Theory of the Earth
Don L. Anderson
Chapter 9. The Source Region
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:
Magmas probably represent partial, rather than complete, melts of the basalt
source region. Much or most of the source
region is left behind when a melt is
extracted from a partially molten sour
ce region. The composition, mineralogy
and depth of the source region are all co
ntroversial. Little progress can be made
without a simultaneous consideration of field and experimental petrology,
isotopes and geophysics. Many plausible
scenarios can be constructed from parts
of the data, but most have little merit when tested against other sets of data.
Magmas are just part of the story. The denser and more refractory parts of the
mantle are the subject of this chapter.
-
-
---
The
Source
Region
Long
is
the
way
And
hard,
that
out
of
hell leads
up
to
light.
-
JOHN
MILTON
M
agmas probably represent partial, rather
than
com
-
plete, melts
of
the basalt source region.
Much
or
most of
the source region
is left
behind
when
a melt
is ex
-
tracted from a partially
molten
source region. The compo
-
sition, mineralogy
and
depth
of
the source region are all
controversial. Little progress
can be made
without a simul
-
taneous consideration
of field
and
experimental petrology,
isotopes and geophysics.
Many
plausible scenarios can
be
constructed from parts of the
data,
but
most have
little merit
when
tested against other sets
of
data.
Magmas
are just
part
of
the story.
The
denser
and
more
refractory parts of
the
mantle are
the
subject
of
this chapter.
BACKGROUND
The
rocks
of
the mantle are compounds
involving
predomi
-
nately
MgO,
SiO,
and
FeO.
These oxides account for more
than
90
percent
of
the mantle. Intuitively, one might expect
that
the bulk
properties
of the
mantle, such as density
and
seismic
velocity and
the locations
of
phase change dis
-
continuities, are controlled only
by
compounds such
as
(Mg,Fe)Si03
and
(Mg,Fe),SiO,
and
their high
-
pressure
phases.
Most
discussions
of mantle
chemistry are
based
on
this assumption.
CaO
and
A1203
are
minor
constituents
in
the Earth as a whole, accounting for less
than
8
percent
of
the mantle;
however, they have an
importance far
beyond
their abundance. In
an
alumina
-
poor mantle the mineral
-
ogy is dominated
by
olivine and pyroxene
and
their high
-
pressure
forms
such
as
p-
and
y
-
spinel,
SiO,
(stishovite)
and majorite.
However,
a small amount
of
A1203
com
-
pletely
changes the mineralogy.
(Mg,Fe)O,
CaO
and
Si02
combine
with
A1203
to form
garnet,
which
may
be
the
dominant mineral
between
400
and
650
km.
Generally,
changes
in
mineralogy
have
a greater
effect
on density
and
seismic velocities
than small
changes
in
composition
of
a
given phase.
CaO
and
Al,O,
change the
melting
point
as
well
as the mineralogy
and
depths
of
phase changes.
They
therefore
influence the
evolution, stratigraphy
and
seismic
profiles of
the mantle. Some regions
of
the mantle
may, in
fact, be
CaO-
and
A1203-rich.
The major shallow
-
mantle minerals,
MgSiO,
(pyrox
-
ene)
and
Mg2Si0,
(olivine), are unstable at
high
pressure
and
transform
to denser phases
in
the
spinel, garnet
(major-
ite),
ilmenite, perovskite, periclase
and
rutile structures.
The small amount
of
FeO
in the upper
mantle
readily
sub
-
stitutes for
MgO
and
serves to increase the density, decrease
the seismic velocities
and lower the
transition pressure for
mantle phase changes.
CaO
and
A1203
drastically alter the
phase assemblages
and
the densities
in
the mantle.
They
also are major constituents
of
the basaltic
or
low
-
melting
fraction
of
the mantle
and
serve
to
decrease the melting
point.
A1203
stabilizes the garnet structure,
which means
that the basaltic fraction
of
the mantle is denser
than
"
normal
"
or
A1203-poor
upper mantle, but less dense
than
the perovskite
-
rich
lower mantle
in
the
pressure interval
over
which
garnet
is
stable. Early differentiation
of
the
Earth tends to separate the
Al,O,-rich
(basalt, eclogite)
and
A1203-poor
silicates (peridotite),
and
the density differential
is such that there should
be
little remixing. It
is
probable
that the major geochemical reservoirs
of the
mantle,
which
have
remained isolated for a large fraction
of
Earth
history,
also differ in
AI2O3
content
and
intrinsic density.
Reservoirs
rich
in
Al,O,
and
CaO
are called
"
fertile.
"
We
do
not know
the composition or
mineralogy
of
the
MORB
reservoir.
At
the depth
of
MORB
genesis,
basalts
would
crystallize as eclogite. Eclogite is a dense
clinopy-
roxene
-
garnet
-
rich
rock that
is the high
-
pressure form
of
MORB
and
picrite,
the
postulated direct parent
of
MORB.
The parent reservoir,
in
addition, probably contains ortho
-
pyroxene
and
olivine, the other
major
minerals
of
the
mantle. The composition
of
the
MORB
reservoir can there
-
fore range from eclogite to garnet peridotite, depending on
depth
and
the
extent
of
partial
melting
and
crystal fraction
-
ation occurring prior
to magma
separation. The
bulk
of
the
material erupting at oceanic islands
and
as
continental
flood
basalts
may
also come from either a garnet peridotite or
eclogite source region. The source
region
for so
-
called
hot-
spot
magmas
and
midocean
-
ridge
basalts
may, in
fact,
be
the same
if
hotspot
magmas
are contaminated
in the
shallow
mantle
prior
to
eruption.
The generally
accepted
view
that
basalts represent
various degrees
of
partial melting,
up
to about
20
percent,
of
a garnet peridotite source
region
is based
on several lines
of
evidence
and
assumptions
(Yoder,
1976). Peridotites are
a common rock
type
and
are
found
as
xenoliths
in
kimber-
lite
and
magma,
in
obducted
sections
of
oceanic lithosphere
and in
dredge hauls
at
fracture zones. The seismic proper
-
ties
of
the shallow mantle are
generally consistent
with
pe
-
ridotite. Garnet peridotites can form basalts
by
partial melt
-
ing,
and some
peridotites appear to
be
the refractory residue
remaining after
basalt
extraction. Other potential source
materials,
such
as
eclogite or pyroxenite,
are not
only rarer
but are
thought
(erroneously)
to
require
more
extensive
melting
in order to
provide
basaltic magmas. Melt
-
crystal
separation
is
believed
to occur before such extensive melt
-
ing can occur. These latter arguments only
hold
if
the
source
region
is identical
in
chemistry
to
the basalts it
pro
-
duces
and
if melt
-
crystal
separation
is more
effective
at the
depth
of
initial
melting than
source
buoyancy.
Eclogites
cover
a wide
range
of
compositions
and
they
melt
in
a
eu-
tectic
-
like fashion, thus
basalts
can form
from
eclogites
over
a wide
range
of
partial
melting
conditions.
There are
several
developments that
have
reopened
the question
of
the
nature and
depth
of
the basalt source
region. Seismic evidence
strongly
suggests that the source
of
midocean
-
ridge
basalts
is
in
the
transition region,
well
below
the
low
-
velocity
zone. The properties
of
the
shallow
mantle
are, therefore,
not
relevant. The low
-
velocity zone
itself
is highly
variable in depth, thickness
and
velocity.
It
locally extends
below
400 kilometers. Midocean
-
ridge
ba-
salts are possibly derivative from
picrites
after extensive
crystal fractionation
and may,
therefore, require more ex
-
tensive melting than
previously thought. Komatiites either
require
more than
60 percent melting, indicating that large
amounts
of
melting
are possible prior
to melt
separation, or
that
they
formed at great depth;
in
either case
high
temper
-
atures are implied.
High
melt
densities
make
melt
-
crystal
separation
more difficult at
great depth.
If
diapiric ascent is
rapid, extensive
melting can occur because
of
adiabiatic de
-
compression. Melt
-
crystal separation, or crystal fraction
-
ation,
may
be
restricted to depths less
than
100
km,
where
diapiric ascent is
slowed
by
high
viscosities, high
-
strength
lithosphere
or buoyancy
considerations. Finally, the elastic
properties
of
the transition zone (see Chapter 4) are com
-
patible
with an
eclogite
-
rich composition.
Basalts are chemically equivalent
to
garnet
plus
clino-
pyroxene (diopside
and
jadeite)
and
melting
of
these min
-
erals is
involved
in basalt genesis. Experimental petrology,
however,
cannot tell
us
if the garnet
and
clinopyroxene are
dispersed,
as in
peridotite, or
how
much
olivine
and
ortho
-
pyroxene are in the source region. The basalt fraction
of
pyrolite, a hypothetical source peridotite, is arbitrary. Ex
-
perimental petrology cannot constrain the bulk composition
of
the source region. The existence
of
meteorites rich in
olivine
has provided
support for the argument that perido
-
tite
is the
parental
basalt material
in
the
Earth.
On
the other
hand, cosmochemical models
of
the
Earth
have
more pyrox
-
ene
and
less olivine
than most
peridotite models
for
the
mantle,
and
partial melting
of
meteorites does
not
yield liq
-
uids
with
appropriate iron
-
magnesium
ratios:
Meteorite ol
-
ivine~
are generally
much
higher
in
FeO
than
mantle
oliv-
ines. There is little doubt that there is abundant olivine
in
the mantle. The issues are
whether
it,
and its
high
-
pressure
phases, are distributed uniformly throughout the mantle
and
whether
it is
the dominant mineral in the basalt
source region.
Observations
that have been
used in
support
of an
olivine
-
rich
(>50
percent) source region
for
basalts-gar-
net
peridotites are stable
in
the
upper mantle, their
bulk
compositions are consistent
with
the materials forming the
Earth,
they
are
prominent among
the
recovered
deep
-
seated
samples,
they
are capable
of
yielding
material
of
basaltic
composition
(Yoder, 1976)
-
deal with
possible mineral as
-
semblages
and bulk
compositions somewhere in the mantle
rather
than
dealing directly
with
the basalt source region.
The
"
conclusion
"
that garnet peridotite
is
the immediate
parent
of
basaltic
or
picritic
magma
is actually a
working
hypothesis,
not well
established
(Yoder,
1976). Melting
of
garnet
and
clinopyroxene, whether embedded in a perido
-
tite or eclogite matrix, will form basalt.
The
eclogite itself
may well
have been
derived from partial melting or crystal
fractionation
of
peridotite at
an
earlier stage
of
Earth
evolution.
Yoder
(1976)
and
Carmichael and others (1974)
have given well
-
balanced
treatments
of
the various possible
source rocks.
A
source
region
with
less than
50
percent ol
-
ivine cannot
be ruled
out.
The
effects
of
melt
compressibility
and
matrix stiffness
(reviewed
in
Chapter
7)
are
such
that separation
of
melt is
difficult
at
high
pressure
(Rigden
and
others, 1984).
When
sufficient melting occurs at depth, the
whole
source region
may
become
unstable,
bringing
melt
plus
matrix to shallow
levels
where
melt separation
can
occur.
In
a chemically
stratified mantle the
high
temperature gradient
in
a thermal
boundary
layer
causes
melting
to initiate in the deeper
layer.
Melt
can be retained
at depth
if the overlying mantle
has
high
viscosity
and low
permeability. The high density,
and
possible high
viscosity,
of
melts at depth also
makes
melt
extraction
more difficult than
at shallower levels. The in
-
creased
buoyancy
of
a partially
molten
source
region
may
be
more important
than
the density contrast
between melts
and
residual crystals. Garnet exsolution at high temperature
can also cause a deeper
layer to become buoyant
relative to
a colder
shallow
peridotite
layer.
These
effects
serve to ac
-
centuate the instability
of
thermal boundary
layers
in a
chemically stratified upper mantle. Such boundary layers
may
occur at the
base
of
the upper mantle (near 650 km)
and
near 400
km.
Of
course, the instability breaks the strict
layering,
and
such a
mantle
will exhibit
both
radial
and
lat
-
eral inhomogeneity, as suggested
by
the seismic results dis
-
cussed
in
Chapter
10.
The previous chapter discussed the properties
of
the
various types
of magmas
that emerge from the mantle.
These
magmas
provide
clues about the chemistry
of
some
parts
of
the mantle
but
do not provide a unique composition
or
mineralogy.
We
do
not even know what
fraction
of
the
mantle the basalts represent or from
what
depth
they
origi
-
nated.
Presumably,
the parts that are left
behind
are denser
and,
possibly,
more
refractory. It is
not
clear that
we
have
samples
of
the original source
rock
or
even
a representative
sample
of
the residual material.
Various
attempts
have been
made
to infer the properties
of
the basalt source region from
rocks
that
are exposed at the surface,
and
we
now
turn our
attention to observed
and
hypothetical rock types that
may
be
important
in
the upper mantle.
ULTRAMAFIC ROCKS
Ultramafic rocks are composed
chiefly
of
ferromagnesian
minerals and have
a low
silicon content. The term
is often
used
interchangeably
with
"
ultrabasic,
"
although
pyroxene-
rich rocks are ultramafic but
not
ultrabasic because
of
their
high
SiO,
content. Peridotites, lherzolite, dunite
and
harz-
burgite are specific
names
applied to ultramafic rocks that
are
chiefly
composed
of
olivine, orthopyroxene,
clinopy
-
roxene
and
an
aluminous phase
such
as
plagioclase, spinel
or garnet. Ultramafic rocks are dense
and mainly
composed
of
refractory minerals
with high
seismic velocities. Most
of
the
shallow
mantle is probably ultramafic
in
composition.
Basic rocks,
such
as
basalts, become dense at
high
pressure
(for example, eclogite)
and
can
have
properties comparable
to
the
more
refractory peridotites. The
main
subdivisions
of
ultramafic rocks,
along with
the dominant minerals, are
garnet peridotite (olivine, orthopyroxene, clinopyroxene,
garnet), lherzolite (olivine, orthopyroxene, clinopyroxene),
harzburgite (olivine, orthopyroxene), wehrlite (olivine,
cli-
nopyroxene), dunite (olivine
?
clinopyroxene), websterite
(orthopyroxene, clinopyroxene)
and
eclogite
(clinopyrox-
ene,
garnet). Some eclogites overlap
basalts
in their
bulk
Garnet
FIGURE
9
-
1
Nomenclature tetrahedron
for
assemblages
of
olivine, clinopy
-
roxene,
orthopyroxene and
garnet.
Dunites and
garnet
peridotites
lie
within the tetrahedron.
chemistry. The relationships
between
these
rocks
are
shown
in
Figure
9
-
1.
Peridotites
can
represent
1.
The refractory residue left after
basalt extraction
2.
Cumulates formed
by
the crystallization
of
a magma
3.
Primitive
mantle that
can
yield basalts
by
partial
melting
4.
Cumulates or residues
that have been
intruded
by
basalt
5.
High
-
pressure or high
-
temperature
melts
(Herzberg,
1984,
1986)
Peridotites are divided into fertile or infertile (or
bar
-
ren) depending on their
A1,0,,
CaO
and
Na,O
content.
Fer
-
tile peridotites can
be
viewed
as
having an
appreciable
ba
-
saltic component. The terms
"
enriched
"
and
"
depleted
"
are often
used
interchangeably
with
"
fertile
"
and
"
infer
-
tile
"
but have
trace
-
element
and isotopic
connotations
that are often inconsistent
with
the
major
-
element chem
-
istry. I use
"
infertile
"
and
"
barren
"
as
attributes
of
a rock
poor or very poor
in
CaO
and
Al,O,
and
"
depleted
"
for
rocks poor
in
the incompatible trace elements
and having
low
ratios
of
RbISr,
LREEIHREE
and
so
on. These distinc
-
tions are necessary since some rocks are fertile
yet
de
-
pleted, for instance, the source region for midocean
-
ridge
basalts.
Table
9
-
1
gives compositions for representative
ul-
tramafic rocks.
Garnet lherzolites are composed
mainly
of
olivine
and
orthopyroxene
(Table
9
-
2).
Olivine is generally in the range
of
60
to
70
volume
percent
and
orthopyroxene
30
to
50
percent. The
average
clinopyroxene
and
garnet proportions
TABLE
9
-
1
Compositions
of Spinel
and
Garnet Lherzolites
Spinel Lherzolite
Continental
Oceanic
Garnet
Oxide
(avg. of
301)
(avg.
of
83)
Lherzolite
SiO,
44.15
44.40
44.90
A1203
1.96
2.38
1.40
FeO
8.28
8.31
7.89
MgO
42.25
42.06
42.60
CaO
,2.08
1.34
0.82
Na,O
0.18
0.27
0.11
K,O
0.05
0.09
0.04
MnO
0.12
0.17
0.11
TiO,
0.07
0.13
'
0.06
P,O,
0.02
0.06
-
NiO
0.27
0.31
0.26
Cr203
0.44
0.44
0.32
Maal~e
and
Aoki
(1977).
are
about
5
percent
and
2 percent, respectively
(Maalgle
and
Aoki, 1977).
The major oxides in peridotites
and
lherzolites gener
-
ally correlate
well
(Figure
9
-
2 and Table
9
-
3).
An
increase
in
MgO
correlates
with
decreases
in
SiO,,
A1,03,
CaO,
TiO,
and
NazO
and
an
increase
in
NiO.
CaO,
A1203
and
Na,O
all
go
to approximately zero at
an
MgO
content
of
about
48
weight
percent.
Cr203,
MnO
and
FeO
are
roughly
constant. The lherzolite trend
can
be
explained
by
variable
amounts
of
clinopyroxene
and
garnet.
Olivine
-
and
orthopyroxene
-
rich rocks, presumably
from the mantle, are found
in
foldbelts, ophiolite sections,
oceanic fracture
zones
and, as xenoliths, in kimberlites
and
alkali
-
rich magmas.
They
are rare
in
less viscous magmas
such
as
tholeiites. Olivine
and
orthopyroxene in varying
proportion are the
most
abundant minerals in peridotites.
These are dense refractory minerals, and peridotites are
therefore generally thought
to be
the residue after melt ex
-
tractions. Some peridotites are shallow cumulates deposited
TABLE
9
-
2
Mineralogy
of Spinel
and
Garnet Lherzolites
from cooling
basalts
and
'are
therefore
not
direct samples
of
the mantle. Alumina
in
peridotites is distributed
among
the
pyroxenes
and
accessory minerals
such
as
plagioclase, spi
-
nel and
garnet.
At
higher pressure
most
of
the
A1203
would
be
in
garnet.
Garnet
-
rich peridotite, or pyrolite, is the com
-
monly assumed parent
of
mantle basalts. This
variety
is fer
-
tile peridotite since it can
provide
basalt
by
partial melting.
Most peridotites,
however, have
relatively
low
Al,03
and
can
be
termed barren. These are commonly thought to
be
residual after melt extraction.
A1203-poor
peridotites are
less dense
than
the fertile
variety
and
should concentrate
in
the
shallow
mantle.
Given sufficient water
at crustal
and
shallow
mantle temperatures, peridotite
may
be
converted
to serpentinite
with
a large reduction
in density
and
seismic
velocity. Hydrated
upper
mantle
may
therefore be
seis-
mically
indistinguishable from
lower
crustal minerals.
Lherzolites typically contain 60
to
80
percent olivine,
20
to
40 percent orthopyroxene, less
than
14 percent
clino-
pyroxene
and
1
to 10 percent
of an
aluminous phase such
.as
spinel
or
garnet. Spinel lherzolites, the lower
-
pressure
assemblages, dredged from the ocean bottom are similar
in
composition to those found
in
alkali basalts
and
kimberlites
on
oceanic islands
and
continents. Garnet lherzolites, the
higher
-
pressure assemblages,
have lower
A1203,
CaO
and
FeO
and
are therefore denser
than
spinel lherzolites only
when they
contain appreciable garnet.
They would become
less dense
at
higher temperature, lower pressure or
if par
-
tially molten.
The major
-
element chemistries
of
lherzolites vary
in
a
systematic fashion.
Most
of
the oxides vary linearly
with
MgO
content.
SiO,,
A1203,
CaO,
Na,O
and
TiO,
decrease
with
an increase
in
MgO.
These trends
aregenerally
consis
-
tent
with variable amounts
of
a basaltic component.
How
-
ever,
the basaltic component, especially for the
MgO-SiO,
trend,
is
not tholeiitic or MORB.
If
lherzolites represent
olivine
-
orthopyroxene
-
rich rocks
with
variable amounts
of
melt extraction or addition,
this melt
component
is andesitic
in
major elements.
The major
-
element trends
of
lherzolites
may
also
be
controlled
by
melt
-
crystal equilibration at various depths
in
Spinel Lherzolite
Average
Range
Mineral
(wt.
pct.)
(vol.
pct.)
Olivine
66.7
65
-
90
Orthopyroxene
23.7
5
-
20
Clinopyroxene
7.8
3
-
14
Spinel
1.7
0.2
-
3
-
-
Garnet
Phlogopite
-
-
Garnet Lherzolite
Average
Range
(vol.
pct.)
(vol.
pct.)
Maaloe
and
Aoki
(1977)
V
MgO
(weight
percent)
FIGURE
9
-
2
SiO,,
CaO,
A1,0,
and
FeO
versus
MgO
for
igneous
rocks. The
basalt source region prob
-
ably
has
a composition
intermediate
between basalt
(MORB)
and harzburgite.
Most petro
-
logical models (PM)
of
the major
-
element
chemistry
of
the
source
region favor a small
basalt
fraction.
STP
(spinifex textured peridotites) are high
-
temperature MgO
-
rich
magmas.
Picrites
(PICR)
are
intermediate
in composition
between
STP
and MORB and may evolve
to
MORB
by
olivine
separation.
Picrites
and
eclogites
(ECL)
overlap
in composition.
Lher-
zolites
(LZ)
contain an orthopyroxene
component,
but
the
other rock
types are
mainly
cli-
nopyroxene
+
garnet
&
olivine. Squares
repres~ent
estimates
of
primitive mantle composi
-
tion
based
on
a chondritic
model.
If
the
upper
mantle
is primarily
lherzolite,
basalt and
harzburgite, the lower mantle (open dot) will be primarily
orthopyroxene. The
composition
of
the MORB source
(piclogite
model) probably
falls
between PICR and PM
or
STP.
the mantle. Lherzolites,
and most
other ultramafic rocks,
are generally
thought to
be
the
refractory residue comple
-
mentary to melts presently
being
extracted from the mantle.
They differ,
however,
from primitive
mantle
compositions.
In particular
they
contain
more
olivine
and
less
orthopyrox-
ene
than would be
appropriate for a chondritic or
"
cosmic
"
mantle. Upper
-
mantle lherzolites
and
basalts
may
be
com
-
plementary to the lower mantle, representing melts from the
original, accretional differentiation
of
the mantle. The
MgO
content
of
melts increases
with
temperature
and
with depth
of
melting.
At
great depth
(>200
km)
relatively low
-
MgO
phases, such as orthopyroxene
and
garnet
-
majorite
may
re
-
main
behind, giving olivine
-
rich melts. The major
-
element
trends in lherzolites
may
therefore represent trends
in
high
-
pressure melts (Herzberg, 1984; Herzberg and others, 1986).
Unserpentinized peridotites
have
seismic velocities
and
anisotropies appropriate for the
shallow
mantle. This
situation is often generalized to the
whole
mantle,
but
seis
-
mic
data
for
depths greater than 400
km
are
not
in agree
-
ment
with
that hypothesis.
It
is not
even
clear
that
peridotite
has
the proper seismic properties for
the lower
lithosphere.
In the depth interval 200
-
400
km
both
eclogite
and
peri-
dotite can satisfy
the
seismic
data.
Garnet pyroxenites
and
eclogites are also found
among
the rocks thought to
have been brought up
from the mantle,
and they have
physical properties that
overlap
those of the
ultramafic rocks. Garnet
-
rich pyroxenites
and
eclogites are
denser than peridotites
and
should therefore occur deeper.
The extrapolation
of
the properties
of
peridotites
to the deep
upper mantle,
much
less the
whole
mantle, should
be
done
with
caution.
Not only
do other rock types emerge from the
mantle, but there
is reason to
believe
that peridotites
will
be
concentrated
in
the
shallow
mantle. The
common belief
that
seismic data
require
a peridotite mantle is simply false. The
V,
and
V,
velocities
in the upper 150
krn
of
the mantle under
stable shield areas are consistent
with
olivine
-
rich
perido-
TABLE
9
-
3
Compositions
of
Peridotites
and
Pyroxenite
Lherzolites
Pyrox-
Spinel
Garnet Dunite
enite
Peridotites
Oxide
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
SiO,
44.15
44.40
44.90
41.20
48.60
44.14
46.36
42.1
A1,0,
1.96
2.38
1.40
1.31
4.30
1.57
0.98
FeO
8.28
8.31
7.89
11.0
10.0
8.31
6.56
7.10
MgO
42.25
42.06
42.60
43.44
19.10
43.87
44.58
48.3
CaO
2.08
1.34
0.82
0.80
13.60
1.40
0.92
Na,O
0.18
0.27
0.11
0.08
0.71
0.15
0.11
K2O
0.05
0.09
0.04
0.016
0.28
-
-
MnO
0.12
0.17
0.11
0.15
0.18
0.11
0.11
Ti0
,
0.07
0.13
0.06
0.06
0.83
0.13
0.05
p205
0.02
0.06
-
0.10
0.10
-
NiO
0.27
0.31
0.26
-
-
-
-
Cr20,
0.44
0.44
0.32
-
-
0.34
0.33
"
20
-
-
-
0.50
0.90
-
-
(I)
Average
of
301 continental spinel lherzolites
(Maaloe
and Aoki, 1977).
(2)
Average
of
83 oceanic spinel lherzolites
(Maaloe
and
Aoki, 1977).
(3)
Average
garnet lherzolite
(Maaloe
and Aoki, 1977).
(4)
Dunite
(Beus,
1976).
(5)
Pyroxenite
(Beus,
1976).
(6) High
-
T peridotites, South Africa (Boyd, 1986).
(7) Low
-
T peridotites, South Africa (Boyd, 1986).
(8) Extrapolated lherzolite trend
(-0
percent
A1,0,,
CaO,
Na,O,
etc.).
tite.
Elsewhere
in
the shallow mantle both
eclogite
and
pe-
ridotite are consistent
with
the data.
It should
be
kept in
mind
that the
most
abundant
rock
type
by
far that emerges
from
the mantle is tholeiite
(equivalent to garnet
plus
clinopyroxene). A garnet
perido-
tite
with
suitable compositions
of
garnet
and
clinopyroxene,
particularly
Na,O
and
TiO,
content, can certainly
yield ba
-
salts
by
partial melting. The
amount
of
melting,
however,
and
the
amount
of
residual olivine
and
orthopyroxene are
unknown. A suitable natural garnet peridotite (pyrolite)
has
not
yet
been
found
that
can provide the major
-
and
trace-
element abundances found
in,
for example, midocean
-
ridge
tholeiites. Such a peridotite
would have
to have
high
Al,O,,
Na,O,
CaO
and
TiO,
and
low
concentrations
of
the incom
-
patible elements. Pyrolite (next section) therefore remains
a hypothetical mantle rock.
If
picrites
are
the
parent for tholeiitic basalts,
then
roughly 30
percent
melting
is implied
for generation from a
shallow
peridotitic parent.
If
the parent is eclogitic,
then
similar temperatures
would
cause
more
extensive melting.
However,
generation
of
basaltic magmas from
an
eclogitic
parent does
not
require extensive melting. Melts
of
basaltic
composition are
provided over
a large range
of
partial
melting.
Basalts
and
peridotites are
two
of
the results
of
mantle
differentiation.
They both
occur near the surface
of
the
Earth
and may
not represent the whole story. They are also
not
necessarily the result
of
a single
-
stage differentiation
process.
In
the
next
chapter I discuss the possible evolution
of
the mantle, from its high
-
temperature birth to the for
-
mation
of
the geochemical reservoirs that
we
sample
today.
This complements the standard approach that attempts
to
take the observed rocks
and work
backward
in
time.
PYROLITE
The general idea of
a peridotitic mantle as a source for
ba
-
saltic
magmas
by
partial
melting
goes
back
at least as
far
as
Bowen
(1928).
Ringwood
(1962) formalized the concept
in
a
series
of
papers starting in 1962. He proposed the exis
-
tence
of
a primitive mantle material
that was defined by
the property
that
on fractional
melting
it would
yield a typi
-
cal basaltic
magma and leave behind
a residual refractory
dunite
-
peridotite. This he called
"
pyrolite
"
(pyroxene-
olivine
-
rock). The initially proposed composition
was
one
part basalt
and
four parts dunite, although this is arbitrary.
Green and
Ringwood
(1967) considered that 20
-
40 percent
melting will be
necessary before liquid segregates and be
-
gins
an
independent existence,
and
this
mechanical idea in
-
fluenced
their choice
of
basalt
-
to
-
dunite ratios. The stable
phase assemblages
in pyrolite, and
the disposition
of
Al,O,,
depend on temperature
and
pressure. Garnet pyrolite is es
-
sentially identical
with
garnet peridotite but is
more
fertile
than most
natural samples (Ringwood, 1962).
Ringwood
earlier
argued
that garnet pyrolite
would not
yield,
by
direct
partial melting, basaltic
magma and
proposed that partial
melting
of
plagioclase pyrolite
must
be
the
main
source
of
basalts. He attributed the low
-
velocity zone to the presence
of
plagioclase
and
assembled
many
arguments
in
favor
of
these conjectures. Plagioclase
is not
in
fact stable at
upper-
mantle pressures,
and
Ringwood
revised his
pyrolite model
and
ideas about the basalt source region
many
times
in
the
subsequent years.
In an
early
model
garnet pyrolite
was
considered to
be
free
of
orthopyroxene. In 1963
Green and
Ringwood
cal
-
culated a pyrolite composition
based on the assumption that
it was
three parts dunite
plus
one
part
of
the averages
of
tholeiitic
and
alkali olivine basalt; it therefore contained
some orthopyroxene.
In
1966
Ringwood
assumed that
py-
rolite
was
a three
-
to
-
one
mix
of
Alpine
-
type peridotite
and
a Hawaiian
olivine
-
tholeiite.
Table
9
-
4 gives
compositions
of
some
of
these pyrolite models.
The cornerstones
of
the pyrolite hypothesis were that
the basaltic magmas
that were
thought
to
represent the
largest degrees
of
partial melting, tholeiites, were
prirni-
tive, unfractionated
melts
and
that
they were
in equilibrium
with
the residual peridotite at the depth
of
melt separation.
O'Hara
(1968)
argued
that extensive crystal fractionation,
at depth, operated on the parent
magma
and,
therefore,
ba-
salts including tholeiites
were
not primary magmas.
He
sug
-
gested that partial
melting and magma
segregation occur at
pressures greater
than 30
kbar, rather
than
the 5
-
25 kbar in
the Green
-
Ringwood models.
Under
these conditions the
parent
magma
is picritic,
and
extensive olivine crystalliza
-
tion
at
low
pressure produces tholeiitic magmas.
He
also
proposed that extensive eclogite fractionation at depths
of
order
100
km occurred.
Ringwood
(1975) argued strongly
against
O'Hara's
model.
Jaques
and
Green
(1980)
on the basis
of
an
extensive
series
of
melting
experiments on synthetic peridotites aban
-
doned
the earlier
model
of
Green
and
Ringwood
(1967)
and
Green
(1971),
which involved
the segregation
of
olivine
tholeiite magmas
at about
30
km
depth. Studies
by
Green
and
others (1979)
and
Stolper (1980)
have
shown
that pri
-
mary
picritic
magmas are likely parents
for
midocean
-
ridge
basalts
by
separation
of
about
15
percent olivine at
about
60
-
70 km depth
and
temperatures
of
about 1400
-
1450°C.
Picritic
and
komatiitic
magmas
require
about
40
-
60 per
-
cent melting
of
peridotite
if
the
melting
initiates at
such
shallow
depths.
Green and
others
showed
that there can
be
no simple genetic relationship
between
the primary
tholei-
itic
picritic
magmas that
are parental
to
midocean
-
ridge
basalts
and
the residual harzburgite
or
accumulate peridotite
of
ophiolite complexes. Thus, these studies
effectively pull
the
rug
out from under the
basic
pyrolite assumptions.
Nevertheless, the same authors
proposed
yet
more
pyrolite
models
that are mixtures
of
picrite
and
varying, arbitrary
proportions
of
olivine (50
-
69 percent), orthopyroxene
(O-
25 percent)
and
spinel
(0.5
to 1 percent). The
new
pyrolites
are richer
in
Al,O,
and
CaO
than previous pyrolite
models
and less rich in
TiO,,
Na,O
and
K,O.
Their preferred
model
generates the primary midocean
-
ridge
basalts
by
24
percent
partial melting.
At
this point
one
is reminded
of
Karl
Popper's (1972)
"
conventionalist stratagem
"
:
"
Some genuinely testable
theories,
when
found to
be
false, are still
upheld
by
their
admirers
by
introducing
ad
hoc
some
auxiliary assumptions
TABLE
9
-
4
Com~ositions
of
Pvrolites
and
Garnet Peridotites
(Weight
Percent)
PysoUite
Garnet Peridotite
Oxide
(1)
(2)
(3)
(4)
6)
(6)
SiO,
TiO,
A1203
Cr203
MgO
FeO
MnO
CaO
Na20
K
20
-
-
(1)
Ringwood
(1979),
p.
7.
(2)
Green
and
Ringwood
(1963).
(3)
Ringwood
(1975).
(4)
Green and
others (1979).
(5)
Boyd
and
Mertzman
(1987).
(6)
Boyd and
Mertzman
(1987).
or
by
re
-
interpreting
the
theory
ad
hoc
in
such a
way
that
it
escapes refutation. Such a procedure is
always
possible, but
it rescues
the theory from refutation only at the price
of
destroying, or at least lowering
its scientific
status.
"
Jaques
and
Green
(1980),
in
common
with
most
pe
-
trologists,
have
restricted their attention to rocks that can
form
basalts
by
a relatively small degree
of
partial melt
-
ing
since
high
degrees
of
partial
melting
may
not be
geo
-
logically feasible.
They
point
to
laboratory measurements
showing
strong crystal settling
of
olivine
when
melting ex
-
ceeds 40 percent,
implying
that conditions
in
a convecting
mantle are
the same
as in
a static laboratory experiment.
The real Earth
has
a cold, high
-
viscosity, buoyant, strong
and
relatively impermeable lithosphere.
An
important ques
-
tion, then,
is how
much melting
can occur
in the underlying
convecting mantle
that
cannot easily lose
its
melt fraction
to the surface?
What
are the likely products
of
partial melt
-
ing
in
a convecting, differentiating mantle?
In
a garnet
-
clinopyroxene
-
rich rock there
is
a small
temperature differential
between
the onset
of
melting
and
the temperature
of
extreme melting. Adiabatic ascent
of
such
a rock
can
lead to
extensive melting. In contrast, in a
garnet peridotite, the eclogite fraction is
melted
near the
solidus
but
the
bulk
of
the
rock
does
not
melt until
much
higher temperatures, near the liquidus. Therefore, garnet
peridotites require a large temperature increase in order to
melt
extensively.
The pyrolite hypothesis
was
based
entirely on major
elements
and
on
several
arbitrary assumptions regarding al
-
lowable amounts
of
basalt
and
melting
in
the source region.
It does
not
satisfy trace
-
element or isotopic data on basalts,
and
it violates chondritic abundances
and
evidence
for
mantle
heterogeneity.
ECLOGITES
Hauy
(1822) introduced the term
"
eclogite
"
for rocks com
-
posed
of
omphacite (diopside plus jadeite)
and
garnet,
occasionally
accompanied
by
kyanite, zoisite, amphibole,
quartz
and
pyrrhotite.
Natural
eclogites
have
a variety
of
associations, chemistries, mineralogies
and
origins,
and
many names have been
introduced
to
categorize these sub
-
tleties. Nevertheless, the term
"
eclogite
"
implies different
things to different
workers.
To
some eclogites
mean
meta
-
morphic
crustal rocks,
and
to others the term implies
bi-
mineralic
kimberlite xenoliths. The chemical similarity
of
some
eclogites to
basalts prompted
early investigators to
consider eclogite
as
a possible source
of
basalts but
more
recently has been taken
as evidence
that
these eclogites are
simply subducted oceanic crust
or
basaltic
melts that have
crystallized
at high
pressure. Some eclogites are demon
-
strably
metamorphosed
basalts, while others appear to
be
igneous
rocks ranging
from
melts
to cumulates. The
trend
in recent
years has
been
toward
the splitters rather
than
the
lumpers,
and
toward
explanations that emphasize the de
-
rivative
and
secondary nature
of
eclogite rather
than
the
possible importance
of
eclogite
as
a source
rock
for basal
-
tic magmas.
Pyroxene
-
garnet rocks
with
jadeite
-
bearing pyroxenes
are found
as
inclusions
in
alkali basalt
flows
as
layers
in
ultramafic intrusions,
as
inclusions
in
kimberlite pipes, as
tectonic inclusions
in
metamorphic terranes associated
with
gneiss
and
schist,
and
as inclusions
in
glaucophane schist
areas. Jadeite
-
poor garnet clinopyroxenites are abundant in
Salt
Lake Crater, Hawaii. Only the eclogites from
kimber-
lite pipes are demonstrably from great depth, that
is,
well
into the upper mantle. Some
of
these contain diamonds.
Peridotites
and
lherzolites predominate over all other
rock
types as inclusions
in
diamond
-
bearing kimberlites.
The presence
of
diamond indicates origin depths
of
at least
130
km,
and
other petrological indicators suggest depths
even
greater.
In
a few
kimberlite pipes eclogites form the
majority
of
inclusions.
The
overwhelming majority
of
per-
idotites
and
lherzolites are infertile, that
is,
very
low
in alu
-
minum,
sodium
and
calcium.
If
the
distribution
of
rock
types
in
kimberlite inclusions is representative
of
the source
region,
the
majority
of
basaltic components
in
the upper
mantle resides
in
eclogites rather
than
in
an
olivine
-
rich
rock
(Smyth
and
others, 1984).
Table
9
-
5
shows
typical clinopyroxene
and
garnet
compositions
of
eclogites
and
peridotites, a synthetic
two-
mineral eclogite
and
for comparison, a typical MORB.
Note
that,
in
general, diopside
plus
garnet from peridotite
does
not
approximate the composition
of
MORB.
In most
cases,
however,
omphacite
and
garnet from eclogite bracket
MORB
compositions,
and
therefore eclogite is a more ap
-
propriate source rock.
Table
9
-
6 gives
comparisons
of
the
bulk
chemistry
of some
South African eclogites
and
MORB
and
an
estimate
of
the average composition
of
the oceanic
crust. There is a close correspondence
between
the compo
-
sition
of
kimberlite eclogites
and
the material
in
the oce
-
anic crust.
Many
kimberlite eclogites show
signs
of
garnet
exso-
lution from clinopyroxene, implying that there can
be
sub
-
stantial changes
in
the density
of
eclogites
as
a function
of
temperature
in
the subsolidus region. Clinopyroxenes
in
eclogites
have
exsolved
20
percent or
more
garnet, implying
a substantial increase
in
pressure or decrease in temperature
(Smyth
and
Caporuscio, 1984). A pressure increase is
un
-
likely.
A representative eclogite can increase
in
density
by
2.5
percent
by
cooling from
1350°C
to
950°C
at
30 to
50
kbar. The reverse process can
happen
as garnet
plus
cli-
nopyroxene
is elevated into a lower pressure regime
along
an
adiabat.
For
example, the density
of an
eclogite can de
-
crease
by about
3 percent simply
by
rising
50 km.
Thus,
garnet exsolution caused
by
pressure release can accom
-
plish more than a
1000°C
rise in temperature,
by
thermal
expansion, all in the subsolidus regime. This plus the
low