The lithospheric diamond (COMSK3) under investigation is octahedral, with two polished parallel surfaces cut along the [100] crystallographic face, measuring approximately 1.3 × 1.6 mm with a thickness of ∼0.6 mm. The diamond contains ten visible colorless and pale-green mineral inclusions (hereafter, Inc) ranging in size from 0.02 to 0.2 mm (Figures 1a,b). The images of the diamond and its inclusions were acquired using an AmScope stereoscope under reflected and transmitted light, while birefringence colors were observed under polarized light with a Leica DM750 microscope (Supplementary Figure S1a-c). Both microscopes are equipped with their own high-resolution camera, and the images were collected with integrated software provided by the manufacturer.

Textural and morphological analyses on the exposed inclusion (Supplementary Figure S2a) were performed with a ZEISS EVO MA10 scanning electron microscope (SEM) at the CNR- Istituto di Geologia Ambientale e Geoingegneria (IGAG; Sapienza University of Rome, Italy), which is equipped with an AZTEC-integrated energy-dispersive X ray spectrometer (EDS). The analyses were performed with 20 kV and 50 pA, using cobalt as standard. The chemical analyses on Inc1 (Supplementary Figure S2b; Supplementary Table S2) were performed using a JEOL JXA-8230 electron probe microanalyzer (EPMA) equipped with five spectrometers at the Department of Earth Sciences of the University of Florence (Italy). An accelerating voltage of 15 kV and 20 nA with a defocused electron beam of 3 µm diameter size were used. Oxides and silicates were employed as standards: albite for Si and Na, olivine for Mg, plagioclase and garnet for Al, bustamite for Mn, diopside for Ca, ilmenite (Smithsonian standard) for Ti and Fe, chromite for Cr, and sanidine for K. Chemical maps were also acquired (Supplementary Figure S3). Trace element analyses of the exposed Inc1 (Supplementary Table S2) were performed by laser ablation inductively coupled plasma mass spectrometry (LA-ICPMS) with an Agilent-7500 CX plasma mass spectrometer at the Istituto Nazionale di Geofisica e Vulcanologia (INGV) of Palermo (Italy). The analyses were performed with a constant laser frequency rate of 10 Hz, fluency of 15 J/cm², and He flux of 0.8 L min^-1 in the ablation cell. A diameter of 32 µm for the laser spot was used during the ablation. One analysis each was conducted in the core and rim of Inc1, while the background was collected from two spots on the surface of the diamond host. The accuracy of the analyses (RSD%) was calculated by repeated analysis of the USGS basaltic reference glass BCR-2G and was ≤ 15% for all the elements.

The Fe oxidation state of three mineral inclusions, one exposed to the polished surface (Inc1) and two entrapped (Inc2 and Inc5), was determined by in situ synchrotron Mössbauer spectroscopy at the beamline ID18 (Rüffer and Chumakov, 1996) of the European Synchrotron Radiation Facility (ESRF) in Grenoble (France). The typical energy for the Mössbauer effect at 14.4 keV was obtained through the (111) Bragg reflection of a ⁵⁷FeBO₃ crystal mounted on a Wissel velocity transducer driven with a sinusoidal wave form (Potapkin et al., 2012). The beam was focused to an area of 6 × 15 μm² using Kirkpatrick–Baez mirrors. The velocity scale of ±6 mm/s was calibrated using an α-Fe foil of 25 μm of thickness. The spectra were acquired within 2.5 and 4.5 h and fitted using a pseudo-Voigt line shape with MossA software (Prescher et al., 2012). Nestola et al. (2016) and Marras et al. (2023) described the advantage of this technique, which consisted of a much higher spatial resolution of the measurement (∼15 μm) than the Mössbauer milliprobe source that, in turn, allows the acquisition of Fe³⁺/ΣFe values from multiple inclusions either entrapped or exposed. Supplementary Figures S4a–c show the spectra of Inc1 and Inc2, collected in approximately 4.5 h and 2.5 h, respectively, and of Inc5, collected in 2.5 h. The Mössbauer spectra of cpx inclusions were fitted with two doublets for Fe²⁺ and one for Fe³⁺, following Canil and O'Neill (1996) and Stagno et al. (2015). The component area ratios of the Fe²⁺ doublets were left unfixed but constrained to the same value to account for the preferred crystal orientation (McCammon et al., 2000). The spectrum of ol was also fitted with two doublets for Fe²⁺ and one for Fe³⁺. The tiny contribution of Fe from the Be window was considered by fitting all the spectra with an extra singlet (central shift of 0.3249 mm/s), the position of which was determined through measurements of the background performed during the analytical session. Supplementary Table S4 summarizes the hyperfine parameters for each of the three analyzed minerals and the calculated Fe³⁺/∑Fe. The quadrupole splitting (QS) and center shift (CS) are reported in Supplementary Figures S5a and b (cpx and ol, respectively) along with the known hyperfine parameter for ol (Canil and O'Neill, 1996; Ejima et al., 2018; Marras et al., 2023), for peridotitic cpx (Canil and O'Neill, 1996), for natural eclogitic cpx, and for synthetic eclogitic cpx (Stagno et al., 2015) (Supplementary Table S4). Hyperfine parameters for eclogitic cpx inclusions in diamonds (McCammon et al., 1998) are also shown in Supplementary Table S4.

The Horiba LabRam HR 800 Raman spectrometer at the University of Roma Tre (Rome, Italy) was used to identify the entrapped mineral inclusions, infer their chemical composition, and verify the presence of residual stress. A total of 57 spectra were acquired in the range of 100–4,000 cm⁻¹ from eight of the ten inclusions, one of which (Inc6) appeared to consist of two smaller inclusions (Inc6a and Inc6b). The instrument is equipped with an attenuated doubled Nd³⁺: YAG laser with a wavelength of 532 nm and an optical microscope. The laser power on the sample surface was 60 mW, and the analyses were performed with a 100× objective and 5 μm² spot size. Pure silicon was used as a standard. Instrumental settings consisted of 600 grooves/mm grating density, a confocal hole of 300 μm, and a slit of 200 μm, with an exposure time of 60 s. Each collected spectrum was the average of three acquisitions to improve the signal-to-noise ratio. The acquired spectra were preliminarily compared with those available from the RUFF database (Lafuente et al., 2015). The precise positions of the spectral bands were determined through deconvolution using PeakFit software. Supplementary Figures S6a–h show the collected spectra along with the zoomed image of the analyzed inclusions. Spectra for Inc8 and Inc9 are not shown due to their low-quality signal caused by the small sizes (less than 20 µm) of the inclusions. To account for the temporal drift in the Raman band position, the collected spectra were corrected with respect to the peak at 1,330.4 cm⁻¹ measured on the diamond surface. In each of the collected spectra, a broad band appears at ∼1,150 cm⁻¹, which we attribute to a residual solvent (i.e., ethanol) used to clean the diamond prior to the Raman analyses. The occurrence of hydroxyl groups could be investigated only in Inc1 by reflectance micro-Fourier-transform infrared spectroscopy (µ-FTIR) using an IR Nicolet iS50 spectrometer coupled with the Nicolet Continuum IR microscope, available at the Department of Earth Sciences, Sapienza University of Rome, Italy. The spectra were collected in the range of 4,000-650 cm⁻¹ with a spatial resolution of 4 cm⁻¹ averaged from 300 scans and a 50 × 50-μm spot size. Additional measurements were conducted by transmission µ-FTIR spectroscopy at the Material Science branch of SISSI beamline of CNR-IOM at Elettra Sincrotrone Trieste (Italy) to characterize the host diamond for possible thermometric applications; a smaller and circular detectable area avoids the inference of small inclusions when collecting transmission-mode spectra. The analyses were carried out between 6,000 and 600 cm⁻¹ using a Vertex 70 v interferometer coupled with a Hyperion2000 microscope (Bruker Optics), liquid N₂-cooled HgCdTe (MCT) detector, and a KBr beam splitter. Measurements were conducted using the synchrotron source with a resolution of 2 cm⁻¹ and an aperture of 30–60 μm by collecting five to ten acquisitions per area of 128 scans each. The µ-FTIR spectra were elaborated using OPUS and Spectragryph (Menges, 2020) software programs.

Structural information on the entrapped mineral inclusions was acquired by in situ synchrotron single-crystal X-ray diffraction (SCXRD) at the GSECARS 13BM-D beamline of the Argonne National Laboratory (Chicago, USA). Prior to the SCXRD measurements, the diamond was scanned, and its inclusions were localized with the help of X-ray computed microtomography (CMT; Supplementary Figure S7) in conjunction with the SCXRD setup at beamline 13BM-D (Sutton et al., 2022). For the CMT measurements, a monochromatic beam at 30 keV was defocused to provide a large field of view on a scintillator plate. A Mitutoyo 5× objective attached to a 75-mm tube allowed a pixel resolution of 2.08 µm. A total of 900 frames with 0.2° step and 2-s exposure was collected as the diamond rotated. Tomographic reconstruction of the projections was carried out with the help of IDL-based software available at the beamline. The reconstructed images were used to accurately place the inclusion of interest in the center of the X-ray rotation. Subsequently, the same monochromatic beam was focused on a 20 × 20 μm beam, and the diffraction setup was automatically railed in. The X-ray diffraction system was equipped with a Pilatus 1M CdTe detector. The sample–detector distance was set to 227.91 mm and the energy at 30 keV (λ = 0.4133 Å). SCXRD data collection was performed as a series of ω scans with a step of 0.5° and 10 s exposure/step over a ±35° range. Single-crystal X-ray diffraction data were collected on Inc1, Inc4, Inc5, Inc6a, and Inc6b; the unit-cell parameters are reported in Supplementary Figure S5. The crystallinity of the targeted inclusion was initially verified using Dioptas (Prescher and Prakapenka, 2015), followed by further analysis of the unit-cell determination, integration of the reflection intensities, and empirical absorption correction using the CrysAlisPro software package. The refinements were performed using SHELXL (Sheldrick, 2015). Charged scattering curves (Prince, 2004) were employed in the refinement of the olivine crystal structures. EosFit-Pinc software (Angel et al., 2017) was employed to determine the final pressure of the inclusion at ambient conditions (P_inc) and pressure of entrapment (P_trap).

The diamond examined under polarized transmitted light revealed a typical birefringence (Lang, 1967; Howell et al., 2012) which consisted of the formation of strain-nodules caused by the stressed Inc4 and Inc5 (Supplementary Figure S1a). Supplementary Figures S1b and c show the (sub)euhedral shape of the inclusions varying from rounded (Inc5; Supplementary Figure S1b) to prismatic (Inc6; Supplementary Figure S1c). These images highlight both the presence of composite inclusions (e.g., Inc6 is composed of two tiny inclusions, Inc6a and Inc6b, positioned ∼120° from each other) and the occurrence of tiny radial fractures departing from the inclusions, which are generally much shorter than the inclusion size, along with a dark brownish and wavy rim. Some inclusions are rimmed by dark halos, interpreted by Agrosì et al. (2019) as the residue of carbonate-bearing fluids from which the diamond might have formed. Striations, commonly referred to as stepped-face growth, are visible on the external surface of the diamond (Supplementary Figure S8) and represent resorption features formed during transport by oxidized ascending magma (Fedortchouk et al., 2019). The textural analysis of the exposed polished inclusion Inc1 shows the presence of radial fractures and a (sub)euhedral shape (Figure 1b). The chemical analysis of the polished inclusion (Supplementary Table S2) appears homogeneous, which is also confirmed by compositional maps (Supplementary Figure S3). The core-to-rim chemical composition of Inc1 shows a slight decrease of MgO (from 14.00 to 13.47 wt%), CaO (from 14.30 to 13.98 wt%), and Na₂O (from 3.79 to 3.68 wt%), while the other oxides remain constant within the uncertainty of the measurement. The obtained average composition calculated from the four analyses, with totals close to 100 wt%, is 54.82 (±0.63) wt% SiO₂, 0.04 (±0.01) wt% TiO₂, 2.57 (±0.05) wt% Al₂O₃, 6.33 (±0.08) wt% Cr₂O₃, 1.82 (±0.06) wt% FeO, 0.05 (±0.01) wt% MnO, 13.61 (±0.32) wt% MgO, 14.11 (±0.23) wt% CaO, 3.73 (±0.07) wt% Na₂O, and 0.50 (±0.01) wt% K₂O. On the basis of six oxygen atoms, the chemical formula of the polished inclusion can be written as Ca 0.56 Na 0.27 K 0.02 Fe 0.05 2 + Mg 0.10 Mg 0.69 Cr 0.19 3 + Al 0.11 Fe 0.01 3 + Si 2.00 O 6 , identified as a diopside with ∼20% of kosmochlor (NaCrSi₂O₆) component and ratios of Cr/(Cr + Al) (Cr) of 0.63 and Mg/(Mg + Fe) (Mg#) of 0.93.

Inc1, Inc2, and Inc4 (Supplementary Figures S6a,b,d) exhibit the characteristic vibrational features of a diopside with major peaks at 333–341, 403–407, and 558–564 cm⁻¹ related to metal-oxygen translation, 672–678 cm⁻¹ (ν11) caused by stretching vibrations of bridging O–Si–O atoms, and 1,014–1,021 cm⁻¹ (ν16) caused by stretching vibrations of non-bridging O–Si–O atoms (Chopelas, 1999; Huang et al., 2000; Wang and Gasparik, 2001). The consistency of the vibrational modes, especially for ν11, between Inc1, Inc2, and Inc4 (Supplementary Table S6) might indicate that they all have the same chemical composition. According to the parametrization of the Raman shift of ν11 as a function of CaO, MgO, Na₂O, Al₂O₃, and Cr# (Kalugina and Zedgenizov, 2020), these Cr-clinopyroxene inclusions have compositions between peridotitic and eclogitic, with low Al₂O₃ content of approximately 3 wt% calculated as Al 2 O 3   wt % = 0.81   ±   0.05 · ν 11   ‐   545.2   ±   37.2 . (1)

The spectra collected from Inc5, Inc6a, and Inc6b (Supplementary Figures S6e–f) are consistent with olivine (Chopelas, 1991). The position of bands at 826 cm⁻¹ and 857 cm⁻¹ correlate with the Mg# (but also with P). According to the available parameterizations (Kuebler et al., 2006; Yasuzuka et al., 2009), these three inclusions have Mg# > 90, which is further supported by the appearance of a band at ∼610 cm⁻¹, typical of forsteritic olivine. The Raman spectra of Inc3, Inc7, and Inc10 (Supplementary Figures S6c, g, h; Supplementary Table S6) are representative of (ortho)enstatite with Mg# of 90 (Huang et al., 2000). The presence of additional bands at 368 cm⁻¹ (Inc3) and 367 cm⁻¹ (Inc10), identifiable only by peak deconvolution, can however be interpreted as characteristic of clinoenstatite (Cen; Ulmer and Stader, 2001). In conclusion, few variations of the main bands in some inclusions are observed with respect to those from the literature; these can be explained as the effect of either the chemical composition or some possible residual stress of those inclusions entrapped in the diamond. The infrared (IR) spectrum (Supplementary Figure S9a) shows no defect-induced absorption in the 1,500-500 cm⁻¹ region, which excludes the presence of N and H in the diamond with the two-phonon absorption characteristic of type II diamonds in the 2,800-1,500 cm⁻¹ spectrum. No evidence of molecular H₂O was found in Inc1 (Supplementary Figure S9b).

The concentration of trace elements in Inc1 (Supplementary Table S3) between core and rim is within uncertainty, such as Sr with 939 ppm (942 ppm at the rim), Ti with 104 ppm (109 ppm at the rim), Mn with 663 ppm, V with 543 ppm (539 ppm at the rim), Ni with 338 ppm (331 ppm at the rim), Ce with 37 ppm, and Sc with 31 ppm. Hence, no significant variations occur between the rim and core of Inc1. Figure 2a shows the abundance of seven analyzed trace elements at the core and rim of Inc1, normalized to the carbonaceous Ivuna chondrite used as reference (Stachel et al., 2022; McDonough and Sun, 1995) and plotted along with those from 30 lherzolitic inclusions with the aim of reconstructing the possible petrological history of the inclusions and, hence, of the diamond. The trace element distribution is consistent with published data for lherzolitic P-type inclusions (Stachel et al., 2022). The (CI chondrite-) normalized rare earth elements (REE) are shown in Figure 2b, where the typical steep trend for light rare earth elements (LREE) agrees with that reported for lherzolitic cpx inclusions and would indicate coexistence with garnet in the diamond-forming mineral assemblage (Stachel et al., 2022).

Very few Fe³⁺/∑Fe data exist for the tiny minerals entrapped in diamonds that might shed light the redox environment where diamonds grew. Here, in situ synchrotron Mössbauer analyses were performed to determine the Fe³⁺/∑Fe ratio of two cpx crystals (Inc1 polished to the surface, Inc2 entrapped) and one entrapped ol (Inc5). The calculated Fe³⁺/∑Fe ratio in Inc1 is 0.16 (±0.03), 0.14 (±0.02) in Inc2, but 0.02 (±0.01) in Inc5 (Supplementary Figure S4). These data agree with the only available data for Fe³⁺/∑Fe, varying from 0.08 to 0.14, in cpx diamond inclusions from George Creek (McCammon et al., 1998). When compared with Fe³⁺/∑Fe ratio of cpx from mantle rocks (Supplementary Table S7), we notice that spinel (spl)-peridotites (95 samples; Luth and Canil, 1993) have cpx with Fe³⁺/∑Fe that ranges between 0.03 and 0.35 (0.24–0.29 is the most frequent interval); cpx crystals in grt-peridotites (22 samples; Luth and Canil, 1993; Woodland et al., 2006) have Fe³⁺/∑Fe between 0.09 and 0.44 (0.24–0.29 is the most frequent interval), while cpx crystals from eclogites (63 samples Mikhailenko et al., 2020) have Fe³⁺/∑Fe varying between 0.05 and 0.63 (0.15–0.24 being the most frequent interval). Finally, Fe³⁺/∑Fe of wehrlitic cpx crystals (3 samples) vary from 0.16 to 0.21. Our measured Fe³⁺/∑Fe in cpx is also compared with that of synthetic cpx equilibrated at high P, T, and fo₂ of dry eclogitic assemblages. The Fe³⁺/∑Fe of eclogitic cpx varies between 0.13 and 0.24 under the carbon–carbon dioxide (CCO) buffer (Purwin et al., 2013) and 0.04 to 0.19 under the carbon–carbonate buffer (Stagno et al., 2015), respectively. It is, therefore, difficult to infer a possible interaction between cpx and oxidized fluid just by looking at the Fe³⁺/∑Fe. The Fe³⁺/∑Fe measured on the ol of Inc5 is consistent with the literature (Ejima et al., 2018; Marras et al., 2023).

Elastic geobarometry allows the calculation of the conditions of entrapment, exploiting the difference in elastic properties between the host and the inclusion starting from the residual pressure on the latter—Pinc (Angel et al., 2022). Olivine inclusions in diamonds have been employed as geobarometers to infer the growth conditions of the diamond host (Nestola et al., 2011). The unit-cell volume that the ol inclusion would have at room pressure V₀ is needed to determine P_inc , and it depends on the Mg# (0.93 for Inc5 and Inc6b, and 0.92 for Inc6a, all obtained here by in situ structure refinement and Raman spectroscopy). For Inc6a, isotropic approximation was employed due to the lower number of available reflections (Supplementary Table S8). We calculated V₀ using two different equations: a linear regression (Wang et al., 2023) and the equation proposed by Schwab and Küstner (1977). The resulting P_inc of 0.4–1.1 GPa was calculated using the third-order Birch–Murnaghan equation of states (EoS) isothermal model for olivine (Angel et al., 2018), taken as minimum and maximum values by combining the two models in Supplementary Table S9. From P_inc , a pressure of entrapment, P_trap , was calculated using the EoS for diamond (Angel et al., 2015). From the three ol inclusions, applying a temperature of entrapment, T_trap , of both 1,100  °C and 1,200  °C results in Ptrap of 5.2 and 5.5 GPa, respectively, using the linear regression, and 4.5 and 4.7 GPa, respectively (Schwab and Küstner, 1977). The available dataset of inclusions in diamond (Stachel et al., 2022) reports 14 diamond samples with only cpx crystals found as inclusion. Among these diamonds, one is type II and four have no analyses of N aggregation. Temperature estimates vary from 1,116  °C to approximately 1,370  °C using the available geothermobarometer of Nimis and Taylor (2000) and Nimis et al. (2020). In this study, thermobarometry on Inc1 gives T of 1,060  °C and P of 4.3 GPa, the latter being consistent with a P range of 4.5 and 5.5 GPa obtained by elastic barometry on Inc5, Inc6a, and Inc6b. By projecting the mid-pressure value of 5 GPa on the 40 mW/m² geotherm of the Amazonian craton, we obtained an independent T of 1,155  °C, which is consistent with a T range of 1,170  °C–1,300  °C proposed for the Juína area aged approximately 2 Ga (Bulanova et al., 2010). These thermobarometric estimates are reported in Figure 3 along with the P-T conditions equally determined for those diamonds where only cpx inclusions have been reported. The P-T conditions for COMSK3 fall below the graphite–diamond transition line (Day, 2012).

Typical diagrams have been used to distinguish peridotitic from eclogitic cpx inclusions in lithospheric diamonds on the basis of their major elements, such as Cr, Ti, Ca, Na, Al, Mg, Fe, and Cr (Figures 4a–c). The deviation from the proposed P- and E-type fields as well as differences with respect to the composition of cpx from mantle peridotites are evidence of metasomatic reactions (Mikhail et al., 2021; Stachel et al., 2022). The concentration of trace elements in diamond inclusions (LREE/HREE, Ce/Nb, Nd/Dy, Zr/Hf, and Ti/Eu) is also indicative of metasomatism, with a strong dependence on the chemical composition of the reactive fluid (Stachel and Harris, 1997; Stachel et al., 2022). The major elements for the exposed Inc1 were compared with the available chemical dataset for cpx inclusions in lithospheric diamonds. While the Mg# and Cr# of Inc1 cpx falls within the characteristic range of the peridotitic suite (Figure 4a), the Na and Al contents (expressed in atoms per formula units, apfu) lie in between peridotitic and eclogitic compositions (Figure 4b; Equation 1). The average values of Cr₂O₃ (6.3 wt%) along with the high CaO content (14.11 wt%) of Inc1 are typical concentrations of some wehrlitic garnets (Sobolev et al., 1973). These values deviate from the typical values found in cpx inclusions in peridotitic and eclogitic suite (Figure 4c), with the exception of few cpx, such as D1364A (Daniels and Gurney, 1999) and Kmsm-6 and Kmsm-7 from Yakutian wehrlitic diamonds (Sobolev et al., 1977; 2004). Both the major (e.g., Na, Cr, and Al) and trace (e.g., V and Sc) element composition of the entrapped cpx suggest a refertilization process similar to that recorded by cpx formed by (re-)fertilization of spl-bearing peridotite (Aulbach et al., 2020). Wehrlitization processes thus occur through the interaction of a mantle grt harzburgite with a metasomatic CO₂-bearing melt at 1.5–2 GPa, leading to the formation of opx poor/free, cpx-rich rocks through the equilibrium reaction (Wallace and Green, 1988; Yaxley et al., 1998): e n s t a t i t e + d o l o m i t e   m e l t = d i o p s i d e + o l i v i n e + C O 2   v a p o r (2)

An analog reaction occurring more in-depth is e n s t a t i t e + d o l o m i t e   m e l t = d i o p s i d e + o l i v i n e + d i a m o n d + O 2 (3) that is, equivalent to the fo₂ buffering equilibrium (the acronym is EDDOD; Stagno and Frost, 2010) and describes the coexistence of carbonate (either liquid or solid) with diamond in the upper mantle. At the P and T determined from Inc1, this equilibrium reaction locally buffers the Δlogfo₂ at values between −1.3 and −1.6, assuming a reacting (metasomatic) melt varying from Na-bearing dolomitic to carbonate–silicate in composition. The lack of a chemical zonation between rim and core in Inc1 would support the cpx origin by direct transformation of a preexisting opx rather than by metasomatic contamination of a lherzolitic cpx. Indeed, a recent experimental study (Wang et al., 2022) helps distinguish peridotitic cpx from cpx formed by wehrlitization (Equation 2) on the basis of the variation of the major element composition, which is affected by the nature of the interacting melt, whether carbonatitic or silicic. In the case of Inc1, the simultaneous increase of the Ca/Al ratio with Mg# opposite the decreasing trend of Al₂O₃ and TiO₂ are proposed evidence of the formation by interaction with coexisting carbonatitic melts (Equation 3). The high kosmochlor component has been classically considered a marker of groups 6 and 7 cpx used as mineral indicators for diamond exploration (Quirt, 2004) and related to lithospheric refertilization by Na-rich carbonatite metasomatism of either Cr-rich, grt-bearing peridotites (Grütter et al., 2024) or Cr-rich, spl-bearing peridotites (Ikehata and Arai, 2004). Notably, the V content is much greater than that reported in P-type inclusions (542 ppm; Figure 5). Such enrichment would require crystallization from (or equilibration with) a parental melt with 100–300 ppm of V (Wang et al., 2019) that, in the absence of data for carbonatites, is consistent with 100–300 ppm measured in natural kimberlitic rocks (Kostrovitsky et al., 2023). Thus, the minerals found entrapped in COMSK3 diamonds are representative of a lherzolitic environment, then P-type; however, the enrichment in kosmochlor is possible evidence of fluid–rock interaction as described by reaction (3). The proposed presence of carbonate halos, by analogy with the literature, might be further evidence of CO₂-rich fluids from which diamonds might have formed.