Seismic evidence for a 1000-km mantle discontinuity under the Pacific

Seismic discontinuities in the mantle are indicators of its thermo-chemical state and offer clues
to its dynamics. Ray-based imaging methods, though limited by the approximations made, have
mapped mantle transition zone (MTZ) discontinuities in detail, but have yet to offer definitive
conclusions on the presence and nature of mid-mantle discontinuities. We use a waveequation-
based imaging method to image both MTZ and mid-mantle discontinuities, and
interpret their physical nature. We focus on precursors to the surface-reflected seismic phases
PP, SS, PS, and SP to produce images of deep reflectors using reverse-time migration (RTM),
employing the full-waveform tomographic model GLAD-M25 for wavefield extrapolation. Our
adjoint-based inverse modeling accounts for more of the physics of wave propagation than raybased
stacking methods, which leads to improved accuracy and realistic precision of the
obtained images. The relative amplitude and location of the imaged reflectors are indeed well
resolved, but an interpretation of absolute amplitudes in terms of reflection coefficients
remains elusive. We observe a thinned mantle transition zone southeast of Mauna Loa, Hawaii,
and a reduction in impedance contrast around 410 km depth in the same area. These
observations coincide with anomalously low S-wavespeeds in the background tomographic
model, suggesting a hotter-than-average mantle in the region. Our new images furthermore
reveal a 4000—5000 km-wide reflector in the mid mantle below the central Pacific, at 950—
1050 km depth. This discontinuity displays strong topography and is marked by a polarity
opposite to that of the 660-km discontinuity, implying an impedance reversal near 1000 km.
We speculate that this mid-mantle discontinuity is linked to the mantle plumes rising from the
large low shear-velocity province (LLSVP) at the base of the mantle below this region. Some
seismic tomography models are in support of this interpretation, while others remain at odds---
a discrepancy that our observations may help resolve.