The most prominent physical discontinuity within the Earth is the core–mantle boundary, which separates the solid silicate mantle from the molten metal alloy comprising the outer core1,2,3. The lowermost mantle, also known as the D″ layer, is defined by seismic velocities as the few hundred kilometres above the core–mantle boundary and controls the heat and material flux across the boundary. Lowermost mantle structures and properties influence various dynamic processes, including mantle circulation and plate tectonics, outer core convection and the geomagnetic field, as well as inner core growth patterns1.

The lowermost mantle is spatially complex, displaying seismic heterogeneities on multiple lengthscales1,2,4 (Fig. 1), that have been linked to thermal or compositional variations as a consequence of heat and material transport. The region also exhibits variable anisotropy (the variation of seismic wave speed with propagation direction)1,2,5,6, which may arise from flow and deformation aligning its mineral constituents. Mineral phase transitions are invoked to explain the layer’s sharp upper boundary1,2.

Fig. 1: Illustration of fine-scale structures in Earth’s lower mantle.
figure 1

The core–mantle boundary (CMB), showing topography and a possible 2-km-thick melt layer, is represented by the red line (melt) with texturing (topography). ULVZs are depicted as red blobs on the CMB, with sizes on the order of tens to hundreds of km in width; the largest ULVZ (~900 km, beneath Hawaii4) is located at the base of an upwelling plume. The LLVP is several thousand km in lateral size3. The green dashed line marks the lowermost mantle (D″) upper boundary, at approximately 100–300 km height above the CMB, with varying strengths of anisotropy beneath (green shaded area). Blue lines indicate entrained basalt.

Full size image

The dominating seismic features in the lower mantle are ‘large low velocity provinces’ (LLVPs), in which seismic wave speeds are unexpectedly low1,2,3. There are two LLVPs — located beneath Africa and the Pacific Ocean — that extend approximately 1000 km in height from the core–mantle boundary. Laterally, they are comparable in size to continents, and comprise approximately 25% of the core–mantle boundary surface3. LLVPs are hotter and likely compositionally distinct to the bulk of the mantle. Dynamically, they appear to contribute to upwelling plumes, observed as surface volcanic activity1, and topography of the core–mantle boundary1,3.

The base of the mantle also contains ‘ultra-low velocity zones’ (ULVZs), in which the seismic velocities are extremely low, up to 50% slower than the bulk of the lower mantle1,2,4,7. ULVZs are typically tens of kilometres in height and width, with varying properties4,8, and are detected globally with inhomogeneous spatial distribution2. Their origins are debated4,7.

Finer-scale features are also detected globally in seismic data, and linked to various types of scattering heterogeneities, including recycled material from ancient subducted slabs9. An extremely thin layer of dense melt may exist atop the core–mantle boundary, as a result of interaction with the core, or fractionation of the lower mantle8,10.

At depths where observational measurements are limited to seismology, understanding the enigmatic heterogeneous lowermost mantle and its global dynamical influence has proven challenging. Advanced analytical techniques are required to resolve lowermost mantle structures and properties. Furthermore, to understand the global impacts of the deepest mantle, seismic measurements must be interpreted in combination with dynamical simulations, mineral physics calculations and experiments, and gravitational modelling.

Extracting more from seismic data

Seismological studies now benefit from extremely large datasets and dense regional coverage4,7. These massive datasets require sophisticated computational methodologies to identify and evaluate the small signals from the lowermost mantle. Recent technical developments allow more information from seismic data to be extracted via waveform modelling4,5,6,7,8,10, statistical analysis11,12, and integration with mineral physics13,14,15, to deliver increasingly detailed models of structures and properties. The following emerging methodological advancements hold promise for improving constraints on Earth’s lowermost mantle:

Forward modelling

Forward modelling describes the process of calculating theoretical seismograms for synthetic velocity models, to determine the influence of 3D structures on seismic waveforms. The method produces the full waveform response to the structures, revealing subtle effects on seismic signals such as distortion or additional features. It therefore provides substantially greater constraints on seismic properties (velocity, attenuation, anisotropy) as compared to traditional body wave studies, which measure discrete parameters such as travel times, amplitudes, and polarities4,5,6,7,8,9,10. An important contemporary development in lowermost mantle studies is the implementation of small-scale 3D velocity anomalies4,5,6,7,8 into waveform simulations, to capture the region’s complexities.

Seismic waves diffracted at the base of the mantle are deflected and distorted by small velocity anomalies including ULVZs and melt4,10. By utilizing very dense ray coverage facilitated by close-packed seismic installations, forward waveform modelling can be used to estimate the size of such anomalies, but this type of modelling is computationally expensive and has limitations in terms of resolution. A complementary technique employs 2D modelling, which allows for rapid analysis of synthetic models and the generation of probabilistic maps of ULVZ lateral size7. However, this 2D method is less useful for constraining ULVZ height. Combining 2D and 3D methodologies may provide a way forward for rapid analysis of fully 3D structures.

Shear waves are a combination of two orthogonal components (vertical and horizontal). When traversing an anisotropic material, these components split into two waves which travel with different velocities. The resultant time delay between the components indicates the strength of anisotropy1,2,5,6,8. Forward modelling of multiple combinations of different shear waves has proven effective to constrain regional variations in lowermost mantle anisotropy6,8. This highlights the importance of integrating multiple data types, particularly as 3D forward modelling incorporates both anisotropy and regional anomalies8. However, the current seismic coverage does not yet provide the necessary range of ray orientations to fully constrain the directional dependence of wave speed. With continuing expansions in available seismic data, the potential improved sampling of different directions means a comprehensive joint inversion of anisotropy and 3D structures is an upcoming possibility.

Tomographic inversions

Tomography is a long-established method for mapping whole-mantle seismic properties1,2,3. The procedure inverts travel times of large, global datasets of body waves to calculate variations in seismic velocity along their ray paths, producing regional scale velocity models. However, resolving the complicated, fine-scale structures of the lowermost mantle is challenging with conventional methods, due to limited and unevenly distributed ray paths traversing these depths3, and model limitations on parameters including cell size and location, damping, and smoothing11.

The relatively recent transdimensional probabilistic approach to deep Earth tomography12 is entirely data-driven, contingent on coverage, noise, and spatial heterogeneity, rather than user-provided parameters11,12. This method is especially useful for regions with uneven data coverage, and thus very well suited for the lowermost mantle. Although such models have a minimum resolution of approximately hundreds of kilometres and may not image the smallest features like ULVZs (ref. 11), they offer a marked improvement upon fixed parameterizations for mapping regional-scale heterogeneity12.

Future directions for lowermost mantle topography include incorporating forward modelling techniques to fully account for deflection and deviation of seismic waves, and thereby better resolve the complex structures5,11.

Need for multidisciplinary approaches

The integration of seismology with other deep Earth disciplines is crucial for interpreting material composition and the dynamic origins of observed structures. Seismic measurements are typically examined against published results from other disciplines and vice versa; however, discrepancies between different data types are common, and can be difficult to resolve2,3. Algorithms to automatically integrate multidisciplinary measurements within joint frameworks are emerging as prevalent and valuable analytical tools13,14,15.

Methods for comparing seismic and mineral parameters can be broadly categorized into three approaches: inverting seismic data to determine the best-matching mineral compositions13; forward modelling of compositions and textures to predict seismic parameters and signatures1,2,14; and extrapolating experimental mineral physics data to estimate material behaviour under high pressure and temperature conditions in the lowermost mantle15. Further integration of these analyses with geodynamical modelling is required to relate observations to flow patterns and convection history1,13.

The comparisons reveal inconsistencies in resultant models and interpretations2,13,14,15 and underscore the importance of interdisciplinary studies and the ongoing effort towards joint inversions that reconcile disparities.

Implications for interior composition and dynamics

Recent methodological advances have yielded a wealth of insights into the composition and dynamics of the lowermost mantle, as well as broader global implications. Strong, short-scale compositional heterogeneity is detected globally, including within the LLVPs11,12. ULVZs display highly variable sizes4,8, with intricate internal structures, and some of them act as roots for upwelling plumes4, potentially sourced from a layer of iron-enriched melt at the core–mantle boundary10. Such a thin melt layer remains indetectable by current seismic methods but could potentially account for geochemical anomalies in oceanic basalts from mantle plumes via ULVZs (ref. 8). This also implies that, at least in some locations, flow persists relatively unimpeded from the base of the mantle to the Earth’s surface.

ULVZs are co-located with anisotropy patterns that indicate strong vertical flow, supporting a connection between the lower mantle and upwelling plumes1,2,8,13, although present data and modelling cannot yet constrain the exact spatial correlation8. Joint analysis of seismological data and mineral physics finds non-unique solutions for lower mantle composition14, highlighting inconsistencies with existing mantle flow models13. Addressing these discrepancies will be crucial.

Advances in observational and analytical methodologies are the driving force for expanding our knowledge of the lowermost mantle. The development of new algorithms and approaches will be essential for efficiently analysing, modelling, and interpreting the growing volume of multidisciplinary data3,7,13,14,15. Current computational limitations are hindering some research goals, emphasizing the need for advancements in this area11,12,15. Near-future objectives include joint analyses of multiple seismic datasets, high-resolution modelling of complex 3D structures, and cohesive integration of seismology with mineral physics and geodynamics. Automated, data-driven, multidisciplinary investigations will be integral to progress in comprehensive models of the lowermost mantle and its implications for the Earth’s interior.