of 12
Chapter
25
Crystallization
of
the
mantle
Rocks,
like
everything
else,
are
subject
to
change and
so
also
are
our
views
on
them.
Franz
Y
Loewinson-
L
essing
The
Earth
is
cooling
and
crystallizing.
The
man-
tle
has
evolved
considerably
from
the
magma
ocean
era
to
the
plate
tectonic
era.
Part
of
the
evolution
is
due
to
igneous
processes,
and
part
is
due
to
plate
tectonic
processes.
The
visible
rocks
are
the
end,
or
present,
product
of
man-
tle
evolution.
If
these
were
our
only
source
of
information,
we
could
come
up
with
a
fairly
simple
scheme
of
magma
genesis,
perhaps
involv-
ing
single-stage
melting
of
a
homogenous,
even
primitive,
mantle.
We
could
design
simple
one-
and
two-layer
mantle
models,
such
as
were
pop-
ular
in
the
last
century.
It
now
seems
unlikely
that
we
will
find
the
'Rosetta
Stone,'
a
rock
or
a
meteorite
fragment
that
represents
'original
Earth'
or
even
the
parent
or
grandparent
of
other
rocks.
Rocks
and
magmas
represent
products
of
complex
multistage
processes,
and
they
are
mix-
tures
of
components
with
various
melting
his-
tories.
As
we
delve
deeper
into
the
Earth
and
further
back
in
time,
we
depend
more
and
more
on
isotopes
and
on
modeling
of
planetary
accre-
tion
and
mantle
processes.
Melts
and
rocks
are
averages
of
various
components
and
processes.
Sampling
theory
and
the
manipulation
of
aver-
ages
are
involved
in
this
modeling.
In
addition
to
recycling,
and
an
intrinsically
heterogenous
mantle,
there
are
igneous
processes
that
cause
chemical
heterogeneity
interior
to
the
mantle.
Melt
trapping
and
melt
transport
can
create
distinctive
chemical
and
isotopic
components.
Chemical
heterogeneity
can
be
formed
internal
to
the
mantle.
Components
vs.
reservoirs
revisited
A
variety
of
studies
have
lent
support
to
the
con-
cept
of
a
ch
e
mi
ca
lly
inh
o mo g e
nou
s
m
an-
tl
e .
The
mantle
contains
a
variety
of
components
that
differ
in
major
elements,
intrinsic
density,
melting
point,
large-ion-lithophile
(LIL)-contents
and
isotope
ratios;
they
have
maintained
their
separate
identities
for
at
least
10
9
years.
These
components
have
been
termed
reservoirs
but
only
very
loose
bounds
can
be
placed
on
the
sizes
of
these
so-called
reservoirs.
They
could
be
grain-size,
or
slab-size,
depending
on
the
ability
of
chem-
ical
species
to
migrate
from
one
to
another
. A
component
can
have
dimensions
of
tens
of
ldlo-
meters,
a
typical
scale
for
subducted
or
delam-
inated
assemblages.
The
melting
process
gath-
ers
together
these
components,
and
partial
melts
therefrom
.
Some
components
or
reservoirs
have
high
values
of
Rb/Sr,
Nd/Sm,
U/Pb
and
3
He
jU.
These
components
are
enriched,
but
for
historic
reasons
they
are
sometimes
called
primitiv
e,
und
e-
gassed
or
more-primitive
(than
MORE).
Lithology
of
the
upper
mantle
Since
the
seismic
velocities
and
anisotropy
of
the
shallow
mantle
are
consistent
with
an
olivine-
rich
aggregate,
and
since
most
mantle
xenoliths
are
olivine-rich,
it
has
been
natural
to
assume
that
the
shallow
mantle,
and
the
source
region
for
various
basalts,
including
OIB
and
NMORB,
are
the
same
and
is
a
peridotite.
Since
the
LIL-
depleted
components
in
the
mantle
have
already
lost
a
melt
fraction,
they
should
be
depleted
in
garnet
-
and
therefore
infertile
-
unless
they
are
eclogite
cumulates,
or
delaminated
lower
conti-
nental
crust.
Peridotites
depleted
in
basalt
have
less
Al
2
0
3
and
garnet
than
nondepleted
or
fer-
tile
peridotites.
But
the
depletion
event
may
have
involved
a
very
small
melt
fraction,
in
which
case
the
incompatible
trace
elements
will
be
affected
more
than
the
major
elements.
The
traditional
emphasis
on
homogenous
olivine-rich
and
peridotite
source
regions
for
mantle
magmas
is
based
on
the
following
arguments.
(1)
Peridotite
is
consistent
with
seismic
ve
l
ocities
for
the
shallow
mantle;
basalts
come
from
the
mantle
;
the
upper
mantle
is
therefore
mainly
peridotite
.
(2)
Garnet
peridotite
is
stable
in
the
upper
man-
tle.
(3)
Garnet
peridotites
have
close
compositional
relationships
to
meteorites
.
(4)
Partial
melts
of
natural
samples
of
garnet
peridotite
at
high
pressure
have
basaltic
com-
positions.
(5)
Eclogites
are
the
high-pressure
chemical
equivalent
of
basalt
and
partial
melting
of
eclogite
does
not
recreate
the
composition
of
basalt.
(6)
Melting
of
eclogites
would
have
to
be
very
extensive,
and
melt-crystal
segregation
would
occur
before
such
extensive
melting
can
be
achieved.
These
arguments
are
all
suggestive
rather
than
definitive.
They
do
not
rule
out
other
lithologies
for
the
upper
mantle
or
for
the
source
regions
of
at
least
some
basalts
.
There
is
increasing
evi-
dence
that
large
parts
of
the
upper
mantle
are
eclogitic
or
composed
of
garnet
pyroxenite
and
are
therefore
more
fertile
than
most
peridotites.
Most
of
the
seismic
information
about
the
upper
mantle
is
derived
from
seismic
waves
with
wave-
lengths
from
20
to
300
km.
Much
of
the
petrologi-
cal
information
comes
from
midocean
ridges
and
large
volcanoes,
which
sample
comparable
size
COMPONEN
TS
VS.
RE
SE
R V O I RS
REVISITED
319
regions
of
the
upper
mantle.
This
averaging
effect
of
geophysica
l
and
geochemical
data
can
distort
views
regarding
homogeneity
of
the
mantle
.
But,
t h
ere
is
no
doubt
that
peridotites
can
and
do
co
m e
from
the
sha
ll
ow
mantle;
some
regions
of
the
upper
mant
le
and
lithosphere
are
proba-
bl
y
most
ly
peridotite.
Eve
n
if
basalts
derive
from
eclogitic
regions
of
the
mant
le
they
traverse
and
evolve
in
peridotitic
surroundings.
The
average
compos
i
tion
of
the
Earth
is
proba-
b ly
close
to
ordinary
chondrites
or
enstatite
mete
-
orites
in
major-element
chemistry
,
the
mantle
therefore
contains
abundant
,
although
not
neces-
sarily
predominant,
olivine
.
By
the
same
reason-
ing
the
mantle
contains
even
more
pyroxene
plus
garnet.
The
above
arguments
do
not
prove
that
the
source
region
of
the
most
abundant
basalt
types,
is
garnet
peridotite
or
that
the
regions
of
the
mantle
that
appear
to
be
peridotitic
,
on
the
basis
of
seismic
velocities,
are
the
regions
where
midocean
basalts
are
generated.
Although
some
of
the
older
ideas
about
source
regions
,
such
as
a
glassy
or
basaltic
shallow
source
,
can
be
ruled
out,
the
possibility
that
basalts
involve
eclogite,
pyroxenite,
recycled
crust
or
cumulates,
cannot
be
ruled
out.
Eclogite,
garnet
pyroxenite,
peridotite-
eclogite
mixtures,
or
piclogite,
are
also
candi-
date
source
'rocks';
the
'grains'
in
such
'rocks'
can
be
tens
of
km
in
extent.
Hand-specimen-sized
rocks
are
a
different
scale
from
what
volcanoes
and
seismic
waves
see.
The
trace-element
inhomogeneity
of
the
man-
tle
plus
the
long
-
term
isolation
of
the
vari-
ous
components
suggests
that
differentiation
has
been
more
effective
in
the
long
run
than
mixing.
Mixing
can
be
avoided
in
a
chemically
inhomoge-
nous
-
or
chemically
stratified
-
mantle
if
the
components
are
large
chunks
and/or
have
large
intrinsic
density
and
viscosity
contrasts.
Garnet
has
the
highest
density
of
any
abundant
upper-
mantle
mineral
and
therefore
plays
a
role
in
determining
the
density
of
various
components,
and
regions,
of
the
mantle.
However
,
the
chemi-
cal
heterogeneity
may
also
be
dispersed
through-
out
the
upper
mantle.
Eclogites
come
in
a
variety
of
compositions
and
densities;
they
all
have
low
melting
points
compared
to
peridotites.
Some
eclogites
have
densities
similar
to
some
upper
mant
le
peridotites;
the
density
of
eclogite
is
very
320
CRYSTALLIZATION
OF
THE
MANTLE
Table
25
. 1
I
Effect
of
eclogite
and
olivine
fractionation
on
primitive
magma
Magma
Si0
2
AI
20 3
FeO
MgO
CaO
Ti0
2
Na
20
K20
I
. Primitive
46.2
11.1
10.8
20.2
9.4
0.77
1.06
0.
08
2.
Extract
46.2
13.9
9.3
1
6.3
11.9
0.
81
1.29
0.02
3.
Picrite
46.2
8.3
12.3
24.
1
6.9
0.74
0.
83
0.
14
Tholeiites
4.
Model
50
.0
13.8
12.4
8.5
I I
.5
1.23
1.38
0.23
5.
Hawaiian
50
.0
14.1
I
1.4
8.6
I
0.4
2.53
2
.1
6
0.
39
6.
Continenta
l
50
.6
13
.6
10
.0
8.5
1
0.0
1.
95
2.90
0.54
7.
Average
oceanic
50.7
15.6
9.9
7.7
I
1.4
1.49
2.66
0.17
1 .
Possib
le
primitiv
e
ma
g
ma.
Th
e
partial
melt
product
of
primitive
mantl
e
diff
erent
iation
(O'Hara
and
others,
1975).
2.
Eclog
it
e
extract
(O'Hara
and
others,
1975).
3.
Residual
liquid
after
50
percent
eclogite
(2)
r
emova
l
from
primitive
magma
(1).
This
is a
mod
el
picritic
primary
magma.
4.
Residua
l
liquid
after
a
furth
er
remova
l
of
40
pe
rcent
oliv
in
e
(Fo
8
.
75
)
from
liquid
(3)
.
5.
Average
Hawaiian
parental
tholeiite
.
6.
Continental
thol
e
iit
e (Tasmania)
(Frey
a
nd
others,
1978).
7.
Average
oceanic
tholeiite
glass
(Elthon,
1979)
.
dependent
on
compos1t10n
a
nd
temperature.
If
the
mantle
is
chemically
stratified,
mixing
will
be
less
vigorous
and
chemically
distinct
compo-
nents
can
survive.
Part
of
this
gravitational
stratification
will
be
irreversible.
The
coefficient
of
thermal
expansion
is
high
at
low-pressure
and
high-temperature.
This
means
that
temperature
can
overcome
intrinsic
density
differences.
However,
at
high
pressure,
this
is
no
longer
possible
and
deep
dense
layers
may
be
trapped.
At
lower
mantle
con
-
ditions,
a
chemically
distinct
layer
with
an
intrin-
sic
density
contrast
of
~
1%
can
be
stable
against
convective
over-turn
and
mixing.
Crystallization
of
a
melt
layer
or
magma
ocean
leads
to
a
series
of
cumulate
layers
,
and
fractionation
of
the
LIL.
Cumulate
layers
originally
contain
interstitial
fluids
that
hold
most
of
the
incompatible
ele-
ments.
As
crystallization
proceeds,
these
melts
may
migrate
upward
. Melts
from
an
eclogite
or
olivine
eclogite
cumulate
have
the
characteristics
of
kimberlites.
Removal
of
late-stage
(kimberlite)
intercumulus
fluids
from
an
eclogite-rich
cumu-
late
layer
will
deplete
it
and
enrich
the
overly-
ing
olivine-rich
layer
.
The
enrichment,
however,
will
be
selective.
It
will
be
uniform
in
the
very
in
compatible
elements,
giving
apparently
primit-
ive
ratios
of
Rb
/Sr,
Sm/Nd
and
such,
but
will
impart
a
pattern
of
depletion
in
the
HREE,
yttr-
itnn,
sodium,
manganese
and
so
on
since
these
are
the
eclogite-compatible
elements.
Partial
melts
from
a
sha
ll
ow
enriched
reservoir
will
therefore
appear
to
have
a
garnet-residual
pat-
tern,
even
if
this
reservoir
contains
no
garnet.
This
pattern
can
be
transferred
to
any
MORB
mag-
mas
interacting
with
this
layer
.
Access
to
deep
layers
Convection
in
a
chemically
stratified
system
causes
lateral
variations
in
temperature,
and
deformation
of
the
interfaces
because
of
the
buoyancy
of
the
uprising
cu
rrents.
If
this
defor-
ITiation
raises
a
chemical
boundary
across
the
solidus,
or
if
the
temperature
is
perturbed
by,
for
example,
continental
insulation
,
then
partial
melting
ca
n
ge
nerate
buoyant
diapirs
,
even
in
a
dense
eclogite-rich
layer.
Subsolidus
reactions
between
garnet
and
clinopyroxene
also
occur
at
EARLY
CHEMICAL
STRATIFICATION
OF
THE
MANTLE
321
high
temperature.
This
results
in
a
temperature-
induced
density
decrease
much
greater
than
can
be
achieved
by
thermal
expansion.
Adiabatic
ascent
of
an
eclogite
blob,
or
a
diapir
from
a
buried
eclogite
or
piclogite
layer,
can
lead
to
extensive
melting.
Crystal
settling
and
melt
extraction
can
be
avoided
in
a
rapidly
rising
diapir
because
of
the
high
temperatures,
temper-
ature
gradients
and
stresses,
and
the
surround-
ing
envelope
of
subsolidus
peridotite.
Formation
of
fertile
regions
in
the
mantle
The
process
of
planetary
accretion
and
melting
during
accretion
is
akin
to
a
zone-refining
pro-
cess
.
The
surface
of
the
planet,
where
the
kinetic
energy
of
accretion
is
turned
into
heat,
acts
as
the
furnace;
refractory,
'purified'
material
is
fed
into
the
planet.
The
incompatible
elements
and
melts
are
preferentially
retained
near
the
sur-
face.
A
deep
magma
ocean
or
whole
planet
melt-
ing
-
at
any
one
time
- is
not
required,
or
even
desirable.
It
is
not
desirable
since
if
the
surface
melt
layer
is
in
equilibrium
with
dense
high-
pressure
phases
such
as
perovskite,
there
should
be
anomalies
in
the
trace-element
patterns
of
upper
mantle
materials
that
are
not
observed
.
Tables
25.1
and
25.2
and
Figure
25.1
illus-
trate
a
petrological
evolutionary
scheme
for
the
mantle.
Early
melting
is
likely
to
be
extensive
since
large
amounts
of
melt,
15-25
%,
occur
in
small
temperature
range
just
above
the
solidus.
Eclogite
and
basalt
extraction
leave
a
peridotitic
residue
deficient
in
the
basaltic
elements,
Ca,
Al
and
Na.
Olivine
cumulates
and
enriched
resid-
ual
fluids
are
complements
to
the
eclogite
and
basalt
regions
. Table
25.2
gives
a
more
detailed
comparison
of
the
possible
products
of
mantle
differentiation.
These
tables
illustrate
the
plausibility
of
eclogite-rich
regions
in
the
mantle
.
An
eclogite
or
basalt
layer,
or
distributed
blobs,
representing
about
10-15
%
of
the
mantle
reconciles
the
major-
element
compositions
inferred
by
cosmochemi-
cal,
geophysical
and
petrological
techniques
,
and
can
account
for
such
elements
as
Nb,
Ti
and
Zr.
HOM
OGE
N
EO
US
UND
E
PL
ET EO
MA
NT
LE
Differentiation
of
a
planet
during
accretion
and
early
high-temperature
evolution
.
E
is
the
enrichment
of
incompatible
elements,
relative
co
the
starting
materials.
These
elements
have
low
crystal-melt
partition
coefficients
and
therefore
readily
enter
the
melt
fraction.
f
is
the
fractionation
factor
and
gives
the
ratio
of
two
incompatible
elements
in
the
melt,
expressed
as
the
difference
from
the
starting
material.
Very
incompatible
elements
occur
in
the
same
ratio
in
melts
as
in
the
original,
or
primitive
,
material.
Isotopic
ratios
of
these
elements
will
evolve
at
the
same
race
as
in
primitive
material.
A
magma
ocean
will
therefore
be
enriched
but
unfraccionated.
As
the
magma
ocean
crysta
lliz
es
,
the
fractionating
crystals
will
either
float
or
sink,
leaving
behind
an
enriched,
fractionated
residual
liquid
layer
.
This
fluid
may
permeate
the
shallow
mantle,
giving
an
enriched
geochemical
signature
co
chis
region.
and
co
the
continental
cruse
.
The
large
difference
in
crystallization
temperature
and
density
of
olivine-orthopyroxene
(
ol-opx).
garnet,
plagioclase
and
so
on,
means
chat
mineralogically
distinct
regions
can
form
in
early
Earth
history.
The
source
regions
for
son1.e
basalts
may
be
eclogite-rich
cumulates
or
blobs
that
have
been
depleted
by
removal
of
a
kimberlite-like
fluid.
Eclogitic
layers
or
blobs
become
unstable
at
depth
as
they
warm
up.
Garnet-clinopyroxene
reactions
and
partial
melting
contribute
to
the
buoyancy
.
Early
chemical
stratification
of
the
mantle
Chemical
stratification
resulting
from
early
dif-
ferentiation
of
the
mantle
,
upward
removal
of
the
melt
and
fractionation
via
crystal
settling
during
cooling
is
one
way
to
explain
chemi-
cally
distinct
reservoirs.
In
the
first
stage
,
proba-
bly
during
accretion,
the
incompatible
elements
(including
Rb,
Sr,
Nd,
Sm
and
U)
are
concen-
trated
into
melts
(zone
refined)
and
the
shallow