Superdeep Diamond Genesis Through Fe-Mediated Carbonate Reduction

Abstract

Superdeep diamonds and their syngenetic inclusions are crucial for understanding Earth’s deep carbon cycle and slab–mantle redox dynamics. The origins of these diamonds, especially their links to iron (Fe) carbides and ferropericlase with varying Mg# [=Mg/(Mg+Fe)_at], however, remain elusive. In this study, we performed high pressure–temperature (P-T) experiments (10–16 GPa and 1200–1700 K) across cold-to-warm subduction zones using a multi-anvil press. The results reveal a stepwise Fe-mediated carbonate reduction process for the formation of superdeep diamonds: MgCO₃ → Fe-carbides (Fe₃C/Fe₇C₃) → graphite/diamond. This mechanism explains two phenomena regarding superdeep diamonds: (1) anomalous ¹³C depletion results from kinetic isotope fractionation during ¹²C enrichment into the intermediate Fe-carbides; (2) nitrogen scarcity is due to Fe-carbides acting as nitrogen sinks. Ferropericlase [(Mg,Fe)O] formed during the reactions in our experiments shows Mg# variations (0.2–0.9), similar to those found in natural samples. High Mg# (>0.7) variants from lower temperature experiments indicate diamond crystallization from carbonatitic melts in the shallow lower mantle, while the broad Mg# range (0.2–0.9) from experiments at higher temperatures suggests multi-depth formation processes as found in Brazilian diamonds. These findings suggest that slab–mantle interactions produce superdeep diamonds with distinctive Fe-carbides and ferropericlase assemblages as inclusions, coupled with their ¹³C- and nitrogen-depleted signatures, which underscore thermochemical carbon cycling as a key factor in deep carbon storage and mantle mineralogy.

1. Introduction

Plate tectonics plays a fundamental role in crust–mantle exchange processes, with subducting slabs acting as conduits for marine sediments and oceanic crustal materials to enter the Earth’s interior. This global carbon conveyor system is pivotal in regulating lithosphere–climate feedback, moderating surface-to-mantle carbon fluxes, and reshaping deep carbon reservoirs. In warm subduction zones such as Cascadia, a substantial portion of carbon is released into the atmosphere through forearc metamorphic and melting decarbonation; in contrast, colder subducting slabs like those in Tonga preserve considerable carbon inventories down to depths exceeding 300 km [1,2]. Subducted carbon, primarily in the form of carbonates within altered oceanic crust [3], introduces distinctive carbon and nitrogen isotopic signatures into the mantle transition zone (MTZ) and lower mantle (LM), where reduction processes facilitate the formation of superdeep diamonds [3,4,5]. These diamonds, through the mineralogy of their inclusions, offer insights into the chemistry and dynamics of the deep mantle. For instance, hydrous ringwoodite [6] and Ice-VII inclusions [7] indicate the presence of water-bearing domains, at least locally, in the MTZ; while stishovite, a new aluminum silicate phase, calcium ferrite phase [5], and davemaoite CaSiO₃ inclusions [8,9,10] indicate the recycling of oceanic crust to extreme depths (≥1000 km). Despite these advances, the origin of superdeep diamonds, and particularly of their diagnostic inclusion assemblages, remains a subject of ongoing investigations.

Iron carbides (Fe-C) inclusions, such as Fe₂C, Fe₃C, and Fe₇C₃, are notable redox-sensitive precursors for diamond formation [11,12]. Their coexistence with native Fe and graphite suggests that diamond formation occurs under oxygen fugacities close to the Fe⁰/FeO buffer [13,14,15,16,17,18]. Fe-C phases can form through solid-state reactions between mantle-derived metallic Fe and subducted carbonates [19], with diamond nucleation driven by carbon supersaturation [20]. This process has been observed in meteorites [21] and replicated experimentally at shallow mantle conditions (5.6 GPa and 1700–1745 K) [22]. Additionally, the nitrogen dichotomy—nitrogen enrichment in cratonic and its depletion in superdeep diamonds [15]—implicates the involvement of Fe-rich precursors as nitrogen sinks. Despite these findings and insights, the role of Fe-C phases in superdeep diamond formation remains underexplored.

Ferropericlase [Fper, (Mg,Fe)O], a dominant inclusion (≥50%) in superdeep diamonds, exhibits perplexing variability in Mg# (=Mg/(Mg+Fe)_at). Contrary to pyrolitic model predictions, which suggest a concentrated Mg-rich composition of Mg# = 0.89–0.92 [23,24], Fper inclusions in natural specimens display a wide range of Mg# from 0.36 to 0.90. Note that specimens with Mg#<0.5, termed magnesiowüstite, are included under the Fper category for simplicity. Extremely Fe-rich varieties (Mg# = 0.36) have been identified in Brazilian diamonds [25]. Proposed explanations for the unusually low Mg# include (a) preferential partitioning of Fe from post-perovskite into Fper, or a deep origin from the Fe-rich D″ layer [26,27]; and (b) redox precipitation of Fper and diamond from Fe-rich carbonate [28]. However, the former is challenged by the absence of a Ni-Mn correlation with Mg# [29], and the latter lacks depth-specific diagnostics. The discovery of super-Mg-rich Fper (Mg# = 0.94) with trace Ni in Guinea diamonds further complicates this issue [30]. Other potential mechanisms, such as slab-induced mantle heterogeneity [31] and kinetic nonequilibrium during inclusion entrapment [32], remain speculative due to insufficient data. Understanding these Mg# anomalies is crucial for unraveling superdeep diamond formation and mantle chemistry.

Magnesite (MgCO₃), a common carbonate in subduction zones, remains stable in oxidized mantle regions up to LM pressures [33]. The redox conditions in these regions are influenced by Fe²⁺ disproportionation, a process that intensifies with depth and produces Fe⁰ and releases Fe³⁺ into majorite garnet or bridgmanite [17,18]. At slab–mantle interfaces, redox reactions between MgCO₃ and Fe⁰ result in the formation of Fe-C, Fper, graphite, and carbonatitic melts, all having been documented as inclusions in superdeep diamonds [13,25,34]. However, the intermediate role of Fe-C and the compositional constraints of Fper on diamond formation require further investigation. This study investigates redox reactions between MgCO₃ and metallic Fe⁰ through nonequilibrium experiments using a multi-anvil press. The experimental conditions span subduction thermal regimes from the deep upper mantle (UM) to the MTZ. Two temperature-dependent reaction mechanisms are identified, producing Fe-C intermediates and Fper with wide Mg# variations dependent on thermal and chemical gradients. These carbon reduction pathways integrate the global diversity of diamond inclusions into a unified redox framework, linking subduction thermochemistry to deep carbon storage.

2. Materials and Methods

2.1. Starting Materials and Multi-Anvil Press Experiments

Synthetic magnesite powder and high-purity Fe-foils (99.95%, 50 μm thick, Materials Research Corporation) were used as the starting materials. High pressure–temperature (P-T) experiments were carried out using the 2000-ton multi-anvil press at the University of Hawaiʻi at Mānoa (UHM). An MgO octahedron (10 mm edge-length) was utilized as the pressure transmission medium. Pressure calibration was achieved by monitoring the electrical resistivity changes associated with phase transitions in Bi [35] and ZnTe [36] at room temperature, and quartz–coesite, olivine–wadsleyite phase transitions at high P-T conditions. A Re heater embedded in LaCrO₃ thermal insulation sleeves was used for generating high temperatures, measured by a W_97%Re_3%-W_75%Re_25% thermocouple at the capsule top. The estimated temperature variation from the capsule’s center to its end was within 40 °C [37]. Hexagonal boron-nitride (hBN) was used as the sample capsule [38]. The capsule was symmetrically aligned relative to the thermocouple junction and thermally insulated from the heater by an Al₂O₃ sleeve. For each run, two or three pieces of Fe-foil were sandwiched between magnesite layers to create oxygen fugacity gradients. Prior to the experiments, the capsules were dried at ~400 K for hours to remove moisture. The samples were compressed to the target pressures at room temperature and then heated to the desired temperatures at a rate of ~1 K/s. After a set duration, samples were quenched by cutting off the heater power. Recovered samples were mounted in epoxy, polished with SiC sandpaper and Al₂O₃ powder (0.3 µm), and examined after ultrasonic cleaning.

2.2. Microanalyses of the Run Products

The recovered run products from the multi-anvil experiments were analyzed using Raman spectroscopy, a scanning electron microscope (SEM) with an energy-dispersive X-ray spectrometer (EDS), and X-ray micro-diffraction (XRD). Raman spectra were obtained by a Renishaw inVia micro-Raman system equipped with a 514.5 nm Ar⁺ laser at UHM. The system achieved a spectral resolution of 2 cm⁻¹ with an accuracy of ±0.1 cm⁻¹. For spatial mapping, an automated XYZ stage enabled line scans and area mappings with sub-micron precision, using a 20× objective, a 10 µm step size, and a 30 s exposure time. Raman data were processed using Grams/AI 8.0 software (Thermo-Fisher Scientific, Inc., Waltham, MA, USA). Additional Raman measurements were conducted at Peking University (PKU) using a Raman system with a 785 nm solid-state laser.

Microstructural and chemical analyses were performed on a JEOL JXA-8500F electron microprobe at UHM, operating at 15 keV and 15 nA. Phase identification and elemental quantification were achieved with an INCA Energy 450 system integrated with an X-Max-80 Silicon Drift Detector (Oxford Instruments, Abingdon, UK), utilizing the XPP quantification procedure. Additional SEM-EDS characterization was carried out at PKU using a FEI Quanta 650 FEG SEM equipped with an Oxford INCA X-MAX50250+ system, operating at 20 kV with 10 nA. High-resolution back-scattered electron (BSE) imaging and semi-quantitative EDS analyses were further carried out at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS), using a FEI Nova NanoSEM 450 under conditions of 15 kV, 5.5 nA, and 6.5 mm working distance.

Phase identification was completed with X-ray diffraction (XRD) analysis using a Bruker D8 Venture Photon 100 CMOS microfocus diffractometer at Yanshan University. Measurements were performed in reflection geometry with Mo Kα radiation (λ = 0.71073 Å) and a beam size of 110 µm in diameter. Three representative regions were analyzed through ϕ-scan measurements with 10–30 s exposure per angle. Data were processed using the APEX3 software package (Bruker AXS Inc., Billerica, MA, USA).

3. Results

High P-T experiments investigating the MgCO₃-Fe⁰ interactions were conducted at 10–16 GPa and 1200–1700 K (Figure 1,

Table A1. Experimental conditions and phase associations of recovered products.

Run No.	Starting Materials	P (GPa)	T (K)	Duration (hr)
PL062	Mgs+3Fe	12	1200	15
PL051	Mgs+3Fe	14	1200	12
PL070	Mgs+3Fe	16	1200	8
PL037	Mgs+2Fe	12	1250	18
PL033	Mgs+2Fe	12	1300	24
PL041	Mgs+2Fe	14	1300	12
PL066	Mgs+2Fe	10	1400	24
PL039	Mgs+2Fe	12	1400	12
PL064	Mgs+3Fe	16	1400	24
PL036	Mgs+2Fe	14	1500	18
PL043	Mgs+3Fe	12	1500	5
PL045	Mgs+3Fe	12	1600	3
PL044	Mgs+3Fe	14	1600	12
PL065	Mgs+3Fe	16	1600	5 min
PL061	Mgs+2Fe	12	1700	4
PL060	Mgs+3Fe	14	1700	20

Note: “Mgs+3Fe” and “Mgs+2Fe” represent three or two pieces of Fe-foil which are loaded in the capsule, respectively. Mgs: MgCO₃; Fper: (Mg,Fe)O; Fe-C: Fe₇C₃ and/or Fe₃C; Dia: diamond; Gr: graphite.

). The run products, including diamond, Fe-C, Fper, graphite (including amorphous carbon), and carbonatitic melts, exhibited a temperature-dependent textural evolution (Figure 2 and Figure A1).

Under subsolidus conditions (1200–1500 K), reaction zones formed around the Fe-foils. These zones consist of an inner layer of Fe-C (Fe₃C or Fe₇C₃) plus Fper, and an outer layer of pure Fper (Figure 3). The intermediate Fe-C reacted incrementally with MgCO₃, facilitating diamond/graphite nucleation along the Fper grain boundaries [42]. An inward-propagating distribution of diamond/graphite is observed, with the highest concentrations at the centers of the Fe-foils (Figure 4 and Figure 5). Concurrently, the Mg# of Fper in the reaction zones varied from 0.16 (at the Fe front) to 0.40 (at the MgCO₃ front) and achieved a compositionally homogeneous (Mg_0.36Fe_0.64)O at distances beyond ~30 µm from the Fe front (Figure 6 and Figure A3). We found that prolonged heating did not lead to further morphological or compositional alterations after the Fe-foils were completely consumed. The persistence of graphite (including amorphous carbon) in the diamond stability field suggests the presence of kinetic barriers and/or redox constraints [38], considering that graphite might serve as a precursor for diamond formation [43]. Elevated temperatures and catalytic conditions can enhance the conversion from graphite to diamond (Figure A2) [19].

At elevated temperatures (1500–1700 K), the system developed dendritic textures containing Fe-C melts [44]. These melts progressively reacted with MgCO₃, producing Fper and diamond/graphite, with only trace amounts of Fe₃C remaining (Figure 2d–f). The irregular pores observed in the Fper grains (red circles in Figure 2f) may have resulted from the formation of CO₂ during the synthesis of Fper and Fe₃C [45], followed by its release upon decompression. This CO₂ degassing promotes Mg-enrichment in the Fper. Fper formed within 30 µm of the Fe fragments exhibits a broad Mg# range of 0.32–0.86, whereas Fper located farther from the Fe fragments shows a narrow Mg# range of 0.35–0.50. Despite the chemical nonequilibrium, the depletion of MgCO₃ halted further compositional evolution, resulting in Fper compositions with Mg# > 0.36, indicative of a more significant contribution from MgCO₃ than Fe⁰. The high temperature conditions enhanced carbon production, facilitating the growth of diamonds over graphite (Figure A1). Additionally, the incorporation of nitrogen led to the formation of Fe₃N (Figure 3), which likely originated from the BN capsule material and/or ambient atmosphere. This contrasts with the minimal nitrogen content in the subsolidus reaction products (Figure A2).

4. Discussion

4.1. Redox Pathways in Diamond Formation

This study elucidates the reaction pathways leading to diamond formation via redox reactions between MgCO₃ and Fe⁰. Under high P-T conditions, Fe⁰ reacts with MgCO₃, extracting its carbon to form the intermediate Fe-C phases. At subsolidus conditions, both Fe₇C₃ and Fe₃C phases are present as run products, whereas only Fe₃C persists once melting occurs. Within the reaction zones, we observe that MgCO₃ continues to react with the intermediate Fe-C phases, producing Fper and elemental carbon. Initially, carbon precipitates as graphite, which subsequently transforms into diamond. Thus, the reduction pathway of carbon can be summarized as follows:

MgCO₃ → Fe-C → C⁰(Gr) → C⁰(Dia)

(1)

The redox mechanism involves the reduction of C⁴⁺ (in MgCO₃) to C⁰ and oxidation of Fe⁰ to Fe²⁺, requiring a stoichiometric ratio of MgCO₃/Fe⁰ = 1/2. At subsolidus conditions, the MgCO₃-Fe⁰ redox reactions yield Fper, Fe-C phases, graphite, and diamond. Equilibrium constraints suggest a Fper composition of (Mg_1/3Fe_2/3)O, which is consistent with the EDS-measured composition of (Mg_0.36Fe_0.64)O in our experiments (Figure 6 and Figure A3). Consequently, the balanced redox reactions can be represented as follows, exemplified using Fe₃C as the intermediate Fe-C phase:

$$\mathrm{M}\mathrm{g}\mathrm{C}{\mathrm{O}}_3+5\mathrm{F}\mathrm{e}0\to 3\left({\mathrm{M}\mathrm{g}}_{1/3}{\mathrm{F}\mathrm{e}}_{2/3}\right)\mathrm{O}+{\mathrm{F}\mathrm{e}}_3\mathrm{C}$$

(2)

$$3\mathrm{M}\mathrm{g}\mathrm{C}{\mathrm{O}}_3+2{\mathrm{F}\mathrm{e}}_3\mathrm{C}\to 9\left({\mathrm{M}\mathrm{g}}_{1/3}{\mathrm{F}\mathrm{e}}_{2/3}\right)\mathrm{O}+5\mathrm{C}0\left(\mathrm{G}\mathrm{r}\right)$$

(3)

$$\mathrm{C}0\left(\mathrm{G}\mathrm{r}\right)\to \mathrm{C}0\left(\mathrm{D}\mathrm{i}\mathrm{a}\right)$$

(4)

In reaction (1), the molar ratio of MgCO₃ to Fe⁰ is 1:5, where 3 out of 5 moles of Fe⁰ participate in forming the intermediate Fe₃C phase. Integrating reaction (1) and (2) gives a molar ratio of MgCO₃/Fe⁰=1:2. Thus, the overall reaction is formulated as MgCO₃ + 2Fe⁰→3 (Mg_1/3Fe_2/3)O + C⁰.

In contrast, at solidus conditions (1500–1700 K), the production of CO₂ results in the generation of Mg-rich Fper (Mg# = 0.32–0.86). Using the end-composition of (Mg_9/100.9Fe_1/100.1)O as a representative to depict the reactions, we can describe the process as follows:

$$18\mathrm{M}\mathrm{g}\mathrm{C}{\mathrm{O}}_3+5\mathrm{F}\mathrm{e}0\to 20\left({\mathrm{M}\mathrm{g}}_{9/10}{\mathrm{F}\mathrm{e}}_{1/10}\right)\mathrm{O}+{\mathrm{F}\mathrm{e}}_3\mathrm{C}+17\mathrm{C}{\mathrm{O}}_2$$

(5)

$$18\mathrm{M}\mathrm{g}\mathrm{C}{\mathrm{O}}_3+5{\mathrm{F}\mathrm{e}}_3\mathrm{C}\to 20\left({\mathrm{M}\mathrm{g}}_{9/10}{\mathrm{F}\mathrm{e}}_{1/10}\right)\mathrm{O}+6[{\mathrm{C}}_{\mathrm{F}\mathrm{e}-\mathrm{C}}+{\mathrm{C}}_{\mathrm{M}\mathrm{g}\mathrm{C}\mathrm{O}3}]+17\mathrm{C}{\mathrm{O}}_2$$

(6)

$$\mathrm{C}0\left(\mathrm{G}\mathrm{r}\right)\to \mathrm{C}0\left(\mathrm{D}\mathrm{i}\mathrm{a}\right)$$

(7)

In reaction (4), the molar ratio of MgCO₃ to Fe⁰ is 18:5; however, only 1 mole of MgCO₃ (with carbon in the C⁴⁺ valence state) and 2 moles of Fe⁰ participate directly in the redox reaction. The remaining 17 moles of MgCO₃ yield 17 moles of CO₂ without undergoing any change in valence state (as evidenced in the micro-texture of the run products in Figure 2 and Figure A1). The notation C_Fe-C + C_MgCO3 represents dissolved carbon in the Fe-C and carbonate melts, respectively. Assuming that the intermediate Fe-C is fully consumed in reaction (5) and all dissolved carbon precipitates as graphite/diamond, the overall reaction can be delineated as 18MgCO₃+2Fe⁰→20(Mg_9/10Fe_1/10)O+C⁰+17CO₂.

Numerous studies have focused on the redox reactions between carbonates and metallic Fe. These reactions, as observed both in previous and present studies, typically do not achieve equilibrium. Nevertheless, the presence of homogeneous (Mg_0.36Fe_0.64)O in our subsolidus experiments suggests that the redox reaction proceeded to completion. Comparative investigations reveal that MgCO₃ reacts more readily with Fe⁰ than CaCO₃ [19,38,46], although complete reduction requires aqueous environments [47]. Elevated temperatures (>~1500 K) significantly enhance carbonate decomposition and diamond formation, accompanied by CO₂ production, independent of pressure or oxygen fugacity conditions. For instance, as shown in Figure A1, at 12 GPa under identical f_O2 conditions, the subsolidus experiments (e.g., PL033, 1300 K) still retain residual MgCO₃ after complete Fe⁰ consumption. In contrast, the solidus experiments (e.g., PL061, 1700 K) completely consume the MgCO₃, leaving behind Fe fragments. It should be noted that diamond growth ceases once either carbonate or Fe⁰ is locally exhausted.

4.2. Role of Intermediate Fe-C Phases in Diamond Genesis

Fe-C phases emerge during the reduction of carbonate by Fe⁰ across cold-to-warm subduction geotherms, acting as critical carbon reservoirs for subsequent crystallization of diamond. These intermediate phases profoundly affect the chemical characteristics of superdeep diamonds.

Firstly, superdeep diamonds exhibit lower δ¹³C values (−24 to −2‰) [5,48,49,50], which deviates significantly from the carbon isotope composition of the normal mantle (δ¹³C = ~–5‰) [51]. While eclogitic and metamorphic diamonds commonly inherit their low δ¹³C signatures (<−15‰) from subducted crustal organic carbon (−27 to −1‰), this mechanism does not adequately reconcile the low nitrogen concentrations observed in superdeep diamonds [52]. Moreover, the rare occurrences of organic carbon within superdeep diamonds suggest an alternative formation pathway involving hydrogen–rock interactions [53]. Although carbonate-derived diamonds exhibit a depletion of 1.6–2.7‰ in ¹³C at 6.3–7.5 GPa and 1573–1973 K [54], this fractionation scale is insufficient to explain the isotopic signatures of superdeep diamonds, even with pressure-enhanced fractionation [55].

Peridotitic diamonds primarily inherit their carbon from primordial mantle reservoirs, exhibiting a narrow δ¹³C range (−4 to −6‰). The minor isotopic deviations in these diamonds are indicative of localized fractionation processes during their growth, a phenomenon particularly evident in the genesis of superdeep diamonds. Here we propose that the intermediate Fe-C phases are instrumental in facilitating large-scale carbon isotopic fractionation. Theoretical calculations predict that substantial carbon isotopic fractionation (1000lnβ~10) occurs between carbonate and Fe-C at mantle conditions (e.g., ~1500 K), with ¹²C preferentially partitioning into the Fe-C phases [56]. Experimental studies further validated a fractionation of ~6.5‰ between Fe-C and their source carbonate [19]. Moreover, an additional fractionation (1000lnβ~8) between Fe-C and diamond could yield δ¹³C values as low as −12‰ relative to the initial carbonate. Furthermore, under solidus conditions, the production of CO₂ exacerbates ¹³C depletion through the preferential removal of ¹³C (1000lnβ~12) [56]. All these processes collectively result in the final diamond retaining a pronounced light carbon isotopic signature.

In addition, superdeep diamonds have markedly lower nitrogen concentrations (~33 ppm) [15] compared to lithospheric diamonds (up to 3830 ppm) [57]. Under solidus conditions, nitrogen diffuses into capsules to form Fe₃N, reflecting nitrogen’s siderophile nature [13]. The low nitrogen partition coefficient between diamond and metallic Fe (0.0005–0.013) [15] is responsible for this depletion, as natural Fe/Fe-C inclusions contain super-higher nitrogen concentrations than their diamond hosts [13,27]. Experimental data further show that diamonds crystallizing from carbonatitic melts contain 1000–1500 ppm nitrogen, in contrast to ~200 ppm from Fe-C melts [19]. This nitrogen-depletion mechanism operates exclusively at Fe-C solidus conditions, as subsolidus experiments reveal no nitrogen incorporation (Figure A1). Notably, the low nitrogen content and low ¹³C signature of superdeep diamonds likely result from a combination of factors, such as the presence of Fe⁰, distinct primordial reservoirs, and contributions from crustal organic carbon [58].

4.3. Temperature Dependence of (MG,Fe)O Compositions

Chemical analyses of recovered products reveal two distinct compositional groups of Fper (Figure 6, Figure A3 and Figure A4,

Table A2. Results of element analysis (wt%) of PL036.

	C	O	Mg	Fe	Total	Phase
spectrum1	17.20	46.20	11.74	23.56	98.70	(Mg0.35Fe0.65)O
spectrum2	11.53	50.01	12.67	24.99	99.20	(Mg0.36Fe0.64)O
spectrum3	9.52	52.95	11.99	23.21	97.67	(Mg0.36Fe0.64)O
spectrum4	17.57	49.20	11.23	21.69	99.69	(Mg0.35Fe0.65)O
spectrum5	6.66	52.88	12.13	27.44	99.11	(Mg0.32Fe0.68)O
spectrum6	8.03	51.21	13.18	25.68	98.10	(Mg0.34Fe0.66)O
spectrum7	6.65	53.40	13.17	25.68	98.90	(Mg0.34Fe0.66)O
spectrum8	74.03	17.28	3.16	4.93	99.40	C

and

Table A3. Results of element analysis (wt%) of PL045.

	C	O	Mg	Fe	Total	Phase
spectrum1	2.18	54.43	28.33	14.97	99.91	(Mg0.65Fe0.35)O
spectrum2	2.69	53.11	27.35	16.10	99.25	(Mg0.62Fe0.38)O

). At subsolidus conditions, Fper forms with lower Mg content (Mg# = 0.16–0.40). The majority of Fper grains are centered around Mg# ~ 0.36 (black dashed line in Figure 6), with local Mg# values dropping to as low as 0.16 within 20 µm of the Fe-foils. These compositions are consistent with previous reports of (Mg_0.22Fe_0.67Ca_0.11)O and (Mg_0.33Fe_0.66)O at 6 GPa under subsolidus conditions [38]. The homogeneous composition of (Mg_0.36Fe_0.64)O is independent of pressure or oxygen fugacity, being governed solely by the valence-state changes of Fe (Fe⁰→Fe²⁺) and C (C⁴⁺→C⁰).

At solidus conditions (1500–1700 K), the decomposition of MgCO₃ releases a substantial amount of Mg²⁺, yielding Mg-rich Fper. Concurrently, the chemical heterogeneity of the reactants within the sample charge leads to a wide variation of Mg# values. Comparable Mg# ranges (0.45–0.66) have been documented at 1673–1873 K and <8 GPa [38]. Under extreme conditions (77–113 GPa and 1900–2500 K), Fper compositions diverge into Mg-rich (Mg# = 0.4–0.6) and Mg-poor (Mg# < 0.15) populations [42]. However, the scarcity of Fper grains large enough for composition analysis results in sporadic reported Mg# values. Moreover, the formation of Fper with Mg# < 0.15 may be attributed to the chemical heterogeneity and/or temperature gradient within laser-heated diamond anvil cells. If the reaction were to reach equilibrium, extremely Mg-rich Fper with narrow compositional ranges [23,24] would be generated. Nevertheless, the majority of Fper inclusions in superdeep diamonds display wide Mg# variations, indicative of nonequilibrium formation processes.

5. Implications

The deep subduction of oceanic crust transports crustal carbon into Earth’s interior, where interactions with the reducing mantle give rise to superdeep diamonds. These diamonds are characterized by low δ¹³C values and low nitrogen concentrations, facilitated by intermediate Fe-C phases, and are accompanied by the formation of Fper with temperature-dependent Mg# variations. The experimentally determined Mg# range here encompasses natural compositions. The end-member (Mg_0.16Fe_0.84)O, formed under subsolidus conditions, coincides with the lowest Mg# value (0.16) reported in Brazilian diamonds [50]. The homogeneous composition (Mg_0.36Fe_0.64)O corresponds to the lower limit of most Brazilian Fper inclusions (Mg# = 0.36–0.90). In contrast, Fper from other occurrences is significantly Mg-rich and has a narrow Mg# range (Mg# = 0.75–0.90) [25]. These variations suggest that Brazilian superdeep diamonds form through slab–mantle interactions spanning the deep UM to the LM (Figure 7).

In subduction zones, redox reactions between subducted carbonates and the reduced surrounding mantle occur under both subsolidus and solidus conditions. The former produce low-Mg# Fper grains (primarily (Mg_0.36Fe_0.64)O), while the latter produce high-Mg# varieties. Low-Mg# Fper formed in the UM could be carried down to the LM and enveloped as inclusions within superdeep diamonds. This is supported by the co-occurrence of Mg-poor and Mg-rich Fper in Brazilian diamonds, along with other minerals typical of the MTZ and LM, such as ringwoodite (Mg,Fe)₂SiO₄, perovskite CaSiO₃, or bridgmanite (Mg,Fe)SiO₃ [25]. Furthermore, during subduction, the Mg# of Fper would change if additional carbonates and/or Fe are involved in the redox reactions. Thomson et al. [31] proposed that the evolution of Fper composition records the progressive reaction between carbonatite melts and the ambient mantle, resulting in Fper with a broad range in Mg#.

In summary, factors such as reaction nonequilibrium, chemical inhomogeneity, and the downward migration of low-Mg# Fper, etc., can lead to the formation of Fper with a range of Mg# values. These Fper grains can become encapsulated in superdeep diamonds. Notably, Fper grains that form at the deep UM and are subsequently enveloped and transported back to the surface would exhibit only low Mg# characters, as observed in Brazilian superdeep diamonds. In contrast, diamonds from other localities likely form under solidus conditions in the topmost LM (~20 km depth) [59], exclusively trapping high-Mg# Fper. The presence of Mg#-rich Fper indicates either a more depleted or hotter source. In all cases, residual carbonates, intermediate Fe-C phases, and graphite are incorporated during diamond growth, as evidenced by natural assemblages of Mg-poor Fper + carbonate, carbonate + Fe-C, and Fe-C + metallic Fe + graphite in superdeep diamonds [13].