Perovskite-bearing crystal-controlled oxide-silicate mantle xenoliths: Resolution to controversial origins?

Abstract

Classic lamellar clinopyroxene-ilmenite intergrowths (type 1) are extended to include discovery of olivine-ilmenite-perovskite-wüstite (type 2) and olivine-spinel-perovskite (type 3) xenoliths in kimberlites from Liberia. Low titanium solubilities in olivine, garnet, and pyroxene cannot account for exsolution-like relations. Because the oxides coexist with high-pressure perovskite-structured silicate minerals in diamond, a permissive conclusion is that type 1 to type 3 xenoliths are of super-deep origin. Phase equilibria and thermodynamic studies show that type 1 xenoliths are stable at P > 80 GPa, with type 2 and type 3 at 35 to 50 GPa consistent with an origin in anomalous large low shear velocity province bodies anchored at the core-mantle boundary. Dissociated precursor perovskite-structured Ca-Fe-Ti bridgmanite is proposed and is indirectly supported by the copresence of type II diamonds with a sublithospheric lower mantle origin.

Perovskite structured mineral precursors from the lower mantle are invoked as the origin of oxide-silicate xenoliths.

INTRODUCTION

Glossary .

Mineral abbreviations (1)

Bdm, bridgmanite; Bri, breyite; Chr, chromite; Cpx, clinopyroxene; Dvm, davemaoite; En, enstatite; Fo, forsterite; Fper, ferro-periclase; Gk, geikielite; Grt, garnet; Hir, hiroseite; Hc, hercynite; Ilm, ilmenite; IW, iron wüstite; Kir, kirschsteinite; Mgs, magnesite; Mfr, magnesioferrite; Mag, magnetite; Maj, majorite; Mtc, monticellite; Ol, olivine; Opx, orthopyroxene; Per, periclase; Phl, phlogopite; Prv, perovskite; Px, pyroxene; Rwd, ringwoodite; Spl, spinel; Tsc, tschaunerite; Uspl, ulvöspinel; Wds, wadsleyite; Wüs, wüstite; WM, wüstite-magnetite.

Other abbreviations

CMB, core-mantle boundary; EDS, energy-dispersive spectroscopy; EMPA, electron microprobe analyses; LIPS, large igneous provinces; LLSVP, large low shear velocity provinces; P, pressure; P-T pressure-temperature; SEM, scanning electron microscopy; TZ, transition zone; UHP, ultrahigh pressure; XRF, x-ray fluorescence.

J. D. Bernal in 1936 first suggested (2) that increases in seismic velocities, deeper than the Moho, may not be due to density changes in petrological rock type but to crystallographic structural changes in mineralogy. A. E. Ringwood confirmed the proposition two decades later in high-pressure, high-temperature experiments, that orthorhombic olivine is structurally transformed to modified cubic and then cubic spinel (2). The former (wadsleyite) defines the upper boundary to the transition zone (TZ), at the 410-km seismic discontinuity; the latter (ringwoodite) becomes stable at 520 km but decomposes to Bdm [high pressure-temperature (P-T) Prv-structured MgSiO₃] + Feper at 660 km, which defines the lower boundary to the TZ (3, 4). Shocked meteorite studies, mineral inclusions in diamonds (5), and advances in mineral physics have established the importance of crystallographic structural changes in an understanding of deep Earth dynamics. Distinctions in diamonds from the subcontinental lithosphere, the TZ, and lower mantle are readily made (6–10). Diamond inclusion stratigraphy, mineral ages, and petrological rock affinities (e.g., eclogitic and peridotitic), evidence for fluids, including ice (11), bear directly on deep Earth C and N cycles, the controversial beginning of Plate Tectonics (12), and the origin of deep focus earthquakes (13). From this backdrop, attention is drawn to the recognition of a suite of unusual xenoliths in a diamond-bearing kimberlite pipe and associated dikes from northwest Liberia. From field and analytical studies (14), the Ol + Ilm and Ol + Spl xenoliths are notable because these have inclusions of Pvr + Ol + Wüs and Pvr + Spl, which are compositionally, texturally, and morphologically distinct from late-stage, low P-T Prv + Ilm/Spl in Ol-rich kimberlites. From a search of numerous rock geochemistry databases, the closest match are nelsonites (Fe-Ti oxide apatite) in anorthosites and Ti-rich orange Si glasses on the Moon. Foskorites (Ol + Mag) in carbonatites and meteoritic pallasites (Ol + Fe) lack Ti and potential xenoliths in the mantle sample, viz., MARID (mica, amphibole, rutile, and diopside), PICs (phlogopite, Ilm, and Cpx), and metasomites have low Ti and high Si (https://doi.org/10.5194/essd-11-1553-2019).

With the discovery of silicate perovskite inclusions in diamond from Brazil, South Africa, and Guinea and the repeated finds of large, irregular-shaped sublithospheric type II diamonds, e.g., The Star of Sierra Leone (969 ct), Peace Diamond (709 ct), Meya Diamond (476 ct), and unnamed 400- to 800-ct stones from neighboring Liberia, the West Africa Man shield is currently of considerable interest as a previously unidentified kimberlite province sourcing magmatic and mantle-derived lithologies from depths of >660 km. These samples (xenoliths) are orders of magnitude more abundant than diamonds (1 to 2 parts per million), so every cratonic kimberlite is predicted to contain some deep Earth minerals that co-equilibrated with diamond. As importantly is that the heat source driving eruption must be deeper than the deepest diamonds recovered. Global synchroneity of the two major kimberlite eruptive events in Earth’s history occurred at ~100 million and 1 billion years, and both are linked to large igneous provinces (LIPS) that erupted >10⁵ km³ of lavas over short periods 1 to 2 million years and superchrons (atypical behavior of Earth’s magnetic field when conventional North-South reversal episodes ceased). The extraordinarily deep (600 to >1000 km) origins of type II diamonds, including microdiamonds (15), coupled with the properties of kimberlite eruptions have led to the suggestion that superdeep global volcanism, the deepest on Earth, was triggered by thermal release in plumes from the D″ layer (16) or from large low shear velocity province (LLSVP) seismic anomalies also at the core-mantle boundary (CMB) (17). The aim of this report is to detail the possible link between the exotic Pvr-bearing xenoliths and a deep-mantle origin by using the robust high P-T experimental data extant in subsystem, mineral equilibria studies of Pvr-Mg-Fe oxides and Pvr silicates. These are foundational to initial estimates of P-T depths of origin. The subsystem approach is consolidated in an existing multicomponent study, which at TZ and lower mantle pressures has the mineral endmembers Pvr-Geik-Brg-Dvm. In Ca-Mg, Ti-Si space, the P-T stabilities of xenolith bulk compositions can be assessed, albeit structurally transformed mineralogically. On the basis of a large analytical database of the new type 2 and 3 xenoliths, the most conservative conclusion drawn here is that the mineral assemblages are reconstituted high P-T rocks of deep-mantle origin. However, to confirm this initial conclusion, state-of-the art analytical studies are required and will be made on the following: search for additional Pvr- bearing xenoliths in heavy mineral mine concentrates to undertake systematic studies of inherited high P-T signatures in Ol, Ilm, and Spl and to substantiate that the xenoliths are reconstituted high P-T protoliths of TZ or lower mantle origin. Other studies will evolve as the significance of hand-on naturally occurring samples are recognized to substantiate the theoretical, seismological, and thermodynamic views of Earth’s true structure. To this end, the following sections are a detailed petrographic study that includes mineral abundances, mineral chemistry, and bulk compositions as a prelude to discussions of the data in an experimental framework of depths of origin.

RESULTS

Xenolith descriptions

Type 1 Cpx-Ilm intergrowths

Among 16 silicate-oxide–rich xenoliths recovered from the Camp Alpha Kimberlite Complex to date are 4 new and exotic (Fig. 1) and 12 relatively common Ilm-Cpx lamellar intergrowths (Fig. 2A). Recognized in kimberlites worldwide (18), the latter are classified here as type 1 and are texturally similar in description and bulk composition to those found globally.

Fig. 1. Mounted, cut, and polished almond-shaped xenoliths (4 to 5 cm × 1 to 2 cm) of lamellar and rod oxide-silicate intergrowths encased in thin veneers of kimberlite.

The dark oxides are ilmenite in (A) to (C) and spinel in (D). The greenish host in the four samples is olivine + metasomatic monticellite.

Fig. 2. Type 1 xenoliths Cpx-lim.

(A) Clinopyroxene-ilmenite intergrowths with subsolidus reduction of ilmenite forming crystallographically controlled lamellar spinel at high (B) and low (C) magnifications and with SEM compositions displayed for individual minerals and for integrated mineral assemblages.

Clinopyroxenes, with few exceptions, are narrow in composition (Wo 35-38, En 52-56, Fs 8-10 mol %), whereas lamellar Ilm has a broader range (Ilm 45-58, Gk 32-36, Hem 4-20 mol %) in composition. The bulk compositions of 2 large (2 to 3 cm) and 40 (0.5 to 2 cm) combined fragments of type 1 xenoliths were determined by x-ray fluorescence (XRF) analyses (Table 1). Ilmenite has undergone subsolidus reduction in all samples studied here with variable concentrations of Spl lamellae along (0001) Ilm planes (Fig. 2B) (19). Reduction-“exsolution” blebs are in the range of <5 to 10 μm. Ferric iron is variable, but because Ilm (48 mol %) and Gk (35 mol %) in the Liberia samples are similar to the suite of intergrowths from other localities, it is reasonable to assume that subsolidus reduction in Ilm-Cpx intergrowths is widespread (19). Subsolidus reduction of Ilm to Ilm + Spl is confirmed, in texture and chemistry, by (20) under controlled fo₂ conditions at 10⁻⁸ bars 1100°C to 10⁻¹⁵ bars 950°C [i.e., magnetite-wüstite (MW)–iron wüstite (IW)].

Table 1. Bulk xenolith XRF analyses.

Type 1 xenoliths (Cpx + Ilm); type 2 (Ol + Ilm) with inclusions of (Pvr + Ol + Wüs); type 3 (Ol + Spl) with inclusions of (Pvr + Spl). Note the differences in SiO₂ and MgO contents between the Cpx-bearing type 1 samples and the Ol-bearing type 2 xenoliths. In addition, type 3 is distinctive in having the highest CaO content in the entire suite, a reflection of high modal perovskite. Trace elements Ni and Cr (last two columns), hallmarks of mantle xenoliths, are low in concentration with minor variations, except for type 3 that is characterized by spinel rather than ilmenite in the primary assemblages. SC and SL refer to sample site designations.

	SiO2	TiO2	Al2O3	Fe2O3	MnO	MgO	CaO	Na2O	K2O	P2O5	SUM	NiO	Cr2O3
Sample														Type
SL25A	37.244	15.024	1.857	17.671	0.215	15.47	11.486	1.039	0.221	0.023	100.25	0.06	0.061	1
SL25B	37.4	14.835	1.858	17.523	0.215	15.47	11.525	1.055	0.227	0.023	100.13	0.062	0.056
Average	37.32	14.93	1.86	17.60	0.22	15.47	11.51	1.05	0.22	0.02	100.19	0.06	0.06
														Type
SL40A	35.017	16.625	1.953	19.981	0.173	15.759	9.23	0.935	0.051	0.016	99.74	0.091	0.299	1
SL40B	34.911	16.758	1.952	20.066	0.178	15.728	9.236	0.936	0.053	0.014	99.83	0.094	0.3
Average	34.96	16.69	1.95	20.02	0.18	15.74	9.23	0.94	0.05	0.02	99.79	0.094	0.30
														Type
SC14-19A	29.102	14.514	1.067	21.024	0.265	24.093	9.92	0.155	0.124	0.136	100.40	0.065	0.016	2
SC14-19B	28.985	14.495	1.058	20.876	0.27	23.956	9.936	0.147	0.125	0.134	99.98	0.062	0.011
Average	29.04	14.50	1.06	20.95	0.27	24.02	9.93	0.15	0.12	0.14	100.19	0.06	0.01
														Type
SC14-29A	25.761	19.21	0.406	20.607	0.243	26.869	6.489	0.283	0.058	0.066	99.99	0.112	0.104	2
SC14-29B	25.693	19.145	0.398	20.657	0.245	26.743	6.605	0.267	0.055	0.074	99.88	0.114	0.101
Average	25.73	19.18	0.40	20.63	0.24	26.81	6.55	0.28	0.06	0.07	99.94	0.11	0.10
														Type
SC14-30A	28.147	15.5	0.433	19.127	0.213	27.203	8.721	0.273	0.056	0.118	99.79	0.116	0.088	2
SC14-30B	27.999	15.627	0.435	19.146	0.208	27.025	8.771	0.282	0.054	0.117	99.66	0.113	0.088
Average	28.07	15.56	0.43	19.14	0.21	27.11	8.75	0.28	0.06	0.12	99.73	0.11	0.09
														Type
SC14-28A	28.831	14.676	0.632	17.37	0.127	17.886	19.54	0.311	0.078	0.047	99.50	0.104	0.365	3
SC14-28B	28.843	14.952	0.633	17.197	0.127	18.024	19.502	0.334	0.074	0.043	99.73	0.105	0.369
Average	28.84	14.81	0.63	17.28	0.13	17.96	19.52	0.32	0.08	0.05	99.61	0.10	0.37

Type 2 Ol-Ilm-Pvr xenoliths

Three (SC14-19, SC14-29, and SC14-30) type 2 xenoliths are similar in mineralogy but differ to a small degree in modal abundance (Ol 70-75, Ilm 20-25, and Prv 5), mineral composition (Table 2), density (3.016 to 3.078 g/cm³), and XRF bulk rock analyses (Table 1). Host Ol (Fo 87-89) has interstitial stubby lamellar (25 to 50 μm) Ilm 55-66 Gk 31-38 Hem 2-7 in parallel flow patterns. Randomly distributed coarser (5 to 10 mm) Ilm 59-64 Gk 28-40 Hem 6-7 grains with reduction-exsolution lamellae of Spl (Uspl 45-49), texturally similar to Fig. 2 (B and C), are also present. Perovskite [~1 weight % (wt %) each of SrO and Na₂O] crystals (1 to 4 mm) are pleochroic and weakly anisotropic (i.e., noncubic) and exhibit inferred orthorhombic cross sections of pyramid {111}, prism {110}, and basal pinacoid {001} faces (Fig. 3). Perovskite crystals have abundant Ol (Fo 90-93) with droplets of Wüs (Fig. 4). Scanned scanning electron microscopy (SEM) bulk compositions are listed in Fig. 4 and plotted in Fig. 5. A common groundmass feature is the presence of secondary, metasomatic monticellite (Mtc 20-30 Fo 60-70 Fa 10-20) that ranges from incipient to advanced replacement of Ol.

Table 2. Representative oxide mineral compositions in Pvr-bearing xenoliths.

Ilm and Spl show the largest ranges in composition and textural relations. Primary Ol and Pvr compositions are given in the text along with the variations in composition of metasomatic monticellite. Gm, groundmass; Geik, geikielite; Hem, hematite; Ilm, ilmenite; Ol, olivine; Pvr, perovskite; Spl, spinel; Usp, ulvöspinel.

ILMENITE
	Type 3				Type 2
Oxide	Ilm in Pvr	Ilm in Pvr	Ilm in Pvr	Ilm in Pvr	Gm Ilm	Ilm in Ol	Ilm ± Spl	Ilm ± Spl	Ilm ± Spl	Gm Ilm
TiO2	55.52	56.06	56.03	56.63	52.42	55.26	52.86	51.36	54.21	56.38
FeO	32.43	31.24	31.23	31.18	37.03	35.46	37.99	39.63	32.73	31.09
MgO	10.28	11.24	11.21	11.59	10.01	9.01	8.68	8.58	11.92	11.68
MnO	0.23	0.7	0.2	0.26	0.47	0.27	0.48	0.39	0.24	0.46
Total	98.46	99.24	98.67	99.66	99.93	100	100.01	99.96	99.1	99.61
Mole%
Ilm	64.1	55.9	56.1	55.9	53.3	66.1	64.7	59.9	56.7	55.1
Geik	35.8	40.8	41	40.9	32.6	31.4	27.9	26.7	38	38.6
Hem	0	3.2	2.9	3.2	14.1	2.5	7.4	13.3	5.2	6.3
SPINEL
	TYPE 3
	Spl in Ilm	Spl in Ilm	Spl in Pvr	Spl in Pvr	Spl in Pvr	Gm Spl	Spl Core	Spl rim	Gm Spl	Spl rim
TiO2	17.85	19.86	14.92	10.65	12.06	8.88	7.62	1.6	8.53	8.06
FeO	77.41	74.64	74.95	80.68	76.51	81.49	81.2	91.33	82.02	83.09
MgO	4.08	4.73	5.09	4.16	4.52	3.96	6.29	2.88	3.69	2.89
MnO	0.38	0.36	0.97	0.21	1.04	0.26	0.22	0.32	0.24	1.04
Cr2O3	0.16	0.03	4.08	2.82	4.49	3.96	4.71	4.71	2.98	3.86
Al2O3	0.12	0.38	0.25	1.47	0.27	0.54	0.14	0.14	1.23	0.7
Total	100	100	100.26	99.99	98.89	99.09	100.18	100.98	98.69	99.64
Fe2+	9.98	10.13	8.88	10.78	8.81	8.44	8.25	7.67	8.2	8.24
Fe3+	8.3	7.46	8.89	8.47	9.43	10.34	12.03	14.85	11.28	11.11
Mole%	Usp 45.3	Usp 49.1	Usp 40.3	Usp 47.1	Usp 38	Usp 36	Usp 44	Usp 21	Usp 35	Usp 25.7

Fig. 3. Type 2 olivine + ilmenite + perovskite ± wüstite and type 3 olivine + spinel + perovskite.

(A to C) Type 2 bluish perovskite with colorless olivine inclusions in a groundmass host of clear olivine, brown monticellite, and black ilmenite. (D) Dense intergrowth of perovskite + spinel in a host of brown monticellite, clear olivine + black spinel. Polished thin sections in transmitted light microscopy; field diameters are 0.5 mm. Ilm, ilmenite; Mtc, monticellite; Ol, olivine; Pvr, perovskite; Spl, spinel; p, density.

Fig. 4. Perovskite inclusion in a type 2 (olivine + ilmenite) xenolith.

(A) A backscattered SEM image of gray perovskite, black olivine, and bright white blebs of wüstite. (B) Element distribution map of Mg-rich olivine (blue), Ti-rich perovskite (yellow), Fe-rich wüstite (orange), and Fe-Ti rich ilmenite (tangerine). (C) Element distribution image of a selected area at a higher magnification of the assemblage with Mg-rich olivine in blue, Fe-rich wüstite in orange, and host perovskite in black. Note that wüstite throughout is exclusively in olivine. Bulk compositions of CaO, MgO, TiO₂, and SiO₂ from SEM area scans are listed with ranges and averages. The overall composition reduces to perovskite 55 olivine 45. Within experimental error and taken at 50:50, the bulk composition reduces to a stoichiometric compound CaMg₂TiSiO₇ discussed in the text and plotted in Fig. 5.

Fig. 5. Experimental phase diagram for the system enstatite-geikielite modified from (35).

Although the end member is a pyroxene, application to the present study, with olivine in type 2 and 3 xenoliths, is warranted because wadsleyite is stable below 20 GPa. Possible reequilibration paths for the SEM determined bulk composition of perovskite 55 olivine 45 and for the ideal stoichiometric compound CaMg₂TiSiO₇ are shown as dashed lines. The vertical solid lines from high to lower pressures are possible equilibration paths for the bulk compositions of type 1 (green), 2 (red), and 3 (blue) xenoliths, stripped of metasomatic monticellite. The starting composition for type 1 is at a higher pressure relative to type 2 and type 3 as discussed in the text and shown graphically in Fig. 8. The relevant major element (Mg, Ca, Ti, and Si) compositions are listed in table S1. Brd, bridgmanite; MST-Brd, high-Ti Brd; Wad, wadsleyite Mg₂SiO₄; Web, weberite-structured MgTiSi₂O₇; Rt, rutile; Px, pyroxene.

Type 3 Ol-Pvr-Spl xenolith

Sample 14-28 has a primary mode of Ol 70 Prv 25 Spl 5 (Fig. 3D). Representative mineral compositions are given in Table 2. Polycrystalline Ol (Fo 85-90) is host to euhedral and subhedral laths of zoned Spl from cores (Uspl 32-44) to rims (Usp 21-24) in bands and as scattered grains throughout the xenolith (Fig. 3D). The most distinctive feature of the type 3 xenolith is the uniformly abundant lamellae of crystallographically controlled Spl (Uspl 35-47) in euhedral Prv (0.5 wt % SrO and 0.7 wt % Na₂O) (Figs. 6 and 7). Scanned SEM bulk compositions reduce to a mineral formula of Ca₂FeTi₂O₇ (Fig. 6). Perovskite with Fe oxide lamellae, and with comparable Sr, Na, and low Fe contents, is known in only one other locality, the Dutoitspan kimberlite, South Africa (21). Discrete Ilm (53 to 55 mol %), as in type 1 and type 2, contains reduction exsolution Spl (Uspl 40-50) lamellae. The groundmass has cloudy and zoned monticellite (Mtc 54-65 Kir 14-38 Fa 8-22) after Ol.

Fig. 6. Type 3 sample SC14-28.

(A) Euhedral perovskite (light gray) with crystal-controlled spinel lamellae (white), mantled by dark gray metasomatic monticellite. Note the noncubic habit of groundmass spinel (white). Mineral compositions are listed. The rectangles (black) delineate SEM scans for bulk composition determinations of the perovskite–lamellar spinel assemblages. Additional examples of scanned areas are shown in (B) and (C). (D) The average bulk composition of 12 determinations reduces to, and is compared with, a stoichiometric compound (Ca₂FeTi₂O₇) discussed in the text and evaluated in Fig. 7.

Fig. 7. Type 3 bulk Pvr + Spl.

(A) Simplified Fe-Ca-Ti oxide ternary: Stable experimental compounds (circles) are shown at 15 to 20 GPa and 1400 to 1600°C (26–31), and bulk compositions of Pvr + Spl (B), star in (A), and the Kao compound CaTiFe₂O₆ (43) are in relation to noncubic spinel habits typical of type 3 xenoliths (C) and tschaunerite (D), the high-pressure analog of ulvöspinel Fe₂TiO₄ (70), along the join FeO-TiO₂ (A). Reduction in the ternary refers to the possible derivation of the type 3 compound Ca₂FeTi₂O₇ (Pvr + Spl) from experimentally determined stable phases along the joins Pvr-Ilm and Pvr-Usp (A). Ilm, ilmenite; Ol, olivine; Spl, spinel; Usp, ulvöspinel.

DISCUSSION

Constraints on the origin of Ilm-Cpx and Ilm-Sp-Ol xenoliths

Perovskite is the distinguishing feature in type 2 and type 3 xenoliths and is found as discrete inclusions in diamonds (22–25). For these reasons, both compositionally and structurally, the xenolith inclusions of Pvr-Ol-Wüs and Pvr-Spl are examined as potential P-T indicators before an evaluation of the origin of the xenolith minerals (Ol + Ilm and Ol + Spl) hosting the inclusions.

Type 2 Pvr-Ol-Wüs inclusions

Perovskite is stable to at least 60 GPa, ~1800 km depth (26, 27). Phase equilibria at high P-T demonstrate extensive to complete solid solution for CaTiO₃-CaSiO₃ (28), CaTiO₃-FeTiO₃ (29, 30), and FeTiO₃-MgTiO₃ (31), is limited for CaSiO₃-MgSiO₃ along an upper mantle adiabat (32), but complete at P > 85 GPa in the presence of Fe and Ti (33, 34). Single-phase En-Gk is stable at P > 25 GPa, 1600°C (35). An inevitable consequence of these experimental results is that with decompression and cooling, any high P-T single supersolvus solid solution compound will decompose to two phases that are crystallographically coherent.

This is a possible origin for the type 2 inclusions by decomposition of a mineral precursor, specifically Ca-Ti-Bdm in the system CaO-MgO-TiO₂-SiO₂ at 20 to 97 GPa and 2000 K (33) and in the solid solution series MgSiO₃-MgTiO₃ (35). In the latter, equilibration at progressively lower P passes through a series of three-phase assemblages that include Bdm, Wds, and structural weberite-type (36) MgTiSi₂O₇ (Fig. 5). Perovskite crystals in type 2 xenoliths have irregular to spherical blebs of refractory Ol (Fo 93-95), reminiscent of melts (Fig. 4). Within the olivine are equally rounded blebs of Wüs (98 wt % FeO with trace contents of Mg, Cr, and Ti). The Wüs is neither the product of oxidation, as in terrestrial basalts with magnetite (37), nor reduction of lunar basalts with metallic Fe (38). A plausible origin is that Wüs is a residual constituent from the decomposition of Brg-Hir solid solution.

$$2{\mathrm{FeMgSiO}}_3={(\mathrm{Fe}\mathrm{Mg})}_2{\mathrm{SiO}}_4+\mathrm{FeO}$$ (1)

If Pvr + Ol is derived from the decomposition of Ca-Ti-Bdm in the system MgO-CaO-TiO₂-SiO₂ (33), progressive decreases in P-T can be estimated from SEM bulk compositions (Fig. 4) and application to (35). From point counting and mineral mode recombination, Ol 55 Pvr 45 is close to ideal CaTiO₃ + Mg₂SiO₄ = CaMg₂TiSiO₇. Structural weberite (MgTiSi₂O₇) forms in a closely related system (MgTiO₃ + MgSiO₃ in which Pvr replaces Gk and Ol replaces En) that is stable at 15 GPa and 1600°C. Wadsleyite (Ol) and Ti-Bdm (MST Bdm in Fig. 5) have stabilities between 17 and 20 GPa at 1600°C.

Type 3 Prv-Spl inclusion

The type 3 xenolith has ~20 modal % Prv in a host of Ol + two textural forms of ~5 modal % Spl in crystal-controlled lamellae in Pvr (Figs. 6 and 7C ) and as euhedral noncubic Spl in the groundmass (Fig. 7D). The average SEM bulk composition of 11 Prv-Spl lamellar assemblages (Figs. 6 and 7B) is plotted in the FeO-CaO-TiO₂ ternary space (Fig. 7A). The compound lies on the Prv-FeO (i.e., redox reduced) solid solution join. High P-T experiments along the join CaTiO₃-FeTiO₃ show that CaFeTi₂O₆ and CaFe₃Ti₄O₁₂ are stable at 15 GPa and 1150° to 1400°C ~450 km depth in the TZ ( 29, 39–43). The bulk composition of Prv-Spl reduces to Ca₂FeTi₂O₇ (Fig. 6) and reconstituted is FeTiO₃ + 2CaTiO₃ placing it along an extension of the Prv-Ilm solid solution join. In a mineralogical study of the Kao kimberlite in Lesotho (43), an intermediate phase of Prv-Ilm (Hem 30–Ilm 70) was reported with 23.8% CaO, 33.2% Fe₂O₃, and 36.3% TiO₂ that at 50 mol % reduces to CaFe₂TiO₆. This is equivalent to the Ca-depleted, redox-reduced phase (CaFeTi₂O₆), labeled “reduced Kao” in Fig. 7A, and is closely allied to CaFe₃Ti₄O₁₂ that is stable at 15 GPa. This is a potential candidate for subsolidus reduction to achieve the type 3 assemblage of Pvr + Spl (Figs. 6 and 7B).

Another candidate is CaFe₂Ti₂O₇, along the join Pvr-Uspl. This phase is attractive because the type 3 xenolith is distinguished by zoned noncubic orthogonal crystals of Uspl-Mt in the xenolith groundmass (Fig. 7D). The end members listed along the ternary joins (Fig. 7A) are Pvr-structured at high P and, in principle, so are all phases along the solid solution joins. Candidate phases, with projected reduction paths, are shown in (Fig. 7A): One is related to the Kao mineral (Pvr-Ilm) and the other to the join Prv-Uspl. Reduction of the Fe-Prv compound to form Pvr + Spl is consistent with the pervasive subsolidus reduction of the three xenolith types (Fig. 2, B and C) and of discrete, macrocrystic Ilm in kimberlites (18, 19). Given the confined compositional space in which both unusual assemblages occur, that both are structurally in accord with high P-T experimentally determined stabilities is permissive evidence for equilibration in the TZ. That the Kao (Lesotho) mineral lies along the Pvr-Ilm join is expected, but that the Liberia Pvr + Spl assemblage, more complex in mineralogy and composition, should fall exactly along the join Pvr-Wüs is unexpected and points to a common origin. A sublithospheric origin is now supported by the recovery of a 108-ct type IIa pink diamond on 21 April 2023 at Kao (https://stormmountaindiamonds.com/108c-pink.html).

An additional P-T estimate for the type 3 Ol-Pvr-Spl assemblage may be estimated from Ol-Mt equilibria at 10.5 GPa in which a series of new spinelloids is reported with the Wds structure (44, 45), demonstrating the importance of Fe³⁺ in the mineralogy of the TZ and potentially in the xenoliths described here. The groundmass xenolith textures in type 2 and type 3 are substantially coarser and less well ordered than the intergrowth silicate-oxide assemblages in type 1 (Fig. 2) or in those from ultrahigh pressure (UHP) metamorphic settings. Slow cooling, subsolvus exsolution from Ol-Spl or Cpx-Ilm is unlikely from mantle melts given the inferred rapid rise of kimberlites and diamond preservation from the source to the surface. The alternative, and more likely, is decompression-induced decomposition of a lower mantle precursor in the TZ or subcontinental lithosphere. The unquenchable nature of Bdm points to spontaneous breakdown, not slow cooling, because even with the high confining P of diamond inclusions, decomposition of Bdm, Bri, and Maj occurs (10, 46–50).

P-T estimates of type 1, type 2, and type 3 xenoliths

Constraints on the origin of Ilm-Cpx and Ilm-Sp-Ol xenoliths

The origin of type 1 Ilm-Px lamellar intergrowths, by exsolution, decomposition, reaction/replacement, or eutectic/peritectic crystallization, is long debated and remains controversial. Lamellar textures are typical of exsolution between two minerals that share a common crystal structure, form a solid solution series at high T, and unmix with cooling below a solvus. Symplectites, by definition, result from the decomposition of a metastable precursor that breaks down to a vermiform, binary mineral complex. This worm-like or sigmoidal texture may also arise from mineral reactions that lead to complex binary mineral intergrowths as in myrmekites. Eutectic or binary crystallization of two minerals at temperatures less than their respective melting points results in graphic or cuneiform textures (51, 52).

Clinopyroxene-Ilm intergrowths have an exsolution-like appearance (Fig. 2A) with sharply tapered, evenly spaced lamellae that are crystallographically related with (0001)_Ilm//(100)_Cpx (53). Exsolution, however, would require endmember crystal structure compatibility, which is not the case. An origin by decomposition of a precursor, mantle-derived mineral has been proposed for Grt (54) and Cpx (55), but both are unlikely because the TiO₂ contents (1 to 4 wt %) are insufficient (56) to produce the high (30 to 50%) modal contents of Ilm.

Binary mineral reactions (e.g., Grt + Ol = Spl + Px) are also unlikely because the symplectic textures are vermiform/sigmoidal, and the complex intergrowths are structurally dissimilar (57). In exploring an origin by eutectic crystallization, the high P-T experiments by (58, 59) produced Cpx-Ilm intergrowths. The former in quenched charges, but with controlled cooling, the texture was sigmoidal, and because the starting material was crushed and separated Cpx and Ilm, the end-product is best described as the relic effect of crystal-bond memory, not primary crystallization. Melting experiments by (59) used crushed kimberlite as the starting material. Isolated Ilm-Cpx intergrowths were produced but with structurally incoherent cuneiform textures in assemblages that included Grt, Ol, Mgs, and K-Ca-Mg quenched liquids. Although not precisely duplicated, the results are potentially valid because short-term experiments do not simulate long-range equilibration in nature.

Recognizing the limitations of exsolution and of the inability of Gt and Px to produce structurally coherent Ilm-Cpx intergrowths, Collerson et al. (60) attempted single-phase homogenization and partially succeeded in producing a multiphase assemblage of a Ca-Si-Ti Pvr at 30 GPa and 1800°C, suggesting that the type 1 intergrowths originated from a high P-T Pvr-structured progenitor. This was considered in a series of innovative high P-T experiments by (33) in the system CaO-MgO-TiO₂-SiO₂ expressed mineralogically as Brg-Dvm-Pvr-Gk (Fig. 8, A and B) and by (35) for Ti-rich Bdm (MgSiO₃-MgTiO₃) in the system MgO-SiO₂-TiO₂ (Fig. 5).

Fig. 8. Microscope images of polished thin sections of the four silicate-oxide assemblages discussed in the text in plane polarized light at a diameter scale of 0.5 mm.

(A to D) Clinopyroxene (yellow and blue in type 1), olivine (clear), monticellite (brown), with black ilmenite in type 1 and type 2 and spinel in type 3; perovskite is bluish in type 2 and is masked in type 3 with intergrown spinel. (E) The experimentally determined mineral template at ambient conditions (enstatite-wollastonite-geikielite-perovskite) is from the multicomponent system CaO-MgO-TiO₂-SiO₂ with solvi determined at 1725°C and 20 to 97 GPa (33). Bulk compositions for global type 1 xenoliths (clinopyroxene-ilmenite, diamond symbols; orthopyroxene-ilmenite, squares) lie respectfully in the 55- and 35-GPa solvi (33). (F) Silicate oxide intergrowths from this study, determined by XRF analyses (red symbols) and stripped of secondary metasomatic monticellite (blue symbols) using the mineral mode/chemistry method as in “Rock Maker” (table S1). The two compositions are joined by dashed lines. Type 1 is unaffected by metasomatism with virtually no difference between quantitative XRF and semiquantitative analytical Rock Maker [circled in (F)]. By contrast, there are expectedly large displacements between the XRF and the removed metasomatic monticellite, placing the primary compositions of type 2 (olivine + ilmenite + wüstite + perovskite) and type 3 (olivine + spinel + perovskite) xenoliths firmly in the 35-GPa solvus. These compositions lie in the same solvus as the orthopyroxene-ilmenite population from other localities (E).

Neither experimental system is fully appropriate to cover the bulk compositions of the three xenolith types: Ca is absent in (35), and both lack Fe. The latter is readily accommodated in Brg as FeSiO₃ (heroseite), with the interesting property of Fe dissociation (3Fe³⁺ = 2Fe²⁺ + Fe⁰) making it a prominent sink for ferric Fe in the lower mantle (61, 62). For Ca, there is a limited solid solubility of davemaoite (CaSiO₃) in Brg (33), but increased miscibility to a single phase can be achieved in the presence of high iron (Fe²⁺ Fe³⁺) concentrations (63, 64) and extreme P (>90 GPa) as determined experimentally by (33) and thermodynamically by (62, 64). A similar effect is demonstrated for Ti in which Ilm-Cpx intergrowths are stable experimentally at ~55 GPa and Opx-Ilm at 35 GPa (33), and thermodynamically for the binary Pvr-Brg with the corresponding values of 80 and 45 GPa, respectively (34). In addition, proof-of-element (Fe-Ti-Al)–enriched Brg is reported as diamond inclusions from Brazil (48).

It is within this comprehensive geochemical framework that the potential P-T origin of the three xenolith types is now considered. Type 1 and type 2 xenoliths differi in mineralogy and modal abundance that is reflected in XRF bulk compositions in SiO₂ and MgO (Table 1), from Cpx in the former and Ol in the latter. The type 3 xenolith is further distinguished by enriched CaO, related to higher modal Prv. From the assemblages of Ol + Ilm and Spl + Prv, the essential elements are Mg, Ca, Ti, and Si and the appropriate molar ratios, in applying the high P-T experimental data of (33, 35) for the three xenolith types, are Ti/(Ti + Si) versus Mg/(Mg + Ca). As expected in Fig. 8 (A and B), type 1 falls in the regional group of xenoliths reported from other localities with a cluster in the 55-GPa solvus. Type 1 is XRF-determined and is unaffected by metasomatic Mtc. The Mtc overprint in type 2 and 3 xenoliths was removed using the mineral mode compositional procedure (see Materials and Methods). The XRF values for the type 2 and type 3 xenoliths also fall in the 55-GPa solvus but stripped of Mtc are embraced by the 35-GPa solvus (Fig. 8B). It is important to note that these xenoliths occupy the same compositional space as the rarer Opx-Ilm xenoliths (Fig. 8A) in kimberlites, again validating the substitution here of Ol for En in exploring the P-T origins of the Pvr-bearing xenoliths.

If a value of 35 GPa is assumed and treating Prv and associated inclusions (Ol in type 2 and Spl in type 3) as internal P-T recorders, the former would have equilibrated at <20 GPa along a decompression trajectory (long dashed line in Fig. 5) to one of the Wds phase fields. The compositional boundary at ~16.5 GPa, where dissociation of Ti-rich Brg results is Wad + Rt + Gk, is important. In the Fe equivalent system, this would be Wad + Ilm or Wad + Spl. This assumes that the precursor was a Ti-rich Pvr-structured silicate, which is in accord with (35) who show that Wds contains 2 wt % TiO₂ at 14 GPa; Bdm has up to 13 wt % at 24 GPa; Mg(SiTiO₃) has from 29 to 49 wt % at P > 17 GPa; and weberite structured (MgTiSi₂O₇) reaches 43 wt % TiO₂ at 18 GPa. The bulk TiO₂ contents of Pvr + Spl (42 to 47 wt %; Fig. 6), Pvr + Ol (24 to 38 wt %; Fig. 4), type 1 (16 wt %; Table 1), type 2 (14 wt %; Table 1), and type 3 (10 wt %; table S1) are all in the range of TiO₂ solubilities in phases determined between 14 and 24 GPa and 1600°C.

With evidence for P at ~18 GPa, the unusual orthogonal morphology of Spl in type 3 (Fig. 7 B and C) can be circumstantially resolved: The magnetically familiar solid solution series Uspl-Mag undergoes a phase transition from cubic to tetragonal to orthorhombic CaFe₂O₄ with increasing P (65–68). Transformation pressures decrease with Ti content (Mag at ~30 GPa and Uspl at ~12 GPa). For compositions of lamellar Spl (Uspl 35-47) in Pvr and groundmass Spl (Uspl 32-44) in Ol, orthorhombic symmetry is expected at P > 25 GPa (66). The entire solid solution series is postspinel orthorhombic and is stable to at least 60 GPa (65, 66).

Titanomagnetite (i.e., Uspl-Mag_ss), reported in a diamond inclusion (69), was considered a breakdown product of a Ti-rich silicate (e.g., Bdm). However, an alternative interpretation is decomposition of an Uspl-Mag precursor (67).

The high P polymorph of Uspl is tschaunerite (Tsc; Fig. 7D) found in the intensely shocked (~25 GPa) Martian meteorite, Shergotty (70). The noncubic morphology and optical anisotropy of Spl in type 3 Prv (Fig. 7C) is consistent with orthorhombic symmetry and compositions for members along the Tsc-Mag solid solution series. Decomposition of Uspl-Mag is not observed, and the noncubic optical anisotropy of Spl is seemingly retained.

With the reduction-exsolution lamellae of Spl in Ilm, and of Wüs in Ol, the prevailing redox state in type 1 and type 2 xenoliths was in the fo₂ range of MW-IW (21). This is equivalent to estimates for the TZ (71–73) but with carbonated domains at the base (74), OH in Rwd (75), Fe³⁺ in Wds (76), and OH, Fe³⁺ in Rwd and Wds (77); the TZ is stratigraphically heterogeneous in redox potential. This is possibly recorded in type 3 with reduction-exsolution Spl lamellae in Ilm and Prv (MW-IW), and with zoned groundmass Spl that has oxidized cores (Mag 62) and even more oxidized rims (Mag 77).

All minerals identified in this study occur in superdeep diamonds, some to depths of 1000 km or more. Central to a more direct comparison is the justification that the origin of the three Ca-Fe-Ti-Mg–rich xenoliths resulted from the disproportionation of Brg. The xenoliths appear to be compositionally unique to the mantle, so this hinges on where a highly modified Brg is likely to be stable. As outlined above, P > 80 GPa is required, but as important: Is this in any way related to kimberlite magmatism? Single-phase Ca-Fe²⁺Fe³⁺Ti Brg-Dvm (davemaoite CaSiO₃) is predicted at high P-T conditions and is modeled around CMB rising plumes (62) or in LLSVP seismic anomalies at the CMB (63, 64). Although the origin and composition of the large (thousands of kilometer wide × tens of kilometer thick) thermochemical bodies remain uncertain (78), Dvm is considered a possible candidate (79), and in combination with Ca, Fe, and Ti, the inferred composition of the Brg progenitor is achieved. The proposition that the solid solution mixing of other elements in Brg invokes LLSVP and plumes at the CMB is in accord with disruption of these bodies as the driving force for the synchronicity of kimberlite intrusions, the eruption of LIPS, and the superchron behavior of Earth’s magnetic field (16, 17).

Supporting evidence for a deep-mantle origin is the copresence of type II sublithospheric diamonds, locally and regionally. These are assumed to be CMB-related and, therefore, of lower mantle origin. From seismic and experimental data, the composition of the lower mantle is controversial and is considered to be either peridotitic (pyrolite) with Mg/Si of ~1.3 or near chondritic with Mg/Si of ~1.0 (80). This difference is reflected in the inferred mineral modes of Brg (77 to 80%), Ca-Pvr (6 to 16%), and Fper (7 to 13%) in the former, and Brg (85 to 90%), Ca-Pvr (5 to 7%), and Fper (5 to 8%) in the latter (81, 82). Dominated by MgSiO₃ and with 93% Brg in the perovskitic model (80), the MgO (~36 wt %) content compares favorably with the MgO (34 to 36 wt %) contents of Ol-bearing type 2 and 3 xenoliths; the Mg/Si ratio (1.2 to 1.6 Av = 1.4), however, is distinctly pyrolytic. With respect to molar Mg# (Fig. 8E), the xenolith ratios are 0.88 and 0.96, well within the uncertainties for pyrolite (0.94) and chondritic (0.99) models [Ca is not considered in (80)]; the xenoliths, however, are depleted in SiO₂ (~25 versus ~48 wt %) and enriched in TiO₂ (~15 versus ~0.2 wt %), possibly indicative of LLSVP heterogeneity. Titanium is essential to the xenoliths discussed and in a pyrolytic lower mantle, where thermodynamic mixing with Ca and Fe is greatly enhanced to the point that solid state Opx-Ilm and Cpx-Ilm are predicted at ~50 GPa, and ~85 GPa along a 2000 K geotherm (34) in accord with the experimental data shown in Fig. 8.

The deep-mantle origin for the xenoliths reacted with and subsequently equilibrated in the TZ at ~15 to 20 GPa en route to the surface. With density phase transformations at this horizon, plumes from below are expected to mushroom and subducted slabs from above to stagnate. As a consequence, the TZ is inevitably a mélange and an effective barrier to rapid, deep-source magma penetration. It is here that the Pvr-bearing oxide-silicate xenoliths are considered to have paused, reacted, and equilibrated before eruption.

MATERIALS AND METHODS

Microscopy

Doubly polished thin sections of the xenoliths were systematically examined using transmitted (×10 to ×100 magnification) and reflected light polarizing microscopes (×10 in air and ×20, ×50, and ×100 oil objectives), carefully documenting assemblages in Ilm and Prv as redox indicators (20, 21). Quantitative compositions were obtained by electron microprobe analyses (EMPAs) and energy-dispersive spectroscopy (EDS) data by SEM. The SEM was indispensable because at high (×1000) or higher magnifications, a plethora of optically undetected mineral intergrowths readily account for equilibration, nonstoichiometry, and aberrant microprobe mineral totals. Type 2 and type 3 xenoliths have undergone Ca-Si metasomatism with the development of Mnt and K with minor Phl. Bulk rock XRF analyses provide a first-order estimate of P-T conditions, but compositions free of the metasomatic overprint are essential and were obtained by optically determined mineral modes in combination with compositions. A 10 × 10 eyepiece reticule and grids on photomicrographs were used to estimate mineral modes at ×10 and ×25 magnification. Approximately 60% of the groundmass is metasomatic, but relic Ol is always present, and in point counting, Mnt is regarded as former Ol. Being cognizant that mineral density is important in mineral mode composition calculations (21), the “Rock Maker” program by (83) was used to obtain a best estimate of premetasomatized compositions. Xenolith densities were determined using the Archimedes method with distilled water and applied temperature corrections and with quartz (2.65 g/cm³), apatite (3.3 g/cm³), topaz (3.5 g/cm³), and corundum (4.0 g/cm³) as standards.

Electron microprobe and SEM

EMPAs were determined on JEOL JXA 8900 with five wavelength spectrometers, three regular and two large crystal (positron emission tomography, LiF, and TAP) spectrometers with two gas flow detectors and three sealed detectors. Ten elements (Si, Ti, Al, Cr, Ca, Fe, Mg, Mn, Na, and K) were typically analyzed using natural and synthetic standards (21). Operating conditions were 15 kV, aperture current of 20 nano-amps, and spot size of 2 to 5 μm with ZAF correction factors. The SEM/EDS JEOL IT500h field-emission SEM with a Bruker Quantax 60-mm window SSD detector using the Esprit software was used for routine standardless (semiquantitative) analysis and calibrated with copper. Standards were periodically used to verify and test detector accuracy. Operating conditions were 15 kV, 20–nano-amp aperture current, with variable spot size and area scans: ZAF corrections were applied. Backscattered images and elemental maps were obtained at magnifications of 1000 to 20,000.

Ferric iron was calculated using (84). Structural formulae for the SEM EDS data used “GABBRO soft,” and endmembers for Mnt analyses were obtained from a University of Alberta program at https://www.eas.ualberta.ca/eml/?page=links#SW.

To judge the reliability of the recombination mineral mode method, Rock Maker compositions were compared with results from classic CIPW Norm calculations, and although far from ideal, because the xenoliths are silica deficient and alkali-free, the correspondence for type 1 is excellent. In addition, note that type 1 xenoliths in Fig. 8B are within experimental error for XRF and mineral mode determinations. For Ol-Ilm-Pvr in type 2 and for Ol-Pvr-Spl in type 3, agreement is within ±5%.

X-ray fluorescence

XRF compositions were determined at the University of Massachusetts (UMass), Amherst. Major elements (Si, Ti, Al, Fe, Mg, Ca, Na, K, and P) were measured as oxides on duplicate fused glass disks, prepared by fusing 0.6 g of sample with 6 g of a prefused Claisse Li-borate flux in an Eagon-2 automated fusion furnace. Powders were pre-ignited between 800° and 1020°C for several hours to remove volatiles (H₂O, CO₂, and S) and oxidize iron to Fe³⁺. Trace elements were measured on pressed powder pellets. Standard correction procedures were applied (https://www.geo.umass.edu/facilities).

Supplementary Materials

This PDF file includes:

Table S1