainFig. 7. Pressure dependence theratioof hydrogen-induced of volume of (9?) and atomicvolumeof Fe (90) in -?- Fe calculated the electron- tain that the overall consistency the resultscannot be accidenby tal; it must be an indication that we have succeeded in band theory. 9? hasbeenobtainedas a difference volumeof Fell in (assumed possess to NaC1structure) -?- Fe (fcc)(K. Terakura, and per- reproducing the real history of the earth rather closely. sonal communication, 1983). Regardingthe multicomponentnature of the core, it was suggestedby Ringwood [1979] that under high pressure,H and O the shockcompression data on FeS0.9[Ahrens, 1982], and the should exsolve out of molten Fe by forming H20 and in this value of fis/fi0 at 135 GPa is obtained as 1.28. This value is, way go into silicatesagain and that the similar thing shouldhaphowever,believedto be overestimated. Considering that the pen to C and O. However, judging from the fact that atomic



atomic volume of O in molten Fe is reduced to -

2/3 of the

volumes of H and O are smaller in molten

Fe than in covalent

valuein crystalline oxide [Ohtaniet al., 1984],it mightbe more appropriate to assumethat fis is also reducednearly by this

Using these results, a density reduction of a liquid alloy
FeHxCyOzSurelative to molten Fe at CMB can be written as

or ionic compounds, it is rather unlikely that the exsolution shouldtake place under high pressure.Thus we believethat the coexistenceof several light elementsin the core is allowed thermodynamically. We shall next examine some implications of the present picture in a broader context of the evolution of the terrestrial


1 + 0.19x


1 + 0.50y


1 + 0.50z


1 + 0.86u

For the outer core composition FeHo.4?Co.o5Oo.?3So.o3 deduced above, the density reduction becomes


= 0.072(H)+ 0.013(C)+ 0.024(0)
+ 0.009(S)-0.118

The evolution of the moon deservesspecial consideration. Although no general consensus has been reached so far on the 7. DISCUSSION geneticprocess the moon, its low densityrequiresthat as long of The process of evolution of the earth described in the foreas it was formed in the feeding zone of the terrestrial planets, in going sectionscan be regarded to be the inevitable consequence it had undergonefractionation process someparent body before it came to exist as such. In fact, most geochemicalevidence of the assumptionthat the primordial material containing both

This result is in reasonable agreement with the value 0.092 + 0.010 obtained from the seismic analysis PREM [Dziewonski and Anderson, 1981]. Thus hydrogen is by far the most important light element for the density reduction of the core. If we ascribethis small discrepancyto a partial lossof hydrogen due to Fe- H20 reaction in the near-surface region, the hydrogen concentration becomesx = 0.25, which implies that --- 40 ?7o of hydrogen was lost into space. The lost fraction of hydrogen has to be increasedif we start with larger amounts of LT component in the primordial material. For the extreme case of 15'85 mixture, incorporation of all the hydrogen in the core leads to the composition FeH0.a3C0.0?O0.23S0.05 a concomitant density reduction of and 0.183. In order to reduce this value to 0.092, we have to assume a hydrogen loss of --- 90 ?7o and the core composition of FeH0.0sC0.0?O0.23S0.0?. lost fraction appearsto be too large, The though not totally impossible.

planets. Among the terrestrial planets, only in Mercury is the Fe- H20 reaction believedto be largely irrelevant; the primordial material of Mercury should have been nearly completely devolatilized. The primordial material of Venus, earth, and Mars can be regarded to consistof similar LT and HT components, differing only in their relative magnitudes. As suggested first by Ringwood and Clark [1971] and later elucidatedby Ringwood and Anderson [1977], relatively low mean density of Venus and Mars may be explained by assuming that Fe is more highly oxidized in theseplanets than in the earth. This amounts to sayingthat the primordial material of theseplanetscontained larger amounts of LT component. The evolution of Venus, which is believedto be very similar to what was described above, will not be considered here. In the case of Mars, the density of the core is estimated to be smaller than Fe by as much as --- 30 070, indicating that much larger amounts of light elementsshould be presentin the core. In addition, the estimated higher moment of inertia requires heavier mantle material [Johnstonand TOksoz, 1977] and hence a larger amount of FeO in the mantle. These observations can be naturally explained by taking account of the Fe- H20 reaction, if the primordial material contained20 --- 30 ?7o LT comof ponent. More quantitative discussionsalong these lines are
believed to be useful as more information
on Mars.

becomes available



appears be in favor of a fissionmodel, which considers to that
the moon was derived, in some way, from the earth's mantle

of NagoyaUniversity valuable for comments the manuscript. on This work has been supported,in part, by a Grant-in-Aid for
ScienticResearchfrom The Ministry of Education, Scienceand

subsequent core formation [Ringwood, 1979]. There is, to however, a strong criticism of the fissionmodel. Nakazawa et al. [1983]pointedout that the angularmomentumof the earthmoon systemis too large to have been possessed their parby ent body and maintained that it was formed by capture. Here we shall examine what little we can say by extending our presentpicture to the evolutionprocess the moon. The only of unambiguousstatementthat can be made irrespectiveof the geneticprocess the moon is that its core, if it ever existed of at all, should not contain any appreciable amount of H or O because pressure the was too low to allow dissolutionof these elements. If we assume that composition of the primordial materialremained samethroughout wholeaccretion the the period, the absenceof metallic Fe on the lunar surfaceby any significant amount indicates that the moon cooled down and its

surface solidified after the accretion in the feeding zone of the
earth had almost ended.

If we shouldadopt the fissionmodel, a little more argument can be made. The fact that the mean density of the moon is closeto that of the earth's mantle suggests that the fission occurred in the final stageof terrestrial accretion. Otherwise, the

primordial material accretedon the moon after the fission
should have made its mean density higher. Evidence of extensive differentiation and melting of the moon suggests that it occurredshortly after the fission, when the temperature of the proto-moon was still high.

Abe, Y., and T. Matsui, The formation of an impact-generated H20 atmosphereand its implications for the early thermal history of the earth, Proc. Lunar Planet. Sci. Conf., 15th, Part 2, J. Geophys. Res., suppl., 90, C545-C559, 1985. Adachi, I., C. Hayashi, and K. Nakazawa, The gas drag effect on the elliptic motion of a solid body in the primordial solar nebula, Prog. Theor. Phys., 56, 1756-1771, 1976. Ahrens, T. J., Constraintson core composition from shock-wavedata, Philos. Trans. R. Soc. London, Ser. A, 306, 37-47, 1982. Anders, E., and M. Ebihara, Solar-systemabundancesof the elements, Geochim. Cosmochim. Acta, 46, 2363-2380, 1982. Anders, E., R. Ganapathy, R. R. Keays, J. C. Laul, and J. W. Morgan, Volatile and siderophile elements in lunar rocks: Comparison with terrestrial and meteoritic basalts, Proc. Lunar Sci. Conf., 2nd, 1021-1036, 1971. Anderson, D. L., The earth as a planet, paradigms and paradoxes, Science, 223, 347-354, 1984. Antonov, V. E., I. T. Belash, V. F. Degtyareva, E.G. Ponyatovskii, and V. I. Shiryaev, Obtaining iron hydride under high hydrogen pressure, Soy. Phys. Dokl., Engl. Transl., 25, 490-492, 1980. Chapman C. R., D. Morrison, and B. Zellner, Surface properties of asteroids: synthesis polarimetry, radiometry and spectrophotomA of etry, Icarus, 25, 104-130, 1975.

Clayton, R. N., and T. K. Mayeda, Oxygenisotopiccompositions of
enstatite chondrites and aubrites, Proc. Lunar Planet. Sci. Conf., 15th, Part 1, J. Geophys. Res. suppl., C245-C249, 1984.

Clayton, R. N., N. Onuma, L. Grossman, and T. K. Mayeda, Distribution of the pre-solarcomponentin Allende and other carbonaceous chondrites, Earth Planet. Sci. Lett., 34, 209-224, 1977. da Silva, J. R. G., and R. B. McLellan, The solubility of hydrogen in super-pure iron single crystals, J. Less-Common Met., 50, 1-5,

Finally,it may be pointedout that the partitionof siderophile
elements between the core and the mantle should have been af-

fectedby the presence water in the primordial material. We of may expect,therefore,that systematic investigation partiof
tion characteristics elementsin the presenceof water should of

provideusefulinformation on the role of water in the evolutionary processand thus serve as a critical test of the view
described in this paper. With this possibilityin mind, we have been engagedin high-

Dziewonski, A.M., and D. L. Anderson, Preliminary reference earth model, Phys. Earth Planet. Inter., 25, 297-356, 1981. Fukai, Y., The iron-water reaction and the evolution of the Earth, Nature, 308, 174-175, 1984a.

Fukai, Y., Site preference of interstitial hydrogen in metals, J. LessCommon Met., 101, 1-16, 1984b.

pressure experiments the partition of somemetallicelements Fukai, Y., Atomistic and electronicapproachesto hydrogen in metals, on Cryst. Lattice Defects Amorphous Mater., 11, 85-111, 1985. in the course suchiron-hydrous of silicate reactions described as Fukai, Y., and S. Akimoto, Hydrogenin theEarth'score:Experimenin this paper. Resultsof theseexperiments will be reported in tal approach, Proc. Jpn. Acad., Ser B, 59, 158-162, 1983. forthcoming publications. Fukai, Y., and H. Sugimoto, Enhanced solubility of hydrogen in metals under high pressure: Thermodynamicalcalculation, Trans. Jpn.

Basedon our experimentalfindings on the Fe-hydroussilicate reactionand concomitantdissolutionof H in Fe under high pressure, have demonstrated we that a consistent scenario of planetaryevolutioncanbe envisaged providedmetallicFe and water were containedin the primordial material. It may be emphasized that the extensive meltingof silicates the evolution in process caused the blanketing by effectof the water atmosphere was of crucial importance for the evolution process,not only for its directconsequences asgravitationaldifferentiation such of the mantle but also for its ability to dissolvea large amount of water, allowingthe reactionof water with metallicFe in the interior of the growing planet. Investigatingthe combination of iron, water, and partially molten silicates believedto shed is light on some important aspects evolution that have been of largely overlooked in the past.
Acknowledgments. We are grateful to S. Akimoto of the Institute for Solid State Physics,University of Tokyo, for his
constantencouragement and useful discussions and H. Mizutani

1983. Fukai, Y., A. Fukizawa, K. Watanabe, and M. Amano, Hydrogen in iron: Its enhanceddissolutionunder pressure and stabilization of Vphase, Jpn. J. Appl. Phys., 21, L318-L320, 1982. Gaffey, M. J., and T. B. McCord, Asteroid surfacematerials: Mineralogical characteristics and cosmologicalimplications, Proc. Lunar Sci. Conf., 8th, 113-143, 1977. Gelatt, C. D. Jr., H. Ehrenreich, and J. A. Weiss, Transition-metal hydrides:Electronic structureand the heatsof formation, Phys. Rev.
B Condens. Matter, 17, 1940-1957,1978. Goldreich, P., and W. Ward, The formation of planetesimals, Astrophys. J., 183, 1051-1061, 1973. Greenberg, R. J., J. F. Wacker, W. K. Hartmann, and C. R. Chipman, Planetesimalsto planets: Numerical simulation of collisional evolution, Icarus, 35, 1-26, 1978. Johnston, D. H., and M. N. T6ksoz, Internal structure and properties of Mars, Icarus, 32, 73-84, 1977. Lange, M. A., and T. J. Ahrens, Impact induced dehydration of serpentine and evolution of planetary atmosphere,Proc. Lunar Planet. Sci. Conf., 13th, Part 1, J. Geophys. Res. suppl., 87, A451-A456,

Inst. Met., 24, 733-740,

Lange, M. A., and T. J. Ahrens, The evolutionof an impact-generated atmosphere, Icarus, 51, 96-120, 1982b. Lange, M. A., and T. J. Ahrens, FeO and H20 and homogeneous accretion of the earth, Earth Planet. Sci. Lett., 71, 111-119, 1984.




Mason, B., The enstatite chondrites, Geochim. Cosmochim. Acta, 30, 23-39, 1966. Nakazawa, K., T. Komuro, and C. Hayashi, Origin of the Moon -

sungim Eisen durch Sauerstoff, Arch. Eisenh?ittenwes.,11, 779-790,

Stevenson,D. J., Hydrogen in the Earth's core, Nature, 268, 130-131.

Capture by gas drag of the earth's primordial atmosphere, Moon
Planets, 28, 311-327, 1983.

Stevenson, D. J., Models of the Earth's core, Science, 214, 611-619,

Ohtani, E., and A. E. Ringwood, Composition of the core, I, Solubility of oxygen in molten iron at high temperatures,Earth Planet. Sci.
Lett., 71, 85-93, 1984. Ohtani, E., A. E. Ringwood, and W. Hibberson, Composition of the core, II, Effect of high pressureon solubility of FeO in molten iron, Earth Planet. Sci. Lett. , 71, 94-103, 1984. Ozima, M., Ar isotopesand Earth-atmosphere evolution models, Geochim. Cosmochim. Acta, 39, 1127-1134, 1975. Peisl, H., Lattice strains due to hydrogen in metals, in Hydrogen in Metals I, edited by G. Alefeld and J. V61kl, Springer-Verlag, pp.53-74, New York, 1978. Ringwood, A. E., Composition and Petrology of the Earth's Mantle, 618pp., McGraw-Hill, New York, 1975. Ringwood, A. E., Composition of the core and implications for origin of the earth, Geochem. J., 11, 111-135, 1977. Ringwood, A. E., Water in the solar system,in Water-Planets, Plants and People, edited by A. Mcintyre, pp. 18-34, Australian Academy of Sciences, Canberra, 1978. Ringwood, A. E., Origin of the Earth and Moon, 295pp., SpringerVerlag, New York, 1979. Ringwood, A. E., and D. L. Anderson, Earth and Venus: A comparative study, Icarus, 30, 243-253, 1977. Ringwood, A. E., and S. P. Clark, Internal constitution of Mars, Nature, 234, 89-92, 1971. Safronov, V. S., Evolution of proto-planetary cloud and formation of the earth and planets, NASA-TT-F-677, 1972.

Strong, H. M., R. E. Tuft, and R. E. Hanneman, The iron fusion curve and ?, -8 - l triple point, Metall. Trans., 4, 2657-2661, 1973. Suzuki, T., S. Akimoto, and Y. Fukai, The systemiron-enstatite-water at high pressuresand temperatures - Formation of iron hydride and some geophysical implications, Phys. Earth Planet. Inter., 36, 135-144, 1984. Syono, Y., K. Kusaba. K. Fukuoka, Y. Fukai, and K. Watanabe, Shock compressionof V2H and V2D to 135 GPa and anomalous decompression behavior, Phys. Rev. B, 29, 6520-6524, 1984. Takahashi, E., Melting of YAMATO L3 chondrite(Y-74191) up to 300 kbar, in Proc. Eighth Symposium on Antarctic Meteorites, pp.168-180, National Institute for Polar Research, Tokyo, 1983. Usselman. T. M., Experimental approach to the state of the core, I, The liquidus relations of the Fe-rich portion of the Fe-Ni-S system from 30 to 100 kb, Am. J. Sci., 275, 278-290, 1975a. Usselman, T.M., Experimental approach to the state of the core, II, Composition and thermal range, Am. J. Sci., 275, 291-303, 1975b.

Y. Fukai, Department of Physics,Chuo University, Bunkyo-ku,
Tokyo 112, Japan. T. Suzuki, Institute for Solid State Physics,University of Tokyo,

Minato-ku, Tokyo 106, Japan.
(Received March 26, 1985; revised December 4, 1985; accepted January 28, 1986.)

Schenck, and H. Wtinsch, H., Ober die Gleichgewichts16slichkeit des
Wasserstoffsim fltissigenreinen Nickel und Eisen und die Beeinflus-



All rights reserved Powered by 甜梦文库 9512.net

copyright ©right 2010-2021。