The large low shear‐wave velocity provinces (LLSVP) are thermochemical anomalies in the deep Earth's mantle, thousands of km wide and ∼1800 km high. This study explores the hypothesis that the LLSVPs are compositionally subdivided into two domains: a primordial bottom domain near the core‐mantle boundary and a basaltic shallow domain that extends from 1100 to 2300 km depth. This hypothesis reconciles published observations in that it predicts that the two domains have different physical properties (bulk‐sound versus shear‐wave speed versus density anomalies), the transition in seismic velocities separating them is abrupt, and both domains remain seismically distinct from the ambient mantle. We here report underside reflections from the top of the LLSVP shallow domain, supporting a compositional origin. By exploring a suite of two‐dimensional geodynamic models, we constrain the conditions under which well‐separated “double‐layered” piles with realistic geometry can persist for billions of years. Results show that long‐term separation requires density differences of ∼100 kg/m3 between LLSVP materials, providing a constraint for origin and composition. The models further predict short‐lived “secondary” plumelets to rise from LLSVP roofs and to entrain basaltic material that has evolved in the lower mantle. Long‐lived, vigorous “primary” plumes instead rise from LLSVP margins and entrain a mix of materials, including small fractions of primordial material. These predictions are consistent with the locations of hot spots relative to LLSVPs, and address the geochemical and geochronological record of (oceanic) hot spot volcanism. The study of large‐scale heterogeneity within LLSVPs has important implications for our understanding of the evolution and composition of the mantle.

Large low shear velocity provinces (LLSVPs), whose origin and dynamic implication remain enigmatic, dominate the lowermost mantle. For decades, seismologists have created increasingly detailed pictures of the LLSVPs through tomographic models constructed with different modeling methodologies, data sets, parametrizations and regularizations. Here, we extend the cluster analysis methodology of Lekic et al., to classify seismic mantle structure in five recent global shear wave speed (VS) tomographic models into three groups. By restricting the analysis to moving depth windows of the radial profiles of VS, we assess the vertical extent of features. We also show that three clusters are better than two (or four) when representing the entire lower mantle, as the boundaries of the three clusters more closely follow regions of high lateral VS gradients. Qualitatively, we relate the anomalously slow cluster to the LLSVPs, the anomalously fast cluster to slab material entering the lower mantle and the neutral cluster to ‘background’ lower mantle material. We obtain compatible results by repeating the analysis on recent global P-wave speed (VP) models, although we find less agreement across VP models. We systematically show that the clustering results, even in detail, agree remarkably well with a wide range of local waveform studies. This suggests that the two LLSVPs consist of multiple internal anomalies with a wide variety of morphologies, including shallowly to steeply sloping, and even overhanging, boundaries. Additionally, there are indications of previously unrecognized meso-scale features, which, like the Perm anomaly, are separated from the two main LLSVPs beneath the Pacific and Africa. The observed wide variety of structure size and morphology offers a challenge to recreate in geodynamic models; potentially, the variety can result from various degrees of mixing of several compositionally distinct components. Finally, we obtain new, much larger estimates of the volume/mass occupied by LLSVPs—8.0 per cent ±0.9 (μ ± 1σ) of whole mantle volume and 9.1 per cent ±1.0 (μ ± 1σ) of whole mantle mass—and discuss implications for associating the LLSVPs with the hidden reservoir enriched in heat producing elements.

The high degree of inter-model agreement shown in the vote maps and cross sections may be misinterpreted to suggest VS (or VP ) uniformity within the clusters. In fact, velocities vary within the fast and slow families, across different anomalies, and with depth in the mantle. To give the reader a sense of these variations, we plot contours for m=4 on top of the mean velocities across the five S wave tomographic models used here and mask out regions with disagreement across votes ( Figures S4, S5 and S6).