Seismic evidence for a 1000 km mantle discontinuity under the Pacific

The study seeks to determine the presence, geometry, and physical nature of mid-mantle discontinuities (~950–1050 km depth) beneath the central Pacific and to reassess mantle transition zone (MTZ) structure (410 and 660 km discontinuities). While MTZ discontinuities are globally recognized and linked to olivine system phase transitions with temperature-sensitive Clapeyron slopes, mid-mantle discontinuities have been observed intermittently and lack a definitive explanation. Traditional ray-based imaging (e.g., receiver functions, CCP stacking) can misplace structures due to simplifying assumptions, and transmission tomography is less sensitive to horizontal layering or sharp impedance contrasts, particularly beneath isolated oceanic stations where arrivals are steep. The authors propose a wave-equation-based reverse-time migration (RTM) imaging of surface-reflected precursors (PP, SS, PS, SP) to more accurately image impedance contrasts, aiming to clarify mid-mantle structure under the central Pacific, especially near the Hawaiian hotspot, and to interpret these features in terms of mantle dynamics and thermo-chemical state.

Prior work has established global MTZ discontinuities at ~410 and ~660 km related to olivine polymorph transitions (with positive and negative Clapeyron slopes, respectively) and variable MTZ thickness sensitive to temperature. Mid-mantle discontinuities (~900–1100 km) have been reported in subduction zones (e.g., Indonesia, South America, NE China) and interpreted as slab stagnation, but similar features have also been detected beneath upwelling regions and stable cratons, challenging a single-process explanation. Around Hawaii and Iceland, receiver-function and underside-reflection studies have suggested reflectors near ~1000–1050 km that often coincide with the top of low-velocity anomalies in tomography, though with polarity inconsistencies between methods. Tomography beneath Hawaii shows model-dependent structures with velocity reversals in some models and limited sensitivity to sharp layering at mid-mantle depths. These discrepancies motivate full-waveform methods that account for 3-D heterogeneity and finite-frequency effects.

• Data: 600 earthquakes (Mw 5.5–7.2) from the global CMT catalog were relocated and had focal mechanisms updated using GLAD-M25. Records from 8,642 seismic stations yielded 838,669 station–event pairs with at least one component, with dense illumination from USArray across the central Pacific.
• Phase windows: Three-component seismograms were windowed to include PP, SS, PS, and SP precursors to the 410 and 660 km discontinuities. Events at epicentral distances 70°–180° were selected to minimize interference from transmitted phases. For each target phase, 300 s windows centered on PREM-predicted arrivals (computed with TauP) were used with cosine tapers.
• Forward and backward modeling: Wavefields were simulated and back-propagated using the spectral-element method in the 3-D transversely isotropic Earth model GLAD-M25, incorporating ellipticity, self-gravitation, rotation, ocean loading, and attenuation. RTM steps: (1) forward model from sources through GLAD-M25; (2) back-propagate time-reversed reflections; (3) apply an impedance-kernel imaging condition (zero-lag cross-correlation of forward and backward wavefields) to yield interpretable images of impedance contrasts.
• Imaging condition and interpretation: Reflectors are identified at zero-crossings between alternating pulses (wavepackets) in the migrated image; polarity indicates the sign of the impedance contrast (product of density and seismic velocity). Due to finite-frequency and propagation effects, pulse pairs broaden with depth. Vertical resolution is ~¼ wavelength: ~75 km at 410 km, ~85 km at 660 km, and ~100 km at 1000 km; hence a 520 km discontinuity is unlikely to be separable in these data.
• Amplitude handling: The imaging operator is a preconditioned adjoint and does not yield absolute true-amplitude contrasts. Relative impedance contrasts were estimated via amplitude rescaling factors derived from synthetic tests in known models (1-D PREM and 3-D GLAD-M25), enabling comparison of the 410, 660, and ~1000 km reflectors.
• Coverage and quality control: Source–receiver midpoint binning (100×100 km^2) showed peak coverage in the central Pacific. SS precursors dominate the final image, with PS/SP contributing additional illumination at bounce points. Synthetic tests confirmed sufficient illumination and assessed sensitivity to mid-mantle structures.
• Analysis products: Vertical cross-sections along the Hawaiian seamount chain and surrounding corridors; extraction of reflector topography and image amplitudes for 410, 660, and ~1000 km horizons; comparison with GLAD-M25 shear-wave speed depth slices; masking of poorly illuminated areas using midpoint counts as a proxy.
• Computational aspects: Global-scale elastic simulations and RTM required substantial computational resources and storage; details and flowchart provided in Supplementary Information.

• Imaging outcome: Three impedance discontinuities are robustly imaged beneath the central Pacific: the MTZ boundaries near 410 and 660 km, and a mid-mantle reflector at ~950–1050 km. The 1000 km reflector exhibits stronger topography and broader apparent width (more diffuse contrast) than the MTZ discontinuities.
• Polarity and impedance contrasts: The ~1000 km reflections have polarity opposite to the 660 km reflections, implying an impedance reversal (a decrease in impedance with depth across the discontinuity). After amplitude correction, the ~1000 km reflector amplitude is about one quarter of the 660 km reflector. Under Hawaii, relative impedance contrasts are estimated as: 410 km ~0.48 of 660 km; 1000 km ~0.23 of 660 km.
• Geographic extent: The mid-mantle discontinuity forms a 4000–5000 km-wide reflector below the central Pacific, aligned with the Hawaiian seamount chain, fading toward the northwest and disappearing more abruptly north, northeast, and east of Mauna Loa. Its depth varies between ~950 and ~1050 km, with deepening to ~1050 km southeast of Mauna Loa.
• MTZ structure and thermal implications: The MTZ is thinned by ~30 km beneath and southeast of Mauna Loa (upper boundary near ~425 km; lower boundary near ~650 km), consistent with a positive thermal anomaly of ~200 K relative to ambient mantle. A reduction in impedance contrast around ~410–435 km southeast of Mauna Loa suggests local ponding of hot material and/or melt. The lower MTZ boundary shows more subdued topography than the upper boundary; depth variations of the two are weakly correlated except near Hawaii.
• Consistency with tomography: GLAD-M25 shear-wave speed slices show low-velocity anomalies co-located with the thinned MTZ and the deeper expression of the ~1000 km discontinuity, supporting a thermal origin. The negative-polarity mid-mantle reflections agree in sign with velocity changes in some tomographic models showing velocity reversals at these depths.
• Data and coverage: 600 earthquakes, 8,642 stations, and 838,669 station–event pairs provided dense illumination; SS precursors dominated the final constraints. Synthetic tests verified sensitivity to mid-mantle structure and supported interpretations of reflector diffuseness and polarity.

The results address longstanding uncertainty about the presence and nature of mid-mantle discontinuities beneath the central Pacific. By using RTM with full-wavefield physics, the study confirms a laterally extensive, topographically variable discontinuity at ~1000 km depth with negative polarity relative to the 660 km reflector, indicating an impedance decrease with depth. This polarity and the spatial correlation with low shear-wave speeds are consistent with the top of a hot upwelling (plume) or a system of deflected plume branches rather than slab-related stagnation in this region. The observed MTZ thinning and reduced 410 km impedance contrasts near Hawaii further support a high-temperature anomaly. Compared to prior receiver-function studies that often reported polarity inconsistent with tomographic velocity changes, these RTM images reconcile the sign of reflectivity with velocity anomalies in some tomography models, suggesting improved physical fidelity when finite-frequency wave propagation and 3-D structure are accounted for. The findings imply that mid-mantle discontinuities can mark thermal and compositional boundaries associated with plume dynamics, with implications for mantle circulation pathways beneath hotspots.

This study demonstrates that reverse-time migration full-waveform imaging in a 3-D Earth model can robustly map both MTZ and mid-mantle impedance discontinuities. Beneath the central Pacific, the MTZ is thinned by ~30 km and exhibits reduced impedance contrast near 410 km southeast of Mauna Loa, consistent with elevated temperatures. A broad (~4000–5000 km) mid-mantle discontinuity between ~950 and 1050 km depth displays strong topography and negative polarity relative to the 660 km discontinuity, indicating an impedance reversal. The spatial relationship with low shear-wave speed anomalies supports interpretation as the upper boundary of deflected mantle plume material. These results strengthen the case that mid-mantle discontinuities can be linked to plume dynamics in upwelling regions. Future work should include iterative waveform-based inversion to refine reflector amplitudes and geometries, expanded coverage with additional arrays (especially ocean-bottom seismometers), sensitivity tests across frequency bands to improve vertical resolution, and integration with mineral physics and geodynamic modeling to constrain the thermochemical origins of the ~1000 km discontinuity.

• Amplitude accuracy: The imaging is based on a preconditioned adjoint operator and does not yield absolute true-amplitude contrasts; relative amplitudes are inferred via synthetic-derived scaling.
• Resolution limits: Long-period data constrain vertical resolution to ~75–100 km, limiting the ability to resolve closely spaced features (e.g., 520 km discontinuity, or a potentially split 660 km boundary).
• Background model dependence: RTM relies on the accuracy of the 3-D model (GLAD-M25) for wavefield extrapolation; errors can affect reflector placement and amplitude.
• Data coverage: Although central Pacific illumination is strong (especially from USArray), coverage is uneven globally and some regions are masked; interpretations are focused where midpoint counts are sufficient.
• Computational demands and noise sensitivity: Global-scale RTM is resource-intensive; precursor amplitudes are small and susceptible to noise, making iterative inversion challenging.
• Phase sensitivity: SS precursors dominate constraints; contributions from PP/PS/SP may vary spatially, potentially biasing illumination angles at depth.