Subducted slabs have been detected in the lower mantle for almost 30 years, yet the presence of foundered cratonic segments in the lower mantle is still unclear and inadequately investigated. We present the first P-wave radial anisotropy tomography of southern Africa (our model SA-RAnis2024), which reveals a contrasting feature of preserved northwest and modified southeast Kalahari cratonic root. Segments from the modified cratonic lithosphere are inferred to have dropped into the shallow lower mantle where seismic evidence of isolated high-velocity anomalies are observed. We detect such a high-velocity anomaly under the southwest margin of the Kalahari craton, which possibly detached from the southeast Zimbabwe craton at ca. 60 Ma based on plate reconstructions. Foundered segments can be partially brought back up to shallow depths, and contribute to the geochemical heterogeneity of younger lithosphere, through large-scale mantle convection.

Cratons are the stable and rigid geological units comprising over 60% of continental landmass, and their modification or destruction have been revealed to play a significant role in continental topographical expression and in the locations of mineral deposits like diamond, gold, and platinum (Hu et al., 2018). High-seismic-velocity anomalies have been found in the upper mantle and mantle transition zone (MTZ) beneath modified cratonic belts and their surrounding oceanic areas (e.g., Shen et al., 2018). These have been interpreted as fragments of detached lithosphere, but whether such blobs can sink into the high-viscosity lower mantle is not well understood because of the relatively low resolution of seismic imaging in the deep mantle. Extensive studies have been conducted within cratonic areas with a history of subduction such as the Wyoming craton (USA) and the North China craton (e.g., Bedle et al., 2021), but studies under cratons with no post-Paleozoic subduction history, especially high-resolution lower mantle tomography studies, are rare.

The absence of subduction beneath southern Africa over the past 500 m.y. (Kröner and Stern, 2004) makes it a natural laboratory to conduct such studies. Multi-episode large igneous intrusion events (Fig. 1) in southern Africa (Zhang et al., 2008; Celli et al., 2020) indicate that cratons present there (i.e., the Kalahari and Tanzanian cratons) have been modified to various degrees. Consistent with geochemical evidence, low seismic velocities reveal areas that are significantly affected by metasomatism. A recent continental-scale Vs model of the African lithosphere (Fig. 2C) reveals the occurrence of widespread lithospheric erosion and thinning under most of the African cratons over the past 200 m.y. (Celli et al., 2020). In contrast, a significant portion of the Kalahari craton is interpreted to retain a depleted and relatively thick lithosphere, as suggested by the latest high-resolution thermochemical model (Afonso et al., 2022). Such a difference is possibly attributed to different seismic data and methods employed. We used all available broadband seismic data to create the first P-wave radial anisotropic tomography model (SA-RAnis2024; Xi, 2024) of the mantle beneath southern Africa (south of 5°S) and further mapped preserved versus eroded lithosphere beneath the Kalahari craton (see the Supplemental Material1). The resulting radial anisotropy in our model serves as an efficient tool to determine the strength (amplitude) and direction (horizontal or vertical) of mantle flow or tectonic deformation.

Depleted mantle in cold, thick lithosphere results in high-velocity anomalies (HVAs) (Fig. 2B; Figs. S1 and S2 in the Supplemental Material). Diamondiferous kimberlites have been used extensively as proxies for detecting thick cratonic lithosphere at the time of their eruption (Celli et al., 2020), as their diamond load indicates a kimberlite origin in the diamond stability field. The stability of the cratonic lithosphere may be enhanced by the attachment of depleted oceanic lithospheric mantle (carrying sublithospheric diamonds) to the continental keel during supercontinent assembly (Timmerman et al., 2023). Such diamonds might subsequently be sampled by kimberlitic magmas that are triggered by lithospheric extension, particularly during supercontinent break-up (Gernon et al., 2023). A direct correlation between the strength and thickness of the lithosphere, kimberlite activity, and seismic velocities can be seen beneath the northwest and southeast parts of the Kalahari craton (Fig. 2B).

Our high-resolution SA-RAnis2024 model has detected several fine-scale features of isolated low-Vp anomalies (LVAs) inside the southeast Kalahari craton at 100–200 km depths whose locations are overlapping with the distributions of diamondiferous kimberlites (Fig. 2B) (Tappe e al., 2018; Özaydın and Selway, 2022), possibly indicating lithospheric modification. Available shear wave splitting null measurements (Silver et al., 2001), indicative of either an isotropic medium or a vertical mantle deformation, are consistently located within the observed isolated LVA regions (Fig. 2A). Fossil fabrics of ancient mantle intrusions, having brought the diamondiferous kimberlites to the surface, would tend to develop dominant anisotropy in a vertical orientation and thus can result in the observed null measurements or small azimuthal anisotropy. This is also supported by the relatively high conductivity associated with rising metasomatic fluids from three-dimensional magnetotelluric models of the southern African mantle (Özaydın et al., 2022). Areas with high surface relief on the southeast Kalahari craton (Fig. 2D; Fig. S4) are interpreted to have positive residual topography after removal of the continental crust (Steinberger, 2016). Since the Cretaceous, there is a strong correlation (Fig. 3D) among the unroofing events from thermochronology studies of the Kalahari craton (Stanley et al., 2013), the sedimentation rate of southern Africa (Guillocheau et al., 2012), and the frequency of kimberlite eruptions (Jelsma et al., 2009). This indicates that the aforementioned positive residual topography anomalies of southern Africa can mainly be attributed to cratonic modifications. The infilling of surrounding decompression-melted asthenosphere into the space generated by sinking dense cratonic segments (Fig. 4) would lead to positive residual topography (Hu et al., 2018). Such a mechanism has been globally proposed to explain the topographic anomalies observed in the other cratonic plateaus such as the Transantarctic Mountains in Antarctica (Shen et al., 2018).

The large topographic anomaly in the southern African Plateau has also been interpreted as the surface expression of the underlying mantle flow associated with the African superplume, which is usually imaged as a large LVA in the lower mantle (e.g., French and Romanowicz, 2015). However, our SA-RAnis2024 model displays generally positive radial anisotropy (indicative of a horizontal mantle flow; Figs. S1and S3), and the absence of widespread LVAs in the MTZ and shallow lower mantle (Figs. 2 and 3; Figs. S2 and S3), combined with the lack of hot thermal anomalies in the MTZ as inferred from the dominantly normal MTZ thickness (Fig. 2E) (Reed et al., 2016; Sun et al., 2018; Yu et al., 2020), jointly rule out the existence of active mantle upwelling in the shallow lower mantle. Furthermore, the short wavelengths of residual topography also preclude a lower mantle origin of surface topography, which should result in a wavelength of >1000 km (Hu et al., 2018). The African superplume, if it currently exists, is not contributing to the surface elevation of southern Africa.

The northwest portion of the Kalahari craton is characterized by continuously thick lithosphere highlighted by a prominent HVA at 200 km depth, relatively low conductivity in the magnetotelluric model (Özaydın et al., 2022), an absence of extensive diamondiferous kimberlites and magmatic activities, and generally negative residual topography (Fig. 2). Together, these are indicative of a preserved or intact cratonic lithosphere. Similar features are also revealed in two other regions (the Irumide Belt and the Mozambique Belt; outlined in Fig. 2B) where diamondiferous kimberlites are exposed at their edges, which may suggest preservation of a craton-like lithosphere.

Cratonic modifications have recently been proposed to widely occur under the African plate based on new shear-wave velocity (Celli et al., 2020) and thermochemical (Afonso et al., 2022) models of the upper mantle. However, the existence and origin of detached cratonic segments remains enigmatic due to the limited resolution with depth. Our SA-RAnis2024 model has detected isolated and prominent HVAs at the top of the lower mantle at 700–850 km depths (Figs. 2 and 3; Figs. S1–S3), which have been verified to be reliable and robust based on extensive resolution tests (Figs. S5–S18). These HVAs can be due to either the existence of fossil slabs or foundered lithospheric segments. The most recent subduction in this area can be dated back to the Pan-African Orogeny at 500 Ma (Kröner and Stern, 2004). As the slab sinking rate is globally estimated to be 1–5 cm/yr (Peng and Liu, 2022), with the period of a slab stagnation in the MTZ no more than 60 m.y. (Goes et al., 2017), the most likely candidate for explaining the HVAs in the lower mantle is foundered lithospheric segments from most recent cratonic delamination.

Given the absence of widespread isolated HVAs in the lower upper mantle (300–400 km depths) (Fig. S2), the observed HVAs in the top of the lower mantle may represent the foundered lithospheric segments from the final phase of cratonic delamination at ca. 60 Ma, after which there seems to be no further cratonic modification. Such a timing is inferred from the absence of a Cenozoic hotspot (Fig. 1), the termination of extensive kimberlite eruption in southern Africa (Fig. 3D; Jelsma et al., 2009), and the synchronous cessation of topography response to lithospheric delamination (Hu et al., 2018) as suggested by the abruptly dropping rates of denudation, unroofing, and offshore sedimentation (Guillocheau et al., 2012; Stanley et al., 2013). This inference is also consistent with the occurrence of most recent lithospheric thinning under the Kalahari craton in the late Mesozoic, as revealed from mantle xenolith samples (Janney et al., 2010) and joint analysis of kimberlites and velocity anomalies (Celli et al., 2020). We have tried to determine the original locations of these HVAs, taking the one under the southwestern margin of the Kalahari craton as an example. Plate reconstructions from three different models (Heine et al., 2013; Müller et al., 2016; Torsvik and Cocks, 2017) point to a similar position at the southeast Zimbabwe craton (Fig. 2F; Fig. S19) where there is extensive kimberlite activity before 80 Ma and a missing cratonic root (Fig. 3A). Thus, the foundered lithospheric segment under the southwestern margin of the Kalahari craton possibly resulted from the cratonic modification in the Late Cretaceous and subsequent detachment of the lithospheric segment from the southeast Zimbabwe craton at ca. 60 Ma. If we assume that this cratonic segment detaches from a depth of 100 km (Fig. 3A), the sinking rate is estimated to be ~1.2 cm/yr. This is comparable to the minimum rate for currently sinking slabs (Peng and Liu, 2022) and numerical models of cratonic delamination (Wang et al., 2022b).

Negative buoyancy is necessary to make these cratonic segments sink into the deep mantle, especially when crossing the 660 km phase boundary (Stixrude and Lithgow-Bertelloni, 2011). The cratonic lithospheric mantle is 0.5%–1.23% denser than the asthenospheric mantle based on analysis of whole lithosphere isostasy (Lamb et al., 2020), topography, and gravity anomalies (Wang et al., 2022a), providing a self-sustained negative buoyancy from gravitational instability. In addition, multiple ancient magmatic events (Fig. 1) have contributed to the refertilization of lithosphere under the southeast Kalahari craton (see the Supplemental Material), in which case cratonic segments with fertile peridotite composition or decreasing depletion degrees tend to sink into the lower mantle as suggested by thermo-chemo-mechanical modeling (Wang et al., 2022b). Some portions of these cratonic segments can be brought back up to shallow depths through mantle convection at hotspots or mid-ocean ridges (Fig. 4). Foundered materials from the Archean Kaapvaal craton have been discovered at the nearby ultraslow-spreading Southwest Indian ridge (Liu et al., 2022).

Our new P-wave radial anisotropy and velocity models of southern Africa show a continuous high-velocity anomaly in the lithosphere of the northwest Kalahari craton, suggesting a preserved lithospheric root. In contrast, the southeast Kalahari craton has been modified based on the isolated low-velocity anomalies in the lithosphere, combined with coherent distributions of diamondiferous kimberlites, shear wave splitting null measurements, and positive residual topography. We find evidence of high-velocity anomalies in the uppermost lower mantle that we interpret as delaminated lithosphere based on plate reconstruction and timing of the kimberlitic volcanism.

1Supplemental Material. Supplemental text and Figures S1–S27. Please visit to access the supplemental material; contact with any questions.

We are grateful to Tim Stern for helpful suggestions on this manuscript. We thank Guoming Jiang, Sinan Özaydın, Bernhard Steinberger, and Trond Torsvik for providing us the modified multi-channel cross-correlation (MMCC) code, kimberlite data, residual topography data, and plate motion model, respectively. Constructive comments from Suzette Timmerman, Nicolas Celli, one anonymous reviewer, and Urs Schaltegger (the editor) greatly improved the manuscript. This research is supported by the National Natural Science Foundation of China (grants 42374054 and 42074052) and the National Key R&D Program of China (grant 2023YFF0803202), and partially funded by the Shanghai Rising-Star Program (grant 22QA1409600), Shanghai Pilot Program for Basic Research, and the Chinese Fundamental Research Funds for the Central Universities.

Gold Open Access: This paper is published under the terms of the CC-BY license.