
The plume undergoes partial melting as it rises, the melt escapes to the
surface, and the solid residue that remains in the plume becomes
progressively depleted in easily fusible components. This process
results in progressive change in the composition of the residue, from
fertile lherzolite at the first, high-pressure stage of melting, to highly
refractory dunite at the final low-pressure stage. As a result of a process
that is not well understood, the residues of melting then accumulate
near the surface to form the subcontinental lithospheric mantle.
There are several obvious advantages to this model: (a) the
composition of the residue ranges from relatively Fe-rich garnet
lherzolite at the base of the melting column to highly refractory Fe-
poor dunite at the top. If incorporated into the lithosphere, the vertical
distribution of lithologies, from relatively dense at the base to buoyant
at the top, is isopycnic, at least qualitatively. (b) If the plume is hot
enough and the melting column long enough, the most refractory
residues, which are produced at the top of the column, will contain
very Fo-rich olivine (±orthopyroxene) whose composition is very like
that in old subcontinental lithospheric mantle. (c) Because the
extraction of melt removes volatiles, the residue is anhydrous. In
other words, melting in a hot mantle plume is capable of producing
the low-density, gravitationally stable, high viscosity material that
assures its long-term stability of the lithosphere.
Lee (2006) criticized two aspects of the model. First he notes that
melting at depth in the lower part of the melting column leaves garnet
in the residue. Through his quantitative modeling in which he
assumed that fertile lherzolite underwent isobaric equilibrium partial
melting, he showed that the residues of high-pressure melting contain
high FeO, Al
2
O
3
and Sc contents. In contrast, peridotites from old
subcontinental lithospheric mantle contain relatively low FeO, Al
2
O
3
and Sc contents, features that correspond either to melting at shallow
depths under conditions in which garnet is absent or to secondary
processes, such as orthopyroxene addition, that decreased the
contents of FeO and the other elements. Second, he notes that the
generation of a large volume of refractory Fe-poor dunite requires the
extraction of a large volume of high-degree melt. This melt would
have the composition of a komatiite, a type of magma that forms only
a small fraction of the Archean volcanic sequences interpreted as the
products of melting in mantle plumes. These aspects of the plume
model are discussed below.
Bernstein et al. (1998) note that the Fo93 peak in abundance plots
from Greenland xenoliths coincides to the extent of melting required
to eliminate orthopyroxene from the residue. At higher degrees of
melting, the melt productivity drops drastically; i.e. the amount of
melt produced for a given increase in temperature decreases
markedly. This effect may explain the peak in olivine compositions
in the range Fo92–94.
3.2. Accretion and stacking of oceanic lithosphere
In this model, advocated originally by Helmstaedt and Schulze
(1989), the subcontinental lithospheric mantle is proposed to have
grown through the accretion of slabs of oceanic lithosphere. The idea
is that portions of lithosphere that originally formed at a mid-ocean
ridge were thrust one beneath another in a subduction zone at the
margin of the growing continent, as shown in Fig. 1b.
The advantages of this model are: (a) it accounts for the presence
within suites of mantle xenoliths of eclogite and garnet pyroxenite,
which, in some cases, have geochemical and isotopic characteristics
that point to their having formed as old oceanic crust (e.g. (Fung and
Haggerty, 1995; Rollinson, 1997; Barth et al., 2001). (b) It explains the
presence of dipping seismic reflectors at the edges of some cratons
(Bostock, 1998; Levander et al., 2006). (c) It is consistent with the
inferred low-pressure origin of cratonic peridotites. Stacking of a
series of slabs made up largely of low-pressure peridotite thereby
provides a means of generating a large volume of subcontinental
lithospheric mantle.
Lee (2006) discussed a major problem of the model, a problem that
centers on the wide dispersion of lithologies and compositions in
oceanic lithosphere. The mantle portion of modern oceanic litho-
sphere is made up of rocks ranging from fertile, Fe-rich garnet- or
spinel lherzolite at the base, to harzburgite at the top (Fig. 2a). The
crustal portion is also stratified, from gabbros and Fe-rich olivine-
pyroxene cumulates in the lower part, to basalt in the upper part. The
fraction of harzburgite and dunite is low (b10%) and material with the
composition of Fe-poor cratonic peridotite is absent. In modern
lithosphere, the proportion of oceanic crust is about 10% (6–9 km thick
crust overlies 60–100 km of lithospheric mantle), significantly higher
than the proportion of eclogite and garnet pyroxenite in most parts of
the subcontinental lithospheric mantle. With such a high proportion
of garnet-rich lithologies it is unlikely that lithosphere formed by
stacking of slabs of oceanic plates would have been sufficiently
buoyant to have survived.
Lee mentions two possible solutions: (i) the more Fe-rich portions
of the oceanic lithosphere could have been removed before or during
accretion; (ii) Archean oceanic lithosphere was derived from hotter,
and perhaps more depleted Archean mantle (Davies, 1992) and it
would have had a different composition from modern oceanic
lithosphere. It would have contained a high proportion of Fe-poor
Fig. 3. Sketch of a subduction zone showing how material in the mantle wedge is drawn down through the melting zone to produce a Fo-rich low density residue at depth. This
material is overlain by denser, more fertile peridotite and by still denser cumulates in sub-crustal magma chambers. Redistribution of lithologies is needed to produce a
gravitationally stable configuration.
64 N.T. Arndt et al. / Lithos 109 (2009) 61–71