+ All Categories
Home > Documents > 1-s2.0-S003192019800123X-main(1)

1-s2.0-S003192019800123X-main(1)

Date post: 14-Apr-2018
Category:
Upload: endimion-simmons
View: 213 times
Download: 0 times
Share this document with a friend

of 19

Transcript
  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    1/19

    .Physics of the Earth and Planetary Interiors 109 1998 179197

    The Gruneisen parameter for iron at outer core conditions and theresulting conductive heat and power in the core

    Orson L. Anderson )

    Department of Earth and Space Sciences and Institute of Geophysics and Planetary Physics, Uniersity of California, Los Angeles, Los

    Angeles, CA 90095-1567, USA

    Received 6 April 1998; accepted 11 August 1998

    Abstract

    The aim of this paper is to find the conductive power in the outer core. Before the heat conduction equation can be

    usefully applied, however, a careful examination of the Gruneisen parameter, g, and the thermal conductivity, k , of the ccore is required. The focus of this paper is on these two parameters at outer core conditions, using primarily experimental

    data and one theoretical evaluation of g. The melting g of the core is found to be 1.3, the adiabatic g , slightly less, andm ad

    k is found to be 44 W my1 Ky1 in the upper limit and 28 W my1 Ky1 in the lower limit. To obtain the coresc

    .temperature, T , from T of pure iron at core conditions , the freezing point depression is found. This requires assumptionsm m

    about the impurities in the iron core. Recent experimental data greatly restrict the type and concentrations of these

    impurities. Several allowable combinations are found, however; all of these lead to a DT of ; 1000 K. At the core side of . y1the core-mantle boundary CMB , the adiabatic temperature is 3900 K and dTrd r is 0.9 K m , resulting in a conductive

    12

    .power of 4.4 " 1 TW 1 TWs10 Watt flowing out of the core along the adiabatic gradient. New experimental results onthe thermal conductivity of mantle minerals and new theoretical insights into the heat flow in mantle plumes support the casethat the conductive power along the cores adiabatic gradient passing to the upper mantle can be transported away by both

    conduction and convection in the upper mantle. q1998 Elsevier Science B.V. All rights reserved.

    Keywords: Gruneisen parameter; Iron; Core conduction; Mantle conduction

    1. Brief description of the model

    This paper examines the conductive power flow-

    ing from the core and especially the parameters

    found in the thermal gradient at the core-mantle

    .boundary CMB and the thermal conductivity. At-tention is focused on finding the values of the pa-

    rameters involved in calculating the conductive heat,

    )

    Corresponding author. Fax.: q1-310-206-3051; E-mail:[email protected]

    especially the Gruneisen parameter and the thermalconductivity. The values of these parameters are

    found to be similar to those by Braginsky and Roberts .1995 , but different from those found in most recent

    reports.

    I disallow the cores thermal gradient from beingsubadiabatic because I find the conductive power

    12from the core to be 3.4 to 5.4 TW 1 TWs10.Watt , which is within the range of power leaving

    the CMB on the mantle side by conduction and

    convection. According to this model, any excess

    power, which will be convective, will be returned to

    0031-9201r98r$ - see front matter q 1998 Elsevier Science B.V. All rights reserved. .P I I : S 0 0 3 1 - 9 2 0 1 9 8 0 0 1 2 3 - X

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    2/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197180

    the core by the mechanism of compositional convec- .tion of Loper 1978a,b .

    In this model, there is no attempt to explore the

    physics of the cores convective power or to find all

    the sources that comprise the cores conductive

    power. I do, however, estimate the power generated

    by the crystallization of the inner core boundary

    since the parameters for this power source have been

    determined.

    2. Parameters needed for the heat flow analysis

    The heat conducted from the core to the mantle is

    given by:

    ETq syk , 1 .c c /Er Swhere k is the thermal conductivity of the core andcr is its radiusthe radial distance from the center of

    Earth to the point considered. The outer core is

    known to be liquid and is ordinarily assumed to be

    under adiabatic compression, so one can obtain the

    slope of the temperature profile by means of the

    adiabatic equation:

    ET ygg T1s , 2 . /Er fS

    where g is the Gruneisen parameter of the liquid1state, g is gravity, and f is the seismic parameter.

    The departure from adiabaticity is considered to be

    small. Although the values of g and f at any radius

    can be readily obtained from seismological tables,

    evaluation of the other parameters requires special . .effort. Values of k and ETrEr in Eq. 1 must bec

    . .found. Before ETrEr can be found, g and T r inl . .Eq. 2 must be evaluated. T r in turn depends on

    g , the gamma of melting. Thus, to begin an evalua-mtion of the power, Q , one must first understand thecGruneisen parameter in the various states and phasesof iron.

    3. g in the solid, liquid, and melting states of iron

    Fig. 1 shows the theoretical and experimentalvalues of g for pure iron in its various phases see

    Fig. 1. Gruneisen parameter, g, vs. density for iron. The top scale,for P in GPa, shows various starting values of P s0 for the bcc

    .phase, the hcp phase, and liquid. The main focus is the g e vs. rs .curve starting at G, as deduced from Mao et al. 1990 EoS

    .measurements, extending along the Hugoniot F from Brown and .McQueen 1986 and terminating at melting at the edge of the

    . .solid phase D calculated by Stacey 1995 . Other lines represent . .g vs. r at low pressure and g a E at low pressure measuredl s

    .by Boehler and Ramakrishnan 1980 . Note that g for the core ismgiven as 1.3"0.3, which coincides with the solid phase D by

    Stacey. The curve g for the core, ---, is coincident with DX

    andmD

    Y. The liquid for the isentrope, shown by P , is slightly lower.

    .phase diagram of iron, Fig. 2 . We now review the

    literature data on g as plotted in Fig. 1. The line forvalues of g has Points A, B, and C Stevenson,l

    1981; Verhoogen, 1980; Chen and Ahrens, 1997,.respectively marked as triangles. This line descends

    rapidly with pressure, P. Point E, showing g in the . .solid state g , represents the a- bcc iron datum ats

    .P s0. The line extending from Point E, called g a ,s

    represents measured values at higher densities see.figure caption for data references . The line emanat-

    ing from E represents the measurements for bcc iron:

    g s 1.66 and q s0.6 in the equation:0

    g r0s , /g r0

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    3/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197 181

    Fig. 2. The phase diagram of iron in V, T space. The shaded areas .represent DV at phase boundaries. The beta phase b is accepted

    as part of the phase diagram, making b the dominant phase at core

    conditions. The Hugoniot passes through the e and b phases well

    below melting, until it emerges into the liquid phase at 240 GPa.

    where the subscript 0 means at P s0. The solid X Y .squares in Fig. 1 D , D, and D represent g of thes

    .core at the core-mantle boundary CMB , the inner .core boundary ICB and the middle of the outer

    .core but in the solid state , as calculated by Stacey .1995 . To these literature data on g is now added

    .the value of g e , Point G, corresponding to the essolid phase.

    .We find g e at P s 0, Point G in Fig. 1, fromsthe relationship between K

    Xand g for hcp iron, as0 0

    . .derived by Stacey 1995 see his Eqs. 34 and ..35 , where:

    g s 1r2 KX y0.95, 3 . .0 0X . Xand where K s EK rET . To obtain K for e0 T Ps0 0

    .iron listed in Table 1 , one uses P, V data on hcp .iron measured to 330 GPa by Mao et al. 1990 at

    Ts 300 K. KX is found from these data by solving0 .the BirchMurnaghan EoS, Eq. 4 . The values of

    K and V used in the calculation of KX

    are listedT 0 00in Table 1:

    22 V 3X ~K s4 q P0 / /3 K VT 0 0

    y1y15 2 V 4 V3 3 y y1 y 1 . / / /V 3 V0 0 4 .

    Table 1 shows that the average value of KX

    for hcp0 .iron is 5.31, as found from Eq. 4 . Using the Mao et

    Table 1

    Values of KX

    for e iron from measured P, V dataa using the0

    BirchMurnaghan EoS whereK

    s165 GPa andV

    s6.73 cm3

    T 00moly1

    X3 y1 y5 3 y1 . . .P GPa V cm mol V =10 m kg K0

    34.8 5.8110 10.408 4.915

    35.6 5.8120 10.410 5.257

    56.1 5.4740 9.805 4.795

    80.0 5.2760 9.450 5.687

    101.6 5.0800 9.099 5.587

    112.4 4.9300 8.831 4.997

    125.4 4.9030 8.872 5.538

    132.1 4.8090 8.614 5.135

    141.2 4.7140 8.443 4.878

    142.2 4.7010 8.420 4.835

    156.7 4.6690 8.362 5.193170.8 4.6130 8.263 5.324

    197.3 4.4750 8.015 5.244

    208.7 4.4580 7.985 5.448

    210.1 4.4680 8.383 5.553

    212.1 4.4170 7.912 5.278

    227.2 4.3850 7.854 5.452

    235.7 4.3260 7.748 5.287

    244.8 4.3210 7.740 5.456

    247.3 4.3100 7.720 5.443

    256.2 4.2540 7.620 5.294

    258.2 4.2870 7.679 5.531

    264.0 4.2890 7.682 5.663

    270.1 4.2430 7.600 5.500

    270.4 4.2130 7.546 5.330277.8 4.1700 7.469 5.223

    292.8 4.1450 7.424 5.338

    293.5 4.1370 7.410 5.306

    294.7 4.1440 7.423 5.364

    304.1 4.0990 7.342 5.268

    a .P and V data from Mao et al. 1990 .

    Average value of KX

    is 5.31.0

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    4/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197182

    al. P, V data with the Vinet EoS, one finds KX s5.310

    and, using same data with the BornMie EoS, KX s0

    5.11. For e iron, KX s5.3 is selected from the Mao0

    .et al. 1990 data, yielding g s1.7 for e iron from0 .Eq. 3 . The value of g for hcp iron is represented0

    .by Point G in Fig. 1. We now seek g hcp at highspressure.

    .Stacey 1995 found theoretically the values of gsfor close packed crystalline iron at core pressures

    X Y .shown by the squares D, D , and D in Fig. 1,

    covering the pressure range of the outer core. They X . . Y .are g D s1.27, g D s1.30, and g D s1.33.

    .Gilvarry 1956 argued that the melting of a solid

    occurs at the limit where TT , e.g., at the meltmboundary of the solid and liquid, but on the solid

    side. Therefore the melting line shown as a dashed.line in Fig. 1 is drawn adjacent to Staceys points

    .solid squares . Noting Staceys value of g at mid-

    core pressure, e.g., near 240 GPa, we now connectthe Hugoniot to Staceys Point D. I assume that for

    the hcp phase, g is independent of T at constant V.s .This makes the Hugoniot value, g P , identical toH

    .the hcp value, g P . I now use the experimentalsobservation that the curve representing the P, V data

    on the Hugoniot egresses from the solid hcp phase at

    P s 240 GPa, because at that pressure, the shearvelocity of pure iron approaches zero, as seen in Fig.

    3. Thus, the onset of melting for the Hugoniot, .T 240 , is at a pressure of 240 GPa. The g of them

    Hugoniot, g , is connected to Staceys midpointHvalue, which is at 240 GPa, the same pressure at

    which the Hugoniot enters the liquid state. This gives . .g 240 sg s1.30. Since g e is a function of PH m s

    but insensitive to T at constant r, the Hugoniots gH . is identical to g e or g of b iron above Ts 1500s

    . .K . Thus the terminus of the g e curve, as well assthat of g , is Staceys Point D. A line connectsH

    . y3 . .g e at P s0 rs8280 kg m and g e ats s y3 . .P s240 GPa rs12200 kg m . From Eq. 3 it

    .is found that qs 0.7, so that g r is:e

    0. 7r0g r s1.7 . 5 . .

    e /rThe gr diagram, Fig. 1, is now complete. What is

    significant for core physics from the diagram is that .g 240 s 1.3, which is identical to the solid statem

    .value of Stacey 1995 and also connects the equa-

    Fig. 3. The Hugoniot experimental P, V data, plotting r vs. P.The sound velocity of iron measured along the Hugoniot path. The

    melting of the Hugoniot, placed at 240 GPa, occurs when ysplunges to zero. Dr of melting is found from the separation

    between the two Hugoniot curves in the liquid state. Data repre- .sented by solid circles are from Brown and McQueen 1986 ; data

    .represented by diamonds are from Altshuler et al. 1962 .

    tion of state of hcp iron and the Hugoniot of iron to

    the melt at midcore pressure.

    We are now in a position to evaluate the meltingtemperature at 240 and 330 GPa. We follow the

    .procedure shown by Anderson and Duba 1997 and .also by Stixrude et al. 1997 evaluating T at itsH

    . .melting pressure 240 GPa and then finding T 330m .from T 240 by the Lindemann law. The tempera-m

    ture profile of the Hugoniot must now be evaluated.

    4. Evaluation of the Hugoniot temperature

    The Hugoniot temperature, T , at any PV pointH .can be calculated McQueen, 1964 from the integra-

    tion of:

    T V yV d P q P yP dV . .S 0 0dT syg dVq ,H s

    V 2CV

    6 .

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    5/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197 183

    Table 2 .Calculation of T and DT the isentropic contribution to DTS S H

    y3 y3 y3 y6 3 y1 . . . . .P J m r =10 kg m V =10 m kg g Drrr g DT K T Kave S S

    0 1.7 a40 9.49 10.5 1.55 650

    60 9.99 10.0 1.49 0.0513 1.52 51 701

    80 10.42 9.6 1.45 0.0490 1.47 50.8 752

    100 10.72 9.31 1.42 0.0280 1.435 30.7 783120 11.04 9.06 1.39 0.0215 1.405 30.46 814

    140 11.3 8.85 1.34 0.0232 1.38 26.15 840

    160 11.54 8.69 1.35 0.0210 1.36 24.09 864

    180 11.75 8.51 1.33 0.018 1.34 20.94 885

    200 11.94 8.38 1.32 0.016 1.325 18.82 904

    220 12.12 8.25 1.31 0.0150 1.315 17.84 922

    240 12.20 8.14 1.30 0.0130 1.305 15.80 937

    . 8 y3 . y5 3 y1P sP 40 s40 =10 J m ; V sV 40 s10.5 = 10 m kg .0 0a .Andrews 1973 .

    or:

    d SdT sdT q , 7 .H S

    CV

    where g represents g of the solid state for hcp iron,sdT is the temperature increment due to isentropicS

    compression, and d SrC is the temperature incre-V

    .ment arising from P yP under compression dV.HThus d S can be considered as a type of work term,

    .dW see also Anderson, 1995, p. 321 . The well-known P and r data along the Hugoniot see Table

    .2 of Brown and McQueen, 1986 are reproduced asthe first two columns in Table 2. The first term on

    el .Fig. 4. The electronic specific heat factor at pressures along the Hugoniot. The inset shows C calculated by Stixrude et al. 1997 forVisochores vs. T. The final CelrR curve was found by assuming a temperature distribution T vs. r so as to calculate Cel vs. pressure andV V

    .then recalculate from Eq. 6 the resulting T vs. P distribution and recycle until convergence has been achieved.

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    6/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197184

    Table 3

    Work contribution to T , DTH Hel 2 aP C in terms C Total =10 Work DT : W DT DT : DT Hugoniot T,V V W S H W

    9 y3 y1 y1 2 y1 . . . . . . .=10 J m of R J kg K =10 J kg %C K K qDT K T KV S H

    0 4.461 0b40 0 650

    60 0.3 4.907 1054.8 215.0 51.0 266.0 916

    80 0.45 5.130 1766.6 344.4 50.8 395.2 1311100 0.60 5.353 2084.2 389.4 30.7 420.1 1731

    120 0.77 5.606 2491.5 444.3 30.5 474.3 2205

    140 0.94 5.859 2729.9 465.9 26.2 492.1 2698

    160 1.10 6.097 2976.2 488.1 24.1 512.1 3210

    180 1.28 6.364 3110.9 488.8 20.9 509.7 3720

    200 1.45 6.617 3245.6 490.5 18.9 509.4 4229

    220 1.55 6.766 3406.0 503.4 17.8 521.2 4750

    240 1.62 6.870 3469.1 505.0 15.8 520.8 5271

    a . . 4P yP dVq V yV d P r2; P, P , V, V from Table 2.0 0 0b .Boundary condition: Andrews 1973 .

    Also listed: DT and T .S H

    .the right of Eq. 7 arises from the adiabatic tempera-ture profile; step-by-step details of its evaluation

    along the Hugoniot are listed in Table 2. This term .may be readily evaluated since g V for e iron is

    .now known from Eq. 5 . The numerator of the .second term in Eq. 7 can be considered to be the

    work necessary to displace the isentrope to the . . 4Hugoniot: Ws 1r2 V y V d P q P y P dV .0 0

    .The chief problem in evaluating Eq. 6 consists of

    finding values of C , composed of the classicalVlattice term 3R, but with an additional term arising

    from the free electrons, Cel

    , which is linear in T andValso depends on density. The calculation of the

    fraction of dT arising from Cel must be doneH Vcautiously because Cel depends on T, and T dependsV

    el .on C . Stixrude et al. 1997 found and plotted CV Ve .of iron vs. T and r their Fig. 2 ; their plot is shown

    as an inset in Fig. 4. A solution was found for CelVvs. P; it is the prominent curve shown in Fig. 4. The

    solution for T vs. P for the Hugoniot is listed inHthe last column of Table 3, the previous columns of

    which give the step-by-step evaluation of d S. The

    integrating constant, T , is the value of T at the0

    phase boundary between the a and e phases, given .as 650 K by Andrews 1973 , Fig. 9. The solution

    . .for T P is given in Table 3, which lists T 240H H . sT 240 s5271 K 5300 K to two significantm

    . . .figures . Boness and Brown 1990 found T 240 sH .5625 K and Brown and McQueen 1986 found

    .T 240 s5510 K. Results found here are differentH

    from theirs because the Cel results of Stixrude et al.V .1997 and a different and more exact solution of

    .g V for e iron were used.

    It is significant that the temperature rise due to

    isentropic compression is less than 10% of the total

    temperature rise. Therefore it would take an unrea-

    sonable value of g to shift the Hugoniot temperaturesby a significant amount. We find that g is exactlys

    .fixed by Eq. 5 .

    5. The value of T of iron at the ICB pressurem

    .Taking T 240 s5300 K for iron, the value ofm .T 330 for iron can be found from the Lindemannm

    equation for melting. The Lindemann formulation in .T, r space is Anderson, 1995, p. 281 :

    E ln Tms2 g y 1r3 . 8 . .m /E ln r

    From Fig. 1, we see that g changes very littlembetween P s135 and 330 GPa. Thus the above

    .equation can be approximated by taking 2 g y1r3mas a number independent of density, e.g., 2 g ym

    . .1r3 s1.9 for g s1.3 across the outer coresm .density range. By integrating Eq. 8 for this special

    case, it is found that:

    1. 9T rm 22 s . 9 . / /T rm 11

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    7/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197 185

    Table 4

    Values of parameters for pure iron and the core

    bRadius and r, pure iron:core r, core A surface M mass of DV pure T purec c m ma c . .pressure km conditions area interior iron iron K

    3 y3 3 y3 13 2 24 3 y1 . . . . .10 kg m 10 kg m =10 m =10 kg m kg

    d .r 135 GPa , CMB 11.28 9.9035 15.26 1.85 4380 .r 200 GPa 12.12 5020

    y6

    .r 240 GPa 12.47 11.29 2 = 10 5300y6 .r 330 GPa , ICB 13.38 12.166 1.9 0.097 0.7 = 10 6100

    a . .For e-iron solid Mao et al. 1990 for room temperature data followed by T correction from aT.b

    PREM.c .Stacey 1992 .d .Melchior 1986 .

    .Values of r P needed for the evaluation of Eq. .9 are listed in Table 4, from which we have

    . . .r 330 rr 240 s 13.38r12.47. Thus, T 330 rm . . .T 240 s1.15 by Eq. 9 , and T 330 s 6095 Km m

    for pure iron, 6100 K to two significant figures. .Using Eq. 9 and the densities listed in Table 4,

    we find the melting point of pure iron at mantle-core .boundary MCB pressure to be about 4380 K. At

    200 GPa, T is 5020 K, considerably larger than them .experimental value reported by Boehler 1993 ,

    . .T 200 f3950 K. Boehler 1993 extrapolated hism .experimental T data and obtained T 330 s 4800m m

    K, whereas 6100 K was obtained here. .The value of T 240 obtained here involves nom

    estimates. The calculations used in finding g for eiron were based on P, V measurements on hcp iron

    .in the diamond cell Mao et al., 1990 . The determi-

    nation that the Hugoniot crosses the melting curve at .240 GPa has been made experimentally see Fig. 3 .

    The Lindemann theory is used to find the value of . .T 330 from T 240 ; g is determined from Fig. 1.m m m

    For the calculation of T along the Hugoniot, CelH Vvalues at all pressures along the Hugoniot see Fig.

    .4 are needed. These were extracted from the paper . elof Stixrude et al. 1997 , who calculated C vs.V

    volume and temperature. .My calculated value of T 330 for pure iron,m

    6100 K, agrees reasonably well with the value calcu- .lated by Poirier and Shankland 1994 using disloca-

    .tion theory, T 330 s6150 K, and the value foundm .by Stevenson 1981 . Many others have found

    . .T 330 to be near 6000 K, including Gilvarry 1956 ,m . . .Zharkov 1962 , Birch 1972 , Boschi 1975 , Liu

    . .1975 , Brown and McQueen 1986 , Anderson and . .Young 1988 , Boness and Brown 1990 , Kerley

    . .1994 , Wasserman et al. 1996 and Anderson and

    .Duba 1997 .

    6. Models of core impurity composition

    .Our value for T 330 is for pure iron. To findm .T 330 for the core, we must calculate the effect ofm

    the cores impurities on the melting temperature of

    pure iron. In looking for a plausible set of impurities

    for the core, one observes wide division of opinion

    on this matter among cosmochemists and geophysi-

    cists. Table 5 is a summary of various past modelsfor core impurity composition.

    The uncertainty in the choice of light element

    impurities has been greatly reduced by several recent

    advances. .First, Poirier 1994 provided equations and fig-

    ures that help restrict the concentration of light ele-

    ments. The concentration of all impurities must pro-

    duce a total impurity density equal to 10% of the

    density of pure iron at core conditions. This is called

    the core mass deficit, e.g., the difference between the

    density of iron and PREM density at core conditions. .The equations of Poirier 1994 allow one to derive

    the contribution of each impurity in the core to the

    10% core mass deficit. The relation between the

    weight percent of an impurity and the core mass

    deficit is shown in Fig. 5. The three lower lines, for

    oxygen, sulfur, and silicon, are taken from the analy- . .sis of Poirier 1994 see his Fig. A1 .

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    8/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197186

    Table 5

    Models of core composition impuritiesa

    Previous chemical models

    Sulfur only

    .Murthy and Hall 1970 .Ahrens 1982

    Small amount of sulfur, large amount of silicon

    .Allegre et al. 1995` .Dreibus and Palme 1996No sulfur, silicon only

    .Saxena and Benimoff 1977 .McDonough and Sun 1995

    Silicon and sulfur

    .MacDonald and Knopoff 1958

    Sulfur and oxygen

    .Boehler 1992

    Features added

    Oxygen and silicon almost mutually exclusive .Sulfur with large DV of mixing Sherman, 1997

    a

    Impurities considered are the abundant light elements Si, S, andO.

    .Second, the paper by ONeill et al. 1998 pro-vides experimental evidence on the solubility of Si,

    S, and O in an Fe-rich metal. Their results restrict

    the amount of S in an Fe-rich metal to less than 6%

    by weight because of high volatility, and they found

    Fig. 5. Curves showing the contributions of a given weight

    percent of an impurity element towards the density deficit of the

    outer core. S represents sulfur assuming an FeS ideal solution.

    S) results from the volume of mixing in the FeS solution . according to Sherman 1997 . The curves for O, S , and Si are

    . )taken from Fig. A1 of Poirier 1994 . The curve for S is found .from data presented by Sherman 1997 .

    O to be even more cosmochemically volatile than S.

    The abundance of Si is tied to the abundance of

    SiO , and there is no evidence that there was a large2enough loss of O from the primitive Earth to free

    sufficient Si for the core. ONeill et al. concluded

    that . . . from the chemical view, none of the

    cosmochemically abundant light elements appears

    able to account for the mystery light component in

    the core, at least when considered individually. .Third, Sherman 1995 pointed out that the pres-

    ence of silicon eliminates oxygen as a core impurity

    candidate because oxygen demands oxidizing condi-

    tions and silicon demands reducing conditions. .ONeill et al. 1998 came to the same conclusion

    about the incompatibility of silicon and oxygen.

    They noted, however, that a small amount of silicon

    and oxygen can coexist in iron, according to results

    known and used in the steelmaking industry, and

    they presented a curve taken from the steelmakingliterature shown in Fig. 6. By this figure, 2-1r2wt.% silicon could coexist with 2-1r2 wt.% oxygen.

    .Fourth, Sherman 1997 discovered that there is a .high excess volume of mixing density deficit in

    FeFeS alloys at 250 GPa in the FeFeS phase .diagram see Fig. 7 . This density deficit is suffi-

    ciently large that it shifts the sulfur curve towards )smaller levels of weight impurity the S curve in

    Fig. 6. The mutual dependence of the solubilities of Si and O in

    liquid Fe at 1 bar, showing that Si and O are mutually exclusive

    except along the two axes. From the steelmaking literature and . .modified from ONeill et al. 1998 their Fig. 10 .

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    9/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197 187

    Fig. 7. Density vs. composition for the FeFeS phases at 250 GPa .after Sherman, 1997 . The calculated nonideal mixing curve

    relative to Fe and FeS is compared with the ideal mixing curve.

    The first up arrow on the left corresponds to 2.3 wt.% sulfur

    projected to the calculated curve. The first down arrow on the left

    represents 4.3 wt.% from the ideal mixing curve. The second up

    arrow on the left represents 5.9 wt.% projected to the calculated

    curve. The second down arrow on the left represents 10 wt.%

    from the ideal mixing curve. A concentration of 5.9 wt.% sulfur

    satisfies the 10% density deficit of the core.

    .Fig. 5 , so that only 5.9 wt.% of sulfur would offset

    all 10% of the outer cores density deficit. ONeill et .al. 1998 conclude that the amount of S in the core

    should be less than 6 wt.% or about half the

    amount required for the density deficit, according .to Fig. A1 of Poirier 1994 .

    However, that conclusion was based on the S

    curve in Fig. 5. Taking into account the discovery of .Sherman 1997 of the high volume of mixing of

    sulfur in FeS, the S) curve in Fig. 5 was con-

    structed; it shows that sulfur could completely ac-

    count for the core deficit and be within the limits .stated by ONeill et al. 1998 .

    My approach is to find combinations of S, O, and

    Si that are within the limits described above: sulfur ) .identified in this paper as S to be held to less than

    6 wt.%, and O and Si in combination to be taken

    from the 2750 K isotherm in Fig. 6.

    Of the six proposed impurity compositional mod-

    els listed in Table 6, only the first four satisfy the

    limits stated above. The last two are for the cases of

    S alone and Si alone. For all six cases considered,

    the value of the weight percent sum is constrained to

    account for the 10% core density deficit using the

    curves in Fig. 5.

    ( )7. Depression of T ICB by impuritiesm

    Finding the temperature depression, DT, is

    straightforward if known impurities with known con-

    centrations in iron form as a solid solution, but the

    problem is complex if the solute forms a eutectic

    with the solvent. The concentrations of the impurities

    in the first four cases listed in Table 6 are suffi-

    ciently small that I assume that the impurities doindeed form a solid solution. Case f probably does

    not form a solid solution, but it is listed in order to

    show, as I do below, that the melting point depres-

    sion, DT, is unreasonably large.

    The equation used for calculating DT for outer

    core conditions is:

    DTsyT) ln 1y x , 10 . .i i ,l

    where T) is the pure iron melting point at the ICB

    pressure and x is the molar fraction of the ithi,l .impurity in the liquid outer core. Eq. 10 is derived

    in Appendix A. .Evaluating Eq. 10 for case a in Table 6, where

    x)s 0.048, x s 0.074, and x s0.032:S Si o) 4DTsyT ln 0.952 q ln 0.926 q ln 0.968

    sy0.159T) .

    Using T)s 6100 K, DTsy970 K for case a.Results from the first four cases listed in Table 6

    vary from y700 to y1070 K. It is interesting that .Poirier 1994 stated that many authors assume that

    the depression of the melting point is 7001000 K,

    but they do not generally explain how this result

    was obtained. Following the comment of Poirier .1994 that the iron core would likely have more

    than one impurity, I average the results from cases a,

    b, and c, and choose DTsy1000 K. Accordingly,6100 K y1000 K s5100 K is assigned as the value

    of T at the ICB Braginsky and Roberts, 1995,m

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    10/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197188

    Table 6

    Calculation of the freezing temperature depression, DT, resulting from various models of the composition of the core

    )Model Comment Impurities DTrT DT )S S Si O

    Case a Si-rich

    Wt.% 3 4 1

    Molar frac. 0.048 0.074 0.032 y0.159 y970

    Case b O-richWt.% 2 0.5 4

    Molar frac. 0.031 0.009 0.125 y0.1755 1070)Case c S -rich

    Wt.% 4 1.2 1.2

    Molar frac. .077 0.022 0.038 y0.143 y871)Case d S alone

    Wt.% 5.9

    Molar frac. 0.109 y0.115 y700Case e S alone

    Wt.% 11

    Molar frac. 0.177 y0.194 y1183Case f Si alone

    Wt.% 18.5

    Molar frac. 0.311 y0.372 y2271

    T) represents melting at the ICB; S represents the case of sulfur with volume of mixing in FeS solution; S) represents sulfur in an ideal

    solution of FeS. .Only light impurities considered. Cases a, b, c, and d constrained by limits set by ONeill et al. 1998 . Cases e and f presented to illustrate

    .extreme cases, though they exceeded limits set by ONeill et al. 1998 .

    estimated 5300 K, and Stacey and Stacey, 1998,.assumed 5000 K .

    The ratio of temperatures at the ICB core to pure

    . .iron , T rT , is therefore 0.936. Eq. 10 is basedm mc ion the assumption of an ideal dilute solution. Since

    both silicon and sulfur have a volume of mixing,

    neither silicon nor sulfur forms an ideal solution with

    iron at high pressure. Thus there may be some error

    in the above DT calculations. Nevertheless, the re- .sults from Eq. 10 listed in Table 6 give a good idea

    of the DT effect since the magnitude of the errors

    arising from lack of ideality is probably small com-

    pared to the variability in DT values arising from the

    choice of impurities themselves. Except for case f,

    concentrations of impurities described in the six

    models satisfy the criteria for a dilute solution. I

    conclude that an FeSi core is improbable because if

    DT due to crystallization were y2270 K, then theadiabatic T of the core at the CMB would be about

    2750 K, close to the adiabatic T of the mantle at the

    CMB. This would result in a negligible thermal

    gradient across DY

    , preventing DY

    from being a

    useful thermal boundary as required by convection .modelling. ONeill et al. 1998 also found an FeSi

    core to be improbable. They stated, . . . neither O

    nor a combination of O with Si can dissolve in liquidFe-rich metal in sufficient amounts to account for the

    presumed density deficit in the Earths outer core.

    8. The melting temperature and the thermal gra-

    dient

    Values of T and g are needed for calculationsm mof the thermal gradient. The Lindemann law in the

    .form given by Eq. 8 is useful because it is ex-

    pressed in terms of the density, values of which are . .listed in Table 4. Using Eq. 9 to find T CMBm

    .from T ICB s5100 K requires the density ratiom . . 9.904r12.166 . Thus, T CMB s 3450 K seem

    ..Table 7 for other values of T P . The equation formfinding the value of T along the isentrope anchoredad

    .at 5100 K is found from Eq. 2 . Rewriting this

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    11/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197 189

    Table 7

    Details of calculating the thermal gradient and values of thermal conductivity

    a b y2 b y1 2 c . . .Radius T , g g m s T , f km s dTrd r, Q TWm l ad cy1 . . .core K core K core K km

    d e .r 135 GPa , CMB 3450 1.33 10.68 3900 67.33 y0.82 3.4 , 5.4 .r 240 GPa 4400 1.30 7.94 4630 89.95 y0.53 .r 330 GPa , ICB 5100 1.27 4.40 5100 105.38 y0.27

    a .From Stacey 1995 .b

    From PREM.c

    Dimensions, W my1 Ky1 .d y1 y1 .ks28.6 W m km from Stacey 1972 .eks43 W my1 kmy1 from this work.

    equation in densitytemperature coordinates, we

    have:

    E ln Tads g . 11 .l

    E ln r

    .Since the volume decreases at melting Fig. 2 , it isexpected that g will be smaller than g . The vol-l mume decrease at melting is on the order of 1%. Thus,

    we take g s1.29. Since g changes very little atl l .outer core conditions, Eq. 11 becomes:

    1.29T r .ad 22s . 12 . /T r .ad 11

    Values of T for three radii in the outer core foundad . .from Eq. 12 , including T ICB , are shown inad

    Table 7. The core temperature at the CMB is 3900

    K. These values are close to those presented by .Braginsky and Roberts 1995 in their Table E2. .With these data, dTrd r can be found using Eq. 1 ;

    .values of dTrd r for the three selected radii arealso shown in Table 7.

    Now that g and T for the core at the ICB havem mbeen found, there is sufficient information to deter-

    mine the heat of crystallization, DH . The formulam .for DH , derived from Anderson and Duba 1997 ,m

    is:

    DV Km S mDH

    s,

    m 2 g y1r3 1 qag T . .m m m

    where DV is the volume change at melting, K atm Sm .330 GPa is 1370 GPa, according to PREM, a ICB

    y5 y1 .s1.7 = 10 K Stacey, 1995 , and, from thediscussion above, g s 1.3 and T s5100 K.m m

    . y6Anderson and Duba 1997 found DVs 2 = 10m3 kgy1 from Hugoniot data, so that DH s 1.38m

    = 106 J kgy1 from the above formula. To getpower, we must know the age of the inner core t in

    . .billion years . Labrosse et al. 1997 use ts1.7, and .Stacey and Stacey 1998 use ts2.3. For ts 2, the

    power, Q , is 1.5 TW. Glatzmaier and Robertsm .1996 estimated Q

    s3 TW. Stacey and Stacey

    m . 1998 estimate a much smaller value of DV they.find DVrVf1.1% , which would depress the value

    of DH to 0.5 = 106 J kgy1 and that of Q to 0.4m mTW.

    9. The thermal conductivity in the core

    We have determined g and dTrd r across thecore, but we also need the value of thermal conduc-

    tivity to find the gradient power, Q , across the corec

    side of the CMB. According to the Fourier law, .power is equal to Eq. 1 multiplied by area:

    dTQ sA k , 13 .c c c

    dz

    where A is the surface area of the core, and k isc cits thermal conductivity. There are two recent esti-

    mates of the value of k . Braginsky and Robertsc . y1 y11995 proposed 40 W m K ; Labrosse et al. . y1 y11997 proposed 60 W m K . Other estimates of

    y1 y1 .k are 35 W m K Buffett et al., 1996 , 40 Wcy1 y1 . y1 y1m K Stevenson, 1981 ; and 28.6 W m K

    .Stacey, 1972; Stacey, 1992, p. 331 .

    In order to justify the magnitude of these values

    of k for the core, we need to explore the theoretical

    and experimental background by which k is ob-ctained. The chief problem for core physics is that

    values of k are needed at temperatures very muchclarger than most measurements can be taken. Data on

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    12/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197190

    .k at P s0 or at small pressure even at high T arebeset by several transitions, as can be seen by exam-

    ining the P s0 isobar in Fig. 2. In order to find kof the core, we need data at high P and T near the

    path shown by the Hugoniot in Fig. 2. A physical

    principle called the WiedemannFranz ratio, by

    which the value of thermal conductivity in metals is

    found from measurements of the electrical conduc-

    tivity, is very helpful.

    Electrical conductivity, s, can be measured more

    easily at high pressure than thermal conductivity, k,

    because the latter is a transport property requiring

    accurate measurements of gradients, whereas the for-

    mer requires the simpler measurements of current

    and field. The WiedemannFranz ratio, by which k

    is calculated from s, is founded in the classical free

    electron model of atoms. A summary of that theory .following the derivation of Joos 1958 is presented

    in Appendix B, and the result is given by:22k p ke

    s . 15 . /s 3 ek is the electronic contribution to the thermal con-eductivity, s is the electrical conductivity, k is the

    Boltzmann constant, and e is the electrons charge. .The quantity on the right of Eq. 15 is often

    called the Lorenz number, KL : note that it is com-

    posed only of fundamental constants and numbers.

    KL can be simplified by including in the thermal

    conductivity equation the magnetic diffusivity . 5 y1Braginsky and Roberts, 1995 , hs8 = 10 s ,where h is in units of m2 sy1 and s is in units of S

    my1. For this case, k sKL = Trh and KL s 0.02eW m sy1 Ky1. The WiedemannFranz ratio is

    remarkable because, though it is derived from free

    electron theory, none of the parameters of that theory

    are retained. Many physicists early in the century did

    experiments to see whether the WiedemannFranz

    ratio is independent of T and of the metal tested. An .example of such a test is given by Sprackling 1991 ,

    p. 305, who showed that sr

    kT is the same for the

    metals copper, silver, lead, and zinc over a broad

    range of temperature. The WiedemannFranz ratio is

    also very useful for core physics because k iseproportional to sT at all pressures and temperatures.

    Thus if s is measured at a high T, then k is

    measured at the same high T. The WiedemannFranz

    ratio provides a very useful method for extrapolation

    of measurements made at accessible temperatures to

    core temperatures. Further, there is a suggestion that

    the measurement of the effect of an impurity on

    electrical conductivity may be independent of the

    kind of impurity, depending only on concentration.

    The electrical conductivity at pressures along the .Hugoniot was measured by Matassov 1977 on sam-

    ples of iron with various levels of silicon impurities.

    His shock wave results are reproduced in Fig. 8. The

    plot shows s vs. P along the Hugoniot up to 140

    GPa, barely within outer core pressure. On the

    Hugoniot the temperature increases with P, but at .maximum 140 GPa , T ; 2100 K for an ironsili-H

    con alloy, while the core temperature is about 3900 .K see Table 7 . I assume that the effect of one mole

    of Si on electrical conductivity is the same as that of

    one mole of S or one mole of oxygen. Thus a mole .fraction impurity level of roughly 0.150 Table 6 is

    Fig. 8. Electrical conductivity of Fe with various levels of silicon

    impurities at high-pressure shock conditions measured by Matassov . .1977 up to 140 GPa . The FeSi curve for x s0.181 is usedSito obtain the electrical conductivity, ss8.5=103 mho cmy1 s105 S my1 , at maximum P. Since T of the Hugoniot is about

    .2200 K, the value of s must be corrected downward to obtain .core temperatures ; 3900 K , and must be further corrected

    downward to go from the solid to the liquid state.

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    13/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197 191

    needed to achieve the 10% core deficit level. The

    FeSi curve in Fig. 8 for x s 0.181 is used to findSiss 8.5 = 10 5 mho my1 s 8.5 = 10 5 S my1.Matassov reports that in the high pressure range s

    scales as 1rT, so s at 140 GPa and the solidus T is5 . 5 y1 8.5 = 10 2100r3900 s4.6 = 10 S m follow-

    .ing Braginsky and Roberts, 1995 , but this is still in

    the solid state. Another reduction in s is required

    because melting always further reduces the conduc-

    tivity. Assume a further 10% drop due to melting,

    obtaining ss4 = 10 5 S my1. Thus, for Ts3900 Kand hs 2 m2 sy1 , k s0.02 = 3.9 = 10 3r2 s39 Wemy1 Ky1.

    .In his Table 18.8, Matassov 1977 lists g for the . 5 y1Earths core as 7.30 4.87 = 10 S m , so appar-

    ently he did not reduce s due to melting as was .done by Braginsky and Roberts 1995 . Conse-

    quently he finds k for the core to be higher, 54 We

    my1 Ky1. .Gardiner and Stacey 1971 , using data then avail-

    able on ironsilicon alloys, extrapolated the resistiv-

    ity of pure liquid iron at low pressure to core temper-

    atures and found, after correcting for the expected

    impurity concentrations, that ss3 = 10 5 S my1 ,y1 .corresponding to k s28.6 W m K Stacey, 1992 .e

    .Secco and Schloessin 1989 made measurements

    of electrical conductivity on solid and molten pure

    iron up to 7 GPa in a large volume press and found

    ss7.8 = 10 5 S my1 for pure iron at this pressure.

    They reasoned that this value could be assumed to bethe same as the conductivity of the iron diluted by

    impurities at outer core P and T. Taken at face

    value, this leads to k f77 W my1 Ky1, which iseprobably too large. References to the value of s at

    .core conditions listed in the table of Matassov 1977 ,

    may not be corrected for temperature between Hugo-

    niot T and core T and for the solidliquid state

    transition. This could lead to values on the order of

    k s 6070 W my1 Ky1.eThe value of k s3 9 W my1 Ky1 seems ae

    reasonable upper limit for the core. To this must be

    added the lattice contribution, k , so that:l

    k s k q k . 16 .c e l

    Taking k to have the same value as the lattice k inly1 y1 the deep lower mantle, k f4 W m K Kieffer,l

    . y1 y11976 . Therefore, I take k s43 W m K .c

    10. Conductive heat from the core to DY

    and the

    mantle

    Choosing k s28.6 W my1 Ky1 as the lowestclimit and k s43 W my1 Ky1 as the highest limitcfor the outer core, one calculates the limits in con-

    .ductive heat flow from the core by Eq. 13 .

    A is listed in Table 4. The two limiting values ofcthe present-day power, Q , leaving the core alongcthe gradient are 3.4 and 5.4 TW, or Q s4.4 " 1cTW. According to the curve in Fig. 2 of Buffett et al. .1992 , the power leaving the core is Q s 5 TWc .corresponding to their ts2 billion year lifetime

    .curve .

    The power from the surface of the Earth is 44 =12 . .10 W 44 TW , according to Pollack et al. 1993 .

    Therefore the conductive power flowing into the

    mantle is about 10% of the surface power leaving

    Earth.c .On the core side of the CMB, T 2970 km s 3900

    K, as given by T in Table 7. Stacey and Loperad . Y . y11983 found dTrdZ D s9.6 K km . The changein temperature across D

    Yis thus f1440 K for a

    thickness of DY s 150 km. Therefore, on the mantle

    m .side of the CMB, T 2820 km f 2460 K, which .agrees with the value Brown and Shankland 1981

    found from their isentrope of the lower mantle. .The Stacey and Loper 1983 thermal gradient in

    DY

    is 12 times the thermal gradient of the core at the

    CMB, and therefore DY

    would conduct the samepower as the core if k Y is 1r12 that of k , e.g.,D ck Y f3.7 W my1 Ky1. This suggests that DY isDcomposed primarily of mantle material with little

    core material contained in it. I therefore conclude

    that DY

    is a thermal boundary layer, not a composi-

    tional boundary layer.

    11. Thermal conductivity and the power flowing

    into the lower mantle

    It is commonly accepted that the thermal conduc-

    tivity in the deep lower mantle is about 45 W my1

    y1 .K . Kieffer 1976 estimated that k s 4.2 Wlmmy1 Ky1 by calculating the heat transfer due to

    phonons in a dielectric solid. If the deep mantle has

    this value of k , then the lower mantle couldlmconduct less than 4.2r43s11% of the cores con-

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    14/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197192

    ductive power, leaving one to find some other mech-

    anism of heat transfer for the large amount of resid- .ual power 3.9 TW . However, in the assumption

    that k is the same as the lattice thermal conductiv-lmity, there is the implicit assumption that conductivity

    is determined by phonon transfer alone. This implicit

    assumption will probably result in an underestima-

    tion of the capacity of the lower mantle to conduct

    power towards the surface. .In addition to lattice thermal conductivity, k ,L

    there is also likely radiative transfer connected with

    infrared electromagnetic waves arising from the opti-

    cal properties of a dielectric solid. The radiative

    thermal conductivity associated with infrared electro-

    magnetic waves, k , is given by Zharkov and Tru-R .bitsyn 1978 , p. 57, according to the equation:

    2 316 s)n T

    k s " ,R 3 a

    where s) is the StefanBoltzmann constant, a is

    the absorption coefficient, and n is the refractive

    index. Note the T3 factor, which has a pronounced

    effect at lower mantle temperatures. In a recent .report, Hofmeister 1998 showed that the optic

    modes are the primary mode of heat transport deter-

    mining thermal conductivity at lower mantle condi-

    tions, contrary to the assumption made in the calcula-

    tion of k where phonons are considered as thelm

    only transport mechanism. Taking into account re-cent infrared reflectivity measurements and correct-

    ing for the pressure dependence of the constants, she

    found that k for the deep lower mantle is suffi-Rciently large that k q k is about 2.3 times theL Rvalue of k .L

    Thus we have sufficient evidence to suppose that

    k is roughly 10 W my1 Ky1 rather than about 4.2lmW my1 Ky1. Taking the thermal gradient of the

    y1 lower mantle to be 0.5 K km Brown and Shank-.land, 1981 , 0.6 TW and perhaps more is transported

    conductively towards the lithosphere. Could the

    residual of the cores conductive power, 3.8 TW, be

    transported upwards in the lower mantle by convec-

    tion? . Tackley et al. 1993, 1994 and Tackley 1995,

    .1996 have shown that the convective heat in the

    mantle is transported to the surface by a number of

    mantle plumes originating at the base of DY

    . When

    these plumes interact with the lithosphere, they cre-

    ate hotspot intrusions that can be studied by geo-

    physical methods. The number of plumes that are

    currently interacting with the Earths surface equals

    the number of known hotspot intrusions. .G. Schubert private communication reported that

    he, Don Turcotte and Peter Olson, who are writing a

    book on mantle geodynamics, have identified a total

    of 38 current hotspot intrusions. They have calcu-

    lated that these intrusions collectively transport 2.3

    TW of power to the Earths surface. Thus the lower

    mantles conductive heat, 0.6 TW, and the convec-

    tive heat from the known hotspot intrusions, 2.3 TW,

    yield 2.9 TW, close to the lower limit of our sug-

    gested core conductive power, 3.4 TW. In addition

    to the plumes evidenced by hotspot intrusions, there

    are, however, plumes that have not yet been made

    evident by surface activity, as revealed by the snap-

    shots of plume activity given by numerical simula-tion done by Tackley and his colleagues. Plumes

    grow from DY

    , slowly ascend to the surface, contact

    the surface, detach from DY

    , and eventually disap-

    pear, but, at any given time there are more active

    plumes than the number of plumes in contact with .the surface giving hotspot intrusions . Thus, a rea-

    sonable assumption is that the number of hot plumes

    that have not yet made contact with the surface is

    about half of the number that have been identified

    through surface activity. I assume that the total

    convective power arising from DY

    via the plumes is1.5 times the power for known hotspot intrusions

    calculated by Schubert and his colleagues that is, a.total of about 3.5 TW . Under this assumption, the

    total estimated power coming from the CMB both.convective and conductive would be about 4.1 TW,

    close to the mean of the upper and lower limits of .conductive core power 4.4 TW . I therefore conjec-

    ture that a rough balance exists between the conduc-

    tive power transmitted from the core to the mantle

    and the power transferred by conduction and convec-

    tion from the upper mantle towards the lithosphere.

    12. The convective heat in the outer core

    The outer core must be in a state of convection so

    that the Earths magnetic field can be maintained.

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    15/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197 193

    Maintenance of the magnetic field requires a smallfraction of a TW 0.2 TW, according to Stacey and

    . .Loper, 1983 in mechanical energy. Loper 1984 .and Loper and Roberts 1983 have proposed that the

    mechanical energy requirements may be largely sup-

    plied by vigorous convective motion driven bycompositional buoyancy of the ejecta enriched

    .with light elements from the inner core to the outer

    core. The convective power arising from the core

    may be small compared to its conductive adiabatic .power. Buffett et al. 1996 emphasized that a

    modest heat flux in excess of that conducted down

    the adiabatic gradient is sufficient to power the

    geodynamo, even in the absence of compositional .convection and latent heat release. Gubbins 1977

    stated that the gravitational energy released by

    rearrangement of matter in the core is completely

    converted to magnetic dissipation enabling a large

    magnetic field to be generated with a low heat flowfrom the core. Although convective heat may be

    transferred from the core to DY

    and then on to the

    lower mantle, there appears to be no convincing

    evidence that the cores convective power is large

    compared to that conducted down the adiabatic gra-

    dient. Any excess power may be returned to the core .by compositional convection Loper, 1978a or con-

    tained by a subadiabatic thermal gradient near the

    boundary of DY

    , as proposed by Labrosse et al. .1997 . In my proposed model, the mantle success-

    fully transports away the conductive power arisingfrom the cores adiabatic gradient. Therefore, I as-

    sume that a large subadiabatic gradient is not neces-

    sary and that any power from core convection alone

    in excess of the power conducted down the adiabatic

    gradient will be returned to the core by composi-

    tional convection.

    Acknowledgements

    I acknowledge helpful comments from Frank

    Stacey, Stanislav Braginsky, Paul Roberts, Gary

    Glatzmaier, Jerry Schubert, and Paul Tackley on the

    geophysical aspects of an early version of this paper.

    I acknowledge helpful comments from Dave Sher-

    man, Giulio Ottonello, Surendra Saxena, Francois

    Robert, and John Wasson on its geochemical aspects.

    The critiques of two unknown reviewers were also

    quite helpful. Support from NSF grant EAR-96-

    14654 acknowledged. Support by ONR acknowl-

    edged. IGPP contribution no. 5073.

    Appendix A

    Calculation of the temperature depression requires

    the assumption of thermodynamic equilibrium be-

    tween the solid and liquid phases. As a beginning,

    assume that all impurities reside in the liquid. Be-

    cause the condition of thermal equilibrium requires

    that the Gibbs free energy be equal for both phases .DG s0 , we must deal with the chemical potentialof the pure liquid and the pure solid. Stevenson .1981 showed that if all the impurities reside in the

    liquid, the thermal equilibrium arising from DG s0

    is given by:

    ml P ,T qRT ln 1y x s ms P ,T , A1 . . . .0 i ,l 0

    where m is the chemical potential, and x is the0 i,lmolar fraction of the ith species in the liquid and

    where the super and subscripts l and s refer to the

    liquid and solid, respectively.

    Note that the weight percent of an impurity ele-

    ment was used in calculating the fraction of the

    cores density deficit that it can account for. But

    now, using thermodynamic functions to obtain freez-

    ing depression, we need to express concentrations inmolar fraction because molar fraction identifies the

    number of atoms on a particular thermodynamic site.

    The molar fraction of the impurities in the six cases

    is listed in Table 6. .Eq. A1 is oversimplified for use in core theory

    because we now are sure that light impurities exist in

    the solid inner core. As the core solidifies, a large .fraction less than unity of the impurities passes into

    the liquid, leaving a small fraction behind in theinner core see, for example, Jephcoat and Olson,

    .1987 . Impurities in the solid inner core are ac-counted for by adding a term to the right of Eq. .A1 , giving:

    ml P ,T q TR ln 1 yx . .0 i ,l

    sms P ,T q TR ln 1 yx , A2 . . .0 i ,s

    where x - x .i,s i,l

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    16/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197194

    Assuming that v is the fraction of all impurities .existing in the outer core, then 1 yv is the fraction

    residing in the inner core. We make the approxima- .tion that the second term on the left of Eq. A2 and

    .the second term on the right of Eq. A2 are multi- . plied by v and 1y v , respectively it would be

    more accurate to multiply all x by v and all xi,l i,s .by 1 yv , but for small concentrations the approxi-

    .mation used here is sufficiently accurate . Thus, Eq. .A2 can be replaced by the approximation:

    ml P ,T q vRT ln 1y x . .0 i ,l

    sms P ,T q 1y v RT ln 1y x . . . .0 i ,s

    Using the general relationship Dm sDHy TDS,where H is enthalpy and S is entropy, and referenc-

    .ing Chapter 9 of Landau and Lifshitz 1958 as the

    . .authority, Stevenson 1981 replaced Eq. A1 elimi-nating ms, leaving the following for the freezing

    depression, DT, of the outer core:

    RT)

    DTsyv ln 1 yx .i ,l /DSvT)

    sy ln 1 yx , A3 . .i ,lln 2

    where DS is the entropy of melting and v has been

    incorporated. For the core, T) is 6100 K. The

    parameter v takes into account the fact that the inner

    core contains some impurities. Thus:

    T)vDTsy ln 1 yx . A4 . .i ,l

    0.693 .

    .Anderson and Duba 1997 calculated that at the . y3ICB, Dr freezing s200 kg m . Masters and

    .Shearer 1990 found from seismic data on the core . y3 that Dr ICB

    s550 kg m , leaving Dr chemical

    . y3differentiation s350 kg m . The outer core has adensity 1270 kg my3 lower than that of pure iron so

    that the concentration of impurities in the outer core

    accounts for 1270y350 or 920 kg my3. Thus vs920r1270s 0.72. Within the approximations cre-ated by the uncertainty in exact chemical composi-

    tion, the value of v is close enough to the value in

    .the denominator of Eq. A4 that vrln 2 can be .replaced by unity. Thus, Eq. A4 is simplified to:

    DTsyT) ln 1 yx . A5 . .i ,l

    Appendix B

    The free electron model of metals, which pre-

    ceded wave mechanics, was very successful in the

    formulation of a number of properties of metals,

    including electrical conductivity. In this classical

    model, the valence electrons are able to move about

    the lattice freely having intermittent reactions with

    the lattice. Five parameters are of significance in the

    free electron theory of metals: the drift velocity of

    the electrons, y; the relaxation time, t, measuring

    the time between collisions with the lattice; thenumber of electrons per unit volume, n; and the

    mass and electrical charge of the electron, m and e.

    The electrical conductivity, s, of a metal is the

    parameter relating the current j to the electrical field

    E, where:

    j ssE. B1 .

    Separate solutions of y with E giving ys . .errm E and y with J Jsney combined with Eq. .B1 eliminate the parameters J, E, and . We then

    find the electrical conductivity in terms of the pa-

    rameters of the free electron model:

    ssne2trm. B2 .

    We proceed to find the thermal conductivity of

    the free electron model. In solids, k is found from .the Fourier law, Eq. 13 , so that:

    Q dzk solid s . . /A dTAnalogous to this is the thermal conductivity of

    gases, which, according to the kinetic theory of

    gases, is:

    k gases s 1r3 Cul, . .

    where u is the average particle velocity in the direc-

    tion of heat flow, l is the mean free path, and C is

    the contribution of each gas molecule to the specific .heat per unit volume. In the equation for k solid ,

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    17/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197 195

    we note that k is inversely proportional to the

    temperature gradient, dTrdz. The thermal gradientis also found in k of gases, located in C, the specific

    . . .heat: Cs dErdT s dErdz dzrdT . Thus k ofgases is also inversely proportional to the thermal

    .gradient. The 1r3 comes from integration of thecosine angle, because the particles will be travelling

    in all directions with the velocity . The need to find

    the vector component in one selected direction leads,

    by integration, to the factor 1r3. The specific heat,C, is the product of n and k, where n is the number

    of molecules per unit volume, and k is the energy of

    one particle with one degree of freedom, the Boltz-

    mann constant.

    The conduction electrons are in a state of chaotic

    thermal agitation, somewhat like the atoms in an

    ordinary gas, except that l of free electrons is the

    measure of the distance between collisions of the

    electron with the lattice. The equation for k for freeeelectrons is the same as for gases:

    1k free electrons s nkul, B3 . .e /3where k is the electronic contribution to the totalethermal conductivity, the subscript emphasizing that

    the lattice conductivity, k , also has yet to be takenl . .into account. Dividing Eq. B3 by Eq. B2 :

    k 1 nkulmes .2s 3 ne t

    Using tslru:

    k 1 ke 2s u m. B4 .2 /s 3 eUsing the theorem of equal partition of thermal

    and kinetic energy:

    12mu skT.

    2

    .Eq. B4 is transformed into the WiedemannFranz

    ratio, giving:

    2k 2 kes . B5 . /sT 3 e

    References

    Ahrens, T.J., 1982. Constraints on core composition from shock-

    wave data. Phil. Trans. R. Soc. London A 306, 3747.

    Allegre, C.J., Poirier, J.-P., Humler, E., Hofmann, A.W., 1995.`The chemical composition of the Earth. Earth Planet. Sci. Lett.

    134, 515526.

    Altshuler, L.V., Sinakov, G.V., Irunin, R.F., 1962. On the com-position of the Earths core. Isv. Earth Phys. 1, 13.

    Anderson, O.L., 1995. Equations of State of Solids for Geo-

    physics and Ceramic Science. Oxford Univ. Press, New York.

    Anderson, O.L., Young, D., 1988. Crystallization of the Earths .inner core. In: Smylie, D.E., Hyde, R. Eds. , Structure Dy-

    namics of Earths Deep Interior, Geophys. Monogr. Ser., 46.

    Am. Geophys. Union, pp. 8389.

    Anderson, O.L., Duba, A., 1997. Experimental melting curve of

    iron revisited. J. Geophys. Res. 102, 2265922669.

    Andrews, D.J., 1973. Equation of state of the alpha and epsilon

    phases of iron. J. Phys. Chem. Solids 34, 825840.

    Birch, F., 1972. The melting relations of iron and temperatures in

    the earths core. Geophys. J. R. Astr. Soc. 29, 373387.

    Boehler, R., 1992. Melting of the FeFeO and the FeFeS

    systems at high pressure: Constraints on core temperatures.

    Earth Planet. Sci. Lett. 111, 217 227.

    Boehler, R., 1993. Temperatures in the Earths core from

    melting-point measurements of iron at high static pressures.

    Nature 363, 534536.

    Boehler, R., Ramakrishnan, J., 1980. Experimental results on the

    pressure dependence of the Gruneisen parameter. A review. J.Geophys. Res. 85, 69967002.

    Boness, D.A., Brown, J.M., 1990. The electronic band structures

    of iron, sulfur and oxygen at high pressures and the Earths

    core. J. Geophys. Res. 95, 2172121730.

    Boschi, E., 1975. The melting relations of iron and temperatures

    in the Earths core. Riv. Nuovo Cim. 5, 501531.Braginsky, S.I., Roberts, P.H., 1995. Equations governing the

    convection in Earths core and the geodynamo. Geophys.

    Astrophys. Fluid Dyn. 79, 197.

    Brown, J.M., Shankland, T.J., 1981. Thermodynamic parameters

    in the earth as obtained from seismic profiles. Geophys. J. R.

    Astr. Soc. 66, 579596.

    Brown, J.M., McQueen, R.G., 1986. Phase transitions, Gruneisenparameter and elasticity for shocked iron between 77 GPa and

    400 GPa. J. Geophys. Res. 91, 74857494.

    Buffett, B.A., Huppert, H., Lister, J.R., Woods, A.W., 1992.

    Analytical model for solidification of the Earths core. Nature

    35, 329331.

    Buffett, B.A., Huppert, H.E., Lister, J.R., Woods, A.W., 1996. On

    the thermal evolution of the Earths core. J. Geophys. Res.

    101, 79898006.

    Chen, G.Q., Ahrens, T.J., 1997. Sound velocities of g- and liquid .iron under dynamic compression, abstract . EOS Trans. Am.

    Geophys. Union 78, F757, Suppl.

    Dreibus, G., Palme, H., 1996. Geochemical constraints on the

    sulfur content of the Earths core. Geochim. Cosmochim. Acta

    60, 11251130.

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    18/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197196

    Gardiner, R.B., Stacey, F.D., 1971. Electrical resistivity of the

    core. Phys. Earth Planet. Sci. 4, 406410.

    Gilvarry, J.J., 1956. The Lindeman and Gruneisen laws. Phys.Rev. 102, 308316.

    Glatzmaier, G.A., Roberts, P.H., 1996. An anelastic evolutionary

    geodynamo stimulation driven by compositional and thermal

    convection. Physica D 97, 8194.

    Gubbins, D., 1977. Energetics of the Earths core. J. Geophys. 43,

    453464.Hofmeister, A.M., 1998. Mantle values of thermal conductivity

    from infrared phonon lifetimes: relationships between trans- .port and thermodynamic properties abstract S31A-08 . EOS

    Trans. AGU Suppl., p. S214.

    Jephcoat, A.P., Olson, P., 1987. Is the inner core of the Earth pure

    iron?. Nature 325, 325335.

    Joos, G., 1958. Theoretical Physics, 3rd edn., Chapter XXIV.

    Blackie, London.

    Kerley, G.I., 1994. A new multiphase equation of state of iron. In: .Schmidt, S.C., Shaner, J.W., Samara, G.A., Ross, M. Eds. ,

    High-Pressure Science and Technology-1993. American Insti-

    tute of Physics, College Park, MD, pp. 903906.

    Kieffer, S.W., 1976. Lattice thermal conductivity within the Earth

    and considerations of a relationship between the pressure

    dependence of the thermal diffusivity and the volume depen-

    dence of the Guneisen parameter. J. Geophys. Res. 81, 30253030.

    Landau, L.D., Lifshitz, E.M., 1958. Statistical Physics. Pergamon,

    London.

    Labrosse, S., Poirier, J.-P., le Mouel, J.-L., 1997. On cooling ofthe Earths core. Phys. Earth Planet. Inter. 99, 117.

    .Liu, L., 1975. On the g,e,l triple point of iron and the Earths

    core. Geophys. J. R. Astr. Soc. 43, 697705.

    Loper, D.E., 1978a. The gravitationally powered dynamo. Geo-

    phys. J. R. Astr. Soc. 54, 389404.

    Loper, D.E., 1978b. Some thermal consequences of a gravitation-

    ally powered dynamo. J. Geophys. Res. 83, 59615970.Loper, D.E., 1984. Structure of the core and lower mantle. Adv.

    Geophys. 26, 134.

    Loper, D.E., Roberts, P.H., 1983. Compositional convection and .the gravitationally powered dynamo. In: Soward, A.M. Ed. ,

    Stellar and Planetary Magnetism. Gordon and Breech, New

    York, pp. 297327.

    MacDonald, G.F.K., Knopoff, L., 1958. On the chemical composi-

    tion of the outer core. Geophys. J. R. Astr. Soc. 1, 284297.

    Mao, H.K., Wu, Y., Chen, L.C., Shu, J.F., Jephcoat, A.P., 1990.

    Static compression of iron to 300 GPa and Fe Ni alloy to0. 8 0.2260 GPa: implications for composition of the core. J. Geo-

    phys. Res. 95, 2173721742.

    Masters, T.G., Shearer, P.M., 1990. Seismological constraints on

    the structure of the Earths core. J. Geophys. Res. 95, 21691

    21695.

    Matassov, G., 1977. The electrical conductivity of ironsilicon

    alloys at high pressures and the Earths core. PhD thesis,

    Lawrence Livermore Laboratory, Rpt. UCRL-52322, Liver-

    more, CA.

    McDonough, W.F., Sun, S.S., 1995. The composition of the earth.

    Chem. Geol. 120, 223253.

    McQueen, R.G., 1964. Laboratory techniques for very high pres-

    sures and the behavior of metals under dynamic loading. In: .Schneider, K.A.G., Hepworth, M.T., Parlee, N.A.D. Eds. ,

    Conference on Metallurgy at High Pressure, Vol. 22. Gordon

    & Breach, New York, pp. 44132.

    Melchior, P., 1986. The Physics of the Earths Core: an Introduc-

    tion. Pergamon, Oxford.

    Murthy, V.R., Hall, H.T., 1970. The chemical composition of the

    Earth core: Possibility of sulfur in the core. Phys. Earth Planet.Inter. 2, 276282.

    ONeill, H.St.C., Canil, D., Rubie, D.C., 1998. Oxide-metal equi-

    libria to 25008C and 25 GPa. Implications for core formation

    and the light component in the Earths core. J. Geophys. Res.

    103, 1223912260.

    Poirier, J.-P., 1994. Light elements in the Earths outer core: a

    critical review. Phys. Earth Planet. Inter. 85, 319337.

    Poirier, J.-P., Shankland, T.J., 1994. Dislocation theory of melting

    for iron, revisited. In: Schmidt, S.C., Shaner, J.W., Samara, .G.A., Ross, M. Eds. , High-Pressure Science and Technol-

    ogy. American Institute of Physics, College Park, MD, pp.

    927930.

    Pollack, H.N., Hurter, S.J., Johnson, J.R., 1993. Heat flow from

    the Earths interior: analysis of the global data set. Rev.

    Geophys. 31, 267280.

    Saxena, S.K., Benimoff, A., 1977. Formation of FeNiSi plane-

    tary cores. Nature 270, 333334.

    Secco, R.A., Schloessin, H.H., 1989. The electrical resistivity of

    solid and liquid Fe at pressures up to 7 GPa. J. Geophys. Res.

    94, 58875893.

    Sherman, D.M., 1995. Stability of possible FeFeS and FeFeO

    alloy phases at high temperature and the composition of the

    Earths core. Earth Planet. Sci. Lett. 132, 87 89.

    Sherman, D., 1997. The composition of the Earths core: new

    constraints on S and Si versus temperature. Earth Planet. Sci.

    Lett. 153, 149155.

    Sprackling, M., 1991. Thermal Physics. American Institute ofPhysics, New York, p. 305.

    Stacey, F.D., 1972. Physical properties of the Earths core. Geo-

    phys. Surv. 1, 99119.

    Stacey, F.D., 1992. Physics of the Earth. 3rd edn. Brookfield

    Press, Kenmore, Queensland, Australia.

    Stacey, F.D., 1995. Theory of thermal and elastic properties of the

    lower mantle and core. Phys. Earth Planet. Inter. 89, 219246.

    Stacey, F.D., Loper, D.E., 1983. The thermal boundary layer

    interpretation of DY

    and its role as a plume source. Phys. Earth

    Planet. Inter. 33, 4555.

    Stacey, F.D., Stacey, C.H.B., 1998. Gravitational energy of core

    evolution: implications for thermal history and geodynamo

    power. Submitted to Phys. Earth Planet. Inter.

    Stevenson, D.J., 1981. Models of the Earths core. Science 214,

    611619.

    Stixrude, L., Wasserman, E., Cohen, R.E., 1997. Composition and

    temperature of Earths inner core. J. Geophys. Res. 102,

    2472924739.

    Tackley, P.J., 1995. Mantle dynamics: Influence of the transition

    zone. U.S. National Report to International Union of Geodesy

    and Geophysics, 19911994 American Geophysical Union,

  • 7/27/2019 1-s2.0-S003192019800123X-main(1)

    19/19

    ( )O.L. AndersonrPhysics of the Earth and Planetary Interiors 109 1998 179197 197

    Washington, DC, Rev. Geophys. Suppl., July 1995, pp. 275

    282.

    Tackley, P.J., 1996. Effects of strongly variable viscosity on

    three-dimensional compressible convection in planetary man-

    tles. J. Geophys. Res. 101, 33113332.

    Tackley, P.J., Stevenson, D.J., Glatzmaier, G.A., Schubert, G.,

    1993. Effects of an endothermic phase transition at 670 km

    depth in a spherical model of convection in the Earths mantle.

    Nature 361, 699704.Tackley, P.J., Stevenson, D.J., Glatzmaier, G.A., Schubert, G.,

    1994. Effects of multiple phase transitions in a 3-D spherical

    model of convection in the Earths mantle. J. Geophys. Res.

    99, 1587715901.

    Verhoogen, J., 1980. Energetics of the Earth. National Academy

    of Sciences, Washington, DC, pp. 5165.

    Wasserman, E., Stixrude, L., Cohen, R.E., 1996. Thermal proper-

    ties of iron at high pressures and temperatures. Phys. Rev. B

    53, 82968309.

    Zharkov, V.N., 1962. Physics of the Earths core. Inst. Earth

    Phys. Acad. Sci. USSR Engl. Transl. 20, 187.

    Zharkov, V.N., Trubitsyn, V.P., 1978. Physics of Planetary Interi-

    .ors. In: Hubbard, W.B. Ed. , Astronomy and AstrophysicsSeries, Vol. 6. Pachart Publishing Housing, Tucson, AZ, p. 57.


Recommended