There appear to be substantial large-scale heterogeneities in
the D
region, with volumetric velocity
fluctuations of 
2% or larger. Our knowledge of velocity
dependence on temperature at such very high pressures is actually
very poor, but extrapolation of experimental constraints from lower
pressures suggests that excessively large variations, on the order
of 1000 K, would be required for a purely thermal explanation (e.g.
Yuen et al., 1993). Coupled with the poor correlation between P
and S velocity perturbations in many regions, it has become
commonplace to assert that there must be a chemical contribution to
the observed velocity heterogeneity. At present, this must be
deemed a soft conclusion, but even if we assume that it is accurate
it is unclear what the nature of the chemical heterogeneity must
be.
The geometry of fast velocity regions beneath the
circum-Pacific and the tendency for these regions to show the
clearest evidence for D
velocity discontinuities
points toward a causal relationship with mantle convection. The
notion that interaction with transition zone phase boundaries tends
to cause episodic flushing events of upper mantle slab
accumulations (e.g. Tackley et al., 1993) offers one means for
accumulating large-scale down-wellings of chemically differentiated
slab materials which may sink rather rapidly to the CMB, retaining
sufficient thermal anomaly to have faster than average seismic
velocities in the `footprint' of descending material. Chemically
distinct components in the slabs, particularly the eclogitic
crustal components may provide strong chemical contrasts that could
contribute to small-scale heterogeneities in D
,
presumably in combination with small-scale thermal irregularities
in the deep boundary layer. Gradual thermal equilibration of the
slab material could weaken the velocity anomalies, leaving only
weak, intermittent chemical contrasts. Other scenarios such as in
situ growth of chemical heterogeneities via core-mantle chemical
reactions can also be invoked, but suffer from lack of any
constraints on the volumes of material that can be produced or on
why there may be any relationship to the configuration of the
present day shallow mantle convection. A concerted effort to map
out the broadband heterogeneity spectrum of the D
region will eventually provide the information for assessing the
configuration of thermal and chemical heterogeneity in this region,
but the effort will be very challenging. Seismologists will need
to develop new processing methods, exploiting the full complexity
of secondary arrivals in broadband seismograms and using stacking
methods to enhance the signal-to-noise ratio for weak arrivals.
New three-dimensional modeling capabilities are also needed to
enable modeling of the complex signals from D
.
In addition, seismological results must be interpreted in the
context of independent constraints on the D
structure provided by other disciplines, if we are to resolve the
processes occurring near the CMB.
Acknowledgments. T. Lay's research on the core-mantle boundary is currently supported by the National Science Foundation under Grant EAR9305894. Ed Garnero and two anonymous reviewers gave useful editorial comments. This is contribution number 248 of the W. M. Keck Seismological Laboratory of the Institute of Tectonics, University of California at Santa Cruz.