Multistep evolution of harzburgitic mantle underneath pipe 200 kimberlite, northern Lesotho: a study on xenoliths and their implication on diamond-barren nature of pipe 200 kimberlite

Abstract

The Pipe 200 kimberlite in northern Lesotho on the southeast margin of the Kaapvaal Craton is a diamond-poor deposit despite its proximity to economically viable kimberlites like the Liqhobong kimberlite. We study harzburgite xenoliths' mineral composition, geochemistry, and pressure–temperature evolution to understand factors influencing diamond destruction. The xenoliths are classified into five types based on their petrography and geochemistry. The diversity in the mineral assemblage correlates with a sampling depth of ~ 100 to 175 km (~ 2.8 to < 5.0 GPa). The signatures of metasomatism are evident in type 3 and 4 xenoliths, where garnet breaks down to form a cluster (henceforth assemblage) of phlogopite, chromite, and diopside. Fine-grained melts associated with an uplift in the mantle to shallower depths of < 90 km (< 2.3 GPa) encompass the minerals in the assemblage, which display resorption at the boundaries. Water contents (structural hydroxyl) of olivine and possibly orthopyroxene are lower in the xenoliths with metasomatism-induced breakdown of garnet. The structural hydroxyl distribution in the nominally anhydrous minerals shows flat distribution profiles of re-equilibration due to residence in the kimberlite magma. It is supported by the disruption of the inter-mineral water partition coefficient due to olivine water diffusion during residence in the kimberlite magma. The barren nature of the Pipe 200 kimberlite is attributed to the signatures of mantle metasomatism and residence in kimberlite magma, which led to the diamond destruction.

Graphical abstract

Introduction

Lithospheric diamonds form in the mantle at depths of approximately 150–200 km (Shirey et al. 2013) and are carried to the surface by deep-seated mantle magmas, such as kimberlites and lamproites (Mitchell et al. 2019). During ascent, kimberlite melt crystallizes primary magmatic minerals (Roeder and Schulze 2008), entrains, and interacts with mantle and crustal rocks (Gurney et al. 1991), and de-gasses CO₂ + H₂O dominated fluid (Wilson and Head III 2007). This changes the chemistry of the primary kimberlite melt (Russell et al. 2012). Further changes in chemistry caused by meteoric water during the late stages of magmatic crystallization cause the formation of secondary minerals (Dalton and Presnall 1998; Becker and Roex 2006). The mantle xenoliths entrained by the kimberlite provide evidence of the mineralogy, chemical composition, and physical state of the upper mantle, the xenoliths' evolution, and physical and chemical changes during ascent (Griffin et al. 2003). Diamonds trapped in mantle xenoliths can be liberated during kimberlite ascent and occur in kimberlites as xenocrysts. These diamonds, stable under the mantle pressure and temperature conditions, are brought to the surface by volatile-rich kimberlite magmas at rapid ascent rates, which prevent the destruction of diamonds (Sparks et al. 2006). Diamond dissolution in kimberlite magma is controlled by temperature, pressure, the magma ascent rate, the magma composition, and oxygen fugacity (fO₂) (e.g., Yamaoka et al. 1980; Robinson et al. 1986; Pal’yanov et al. 1995; Sonin et al. 2000; Fedortchouk et al. 2005, 2012). When present, the fluid phase in kimberlite also controls diamond dissolution wherein CO₂ and H₂O produce different surface features of resorption, and the amount of water controls the resorption morphology (Fedortchouk et al. 2007; Khokhryakov and Pal’yanov 2010).

Kimberlites are volatile-rich magmas, with H₂O being an important volatile component. In the mantle, H₂O primarily resides in the nominally anhydrous minerals (NAMs), such as olivines, pyroxenes, and garnets, in the form of hydrogen (H) attached to structural oxygen (O) in intrinsic crystal defects (Miller et al. 1987; Bell and Rossman 1992; Peslier 2010). Several studies have examined the concentration of OH defects in NAMs found as macrocrysts or inside xenoliths entrained in kimberlites to estimate the water (H₂O) content of the lithospheric roots of the Kaapvaal Craton and proposed a possible correlation between the OH content of NAMs and diamond preservation in kimberlites (Bell and Rossman 1992; Grant et al. 2007; Peslier 2010; Baptiste et al. 2012; Peslier et al. 2012, 2015; Doucet et al. 2014; Hilchie et al. 2014; Demouchy and Bolfan-Casanova 2016; Schmadicke et al. 2013). According to studies on the OH defects in the rims of these NAMs, primarily olivines, kimberlites with pronounced OH zoning in olivines have low diamond content (Hilchie et al. 2014). The core-to-rim OH zonation/diffusion in olivines indicates H loss. It has been used to estimate the ascent rate of kimberlites (5–37 m s⁻¹) and compared with that of slower alkali basalts (Peslier et al. (2008) and references therein). This observed OH zonation can be due to the slow emplacement of magma. The solubility of hydrogen in NAMs depends on the pressure, temperature, and fH₂O, among others (Mackwell and Kohlstedt 1990; Kohlstedt and Mackwell 1998; Demouchy and Mackwell 2006; Fei and Katsura 2020). During magma ascent and depressurisation, H diffusivity increases with a decrease in pressure, causing the olivine rim to have lower H contents than the core, suggesting that hydrogen is lost to the kimberlite magma (Peslier et al. 2008).

Water from NAMs is usually lost to kimberlite magmas, as indicated by the abundance of water loss profiles over water gain. Peslier and Luhr (2006) stated that the adding water to xenolithic olivines is improbable. The absence of diffusion-related OH profiles can imply the preservation of the original water content by NAMs or a complete re-equilibration between cores and rims due to diffusion. Most studies on the water concentration in NAMs (Bell and Rossman 1992, 2015; Grant et al. 2007; Peslier et al. 2010, 2012; Peslier 2010; Baptiste et al. 2012; Doucet et al. 2014; Demouchy and Bolfan-Casanova 2016; Schmadicke et al. 2013) focus on the H (structural hydroxyl) distribution, calculation of kimberlite ascent rates, and their correlation with diamond preservation, especially for economic kimberlites with variable diamond grades. The calculated water contents spread over a wide range (olivine: 0–86 ppm, orthopyroxene: 40–250 ppm, clinopyroxene: 150–400 ppm, garnet: 0–20 ppm) due to differences in the nature of metasomatic agents (hydrous, alkaline, siliceous, or ultramafic) (Peslier et al. 2012), temporal variations (Ingrin and Grégoire 2010), and sampling depths (Peslier et al. 2010; Doucet et al. 2014), among others. Despite diffusion, olivine cores preserve their mantle water contents (Peslier et al. 2008, 2010). The base of the Kaapvaal Craton lithosphere contains a water-poor layer characterized by dry olivines, whereas olivines derived from shallower depths are enriched in OH defects (Doucet et al. 2014). Using the diffusion profiles of kimberlitic olivines from the Kaapvaal Craton, Peslier et al. (2008) found minimum ascent rates of 5–37 ms⁻¹.

Many kimberlites punctuate the Kaapvaal Craton, which hosts over 1,000 (De Wit 2010) diamond deposits spreading across parts of the Republic of South Africa, southern Zimbabwe, southern Mozambique, southeast Botswana, Namibia, Swaziland, and northern Lesotho. The first account of kimberlite was by (Lewis 1898). Kimberlites from the Kaapvaal Craton are extensively studied and mined because of their large diamond deposits (Field et al. (2008) and references therein). The cluster of kimberlites from northern Lesotho forms a kimberlite province with a high density of kimberlites per unit area (Lock 1980). Four mines from northern Lesotho kimberlite pipes, namely Kao, Letšeng-le-Terae, Mothae, and Liqhobong (Fig. 1), produce some of the largest, most valuable lowest-grade diamonds. However, most other Lesotho kimberlites are barren, such as the Pipe 200 kimberlite (Figs. 1 and S1).

Fig. 1

Geological map of Lesotho showing Pipe 200 kimberlite (red circle) along with Karoo basalts of Drakensberg Group and a smaller portion, covered by clastic rocks of Stormberg and Beaufort Group. Other kimberlite pipes (gray circle) and kimberlite dykes (solid blue line) in northern Lesotho are also shown. Modified from (Shor et al. 2015; Rapopo and Sobie 2018)

One of the first descriptions of Pipe 200 kimberlite by Kresten and Dempster (1973) mentions its disparate textural characteristics (volcaniclastic and coherent) and describes the geology of the pipe and its association with the Malibamatso dike swarm. Carswell et al. (1977) classified the mantle peridotite xenoliths from the Pipe 200 kimberlite into five types (garnet lherzolite, garnet–chromite–lherzolite, chromite–lherzolite, chromite–harzburgite/–lherzolite, and spinel–harzburgite). The major element geochemistry of Pipe 200 kimberlite xenoliths indicates that the mantle beneath northern Lesotho is depleted in incompatible elements, including Al and Fe (Carswell et al. 1979). Geothermometry and geobarometry (Mitchell et al. 1980) findings on chromite and garnet harzburgite xenoliths show a narrow depth range of 90–110 km, and their geochemistry demonstrates the heterogeneous nature of the mantle beneath northern Lesotho over short distances. Hilchie et al. (2014) conducted a detailed study on H zonation in kimberlite-derived olivine macrocrysts and olivines in xenoliths from kimberlites from Lesotho (including Pipe 200) and Canada. They found zonation due to H depletion in the group 1 region (~ 3420–3,700 cm⁻¹) of the Fourier-transform infrared spectroscopy (FT-IR) spectrum of Pipe 200 olivines. This H zonation in olivines is attributed to a slow ascent, allowing H more time to diffuse out of the crystals.

The NAM structural hydroxyl distribution and evolutionary history of mantle xenoliths from uneconomic kimberlites should be studied concurrently to understand and distinguish their geochemistry and evolution from those of economic kimberlites. Uneconomic kimberlites with no or low diamond concentrations can be analyzed to help elucidate the evolutionary processes leading to diamond destruction and further support the findings of diamond preservation/destruction studies.

This study focuses on the petrographic observation, geochemistry (including the structural hydroxyl distribution in NAMs), and P–T evolution of harzburgite xenoliths from the barren, uneconomic Pipe 200 kimberlite, which sampled the mantle beneath northern Lesotho. By studying these uneconomic kimberlites, we aim to better understand the diamond destruction factors of evolution and explain the mantle processes that have led to the barren nature of the Pipe 200 kimberlite. We test whether the observed variations in its geochemistry and hydrogen concentration provide information about the petrological processes responsible for the barren nature of Pipe 200.

Geological setting

The Kaapvaal Craton (Fig. S1), located in southern Africa, covers an area of about 1.2 × 10⁶ km² and consists of four terranes: The Western, Southern, Central, and Pietersburg Terranes (Donnelly et al. 2012 and references therein). The Craton formed and stabilized at approximately between 3.7 and 2.6 Ga and represents an assemblage of tonalitic gneisses and early Archean granite–greenstone terranes intruded by different granitic plutonic rocks. These rocks are overlain by basins filled with thick sequences of sedimentary and volcanic rocks (Gregoire et al. 2005). In the north, the Late Archean Limpopo belt welded the Kaapvaal Craton to the Zimbabwean Craton at 2.8–2.5 Ga (de Wit et al. 1992; Gregoire et al. 2005). The Kaapvaal Craton is bounded south by the 1.2–2.0 Ga Proterozoic Namaqua Natal orogenic belt, formed via continental collision and arc accretion (Eglington 2006; Burness et al. 2020). To the west, it is bordered by the Khesi Orogenic Belt (Fig. S1; 2.0–1.7 Ga), and to the east, by the Lebombo Monocline (Donnelly et al. 2011). The kimberlite intrusions punctuated the Kaapvaal Craton between 1.8 Ga and 60 Ma, with main pulses around the Mesozoic (Jelsma et al. 2009).

The Lesotho Kimberlite Province lies at the margin of the Kaapvaal Craton, where kimberlites intrude Kaapvaal Craton rocks (Bloomer and Nixon 1973). Pipe 200 kimberlite lies in northern Lesotho, on the western bank of the Malibamatso River (Fig. 1). The only account of the estimated ~ 90 Ma age of the Pipe 200 kimberlite is that of Davis (1977), who used uranium contents in zircons for estimation. Pipe 200 contains a suite of ultramafic mantle and crustal xenoliths (Mitchell et al. 1980). It is characterized by multiple phases (intrusions) of kimberlite magmas, showing textural differences, such as magmatic and volcaniclastic-type kimberlite (Hilchie 2011). Two of the three kimberlite intrusions do not contain diamonds (Kresten and Dempster 1973).

Materials and methods

Sample preparation and petrography

The samples used in this study were collected during the 1st International Kimberlite Conference by Barrie Clarke in 1973. The sample preparation for FT-IR analyses is detailed in Hilchie et al. (2014). One to three thick (< 400 μm) sections were prepared by polishing both sides of each xenolith. The samples are described in Table 1. The thick sections chosen for FT-IR analyses contained unaltered, predominantly unfractured, inclusion-free clear minerals. The thick sections were removed from their glass slides via immersion in acetone and glue dissolution for FT-IR analyses.

Table 1 Xenolith thin and thick sections from Pipe 200 kimberlite, Kaapvaal Craton selected for this study

Thin and thick sections of each selected Pipe 200 xenolith (Table 1) were carefully studied using Brunel SP300P and Olympus BX41 optical microscopes at the Department of Mineralogy, Geochemistry, and Petrology, University of Szeged, Hungary, to determine their lithological types based on their mineral compositions and microtextures.

Micro-X-ray fluorescence (μXRF)

Energy-dispersive (EDS) multi-element μXRF mapping was performed on the xenolith thin sections at the Department of Mineralogy, Geochemistry, and Petrology, University of Szeged, Hungary, using a Horiba Jobin Yvon XGT-5000 X-ray fluorescence (XRF) spectrometer equipped with a Rh X-ray source. The measurement conditions were 30 kV beam voltage, 0.5 mA beam current, and 10 μm beam spot diameter. The intensity of each element was measured in counts per second (cps). The measured elements were Si, Ti, Al, Cr, Fe, Mn, Mg, Ni, Ca, K, and Ba (Yu et al. 2019).

Scanning electron microscopy (SEM)

The major element contents and compositional variations within and between the minerals were obtained from carbon-coated thin sections using a Quantax75 EDS–silicon drift detector combined with a Hitachi TM4000Plus microscope in the Faculty of Science Research and Instrument Core Facility, Eötvös Loránd University (ELTE), Budapest, at a 15-kV accelerating voltage, 833 pA beam current, and 15 s counting time. The measurements were carried out using the technique of Berkesi et al. (2020) to calculate the mass proportions (m/m %) of major elements in the cores and rims of the studied minerals. The difference in concentration measurement between the EPMA and EDS is 0.1 m/m % for major elements. To avoid edge problems, we avoided cracks. The analyzed elements were Si, Ti, Al, Cr, Fe, Mn, Mg, Ni, Ca, Na, and K. The measured mineral chemistry is reported in Supplementary Table S1, where the standard deviations are written in parentheses.

Electron probe micro-analyser (EPMA)

The olivines and orthopyroxenes in all xenolith types were analyzed by EPMA using a JEOL JXA 8,200 in the Chair of Resource Mineralogy, Montanuniversität Leoben, Austria, for thermobarometric calculations only. The polished sections were carbon coated (EmiTech K950X) to minimize charging under the electron beam. Spot measurements were conducted via WDS at a 50 nA beam current and 20 kV accelerating voltage, and the beam diameter was set to a spot size of ~ 1 μm. For all quantitative analyses through EPMA, Kα lines were used; the elements measured in the olivines and orthopyroxenes, used standards, detection limits, and peak and background counting times are summarized in Supplementary Table S2: sheet-conditions.

Following the recommendations of Nimis (2022) and references therein for precise Al-in-olivine content measurements, very high probe currents (200 nA) and long count times (100 s) were used. However, the Al₂O₃ content did not significantly exceed the detection limit of 35 ppm. Thus, the rest of the measurements were conducted using the earlier setting (50 nA beam current).

Fourier-transform infrared spectroscopy

FT-IR microscopic analysis was conducted on unoriented samples to study their structural hydroxyl distributions in the olivine, orthopyroxene, clinopyroxene, and garnet and calculate the water contents in the NAMs. Point and profile measurements were carried out in these nominally anhydrous minerals. The majority of analysis was performed in the Institute for Geological and Geochemical Research, Budapest, using a Bruker FT-IR Vertex 70 spectrometer equipped with a globar light source, and an MCT-A detector was coupled to a Bruker Hyperion 2000 microscope. Additional measurements were carried out at the Budapest University of Technology and Economics, Budapest, using a PerkinElmer Spectrum 400 infrared spectrometer with a coupled Spotlight 400 FT-IR imaging system. The double-polished thick sections had thicknesses of 256–429 μm (Table 1), and all analyses were conducted using unpolarised light. The olivine and both pyroxene measurements were collected between 4,000 and 400 cm⁻¹ using an aperture of 50 × 50 μm. These measurements were performed using a globar light source, a KBr beam splitter, and an MCT detector. Background and sample scans were conducted at 128 scans at 4 cm⁻¹ resolutions. The data correction and processing details are described in Appendix 1.

Thermobarometry

In this study, we obtained the pressures and temperatures of harzburgite xenoliths in this study using various methods. For type 5 xenoliths that contain diopsides and/or garnets, we used thermometers (Harley 1984; Brey and Köhler 1990; Taylor 1998; Nimis and Taylor 2000) and barometers (Brey and Köhler 1990; Taylor 1998; Nimis and Taylor 2000; Grütter et al. 2006). To compare all xenolith types, we calculated the temperatures using monomineralic olivine and orthopyroxene thermometers Brey and Köhler (1990) modified by Nimis and Grütter (2010) and Witt-Eickschen and Seck (1991) on the EPMA data. The median of all calculated temperatures (in °C) and pressures (in kbars), along with the interquartile range for all thermometers and barometers, is presented in Table 3.

Thermodynamic modeling

Geothermobarometric modeling was performed using Theriak-Domino (de Capitani and Petrakakis 2010) to determine the pressure and temperature conditions of the stable mineral paragenesis observed in the different microtextures of our samples. The program uses Gibbs free energy minimisation to determine equilibrium mineral assemblages. Of the many databases available for Theriak-Domino, we used the HoPomelt database, which allows using liquids (melts) in the computation. We used a simplified system comprising SiO₂–Al₂O₃–FeO–MgO–CaO–K₂O–H₂O. Other oxides (Na₂O, TiO₂, MnO, and BaO, among others) were omitted due to their small quantities and insignificant roles in the stability of the minerals in this system. Chromium (Cr₂O₃) is a significant oxide in the chemical system, but Theriak-Domino databases do not include Cr in the models. However, Al behaves similarly to Cr and substitutes Cr in the chemical system. Pseudosections were calculated using the measured bulk compositions over a large temperature range of 1,000 °C–1,500 °C at 15–45 kbars, which transgresses the diamond stability field.

Results

Petrography

The peridotites have the modal composition of harzburgite (Table 1) based on the classification of Streckeisen (1976). Following the classification of xenolith textures by Mercier and Nicolas (1975), all xenoliths have protogranular textures (Figs. 2, 3, 4, 5, 6). Nine xenoliths from Pipe 200 were classified into five types based on variations in the mineralogy and textural associations. The types are harzburgites (Type 1), chromite harzburgites (Type 2), chromite harzburgite with melt pocket (Type 3), garnet–chromite–harzburgite with melt pocket (Type 4) and garnet harzburgite (Type 5).

Fig. 2

Characteristic petrographic images of the type 1 (a, b) and type 2 (c, d) harzburgite showing protogranular texture in cross-polarized light (a, b, d) and plane-polarized light (c). Serpentine fills along the grain margins. a Poikilitic enstatite in large coarse granular olivine and in a large orthopyroxene host crystal. b Poikilitic orthopyroxene in large coarse granular olivine. c Holly leaf chromite in fractured olivine showing serpentine mesh. d Poikilitic olivine in large orthopyroxene grain and a large olivine crystal. Opx orthopyroxene, Ol olivine, Chr chromite

Fig. 3

Characteristic petrographic images of the type 3 harzburgite showing protogranular texture in plane-polarized light (a, b) and back-scattered electron images (c, d). a Chromite with small bright green clinopyroxene. b, c phlogopite flake, chromite, and tiny olivine crystals in the fine-grained melt showing irregular grain boundaries. d Sieve texture and zoning in chromite rim at the contact with fine-grained melt and with crystallization of poikilitic euhedral grains. ol olivine, opx orthopyroxene, cpx clinopyroxene, chr chromite, phl phlogopite

Fig. 4

Characteristic petrographic images of the type 4 harzburgite showing protogranular texture in plane-polarized light (a, b, d) and back-scattered electron image (c). a Fine-grained melt around phlogopite, chromite, olivine and garnet with bright green clinopyroxene occurring between grains as veins. b Chromite and small clinopyroxene unaffected by melt. c Chromite, phlogopite and olivine in melt showing irregular grain boundaries. d Phlogopite with more significant resorption in the melt. ol olivine, opx orthopyroxene, cpx clinopyroxene, chr chromite, phl phlogopite

Fig. 5

Characteristic petrographic images of the type 5 harzburgite showing protogranular texture in plane-polarized light (a, b), cross-polarized light (d) and back-scattered electron image (c). a Primary clinopyroxene in the matrix of olivine and orthopyroxene affected by cracks filled with serpentine. b Garnet with feathery inner rim and coarse-grained outer rim with euhedral spinel. c Primary clinopyroxene with the inclusion of orthopyroxene. d Orthopyroxene inclusion in olivine with serpentine at the margins. ol olivine, opx orthopyroxene, cpx clinopyroxene, sp spinel, grt garnet

Fig. 6

SiO₂ vs MgO m/m % variation diagram showing analyses for olivine core, rim, olivine inclusion in orthopyroxene and olivine near assemblage along with those from Carswell et al. (1979)

All types are marked by large (up to 4 mm) olivine and orthopyroxene grains with smooth, anhedral grain boundaries. Olivine forms intergrowth in orthopyroxene and vice versa, which imparts poikilitic texture (e.g., Fig. 2a, b). Type 1–3 xenoliths have high olivine content (71–85 vol. %), and type 4 and 5 xenoliths have lower olivine content (57–60 vol. %) (Table 1). The orthopyroxene content is the lowest in types 2 and 3 (14–15 vol. %) but variable in other types (23–36 vol. %) (Table 1). In all xenoliths, olivine is more fractured than orthopyroxene, where serpentine veins run through the cracks and grain margins (e.g., Fig. 2c, d).

The chromite in type 2 xenoliths is interstitial and slightly elongated (up to 1.5 mm), irregular-shaped reddish-brown anhedral magnesiochromite (< 1 vol. %) with smooth boundaries (Fig. 3c). In comparison, the magnesiochromite in type 3 (< 1 vol. %) and 4 (3 vol. %) xenoliths has curved, irregular boundaries and often occur in conjunction with large (up to 1 mm) flakes of pale-orange colored phlogopite (< 1 vol. %) (e.g., Fig. 3b, c). In addition, the magnesiochromite and phlogopite are surrounded by a fine-grained matrix (Figs. 3b–d, 4a, c, d) interpreted as melt, and at its contact, they have irregular grain boundaries. The magnesiochromite shows zoning, spongy rims, and crystallization of new idiomorphic crystals when in contact with the melt (Fig. 3d). The shape of the assemblage (magnesiochromite, phlogopite, melt) is mostly irregular and resembles that of ghost minerals that have entirely broken down (e.g., garnet). Furthermore, in type 3 and 4 xenoliths, the clinopyroxene occurs as discrete tiny crystals (up to 100 μm) around the assemblage (Figs. 3a, 4b) and in veins at grain boundaries of matrix minerals or those cross-cutting the matrix minerals. Type 4 xenoliths differ from type 3 in having larger (> 2 mm) clinopyroxene (2 vol. %) around the assemblage between primary minerals (Fig. 4a). Furthermore, a small (~ 200 μm) relict pink chrome-pyrope garnet (< 1 vol. %) (Fig. 4a) is present in the melt in Type 4 xenoliths. Type 4 xenoliths are classified as garnet–chromite–harzburgite because the presence of relict garnet and the ghost mineral-like shape of the assemblage indicates its previous presence. The melt surrounding the phlogopite and chromite in Type 3 and 4 xenoliths is heterogenous (Figs. 3c, 4c) contains grains of olivine, chromite, etc. and under SEM shows irregular elongate streaks or flakes-like crystals (Fig. 3d). Type 5 xenoliths differ from other xenoliths in having interstitial clinopyroxene and garnet. The clinopyroxene is elongated (> 3 mm) and bright green anhedral Cr–diopside (3 vol. %). The garnet is purple-colored (1 vol. %) chrome-rich pyrope (< 1 mm) surrounded by a semi-opaque kelyphitic rim (Fig. 5b). The kelyphitic rim is made up of two rims where the inner rim resembles a graphic or feather texture due to the habit of spinel (Špaček et al. 2013), and the outer rim is coarser and associated with phlogopite. Numerous euhedral small (up to 100 μm) yellowish to reddish brown spinel grains impart a coarse texture.

Mineral chemistry

Olivine

The Fo content of olivine is the highest in type 1–3 xenoliths (> Fo₉₄) and the lowest (< Fo₉₂) in type 5 (Fo_{90.05–92.02}) xenoliths (Supplementary Table S1: olivine). The type 4 olivines have intermediate Fo content (Fo_{92.03–94.67}) varying between type 5 and type 1–3. The NiO does not show any correlation between the different types (or core-rim relation) and is variable (0.1–0.78 m/m %). Olivine grain in chromite in type 3 xenolith shows up to 0.92 m/m % Cr₂O₃ (Supplementary Table S1: olivine).

Orthopyroxene

The orthopyroxenes are enstatite. The Mg# in orthopyroxene shows a similar trend as Fo in olivine. The Mg# is highest in type 1–3 xenoliths (> 95) and the lowest (< 93.2) in type 5 (92.2–93.2) xenoliths (Supplementary Table S1: orthopyroxene). Orthopyroxene from type 4 has intermediate Mg#, usually varying between type 5 and type 1–3. The Al₂O₃ (0.69–1.540.69 m/m %) and Cr₂O₃ (0.4–0.64 m/m %) contents are the highest in type 2 chromite harzburgites (Supplementary Table S1: orthopyroxene). The CaO content is the highest (0.42–0.7 m/m %) in type 5 xenoliths with primary garnet and clinopyroxene (Fig. 7; Supplementary Table S1: orthopyroxene).

Fig. 7

Major element variation diagram for CaO vs Al2O3 m/m % showing analyses for orthopyroxene core, rim, orthopyroxene inclusion in olivine, orthopyroxene inclusion in orthopyroxene and orthopyroxene profile near assemblage compared with compositional of Pipe 200 orthopyroxene inclusion in orthopyroxene and orthopyroxene profile near assemblage compared with compositional of Pipe 200 orthopyroxene from Carswell et al. (1979). The field for garnet peridotites in Kaapvaal kimberlites and spinel peridotites kimberlites from Premier and Kimberley, Kaapvaal Craton are from Gregoire et al. (2005) and references therein

Chromian spinels

The spinels from type 2–4 are magnesiochromites, whereas those from type 5 xenoliths occurring at the kelyphitic rim of garnet are spinels (sensu stricto). The spinels in all investigated harzburgite types show an inverse relation between the Al₂O₃ and Cr₂O₃ content. The types 3 (core Cr#–84.0 and Mg#–76.2) and 4 (core Cr#–81.6 and Mg#–62.7) xenolith chromites, which occur in conjunction with phlogopite and are surrounded by the melt (Supplementary Table S1: melt) and clinopyroxene, have the lowest Al₂O₃ content at 6.82–7.92 and 7.93–9.40 m/m %, respectively (Fig. 8 and Supplementary Table S1: spinel). The chromites in type 2 (core Cr#–78.4 and Mg#–74.3) xenoliths have a higher average Al₂O₃ at 13.23 m/m %, varying up to 25.75 m/m % (Fig. 8; Supplementary Table S1: spinel). The Type 5 spinels (core Cr#–50.7 and Mg#–68.9), which occur at the kelyphitic rim, differ from type 2–4 chromites (Fig. 8) in having the highest Al₂O₃ (17.1–43.47 m/m %) at the lowest Cr₂O₃ (20.61–45.79 m/m %).

Fig. 8

Major element variation diagram for Cr₂O₃ vs Al₂O₃ m/m % showing chromian spinel core and rim analyses. The analyses of Pipe 200 spinel from Carswell et al. (1979) are also shown for comparison

Clinopyroxene

The clinopyroxene from all types is chrome-diopside and can be differentiated based on CaO, Cr₂O₃, and NaO contents (Fig. 9; Supplementary Table S1: clinopyroxene). The interstitial clinopyroxene in type 5 xenoliths has higher CaO (20.65–21.3 m/m %) content than those in type 3 (19.43–19.56 m/m %) and type 4 (16.75–19.36 m/m %) xenoliths which are restricted to veins and grain boundaries. Furthermore, the Cr₂O₃ content of clinopyroxene in type 3 and 4 xenoliths is always higher than 2.07 m/m % but lower than 1.87 m/m % in the case of type 5 xenoliths (Supplementary Table S1: clinopyroxene).

Fig. 9

Major element variation diagram for Na₂O vs CaO m/m % showing clinopyroxene core and rim analyses. The analyses of Pipe 200 clinopyroxene from Carswell et al. (1979) are also compared

Phlogopite

The phlogopite in the type 3 (Mg# 95.53–96.0) and 4 (Mg# 93.87–94.81) xenoliths has similar TiO₂ = 0.0–0.22 m/m %, Cr₂O₃ = 1.40–1. 69 m/m %, and K₂O = 9.9–11.0 m/m % concentration (Supplementary Table S1: phlogopite). The Al₂O₃ concentration is slightly higher (Fig. 10) in type 4 (13.52–14.22 m/m %) xenoliths than in type 3 (12.78–13.34 m/m %).

Fig. 10

Major element variation diagram for Al₂O₃-Cr₂O₃ vs TiO₂ m/m % showing analyses for phlogopite in the assemblage (Type 3 and 4) from Pipe 200 xenolith types along with that for Pipe 200 phlogopite from Carswell et al. (1979). Fields for phlogopite from metasomatised peridotite xenoliths, kimberlite groundmass and high-Ti-Cr phlogopite from mantle xenoliths are from Kargin et al. (2019) and the references therein

Garnet

The type 4 garnet in the melt pocket and the interstitial garnet with kelyphitic rim in type 5 xenoliths do not vary much in their CaO and Cr₂O₃ concentrations and plot in G9 (lherzolitic garnets) composition field of the classification from Grütter et al. (2004) (Fig. 11). However, the pyrope content of type 4 garnet is slightly lower (70.57) than that of type 5 (74.78–77.63), and the almandine content of type 4 garnet (10.25%) is higher than that in type 5 xenoliths (up to 5.78%) (Supplementary Table S1: garnet).

Fig. 11

Major element variation diagram for Cr₂O₃ vs CaO m/m % showing analyses for garnet in the assemblage (Type 4) and primary garnet (Type 5) from Pipe 200 xenolith types along with the garnet from Carswell et al. (1979). The composition field classification from Grütter et al. (2004) where- G1 low-Cr megacrysts, G3 eclogitic garnets, G4, G5 pyroxenitic garnets, G9 lherzolitic garnets, G10 harzburgitic garnets, G12 wehrlitic garnets, G0 unclassified category and GDC graphite–diamond constraint

μXRF

μXRF maps were used to characterize the element distributions in the garnet and the kelyphitic rim (Fig. 6b). The Al, Ca, Fe, and Mn concentrations are pronounced in the garnet cores. The kelyphitic rim is marked by a zone (particularly around the spinel grains in Fig. 12b, c) enriched in K and Ca (Fig. 12b–d). The neighboring olivine can be distinguished because of the high Ni and Mg concentrations. The Ca concentration is higher in the cracks and veins in the garnet, whereas Fe is present in all minerals. The higher Fe and Mg concentrations in olivine and Si-rich orthopyroxene are used to enhance the differences between olivine and orthopyroxene further which was used for estimation of modal content.

Fig. 12

a Type 5 harzburgite garnet; major element distribution map of Type 5 harzburgite garnet kelyphitic rim where the R:G:B sequence of the elements is b Ca:K:Ba; c Ti:Ca:Mg and d K:Ni:Al

FT-IR

FT-IR analyses were conducted on the olivine, orthopyroxene, clinopyroxene, and garnet from the type 2–5 xenoliths (Table 1). The olivine shows absorbance bands in Group 1 of the OH infrared absorbance bands (Fig. 13a) at 3415–3,653 cm⁻¹ (Matveev 2001). These absorbance bands are transitional between Group 1A (P > 2 GPa) and Group 1B (1–2 GPa) pressures of hydrogenation (Matveev and Stachel 2007). The absorbance bands appear in all types at ~ 3,571, ~ 3,592, ~ 3,638, and ~ 3,624 cm − 1 (Fig. 13a). The low-intensity absorbance bands include those at ~ 3,534, ~ 3,624, and ~ 3,612 cm⁻¹. Some samples have a weak band at ~ 3,328 cm⁻¹. In the type 3 and 4 harzburgites, the ~ 3,638 cm⁻¹ band is stronger than the ~ 3,592 cm⁻¹ band and, at times, is as pronounced as the ~ 3,571 cm⁻¹ band. The ~ 3,534 cm⁻¹ band is more prominent in types 3 and 4 (with melt pocket assemblage) (Fig. 13a) and is associated with H in a Ti–clinohumite point defects (Berry et al. 2005; Mosenfelder et al. 2006; Walker et al. 2007) which also show a strong absorbance band at ~ 3,672 cm⁻¹, representing the inclusion of talc (Matsyuk and Langer 2004). The type 2 xenoliths (chromite harzburgites) and type 5 xenoliths (garnet harzburgites) differ from type 3 and 4 xenoliths by showing a higher ~ 3,571 cm⁻¹ band intensity. Utmost care was taken to avoid serpentines, and the data were checked to discard spectra with serpentine absorbance bands at ~ 3,690 cm^{− 1} (Jamtveit et al. 2001). The profiles taken across the olivine grains are flat for the group 1 absorbance band intensities (A_int) in all types (Fig. S2). The water concentration in the olivine is 17–33 ppm, as calculated using analysis from five differently oriented grains (Table 2). The type 2 xenoliths (chromite harzburgites) have the highest olivine water concentration, and those with assemblages surrounded by melt pockets (types 3 and 4) have the lowest water concentration. The type 5 xenoliths (garnet harzburgites) have intermediate water concentrations.

Fig. 13

Representative unpolarised infrared spectra for a olivine; b orthopyroxene; c clinopyroxene and d garnet in the Pipe 200 xenoliths. The absorbance band at ~ 3690 cm⁻¹ is characteristic of serpentine

Table 2 Water concentration (ppm wt.) calculated from Pipe 200 xenolith NAMs

The orthopyroxene absorbance bands appear at ~ 3,517, ~ 3,409, ~ 3,546, and ~ 3,596 cm⁻¹, and a wide band with less intensity appears at ~ 3,060– ~ 3,070 cm⁻¹ (Fig. 13b). In the type 5 xenoliths, a high-intensity band with an asymmetric shoulder appears at ~ 3,690 cm⁻¹ due to serpentines (Post and Borer 2000). The type 2 and 5 orthopyroxenes show a small band at ~ 2,925 cm⁻¹. The orthopyroxene water concentration is calculated using one or two differently oriented grains (Table 2) and is underestimated because of the removal of the serpentine bands. The computed water concentration in orthopyroxene is 21–92 ppm wt. (Table 2). Similar to the olivines, the type 2 orthopyroxenes (chromite harzburgites) have the highest water concentration, followed by the type 5 orthopyroxenes (garnet harzburgites) and type 4 orthopyroxenes (harzburgites with assemblages surrounded by melt), which have the lowest water concentration.

The main clinopyroxene absorbance bands are at ~ 3,690 and ~ 3,640 cm⁻¹ (Fig. 13c). The other band positions are ~ 3,555 and ~ 3,455–3,465 cm⁻¹. The clinopyroxene water concentrations in the type 5 and 4 harzburgites are 737 and 833 ppm wt., respectively (Table 2). However, the serpentine bands may cause an overestimation of the type 5 clinopyroxene water concentration. The garnets show only bands associated with contamination and are assumed to be dry (Fig. 13d).

Discussion

P–T reconstruction

Thermobarometry

The five harzburgite xenolith types show variations in their accessory minerals (chromite, garnet, and clinopyroxene) and their orthopyroxene and olivine mineral chemistry (Figs. 6, 7, 8, 9, 10, 11). The mineralogy and geochemistry variations are due to the sampling depth of the heterogeneous mantle beneath the Kaapvaal Craton. The same was observed in Kaapvaal Craton harzburgites and lherzolites by (Mitchell et al. 1980; O’Reilly and Griffin 2006), and others and has been established for all cratons. Except for the type 5 xenoliths (containing primary garnet and clinopyroxene; Fig. 5), all studied rock types lack the minerals required for barometry, making pressure estimation difficult. We used the garnet–orthopyroxene barometer of (Brey and Köhler 1990) (P_BKN) for the type 5 sample, which has a coherent equilibration pressure of 44 ± 5 kbars. In addition, the single-clinopyroxene barometer of Nimis and Taylor (2000) (P_NT00) provides a comparable pressure of 46 ± 2 kbars. The single-garnet barometer of Grütter et al. (2006) (P_HG06) indicates a much lower pressure of 30 ± 1 kbars. The pressure was calculated using the garnet–orthopyroxene barometer (P_BKN) of Brey and Köhler (1990) and the single-clinopyroxene barometer of Nimis and Taylor (2000) (P_NT00) giving comparable pressures of ~ 45 kbars, we use them in this study (Fig. 14). The presence of Cpx in the type 5 samples allowed us to use the single-clinopyroxene thermometer of Nimis and Taylor (2000) (T_NT00); the median temperature of the clinopyroxene in the type 5 xenoliths is 1,050 ± 48 °C. The clinopyroxene–orthopyroxene thermometers of Taylor (1998) (T_TA98) and Brey and Köhler (1990) (T_BKN) give median temperatures of 993 ± 17 °C and 1,010 ± 12 °C, with < 0.05 atoms per 6-oxygen formula unit (apfu) Na content. The P–T values of the type 5 xenoliths cluster along a conductive geotherm of 40.0 mWm⁻² (Fig. 14). This is similar to the Kalahari Craton conductive continental geotherm (40.0 ± 3.0 mWm⁻²) (Hasterok and Chapman 2011). Smildzins (2015) reported a higher conductive Kaapvaal Craton continental geotherm (44.0 ± 2.0 mWm⁻²) for the Premier and Frank Smith kimberlites from the RSA, shown in Fig. 14 for comparison.

Fig. 14

P–T-depth diagram for Pipe 200 harzburgite xenolith types from this study. The pink rhombus denotes the opx–grt thermometer (T_Ha84) combined with the opx–grt barometer (P_BKN) and the blue rhombus single-pyroxene thermobarometer (P_NT00) and (T_NT00). Temperature from two pyroxene thermometers (T_BKN)-yellow rhombus, (T_TA98)-green rhombus is presented with the median pressure of ~ 44 Kbars calculated from opx–grt barometer (P_BKN). The temperature from Ca in Opx thermometer (T_Ca-in-opx) is shown along with median pressure of ~ 46 kbars calculated from a single-pyroxene barometer (P_NT00). The data for economic Kaapvaal Craton kimberlite pipes- Premier (blue circle) and Frank Smith (yellow circle), graphite-diamond stability field and the conductive continental geotherms is taken from Smildzins (2015) and references therein

As all the rock types have olivine and orthopyroxene, applying thermometers based on these two minerals would allow us to estimate the temperature of all five rock types (Fig. 14; Table 3). Single olivine-based thermometers have been developed by De Hoog et al. (2010), and Al-in-olivine thermometers (T_Al-ol) have been recalibrated by Bussweiler et al. (2017). It is suggested to be applicable to garnet-bearing peridotites only (Bussweiler et al. 2017) and not applied to the studied xenoliths because of the dominance of chromite (Type 2–4) in the xenoliths and low Al (Supplementary table S2: olivine) content in olivine. The well-studied and robust single orthopyroxene (Ca-in-Opx) thermometer by Brey and Köhler (1990) modified by Nimis and Grütter (2010) (for temperature < 900 °C) is applied to the orthopyroxene in all the xenolith types (EPMA data; Supplementary Table S2: orthopyroxene) at the calculated pressure of the type 5 xenoliths (45 kbars). The T_Ca-in-Opx thermometer gives 1,013 ± 6 °C temperature (Table 3) for the type 5 xenoliths, which is comparable to the 993 ± 17 °C and 1,010 ± 12 °C temperatures using the two pyroxene thermometers of Taylor (1998) (T_TA98) and Brey and Köhler (1990) (T_BKN), respectively. The temperatures of all xenolith types calculated using the T_Ca-in-Opx thermometer are in Table 3. The interquartile range is higher (46 °C–93 °C) for the type 1–3 xenoliths, which do not show garnet and have lower temperatures than the type 4 and 5 xenoliths.

Table 3 The median of calculated temperature (in ^oC) and pressure (in kbar) for pipe 200 harzburgite types 1 to 5 (this study), along with those from Carswell et al. 1979 (for comparison), are presented

A thermometer (T_{Al–Cr-in-Opx}) was derived for the Cr and Al contents in orthopyroxene by Witt-Eickschen and Seck (1991) for spinel peridotites, which was applied using the EPMA data of xenoliths. The temperatures calculated using T_{Al–Cr-in-Opx} (897.7 ± 9.8 °C) are similar to those of the T_Ca-in-Opx thermometer for the type 2 xenoliths but much lower for the type 3 and 4 xenoliths (Table 3). The reason is the much lower Al and Cr contents at higher Si and Ca contents, which is attributed to the composition change due to metasomatism (discussed in the Theriak-Domino modeling section).

This metasomatism of type 3 and 4 xenoliths affects the determination of temperature using T_{Al–Cr-in-Opx}; thus, this temperature was not considered in this study. Other thermometers, such as T_{Ca- ol} (De Hoog et al. 2010) and the spinel–olivine thermometers of O’Neil and Wall (1987) and Wan et al. (2008), were also applied to the xenoliths but gave erroneous results due to the low Ca and Al contents of the olivine and thus were not considered in this work. Temperature estimation was difficult due to the lack of well-calibrated thermometers for clinopyroxene-free spinel (chromite) harzburgites (single olivine and orthopyroxene).

Furthermore, the well-known T_Ca-in-Opx thermometer (Table 3) may give lower T estimates in clinopyroxene-free (harzburgitic) rocks (Nimis 2022). However, despite the evidence of metasomatism for the type 3 and 4 xenoliths, the temperature difference between the types is apparent, with the garnet harzburgites having a higher temperature than the metasomatised chromite harzburgites. Thus, this thermometer was considered in this study.

The P–T calculation results are linked to a difference in mineral compositions of the five xenolith types and reflect their equilibration in a wide pressure–temperature interval (Fig. 14) of the mantle segment. Therefore, the mineralogical and mineral chemical diversity of the harzburgitic xenoliths (harzburgite, chromite harzburgite, garnet–chromite harzburgite, and garnet harzburgite) in the studied xenoliths is a result of the sampling depth. Projecting the temperatures calculated using T_Ca-in-Opx on the conductive continental geotherm of 40.0 ± 3.0 mWm⁻², we obtain a pressure range of ~ 28– < 50 kbars (or ~ 2.8– < 5.0 GPa) (Fig. 14). These estimates correspond to a depth range of > 100 to 175 km, which transgresses the diamond stability field.

Theriak-Domino modeling

The type 3 and 4 xenolith types show clear textural evidence for a two-step evolution, with, in particular, the association of chromite, phlogopite, olivine and clinopyroxene (Figs. 3, 4). Although the presence of chromite, olivine, and clinopyroxene is not uncommon in the primary paragenesis of the studied assemblages, phlogopite exclusively appears in conjunction with chromite in the investigated samples. It has low TiO₂ (< 0.22 m/m %) at 1.4–1.69 m/m % Cr₂O₃ (Fig. 10, Supplementary Table S1: phlogopite), similar to low-Ti-Cr phlogopite from garnet peridotite assemblage (Carswell 1975). The orthopyroxene has low alumina content, unlike the Kaapvaal Craton chromite peridotite and similar to the orthopyroxene in the Kaapvaal Craton garnet peridotite (Fig. 7), as alumina is preferentially incorporated into garnet. Therefore, the chromite harzburgites were originally garnet peridotites. The presence of relict garnet in the assemblage confirms the original presence of garnet as a primary phase that was later destroyed during the metasomatic event(s) (Fig. 5). The K and Ba enrichment of the kelyphitic rim around garnet (type 5) is evident on the μXRF maps (Fig. 12) is related to phlogopite formation as a product of garnet breakdown. The composition of this phlogopite is similar to that of the metasomatised peridotite xenoliths (Fig. 10). Phlogopite growth associated with garnet breakdown in mantle peridotites was previously reported for kimberlite-entrained mantle xenoliths by (Kargin et al. 2019), (Safonov et al. 2019), and others.

The approximate triangular shape of this assemblage (Fig. 3c) indicates that they probably recrystallised in already available spaces and represented the original garnet's dissolution and replacement product. The composition of clinopyroxene associated with the assemblage differs from that of the primary clinopyroxene in the harzburgite. This clinopyroxene has higher Na₂O and Cr₂O₃ at lower CaO content than the primary clinopyroxene (Fig. 9; Supplementary Table S1: clinopyroxene) in the type 5 xenoliths. The garnet in the assemblage has a low TiO₂ content of 0.13 m/m% (Supplementary Table S1: garnet), characteristic of fluid-related metasomatism (O’Reilly and Griffin 2013).

The garnet rim in the type 5 xenoliths shows K enrichment (Fig. 12), and the olivine and orthopyroxene rims in the matrix show elevated CaO, Cr₂O₃, and Al₂O₃ (Supplementary Table S1 and Fig. 7). These signatures suggest the metasomatism of garnet peridotite. Petrography suggests simultaneous assemblage (phlogopite, chromite, and olivine) crystallization. The addition of K to the garnet rims (Fig. 12) and the crystallization of phlogopite implies a reaction with melt or fluid enriched in K and H₂O where the garnet breaks down via fluid (or melt) metasomatism to form phlogopite, chromite, and olivine. The olivine formed due to metasomatism-induced breakdown of garnet in Eq. (1) is different from the primary olivine and hereafter referred to as olivine2. The peridotite changes from Grt harzburgite to Grt free (Phl) harzburgite on metasomatism by melts or fluids.

$${\rm fluid1}+{\rm garnet}={\rm phlogopite}+{\rm chromite}+{\rm olivine2}+{\rm fluid2}$$

(1)

The phlogopite and chromite in the assemblage formed due to the metasomatism-induced breakdown of garnet are surrounded by a fine-grained melt and show resorption due to partial melting (Figs. 3, 4). The host chromite grains show a spongy rim or sieve-like texture with the crystallization of tiny idiomorphic crystals at the rims of these chromite crystals (Fig. 3d). The phlogopite and chromite show irregular grain boundaries (Figs. 3, 4). The spongy rim texture is attributed to partial melting or mineral breakdown due to decompression (Carswell 1975; Nelson and Montana 1992; Carpenter et al. 2002; Guzmics et al. 2008) and is the most probable explanation for those textures. Surrounded by fine-grained melt, the equilibrium assemblage of phlogopite + chromite + olivine2 (Eq. (1)) formed due to the garnet breakdown suggests its partial melting.

We conducted the thermodynamic modeling of this equilibrium assemblage using Theriak-Domino to understand the P–T partial melting conditions. The measured type 5 non-metasomatised garnet composition (Supplementary Table S1: garnet) was modified by adding CaO, K₂O, H₂O, and SiO₂ (common in minerals found in melt pockets) until the desired mineralogical series observed in the melt pockets were achieved. This estimated bulk composition (Fig. 15) best produces the observed paragenesis (phlogopite + spinel + olivine2 + melt + clinopyroxene); other bulk compositions do not produce the same set of minerals.

Fig. 15

P–T diagram for the Theriak-Domino thermodynamic modeling of metasomatised garnet bulk composition to form melt pocket assemblage

The melt composition measured in the type 3 xenoliths (Supplementary Table S1: melt) best coincides with the modeled melt composition at 1,250 °C and 3.55 GPa and has 13.8 wt % water. However, the heterogenous nature of this melt under higher magnification can be used to explain the slight variation in measured (e.g., MgO, FeO, CaO) and estimated composition. The melt composition estimated using Theriak-Domino modeling has higher CaO and FeO contents than the measured melt composition. However, in type 3 and 4 xenoliths, the clinopyroxene crystals appear as small, discrete grains (Fig. 3a) or occupy the cracks (Fig. 4a) surrounding the melt; they do not show resorption at the grain boundaries when in contact with the melt. These signatures suggest later crystallization of clinopyroxene from the melt. This is further confirmed by the Theriak-Domino modeling, where the measured melt composition is plotted at a higher pressure than clinopyroxene formation in the P–T window (Fig. 15). The later crystallization of clinopyroxene can lower the CaO content of the melt so that it is closer to the measured melt composition. Thus, the P–T path suggests a further decreased approximate pressure of < 2.3 GPa at ~ 1,360 °C (Fig. 15). This lower pressure estimate corresponds to more significant partial melting (33.4 vol % melt in Fig. 15). It correlates with the larger melt volume observed in the investigated samples (Figs. 5c, 6c). This pressure decrease is associated with an uplift within the mantle by a fluid or melt, probably the kimberlite itself. The P–T window covers the diamond-to-graphite and garnet-to-spinel transition boundaries and the stability field of the spinel and garnet peridotites.

Interpretation of water contents

The olivine water content in our samples is 17.0–32.5 ppm wt. (Table 2). It falls well within the range (0–86 ppm) reported in the literature (Peslier et al. 2008, 2010) for the Kaapvaal Craton peridotite xenoliths. However, they are much lower than the 70 ppm reported by Kurosawa et al. (1997) for Pipe 200 olivine. The olivine water concentration is the lowest in type 3 and 4 harzburgites, which show signatures of metasomatism (Fig. 3, 4). In contrast, the type 2 harzburgites with primary chromite have the highest measured water concentration. The water concentration of orthopyroxenes is 21–92 ppm wt. It is lower than the 40–250 ppm range mentioned in Peslier et al. (2012) for Kaapvaal Craton peridotite xenoliths. The orthopyroxene water concentration shows a similar trend to olivine (Table 2), with the type 2 (chromites) harzburgites having the highest water concentration, followed by the type 5 (garnet) harzburgites, whereas the type 3 and 4 (garnet breakdown assemblage) harzburgites have the lowest water contents. However, the underestimation due to the removal of serpentine absorbance bands from the spectra (Fig. 13b) may be the reason.

The clinopyroxene H₂O contents in the type 5 and 4 harzburgites are 737 and 833 ppm wt., respectively, which exceed the 150–400 ppm range stated by Peslier et al. (2012). The clinopyroxene water concentrations are overestimated because of the removal of spectra affected by serpentine group minerals (band at ~ 3,690 cm⁻¹) (Fig. 13c). The studied garnets are completely dry and show bands for contamination only (Fig. 13d). This agrees with the observations of Doucet et al. (2014), who stated that half of the Kaapvaal Craton garnets have water concentrations < 5 ppm or below the detection limit. Metasomatism evidenced in type 3 and 4 harzburgites is associated in particular with the crystallization of phlogopite (with Si addition) breakdown in the melt pocket assemblage (Figs. 3, 4), suggesting hydrous metasomatism (Eq. 1). On the contrary, the decomposition of garnet into kelyphitic rims (Fig. 5) observed in type 5 xenoliths is probably related to the ascent or uplift of the kimberlite to a shallower depth.

Moreover, the mineral chemistry suggests metasomatism: low-Ti phlogopite (Fig. 10); low-Ti garnet in melt pockets; low orthopyroxene alumina content (Fig. 7); elevated Ca, Cr, and Al contents in olivine and orthopyroxene (Supplementary Table S1: olivine and orthopyroxene); and elevated K and Ba in garnet rims (Fig. 12). This metasomatism may have lowered the original water content of the NAMs, as the metasomatised type 3 and 4 harzburgites' (garnet metasomatism) olivine have the lowest water contents (Table 2). The measured water concentration represents the hydration state of the Kaapvaal Craton mantle beneath the Pipe 200 kimberlite. However, this metasomatism is different from the mantle metasomatism at low pressures caused by hydrous alkaline silicic melts or fluids in Liqhobong (close to Pipe 200) and Kimberley kimberlites (Peslier et al. 2012), where metasomatic melts add water to the minerals.

The profiles taken across olivine (Fig. S2) are flat and show no core-to-rim variation in water content. These flat profiles can be due to the shortness of the profiles affected by fractures. However, Hilchie et al. (2014) found core-to-rim variations in the integrated absorbance intensities (A_int) in a few profiles. The integrated absorbance intensities (A_int) found by Hilchie et al. (2014) at the profile core are higher than the A_int measured in all profiles from our study (Table 2). Their core-to-rim variation in A_int is attributed to the diffusive loss of H to the kimberlite magma. Combined with the low A_int, the absence of this rare core-to-rim variation in Fig. 5d in Hilchie et al. (2014) in our samples suggests the diffusion of water to the kimberlite magma and the possible reset of H₂O at low pressures to the kimberlite. The petrological features evidence xenolith residence in the kimberlite magma at low pressures (uplift); the kelyphitisation of garnet in the type 5 xenoliths (Fig. 5) and the partial melting of the metasomatic assemblages from the melt pockets (Figs. 3, 6) imply a temperature increase and a pressure decrease, as evidenced by the Theriak–Domino modeling (Fig. 15).

The inter-mineral H₂O partition coefficient D_H2O^opx/ol indicates xenolith residence in the kimberlite magma at low pressures (Fig. S3). D_H2O^opx/ol is the ratio of the water contents in orthopyroxene and olivine from the same sample. This coefficient has a value of 1.24–2.84 for the measured Pipe 200 xenoliths (Fig. S3). The average D_H2O^opx/ol of 1.91 is higher than the ratio determined by Bell et al. (2004) from Kaapvaal megacrysts formed at 5 GPa (1.6 ± 0.2) and the experimental estimation of Withers et al. (2011) (0.61). This variation in D_H2O^opx/ol can be attributed to the disruption of mineral water contents (i.e., the addition of water to orthopyroxene or water loss from olivine).

The indication of diffusive water loss in olivine and re-equilibration can explain the higher ratio. Our samples have variable olivine Fo contents Fo_88.9–97.1 for equilibration P–T conditions of 2.8 to < 5.0 GPa and temperature varying from ~ 830 °C to 1,015 °C. Olivine having a high Fo content is representative of a mantle affected by a high degree of partial melting (Peslier et al. 2010). The low olivine Mg# of < 0.92 and the small amounts of modal primary garnet and clinopyroxene in the type 5 xenoliths unaffected by metasomatism contradict high degrees of melting (Doucet et al. 2014 and references therein). Thus, the type 5 xenoliths probably represent the starting stage of the evolutionary processes.

Subcontinental lithospheric mantle (SCLM) evolution

The Pipe 200 kimberlite mantle xenolith types offer insights into the processes occurring in the mantle before kimberlite eruption and emplacement (Fig. 16). The mantle sampled by the xenoliths represents different stages of mantle evolution beneath Lesotho and the Kaapvaal Craton. The initial step (Fig. 16a), represented by the type 5 lherzolite (now harzburgite; composed of olivine + orthopyroxene + clinopyroxene ± garnet and/or chromite), was poorly affected by melt extraction event(s). The melting residues were exhausted in clinopyroxenes to form harzburgites (Fig. 16b) made up of olivine + orthopyroxene ± garnet and/or chromite (types 1–4). The equilibrated mantle sampled by the Pipe 200 kimberlite covering a significant depth range (> 100 to 175 km) was later metasomatised by hydrous alkaline silicic melts or fluids, which led to the garnet breakdown and the co-crystallization of melt pocket assemblages (Fig. 16c) of phlogopite + chromite + olivine2 (Figs. 3, 4). It also modified the minerals' rim compositions (elevated Ca, Cr, Al, and Si, among others). Ultimately, the rock from garnet–harzburgite became phlogopite–chromite harzburgite (types 3 and 4). Metasomatism lowered the water concentration (Table 2), possibly disturbing the OH distribution in the different NAMs (Fig. S3). The coeval fluid/melt enrichment and Karoo magmatism have been linked to explain some mantle metasomatism events recorded in xenoliths from the intrusive kimberlites of the Kaapvaal Craton (e.g., Giuliani et al. 2014; Woodhead et al. 2017; Jollands et al. 2018). These Karoo magmatic events may be the source of metasomatism of the Pipe 200-entrained xenoliths (types 3 and 4). A melt or a fluid, possibly the kimberlite itself, uplifted the xenoliths (Fig. 16d) from the diamond stability field to a lower pressure (2.3 GPa; Fig. 15) across the diamond–graphite boundary. This uplift caused the partial melting of the chromite and phlogopite in the melt pockets (Figs. 3, 4), and later, clinopyroxene crystallized. Furthermore, the pressure decrease caused the formation of the kelyphitic rims of the garnet (Figs. 5b, 12a–d) in the type 5 harzburgite. It also led to the diffusion of olivine water to the kimberlite magma, which disturbed the water distribution in the NAMs through diffusion. The olivine water content's core-to-rim transition was re-equilibrated during the mantle residence (Fig. S2). The change in P–T conditions may have destroyed the diamonds. Consequently, during or after the uplift, the kimberlite melt infiltrated the rock body and exhumed it to the surface (Fig. 16e).

Fig. 16

Sketch showing the evolution of mantle xenoliths from Pipe 200 kimberlite. a The original starting material is assumed to be garnet–spinel lherzolite, b which is depleted to form garnet–spinel harzburgite, c then metasomatised by a fluid or melt (garnet breakdown assemblage of phlogopite, spinel, and clinopyroxene). D Then, part of the mantle is uplifted to shallower depths, which causes partial melting of the assemblage and kelyphitisation of garnet. e Eventually, the xenoliths are sampled by a kimberlite and placed on the surface

Consequences on diamond preservationThe Pipe 200 mantle xenoliths show diversity in harzburgite mineral assemblage, which is reflected in the large equilibration temperature and pressure interval and is linked to the great sampling depth and evolutionary processes in the mantle beneath Lesotho and, in turn, the margin of the Kaapvaal Craton. The petrographic features of melt pocket assemblage, chromite and phlogopite resorption along with the geochemical features of elevated rim Ca and K concentration, H₂O loss and re-equilibration suggest metasomatism and the residence of harzburgitic mantle materials at lower pressures. The processes recorded in the P–T paths in the mantle columns indicate that the lithospheric mantle was affected by melt extraction process(es), metasomatism, and residence in the mantle, which transgresses the diamond–graphite transition boundary. The timescales of H₂O content loss and re-equilibration suggest extended interaction between xenolith and the kimberlite magma, which may be due to the stalling of the kimberlite. These multistep processes of evolution of the harzburgite xenoliths within and across the diamond stability field may be responsible for the destruction of diamonds and the barren nature of two of the three Pipe 200 kimberlite phases. Metasomatism caused by melts (hydrous alkaline silicic melts or fluids) in the mantle itself or pressure decrease combined with interaction with fluids during extended residence in the mantle may be the reason for diamond destruction in the Pipe 200 kimberlite relative to the nearby economic Lesotho kimberlites.

Conclusions

• The studied Pipe 200 xenoliths consists of various types of harzburgites showing signatures of metasomatic enrichment in the form of elevated Ca and K at the rims and garnet breakdown, which formed assemblages of chromite and phlogopite.

• The Cratonic lithosphere sampled by kimberlite represents a pressure of ~ 3.5–7 GPa and lies close to the estimated geotherm of 40 mW m⁻², typical of the Kaapvaal Craton (Hanger et al. 2015).

• The xenoliths (metasomatic assemblages) were affected by subsequent partial melting events during uplift in the mantle at a lower pressure of < 2.5 GPa.

• The structural hydroxyl distribution pattern suggests the diffusion and then re-equilibration of H₂O from xenolith olivine during the brief stalling of the kimberlite during its ascent, leading to disruptions in the inter-mineral H₂O partition coefficient.

• The multistep processes of metasomatism during residence in the mantle and partial melting at shallower depths before kimberlite emplacement possibly led to the destruction of diamonds.

• The local processes in the Pipe 200 kimberlite led to the destruction of diamonds relative to the nearby economic Lesotho kimberlites.