Download EARTH AND PLANETARY SCIENCE LETTERS 15 (1972) 286

EARTH AND PLANETARY SCIENCE LETTERS 15 (1972) 286-290. NORTH-HOLLAND PUBLISHING COMPANY
METAL/SILICATE
FRACTIONATION
IN THE SOLAR SYSTEM
John S. LEWIS
Planetary Astronomy Laboratory, *
Department of Earth and Planetary Sciences,
and
Department of Chemistry, Massachusetts Institute of Technology,
Cambridge, Mass. 02139, USA
Received 2 February 1972
Fractionation between the metal and silicate components of objects in the inner solar system has long been recognized as a necessity in order to explain the observed density variations of the terrestrial planets and the H-group,
L-group dichotomy of the ordinary chondrites. This paper discusses the densities of the terrestrial planets in light of
current physical and chemical models of processes in the solar nebula. It is shown that the observed density trends in
the inner solar system need not be the result of special fractionation processes, and that the densities of the planets
may be direct results of simultaneous application of both physical and chemical restraints on the structure of the
nebula, most notably the variation of temperature with heliocentric distance. The density of Mercury is easily attributed to accretion at temperatures so high that MgSiO3 is only partially retained but Fe metal is condensed. The
densities of the other terrestrial planets are shown to be due to different degrees of retention of S, O and H as FeS,
FeO and hydrous silicates produced in chemical equilibrium between condensates and solar-composition gases. It is
proposed that Mercury and Venus Have cores of Fe °, Earth has a core of Fe ° containing substantial amounts of FeS,
and Mars has a quite small core of FeS with more FeO in its mantle than in Earth's. Geophysical and geochemical
consequences of these conclusions are discussed.
The densities of the terrestrial planets can be most
accurately determined by radar tracking o f Mercury,
Venus, and Mars, and of spacecraft in interplanetary
space. The most recent estimates o f the densities o f
the planets, by Ash et al. [ 1] are : Mercury, 5.42
g/cm 3; Venus, 5.25; Earth, 5.51 ; and Mars, 3.96.
Corrected to zero pressure, the densities become
~5.3, 4.40, 4.45, and ~3.85 respectively. Anderson
and Kovach [2] have shown that the density may be
converted into an equivalent parameter, the mean
atomic weight (#). Their method applied to the data
of Ash et al. yields mean atomic weights o f ~35.4,
26.6, 26.9, and 25.0, respectively, with errors of
~+-0.15 at worst.
Two major theories have been proposed to account
for the densities of the terrestrial planets. Urey [3]
showed that, with the solar abundances o f the ele* Contribution no. 45 of the Planetary Astromony Laboratory.
ments then accepted, profound fractionation of metalhc iron relative to silicates was required in order to effect the observed variations in density. Because of the
inaccuracies in available estimates of the density o f
Venus, Urey was unable to conclude that Earth was
in fact more dense than Venus. A very different view
was that of Ringwood [4] who hypothesized that the
abundances o f the major rock-forming elements were
the same in the various terrestrial planets, and that the
density differences were caused by differences in oxidation state due to imcomplete reduction of primitive
Fe 3+ to Fe ° by carbon. The density o f Mercury was
attributed to volatilization of silicates by the Sun
during its superluminous pre-main sequence phase.
My present purpose is to reassess this problem in
light of the recent upward revision in the iron abundance [ 5 - 1 1 ], Cameron's detailed physical models o f
the solar nebula [12], and recent progress in understanding the chemistry of solar material [ 1 4 - 2 0 ] .
J.S. Lewis, Metal/silicate fractionation in the solar system
Cameron has presented calculations on the evolution of the primitive solar nebula in which the temperature, pressure, and density profiles are computed
for the early nebula [ 12]. The structure of the inner
portions of the nebula is governed by rapid turbulence,
and the temperature-pressure profile is closely adiabatic. Within the central plane of the nebula the cooling
process is essentially isobaric at any given point. I will
present here a subset of my results from chemical
equilibrium calculations on solar material. For the
present it will suffice to simplify the chemical system
under consideration to H, C, O, S, Si, Mg, Ca and Fe
I
I
I
I
I
I
1
28 7
in their solar proportions. Because there is still an uncer
tainty of ~-+20% in the solar Fe: Si ratio, we will calculate the density of the condensate (as ~) vs. T for
several different acceptable values of this solar abundance ratio. Fig. 1 presents the major features of the
chemistry of solar material as a general function of
both temperature and pressure, and is marked with
an adiabat schematically indicating the temperaturepressure profile through the-nebula early in its history. Pressure can be directly related to heliocentric
distance through Cameron's models, and the points
of origin of the planets are indicated [ 12, 13]. Fig. 2
presents the density of the condensate along an isobaric section through fig. 1 at 10 .3 bar total pres-
2000
36
,
T
'
1
'
I
'
T ~ I
'
I
'
'
Mercury
CaTiO 3
/
150C
MgSi03
:54--
?
o
W
~
•
d
Venus
Earth
Mars
::k
g
Fe
o52
g
g
IOOC
500
'o~c
"[ 28
I
-6
condensation
FeO
/
-7
MgSi03
50
FeS
0
K
I
I
I
ice
I
-5
-4
-3
-2
-I
Ioglo PRESSURE ( b a r s }
1
g
I
0
0
o~
+1
FeS
format ion
26--
-0.86
Fig. l. Some major features o f the chemistry o f solar material,
0 to 2000°K and 10 7 to 101 bar total pressure. The condensation curves of CaTiO3 (a representative refractory mineral),
Fe, MgSiO3 and ice; the appearance temperatures of FeS, trem o h t e (a h y d r o u s calcium silicate) and talc (a h y d r o u s magnesium silicate); and the line at which Fe metal is wholly oxidized to FeO (as FeSiO3 and Fe2SiO4) are indicated. An adiabat
for the nebula is drawn in, and nebular pressures corresponding to the formation conditions o f the planets in Cameron's
models are marked on the adiabat. The symbols, reading from
the high-temperature end downward, are for Mercury, Venus,
Earth, Mars, the asteroids, Jupiter, Saturn, Uranus, and Neptune. Temperatures of formation deduced from correlating
the density data in fig. 2 with the observed densities of the
terrestrial planets are in excellent agreement with this adiabat.
2400~16
~-
I
~ Fe/Si = 1,06 ~ _ ~ _ ~ . / " ~
k Fe/Si = 0 96
-'~/-~T
/-~\/0t
I ~ I ~ L ~ I ~ I
1400 1200 I 0 0 0 8 0 0 6 0 0
Ternperature,
°K
geO ond
./tremolile
formation
'Orrn°tiOn i
400
Fig. 2. Density of condensed material inequilibrium with a
solar-composition gas, 4 0 0 - 1 6 0 0 ° K at 10 - 3 bars. A simplified chemical system (the 20 m o s t a b u n d a n t elements) is employed for three different values of the Fe: Si ratio. The densities of the planets are excellently consistent with an Fe: Si
ratio of 1.08, b u t the omission of rare elements and uncertainties in the abundances of major elements could displace the
entire manifold of curves shghly. A true Fe: Si ratio below
1.0 is still possible.
288
.LS. Lewis, Metal/silicate fractionation in the solar system
sure, ranging from the condensation temperature of
Fe ° down to 400°K. Three different solar Fe: Si ratios
are indicated. The most important features of these
graphs are (1) condensation of MgSiO 3 near 1400°K,
(2) removal of the H2S from the gas by reaction with
solid Fe to make FeS, and (3) oxidation of the remaining elemental iron to FeO-bearing silicates. The density of the condensate drops rapidly from that of
pure iron (for which B = 55.8) during MgSiO 3 condensation, rises briefly-by ~1% when FeS forms, and then
drops by ~5% when Fe ° is oxidized to FeO. Thus the
density of condensed material in chemical equilibrium
with the nebula shows the same general features that
are observed in the densities of the planets. The
striking 1% increase in the density of Earth relative to
Venus is attributed to the retention of the heavy
volatile element sulfur by the Earth. Since the mean
atomic weight of sulfur is 32, retention of sulfur in
the Earth will cause a 1% increase in tt for each 5% of
sulfur content.
The correlation between the observed densities of
the terrestrial planets and those predicted by the equilibrium theory is quite good. It is interesting that so
good a fit can be obtained without invoking any
mechanism for m e t a l - s i l i c a t e fractionation. The relative abundances of silicon, magnesium, iron, calcium,
aluminum, etc. are exactly the same in Venus, Earth
and Mars for this model. Mercury is seen to condense
at such a high temperature that it is depleted in magnesium, silicon, and the alkali metals relative to iron,
calcium, aluminum, titanium, nickel and other refractory materials. We conclude that Urey's [3] proposal
of iron-silicate fractionation involving all the terrestrial
planets is an artifact of the low cosmic abundance of
iron then accepted, and that m e t a l - s i l i c a t e fractionation by physical means (magnetic forces, ease of fragmentation, etc.) need not be postulated in so far as
the densities of the inner planets are concerned.
Ringwood's [4] model for producing the terrestrial
planets by in situ reduction of Fe304 to Fe ° by carbon
involves several acl hoc hypotheses. In particular we
note that Ringwood requires carbonaceous chondrite
parent material, and faces the embarrassing task of removing a mass of CO from Earth roughly equal to the
mass of Mars; that Ringwood postulates two separate
processes to account for the densities of the terrestrial planets; that the density of Venus is explained as
being less than that of Earth because of the "higher
degree of oxidation of Venus", for which no cause is
suggested. Ringwood's basic assumption that differences in oxidation state and in the degree of retention
of volatiles can explain the densities of Venus, Earth
and Mars is valid, but we find that, because of the
high atomic weight of sulfur, density is not a monotonically decreasing function of oxidation state and
volatile retention. In the present model, oxidation
state and volatile retention do correlate monotonically with heliocentric distance, and the reason for
this correlation is that the chemical composition of
the condensate is a function of its temperature of formation.
Because of the absence of any p r o o f that m e t a l silicate fractionation influences the density of the
terrestrial planets, any hypothesis which requires that
these planets have accumulated involatile chemical
components incompletely [12], in a disequilibrium
manner [4, 21 ], or with physical fractionation between components [3] must explain why the equilibrium theory with no fractionation does such a satisfactory job of explaining the data with far fewer arbitrary assumptions. The variation in Fe: Si ratio in
the several families o f chondrites is attributed to their
origin in parent bodies as small as 10 -12 times the
mass of the Earth, in which statistical fluctuations in
composition or local variations in conditions of accretion might be enormously larger than for objects of
planetary mass.
The models for the terrestrial planets which emerge
from the present work are as follows:
Mercury. A massive core of Fe-Ni alloy is surmounted by a small mantle o f Fe2+-free magnesium silicates.
Ca, A1, Ti and other refractory oxides are present, but
only traces of alkali metals, sulfur, FeO, etc. (due
perhaps to "contamination" b y infalling debris) could
be present. This agrees with Ringwood's picture of
Mercury, but arises not as a result of a special additional postulate, but as an integral part of the model
for formation of the planets.
Venus. A massive core of Fe-Ni alloy is surmounted b y a massive mantle of Fe2+-free magnesium cilicates. Iron, magnesium, and silicon are present in cosmic proportions. A silica-rich crust similar in composition to Earth's is present [22]. Sulfur is probably
virtually absent from the planet [23].
Earth. An inner core of Fe-Ni alloy and outer core
of Fe-FeS melt are present. Certain chalcophile ele-
J.S. Lewis, Metal/silicate fraetionation in the solar system
ments are deficient in the mantle and crust but enriched in the outer core [ 2 4 - 2 6 ] . These elements,
which are chalcophile above the Fe-FeS eutectic temperature at oxygen fugacities typical of ordinary chondrites, may include potassium, rubidium and cesium.
Earth's mantlle contains ~10% FeO. The crust and
upper mantle are oxidized relative to the rest of the
Earth, due to preferential enrichment of Fe 3+ in upwardflowing differentiate melt relative to solid olivine, and
by the effects of the Earth surface environment. Depletion of chalcophiles in the crest and the marked deficiencies of S, K, Rb and Cs in the crust and upper mantle
are due to their extraction into an FeS-rich melt. The
overall composition of the Earth is rather close to
that of the H-group chondrites, but is not identical to
nor derived from them or any other class o f meteorites.
Mars. Mars is essentially devoid of free iron. It may
contain a core of FeS with or without a small amount
of Fe ° [27]. Its mantle is rich in FeO [FeO/(FeO +
+ MgO) ,-~ 0.5]. Hydrous minerals were retained in
appreciable quantities during accretion. The present
crust (assuming differentiation has taken place and
been reasonably complete) should be similar to but
more iron-rich than Earth's.
Of the terrestrial planets, only Earth is likely to
have accreted at a relatively low degree o f oxidation
with both abundant Fe ° and FeS present. Only Earth
is likely to have differentiated so as to extract the
heavy alkali metals into the core. Thus the very large
heat source of ~1020 erg/sec possible within the outer
core of the Earth is impossible on Mercury or Venus
and unlikely on Mars. The presence of a planetary magnetic dipole field on Earth and its absence on Mars is
attributed to the difference in core composition.
Heat flow through the crusts of Mars, Earth, and
Venus should closely approximate that predicted by
the chondritic model. On earth, deep convection of
the mantle delivers heat from decay of deep-seated
40K to the base of the thin oceanic crust, and the continental crust is driven to positions over the subsiding
parts of the upper mantle. The near-equality of the
continental and oceanic heat flow and the close correspondence with the chondritic model are direct resuits of this model. In addition, the presence of the FeFeS eutectic melt very early in the thermal history of
the Earth provides a simple explanation o f early differentiation of the Earth without the necessity of
289
postulating very rapid accretion. Because o f the postulated role of an Fe-FeS eutectic melt in permitting
early differentiation of the Earth, the geochemical
differentiation of Earth will differ in several respects
from that of the other planets. The most important
distinctions are in the potassium distribution and in
the time of differentiation. Venus, Earth's near-twin,
may take much longer to differentiate than Earth. In
fact, Earth may differentiate before it is completely
accreted, because adiabatic compression o f cold material is by itself adequate to heat most of the Earth's
interior to above the Fe-FeS eutectic temperature.
Heat flow through the crust of Mercury should be
higher than that calculated for a potassium-free chondritic model because we expect complete retention of
U and Th relative to Fe, but less dilution b y magnesium silicates. A heat flow value close to that for a
chondritic model could result from this accidental
compensation.
The densities of the planets are not the only observable properties which may be strongly affected by
the temperature of formation and accretion: a very
interesting question is the different degree of retention of volatiles, such as H20, by the terrestrial planets.
The data presented herein suggest that Earth accreted
on the edge of the tremolite [Ca2-MgsSi8022(OH)2 ]
stability field and Mars accreted well within it. Mars
should thus contain roughly one water molecule for
every two calcium atoms, or ~0.3% of the mass of
Mars, and Earth should contain several times less
water. A component of volatiles from accretion of
a small proportion (<~ 1%) of cometary or carbonaceous chondrite material cannot be ruled out [29]. In
this model the primitive material present throughout
the asteroid belt is of carbonaceous chondrite composition.
Further evidence for the existence of a significant
radial temperature gradient and rather low temperatures in this region can be found in the densities of
the satellites of Jupiter. The temperature gradient
must not be substantially less than adiabatic or it becomes impossible to get accretion temperatures for
the Galilean satellites low enough to explain their observed low densities [20]. The evidence from condensation models for the terrestrial planets combined
with similar work on the satellites of the outer planets
combine to suggest that the ordinary chondrites originated inside the orbit of Mars. The present orbits of
29(/
J.S. Lewis, Metal/silicate fraetionation in the solar system
the o r d i n a r y c h o n d r i t e p a r e n t b o d i e s are speculative.
A n d e r s a n d c o w o r k e r s [28, 29] have argued t h a t the
E a r t h a n d t h e o r d i n a r y c h o n d r i t e s o r i g i n a t e d at very
nearly the same t e m p e r a t u r e a n d pressure. I w o u l d
claim t h a t this is n o t e v i d e n c e for a great u n i f o r m i t y
o f c o n d i t i o n s over a wide range o f h e l i o c e n t r i c dist a n c e , b u t r a t h e r suggests an origin for o r d i n a r y c h o n dritic m a t e r i a l near 1.2 A U f r o m the S u n .
A d e t a i l e d s t u d y o f the c h e m i s t r y o f solar m a t e r i a l
is in p r e p a r a t i o n , and will b e p r e s e n t e d elsewhere.
Acknowledgements
I am g r a t e f u l to Prof. A.G.W. C a m e r o n for m a n y
h e l p f u l discussions o f p h y s i c a l m o d e l s o f t h e p r i m i t i v e
solar n e b u l a , t o Prof. H.C. U r e y for r e p e a t e d l y stimulating m y i n t e r e s t in this p r o b l e m over t h e past seven
years, and t o Prof. E. A n d e r s for his h e l p f u l c o m m e n t s
on t h e m a n u s c r i p t o f this p a p e r . This w o r k was supported by NASA under grant NGL-22
009-521.
References
[ 1 ] M.E. Ash, I.I. Shapiro and W.B. Smith, The system of
planetary masses, Science 174 (1971) 551.
[2] D.L. Anderson and R.L. Kovach, The composition of the
terrestrial planets, Earth Planet. Sci. Letters 3 (1967) 19.
[3] H.C. Urey, The origin and development of the Earth and
other terrestrial planets, Geochim. Cosmochim. Acta 1
(1951) 209.
[4] A.E. Ringwood, Chemical evolution of the terrestrial
planets, Geochim. Cosmochim. Acta 30 (1966) 41.
[5] T. Garz, H. Holweger, M. Kock and J. Richter, Revised
solar iron abundance and its influence on the photospheric model, Astron. Astrophys. 2 (1969) 446.
[6] T. Garz, M. Kock, J. Richter, B. Baschek, H. Holweger
and A. Uns61d, Abundances of iron and some other elements in the Sun and in meteorites, Nature 223 (1969)
1254.
[7] N. Grevesse and J.P. Swings, Forbidden lines of FelI in
the solar photospheric spectrum, Astron. Astrophys. 2
(1969) 28.
[8] B. Baschek, T. Garz, H. Holweger and J. Richter, Experimentelle Oszillatorenst~irken yon FeIl Linien und die
solare Eisenh~ufigkeit, Astron. Astrophys. 4 (1969) 229.
[9] J.M. Bridges and W.L. Wiese, The oscillator-strength
scale for FeI, Astrophys. J. 16 (1970) L71.
[10] T. Garz, J. Richter, H. Holweger and A. Uns61d, Abundance of iron in the solar photosphere, remarks concerning a paper by John E. Ross (1970), Astron. Astrophys.
7 (1970) 336.
[ 11 ] S.J. Wolnik, R.O. Berthel and G.W. Wares, Shock-tube
measurements of absolute gf-values for FeI, Astrophys.
J. 162 (1970) 1037.
[ 12] A.G.W. Cameron, Physical conditions in the primitive
solar nebula, in: Meteorite Research, ed. P.M. Millman,
(Reidel, Dordrecht, 1969).
[13] A.G.W. Cameron, personal communications (1971).
[ 14] H.C. Lord, Molecular equilibria and condensation in a
solar nebula and cool stellar atmospheres, Icarus 4
(1965) 279.
[ 15] J.W. Larimer, Chemical fractionations in meteorites
I. Condensation of the elements, Geochim. Cosmochim.
Acta 31 (1967) 1215.
[16] J.W. Latimer and E. Anders, Chemical fractionation in
meteorites - I1. Abundance patterns and their interpretation, Geochim. Cosmochim. Acta 31 (1967) 1239.
[17] J.W. Larimer and E. Anders, Chemical fractionations in
meteorites - III. Major element fractionations in chondrites, Geochem. Cosmochim. Acta 34 (1970) 367.
[ 18] L. Grossman, Ph.D. Dissertation, Yale University (1971 ).
[ 19] L. Grossman, Condensation in the primitive solar nebula,
Geochim. Cosmochim. Acta (1972) in press.
[20] J.S. Lewis, Low temperature condensation from the solar
nebula, Icarus 16 (1972) in press.
[21] K.K. Turekian and S.P. Clark, Jr., Inhomogeneous accretion of the Earth from the primitive solar nebula, Earth
Planet. Sci. Letters 6 (1969) 346.
[22] J.S. Lewis, An estimate of the surface conditions of
Venus, Icarus 8 (1968) 434.
[23] J.S. Lewis, Venus: Atmospheric and lithospheric composition, Earth Planet. Sci. Letters 10 (1970) 73.
[24] V. Rama Murthy and H.T. Hall, The chemical composition of the Earth's core: Possibility of sulfur in the core,
Phys. Earth Planet. Int. 2 (1970) 276.
[25] J.S. Lewis, Consequences of the presence of sulfur in
the core of the Earth, Earth Planet. Sci. Letters 11
(1971) 130.
[26] H.T. Itall and V. Rama Murthy, The early chemical history of the Earth: Some critical elemental fractionations, Earth Planet. Sci. Letters 11 (1971) 239.
[27] D.L. Anderson and T.H. Jordan, The composition and
evolution of Earth and Mars, Trans. Am. Geophys.
Union 52 (1971) 349 (Abstract).
[28] J.S. Larimer, Composition of the Earth: Chond~itic or
achondritic?, Geochim. Cosmochim. Acta 35 (1971)
769.
[29] E. Anders, Chemical processes in the early solar system
as inferred from meteorites, Accts. Chem. Res. 1 (1968)
289.