Theory of the Earth
Don L. Anderson
Chapter 11. Evolution of the Mantle
Boston: Blackwell Scientific Publications, c1989
Copyright transferred to the author September 2, 1998.
You are granted permission for indi
vidual, educational, research and
noncommercial reproduction, distribution, display and performance of this work
in any format.
Recommended citation:
Anderson, Don L. Theory of the Earth.
Boston: Blackwell Scientific Publications,
1989.
http://resolver.calt
ech.edu/CaltechBOOK:1989.001
A scanned image of the entire book may
be found at the following persistent
URL:
http://resolver.caltech.edu/CaltechBook:1989.001
Abstract:
I have now discussed the various kinds of
magmas and refractory rocks that are
important in the mantle, or at least the upper mantle, and some isotopic and
seismic properties of the mantle. The vi
sible rocks are the end, or present,
product of mantle evolution. If these were our only source of information, we
could come up with a fairly simple sch
eme of magma genesis, perhaps involving
single-stage melting of a homogeneous,
even primitive, mantle. It now seems
unlikely that we will find the "Rosetta Ston
e” -- a rock fragment that represents
"original Earth" or even the parent or
grandparent of other rocks. Rocks and
magmas represent products of complex multistage processes, just as the crust
and mantle do. As we delve deeper into
the Earth and further back in time, we
depend more and more on isotopes and
on modeling of planetary accretion and
mantle processes.
Evolution
of
the
Mantle
Rocks, like everything
else,
are
subject
to
change
and so
also are our
views
on
them.
-FRANZ
Y.
LOEWINSON
-
LESSING
I
have
now
discussed the various kinds
of
magmas
and
refractory rocks that are important
in
the mantle, or at
least the upper mantle,
and
some isotopic
and
seismic
prop
-
erties
of
the mantle. The visible rocks are the
end,
or pres
-
ent,
product
of
mantle evolution.
If
these
were
our only
source
of
information,
we
could
come
up with
a fairly sim
-
ple scheme
of
magma
genesis, perhaps
involving
single-
stage melting
of
a homogeneous,
even
primitive, mantle.
It
now
seems unlikely that
we
will
find
the
"
Rosetta Stone
H
-
a
rock
fragment that represents
"
original Earth
"
or
even
the
parent or grandparent
of
other rocks.
Rocks
and
magmas
represent products
of
complex multistage processes,
just
as
the crust
and
mantle do.
As we
delve deeper into the Earth
and
further
back
in
time,
we
depend
more and
more on
isotopes
and
on
modeling
of
planetary accretion and mantle
processes.
A variety
of
studies
have
lent support to the concept
of
a chemically inhomogeneous mantle. The mantle contains
at least
two
distinct reservoirs that differ
in
their
large-ion-
lithophile
(LIL) contents
and that have
maintained their
separate identities for
at least
lo9
years. One reservoir,
itself
heterogeneous,
has high
values
of
Rb/Sr,
Nd/Sm,
U/Pb
and, possibly,
3HelU.
A
variety
of
LIL
-
enriched basalts
ranging from basanites, nephelinites
and
alkali basalts
to
tholeiites either form or
evolve
in,
or become contaminated
by,
this reservoir. There
may
be several
enriched reservoirs
and several
mechanisms
of
enrichment (subduction, trap
-
ping
of
melts
or gas). The simplest hypothesis,
however,
is
that
there is one enriched reservoir
that may
be radially
and
laterally inhomogeneous,
having been
enriched at various
times
and by
various mechanisms. The other major
reser
-
voir
provides fairly uniformly depleted tholeiites
(MORW)
to the
midocean
ridges. The incompatible
-
element patterns
for
MORB
and
enriched or alkali
-
rich
basalts
are comple
-
mentary, and
their ratios parallel the partition
coefficient
trend
of
garnet. This suggests
that
neither reservoir is primi
-
tive
and that they
are complementary products
of
a more
primitive reservoir. The LIL
-
depleted reservoir appears
to
be
garnet
-
rich, or fertile, relative to the enriched reservoir.
The high density
of
garnet suggests that
the
depleted reser
-
voir
is the deeper
one,
and that ancient
depletionlenrich-
ment
processes
involved
the
upward
transport
of
fluids that
have been
in
equilibrium
with
garnet. The recycling
of
sed
-
iments
and
hydrothermally
altered
oceanic crust,
too,
has
probably
been
important throughout
most
of
Earth history.
This material
may
contribute to the chemical
and
isotopic
heterogeneity
of
the enriched source region,
or
it may
be
preferentially concentrated
in
the shallow mantle and
act
as
a separate source.
From a geophysical
and
tectonic point
of
view
the fol
-
lowing possible sources
can be
identified. Some
may
be
fer
-
tile sources
of
basalts, while others
may
simply interact
with
magmas
generated elsewhere.
1.
The
continental
shield
lithosphere
2.
The shallow mantle, sometimes called the asthenosphere.
3.
The transition region.
4.
Subducted sediments.
5.
Subducted oceanic crust.
In addition,
magmas
may
interact
with the
continental
crust, the oceanic crust, the oceanic lithosphere, seawater
and
the atmosphere.
Until
all
of
these sources are evaluated,
it seems
premature
to
invoke such remote sources as
the
lower mantle
and
D
"
.
Most models
of
petrogenesis assume that
basic and
ultra-
basic
magmas
are formed
by
varying degrees
of
partial
melting
of
fertile peridotite
in
the upper mantle. In these
models
the degree
of
partial melting
and
crystal fraction
-
ation
and
the depth
of
origin are
the main
variables in con
-
trolling the composition
of
the melt.
Variations
in
volatile
content
and
mineralogy
of
the source region,
and
crustal
contamination are additional variables
that have been
used
to explain the full range
of
magma compositions. Tholeiites
are generally regarded as representing large degrees
of
par
-
tial melting
of
a shallow
source,
and the more
K20-rich
magmas,
such
as
alkali basalts
and
nephelinites, are
re
-
garded as resulting from relatively small amounts
of
melting
at
deeper levels. The incompatible
-
element
-
enriched mag
-
mas with
high LREE content,
such
as melilites,
nephelin-
ites, basanites
and
kimberlites, are
assumed to
result from
extremely small degrees
of
partial melting from a deep
garnet
-
rich source region.
Isotopic studies require at least
two
source regions
in
the mantle. The
most
voluminous
magma
type,
MORB has
low
LIL contents and
low
87Sr/86Sr,
143Nd/144Nd
and
207Pb/
204Pb
ratios. The
Rb/Sr,
NdISm
and
UIPb
ratios
in
this res
-
ervoir
have
been low
for at least
2.5
Ga,
and
perhaps
were
set
by
accretional differentiation. Since these ratios are
high
in
melts and
low
in residual crystals,
the
implication is that
the
MORB
reservoir is a cumulate or a crystalline residue
remaining after the
removal
of
a melt fraction. The enriched
melt
fraction is commonly assumed to
efficiently
enter the
continental crust. The continental crust is therefore re
-
garded as the only complement to the depleted
MORB
source. Mass
-
balance calculations
based on
this
premise in
-
dicate that
most
of
the mantle is undepleted or primitive,
and.the
depleted reservoir is assumed
to
occupy
most
or
all
of
the upper mantle. Since the continental crust is the only
enriched reservoir
in
this model, magmas that
have high
LIL contents
and
87Sr/86Sr
ratios are assumed
to be
con
-
taminated
by
the continental crust, or
to
contain a recycled
ancient crustal component.
However,
enriched magmas
such
as
alkali
-
olivine
ba-
salts (AOB), oceanic
-
island basalts
(OIB) and
nephelinites
also occur
on
oceanic islands
and have
similar LIL
and
iso
-
topic ratios
to continental
flood
basalts (CFB) and
continen
-
tal alkali basalts. Continental contamination is
unlikely
in
these cases.
Veins
in mantle peridotites
and
xenoliths con
-
tained in alkali
basalts and
kimberlites are also commonly
enriched and, again, crustal contamination is
unlikely. In
many
respects these enriched magmas and xenoliths are
also complementary to
MORB (in LIL contents
and
isotopic
ratios), suggesting that there
is ancient enriched material in
the mantle. Island
-
arc basalts are also high
in LIL,
87Sr/86Sr,
143Nd/144Nd
and
206Pb/2"Pb,
suggesting that there is a shal
-
low and
global enriched reservoir. Back
-
arc
basin basalts
(BABB) are closer
to
MORB
in composition and,
if
the
depth
of
the
low
-
velocity
zone
and
the
depths
of
earth
-
quakes can be
used as
a guide, tap a source deeper
than
150
km.
Many
BABBs
are intermediate
in
chemistry
to
MORB
and
OIB. This
and
other evidence indicates
that
the en
-
riched reservoir is shallow
and
the depleted
MORB
reser
-
voir
is deeper.
Since the seismic velocities
and
anisotropy
of
the
shal
-
low
mantle,
above
220
km,
are consistent
with an
olivine
-
rich aggregate,
and
since
most mantle xenoliths are olivine
-
rich,
it is natural
to
assume
that
the enriched reservoir
is a
peridotite. This peridotite
can be
infertile
(low
in
Al2O3)
yet
enriched in LIL. Since the depleted reservoir
has
already
lost a
melt
fraction, it
should be
depleted
in
garnet,
unless
it is an eclogite cumulate. Peridotites depleted
in
basalt
have
less
A120,
and
garnet
than
nondepleted or fertile
peridotites.
The trace
-
element signatures of
enriched
magmas
are
consistent
with
derivation from a reservoir that
has
experi
-
enced eclogite fractionation or metasomatism
by
melts
from
an
eclogite
-
rich source.
Magmas
from
this
reservoir
will
therefore
be
LREE
-
enriched,
not
because there is garnet
in
the residue but because
the
reservoir
itself
had
experienced
a prior stage
of
garnet
removal
or secondary enrichment
by
a fluid
from a garnet
-
rich reservoir. This eliminates one
of
the
arguments for derivation
of
LIL
-
and
LREE
-
enriched
magmas,
such
as
alkali
olivine basalts,
by
small degrees
of
partial melting from a deep garnet
-
rich peridotite layer.
In
general, diapirs from great depth will experience a greater
degree
of
partial melting
than
diapirs originating at
shallow
depths. This is another argument against a
shallow source
for tholeiites. Extensive
melting
probably requires adiabatic
ascent from a deep thermal boundary layer.
Kimberlites are
among
the
most
enriched magmas.
Al
-
though
they
are rare, the identification
of
a kimberlite
-
like
component
in
enriched
magmas means that they
may
be
volumetrically more important
than
generally appreciated.
The eclogite
-
fractionation,
magma
-
mixing
hypothesis
for the evolution
of
the
mantle
(Anderson,
1982a,b,c)
dif
-
fers from conventional petrological assumptions in
several
ways:
(1)
The
depleted
source
region is
an
eclogite
-
rich
cumulate
or
a piclogite rather
than
a garnet peridotite;
(2) enriched
magmas
are
blends
or hybrids
of
a depleted
magma (MORB
or
picrite)
and
a melt
from a
shallow
en
-
riched peridotite reservoir;
(3)
the LIL pattern
of
enriched
magmas,
in
particular LREE enrichment, is
due
to
small
degrees
of
partial melting
of
a garnet peridotite, or
garnet-
clinopyroxene fractionation from a
picritic
or
tholeiitic
magma. This pattern
may
also
have been
introduced into
the source at some earlier time
by
metasomatic fluids.
There is general consensus
that
the depleted,
or
MORB, reservoir is depleted
in
the incompatible elements
such
as rubidium, neodymium, uranium
and
LREEs
and
that the very incompatible elements are depleted
more than
MODELS
OF
THE
MANTLE
PETROGENESIS
the less incompatible elements, thereby giving
low
RbISr,
NdlSm
and
UIPb
ratios. The corresponding isotopic ratios
indicate that the depletion
event was
ancient, perhaps dating
back
to the formation
of
the continental crust. The com
-
plementary enrichment events
would
likewise
have
been
ancient.
We
use
"
depleted
7
' to describe basalts
and
reser
-
voirs that have
low
LIL contents
and
low
LREEIHREE,
RbISr,
NdISm,
U/Pb,
s7Sr/86Sr,
143Nd/L44Nd,
206Pb/204Pb
and
other ratios. A depleted reservoir can still
be
fertile,
as it can provide basalts
by
partial melting. A garnet
-
clinopyroxenite cumulate, for example, can
be
depleted but
fertile. Similarly,
an
enriched reservoir can
be
infertile,
being
low
in
CaO,
A1203,
Na20
and
so
on.
What are
not agreed upon
are
the
following:
1.
The location
of the MORB
reservoir. It
has
been
vari
-
ously
placed within
the
low
-
velocity
zone,
at the
bottom
of
the low
-
velocity zone,
and
in the transition region,
that
is,
anywhere
between
220
lun
and
650
km.
The ho
-
mogeneity
of
MORB,
on
a global basis, the arguments
that
attribute it to large degrees
of
partial melting, the
tendency
of
enriched interstitial
fluids
to migrate
up
-
ward,
and
the possible contamination
of
the
shallow
mantle
by
subducted sediments
and
altered oceanic crust
all argue for a relatively deep origin for the
MORB
res
-
ervoir. The time sequence
of
erupted magmas
at
island
arcs, continental rifts
and
oceanic islands
is
consistent
with a
shallow
enriched reservoir
and
a deeper
LIL-
depleted reservoir.
2.
The composition
of
the
MORB
reservoir. The conven
-
tional petrological
view
is that
MORBs
result from large
degrees
of
partial
melting
of
a garnet peridotite. It
now
appears that midocean
-
ridge tholeiites are
not
primitive
magmas but are the result
of
extensive crystal fraction
-
ation
of
a more
picritic
parental
melt
of
a nonchondritic
reservoir.
Picrites
imply
that
the parent peridotite
be
melted
by 30 percent or more. The
high
degree
of
partial
melting
is required
in order to generate high
-
MgO
picri-
tic magmas at
shallow
depths. The alternative point
of
view
is that
the
depleted source
region
is a deep garnet
-
rich pyroxenite cumulate,
such
as
a piclogite (olivine
eclogite).
3.
The nature
of
the depletion process.
All workers
agree
that
the source
region of
midocean
-
ridge tholeiites
is de
-
pleted
in
the LIL elements
and
that this
depletion
was an
ancient event.
In
the conventional
view
this depletion
was
the result
of
a small
amount
of
partial melting
and
melt
removal to
the continental crust. Since the
MORB
source is obviously still fertile,
in
the sense that
it can
provide basalts, it
must
be
garnet
-
rich in spite of having
lost a basaltic component. The
MORB
source, in
this
model,
must have
remained homogeneous and
must
have
escaped
the
early
high
temperatures implied
by
thermal history calculations. Alternatively, the depletion
of
the
MORB
source
in
a cooling Earth
could be
ex
-
plained
if it is
a garnet
-
pyroxenite cumulate. The
late-
stage residual
fluids
in
equilibrium
with
such a
layer
would be LIL
-
and
LREE
-
enriched and,
in
this
respect,
kimberlitic. Such fluids,
if
buoyant, could
form
the
proto
-
crust
.
Fluids
and
melts
are LIL
-
enriched,
and
they tend to
migrate upward. Sediments
and altered
ocean crust, also
LIL
-
enriched, re
-
enter the upper
mantle
at subduction
zones. Thus there are several reasons to
believe
that the
shallow
mantle
serves
as
a scavenger
of
incompatible ele
-
ments, including the radioactive elements (U, Th
and
K)
and
the
key
tracers (Rb,
Sr,
Nd, Sm and, possibly, Pb
and
3He).
The continental crust
and
lithosphere are
commonly
assumed
to be
the
main
repositories
of
the incompatible ele
-
ments, but oceanic islands, island arcs
and
deep
-
seated
kimberlites also
bring
LIL
-
enriched material to the surface.
This
is
one
reason
for
invoking
a
global upper
-
mantle en
-
riched layer
and
for investigating reservoirs other
than
the
continental crust
in
compiling
inventories
of
the incompat
-
ible elements. Even a moderate amount
of
LIL
in
the upper
mantle will destroy the arguments for a primitive
lower
mantle.
Recent metasomatism
of
an
upper
-
mantle source
re
-
gion is sometimes
invoked
to explain LIL
-
enriched
mag
-
mas, particularly those
with
low
87Sr/86Sr
ratios. The source
of the enriching
fluid
is
seldom addressed. Recycling of
sediments or remelting
of
subducted oceanic crust
may
ex
-
plain some enriched magmas. These mechanisms do
not
ex
-
plain
the
high
3He/4He
ratios found in
some
oceanic
-
island
basalts.
The continuum
in
trace elements
and
isotopic ratios
between such
enriched magmas as
nephelinites
and
alkali
-
olivine basalts
and
depleted magmas such
as MORB
sug
-
gests that
most
mantle basalts represent mixtures
between
MORB, or its
picritic
parent,
and
partial
melts
from a
shal
-
low
enriched reservoir.
Mixing
or
contamination
upon as
-
cent is an alternative to recent metasomatism
of
the
source
reservoir.
The study
of
basalts combined
with
experimental
pe
-
trology
allow
one
to
estimate the temperatures
and pres
-
sures
at which
basalts
might have
separated from their
source region. The amount
of
melting
involved and the
composition
of the
source region,
however,
cannot
be
de
-
termined
by
these means. The emphasis on olivine
-
rich
and
peridotite source
regions
is
based
on the
following
arguments:
1.
Peridotite is consistent
with
seismic velocities for
the
shallow mantle,
and basalts
come from
the
mantle. The
seismic arguments for
an
olivine
-
rich
shallow mantle
in
-
clude
both
the velocities
and
the
anisotropy.
However,
peridotites rarely
have
either the major
-
element or
trace-
element chemistry necessary
to provide basalts
of
the
required composition.
Most
of
these
rocks are
refractory
220
EVOLUTION
OF
THE
MANTLE
NNAG:
VSV
lat=
30,
lon=
-
30,
az=
155
FIGURE
11
-
1
Seismic shear velocities
from
50
to
550
km
depth (bottom) along great
-
circle
path
shown in
center of top panel. Low velocities are
open circles, fast
velocities
are
filled
circles.
Verti
-
cal exaggeration is
24:
1
in
bottom panel and
4:
1
in center
panel.
Note the slow velocities
in
the
North
and
South Atlantic, where
the cross section
cuts across
the
Mid
-
Atlantic
Ridge.
Dashed lines are plate boundaries. Circles in upper panel are
hotspots.
(Model
from
Nataf and
others,
1986.)
residuals or cumulates.
In
regions
of
very
low
seismic
velocity, such
as
under tectonic regions
and
midocean
ridges, the velocities are
much
lower
than
in
peridotites.
Partial
melting
or high
-
temperature grain boundary
relaxation
can
reduce the velocities considerably, but
when this
occurs
one
can
have
almost
any
major
-
element
chemistry for the matrix; that is, seismic velocities are
no
longer a constraint. The
slow
velocities associated
with
midocean
-
ridge upper
mantle
now
appear to extend
much
deeper
than
200
km,
and,
in
some
cases, into
the
transition
region
(Figure
11
-
1).
Velocities
in the transi
-
tion region
do
not
seem to
be
appropriate for peridotites.
Thus, although arguments can
be made
that some parts
of
the
shallow mantle
have
seismic velocities appropri
-
ate for peridotite, the connection
between
these regions
and
the source regions for basalts
has not
been made.
2.
The occurrences
of
garnet peridotite are appropriate to
deep
-
seated environments. Garnet peridotite is stable
in
the upper
mantle
but, again, this does
not
prove that
it
occurs in the basalt source region or is the immediate
parent for basalts. Most natural peridotites, in fact,
are
buoyant
relative to fertile peridotites, eclogites
and
piclogite.
3.
C3arne.t
peridotites
have
close compositional relation
-
ships to meteorites. This is
an
argument that the average
Earth
or
even
the
bulk
composition
of
the mantle can
be
inferred from meteoritic abundances. Midocean
-
ridge
PETROLOGICAL
EVOLUTION
OF
THE
MANTLE
221
basalts
clearly
do
not
come from a chondritic source; the
LIL elements are depleted
in
MORB
relative to
chon-
dritic abundances. The
MORB
source is at least one
generation
removed
from a chondritic ancestor. This ar
-
gument,
and many
like
it,
confuse
what
may
be
in
the
mantle
or
what
may
be
the
average
composition
of
the mantle
with what
is actually required
of
the imme
-
diate parent
of
basalts.
A
single-stage
basalt forming
process is
implied whereas
isotopic
and
trace
-
element
data clearly require a multistage process.
Partial
melts of
natural samples
of
garnet peridotite
at
high
pressure
have
basaltic compositions. Actually, par
-
tial melting of
natural peridotites
has not provided
mag
-
mas with
MORB
compositions, particularly in the trace
elements. Synthetic peridotitic aggregates (minus
ojiv-
ine) come close
to
matching
inferred
compositions
of
parental picrites, except for
K,O
(Green and
others,
1979). Melting
of
eclogites,
of
course, also
gives
basal
-
tic composition melts, so the
above
argument, at best,
is
permissive rather
than
persuasive.
Melting
of
eclogites
would have
to be very extensive,
and
melt
-
crystal segregation
would
occur before
su~ch
extensive
melting
can
be
achieved. Actually, eclogites
provide basaltic melts
over
a wide range
of
melting,
and
large amounts
of
melting
may
be
required
for
melt
separation.
There is
no
doubt
that
garnet peridotites
can and
do
come from the
shallow
mantle,
and
some regions
of
the
upper
mantle
are probably
mostly
garnet peridotite. Some
garnet peridotites can
provide
basalts
by
partial melting,
as
can pyrolite
by
definition. The
average
composition
of
the
Earth
is
probably
close
to
chondritic
in
major-element
chemistry,
and
the mantle therefore contains abundant, al
-
though not
necessarily predominant, olivine.
By
the same
reasoning
the
mantle
contains
even more
pyroxene and gar
-
net. These arguments
do
not
prove that
the source
region
of
the
most
abundant basalt type, midocean
-
ridge basalts,
is
garnet peridotite
or that
the regions
of
the mantle that ap
-
pear to be
peridotitic,
on
the basis
of
seismic velocities, are
the regions where
midocean
basalts are generated.
Although some
of
the older ideas about the source re
-
gions,
such
as
melting
of
a glassy
or basaltic
shallow
source,
can be ruled
out,
the possibility that a
deep
eclo-
gite,
or
picritic
eclogite,
is the basalt source region cannot
be
ruled
out. Olivine eclogite, or piclogite, the inferred
composition
of
the transition region, is also a candidate
source rock.
PETROLOGICAL
EVOLUTION
OF
THE MANTLE
Most models
of
petrogenesis assume
that
so
-
called primary
magmas are
the result
of
varying
degrees
of
partial melting
of
peridotite.
Abyssal
tholeiites, because
of
their relative
depletions
in
the LIL
elements,
have
had
a more
complex
history, although it seems likely that partial melting
of
a
peridotite
was involved
at
an
early stage
of
their evolution.
The trace
-
element inhomogeneity
of
the mantle (Chapter
13)
plus
the
long
-
term
isolation
of
the major reservoirs sug
-
gests that differentiation
has been
more effective
in
the
long
run than
mixing. Mixing can
be
avoided
in
a chemically
stratified mantle
if the layers
have
a large intrinsic density
contrast. Garnet
has
the highest density
of
any abundant
upper
-
mantle mineral and therefore
plays
a role
in
deter
-
mining
the density
of
various
regions
of
the mantle.
The conventional
model
for the origin
of
magmas
might
well
be
designated
PM
3
for
"
primitive mantle partial
melt pyrolite model.
"
In
this model,
tholeiitic
basalts are
considered to
be
primary, unfractionated
melts
resulting
from about
20
percent melting
of
fertile garnet peridotite.
In
recent variations, oceanic tholeiites are treated
as
partial
melts from a reservoir that
has been
depleted
by
removal
of
the continental crust. Continental
and
oceanic
-
island
ba
-
salts
(OIB)
are
assumed
to be melts from
a
primitive undif
-
ferentiated lower mantle. The oceanic lithosphere
is
mod
-
eled
as
6
km
of
basalt,
the
primary melt,
and
24
km
of
depleted residual harzburgite (pyrolite
=
1
part basalt, 4
parts depleted peridotite).
Upon
subduction, the lithosphere
sinks to the core
-
mantle boundary (Ringwood, 1975).
Ex
-
cept
for a
layer
of
depleted peridotite
in
the upper mantle,
with
perhaps some isolated blobs
of
eclogites, the mantle is
uniform
in
composition
and
composed
of
pyrolite, which,
by
definition, can
yield
basaltic magmas
by
partial melting.
There are several problems
with
this model. There is
increasing evidence that midocean
-
ridge tholeiites are
not
primary magmas
but
are the result
of
olivine fractionation
from a more
MgO-rich
picritic
parent.
The
harzburgite resi
-
due after
removal
of
partial melt from a garnet peridotite
is
less dense
than
the parent
and will
remain in the upper
mantle. Several
billion years
of
seafloor
spreading
will
fill
up
the entire upper
mantle
with this depleted residue.
In
addition, harzburgites do not appear to
have been
in equilib
-
rium
with
MORB
(Green and
others, 1979).
The
Moon
differentiated
early
in
its
history
and
evolved
into a series
of
cumulate layers. It
is likely that the
Earth did
as
well.
Even
if
a large fraction
of
the heat
of
accretion
was
radiated
away
or
was
convected
efficiently
to
the surface,
it is difficult
to construct geotherms that remain
below
the
solidus
of
silicates during
most of the
accretion
of
the
Earth
(Chapter
1).
Gravitational separation
of
the
melt
will
concentrate the
LIL
elements and
Al,O,,
CaO
and
SiO,
into the upper mantle. In contrast to the conventional
model, the
"
primitive mantle,
"
which
can
yield
basalt
by
partial melting, has
already been processed
into a depleted
lower
mantle
and an
upper mantle that
on
average is en
-
riched. This processing occurs near the surface where
ma
-
terial delivered
by
accretion,
or
by
whole
-
mantle convec
-
tion, crosses the melting zone.
If olivine
and
orthopyroxene
are the
main
residual phases, the incompatible
-
element
en-