Chapter
23
The
upper
mantle
Let's
descend
to
that
blind
world
below.
I'll
go
first,
and
you
can
follow
Dante
Overview
The
composition
of
the
crust
and
the
upper
man
-
tle
are
the
results
of
a
series
of
meltin
g
and
frac-
tionation
events,
including
the
high-temperature
accretion
of
the
planet.
Early
attempts
to
esti-
mate
upper
mantle
(UM)
chemistry
started
from
the
assumption
that
UM
initially
was
the
same
as
bulk
silicate
Earth
(BSE)
and
differed
from
it
only
by
the
extraction
of
the
continen
-
tal
crust,
or
that
the
most
depleted
-
low
in
LIL,
U,
Th,
K
-
midocean-ridge
basalts
plus
their
depleted
refractory
residues
-
unaltered
abyssal
peridotites-
constitute
the
entire
upper
mantle
.
Traditionally,
geochemists
have
assumed
that
the
lower
mantle
is
und
egassed
BSE
or
primitive
mantle
(PM).
Geodynamicists
have
noted
that
there
are
heat
flow
problems
with
this
model,
which
they
fixed
up
by
putting
a
large
amount
of
U,
Th
and
Kin
a
lower
mantle
stealth
layer.
Large-scale
melting
and
differentiation
upon
accretion
probably
pre-enriched
the
upper
man-
tle
with
incompatible
elements,
including
the
radioactive
elements;
the
crust
and
the
vari-
ous
enriched
and
depleted
components
sam-
pled
by
current
melting
events
were
a
lr
eady
in
the
upper
mantle
shortly
after
accretion
and
solidification
.
Recycling
of
crust
into
the
upper
mantle
is
an
important
current
process.
It
is
possible
to
esti-
mate
the
compos
ition
of
the
fertile
upper
mantle
by
combining
known
components
of
the
upper
mantle-
basalts
,
peridotites,
recycled
crust
and
so
on.
The
MORB
source
is
just
part
of
the
upper
mantle
and
it
is
not
the
only
LIL-depleted
part
of
the
mantle.
It
is
not
necessarily
convec-
tively
homogenized.
Attempts
to
establish
an
average
composi-
tion
for
'the
upper
mantle'
focus
on
the
most
depleted
MORB
lavas
or
abyssal
peridotites
and
involve
major
assumptions
about
melt
genera-
tion,
melt
transport
and
differentiation
processes
that
have
affected
these.
The
depleted
upper
mantle,
that
part
of
the
mantle
that
is
assumed
to
provide
MORB
by
partial
melting
is
variously
called
DUM,
OM,
DMM
and
the
convecting
upper
mantle.
Simplified
mass
balance
cal-
culations
suggested
to
early
workers
that
this
depleted
mantle
constituted
only
~
30
%
of
the
mantle;
the
650-670
lan
discontinuity
was
there-
fore
adopted
as
the
boundary
between
DUM
and
't
he
primitive
undepleted
undegassed
lower
man
-
tle.'
The
starting
condition
for
the
upper
mantle
(UM)
was
taken
as
identical
to
primitive
man-
tle
(PM)
-
or
bulk
silicate
Earth
(BSE)
-
and
the
present
lower
mantle
(LM).
Estimates
of
PM
and
BSE
are
based
on
cosmological
and
petro-
logical
considerations.
The
hypothetical
prim-
itive
upper
mantle
-crust
plus
DUM-
is
labeled
PUM
.
It
is
further
assumed
that
the
upper
mantle
-
from
the
base
of
the
plate
to
650
Ian
-
after
extraction
of
the
crust,
became
well-stirred
and
chemically
homogenous
by
vig-
orous
convection
.
The
non-MORB
basalts
that
occur
at
the
initiation
of
spreading
and
at
various
location
s
along
the
global
spreading
302
THE
UPPER
MANTLE
system
are
attributed
to
plumes
from
the
chem-
ically
distinct
lower
mantle.
However,
it
can
be
shown
that
most
or
all
of
the
mantle
needs
to
be
depleted
and
degassed
to
form
the
crust
and
upper
mantle
and
to
explain
the
amount
of
40
Ar
in
the
atmosphere;
this
was
a
major
theme
in
the
first
edition
of
The-
ory
of
the
Earth
(Anderson,
1989).
Deple-
tion
and
degassing
of
the
upper
mantle
alone
cannot
explain
the
observations.
In
addition,
the
MORB
reservoir
and
the
CC
are
not
exactly
complementary;
another
enriched
component
is
needed.
When
this
(Qcomponent)
is
added
in,
MORB
+
CC
+
Q
require
that
most
of
the
mantle
must
be
processed
and
depleted
.
There
must
be
other
components
and
processes
beyond
sing
le
stage
small-degree
melt
removal
from
part
of
the
primordial
mantle
to
form
CC.
There
are
other
enriched
components
in
the
mantle,
probably
in
the
shallow
mantle,
and
other
depleted
residues
over
and
above
the
MORB-source.
The
upper
man-
tle
(UM)
is
still
generally
treated
as
if
its
composi-
tion
can
be
uniquely
determined
from
the
prop-
erties
of
depleted
MORB
-
NMORB
or
DMORB
-
and
depleted
peridotites,
continental
crust,
and
an
undifferentiated
starting
condition.
The
traditional
hotspot
and
plume
models
of
OIB
and
enriched
magma
genesis
and
mantle
heterogeneity
are
unsatisfactory
[mantle
plu-
mes,
plume
paradoxes].
Traditional
models
for
both
OIB
and
MORB
genesis
involving
only
peridotitic
protoliths
are
also
being
re-
evaluated
[
olivine-
free
mafic
sources
].
Although
recycling
has
lon
g
been
used
as
a
mech-
anism
for
modifYing
the
isotopic
character
of
OIB,
it
is
now
becoming
evident
that
it
can
also
create
melting
anomalies
.
The
roles
of
eclog-
ite
and
garnet
pyroxenite
in
petrogenesis
and
in
the
formation
of
melting
anomalies
are
becom-
ing
evident
(Escrig
et
al.,
2004,
2005,
Gao
et
a!.,
2004,
Sobolev
eta!.,
2005)
[
delamination
man-
tle
fertility
].
Midocean-ridge
basalts
rep-
resent
large
degrees
of
melting
of
a
large
source
volume,
and
involve
blending
of
magmas
having
different
melting
histories
.
The
Central
Limit
Theorem
explains
many
of
the
differences
between
MORB
and
other
kinds
of
melts
that
sample
smaller
volumes
of
the
heterogenous
mantle
.
Observed
isotopic
arrays
and
mixing
curves
of
basalts,
including
ocean-island
basalts
(OIB),
can
be
generated
by
various
stages
of
melting,
mix-
ing
,
melt
extraction,
depletion
and
enrichment
and
do
not
require
the
involvement
of
unfrac-
tionated,
primitive
or
lower-mantle
reservoirs
.
However,
the
first
stage
of
Earth
formation
-
the
accretional
stage
-
does
involve
lar
ge
degrees
of
melting
that
essentially
imparts
an
unfraction-
ated
-
but
enriched
-
chondritic
REE
pattern
to
the
upper
mantle
.
Small-degree
melts
from
this
then
serve
to
fractionate
LIL
What
is
the
upper
mantle?
On
the
basis
of
seismic
data
Bullen
divided
the
mantle
into
regions
labeled
B.
C
and
D.
Region
B
is
the
upper
mantle
and
C is
the
Transition
Zone
(410
to
1000
km).
D.
the
lower
mantle
,
starts
at
1000
km
depth.
The
upper
mantle
(Region
B)
was
s
ubs
eq
uently
found
to
contain
a
high-velocity
lid
and
a low-velocity
zone
(LVZ),
generally
asso-
ciated
with
the
asthenosphere.
Region
C,
the
mantle
transition
region
(TR),
was
subse-
quently
found
to
contain
two
abrupt
seismic
discontinuities
ne
ar
the
depths
of
410
and
650
km,
and
a
region
of
high
and
variable
gradi-
ent
below
650
km
depth
. A
depth
of
670
km
was
found
for
the
deeper
discontinuity
in
west-
ern
North
America
and
this
was
adopted
for
the
PREM
model.
However
,
the
average
depth
of
the
discontinuity
, globally,
is
650
km,
with
a varia-
tion
of
about
30
km.
Some
authors
have
referred
to
the
650
(or
670)
km
discontinuity
as
the
base
of
the
upper
mantle
and
have
suggested
that
this
represents
a
profound
chemical
and
isotopic
boundary
between
depleted
upper
mantle
and
primitive
lower
mantle,
rather
than
primarily
an
isochemical
phase
change,
as
originally
inferred
(Anderson,
1967).
Others
have
suggested
that
the
1000
km
level
is
a
chemical
boundary
and
should
be
retained
as
the
definition
of
the
top
of
the
low
er
mantle
.
Sometimes
the
TR
is
included
as
part
of
the
upper
mantle
;
sometimes
it
is
defined
as
a
separate
region,
Bullen's
Region
C.
This
con-
fusion
in
terminology
about
what
constitutes
the
upper
mantle
and
the
lower
mantle
has
led
to
the
widespread
view
that
there
are
fundamental
conflicts
between
isotope
geochemistry
and
geophysics,
and
confu-
sion
about
whether
slabs
penetrate
into
'the
lower
mantle'
or
not.
It
appears
that
the
upper
1000
km
of
the
mantle
-about
40%
by
mass
of
the
mantle
-
differs
from
the
rest
of
the
mantle
and
this
also
appears
to
be
the
active
layer
for
plate
tectonics.
The
majority
of
the
incompat-
ible
trace
elements
that
are
not
in
the
crust
may
be
confined
to
an
even
shallower
depth
range.
Bullen's
nomenclature
is
precise
and
use-
ful
and
I
will
follow
it.
The
terms
upper
mantle
and
lower
mantle
are
now
fuzzy
concepts
because
of
usage
in
the
geochemical
literature
and
the
decoupling
of
this
usage
from
seismological
data.
These
terms
will
be
used
when
precise
depths,
volumes
or
masses
are
not
needed.
The
term
mesosphere
has
also
been
used
for
the
mid-
mantle,
Bullen's
Region
D'.
The
terms
shallow
mantle
and
deep
mantle
will
be
used
to
avoid
the
conflict
between
the
precise
seismological
definitions
of
mantle
regions
and
geochemical
usage.
Depleted
upper
mantle;
the
DUM
idea
Geochemists
have
ideas
different
from
seismo-
logical
conventions
about
what
constitutes
the
upper
mantle.
They
are
based
on
compositions
of
depleted
midocean-ridge
basalts,
DMORB,
and
assumptions
about
how
these
form.
The
defini-
tion
of
the
upper
mantle
adopted
by
isotope
geo-
chemists
is
the
following.
The
upper
mantle
is
that
part
of
the
mantle
that
provides
uniform
and
depleted
midocean
ridge
basalts
and
that
formed
by
removal
of
the
continental
crust;
it
extends
from
the
Moho
to
the
670-lan
mantle
discontinuity.
It
is
also
called
'the
convecting
mantle.'
The
upper
mantle
is
usually
viewed
by
geochemists
as
homogenous
because
MORB
are
relatively
homogenous.
The
assumption
is
that
homogenous
products
require
homogenous
sources.
The
composition
of
the
inferred
MORB
reservoir
has
been
attributed
to
the
whole
upper
mantle.
Since
MORB
is
depleted
in
LIL
compared
to
other
basalts,
the
above
assumptions
have
led
to
the
Depleted
Upper
Mantle,
or
DUM,
con-
cept.
DUM
is
a
two-component
system,
DMORB
WHAT
IS
THE
UPPER
MANTLE?
303
and
a
residual
depleted
peridotite
or
an
unal-
tered
abyssal
peridotite.
It
has
a
simple
one-stage
history.
In
some
models
of
upper-mantle
chemistry
only
the
most
depleted
materials
are
used
in
its
construction;
hence
Depleted
Upper
Mantle.
Hete
r
og
en
o us
uppe
r
man
tl
e
Convective
stirring
takes
large
blobs
and
shears
and
stretches
them,
folds
them
and
stretches
them
more,
repeatedly,
until
the
dimensions
are
very
small;
for
obvious
reasons
this
is
known
as
the
Baker's
transformation,
a
funda-
mental
result
of
chaotic
advection
the-
ory.
This
theory
may
not
be
appropriate
for
the
mantle.
It
implies
high
Rayleigh
number,
stirring
or
folding
in
one
direction
and
low-viscosity
pas-
sive
particles.
We
must
therefore
pay
attention
to
the
materials
that
enter
the
mantle
instead
of
relying
on
averages
of
the
magmatic
products.
We
must
keep
an
open
mind
about
the
possi-
bility
of
large
fertile
blobs
in
the
man-
tle,
and
extensive
regions
having
high
homologous
temperature.
From
a
petrological
point
of
view,
the
man-
tle
can
be
viewed
as
a
multi-component
sys-
tem.
The
known
components
of
the
upper
man-
tle
are
recycled
continental
crust
and
other
mafic
components,
ultramafic
rocks,
MORB
and
other
basalts,
depleted
refractory
residues
and
enriched
components
(Q,
quintessence
or
fifth
component)
such
as
kimberlite;
mixtures
of
these
satisfY
most
chemical
constraints
on
the
composition
of
the
mantle
or
BSE,
including
major
and
trace
elements.
The
compositions
of
basalts
and
the
compositions
of
con-
tinental
and
abyssal
peridotites
are
available
in
petrological
databases.
These
can
be
used
to
reassemble
the
original
petrology
and
composition
of
the
mantle,
and
with
a
few
other
assumptions,
the
composition
of
the
upper
mantle.
Basalt
and
peridotite
compositions
represent
the
culmination
of
melt
depletion
and
enrich-
ment
processes
over
the
entire
history
of
the
mantle,
including
the
accretional
process.
There
are
a
variety
of
basalts
and
peridotites.
The
com-
positions
of
many
basalts
and
peridotites
appear
to
lie
along
mixing
lines
and
the
end-member
304
THE
UPPER
MANTLE
compositiOns
have
been
interpreted
as
trapped
melts.
depleted
residues
and
recycled
and
dela-
minated
materials
.
On
major
element
plots
(Chapter
15)
the
end-members
are
harzburgite
and
MORE
,
or
eclogite.
Picrites.
komatiites
and
primitive
mantle
have
intermediate
composi-
tions
.
On
LIL
and
REE
plots,
the
extreme
com-
positions
are
kimberlites
and
DMORE
or
abyssal
peridotite.
From
an
isotopic
point
of
view,
oceanic
basalts
are
also
treated
as
multi-component
sys-
tems
involving
mixtures
of
DMM,
EMl
,
EM2
and
HIMU
and
c
or
FOZO.
These
are
short-
hand
names
for
what
are
thought
to
be
the
various
enriched
(EM)
and
depleted
(OM)
iso-
topic
end-members
of
the
mantle
and
there
is
a
large
literature
on
each.
There
is
no
agree-
ment
regarding
the
lithology
or
history
that
goes
with
each
component.
Trends
of
OIB
and
MORE
isotopic
compositions
approach
-
or
con-
verge
on-
a
hypothetical
component
of
the
man-
tle
referred
to
as
FOZO
(focal
zone)
or
C
(common)
.
It
has
been
assumed
that
this
reflects
the
composition
of
the
lower
mantle.
It
is
not
clear
why
the
most
common
component
in
basalts
should
represent
the
deepest.
rather
than
the
shallowest,
mantle.
Melts
pond
beneath
,
and
percolate
through,
the
lithosphere
,
and
inter-
act
with
it.
A
lithosphere
or
harzbur
g
ite
com-
ponent
may
therefore
be
involved
in
most
mag-
mas
.
If
so,
ultramafic
rocks
(UMR)
may
anchor
the
ends
of
both
major
element
and
isotope
mixing
arrays
.
UMRs
are
certainly
the
most
common
or
prevalent
lithology
of
the
shallow
mantle
.
The
average
or
prevalent
mantle
com-
position
has
been
referred
to
as
PREMA
in
the
isotope
literature.
In
contrast
to
the
end-member
components,
PREMA,
c
and
FOZO
are
interior
components
-
on
isotope
diagrams
-
and
are
therefore
either
mixtures
or
sources.
The
one
extreme
attribute
of
these
average
compositions
is
that
basalts
falling
near
these
compositions
tend
to
have
higher
variance
in
their
3
He/
4
He
ratios
,
and
therefore
contain
some
high
3
Hej4He
samples.
The
most
prevalent
lithology
of
the
mantle
-
peridotite
-
may
be
implicated
in
this
component
while
EM
and
HIMU
may
reside
in
the
fertile
or
mafic
components.
Enrichment
and
depletion
processes
Vigorous
stirring
can
homogenize
a
fluid
by
a
process
!mown
as
chaotic
advection.
Dif-
fusive
and
thermodynamic
processes,
in
the
absence
of
gravity,
are
homogenizers.
Large-
scale
melting
can
be
a
homogenizer.
Plate
tec-
tonic
and
other
petrological
processes,
however.
create
heterogeneities.
Removal
of
small-degree
melts
causes
depletion
of
fertile
regions
(basalt
sources)-
or
components-
and
refractory
regions
(depleted
peridotites
,
cumulates).
Sm
.
all-degree
melts
are
enriched
in
LIL
and
become
the
crust
and
the
enriched
components
(EM)
in
the
upper
mantle
,
such
as
ldmberlite
and
carbonatites,
and
in
the
sources
of
enriched
magmas
such
as
EMORE
and
OIB.
Large-degree
melting
occurs
at
spreading
centers
and
at
thin
spots
of
the
litho-
sphere
from
upwelling
mantle
that
has
been
depleted
(NMORE
source)
or
enriched
(EMORB
source)
by
the
transfer
of
these
small
melt
frac-
tions.
Large-degree
and
large-volume
melts
blend
together
large-
and
small-degree
melts
from
a
large
volume
of
the
mantle
and
give
fairly
uniform
magmas
with
small
variance
in
trace-
element
and
isotopic
ratios
,
even
if
the
shallow
mantle
is
heterogenous
.
This
is
called
melt
aggre-
gatio
n
or
blending.
Melt
extraction
from
partially
molten
rocks
or
crystallizing
cumulates
is
not
100%
effi-
cient
and
residual
melts
help
explain
some
of
the
trace
element
and
isotopic
paradoxes
of
mantle
magmatism,
such
as
apparent
contradic-
tions
between
the
elemental
and
isotopic
com-
positions.
Other
sources
of
magmatic
diversity
include
recycling,
delamination
and
melting
of
diverse
lithologies
such
as
eclogite
and
peri-
dotite
,
which
also
have
experienced
various
levels
of
melt
extraction
and
infusion.
This
heteroge-
nous
upper
mantle
or
statistical
upper
mantle
assemblage
gives
relatively
homoge-
nous
products
when
it
experiences
large
degree
melting
.
Magmas
at
ridges
and
thin
spots
represent
blends
of
melts
from
vari-
ous
depths,
lithologies
and
extents
of
par-
tial
melting.
The
compositions
of
ocean
island
basalts,
ocean
ridge
basalts,
and
residual
mantle
reflect
upper
mantle
processes
of
melt
extraction,
migration
and
trapping
-
as
well
as
recycling
from
the
surface
-
and
the
sampling/melting