Chemical Composition of the Earth’s Lower Mantle: Constraints from Elasticity - Physics and Chemistry of the Deep Earth - page 184

(Mg, Fe) 2 SiO 4 . Since isochemical phase transi-
tions in (Mg, Fe) 2 SiO 4 will explain the existence
of global seismic discontinuities at

et al ., 1991; Fukao et al ., 1992; van der Hilst et al .,
1997). Detailed 3-dimensional numerical model-
ing of the mantle convection with an endothermic
phase change at the boundary between upper and
lower mantle demonstrated that the convection
style could be dominated by accumulation of
downgoing slabs at the boundary, resulting in
frequent avalanches of the stagnant slabs into
the lower mantle (Tackley et al ., 1993). This
dynamical image of the mantle may help to inter-
pret the global seismic tomography observations.
More importantly, Christensen and Yuen (1984)
pointed out that the flow pattern may evolve from
layered to whole mantle convection with cooling
history of the Earth's interior, based on the nu-
merical simulation including the effect of phase
change and chemistry on the density difference at
the upper and lower mantle boundary. Therefore,
elucidation of the chemical composition of the
lower mantle also provides an essential key to
understanding the evolution of the mantle con-
vection style and its mixing efficiency throughout
the Earth's history.
Due to the lack of tangible natural samples
directly derived from the lower mantle, the
chemical composition of the lower mantle must
be inferred from indirect geophysical observa-
tions and/or experimental data. Our knowledge of
the lower mantle primarily comes from seismic
observations, which are in principle a rich source
of information such as the density and elastic
properties of the Earth's interior (Dziewonski
& Anderson, 1981). Direct comparison of seis-
mological and laboratory-based elasticity data
can therefore give us one of the most severe
constraints on the mantle mineralogy. For this
purpose, densities/bulk moduli of the lower
mantle minerals have been extensively measured
under relevant high-pressure and temperature
conditions. Based on the density data of silicate
perovskite and feropericlase at simultaneous
high temperature and pressure (Stixrude et al .,
1992) proposed the perovskitic lower mantle
model. They also demonstrated that the trade-off
relationship between composition, temperature
and elastic modulus, showing that the density is
much more insensitive to the mineral proportion

∼ 410 and

∼ 660 km (Ito & Takahashi, 1987), the chem-

ically homogeneous pyrolite model has so far
been considered as the lower mantle model. How-
ever, such Mg/Si ratio is significantly higher than
that of chondritic meteorites of

1.0, which is
usually assumed to be similar in composition
to the whole Earth. The apparent depletion of
Si in the mantle has provoked intense debates
that the upper mantle is not representative of
the entire mantle, or is balanced by the rela-
tive Si-enrichment in the lower mantle or the
core. In contrast, crystal fractionation in an early
terrestrial magma ocean strongly suggests the
possibility of chemical stratification of the man-
tle (Allegre et al ., 1995; Stevenson, 1990; Tonks
& Melosh, 1993; Ohtani, 1985), with a distinct
perovskite-rich lower mantle for which the Mg/Si
ratio is

∼

1.0, although it is still a matter of
debate as to whether such primordial chemical
stratification is preserved throughout the subse-
quent solid-state convection process until present
day. The perovskite-rich lower mantle model is
markedly different from the conventional peri-
dotitic model and it significantly changes our
view of Earth's mantle and its evolution, it is
however challenging from both the seismological
and laboratory mineral physics perspectives to
identify as to whether the chemical stratification
is preserved in the current mantle.
A global seismic tomography technique is revo-
lutionizing our knowledge of the Earth's interior,
providing us the snapshot of the large scale seis-
mic structure of the present-day mantle. How-
ever, decades of research and spirited debate have
not led to a consensus on the scale of mantle
convection. The improved tomographic methods
based on P-wave travel times using a more real-
istic background Earth model or surface-reflected
seismic phases show that slabs in the Western
Pacific (beneath the Japan and Izu Bonin island
arcs) tend to be stagnant at the boundary between
upper and lower mantle, whereas those beneath
northern Kuril and Mariana arcs, and Central
America sink into the lower mantle (van der Hilst

∼