PERSPECTIVE
doi:10.1038/nature10901
Evidence against a chondritic Earth
Ian H. Campbell
1
& Hugh St C. O’Neill
1
The
142
Nd/
144
Nd ratio of the Earth is greater than the solar ratio as inferred from chondritic meteorites, which challenges
a fundamental assumption of modern geochemistry—that the composition of the silicate Earth is ‘chondritic’, meaning
that it has refractory element ratios identical to those found in chondrites. The popular explanation for this and other
paradoxes of mantle geochemistry, a hidden layer deep in the mantle enriched in incompatible elements, is inconsistent
with the heat flux carried by mantle plumes. Either the matter from which the Earth formed was not chondritic, or the
Earth has lost matter by collisional erosion in the later stages of planet formation.
T
he paradigm that underpins much of modern geochemistry is
that the integrated chemical composition of the whole Earth
should be that of the Sun, except for depletion in volatile ele-
ments, according to their volatility under the conditions of the solar
nebula. Similar solar-related compositions are found in ‘chondritic’
meteorites, which are fragments of small rocky bodies that escaped
the usual course of planetary differentiation into a metallic core, silicate
mantle and crust. The composition of a chondritic meteorite is therefore
presumed to reflect its entire parent body. Although the solar composi-
tion can be determined from spectroscopic measurements of the solar
photosphere, measurement is not possible or imprecise for many ele-
ments, is model-dependent and does not give information on the iso-
topic make-up of the elements
1
. Instead, a more complete picture of the
solar composition is inferred from chemical analyses of chondrites. The
compositions of the chondrites vary, with at least 27 parent bodies
sampled
2
, reflecting local differences in the solar-nebula-like density
or proportions of gas to solids, or different accretion processes. The
various chondrite compositions are distinguished by enrichment or
depletion of refractory elements, ratio of lithophile to siderophile ele-
ments (for example, Mg/Fe), oxidation state, oxygen isotopic composi-
tions, and their patterns of depletion of the volatile elements. No
examples with volatile-element enrichment are known, except for a
slight enrichment in a few of the least volatile of these elements in the
highly reduced enstatite chondrites. Yet all chondrites share one dis-
tinctive compositional feature: their refractory lithophile elements
(RLEs) are present in the same ratio relative to each other and to the
solar composition. The RLEs are defined by two properties: they are
refractory, because they condense from a gas of solar composition at
temperatures higher than the main constituents of rocky planets, the
magnesium silicates and iron metal; and they are lithophile, because
they do not enter metal or sulphide phases, either in chondrites or into
the metallic cores formed during planetary differentiation. There are 28
RLEs that are stable or have long half-lives; they include Ca and Al
among the major elements, the entire suite of rare earth elements
(REEs), and the radiogenic heat-producing elements U and Th.
The constant RLE ratio rule is ever challenged on several fronts: by
exceptions due to terrestrial weathering or ad hoc effects on parent
bodies such as impact brecciation, incipient melting or aqueous altera-
tion; by the scale of heterogeneity in chondrites relative to the small
sample sizes available for analysis; by improvements in the precision of
chemical analysis; and by the increasing numbers of chondritic meteorites
available for analysis. It is therefore difficult to quantify the precision to
which the rule holds, but variations from the solar ratios reflecting whole-
body chemistry that are larger than a few per cent are exceptional. The
REE pattern in the RLE-rich CV chondrite Allende is perhaps the largest
well-attested deviation
3
. New techniques of isotopic analysis are revealing
small anomalies in the isotopic make-up of heavy elements in bulk
samples of chondrites, ascribed to less than perfect homogenization of
different nucleosynthetic components in the solar nebula, such as Ti
(ref. 4), Ni (ref. 5), Ba (refs 6 and 7) and Mo (ref. 8). This evidence
challenges the conceptual basis behind the constant RLE ratio rule, but
as yet at no more than the few-per-cent level already accepted. For
example, although the range in Lu/Hf and Sm/Nd in unequilibrated
carbonaceous, ordinary, and enstatite chondrites is as much as 7.9%
and 3.5% respectively, the average Lu/Hf and Sm/Nd values for these three
main classes of chondrites agree within 1% and 0.3% respectively
9
.
Although most geochemists long ago abandoned the notion that the
Earth’s composition mimics any particular type of chondrite
10
,theidea
that the Earth has solar ratios of RLEs has persisted, providing the fun-
damental reference frame against which trace element and radiogenic iso-
topic ratios are compared. This reference frame has usually, if confusingly,
been termed ‘chondritic’ rather than ‘solar’ in the literature, because of the
history of fine-tuning the presumed solar composition to the compositions
of chondrites. Emphasis has been placed on the CI chondrites, which match
the solar composition within uncertainty for many elements irrespective of
their chemical properties, except for the most volatile elements; but CIs are
rare, and useful analyses come from just three falls, Orgueil, Ivuna and
Alais
11
,andmostlyfromOrgueil
1
. The implicit assumption is that the bulk
chemical composition of rocky bodies is established at the earliest stages in
the planet-building process, of which the chondrite parent bodies are
relicts. In this view, the subsequent stages by which these small chondritic
bodies collide and merge to form planetary embryos and ultimately Earth-
sized planets, while resulting in extensive differentiation associated with
melting, does not affect the integrated whole-planet compositions: these, it
is assumed, remain ‘chondritic’.
Planetary accretion
The ‘chondritic’ hypothesis for the Earth’s composition is a survivor
from times when the understanding of how terrestrial planets form
was quite different. Current models can be traced back to Safronov,
whose monograph on the subject was only translated into English in
1974 (ref. 12). Our present understanding is that planetary accretion
proceeds through several stages of ever increasing average size. Initially,
runaway growth is from kilometre-sized bodies (approximately the size
of parent bodies of chondrite meteorites) to form planetesimal oligarchs
about a thousand kilometres in diameter. This is the approximate size of
the differentiated bodies that are thought to be parental to the achondrite
meteorites, like Vesta (diameter 500 km). The achondrite parent bodies
1
Research School of Earth Science, Australian National University, Canberra, Australian Capital Territory 0200, Australia.
29 MARCH 2012 | VOL 483 | NATURE | 553
Macmillan Publishers Limited. All rights reserved
©2012
all show non-chondriticpatterns of volatileloss,forexample, in their Mn/
Na ratios
13
. This post-nebular volatile loss can be dated using the Rb/Sr
chronometer to several million years after the chondrite stage
14
. The last
stages of planetary accretion see the assembly of these planetesimals into
Moon-to-Mars-sized planetary embryos, which then merge through
highly energetic collisions to form a few terrestrial planets. The giant
impact that formed the Moon is the last of significance in the formation
of the Earth. None of this was envisaged when the chondritic hypothesis
was first advanced. Maintaining the hypothesis in its current form
requires that the collisions between bodies during the several stages of
accretion result in no net fractionation of RLEs, despite the bodies being
already differentiated at the earliest stages, which does not seem probable.
Challenge to the chondrite paradigm
The challenge to this paradigm started with two landmark papers showing
that the
142
Nd/
144
Nd ratio of chondritic meteorites is 20 6 5partsper
million less than rocks of terrestrial mantle origin
15,16
.
142
Nd is the daughter
of
146
Sm, with a half-life generally assumed to be 103 million years (Myr)
but which could be as short as 68 Myr (ref. 17). Sm and Nd are two typical
RLEs, and their isotopic relationships provide particularly powerful
constraints on Earth differentiation models, because the short-lived
146
Sm/
142
Nd system is complemented by the long-lived
147
Sm/
143
Nd
system (half-life 106 billion years, Gyr), which has long been used
to constrain the sizes of Earth reservoirs. Although there is some
nucleosynthetic variability in the isotopic composition of Nd among
chondrites, this is unlikely to explain the difference in
142
Nd/
144
Nd
(ref. 18), which therefore requires the ratio of Sm to Nd to be about 6%
above the average chondritic value
15,16
; because of its isotopic signifi-
cance, this value is known to within 0.3% (refs 9 and 19).
There are two possible interpretations
15
. The simplest is that the Earth is
not chondritic after all, and the measured
142
Nd/
144
Nd ratio of terrestrial
samples is that of the bulk silicate Earth (BSE). Many geochemists have
opted for an alternative scenario, in which the Earth’s mantle underwent
an early fractionation event into an early-enriched reservoir with low Sm/
Nd and an early-depleted reservoir (EDR) with high Sm/Nd
15,16,20–22
.
Boyet and Carlson (refs 15 and 16) suggest that the low-Sm/Nd reservoir
was an incompatible-element-enriched basaltic crust that sank to the
core–mantle boundary, becoming isolated in the seismically anomalous
region at the base of the mantle called D99, an irregular 200–250-km-thick
layer overlying the core. The seismic properties of D99 make it unlikely to
be a simple thermal boundarylayer and it is interpreted to be a stable layer
with a density greater than that of the overlying mantle
23,24
. Furthermore,
because the half-life of
146
Sm is no more than 103 Myr, the early frac-
tionation event must have occurred well within the first 10 Myr of Solar
System formation to prevent the average e
143
Nd value of the comple-
mentary EDR rising above about 10, the observed average value of the
mid-ocean-ridge basalt (MORB)
25,26
.
The
142
Nd/
144
Nd ratios of lunarsamples, however, are indistinguishable
from terrestrial values
27–29
, so the Moon and the Earth developed their
142
Nd/
144
Nd enrichment before the Moon’s formation
30
. If the Moon
formed by a giant impact, the collision would have melted and
homogenized the Earth’s mantle
31
, which would have destroyed any
hypothetical early-enriched reservoir
30
. Furthermore, the effect of extrac-
tion of the continental crust from the mantle on the Sm–Nd mass balance
can be modelled, assuming that the BSE has a composition similar to the
EDR, without a hidden reservoir
30
.
The heat carried by mantle plumes provides an equally compelling
argument against a hidden early-enriched reservoir
32
. Most hidden-
reservoir advocates
15,16,33,34
equate it with D99.IfD99 is the hidden
early-enriched reservoir, then under the chondritic assumption, a mass
balance requires it to contain over 40% of the Earth’s heat-producing
elements (U, Th and K), which produce 9 terawatts (TW) of heat
15
.
Regardless of whether D99 is a stable layer or forms a double-diffusive
convecting layer
35
, the heat it liberates can only be transmitted through
the overlying mantle in plumes. The amount of heat transferred by
mantle plumes can be estimated from the dynamic topography they
generate on the sea floor. Early estimates
36,37
placed the heat flow carried
by plumes at 3.5 TW, which has been revised
38
to 7 TW or 15% of the
Earth’s total heat flow
39
of 47 6 2 TW, to take into account the mantle’s
subadiabatic thermal gradient. However, models of mantle heat transfer,
which takes into account both the heat required to warm subducted
lithosphere and the additional heat required to lift compositionally
dense plumes, suggest a higher figure
38,40–42
of 7–14 TW. From this must
be subtracted the heat transferred from the core to the mantle, which
41
must be at least 3–4 TW, the minimum amount of heat required to
sustain the geodynamo, but is more likely to lie between 5–7 TW and
12–14 TW (ref. 38). As a consequence, if heat transfer from the core to
the mantle is greater than the low estimate of 5 TW or if the heat carried
by mantle plumes is less than the high estimate of 14 TW, 40% of the
Earth’s heat-producing elements cannot be hidden in the D99 layer.
A sudden drop in the maximum MgO content of plume-related
komatiites and picrites 2.5 Gyr ago from 30–35 wt% to 18–23 wt%,
which implies a temperature drop of 200 to 250 uC, has also been used
to argue that D99 did not form until the end of the Archaean eon
35
. The
simplest explanation for the observed drop in MgO is that D99 formed as
a stable layer about 2.5 Gyr ago, which insulated the mantle from the
core. The predicted drop in plume temperatures, depending on assump-
tions such as whether D99 is a stable layer or formed a double-diffusive
convecting layer, lies within the range 33% to 50%, which is consistent
with the observed MgO drop
35
.IfD99 formed after the first 10 Myr of the
Earth’s evolution it cannot be responsible for the Earth’s
142
Nd/
144
Nd
anomaly.
If the composition of the EDR is that of the BSE
20
, or if it formed
within 10 Myr of the Solar System
25,26
and is unaffected by subsequent
events, then the e
143
Nd value for the EDR today
30,43
is 7.0 6 2.0. These
values are remarkably similar to the prevalent mantle (PREMA) value of
7 6 1 in ocean island basalts (OIBs)
44,45
. A component with this Nd
characteristic can be recognized in four flood basalts
43
— the Baffin
Island–West Greenland province, the Antarctic Karoo, the Siberian
Traps and the Deccan Traps—and in two oceanic plateaus—the
Kerguelen and Ontong Java plateaus
35
. The EDR component plots
within 150 Myr of the 4.5-Gyr geochron on a
207
Pb/
204
Pb versus
206
Pb/
204
Pb diagram, which suggests that it developed early in the
Earth’s history
43
. It also has enriched
3
He/
4
He, showing that the source
region to these plumes is less degassed than the MORB source
30,43
.
Torsvik et al. suggest that most plume-related basalts that erupted
over the last 320 Myr were above one of two large low-shear-wave-
velocity provinces, which form part of D99 and make up about 2% of
the mantle’s mass
46
. However, this primitive component in plumes may
be much more extensive
35,45
. The isotopic trend in a number of ocean
island suites, produced by melting plume tails, converges towards a
common point called FOZO
44
. A remarkably similar component was
identified in the two oceanic plateaus that are free from contamination
by continental crust, the Ontong Java and the Caribbean plateaus
35,43
.
Basalts from these plateaus are characterized by flat REE patterns and in
this respect they are dissimilar to OIBs and MORBs but are isotopically
similar to the EDR recognized in flood basalts
43
. Archaean basalts asso-
ciated with komatiites, whose Nb/U ratios indicate that they are free of
crustal contamination
47–50
, commonly have flat REE patterns
48–50
, and
basalts from Kambalda in Western Australia have an e
143
Nd value of 3
(ref. 47), which lies on the EDR growth curve for Nd at 2.7 Gyr ago.
Jackson and Carlson
43
interpret the EDR component to originate from
the boundary layer source of plumes, whereas Campbell and Griffith
35
suggest that it is lower mantle that was entrained into plumes during
their ascent. The first interpretation allows the early-enriched reservoir
to be a minor component in plumes, whereas the second requires it to be
a dominant component in the lower mantle. Because this difference is
critical to the debate on whether the Earth has chondritic RLE ratios, we
now summarize the basis for the second interpretation
35,51
.
Plumes must originate from a thermal boundary layer; in the Earth,
only the core–mantle boundary has the properties required to sustain the
temperature drop implied by plume activity on geological timescales
52
.A
RESEARCH PERSPECTIVE
554 | NATURE | VOL 483 | 29 MARCH 2012
Macmillan Publishers Limited. All rights reserved
©2012
new plume has a large head, which is followed by a smaller tail (Fig. 1). As
the head rises through the mantle it heats the adjacent mantle, lowering
its density so that it is swept into the plume head by its recirculating
motion. The plume head is therefore a mixture of material from the hot
boundary-layer source of the plume (dark material in Fig. 1) and cooler
entrained material (light material in Fig. 1). This entrained material,
which makes up a large fraction of the head, comes from the lower half
of the lower mantle
51,52
. When a plume head reaches the top of the mantle
it melts to produce a continental flood basalt or an oceanic plateau
52,53
.The
first material to reach the top of the mantle and undergo decompressional
melting is the hot mantle that originated from the boundary layer above
the core (Fig. 1), which melts to produce picrites or komatiites. Later,
when the head flattens against the overlying lithosphere, the cooler
entrained material from the lower mantle, which is calculated to be three
times as thick as the hot material from the boundary layer
54
,entersthe
melting zone. Production of high-temperature picrites or komatiites
should therefore be followed by lower-temperature basalts formed by
melting a mixture of boundary layer and entrained lower-mantle with
the latter becoming increasingly important with time as the plume head
continues to rise and flatten. It is this primitive entrained lower-mantle
component that has been identified in the Ontong Java and Caribbean
plateaus
35,43,51
and in several flood basalt provinces
43
.
Lower-mantle material is also entrained along the side of rising plume
tails by viscous drag (Fig. 1). As a consequence, when plume tails melt to
produce a chain of ocean islands, which progressively increase in age away
from the current position of the plume,the basalts that make up the islands
may contain a component from the lower mantle. The existence of FOZO
as a component in plume tail basalts is therefore consistent with the
interpretation
45
that it was entrained from the lower mantle during ascent
of a plume tail (Fig. 1). Its presence in all or most plume tails provides
further support that FOZO is a major component in the lower mantle.
FOZO
45
,EDR
15
,SCHEM
30
(super-chondritic Earth model), and
NCPM
43
(non-chondritic primitive mantle) are therefore all the same:
theprimordial component of theBSE. The evidence from oceanic plateaus,
flood basalts and OIBs is that this is a major component in the lower
mantle. The high
3
He/
4
He ratios recognized in the EDR component are
consistent with this interpretation
30
. The displacement of most oceanic
plateaus and flood basalts, with e
143
Nd between 5 and 9, slightly to the
right of the geochron on a plot of
207
Pb/
204
Pb against
206
Pb/
204
Pb, shows
that the lower mantle, as sampled by Phanerozoic plumes, includes some
recycled oceanic crust
43
. Basalts from Malaita Island, part of the Ontong
Java Plateau, have an average Nb/U of 42, which is between that of the
average for both the OIB- and MORB-type mantles (47) and the BSE
value of 34, indicating that some continental crust has also been extracted
from the lower mantle
51
.
The picture of mantle convection that is emerging is one in which the
upper mantle differentiates into harzburgite and basalt at mid-ocean
ridges, which are subducted to the thermal boundary layer above the
core and returned to the upper mantle in plumes, mixed in varying
proportions, with perhaps a small sedimentary component. It appears
that the bulk of the lower mantle is largely bypassed by this process, so
that it is less degassed, has had less subducted basalt mixed into it, and
less continental crust extracted from it, compared to OIB- and MORB-
source mantles.
Volatile elements
Although the comparison of Earth’s composition with chondrites has
focused on the RLEs, the differences in the abundances of the moderately
volatile elements between the Earth and chondrites provide another line
of evidence that chemical fractionations, which occurred in the latter
stages of accretion, moved the final compositions of terrestrial planets
away from the compositions encompassed by chondritic meteorites
13,14
.
The moderately volatile elements are those calculated to condense from
the solar nebula after magnesian silicates and Fe metal, but at tempera-
tures above the ice-formingelements (H, C, N and the noble gases). Many
moderately volatile elements are also siderophile and were depleted in the
BSE by partitioning into the core, making their bulk Earth abundances
inaccessible, but among the lithophile elements are the alkali metals, the
halogens and boron. For the other potentially siderophile elements like
Zn and In, their BSE abundances provide useful constraints on minimum
Earth abundances. The pattern of depletion of the moderately siderophile
elements is unlike that found in any chondrite group (Fig. 2). In the
chondrites, depletion correlates with calculated condensation tempera-
tures, but in the Earth, elements that are very highly incompatible during
partial melting, namely the heavy halogens and Cs, are much more
depleted than the nearly compatible Zn and In. This is highlighted by
Zn/Cl ratios, which are an order of magnitude greater in the BSE than in
any chondrite group (Fig. 2). The dependence of volatility on incom-
patibility suggests volatilization from early-formed crusts during the
latter stages of accretion
13
. In the case of the halogens, their volatilities
are enhanced by relatively oxidizing conditions that prevailed after dis-
persion of the H-rich solar nebula. Post-nebular oxidizing conditions are
reflected in the ubiquitous fractionation of Na from Mn in small-
differentiated planetary bodies
13
, although this particular fractionation
is not seen in the nearly chondritic Na/Mn ratios of the BSE, perhaps
because some Mn has also been partitioned into the core. Attempts to
define simple volatile-element depletion trends in the BSE have invari-
ably omitted key elements from consideration, like the heavy halogens.
For elements that are both volatile and siderophile, the complexities
of the volatile-element depletion in the Earth, as indicated by the full
pattern (Fig. 2), prevents us from assigning how much of an element’s
loss is due to volatility and how much is due to partitioning into the core.
This is particularly vexing for Pb. Was the extent of Pb depletion by
volatility in the material that formed the Earth like that of Zn and In, or
was it like Cl and the other heavy halogens, or was it somewhere in
between? On the
207
Pb/
204
Pb versus
206
Pb/
204
Pb diagram, it is well
established that both the mantle and crust of the Earth plot well to the
right of the 4.57-Gyr geochron, which, neglecting the ‘hidden reservoir’
explanations, is usually taken to imply loss of Pb relative to parental U
sometime later, at about 4.45 Gyr ago
55
. Because feasible hypotheses can
be constructed to argue for Pb loss by both mechanisms at several stages
Hot mantle from
boundary-layer source
(OIB-type mantle)
Entrained
lower mantle
Hot mantle from
boundary-layer source
(OIB-type mantle)
Figure 1
|
Laboratory model of a mantle plume. The dark-coloured fluid is
from the hot boundary-layer source of the plume, whereas the light material is
cooler overlying fluid that was entrained into the rising plume. In the case of the
mantle the dark-coloured material is from the thermal boundary layer above
the core, whereas the light material is entrained lower mantle (after ref. 54). The
entrained material makes up a large fraction of the plume head
54
and comes
mainly from near the bottom of the lower mantle but above the boundary-layer
source of the plume. The critical parameter in the present context is the
thickness of the upper entrained layer (light material near the top of the plume
head), which is about three times the thickness of layer at the top of the head
(dark material) that originates from the boundary-layer source of the plume
54
.
PERSPECTIVE RESEARCH
29 MARCH 2012 | VOL 483 | NATURE | 555
Macmillan Publishers Limited. All rights reserved
©2012
of the planet-building process, the significance of this age information
remains ambiguous.
Likewise, the interpretation of other early-Earth geochronometers is also
affected if the Earth has a non-chondritic composition. For example, short-
lived
182
Hf (half-life 9 Myr) decays to
182
W, providing a chronometer with
which to constrain the duration of core formation, because W is a
moderately siderophile element that partitioned incompletely into the
core. Evaluating the time significance of this chronometer for the Earth
depends on knowing the W/Hf ratio in the BSE
56
. The W content of the
BSE has been estimated by noting that W/Th (or W/Ba) ratios remain
constant in igneous processes, hence W/Hf 5 (W/Th) 3 (Th/Hf).
Because Hf and Th (or Ba) are both RLEs, it is then assumed that their
ratio is the same as in chondrites. But the non-chondritic Earth model of
ref. 13 predicts that Th/Hf would be only around 70% of the chondritic
ratio; for a simple two-stage model of core formation, this would increase
the calculated time from about 30 to about 35 Myr.
Alternative hypotheses
There are two classes of explanation for the Earth not being chondritic.
Most simply, the compositions of chondrites may not reflect that of the
Solar System precisely enough to deduce detailed element ratios for RLE.
It needs to be remembered, however, that the average Sm/Nd ratio of the
three main classes of chondrites agrees to within 0.3%.
Alternatively, the Earth could have been assembled from initially
chondritic material that was then modified during the subsequent stages
of the planet-building process by collisional erosion
13,57,58
. Current esti-
mates of the Earth’s Fe/Mg ratio are consistent with about 10% of its
silicate part having been lost by this mechanism relative to its Fe-rich
metallic core. The meteorite record attests to the differentiation of small
rocky bodies into metal and silicate being inevitably associated with
partial melting and hence also the formation of an incompatible-
element-enriched crust. If material from these crusts were preferentially
lost during the collisions, it would deplete the Earth systematically in
incompatible elements according to their incompatibility (Fig. 3). One
weakness of the hypothesis is that it implies the loss of incompatible-
element-enriched material to space that no class of meteorite has
sampled. However, no meteorites sample the moderately volatile ele-
ments missing from all chondrites (apart from the CIs) and from the
achondrites and terrestrial planets. Did the gravitation field of the Sun or
Jupiter capture this missing material?
The pattern of depletion caused by preferential collisional erosion is
geochemical rather than cosmochemical, and its effect on the composi-
tion of the Earth’s mantle is essentially the same as that which subse-
quently occurred throughout the Earth’s history by crust formation.
Detecting the effects of collisional erosion therefore depends on obser-
vations that can sum all the reservoirs in the BSE to see whether they add
up to chondritic ratios of RLEs. The Sm/Nd ratio provides the most
compelling evidence, but once the idea of a non-chondritic Earth is
allowed, the resolution of other so-called geochemical paradoxes
becomes achievable.
Geochemists are fascinated by the many paradoxes of the Earth’s man-
tle,which aresummarized in Table1.All of theseparadoxesare predicated
on geochemistry’s most fundamental paradigm; that the Earth was pro-
duced by the accretion of meteorites with the same ratios of RLEs as in
chondrites. Most of these paradoxes disappear if this assumption is
relaxed, but one existing paradox becomes worse: the low value of the
Earth’s Urey ratio. The Urey ratio is the Earth’s radiogenic heat produc-
tion divided by its surface heat flux, which—under the assumption of BSE
U, Th and K concentrations given by the chondritic hypothesis—is about
0.5. The difference must be accounted for by secular cooling. Collisional
erosion lowers the heat-producing element content of the BSE by up to
Mantle/crust partition coefcient
U
La
Sm
Hf
Nd
Th
Lu
Ca Al
10
–4
110
–3
10
–2
10
–1
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Ba
Normalized depletion
Figure 3
|
Depletion of some RLEs in the BSE by preferential collisional
erosion of early-formed basaltic crust during accretion of planetesimals and
planetary embryos. The figure is based on the model of ref. 13, assuming three
constraints: (1) loss of silicate (crusts plus mantles) relative to metallic cores is
10%; (2) the most incompatible RLEs (here represented by Ba) are depleted to
50% of their chondritic abundance; and (3) Sm/Nd is 6% above the chondritic
ratio. The partition coefficients during crust formation are from ref. 67 for the
production of oceanic crust.
Pb
0.001
0.01
0.1
1
400 600 800 1,000 1,200
CV
EH
50% condensation temperature (K)
10
100
1,000
0.5 1 1.5
2
2.5 3
Zn/Mg
Cd
Cl
In
Rb
B
Li
Mn
Na
K
I
Zn
Cs
Br
F
?
Chondrites
BSE
Zn/BrBSE/Mg
Zn/Mg
Figure 2
|
The pattern of volatile element depletion in the BSE for lithophile
elements compared to CV carbonaceous chondrites and EH enstatite
chondrites. CV carbonaceous chondrites are the most volatile-depleted of the
chondrites and EH enstatite chondrites are a class of chondrites sometimes
considered to have affinities to the Earth because of their stable isotope ratios.
Elements are normalized to CI abundances and Mg. The calculated 50%
condensation temperatures of elements for the solar nebula are from ref. 10.
Some elements inthe BSE mayhave been additionally depleted by core formation
(for example, Zn, In and Pb), in which casetheir depletions due to volatility alone
will be overestimated. The chondrites form smooth depletion trends with
calculated condensation temperatures, but the BSE is not only more depleted in
moderately volatile elements than any known chondrite, but the pattern of
depletion is qualitatively different, probably owing to post-nebula volatile loss
under more oxidizing conditions
13
. This is shown, for example, by the BSE Zn/Br
ratios (inset), which are about an order of magnitude greater than that found in
any class of unmetamorphosed chondrites (black circles). The lack of any clear
volatility trendin the BSE means thatit is notat present possible to constrain how
much Pb, if any, was partitioned into the Earth’s core, making the interpretation
of Pb isotopic systematics in terms of Earth accretion and core formation
uncertain. Meteorite abundances are from ref. 66, and BSE abundances from ref.
13, which was derived under the ‘chondritic assumption’. A non-chondritic BSE
would have lower abundances of B, K, Rb, Cs, Cl, Br and I but not Mn, Na, F, Zn
or In (ref. 12), increasing the discrepancy with the chondritic trends.
RESEARCH PERSPECTIVE
556 | NATURE | VOL 483 | 29 MARCH 2012
Macmillan Publishers Limited. All rights reserved
©2012
half
13
, which halves the already low Urey ratio, implying unlikely cooling
rates extrapolated over geological time. Perhaps the Earth is currently in a
phase of abnormally fast ocean crust formation and subduction
59
.
It is apparent that the only reliable way of determining the composi-
tion of the Earth is by sampling the Earth itself. As argued in this study,
the heads of mantle plumes entrain primitive lower mantle. By studying
basalts produced by melting this material, especially Archaean basalts
associated with komatiites, provided they are not affected by crustal
contamination, we are sampling basalts derived from the Earth’s earliest
and most primitive mantle. It may also be possible to obtain the inte-
grated BSE composition of two of the RLEs most susceptible to col-
lisional erosion—Th and U—from the geoneutrino flux
60
. It would
then be possible to see whether the ratios of these elements with other
RLEs, such as Ca and Al, were indeed within the range found in chon-
dritic meteorites or that predicted by collisional erosion.
1. Lodders, K., Palme, H. & Gail, H.-P. in Landolt-Bo
¨
rnstein New Series (ed. Trumper, J.
E.) Vol. VI/4B, Ch. 4.4 560–630 (Springer, 2009).
2. Meibom, A. & Clark, B. E. Evidence for the insignificance of ordinary chondritic
material in the asteroid belt. Meteorit. Planet. Sci. 34, 7–24 (1999).
3. Pourmand, A., Dauphas, N. & Ireland, T. J. A novel extraction chromatography and
MC-ICP-MS technique for rapid analysis of REE, Sc and Y: revising CI-chondrite
and Post-Archean Australian Shale (PAAS) abundances. Chem. Geol. http://
dx.doi.org/10.1016/j.chemgeo.2011.08.011(in the press).
4. Trinquier, A. et al. Origin of nucleosynthetic isotope heterogeneity in the solar
protoplanetary disk. Science 324, 374–376 (2009).
5. Regelous, M., Elliott, T. & Coath, C. D. Nickel isotope heterogeneity in the early Solar
System. Earth Planet. Sci. Lett. 272, 330–338 (2008).
6. Andreasen, R. & Sharma, M. Mixing and homogenization in the early solar system:
clues for Sr, Ba, Nd, and Sr isotopes in meteorites. Astrophys. J. 665, 874–883
(2007).
7. Carlson, R. W., Boyet, M. & Horan, M. Chondritic barium, neodymium, and
samarium isotopic hetrogeneity and early Earth differentiation. Science 316,
1175–1178 (2007).
8. Burkhardt, C. et al. Molybdenum isotope anomalies in meteorites: constraints on
solar nebula and origin of the Earth. Earth Planet. Sci. Lett. 312, 390–400 (2011).
9. Bouvier, A., Vervoort, J. D. & Patchett, P. J. The Lu–Hf and Sm–Nd isotopic
composition of CHUR: constraints from unequilibrated chondrites and
implications for the bulk composition of terrestrial planets. Earth Planet. Sci. Lett.
273, 48–57 (2008).
10. Drake, M. J. & Righter, K. Determining the composition of the Earth. Nature 416,
39–44 (2002).
11. Lodders, K. Solar system abundances and condensation temperatures of the
elements. Astrophys. J. 591, 1220–1247 (2003).
12. Safronov, V. S. Evolution of the Protoplanetary Cloud and Formation of the Earth and
the Planets (Israel Program for Scientific Translations, 1972).
13. O’Neill, H., St C.., &. Palme, H. Collisional erosion and the non-chondritic
composition of the terrestrial planets. Phil. Trans. R. Soc. A 366, 4205–4238
(2008).
This paper discusses how the compositions of terrestrial planets may change
during the later stages of their accretion.
14. Halliday, A. N. & Porcelli, D. In search of lost planets—the paleocosmochemistry of
the inner solar system. Earth Planet. Sci. Lett. 192, 545–559 (2001).
15. Boyet, M. & Carlson, R. W.
142
Nd evidence for early (.4.53 Ga) global
differentiation of the silicate Earth. Science 309, 576–581 (2005).
Here the
142
Nd/
144
Nd for chondritic meteorites is shown to be 20 parts per
million less than that of most terrestrial rocks.
16. Boyet, M. & Carlson, R. W. A new geochemical model for the Earth’s mantle inferred
from
146
Sm–
142
Nd systematics. Earth Planet. Sci. Lett. 250, 254–268 (2006).
This paper discusses the implications of excess
142
Nd/
144
Nd in terrestrial
samples in terms of a non-chondritic mantle and a hidden reservoir.
17. Kinoshita, N. et al. Geocosmochronometer
146
Sm: a revised half-life value. Mineral.
Mag. (Goldschmidt Conf. Abstr.) 75, 1191 (2011).
18. Qin, L.-P., Carlson, R. W., Alexander, C. M. & O’D.. Correlated nucleosynthetic
isotopic variability in Cr, Sr, Ba, Sm, Nd and Hf in Murchison and QUE 97008.
Geochim. Cosmochim. Acta 75, 7806–7828 (2011).
19. Patchett, P. J., Vervoort, J. D., So
¨
derlund, U. & Salters, V. J. M. Lu–Hf and Sm–Nd
isotopic systematics in chondrites and their constraints on the Lu–Hf properties of
the Earth. Earth Planet. Sci. Lett. 222, 29–41 (2004).
20. Andreasen, R. Sharma, M. Subbarao, K. V. & Viladkar, S. G. Where on Earth is the
enriched Hadean reservoir? Earth Planet. Sci. Lett. 266, 14–28 10.1016/
j.epsl.2007.10.009 (2008).
21. Caro, G., Bourdon, B., Birck, J.-L. & Moorbath, S. High-precision
142
Nd/
144
Nd
measurements in terrestrial rocks: constraints on the early differentiation of the
Earth’s mantle. Geochim. Cosmochim. Acta 70, 164–191 (2006).
22. Labrosse, S., Herlund, J. W. & Coltice, N. A crystallizing dense magma ocean at the
base of the Earth’s mantle. Nature 450, 866–869 (2007).
23. Davies, G. F. Mantle plume, mantle convection and hotspot chemistry. Earth Planet.
Sci. Lett. 99, 94–109 (1990).
24. Loper, D. & Lay, T. The core-mantle boundary region. J. Geophys. Res. 100,
6397–6420 (1995).
25. Korenaga, J. A method to estimate the composition of the bulk silicate Earth in the
presence of a hidden geochemical reservoir. Geochim. Cosmochim. Acta 73,
6952–6964 (2009).
26. Bourdon, B., Touboul, M., Caro, G. & Kleine, T. Early differentiation of the Earth and
the Moon. Phil. Trans. R. Soc. A 366, 4105–4128 (2008).
27. Nyquist, L. E. et al.
146
Sm–
142
Nd formation interval for the lunar mantle. Geochim.
Cosmochim. Acta 59 (13), 2817–2837 (1995).
28. Boyet, M. & Carlson, R. W. A highly depleted moon or a nonmagma ocean origin for
the lunar crust? Earth Planet. Sci. Lett. 262, 505–516 (2007).
29. Brandon, A. D. et al. Re-evaluating
142
Nd/
144
Nd in lunar mare basalts with
implications for the early evolution and bulk Sm/Nd of the Moon. Geochim.
Cosmochim. Acta 73, 6421–6445 (2009).
30. Caro, G. & Bourdon, B. Non-chondritic Sm/Nd ratio in the terrestrial planets:
consequences for the geochemicalevolution of the mantle-crust system. Geochim.
Cosmochim. Acta 74, 3333–3349 (2010).
31. Tonks, W. B. & Melosh, H. J. Magma ocean formation due to giant impacts.
J. Geophys. Res. 98, 5319–5333 (1993).
32. Campbell, I. H. The identification of old mantle plumes. Geol. Soc. Am. Spec. Publ.
352, 5–21 (2001).
33. Tolstikhin, I. N., Kramers, J. D. & Hofmann, A. W. A chemical Earth model withwhole
mantle convection: the importance of a core-mantle boundary layer (D99) and its
early formation. Chem. Geol. 226, 79–99 (2006).
34. Labrosse, S., Hernlund, J. W. & Coltice, N. A crystallizing dense magma ocean at the
base of the Earth’s mantle. Nature 450, 866–869 (2007).
35. Campbell, I. H. & Griffiths, R. W. The changing nature of mantle hotspots through
time: implications for the geochemical evolution of the mantle. J. Geol. 100,
497–523 (1992).
This paper uses the melting products of mantle plumes in conjunction with
knowledge of their dynamics to determine the chemical structure of the mantle
and shows how it has evolved with time.
36. Davies, G. F. Ocean bathymetry and mantle convection. 1. Large-scale flow and
hotspots. J. Geophys. Res. 93, 10467–10480 (1988).
37. Sleep, N. H. Hotspots and mantle plume: some phenomenology. J. Geophys. Res.
95, 6715–6736 (1990).
Table 1
|
The geochemical paradoxes of the mantle
Paradox Chondritic solution Non-chondritic solution
The
142
Nd/
144
Nd ratio of chondritic meteorites is
206 5 parts per million less than that of rocks of
terrestrial mantle origin.
A low-Sm/Nd hidden reservoir became isolated from the
convecting mantle within 10 Myr of the Earth’s formation
15,61
.
The Sm/Nd ratio of the primitive Earth was about
6% above the chondritic value
13
.
Earth’s oldest rocks show evidence of being derived
from a mantle with positive e
Nd
and e
Hf
before the
formation of the first preserved continental crust.
Extensive continental crust formed before the first preserved
continental crust and was recycled through the mantle
62
or
there is a hidden basaltic low-Sm/Nd reservoir
15
.
The Sm/Nd ratio of the primitive Earth was about
6% above the chondritic value
13
.
The Ar concentration in the mantle is about half
the value predicted from the chondritic model
63
.
Only half of the mantle is degassed
63
. The collisional erosion hypothesis
13
predicts a K
content of the mantle appreciably below that
expected from the chondritic model.
Alternatively, the Earth is not chondritic
16
.
Nb/Ta and Nb/La values of both continental crust
and depleted mantle lie below (Nb/Ta) and above
(Nb/La) the primitive mantle values of 17.5 for
Nb/Ta and 0.9 for Nb/La.
Hidden reservoir enriched in Nb, Ta and Nb with super-
chondritic Nb/Ta and sub-chondritic Nb/L a
64
.
The Nb/Ta and La/Nb values of the primitive
mantle lie between those of the depleted mantle
(15.5 and 1.2) and the continental crust (12.5
and 2.2).
4
He production in oceans is less than that
predicted from observed heat flow and about
half that predicted from chondritic Earth model.
4
He stored in lower mantle that is separated by a boundary
layer that transmits heat but not
4
He (ref. 65).
Collisional ero sion model predicts the Th–U
content of the BSE to be about half the chondritic
value
13
.
PERSPECTIVE RESEARCH
29 MARCH 2012 | VOL 483 | NATURE | 557
Macmillan Publishers Limited. All rights reserved
©2012
