poirier 1994 pepi

Upload: gabrielignaciofuentealbaumanzor

Post on 06-Jul-2018

225 views

Category:

Documents


0 download

TRANSCRIPT

  • 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

  • 8/17/2019 Poirier 1994 PEPI

    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 

  • 8/17/2019 Poirier 1994 PEPI

    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

  • 8/17/2019 Poirier 1994 PEPI

    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

  • 8/17/2019 Poirier 1994 PEPI

    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

  • 8/17/2019 Poirier 1994 PEPI

    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).

  • 8/17/2019 Poirier 1994 PEPI

    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

  • 8/17/2019 Poirier 1994 PEPI

    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

  • 8/17/2019 Poirier 1994 PEPI

    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

  • 8/17/2019 Poirier 1994 PEPI

    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.

  • 8/17/2019 Poirier 1994 PEPI

    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.

  • 8/17/2019 Poirier 1994 PEPI

    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).

  • 8/17/2019 Poirier 1994 PEPI

    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,

  • 8/17/2019 Poirier 1994 PEPI

    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.

  • 8/17/2019 Poirier 1994 PEPI

    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-

  • 8/17/2019 Poirier 1994 PEPI

    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.

  • 8/17/2019 Poirier 1994 PEPI

    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.

  • 8/17/2019 Poirier 1994 PEPI

    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