Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
Phil. Trans. R. Soc. A (2008) 366, 4315–4337
doi:10.1098/rsta.2008.0147
Published online 30 September 2008
The redox state of the mantle during and just
after core formation
B Y D. J. F ROST *, U. M ANN † , Y. A SAHARA ‡
AND
D. C. R UBIE
Bayerisches Geoinstitut, University of Bayreuth, Bayreuth 95440, Germany
Siderophile elements are depleted in the Earth’s mantle, relative to chondritic
meteorites, as a result of equilibration with core-forming Fe-rich metal. Measurements
of metal–silicate partition coefficients show that mantle depletions of slightly siderophile
elements (e.g. Cr, V ) must have occurred at more reducing conditions than those
inferred from the current mantle FeO content. This implies that the oxidation state (i.e.
FeO content) of the mantle increased with time as accretion proceeded. The oxygen
fugacity of the present-day upper mantle is several orders of magnitude higher than the
level imposed by equilibrium with core-forming Fe metal. This results from an increase in
the Fe2O3 content of the mantle that probably occurred in the first 1 Ga of the Earth’s
history. Here we explore fractionation mechanisms that could have caused mantle FeO
and Fe2O3 contents to increase while the oxidation state of accreting material remained
constant (homogeneous accretion). Using measured metal–silicate partition coefficients
for O and Si, we have modelled core–mantle equilibration in a magma ocean that became
progressively deeper as accretion proceeded. The model indicates that the mantle would
have become gradually oxidized as a result of Si entering the core. However, the increase
in mantle FeO content and oxygen fugacity is limited by the fact that O also partitions
into the core at high temperatures, which lowers the FeO content of the mantle.
(Mg,Fe)(Al,Si)O3 perovskite, the dominant lower mantle mineral, has a strong affinity
for Fe2O3 even in the presence of metallic Fe. As the upper mantle would have been poor
in Fe2O3 during core formation, FeO would have disproportionated to produce Fe2O3
(in perovskite) and Fe metal. Loss of some disproportionated Fe metal to the core would
have enriched the remaining mantle in Fe2O3 and, if the entire mantle was then
homogenized, the oxygen fugacity of the upper mantle would have been raised to its
present-day level.
Keywords: perovskite; oxygen fugacity; lower mantle; accretion; magma ocean;
element partitioning
* Author for correspondence ([email protected]).
†
Present address: Institute for Mineralogy and Petrology, ETH Zurich, Clausiusstrasse 25, NW,
8092 Zurich, Switzerland.
‡
Present address: Spring 8 Material Structure Group I, Japan Synchrotron Radiation Research
Institute, Kouto, Sayo-cho, Sayo-gun, Hyogo 679-5198, Japan.
One contribution of 14 to a Discussion Meeting Issue ‘Origin and differentiation of the Earth: past
to present’.
4315
This journal is q 2008 The Royal Society
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
4316
D. J. Frost et al.
1. Introduction
The redox state within the proto-Earth and its constituent planetary embryos
controlled the partitioning of siderophile elements between the mantle and core.
The cumulative effect of this partitioning process is reflected in the present-day
mantle concentrations of siderophile elements when compared with those in
chondritic meteorites. It has been frequently noted that siderophile element
depletions in the mantle are inconsistent with equilibration with Fe metal at a
single oxygen fugacity and, in several studies, it has been proposed that oxygen
fugacity evolved with time during accretion from reduced conditions at the start
to more oxidized towards the end (Wänke 1981; O’Neill 1992; Wood et al.
2006). Seemingly in agreement with this evolutionary trend, the oxygen fugacity of
the present-day upper mantle is several orders of magnitude higher than that
imposed by equilibrium with metallic Fe during core formation (O’Neill & Wall
1987; Wood 1991). The timing and extent to which the Earth’s mantle became
oxidized would have had a major impact on its evolution, and identifying the
mechanisms by which this occurred should place important constraints on
planetary formation processes. The redox path taken during accretion would
have strongly influenced the nature of the light alloying element (or elements) in the
Earth’s core (O’Neill & Palme 1998). The speciation of degassing volatiles from the
mantle would have also been a function of the oxidation state of the early mantle,
which may have consequently influenced the development of the hydrosphere and
atmosphere (Kasting et al. 1992). Major chemical and physical differences between
terrestrial planets were probably dictated by differences in their redox evolution,
placing great importance on the comparative study of planetary redox states.
An important underlying question is whether mantle oxidation occurred as a
result of a change in the redox state of material that was being accreted to the
Earth, i.e. heterogeneous accretion (Wänke 1981; O’Neill 1992), or whether
there are internal processes that could have led to an increase in mantle oxygen
fugacity while retaining a single homogeneous accreting composition (Wade &
Wood 2005). Integral to most ‘homogeneous’ accretion models is the concept that
changes in metal–silicate partition coefficients at high pressures and temperatures
may negate the necessity for changing fO2 during core formation (Li & Agee 1996;
Chabot & Agee 2003; Wade & Wood 2005). In addition, however, several
fractionation processes involving Fe, O and Si have been identified, which could
have changed the oxidation state of the mantle, even if the redox state of accreting
material remained essentially constant (Javoy 1995; Frost et al. 2004; Rubie et al.
2004; Galimov 2005). It is, therefore, important to determine the effectiveness of
these processes and the conditions under which they may have operated.
Here we examine internal processes that could have led to changes in the redox
state of the mantle during and after core formation. Throughout this analysis, we
pay special attention to determining whether fO2 varied due to heterogeneous or
homogeneous processes and how the mantle attained its present-day ferric and
ferrous Fe concentrations.
2. The evolution of mantle fO2
Siderophile elements are generally divided into three groups based on their
1 bar metal–silicate partition coefficients (figure 1), slightly siderophile
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
4317
Core formation and redox state of mantle
10.00
transitional
silicate earth /CI chondrite
(Ti normalized)
refractory
1.00
0.10
moderately
volatile
lithophile
Ta
Zr Al REE TiCa
slightly
siderophile
vola
t
Mg Si
Nb
V
moderately
siderophile
W
Cr
Fe
Co
Li
Mn Rb
Ni
Mo
ility
tren
d
F
B
As
0.01
0.001
2000
K
Na Ga
Cu
P
highly siderophile
Re
PGE
volatile
Zn
Sb
Cl
Sn
Ge
Br
Ag
Au
Se
Te
1800
In
1600
1400
1200
1000
800
50% condensation temperature (K) 10 – 4 bar
S
600
400
Figure 1. Mantle element abundances (McDonough & Sun 1995) normalized to those in CI
chondrites and Ti. Siderophile elements have metal–silicate partition coefficients that are greater
than 1 and were therefore depleted from the mantle during core formation. Transposed upon the
effects of siderophile behaviour is an additional depletion trend resulting from volatility (circles),
which is a broad function of the 50% condensation temperature. Siderophile elements can be
divided into three basic groups, slightly siderophile (unfilled symbols), moderately siderophile
(grey) and highly siderophile (black squares).
(e.g. V, Cr and Mn), moderately siderophile (e.g. Ni, Co and W) and highly
siderophile (e.g. Pt and Ir). Based on these 1 bar partition coefficients, however,
the mantle depletions of these three element groups cannot be reconciled with
core–mantle fractionation at a single oxygen fugacity (Ringwood 1966; Wänke
1981; O’Neill 1992). The mantle depletions of slightly siderophile elements such
as V or Cr would have required core–mantle equilibration at conditions more
reducing than those required for the siderophile or highly siderophile depletions.
Even within the group of siderophile elements, Ni and Co 1 bar partition
coefficients are inconsistent with their concentrations in the mantle. To explain
these inconsistencies, homogeneous accretion models require that core–mantle
equilibration occurred at high pressures and temperatures where element metal–
silicate partition coefficients reached values that can explain mantle abundances
(Li & Agee 1996; Righter et al. 1997), even if the Earth accreted from material
with a broadly constant composition and redox state. Metal–silicate partitioning
has, therefore, been envisaged to have occurred at the base of a deep silicate
magma ocean. Currently, however, it is not possible to find either a set or range
of pressure–temperature conditions where metal–silicate partitioning at a single
f O2 can explain concentrations of all three groups of siderophile elements (Wood
et al. 2006). Highly siderophile elements, for example, remain excessively
depleted in a homogeneous accretion scenario and require the addition of a final
late veneer of material that did not undergo core formation (Holzheid et al.
2000). Although it has been proposed that at high temperatures and pressures
slightly siderophile (e.g. V, Cr) and siderophile element depletions could be
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
4318
D. J. Frost et al.
produced at a single fO2 (Gessmann & Rubie 2000; Chabot & Agee 2003; Wade &
Wood 2005), the temperatures required are much higher than the liquidus of a
likely magma ocean composition and therefore cannot realistically describe the
conditions at the base of a magma ocean (Wade & Wood 2005). Recent data can also
be used to challenge the concept that conditions exist where the mantle
concentrations of Ni, Co and Fe could have been obtained simultaneously
through core–mantle equilibration (Kegler et al. 2008). To overcome this
problem, recent models (Wade & Wood 2005; Wood et al. 2006) have returned
to the concept that fO2 increased during accretion from initially reducing
conditions, to explain slightly siderophile element depletions, to more oxidized
conditions towards the end. This would require initially FeO-poor silicate
material to become FeO rich with time while silicate–metal equilibration
continued. This concept is similar to the traditional heterogeneous models
(Wänke 1981; O’Neill 1992), which argued for initially very reducing conditions
to account for slightly siderophile element depletions, followed by accretion of
more oxidized material, where either a sulphide liquid or no liquid separated to
the core. It has been argued in recent studies, however, that changes in the
Earth’s redox state were caused by internal fractionation processes rather than
the accretion of more oxidized material (Galimov 2005; Wade & Wood 2005).
Core formation models, therefore, seem to have converged on the idea that the
redox state, i.e. the FeO content (or FeO/MgO ratio) of silicate material,
increased with time during core formation.
In addition to an increase in FeO, the Fe2O3 content of the mantle must have
also increased towards the end of, or just after, core formation (O’Neill 1992;
O’Neill et al. 1993a,b; Delano 2001). Equilibrium with Fe-rich metal should have
ensured that the oxygen fugacity of the silicate mantle remained at values below
the iron–wüstite (IW) oxygen buffer during core formation. Taking into account
the current concentration of FeO in the mantle, the fO2 is constrained during the
final stages of core formation to approximately 2 log units below the IW-buffer
(approx. DIWK2) as shown in figure 2. The oxygen fugacity of the present-day
upper mantle, as determined through oxy-thermobarometry, is between K2.5
and C1 log units relative to the fayalite–magnetite–quartz oxygen buffer
(DFMQ) or between DIW C2 and C5.5 (Wood 1991). Mantle samples with
closer links to the asthenosphere fall at the more reduced end of this range, i.e.
DIW C2 to C4.5, but this still requires that the present-day asthenospheric
mantle is at an fO2 at least 3 log units above the level that prevailed during core
formation (Wood et al. 1990; Wood 1991). Recent calibrations of V partitioning
during basalt genesis have allowed the fO2 of the Archaean mantle to be assessed
using analyses from ancient basaltic and ultramafic rocks. Such studies indicate
that the fO2 of the mantle has remained within approximately 0.3 log units of its
present-day value over the last 3.5 Gyr of the Earth’s history (Li & Lee 2004;
Canil et al. 2006). Delano (2001) reached a similar conclusion by studying Cr
partitioning.P
In fact, as will be seen, mantle fO2 is not constant with depth. Even
if the Fe3C/ Fe or bulk oxygen content of the mantle is constant, the fO2 will be
much lower in the deeper mantle as a result of crystal chemical effects (Ballhaus 1995;
Gudmundsson & Wood 1995; Woodland & Koch 2003; Frost & McCammon 2008).
The rise in the redox state of the mantle
since core formation, therefore, is
P
more accurately reflected in the Fe3C/ Fe of the upper mantle (or better still the
O/Fe ratio), which is in the range of 2–3 per cent in the present-day mantle
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
Core formation and redox state of mantle
4319
6
IW
4
FMQ
NNO
Earth
log fO
2
2
0
Mars
IW
Moon
–2
–4
0
MgO
0.2
0.4
0.6
Fe/(Fe + Mg)
0.8
1.0
FeO
Figure 2. The two phases of the redox evolution of the Earth’s upper mantle are indicated by two
arrows, with the fO2 of the mantle calculated relative to the DIW buffer. During core formation, the
fO2 of the mantle was constrained by the presence of metallic Fe to lie close to the thick solid curve
and would vary depending the Fe/(FeCMg) ratio of the mantle. An early stage of reducing core
formation would have left the Earth’s mantle depleted in FeO, but towards the end of core
formation the Earth must have attained its present-day FeO content (indicated by the open circle).
At some point towards the end of, or just after, core formation, the fO2 of the mantle was further
raised above the level of metallic Fe equilibrium to its present-day value as reflected in samples
from the upper mantle (indicated by the open bar). Estimates for the fO2 of the Martian mantle
( Wadhwa 2001; Herd et al. 2002) and for Lunar basalts (Papike et al. 2004) are also shown. IW,
iron–wüstite; FMQ, fayalite–magnetite–quartz; NNO, Ni–NiO redox buffer.
(Canil & O’Neill 1996), but would have been much lower than this when in
equilibrium with metallic Fe. In figure 2, estimates for the fO2 of the Martian
mantle from SNC meteorites (Wadhwa 2001; Herd et al. 2002) and the fO2 of
lunar basalts (Papike et al. 2004) are also shown. The simplest explanation
would seem to be that these bodies did not undergo the same level of mantle
oxidation as experienced by the Earth.
There is, therefore, evidence for two phases of mantle oxidation. The first
apparently raised the FeO content of the mantle during accretion, and the
second raised the Fe2O3 or O/Fe ratio within the first 1 Ga of the Earth’s history.
To examine the possible mechanisms for these oxidation events, it is important
to first study the present-day redox state of the mantle.
3. The redox state of the present-day mantle
The dominant minerals of the upper mantle, olivine and orthopyroxene, have low
Fe2O3 solubilities, which means that in most mantle xenoliths found in alkaline
basalts Fe2O3 is concentrated in (Mg,Fe)Al2O4 spinel (Canil & O’Neill 1996),
a mineral ofPlow modal abundance (2–3% in fertile peridotites). Even though the
bulk Fe3C/ Fe content of the mantle is relatively low (2–3%), oxygen fugacities
calculated from ferric–ferrous equilibria involving spinel are high because fO2
is proportional to the activity of Fe2O3 in spinel (O’Neill et al. 1993b). At higher
pressures, this is no longer the case and ferric Fe becomes dissolved in minerals
such as garnet, which have larger modal abundances (O’Neill et al. 1993b;
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
4320
D. J. Frost et al.
depth (km)
100
150
200
250
300
350
400
5
FMQ
4
2
2
log fO ( IW)
3
1
0
Ni–Fe alloy precipitation
–1
–2
2
4
6
8
10
pressure (GPa)
12
14
Figure 3. Oxygen fugacity relative to the IW buffer determined for garnet peridotite xenoliths from
the Kaapvaal craton as a function of pressure of origin ( Woodland & Koch 2003). The
Psolid curve
shows the fO2 calculated for a typical garnet peridotite assemblage, assuming a Fe3C/ Fe ratio of
2 per cent along a cratonic geotherm using equilibrium (3.1). The Ni–Fe alloy precipitation curve
indicates the fO2 at which Ni–Fe-rich metal will exsolve from silicate minerals (O’Neill &
Wall 1987). At pressures above 8 GPa, a typical mantle assemblage will contain Ni–Fe alloy and
the calculated fO2 follows a less pressure-dependent path close to the IW buffer.
Ballhaus 1995). Figure 3 shows oxygen fugacities calculated for garnet peridotite
xenoliths from kimberlites by Woodland & Koch (2003) using the equilibrium
(Gudmundsson & Wood 1995)
ð3:1Þ
2Fe3 Fe3C
2 Si3 O12 Z 4Fe2 SiO4 C2FeSiO3 CO2 :
garnet
olivine
opx
The decrease in fO2 with depth is mainly due to the positive volume
change of equilibrium (3.1) (8.6 cm3 molK1), which favours the stability of the
Fe3 Fe3C
2 Si3 O12 garnet component (skiagite) with increasing pressure. For a fixed
garnet peridotite composition, pressure alone will drive fO2 to lower levels. Using
a fertile peridotite bulk composition with a fixed Fe2O3 content, and accounting
for interphase partitioning, the fO2 of a mantle assemblage can be calculated at
conditions beyond the depth sampled by xenoliths, shown as the solid curve in
figure 3 (Frost & McCammon 2008). The fO2 is predicted to decrease further due
to the volume change of equation (3.1), but it cannot fall indefinitely because a
level is eventually reached where NiO in silicates will be reduced to form Ni–Fe
alloy (O’Neill & Wall 1987; Ballhaus 1995) via the following reactions:
Ni2 SiO4 Z 2Ni CSiO2 C O2
olivine
ð3:2Þ
metal
and
Fe2 SiO4 Z 2Fe CSiO2 C O2 :
olivine
Phil. Trans. R. Soc. A (2008)
metal
ð3:3Þ
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
Core formation and redox state of mantle
4321
The silica activity is determined from the coexistence of olivine and enstatite.
This Ni precipitation curve occurs at an fO2 slightly below the IW buffer (O’Neill &
Wall 1987). The fO2 of a garnet peridotite assemblage will cross this curve at
approximately 8 GPa or a depth of 250 km, where the metal formed will contain
approximately 60 mol% Ni. Although the formation of metal is instigated by
the reduction of NiO, approximately 40 per cent Fe metal will also partition into the
alloy. Reactions (3.2) and (3.3) are driven to the right-hand side as reaction (3.1) is
driven to the left-hand side with increasing pressure. A simplified combination of
reactions (3.1) and (3.3) is therefore
3FeO Z Fe CFe2 O3 ;
olivine
metal
ð3:4Þ
garnet
which is effectively
FeO disproportionation. The precipitation of Ni–Fe alloy raises
P
the Fe3C/ Fe ratio of the assemblage, which will cause the fO2 to remain close to the
Ni-precipitation curve, although the volume change of equation (3.1) will continue to
lower the fO2 to some extent with pressure, causing more metal to precipitate and
to become increasingly Fe rich. By 14 GPa, a typical upper mantle fertile peridotite
should have exsolved 0.1–0.2 wt% metal and the fO2 will be wIW (Frost &
McCammon 2008). The fO2 in the upper mantle will therefore drop by at least 3
log units between depths of 60 and 250 km, even if the bulk oxygen content
remains constant.
A major testable implication of this calculation is that garnet with a
significant Fe2O3 content should coexist with Ni–Fe alloy at high pressures and
temperatures. As shown in figure 4a, garnets equilibrated in the presence
P of pure
Fe metal at high pressure and temperature indeed have high Fe3C/ Fe ratios
(O’Neill et al. 1993a; McCammon & Ross 2003; Rohrbach et al. 2007). For a
given mineral FeO content, equilibrium with metallic Fe defines the lowest
possible fO2, which means that these Fe2O3 contents are also at their lowest
possible level. Lower pressure minerals such as (Mg,Fe)Al2O4 spinel have
negligible Fe2O3 contents at conditions of IW (Ballhaus et al. 1991), which
indicates a change in mineral behaviour, driven by volumetric effects that favour
Fe3C- over Fe2C-bearing components at high pressure.
The (Mg,Fe,Al)(Al,Si)O3 perovskite dominates the mineralogy of the lower
mantle. As shown in figure 4b, samples of Al-free (Mg,Fe)SiO
P 3 perovskite
synthesized in equilibrium with metallic Fe also have high Fe3C/ Fe ratios that
increase with Fe/(FeCMg). In addition, however, the Fe2O3 content of
perovskite increases strongly with Al2O3 concentration when in equilibrium
with metallic Fe (figure 4c).
P At the Al2O3 content expected for lower
mantle perovskite, the Fe3C/ Fe ratio is over 50 per cent. Samples synthesized
at or close P
to the peridotite solidus (approx. 22008C) have lower Fe2O3 contents
but Fe3C/ Fe ratios are still above 20 per cent. Perovskite has an eightfold
coordinated site mainly occupied by Mg and a sixfold coordinated Si site.
The strong affinity for Fe3C in perovskite probably results from an energetically
favourable coupled substitution of Fe3C onto the eightfold site, which charge
balances Al on the sixfold site. Trace-element partitioning experiments indicate
that the Al2O3 content of perovskite is also coupled to a range of trivalent
trace-element concentrations (Liebske et al. 2005). Elements such as Sc, for
example, which are incompatible in Al-free perovskite during partial melting,
become compatible in Al-bearing perovskite.
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
4322
(a)
D. J. Frost et al.
(b)
0.5
0.30
0.4
Fe3+/(total Fe)
0.35
0.25
0.3
0.20
0.2
0.15
0.10
0.1
0
0.05
0.1
0.2 0.3 0.4
Fe / (Fe + Mg)
0.5
0
0.6
0.05
0.10
Fe/(Fe + Mg)
0.15
Fe3+ (per 3 oxygen formula unit)
(c)
0.12
0.10
0.08
0.06
0.04
0.02
0
0.02 0.04 0.06 0.08 0.10 0.12
Al (per 3 oxygen formula unit)
P
Figure 4. Fe3C/ Fe ratios for (a) garnet and (b) (Mg,Fe)SiO3 perovskite equilibrated with
metallic Fe (or Fe–S liquid in the case of Terasaki et al. 2007) as a function of the Fe/(FeCMg)
ratio. (a) Circles, O’Neill et al. (1993b); diamonds, McCammon & Ross (2003); triangles, Rohrbach
et al. (2007); squares, this study. (b) Circles, Lauterbach et al. (2000); squares, McCammon (1998);
triangles, Terasaki et al. (2007). (c) The Fe3C versus Al3C content in atoms per formula unit of
(Mg,Fe)(Al,Si)O3 perovskite in equilibrium with Fe metal. A likely lower mantle perovskite Al
content is shown by the dashed line. Samples were synthesized between 1400 and 16008C between
24 and 26 GPa. Squares, Frost et al. (2004); circles, Lauterbach et al. (2000).
The equilibrium between Al-bearing perovskite and metallic Fe can be
described by the equilibrium
Al2 O3 C 3FeO Z Fe C2AlFeO3 :
pv
Mgwü
usite
metal
ð3:5Þ
pv
As high concentrations of the FeAlO3 component are measured in perovskite in
the presence of Fe metal at 25 GPa (Frost et al. 2004), this equilibrium must lie
significantly to the right-hand side within a typical lower mantle assemblage.
If Fe2O3-poor material enters the stability field of Al-bearing silicate perovskite,
equilibrium
(3.5) should cause FeO to disproportionate to Fe and Fe3C. If the
3C P
Fe / Fe ratio, or more precisely the bulk oxygen content, of the upper mantle
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
4323
Core formation and redox state of mantle
25.6
8
molar volume (cm3mol–1)
25.4
25.2
5
25.0
4
24.8
Fe3+AlO3
0
Al2O3
3 3
FeSiO3
24.6
24.4
0
MgSiO3
5
10
15
20
25
mol% Al2O3, FeSiO3, FeAlO3
30
Figure 5. Molar volumes of perovskites synthesized along the MgSiO3–FeSiO3, MgSiO3–Al2O3 and
MgSiO3–Fe3CAlO3 joins. Numbers next to symbols indicate the proportion of the FeSiO3
component (mol%) in perovskites nominally along the MgSiO3 –Fe3CAlO3 join. The addition of
FeSiO3 to Al2O3-bearing perovskite has a much larger effect on the perovskite molar volume than
when FeSiO3 is added to Al2O3-free MgSiO3 perovskite. Dashed line, Al2O3, Walter et al. (2004,
2006); dotted line, FeSiO3, Yagi et al. (1979); solid line; Fe3CAlO3; circles, Saikia et al.
(submitted); downtriangles, Vanpeteghem et al. (2006); uptriangles, Nishio-Hamane et al. (2005).
is the same as that of the lower mantle, then the experimental trend in figure 4c
indicates that approximately 1 wt% Fe metal should coexist with FeAlO3bearing perovskite, at least in the top of the lower mantle (Frost et al. 2004). An
indication as to whether disproportionation will occur at higher pressures can be
obtained by considering the likely volume change of equation (3.5), which can
be estimated by assessing the effects of Al2O3, Fe3CAlO3 and FeSiO3
substitutions on perovskite volumes. Figure 5 shows the effects of major substitutions on perovskite molar volumes. The effects on the volume of MgSiO3
perovskite from the substitution of Al2O3 and FeSiO3 individually are quite
similar. Determining the effects of Fe3C
PAlO3 substitution is more challenging as
it requires accurate volume and Fe3C/ Fe ratio measurements, the latter being
possible only through electron energy loss or Mössbauer spectroscopy. The
only complete measurements
to date are those of Saikia et al. (submitted),
P
although Fe3C/ Fe ratios in some other samples can be estimated by assuming
that Al is always charge balanced by Fe3C and that the remaining Fe is present
as Fe2C. The resulting data show that Fe3CAlO3 substitution causes a much
larger increase in volume than either individual Al2O3 or FeSiO3 substitutions,
but the trend is complicated by the fact that most samples also contain a FeSiO3
component, indicated by the number next to each symbol in figure 5. It is
interesting to note, however, that the largest volumes occur for Fe3CAlO3bearing samples with high FeSiO3 concentrations, while the FeSiO3-free sample
of Nishio-Hamane et al. (2005) has the lowest volume for a given Fe3CAlO3
content. FeSiO3 substitution apparently has a much larger effect on the volume
of FeAlO3-bearing perovskites than on MgSiO3 perovskite. As a result of this
strong effect on volume, increasing pressure should favour a low proportion of the
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
4324
D. J. Frost et al.
O
FeO1.5
MgO
FeO
PV
BSE
Fe
MW
Mg
Figure 6. The Fe–O–Mg ternary provides a simple model to describe the redox state of the mantle.
During accretion, equilibrium with metallic Fe at upper mantle pressures would have buffered the
Fe2O3 content of the bulk silicate earth (BSE) at negligible levels, as indicated by the large black
circle. However, as the lower mantle formed, the affinity of (Mg,Fe)(Al,Si)O3 perovskite (PV ) for
Fe2O3 would have forced FeO to disproportionate. As shown in the enlarged inset, loss of
disproportionated Fe metal to the core would have shifted the composition of the BSE off the FeO–
MgO join towards a Fe2O3-bearing composition (open circle). This oxidation mechanism lowers
slightly the total Fe content of the BSE. MW, magnesiowüstite.
FeSiO3 component in perovskite. By linearly extrapolating molar volumes along
the MgSiO3–Al2O3 and MgSiO3–Fe3CAlO3 joins gives a volume change for
equilibrium (3.5) of K2 cm3 molK1, indicating that pressure favours the stability
of coexisting Fe metal and Fe3CAlO3 perovskite. The net result should be an
increase in the Fe3C/Fe2C ratio of perovskite with pressure.
4. Disproportionation during core formation and oxidation of the mantle
Silicate minerals and melts in equilibrium with core-forming Fe-rich metal at
pressures less than 8 GPa contain low levels of the Fe2O3 component (Ballhaus
et al. 1991; Jayasuriya et al. 2004). During the early stages of accretion, the
silicate mantle would, therefore, have been very low in Fe2O3. As pressures inside
the Earth increased, however, and entered the stability field first of the Fe3 Fe3C
2
Si3 O12 garnet component and then the AlFe3CO3 perovskite component, FeO
would have been forced to disproportionate, thus creating Fe–Ni metal. Unless the
mantle was completely molten throughout core formation, which seems unlikely,
then the solid lower mantle, and to a lesser extent the transition zone and lower
part of the upper mantle, would have contained disproportionated metallic Fe–Ni
while the core was still forming. It is important to emphasize that there would,
therefore, have been two types of Fe-rich liquid metal present in the mantle at this
time: (i) ‘core-forming’ metal that accreted and separated to the core and
(ii) ‘disproportionated’ metal that formed as a result of Fe2O3-bearing perovskite
or garnet crystallization. If some of the disproportionated metal also separated to
the core, then the remaining silicate material in the mantle would have been left
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
Core formation and redox state of mantle
4325
with a greater bulk oxygen content, i.e. a higher O/Fe ratio, as can be seen in
figure 6. A mantle composition that initially contains Fe as FeO only would, in the
lower mantle, be in equilibrium with Fe2O3-bearing perovskite and metallic Fe as
a result of disproportionation. Loss of some metallic Fe to the core drives the bulk
composition off the MgO–FeO tie line towards a present-day mantle composition
with a higher O/Fe ratio. Homogenization of the entire mantle by convection,
perhaps aided by one or more giant impacts, would have left the entire mantle
with a higher O/Fe ratio. The degree to which the mantle became oxidized would
have depended on the proportion of disproportionated metallic Fe that separated
from the
P mantle. There is some uncertainty in calculating this proportion as the
Fe3C/ Fe ratio of perovskite in equilibrium with metallic Fe throughout the
entire lower mantle during core formation is not well constrained. However, if of
the order of 5–20 per cent of the disproportionated metal in the entire lower
mantle separated to the core and mantle homogenization followed, then the
mantle could have achieved its present-day O/Fe ratio as reflected in the Fe3C/
P
Fe ratio and fO2 (Frost et al. 2004) of the sampled upper mantle.
There are a number of slightly different ways by which disproportionated
metal could have been lost from the lower mantle. Many core formation models
propose that core-forming metal rained down through a deep magma ocean,
ponding at its base before descending through the solid lower mantle as diapirs
(Li & Agee 1996; Righter et al. 1997; Rubie et al. 2003). The descending diapirs
could have entrained disproportionated metal as they passed through the lower
mantle (figure 7). Another possibility is that disproportionated metal, formed as
perovskite crystallized from a silicate magma ocean, also ponded at the base of
the ocean and separated to the core (Wood et al. 2006). As one preliminary study
indicates that small degree Fe liquids can wet mineral grain boundaries at lower
mantle conditions (Takafuji et al. 2004), a final possibility is that disproportionated metal separated by percolative flow to the core.
As mentioned above, towards the end of accretion, global melting during the
Moon-forming giant impact (Canup & Asphang 2001) may have homogenized
the mantle, mixing back the oxygen-enriched lower mantle with the upper
mantle. Disproportionated metal would not have been lost as the mantle melted
because Fe and Fe2O3 in perovskite would have recombined to produce FeO in
the liquid as melting occurred. Metal would then form again and be trapped in
the lower mantle as the silicate mantle crystallized.
As shown in figure 6, loss of disproportionated metal raises the O/Fe ratio of
the mantle, but it does so at the expense of FeO and thus lowers the FeO/MgO
ratio of the mantle. The approximate FeO/MgO ratio of the mantle must have
been attained, therefore, before the O/Fe ratio was raised. It has been argued
that the depletions of certain slightly siderophile elements from the mantle, such
as V and Cr, required a phase of core formation to take place under very reducing
conditions, compatible with a much lower mantle FeO/MgO ratio than at
present (Wänke 1981; O’Neill 1992; Wade & Wood 2005). Both the FeO/MgO
and O/Fe ratios of the mantle were, therefore, probably raised during or just
after accretion, but these two types of oxidation are quite different and their
mechanisms are unlikely to be related. Disproportionation followed by the
loss of Fe to the core could not have raised the FeO/MgO ratio of the mantle
through the reaction of the resulting Fe2O3 with additional accreting metallic
Fe. This is because producing Fe2O3 by disproportionation consumes the same
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
4326
D. J. Frost et al.
surface
magma
ocean
silicate
solidus
solid
lower
mantle
liquid
core-forming
metal
rains through
magma ocean
metal ponds
at silicate solidus
disproportionated
liquid metal
trapped in the
lower mantle
region of
raised O/ Fe
core
Figure 7. Schematic showing a mechanism to oxidize the mantle during a magma ocean stage of
core formation. As the lower mantle crystallizes, Fe metal and Fe2O3-rich perovskite are produced
by FeO disproportionation. In the overlying magma ocean, core-forming Fe metal rains out and
ponds at the base. The ponds of Fe metal form diapirs that descend through the solid lower mantle
and remove some disproportionated Fe to the core, thereby raising the Fe2O3 content of
the mantle.
amount of FeO as is formed when Fe2O3 reacts with Fe metal. While FeO
disproportionation can explain the increase in the O/Fe ratio of the mantle to
levels above those of equilibrium with Fe metal, it cannot explain increases in the
FeO/MgO ratio of the mantle during accretion.
Loss of disproportionated Fe metal to the core followed by mantle
homogenization could have raised the Fe2O3 content of the upper mantle.
Raised upper mantle Fe2O3 concentrations could have only been sustained,
however, once the proportion of infalling–accreting material equilibrating with
the upper mantle declined, otherwise Fe2O3 would have been reduced by
accreting Fe metal. In the final stages of accretion, however, levels of Fe2O3 may
have been sufficient to oxidize small amounts of incoming metal, causing the
mantle to retain the full compliment of siderophile elements in the accreting
material. This ‘late veneer’ phase of accretion, which did not undergo metal–
silicate fractionation, is required to explain the high levels of highly siderophile
elements in the mantle. The higher fO2 of the upper mantle would have trapped a
potentially reduced late veneer by oxidizing it. The redox state of the late veneer
need not, therefore, have been any different from that of the bulk of the material
from which the Earth formed.
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
Core formation and redox state of mantle
4327
5. Oxygen partitioning and its effect on the FeO/MgO ratio of
the mantle
The partitioning of oxygen during core–mantle equilibration could have also
influenced the redox state of the mantle in addition to potentially contributing at
least part of the light element budget of the core (Ringwood 1977; O’Neill et al.
1998; Rubie et al. 2004). The solubility of oxygen in liquid Fe metal is relatively
low at room pressure. At the 1 bar Fe–FeO eutectic, Fe metal contains only
0.16 wt% oxygen (Ringwood 1977). However, experiments have shown that
solubility increases with both pressure and temperature (Kato & Ringwood 1989;
Takahashi et al. 2004). The oxygen concentration of liquid metal during core
formation will depend on the partitioning of oxygen between silicate melt and Fe
metal, rather than its solubility in the Fe–FeO system. Although such
partitioning cannot result in more oxygen in Fe metal than Fe–FeO solubility
experiments would indicate, the sign of the pressure and temperature
dependences of oxygen partitioning is not necessarily the same as those measured
in Fe–FeO solubility measurements.
To date, the partitioning of oxygen has been studied mainly between liquid Fe
alloy and magnesiowüstite (Ohtani & Ringwood 1984; O’Neill et al. 1998; Rubie
et al. 2004; Asahara et al. 2007), although some data also exist for partitioning
between Fe metal and silicate perovskite (Takafuji et al. 2005; Kawazoe &
Ohtani 2006). Experiments on magnesiowüstite or perovskite are technically
much easier to perform than those on silicate melts, which are hard to encapsulate at very high temperatures and generally react with any capsule material
that does not react with liquid metal. Results can be applied to silicate melts
using data on the partitioning of FeO between such melts and magnesiowüstite
or perovskite. Oxygen partitioning between Fe liquid and magnesiowüstite is a
function of fO2, but this can be accounted for by the reaction
FeO
magnesiowü
ustite
Z Fe C O
metal
ð5:1Þ
and defining a distribution coefficient Kd
Kd Z
met
Xmet
Fe XO
;
Xmw
FeO
ð5:2Þ
met
met
where Xmw
FeO , XFe and XO are the mole fractions of FeO in magnesiowüstite, Fe in
metal and O in metal, respectively. Experiments have shown that Kd is ostensibly
independent of oxygen fugacity over the range of conditions and metal oxygen
contents so far studied (Asahara et al. 2007) and Kd can therefore be modelled as a
function of pressure and temperature alone. As a result of this simplifying
assumption, inaccuracies may arise particularly when resulting models are
extrapolated to high temperatures (higher than 3000 K), where metal oxygen
contents become very large, and more realistic models are required, which treat
oxygen partitioning as a function of oxygen fugacity.
Figure 8 shows ln Kd determined in experiments performed up to 25 GPa and
3150 K (Asahara et al. 2007). Oxygen partitioning is a strong function of
temperature but a much weaker function of pressure. Oxygen contents of liquid
Fe of approximately 7 wt% were observed in equilibrium with (Mg0.9Fe0.1)O at
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
4328
D. J. Frost et al.
2373 K
2473 K
2573–2773 K
2873 K
3073 K
3173 K
1
3400 K
ln Kd
0
3073 K
–1
2673 K
2473 K
–2
–3
0
5
10
15
pressure (GPa)
20
25
Figure 8. The oxygen exchange coefficient Kd (equation (5.2)) as a function of pressure determined
from experiments on magnesiowüstite and liquid Fe metal performed at various temperatures
(Asahara et al. 2007). Curves are calculated at the indicated temperatures from a global fit to the
experimental data.
the highest temperatures examined. The pressure dependence of the data is much
harder to determine and an adequate fit would be obtained with no pressure
dependence over the experimental pressure range examined. Asahara et al.
(2007) used a thermodynamic model based on Fe–FeO solubility measurements
to estimate Kd at room pressure, and estimated Kd at pressures up to 97 GPa
using published data from diamond anvil cell experiments on the oxygen
contents of liquid metal in equilibrium with (Mg,Fe)SiO3 perovskite. These highpressure data indicate that oxygen contents do increase at very high pressure
and Asahara et al. (2007) consequently proposed a model in which metal–
magnesiowüstite oxygen partitioning initially decreases with pressure, but
reaches a minimum in the range of 20–30 GPa, above which oxygen contents
in the metal start to increase. This can be explained if the partial molar volume
of oxygen in Fe metal is relatively high at 1 bar but has a high compressibility.
The volume change of reaction (5.1) is therefore initially negative but the high
compressibility of oxygen eventually results in a sign change in the volume and
oxygen starts to preferentially dissolve in Fe metal. This is in surprisingly good
agreement with the predictions made by Walker et al. (2002) on the FeO content
of Fe metal based on the estimates for the molar volume of O2. The model of
Asahara et al. (2007), however, indicates that the effect of pressure on Kd over
the entire range of pressure encompassed by the present-day mantle is not
greater than that which would be caused by a change in temperature of 200 K.
This model is in reasonable agreement with the recent experiments on
partitioning between Fe liquid and magnesiowüstite performed in the diamond
anvil cell up to 134 GPa (Ozawa et al. 2008). The model does, however, assume
that Kd remains independent of the composition over wide ranges of temperature
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
Core formation and redox state of mantle
4329
and pressure and therefore ignores activities of the components. While Asahara
et al. (2007) observed Kd to be indeed independent of the composition up to
25 GPa, this is somewhat unusual given the large liquid immiscibility in the Fe–
FeO system and this may not be the case at extreme pressures and temperatures.
More complex models could be based on the determinations of oxygen activities
in Fe metal assessed from the miscibility gap in the Fe–FeO system.
Rubie et al. (2004) proposed that the mantles of both the Earth and Mars,
which have low and high FeO contents, respectively, could have been produced
from material with an identical bulk composition (particularly in terms of
bulk O/Fe ratio), as long as core–mantle equilibration occurred at conditions
that promoted more oxygen to partition into the core of the Earth in comparison
with that of Mars. Higher oxygen partitioning into the Earth’s core would
have depleted the mantle of FeO, whereas if relatively little oxygen partitioned
into the Martian core its mantle would have retained a higher FeO content. The
results of Asahara et al. (2007) are consistent with this model and indicate that
higher temperature core–mantle equilibration on the Earth than on Mars could
have caused the required difference in oxygen partitioning. This implies that
accretion was more energetic on Earth than on Mars, which is quite likely, given
the differences in final masses and released gravitational energy during accretion.
6. Silicon and oxygen partitioning and their effects on the FeO/MgO
ratio of the mantle
The partitioning of Si into the core could have also raised the FeO content
of the mantle, and hence the fO2, during core–mantle equilibration (Javoy 1995).
The partitioning of Si and Fe between liquid Fe metal and silicate melt is
written as
SiO2 C 2Fe Z Si C2FeO;
ð6:1Þ
silicate
metal
metal
silicate
where the fO2-independent exchange coefficient is defined as
Kd Z
2
½Xmetal
½Xsilicate
Si
FeO :
metal 2
½Xsilicate
SiO2 ½XFe ð6:2Þ
Figure 9 shows determinations of Kd from analyses of silicate melt and liquid
Fe metal equilibrated using a multi-anvil apparatus between 2 and 25 GPa at
1750–24008C (Wade & Wood 2005; U. Mann, D. J. Frost & D. C. Rubie 2008,
unpublished data). In addition, Kd has also been calculated from the diamond
anvil cell experiments of Takafuji et al. (2005) in which the Si content of Fe metal
in equilibrium with (Mg,Fe)SiO3 perovskite was investigated between 24 and
97 GPa. Although these experiments involved perovskite and not silicate melt,
the differences in the derived Kd values are likely to be small. In figure 10, all
experimental data have been corrected to 2200 K and are plotted as a function of
pressure. These data indicate that Kd decreases with pressure up to
approximately 25 GPa but is apparently almost pressure independent at higher
pressures. A model has been fit to the experimental data with the temperature
dependence guided by thermodynamic data on the pure oxides, as described by
Wade & Wood (2005). These data indicate that, if temperatures are sufficiently
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
4330
D. J. Frost et al.
T (°C)
2800
2400
2000
1600
–2.0
Si
–3.0
log Kd
–4.0
–5.0
2G
6 GPa
Pa
18
25 GPa
GP
a
–6.0
–7.0
–8.0
3.0
3.5
4.0
4.5
10 000/T (K)
5.0
5.5
Figure 9. The Si–Fe exchange coefficient Kd (equation (6.2)) between liquid Fe metal and silicate
melt determined as a function of pressure and temperature in a series of experiments (U. Mann,
D. J. Frost & D. C. Rubie 2008, unpublished data; black symbols (triangles, 2 GPa; diamonds,
6 GPa; squares, 18–20 GPa; circles, 24–25 GPa)). Results of similar experiments by Wade & Wood
(2005) are also shown (open (24–25 GPa) and grey (55 GPa) circles) in addition to data from
Takafuji et al. (2005) on the Si–Fe Kd between liquid Fe metal and (Mg,Fe)SiO3 perovskite
(crosses (93–97 GPa)). The study data of this literature are shown by black symbols. (Triangles,
2 GPa; diamonds, 6 GPa; squares, 18–20 GPa; circles, 24–25 GPa.)
–4.0
log Kd (Si–Fe)
–4.5
–5.0
–5.5
–6.0
–6.5
0
20
40
60
pressure (GPa)
80
100
Figure 10. The data shown in figure 9 normalized to 2200 K and plotted as a function of pressure
(triangles, this study). Multi-anvil data are from Wade & Wood (2005) and U. Mann, D. J. Frost &
D. C. Rubie 2008, unpublished data. Diamonds, Takafuji et al. (2005).
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
4331
Core formation and redox state of mantle
high in a magma ocean, then Si will enter the core and thus raise the FeO content
of the mantle (reaction (6.1)). The effectiveness of this process, however, must be
considered in conjunction with the partitioning of oxygen into Fe metal at high
temperatures. Oxygen released by Si entering the core will therefore partition
between the mantle and core (Javoy 1995) and this partitioning will be strongly
temperature dependent.
The partitioning of Si into the core with a consequent increase in the FeO
content of the mantle, as proposed by Javoy (1995), is a process that can
potentially explain both the initially reducing conditions of core formation and
the subsequent high (8 wt%) mantle FeO concentration. The effectiveness of this
mechanism can be tested by combining mass balance calculations with high P–T
partitioning models for Si and O. The core formation–accretion model of Rubie
et al. (2007) adopts this approach and preliminary results are presented here in
order to demonstrate its potential. The model involves continuous core formation
during accretion whereby the Earth accretes through a series of collisions with
planetary bodies, each of which has a mass equal to 10 per cent of the Earth’s
mass at the time of impact. Following each impact, metal and silicate equilibrate
in a resulting magma ocean at some specified pressure, which is a fraction of
the core–mantle boundary pressure, and therefore increases during accretion.
The temperature of equilibration is given by the peridotite solidus at the
particular pressure (Zerr et al. 1998). Following the approach of Rubie et al.
(2004), metal–silicate equilibration is modelled for a chondritic bulk composition
(McDonough & Sun 1995) in which the oxygen content can be varied. The bulk
composition can be varied anywhere between two extreme cases in which (i) all
Fe is present as FeO (oxygen-rich composition) and (ii) all Fe is present as metal
(oxygen-poor composition). The equilibrium compositions and relative proportions of coexisting silicate and metal liquids are determined at high P–T
½ðFeOÞx ðNiOÞy ðSiO2 ÞðMgu Alm Can ÞO C½Fea Nib Oc Sid :
silicate
liquid
metal
liquid
In order to solve for the seven unknown coefficients (x, y, z, a, b, c and d ), four
mass balance equations (for Fe, Ni, Si and O, respectively) and three Kd
expressions that describe the partitioning of Si, Ni and FeO are used.
Partitioning data described above are used together with the results of Kegler
et al. (2008) for Ni. After equilibrating with the mantle at the base of the magma
ocean, the metallic component of the impactor sinks to merge with the Earth’s
proto-core without further equilibration.
Figure 11 shows preliminary results of calculations to determine the effect of O
and Si partitioning on the evolution of the FeO content of a magma ocean during
accretion. The lower curve [1] shows a model in which the Earth accretes from
reduced (oxygen-poor) material. Initially, the FeO content of the magma ocean
reflects that of the proto-Earth and impactors and the fO2 is sufficiently reducing
for slightly siderophile elements, such as V, Cr and Si, to enter the core. As the
magma ocean deepens and the pressure and (more importantly) the liquidus
temperature increase, Si partitions increasingly into the core and the FeO
content of the mantle increases. However, even with a very deep magma ocean,
the FeO content of the mantle does not approach that of the present-day Earth.
Curve [2] shows a model in which the bulk composition of accreting material was
slightly more oxidized in an attempt to approach an Earth-like final mantle FeO
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
4332
D. J. Frost et al.
8
7
FeO wt%
6
[3]
5
[2]
4
3
[1]
2
1
0
20
40
60
pressure (GPa)
80
100
Figure 11. Results of modelling calculations on the evolution of the FeO content of the mantle
during accretion. The Earth grows by the accretion of a series of impactors with a constant
composition and metal–silicate ratio. During each impact, the metal core of the impactor
equilibrates with the entire silicate mantle of the Earth at the base of a magma ocean. The FeO
content of the mantle is plotted against the pressure of equilibration, which is here taken to be 68
per cent of the pressure at the core–mantle boundary, while the temperature is determined from
the silicate solidus. The initial FeO content of the mantle reflects the FeO content of the impactors
with the lowest curve [1] having less FeO in the impactor mantle (i.e. more reduced) than [2] or [3].
Curve [3] is calculated for a peridotite solidus that is 2008 higher than that experimentally
determined. The grey shaded bar shows the FeO content of the present-day mantle.
content. As the magma ocean deepens and Si enters the core, the FeO content of
the mantle increases, however, at approximately 70 GPa the proportion of O
entering the core increases to the point where the FeO content of the mantle
starts to decrease. The back bend of the curve is controlled by not only the
temperature but also the FeO content of the mantle, because more O enters
the core at higher fO2. Curve [3] shows results for which the composition of the
impactor is the same as for model [2] but the temperature of the peridotite
solidus is raised by 2008. Uncertainties in the peridotite melting temperatures
make this quite plausible. Here, the FeO content of the mantle also initially
increases but backbends at even lower pressures as significant O starts to enter
the core. Thus, although Si entering the core increases the FeO content of the
mantle, the effect is limited at high temperatures by the partitioning of O into
the core, which lowers the FeO content of the mantle. The largest increase in
FeO content of the mantle was obtained with the first model in which fO2
increases from DIW K4.5 to K2.5 during accretion. However, although this
model can adequately describe the depletion of slightly siderophile elements, it
does not result in an increase in the FeO content of the mantle to the level found
in the present-day Earth. The models do predict, however, that, for significant
internal oxidation of the mantle to have occurred, metal–silicate equilibration in
deep magma oceans that extended significantly into the Earth’s lower mantle
would have been necessary (i.e. pressures far higher than 25 GPa).
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
Core formation and redox state of mantle
4333
There are, however, large uncertainties in the current model particularly
at pressures above 25 GPa, where the diamond anvil cell experimental data
have large temperature and compositional uncertainties. A weak element of the
current model is that data for Si are taken predominantly from experiments on
silicate melts, whereas the O model relies mainly on the data from experiments
on magnesiowüstite and perovskite. The conclusions would be more reliable if
data on O and Si were obtained in the same experimental study on silicate melts.
Whether the mantle can be internally oxidized depends critically on the relative
difference in partitioning between O and Si, which is currently poorly determined
at the very high pressures where much of the oxidation is predicted to take place.
Improved O and Si partitioning data between liquid Fe metal and silicate melts
are required to conclusively exclude the possibility of complete internal oxidation
of the mantle from reduced impactors, via reaction (6.1).
7. Discussion and conclusions
The depletion of slightly siderophile elements (V, Cr) from the mantle required
more reducing conditions than those at which depletions of moderately siderophile
elements (Fe, Ni and Co) appear to have been established (Wade & Wood 2005).
A good explanation for this involves the oxygen fugacity at which core–mantle
equilibration occurred increasing during accretion from initially very reducing
conditions, where slightly siderophile elements (V,Cr) would have entered the core,
to more oxidizing conditions where the mantle’s Fe, Ni and Co abundances were
attained (Wänke 1981; O’Neill 1992; Wade & Wood 2005). Such a change could
have been caused by an increase in the redox state of the material that accreted the
Earth with time (Rubie et al. 2007). The variation in FeO contents of chondritic
meteorites indicates that condensation processes and reactions involving H2O could
have caused very large variations in the redox state of accreting material (Krot et al.
2000). However, if accretion occurred from reduced material alone and deep magma
oceans developed, the partitioning of Si into the core would have led to an increase
in the FeO content of the mantle. The question then arises as to whether it is
possible for the Earth to have accreted solely from reduced impactors in a
homogeneous accretion scenario, with internal oxidation of the mantle occurring to
the extent that the present-day upper mantle FeO content was ultimately achieved.
Current models based on existing experimental O and Si partitioning data
indicate that the FeO content of the mantle could have increased from less than
1 wt% to approximately 5 wt% as Si entered the core during metal–silicate
equilibration at the base of a progressively deepening magma ocean, which
corresponds to an increase in fO2 of approximately 2 log units. As a result of O
entering the core at high temperatures, the models fail to predict conditions in
which the present-day mantle FeO content is reached. However, given the
current large uncertainties in the model parameters, particularly at very high
pressures and temperatures, we cannot at present exclude the possibility that
complete internal oxidation of the mantle is a viable mechanism. The models
indicate that very high temperatures are required for this oxidation mechanism to
proceed, corresponding to conditions in a magma ocean that extended deep into
the present-day lower mantle. Further data on Si and O partitioning at deep
lower mantle conditions are required in order to further test this hypothesis.
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
4334
D. J. Frost et al.
Once a solid lower mantle formed, disproportionation of FeO to produce
Fe2O3-bearing perovskite and Fe metal would have occurred. Loss of the
disproportionated Fe to the core by entrainment in the final stages of core
formation may have then raised the Fe2O3 content of the lower mantle. After
mantle homogenization, either by convection or by a Moon-forming impact, the
Fe2O3 content of the whole mantle could have attained its present-day level. This
mechanism raises the Fe2O3 content of the mantle (i.e. the Fe/O ratio) but
lowers the total FeO content. It therefore cannot explain increases in the FeO
content of the mantle during core formation from the very low levels compatible
with the loss of slightly siderophile elements (i.e. V and Cr) from the mantle
because Fe2O3 formed by disproportionation would have reacted with accreting
Fe metal to produce exactly the same amount of FeO in the mantle, which had
initially disproportionated. Whole mantle oxidation by disproportionation could
have operated only towards the end of core formation where levels of accreting
Fe metal in the upper mantle became low and insufficient to reduce Fe2O3
back to FeO.
The authors are grateful to C. McCammon and T. Boffa Ballaran for their constructive discussions.
U.M. acknowledges the support of the Elitenetzwerk Bayern Graduate programme. The
manuscript was improved through the comments of two anonymous reviewers.
References
Asahara, Y., Frost, D. J. & Rubie, D. C. 2007 Partitioning of FeO between magnesiowüstite and
liquid iron at high pressures and temperatures: implications for the composition of the Earth’s
outer core. Earth Planet. Sci. Lett. 257, 435 – 449. (doi:10.1016/j.epsl.2007.03.006)
Ballhaus, C. 1995 Is the upper mantle metal-saturated? Earth Planet. Sci. Lett. 132, 75–86. (doi:10.
1007/BF00311183)
Ballhaus, C., Berry, R. F. & Green, D. H. 1991 High pressure experimental calibration of the
olivine–orthopyroxene–spinel oxygen geobarometer: implications for the oxidation state of
the upper mantle. Contrib. Miner. Petrol. 107, 27–40. (doi:10.1016/0012-821X(95)00047-G)
Canil, D. & O’Neill, H. St. C. 1996 Distribution of ferric iron in some upper-mantle assemblages.
J. Petrol. 37, 609–635. (doi:10.1093/petrology/37.3.609)
Canil, D., Johnston, S. T. & Mihalynuk, M. 2006 Mantle redox in Cordilleran ophiolites as a record
of oxygen fugacity during partial melting and the lifetime of mantle lithosphere. Earth Planet.
Sci. Lett. 248, 106–117. (doi:10.1016/j.epsl.2006.04.038)
Canup, R. M. & Asphang, E. 2001 Origin of the Moon in a giant impact near the end of the Earth’s
formation. Nature 412, 708–712. (doi:10.1038/35089010)
Chabot, N. L. & Agee, C. B. 2003 Core formation in the Earth and Moon: new experimental
constraints from V, Cr, and Mn. Geochim. Cosmochim. Acta 67, 2077–2091. (doi:10.1016/
S0016-7037(02)01272-3)
Delano, J. W. 2001 Redox history of the Earth’s interior since 3900 Ma: implications for prebiotic
molecules. Orig. Life Evol. Biosph. 31, 311–341. (doi:10.1023/A:1011895600380)
Frost, D. J. & McCammon, C. A. 2008 The redox state of earth’s mantle. Annu. Rev. Earth Planet.
Sci. 36, 389– 420. (doi:10.1146/annurev.earth.36.031207.124322)
Frost, D. J., Liebske, C., Langenhorst, F., McCammon, C. A., Tronnes, R. G. & Rubie, D. C. 2004
Experimental evidence for the existence of iron-rich metal in the Earth’s lower mantle. Nature
248, 409 – 412. (doi:10.1038/nature02413)
Galimov, E. M. 2005 Redox evolution of the Earth caused by a multi-stage formation of its core.
Earth Planet. Sci. Lett. 233, 263–276. (doi:10.1016/j.epsl.2005.01.026)
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
Core formation and redox state of mantle
4335
Gessmann, C. K. & Rubie, D. C. 2000 The origin of the depletions of V, Cr, and Mn in the mantles
of the Earth and Moon. Earth Planet. Sci. Lett. 184, 95–107. (doi:10.1016/S0012821X(00)00323-X)
Gudmundsson, G. & Wood, B. J. 1995 Experimental tests of garnet peridotite oxygen barometry.
Contrib. Mineral. Petrol. 119, 56–67. (doi:10.1007/BF00310717)
Herd, C. D., Borg, L. E., Jones, J. H. & Papike, J. J. 2002 Oxygen fugacity and geochemical
variations in martian basalts: implications for martian basalt petrogenesis and the oxidation
state of the upper mantle of Mars. Geochim. Cosmochim. Acta 66, 2025–2036. (doi:10.1016/
S0016-7037(02)00828-1)
Holzheid, A., Sylvester, P., O’Neill, H. St. C., Rubie, D. C. & Palme, H. 2000 Evidence for a late
chondritic veneer in the Earth’s mantle from high-pressure partitioning of palladium and
platinum. Nature 406, 396 –399. (doi:10.1038/35019050)
Javoy, M. 1995 The integral enstatite chondrite model of the Earth. Geophys. Res. Lett. 22,
2219 –2222. (doi:10.1029/95GL02015)
Jayasuriya, K. D., O’Neill, H. S., Berry, A. J. & Campbell, S. J. 2004 A Mossbauer study of the
oxidation state of Fe in silicate melts. Am. Mineral. 89, 1597–1609.
Kasting, J. F., Eggler, D. H. & Raeburn, S. P. 1992 Mantle redox evolution and the oxidation state
of the archean atmosphere. Geology 101, 245–257.
Kato, T. & Ringwood, A. E. 1989 Melting relationships in the system Fe–FeO at high pressures:
implications for the composition and formation of the earth’s core. Phys. Chem. Minerals 16,
524–538. (doi:10.1007/BF00202207)
Kawazoe, T. & Ohtani, E. 2006 Reaction between liquid iron and (Mg,Fe)SiO3-perovskite and
solubilities of Si and O in molten iron at 27 GPa. Phys. Chem. Miner. 33, 227–234. (doi:10.1007/
s00269-006-0071-4)
Kegler, P., Holzheid, A., Frost, D. J., Rubie, D. C., Dohmen, R. & Palme, H. 2008 New Ni and Co
metal–silicate partitioning data and their relevance for an early terrestrial magma ocean. Earth
Planet. Sci. Lett. 268, 28–40. (doi:10.1016/j.epsl.2007.12.020)
Krot, A. N., Fegley, B., Palme, H. & Lodders, K. 2000 Meteoritical and astrophysical constraints
on the oxidation state of the solar nebula. In Protostars and planets IV (eds V. Manning, A. P.
Boss & S. S. Russell), pp. 1019–1054. Tucson, AZ: University of Arizona Press.
Lauterbach, S., McCammon, C. A., van Aken, P., Langenhorst, F. & Seifert, F. 2000 Mössbauer
and ELNES spectroscopy of (Mg,Fe)(Si,Al)O3 perovskite: a highly oxidised component of the
lower mantle. Contrib. Mineral. Petrol. 138, 17–26. (doi:10.1007/PL00007658)
Li, J. & Agee, C. B. 1996 Geochemistry of mantle–core differentiation at high pressure. Nature 381,
686–689. (doi:10.1038/381686a0)
Li, Z.-X. A. & Lee, C.-T. A. 2004 The constancy of upper mantle fO2 through time inferred from
V/Sc ratios in basalts. Earth Planet. Sci. Lett. 228, 483–493. (doi:10.1016/j.epsl.2004.10.006)
Liebske, C., Corgne, A., Frost, D. J., Rubie, D. C. & Wood, B. J. 2005 Compositional effects on
element partitioning between Mg-silicate perovskite and silicate melts. Contrib. Miner. Petrol.
149, 113–128. (doi:10.1007/s00410-004-0641-8)
McCammon, C. A. 1998 The crystal chemistry of ferric iron in Mg0.05Fe0.95SiO3 perovskite as
determined by Mössbauer spectroscopy in the temperature range 80–293 K. Phys. Chem. Miner.
25, 292–300. (doi:10.1007/s002690050117)
McCammon, C. A. & Ross, N. L. 2003 Crystal chemistry of ferric iron in (Mg,Fe)(Si,Al)O3
majorite with implications for the transition zone. Phys. Chem. Miner. 30, 206–216. (doi:10.
1007/s00269-003-0309-3)
McDonough, W. F. & Sun, S.-s. 1995 The composition of the Earth. Chem. Geol. 120, 223–253.
(doi:10.1016/0009-2541(94)00140-4)
Nishio-Hammane, D., Nagai, T., Fujino, K., Seto, Y. & Takafuji, N. 2005 Fe3C and Al solubilities
in MgSiO3 perovskite: implication of the Fe3CAlO3 substitution in MgSiO3 perovskite at the
lower mantle condition. Geophys. Res. Lett. 32, L16306. (doi:10.1029/2005GL023529)
Ohtani, E. & Ringwood, A. E. 1984 Composition of core. I. Solubility of oxygen in molten iron at
high temperatures. Earth Planet. Sci. Lett. 71, 85–93. (doi:10.1016/0012-821X(84)90054-2)
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
4336
D. J. Frost et al.
O’Neill, H. St. C. 1992 The origin of the Moon and the early history of the Earth: a chemical
model. Part 2. The Earth. Geochim. Cosmochim. Acta 55, 1159–1172. (doi:10.1016/00167037(91)90169-6)
O’Neill, H. St. C. & Palme, H. 1998 Composition of the silicate earth: implications for accretion
and core formation. In The Earth’s mantle:composition, structure and evolution (ed. I. Jackson),
pp. 3–126. Cambridge, UK: Cambridge University Press.
O’Neill, H. St. C. & Wall, V. J. 1987 The olivine–orthopyroxne–spinel oxygen geobarometer, the nickel
precipitation curve, and the oxygen fugacity of the Earth’s upper mantle. J. Petrol. 28, 1169–1191.
O’Neill, H. St. C., McCammon, C. A., Canil, D., Rubie, D. C., Ross II, C. R. & Seifert, F. 1993a
Mössbauer spectroscopy of mantle transition zone phases and determination of minimum Fe3C
content. Am. Miner. 78, 456–460.
O’Neill, H. St. C., Rubie, D. C., Canil, D., Geiger, C. A., Ross II, C. R., Seifert, F. & Woodland, A. B.
1993b Ferric iron in the upper mantle and in transition zone assemblages: implications for relative
oxygen fugacities in the mantle. In Evolution of the Earth and planets (eds E. Takahashi,
R. Jeanloz & D. C. Rubie). Geophysics Monograph, 74, pp. 73–78. Washington, DC: International
Union of Geodesy and Geophysics and American Geophysics Union.
O’Neill, H. St. C., Canil, D. & Rubie, D. C. 1998 Oxide-metal equilibria to 25008C and 25 GPa:
implications for core formation and the light component in the Earth’s core. J. Geophys. Res.
103, 12 239–12 260. (doi:10.1029/97JB02601)
Ozawa, H., Hirose, K., Mitome, M., Bando, Y., Sata, N. & Ohishi, Y. 2008 Chemical equilibrium
between ferropericlase and moltern iron to 134 GPa and implications for iron content at the
bottom of the mantle. Geophys. Res. Lett. 35, L05308. (doi:10.1029/2007GL032648)
Papike, J. J., Karner, J. M. & Shearer, C. K. 2004 Comparative planetary mineralogy: V/(CrCAl)
systematics in chromite as an indicator of relative oxygen fugacity. Am. Miner. 89, 1557–1560.
Righter, K., Drake, M. J. & Yaxley, G. 1997 Prediction of siderophile element metal/silicate
partition coefficients to 20 GPa and 28008C: the effects of pressure, temperature, oxygen
fugacity, and silicate and metallic melt compositions. Phys. Earth Planet. Inter. 100, 115–134.
(doi:10.1016/S0031-9201(96)03235-9)
Ringwood, A. E. 1966 The chemical composition and origin of the Earth. In Advances in earth
science (ed. P. Hurley), pp. 357–398. Boston, MA: MIT Press.
Ringwood, A. E. 1977 Composition of the core and implications for origin of the Earth. Geochem.
J. 11, 111–135.
Rohrbach, A., Ballhaus, C., Golla-Schindler, U., Ulmer, P., Kamenetsky, V. S. & Kuzmin, D. V.
2007 Metal saturation in the upper mantle. Nature 449, 456–458. (doi:10.1038/nature06183)
Rubie, D. C., Melosh, H. J., Reid, J. E., Liebske, C. & Righter, K. 2003 Mechanisms of metal–
silicate equilibration in the terrestrial magma ocean. Earth Planet. Sci. Lett. 205, 239–255.
(doi:10.1016/S0012-821X(02)01044-0)
Rubie, D. C., Gessmann, C. K. & Frost, D. J. 2004 Partitioning of oxygen during core formation on
the Earth and Mars. Nature 429, 58–62. (doi:10.1016/S0012-821X(02)01044-0)
Rubie, D. C., Mann, U., Frost, D. J., Kegler, P., Holzheid, A. & Palme, H. 2007 Heterogeneous
earth accretion and incomplete metal–silicate reequilibration at high pressure during core
formation. EoS Trans. AGU 86, Fall Meet. Suppl., Abstract U11A-0015.
Saikia, A., Boffa Ballaran, T. & Frost, D. J. Submitted. The effect of Fe and Al substitution on the
compressibility of MgSiO3 perovskite determined through single-crystal X-ray diffraction.
Takafuji, N., Hirose, K., Ono, S., Xu, F., Mitome, M. & Bando, Y. 2004 Segregation of core melts
by permeable flow in the lower mantle. Earth Planet. Sci. Lett. 224, 249–257. (doi:10.1016/
j.epsl.2004.05.016)
Takafuji, N., Hirose, K., Mitome, M. & Bando, Y. 2005 Solubilities of O and Si in liquid iron in
equilibrium with (Mg,Fe)SiO3 perovskite and the light elements in the core. Geophys. Res. Lett.
32, 6313. (doi:10.1029/2005GL022773)
Takahashi, E., Nakayama, K. & Okamoto, Y. 2004 Phase relation in the system Fe–FeO up to
25 Gpa. Spec. Issue Rev. High Press. Sci. Technol. 14, 27.
Phil. Trans. R. Soc. A (2008)
Downloaded from http://rsta.royalsocietypublishing.org/ on June 16, 2017
Core formation and redox state of mantle
4337
Terasaki, H., Frost, D. J., Rubie, D. C. & Langenhorst, F. 2007 Interconnectivity of Fe–O–S liquid
in polycrystalline silicate perovskite at lower mantle conditions. Phys. Earth Planet. Int. 161,
170–176. (doi:10.1016/j.pepi.2007.01.011)
Vanpeteghem, C. B., Angel, R. J., Ross, N. L., Jacobsen, S. D., Dobson, D. P., Litasov, K. D. &
Ohtani, E. 2006 Al, Fe substitution in the MgSiO3 perovskite structure: a single crystal X-ray
diffraction study. Phys. Earth Planet. Inter. 155, 96–103. (doi:10.1016/j.pepi.2005.10.003)
Wade, J. & Wood, B. J. 2005 Core formation and the oxidation state of the Earth. Earth Planet.
Sci. Lett. 236, 78–95. (doi:10.1016/j.epsl.2005.05.017)
Wadhwa, M. 2001 Redox state of Mars’ upper mantle and crust from Eu anomalies in shergottite
pyroxenes. Science 291, 1527–1530. (doi:10.1126/science.1057594)
Walker, D., Clark, S. M., Cranswick, L. M. D., Johnson, M. C. & Jones, R. L. 2002 O2 volumes at
high pressure from KClO4 decomposition: D 00 as a siderophile element pump instead of a lid on
the core. Geochem. Geophys. Geosyst. 3, 1070. (doi:10.1029/2001GC000225)
Walter, M. J., Kubo, A., Yoshino, T., Brodholt, J., Koga, K. T. & Ohishi, Y. 2004 Phase relations
and equation-of-state of aluminous Mg-silicate perovskite and implications for Earth’s lower
mantle. Earth Planet. Sci. Lett. 222, 501–516. (doi:10.1016/j.epsl.2004.03.014)
Walter, M. J., Tronnes, R. G., Armstrong, L. S., Lord, O. T., Caldwell, W. A. & Clark, S. M. 2006
Subsolidus phase relations and perovskite compressibility in the system MgO–AlO1.5–SiO2 with
implications for Earth’s lower mantle. Earth Planet. Sci. Lett. 248, 77–89. (doi:10.1016/j.epsl.
2006.05.017)
Wänke, H. 1981 Constitution of terrestrial planets. Phil. Trans. R. Soc. A 303, 287–302. (doi:10.
1098/rsta.1981.0203)
Wood, B. J. 1991 Oxygen barometry of spinel peridotites. In Oxide minerals: petrologic and
magnetic significance (ed. D. H. Lindsley). Reviews in mineralogy, pp. 417–431. Washington,
DC: Mineral Society of America.
Wood, B. J., Bryndzia, L. T. & Johnson, K. E. 1990 Mantle oxidation state and its relationship to
tectonic environment and fluid speciation. Science 248, 337–345. (doi:10.1126/science.248.4953.
337)
Wood, B. J., Walter, M. J. & Wade, J. 2006 Accretion of the Earth and segregation of its core.
Nature 441, 825–833. (doi:10.1038/nature04763)
Woodland, A. B. & Koch, M. 2003 Variation in oxygen fugacity with depth in the upper mantle
beneath Kaapvaal craton, South Africa. Earth Planet. Sci. Lett. 214, 295–310. (doi:10.1016/
S0012-821X(03)00379-0)
Yagi, T., Mao, H. K. & Bell, P. M. 1979 Lattice parameters and specific volume for the perovskite
phase of ortho-pyroxene composition, (Mg,Fe)SiO3. Year Book Carnegie Inst. Wash. 78,
612–625.
Zerr, A., Diegeler, A. & Boehler, R. 1998 Solidus of Earth’s deep mantle. Science 281, 243–246.
(doi:10.1126/science.281.5374.243)
Phil. Trans. R. Soc. A (2008)