of 32
Theory of the Earth
Don L. Anderson
Chapter 8. Chemical Composition 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:
Considerations from cosmochemistry and
the study of meteorites permit us to
place only very broad bounds on the chem
istry of the Earth's interior. These tell
us little about the distribution of elemen
ts in the planet. Seismic data tell us a
little more about the distribution of th
e major elements. General considerations
suggest that the denser major elements will be toward the center of the planet
and the lighter major elements, or those
that readily enter melts or form light
minerals, will be concentrated toward th
e surface. To proceed further we need
detailed chemical information about crusta
l and mantle rocks. The bulk of the
material emerging from the mantle is in
the form of melts, or magmas. It is
therefore important to understand the
chemistry and tectonic setting of the
various kinds of magmatic rocks and the kinds of sources they may have come
from.
Chemical
Composition
of
the
Mantle
These rocks
.
.
.
Shall yet
be
touched with
beau@,
and
reveal
the
secrets
of
the
book
of
earth
to man.
-
ALFRED
NOYES
C
onsiderations from cosmochemistry
and
the study
of
meteorites permit
us
to place only
very broad
bounds
on
the chemistry
of
the
Earth's
interior. These tell
us
little
about
the distribution
of
elements
in
the planet. Seismic
data
tell
us
a little
more
about the distribution
of
the major
elements. General considerations suggest that the denser
major
elements
will
be
toward
the center
of
the planet
and
the lighter major elements, or those that readily enter
melts
or form light minerals, will
be
concentrated
toward
the sur
-
face.
To
proceed
further
we
need
detailed chemical infor
-
mation
about
crustal
and
mantle
rocks. The
bulk
of
the
ma
-
terial emerging from the
mantle
is in
the form
of
melts, or
magmas.
It is therefore important to
understand
the chem
-
istry
and
tectonic setting
of
the various kinds
of
magmatic
rocks
and
the kinds
of
sources they
may
have
come from.
METHODS
OF
ESTIMATING
MANTLE CHEMISTRY
The chemical composition
of
the mantle is one
of
the
most
important
yet
elusive properties
of
our planet. Attempts to
estimate mantle composition fall into
two broad
categories.
Cosmochemical
approaches take meteorites or
mix
-
tures
of
meteoritic material
as
the
basic
building blocks,
and
mixing ratios are adjusted
to satisfy
such
constraints
as
core
size, heat
flow
and
crustal ratios
of
certain elements.
An
example is the six
-
component model
of
Morgan
and Anders
(1980).
Cosmochemical
models
constrain the
bulk
chemis
-
try
of
the Earth rather than that
of
the mantle alone;
never
-
theless
they
provide important
input
into
models
of
the
bulk
chemistry
of
the Earth.
Petrological
models begin with
the reasoning that
since basalts represent melts,
and
peridotites are
thou@
to
be
residues,
some mixture
of
these should approximate the
composition
of
the
upper mantle.
With only two
compo
-
nents this
approach does
not
yield
chondritic ratios for
many
key
elements,
most notably
Si/Mg.
An
alternate ap
-
proach is to search for
the most
"
primitive
"
ultramafic rock
(that
is,
the
one
with most
nearly
chondritic ratios
of
the
refractory elements)
and
attribute
its
composition
to the
whole
mantle.
Unfortunately, even
the
most
"
primitive
7
'
ul-
tramafic nodules are depleted
in
many
of
the trace elements
and have
nonchondritic rare
-
earth ratios,
but
theorizing
has
proceeded
beyond
the
first
crude models.
Both
approaches utilize terrestrial
and
meteoritic data.
The common
theme
is that the Earth should
have
an
unfrac-
tionated chondritic pattern
of
the refractory elements. This
assumption,
justified
by
the observation that these elements
occur
in
roughly constant proportions
in
the
various mete
-
orite classes,
has led
to
the
generally accepted hypothesis
that the refractory elements do
not suffer any
preaccretional
fractionation. This can
be used
as
a formal a priori constraint
in
geochemical modeling of
the composition
of
the Earth.
Some recent estimates
of
the composition
of
the
mantle
are
given in
Table
8
-
1.
Once
they
are
in
a planet, the refractory
elements become fractionated
by
a variety
of
processes.
The refractory siderophiles enter the core,
the
compatible
refractories are retained
in
mantle
silicates
and the
incom
-
patible refractories preferentially enter
melts and the
crust
along with the more
volatile elements.
The volatile elements are fractionated
by
preaccre-
tional processes,
and
they
exhibit a wide range
in
meteor
-
ites.
It is therefore
difficult to
estimate the volatile content
of
the Earth
or to
obtain estimates
of
such
key
volatile-to-
refractory ratios
as
K/U,
Rb/Sr,
Pb/U
and
others. It is often
assumed
that ratios
of
this type are the same in
the
Earth or
in primitive mantle
as they
are
in
the continental crust. The
TABLE
8
-
1
Estimates
of
Average
Composition
of
Mantle
Oxide
(1)
(2)
(3)
(4)
(5)
SiO,
45.23
47.9
44.58
47.3
45.1
A1,03
4.19
3.9
2.43
4.1
3.9
MgO
38.39
34.1
41.18
37.9
38.1
CaO
3.36
3.2
2.08
2.8
3.1
FeO
7.82
8.9
8.27
6.8
7.9
Ti0
,
-
0.20
0.15
0.2
0.2
Cr203
-
0.9
0.41
0.2
0.3
Na,O
-
0.25
0.34
0.5
0.4
K
20
-
0.11
0.2
(0.13)
(1)
Jacobsen
and
others (1984): extrapolation
of
ultramafic and chondritic
trends.
(2)
Morgan
and
Anders (1980): cosrnochemical model.
(3)
Maal~e
and
Steel (1980): extrapolation
of
lherzolite trend.
(4)
20 percent eclogite,
80
percent garnet lherzolite
(Anderson, 1980).
(5)
Ringwood
and Kesson
(1976,
Table
7):
pyrolite adjusted to have
chondritic
CaOlAl,03
ratio and
Ringwood
(1966)
for
K,O.
crust,
of
course,
is
just
one repository
of
the incompatible
elements
and
is
less
than
0.6
percent
of
the mass
of
the
mantle.
The validity
of
this assumption is therefore
not
ob
-
vious
and
needs to
be
tested
by
an
independent approach.
"
Primitive
mantle
"
as
used
here
is the silicate fraction
of
the Earth, prior to differentiation
and removal of
the crust
and any
other parts
of
the present mantle that are the result
of
differentiation,
or
separation, processes.
In
some geo
-
chemical
models
it is
assumed
that large parts
of
the Earth
escaped partial melting,
or melt
removal,
and
are therefore
still
"
primitive.
"
Some petrological
models
assume that
melts being delivered to
the
Earth's
surface are samples
from
previously unprocessed material. I
find
it difficult to
believe
that
any part
of
the Earth
could have
escaped
pro
-
cessing
during
the high
-
temperature accretional process.
"
Primitive mantle,
"
as
used
here,
is a hypothetical material
that
is the
sum
of the
present crust
and
mantle.
"
Primitive
magma
"
is a hypothetical magma, the parent
of
other
mag
-
mas,
which
formed
by
a single
-
stage
melting
process
of
a
parent
rock
and
has not
been
affected by loss
of
material
(crystal fractionation)
prior
to sampling.
The
view
that
there
is a single primitive mantle
magma
type that
leaves behind
a single depleted peridotite, the es
-
sence
of
the pyrolite model, is clearly oversimplified. There
is
increasing evidence that ophiolitic peridotites, for ex
-
ample, are
not
simply
related
to the
overlying basalts. Iso
-
topic
data
on
basalts
and
nodules
show
that there are at least
two major mantle
reservoirs. The identification
of
a com
-
ponent
in
ocean
-
island
tholeiites
and
alkali olivine basalts
that
is
"
enriched,
"
both
chemically
(in
LILs)
and
isotopi
-
cally,
also is
not
adequately
accounted
for in single mantle
reservoir models. I
will
refer
to
this enriched component
as
Q.
The use
of
three components
of
primitive mantle is con
-
ventional:
basalts, ultramafic rocks,
and
continental crust.
The
assumption that the crust
and
depleted mantle are
strictly complementary
and
are together equivalent to the
bulk Earth,
however,
is not
consistent
with
isotopic results.
Basalts cover a
broad
compositional range, from LIL
-
poor
to LIL
-
rich.
"
Large
-
ion
-
lithophile
"
(LIL)
is commonly, al
-
though loosely,
used
to
refer to elements (including small
high
-
charge elements!) that do
not
substitute readily for
magnesium or iron
and
are therefore excluded from olivine
and
orthopyroxene. Some LILs are relatively compatible in
garnet
and
clinopyroxene. One recent proposal is that
most
mantle
magmas
are composed
of
a depleted
MORB
com
-
ponent
and an
enriched component (Q)
with
high potas
-
sium,
LIL,
s7Sr/86Sr,
1MNd/143Nd
and
zosPb/204Pb.
Mid
-
ocean
-
ridge basalt (MORB) represents the
most
uniform
and
voluminous magma
type
and
is an
end
member for LIL
concentrations
and
isotopic ratios. This
is logically
taken
as
one
of
the components
of
the mantle. The
MORB
source
has been
depleted
by
removal
of
a component
-
Q
-
that
must be
rich
in
LIL but
relatively poor
in
Na
and
the
garnet-
clinopyroxene
-
compatible elements (such as
Al,
Ca,
Yb,
Lu
and
Sc).
Kimberlitic magmas
have
the required comple
-
mentary relationship to MORB,
and
I adopt them here as
the
Q
component. Peridotites are the
main
reservoirs for
elements
such
as
magnesium, chromium, cobalt, nickel, os
-
mium and
iridium. The continental crust is
an
important
reservoir
of
potassium, rubidium, barium, lanthanum, ura
-
nium
and
thorium. Thus, each
of
these components
plays
an
essential role in determining the overall chemistry
of
the
primitive mantle.
It is conventional
to
adopt a single
lher-
zolite or harzburgite
as
the dominant silicate portion of
the
mantle.
An
orthopyroxene
-
rich component is also present
in
the mantle
and
is required
if the
MgISi
and
CaIA1
ratios
of
the Earth are
to be
chondritic. Some peridotites appear
to
have been
enriched (metasomatized)
by
a kimberlite
-
like
component.
Figure
8
-
1
shows
representative compositions of
kim-
berlite,
crust,
MORB,
and ultramafic
rock. For
many
re
-
fractory elements kimberlite
and
crust
have
a similar enrich
-
ment
pattern.
However,
the
volatilelrefractory
ratios are
quite different,
as
are ratios involving strontium, hafnium,
titanium, lithium, yttrium, ytterbium
and
lutetium.
Kim-
berlite
and MORB
patterns are
nearly
mirror images for the
refractory elements,
but
this is only approximately true for
MORB
and
crust, especially
for
the HREE,
and
the small
ion
-
high
charge elements.
MORB
and
kimberlite also rep
-
resent extremes
in
their strontium
and
neodymium isotopic
compositions.
An
important development in recent
years has been
the
recognition
of
an
LIL
-
enriched
"
metasomatic
"
component
in
the mantle. The
most
extreme
magmas
from the mantle
(high LIL,
high
LREEIHREE,
high
87Sr/86Sr,
low
'43Nd/
144Nd)
are kimberlites
and
lamproites
(McCulloch
and
others, 1982, 1983).
When
these are
mixed
with a depleted
ALKALI
0
1.5
--
1.0
C
I
FIGURE
8
-
1
Trace
-
element
concentrations
in the
continental
crust
(dots),
con
-
tinental
basalts
and
mid
-
ocean
ridge
basalts
(MORB),
normalized
to average mantle compositions derived
from
a
chondritic
model.
Note
the
complementary
relationship
between
depleted basalts
(MORB)
and
the
other
materials.
MORB
and
continental
tholei-
ites
are
approximately symmetric
about
a
composition
of
7
x
C1.
This suggests that
about
14
percent
of the
Earth
may
be
ba
-
salt.
For
other
estimates,
see
text.
magma
(MORB), the resulting blend can
have
apparently
paradoxical geochemical properties.
For
example, the
hy
-
brid magma can
have
such
high
LaIYb,
Rb/Sr
and
206Pb/
204Pb
ratios that derivation
from
an
enriched source is indi
-
cated,
but
the
'43Nd/144Nd
and
87Sr/86Sr
ratios imply deri
-
vation from
an
ancient depleted reservoir.
Many
ocean-
island, island
-
arc
and
continental
basalts have
these
chas-
acteristics. These
apparently
paradoxical results simply
mean
that ratios do
not
average
as do concentrations.
In order
to estimate the volatile
and
siderophile content
of
the mantle,
we
seek
a linear combination
of
components
that gives chondritic ratios for the refractory elements.
We
can
then
estimate
such key
ratios as
RbISr,
KIU,
and
U/
Pb.
In
essence,
we
replace the
five
basic
building
blocks
of
mantle chemistry
(01,
opx,
cpx,
ga,
and
Q,
or their
high-
and
low
-
pressure equivalents)
with
four composites;
peri-
dotite
(01
.t
opx),
orthopyroxenite (opx
+
ol),
basalt
(cpx
-+
ga),
and
Q.
In
practice,
we
can
use two
different
ultra-
mafic
rocks
(UMR
and
OPX)
with
different
ollopx
ratios,
to decouple the
01
+
opx
contributions. The chemistry
of
the
components (MORB,
UMR,
KIMB,
OPX,
and
crust)
are
given
in
Table
8
-
2.
Having
measurements
of
m
elements
in
n
components
where
m
far exceeds
n,
we
can
find
the
weight
fraction
x,
of
each component, given the concentration
C,
of
the jth
element in the
ith
component, that
yields
chondritic ratios
of
the refractory oxyphile elements.
In
matrix
form,
where
Cj
is
the
chondritic abundance
of
element
j
and
k
is
a dilution
or
enrichment factor,
which
is also to
be
deter
-
mined. The least
-
squares solution is
where
C
T
is the
matrix transpose
of
C,
with
the constraints
Ex,
=
1
and
k
=
constant
(3)
When
x,
and
k
are found, equation 1
gives the mantle con
-
centrations
of
the volatile
and
siderophile elements, ele
-
ments not used
in
the
inversion.
The
mixing
ratios found from equations 2
and
3 are
UMR,
32.6
percent, OPX,
59.8
percent, MORB,
6.7
per
-
cent,
crust,
0.555
percent,
and
Q,
0.11
percent. This is a
model
composition for primitive mantle, that is,
mantle
plus crust. The
Q
component is equivalent
to
a global
layer
3.6
km
thick. The
MORB
component represents
about
25
percent
of
the upper mantle. This solution
is
based
on
18
refractory elements and, relative to carbonaceous chondrite
(Cl)
abundances,
k
=
1.46
k
0.09.
The result is given
in
Table
8
-
3 under
"
Mantle
plus
crust.
"
Concentrations nor
-
malized to
Cl
are also given.
Note
that
the high
-
charge,
small ionic radius elements
(Sc, Ti,
Zr,
Nb,
and
Hf)
have
the
highest
C1
-
normalized ratios.
Taking
all
the
refractory
oxyphile elements into account (23 elements),
the
C1-
normalized enrichment
of
the refractory elements
in
the
mantle
plus
crust
is
1.59
-+
0.26.
If
six
elements (Fe,
S,
Ni,
Co,
P,
and
0)
are
removed
from
C1
to
a core
of
appro
-
priate size
and
density, the
remaining
silicate fraction
will
be
enriched in the oxyphile elements
by
a factor
of
1.48
relative to the starting
C1
composition. This factor
matches
the value for
k
determined from the inversions. Thus
it ap
-
pears that the Earth's mantle can
be
chondritic
in major
-
and
refractory
-
element chemistry
if an
appreciable
amount
of
oxygen
has
entered the core (see Chapter
4).
Table
8
-
3 also compares these results
with
Morgan
and
Anders's
cosmochemically
based
model. This model can
be viewed
as
providing
a first
-
order
correction for
vola
-
tile
-
refractory fractionation
and
inhomogeneous accretion.
Rather than treating each element separately,
Morgan
and
Anders
estimated abundances
of
groups
of
elements: refrac
-
tories, volatiles,
and
so
on.
There is strong support for
un
-
fractionated
behavior
of
refractories prior
to
accretion,
but
the volatile elements are likely
to
be
fractionated.
Both
the
volatile elements
and
the
siderophile elements are
strongly
depleted
in
the crust
-
mantle
system
relative to cosmic
abundances.
In
the pyrolite models, it is
assumed that
primitive
mantle is a
mix
of
basalt
and
peridotite
and
that one
knows
the
average compositions
of
the basaltic
and
ultramafic
components
of
the
mantle. These are
mixed
in
somewhat
arbitrary proportions
and
the
results are comparable
with
150
CHEMICAL COMPOSITION
OF
THE
MANTLE
TABLE
8
-
2
Chemical Composition of Mantle Components (ppm)
Ultramafic
Morgan
MORB
Rocks
KIMB
Crust
Picrite
OPX
and Anders
Li
F
Na
Mg
*
A1
*
Si
*
P
S
C1
K
Ca
*
Sc
Ti
v
Cr
Mn
Fe
*
Co
Ni
Cu
Zn
Ga
Ge
Se
Rb
Sr
Y
zr
Nb
Ag
Cd
In
Sn
Cs
Ba
La
Ce
Nd
Srn
Eu
Tb
Yb
Lu
H
f
Ta
Re
'
0s
+
Ir
+
Au'
Tl
+
Pb
Bi
Th
u
*Results
in percent
'Results
in ppb.
Anderson
(1983a).
TABLE
8
-
3
Chemical Composition of Mantle
(ppm)
Morgan
Upper
Lower
Mantle
Normalized
to
C1
and Anders
UDS
Mantle
Mantle
+
Crust
M&A
C1
Li
F
Na
Mg
*
A1
*
Si
*
P
S
C1
K
Ca
*
Sc
Ti
v
Cr
Mn
Fe
*
Co
Ni
Cu
Zn
Ga
Ge
Se
Rb
Sr
Y
Zr
Nb
Ag
Cd
In
Sn
Cs
Ba
La
Ce
Nd
Sm
Eu
Tb
Yb
Lu
Hf
Ta
Re'
0s
.r
Ir
'
Au
TI'
Pb
Bi
Th
u
*Results
in
percent.
?Results
in
ppb.
Anderson
(1983a).
chondritic
abundances
for some
of
the
major elements.
Tholeiitic
basalts
are
thought
to represent the largest degree
of
partial
melting among common
basalt types
and to most
nearly reflect
the trace
-
element chemistry
of
their mantle
source. Tholeiites,
however,
range
in composition from de
-
pleted
midocean
-
ridge
basalts to
enriched ocean
-
island
ba-
salts (OIB)
and
continental
flood
basalts
(CFB). Enriched
basalts
(alkali
-
olivine, OIB,
CFB) can be modeled
as
mix
-
tures
of
MORB and
an
enriched
(Q)
component that ex
-
perience
varying
degrees
of
crystal fractionation prior
to
eruption.
Ultramafic
rocks
from
the mantle, likewise,
have
a
large compositional range. Some appear to
be
crystalline
residues after basalt extraction, some appear to
be
cumu
-
lates,
and
others appear to
have been
secondarily enriched
in
incompatible elements (metasomatized). This enriched
component
is similar to the Q component
of
basalts. Some
ultramafic rocks have high
Al,O,
and
CaO
contents
and
are
therefore
"
fertile
"
(they
can
yield basalts upon
partial
melt
-
ing), but
they
do
not have
chondritic ratios
of
all
the refrac
-
tory elements.
In the above
calculation I have, in essence, decom
-
posed
the basaltic component
of
the mantle into a depleted
(MORB)
and
LIL
-
enriched
(Q)
component. These can
be
combined
and
compared with
the basaltic component
of
previous two
-
component
models,
such
as
pyrolite,
by
com
-
paring
"
undepleted basalt
"
(MORB
+
1.5
percent
KIMB)
with
the basalts
chosen
by
Ringwood
(Hawaiian
tholeiite or
"
primitive
"
oceanic tholeiite,
KD11).
Thus the present
model's
undepleted
basaltic component
(MORB
+
Q)
has
914 ppm
potassium,
0.06
ppm
uranium,
and
0.27
ppm tho
-
rium.
KDll
has, for these elements, 1400,
0.152,
and
0.454
ppm,
respectively, and
other incompatible elements
are correspondingly high. Refractory ratios
such
as
U/Ca
and
Th/Ca
for
this
pyrolite are
about
50
percent higher
than
chondritic because
of
the
high
LIL content
of
the chosen
basalt.
Hawaiian
tholeiites, the basis
of
another pyrolite
model
(Ringwood,
1975),
have
LIL concentrations
even
higher
than
KD11.
One disadvantage
of
pyrolite
-
type
mod
-
els
is
that the
final
results are controlled
by
the arbitrary
choice
of
components, including basalt. Indeed, the various
pyrolite
models differ
by
an
order
of
magnitude
in,
for
ex
-
ample, the
abundance
of
potassium.
By
and
large,
there
is excellent agreement
with
the
Morgan and Anders
cosmochemical model
(Table
8
-
3).
In
the present
model, the alkalis lithium, potassium, rubidium
and
cesium are
somewhat more
depleted,
as
are volatiles
such
as chlorine,
vanadium and
cadmium. The
Rb/Sr
and
K/U
ratios
are
correspondingly reduced. The elements that
are excessively depleted
(P,
S,
Fe,
Co, Ni, Ge, Se, Ag, Re,
Os, Ir,
and Au)
are plausibly
interpreted
as
residing
in
the
core.
Note
that the chalcophiles are not all depleted.
In
par
-
ticular, lead is
not
depleted relative
to
other volatiles such
as
manganese,
fluorine
and
chlorine,
which
are
unlikely to
be
concentrated
in
the
core, or
to the alkali metals. The less
Li
e
Refroctor~es
Volotiles
700
-
1300
K
(
700
K
V
Cholcophile
A
Siderophile
FIGURE
8
-
2
Abundances
of elements
in
"
primitive
mantle
"
(mantle
+
crust)
relative
to
C1,
derived
by
mixing
mantle components
to obtain
chondritic ratios
of
the
refractory lithophile elements.
volatile chalcophiles (Bi,
Cu,
Zn) are slightly depleted. The
most
depleted siderophiles
and
chalcophiles are those that
have
the highest ionization potential
and
may,
therefore,
have suffered
preaccretional sorting
in
the
nebula. There
is
little support for the conjecture that
the
chalcophiles are
strongly partitioned into the core.
The composition
of
primitive mantle, derived
by
the
above
approach,
is given in Figure 8
-
2.
Some elements are extraordinarily concentrated into
the crust. The
above
results give
the
following proportions
of
the total mantle
-
plus
-
crust inventory
in
the continental
crust; rubidium, 58 percent; cesium,
53
percent; potassium,
46
percent; barium, 37 percent; thorium
and
uranium,
35
percent; bismuth,
34
percent; lead,
32
percent; tantalum,
30
percent; chlorine,
26
percent; lanthanum,
19
percent
and
strontium, 13 percent.
In
addition, the atmospheric
argon-
40 content represents
77
percent
of
the total
produced
by
15
1
ppm
potassium
over
the age
of
the Earth. These results
all
point
toward
an
extensively differentiated Earth
and
ef
-
ficient upward
concentration
of
the incompatible trace ele
-
ments. It is
difficult
to imagine
how
these concentrations
could
be achieved
if the
bulk
of
the mantle is still primitive
or
unfractionated.
If
only the upper mantle provides these
elements
to the
crust,
one would
require
more than
100 per
-
cent
removal
of
most
of
the list
(U,
Th,
Bi, Pb, Ba,
Ta,
K,
Rb,
and
Cs). More
likely,
the whole mantle has contributed
to
crustal,
and
upper
-
mantle, abundances.
The crust,
MORB
reservoir
and
the Q component
account for a large fraction
of
the incompatible trace ele
-
ments. It is likely, therefore, that the
lower
mantle is
de-
pleted
in these elements, including the heat producers
po
-
tassium, uranium
and
thorium.
In
an
alternative approach
we
can replace
UMR and
OPX
by
their primary constituent minerals, olivine, ortho
-
pyroxene,
and
clinopyroxene. The present
mantle
is there
-
fore
viewed
as a
five
-
component system involving
olivine,
orthopyroxene, clinopyroxene,
MORB
(cpx
and
ga),
and
Q.
In
this case the LIL inventory
of
the primitive mantle is
largely contained
in
four components: MORB,
Q,
clinopy-
roxene, and
CRUST.
The results are: olivine,
33.0
percent,
orthopyroxene,
48.7
percent, clinopyroxene,
3.7
percent,
MORB,
14.0
percent,
Q,
.085
percent
and CRUST,
0.555
percent. Concentrations
of
certain
key
elements are so
-
dium, 2994 ppm, potassium,
205
ppm, rubidium,
0.53
ppm, strontium, 25 ppm,
and
cesium,
0.02
ppm.
The al
-
kalis are generally
within
50
percent
of
the concentrations
determined
previously.
Key
ratios are
RbISr,
0.021,
K/U,
4323,
and
KINa,
0.07.
The
Rb/Sr
and
K/Na
ratios are es
-
sentially the same as
those
determined previously; the
KIU
ratio
is 44
percent
lower.
In
summary, a four
-
component (crust, basalt,
perido-
tite
and
Q)
model for the
upper
mantle can
be
derived
that
gives
chondritic ratios for the refractory trace elements.
The model gives predictions for volatilelrefractory ratios
such as
KIU,
RbISr
and
PblU.
An
orthopyroxene
-
rich
component is required
in
order
to match
chondritic ratios
of
the
major elements. Such a component is found
in the
up
-
per
mantle
and
is implied
by
the seismic data for the
lower
mantle. The abundances in the mantle
-
plus
-
crust
system
are
15
1 ppm potassium, 0.0197
ppm
uranium
and
0.0766
ppm
thorium, giving a
steady
-
state
heat
flow
of
0.9
pcallcm2
s.
This implies that slightly
more than
half
of
the terrestrial
heat
flow
is due
to cooling
of
the Earth, consistent
with
convection calculations
in
a stratified Earth
(McKenzie
and
Richter, 1981).
Primitive
mantle
can
be
viewed as
a five
-
component
system; crust, MORB, peridotite, pyroxenite
and
Q
(quin
-
tessence, the
fifth
essence)
or,
alternatively, as
olivine,
or-
thopyroxene, garnet
plus
clinopyroxene (or basalt)
and
in
-
compatible
and
alkali
-
rich material (crust
and
kimberlite or
LIL
-
rich magmas).
THE
UPPER
MANTLE
The mass
-
balance
method
gives
the
average composition
of
the mantle
but makes
no
statement about
how
the compo
-
nents
are distributed
between
the upper
and
lower mantle.
If we
assume that the only
ultramafic
component
of the
up
-
per
mantle is UMR,
we
can estimate the composition
of
the
upper
and
lower mantles and,
as
an
intermediate step, the
composition
of
the
MORB
source
region
prior
to
extraction
of
crust
and
Q.
The
lower mantle
is UMR
plus
OPX.
Ortho-
pyroxenite, the
most
uncertain
and
to
some extent arbitrary
of
the components,
plays
a minor role
in
the mass
-
balance
calculations for the trace refractories
and
is required mainly
to obtain chondritic ratios
of
CafA1
and
MgISi.
My
approach
to
the
upper mantle is similar to the con
-
ventional approach
in
that
I consider a basaltic
and
an
ultra-
mafic
component.
However, instead
of
making an
a
priori
selection
of
basalt,
I
have
decomposed it into a depleted
(MORB)
and an
enriched
(Q)
component. These represent
extremes
in
both
LIL contents
and
isotopic ratios.
For
ex
-
ample, fresh
MORB has
87Sr/86Sr
as
low
as
0.7023
and
kimberlite
usually
has
87Sr186Sr
well
above
0.704;
alkalic
basalts are intermediate in
both
LIL contents
and isotopic
ratios. The
procedure
is as
follows. The
mixing
ratios
of
MORB, crust
and
Q are
known from the previous
section,
and
these ingredients are
assumed to be
entirely contained
in
the primitive upper mantle. The absolute sizes
of
the
crustal
and
upper
-
mantle reservoirs
(above
650
km)
are
known,
so
we
know both
the relative
and
absolute
amounts
of
each component.
As
an
intermediate step,
we
estimate
the composition
of
a possible
picritic
parent
to MORB. The
relation
PICRITE
=
0.75
MORB
+
0.25
UMR
gives
the
results tabulated
under
"
picrite
"
in
Table
8
-
2 and
8
-
4.
The
mixing
ratios
which were
determined to
give
a
chondritic pattern for
the
refractory elements
yield
UDS
=
0.9355PICRITE
+
0.0106Q
+
0.0559CRUST
The composition of
this
reconstructed
Undepleted
Source
Region
is tabulated under
UDS.
The fraction
of
the crustal
component is
about 10 times
the
crust/mantle
ratio, so
UDS
accounts for 10 percent
of
the mantle. The remainder
of
the
upper mantle
is assumed
to
be ultramafic rocks UMR.
This
gives
the composition tabulated
under
"
Upper Mantle
"
in
Tables
8
-
3 and
8
-
4.
This
region
contains 23.4 percent
ba
-
salt (MORB). The resulting upper
mantle has
refractory-
element ratios (Table
8
-
5),
which,
in
general, are
in
agree
-
ment with
chondritic ratios. The
La/Yb,
AlICa
and
SilMg
ratios,
however,
are too high. These are
balanced
by
the
lower
mantle
in
the full calculation. The solution for the
lower
mantle
is
0.145
UMR and
0.855
OPX. This
gives
chondritic ratios for
MgISi
and
CdA1
for the Earth as a
whole.
An
orthopyroxene
-
rich
lower mantle
is expected for
a chondritic
model
for the major elements, particularly
if
the upper mantle is olivine
-
rich.
At
low
pressure olivine
and
orthopyroxene are refractory phases
and
are
left
behind
as
basalt is removed.
However,
at
high
pressure
the
ortho
-
pyroxene
-
rich phases, majorite
and
perovskite,
are both re
-
fractory
and
dense.
If melting during
accretion extended
to
depths greater
than about
350 km,
then the melts would
be
olivine
-
rich
and
separation
of
olivine from orthopyroxene
can
be
expected.
Figure 8
-
3 shows the
concentrations
of
the lithophile
elements
in
the
various components, upper
mantle
and
mantle
-
plus
-
crust, all normalized to the
Morgan
and
Anders
mantle equivalent concentrations. The refractory elements
in
the upper
mantle have
normalized concentrations
of