Experimental Modeling of the Mantle–Slab Interaction in the Metal–Carbonate System, Conditions of Crystallization and Indicator Characteristics of Diamond

Subduction processes determine the mineralogy, petrology, and geochemistry of the deep zones of the Earth to a large extent and significantly affect the global cycles of carbon and other elements. In modern theoretical and experimental works revealing the regularities in the distribution of ƒ_O2 values in mantle rocks (Ballhaus and Frost, 1994; Frost et al., 2004; Rohrbach et al., 2007; Rohrbach and Schmidt, 2011; Shirey et al., 2013, 2019), it is shown that at depths of ≥250–300 km, where a pressure of 7.5 GPa is reached, the mantle begins to be reduced, with ƒ_O2 values 5 logarithmic units below the level of the fayalite–magnetite–quartz (FMQ) buffer. Under these conditions metallic iron, Fe,Ni alloys, carbides, and iron–carbon melts become stable in mantle rocks. According to experimental data, at depths of more than 250 km, the concentration of metallic iron in mantle rocks can reach 1400 ppm (Rohrbach et al., 2007; Rohrbach and Schmidt, 2011). Subducted carbonate-bearing rocks, on the contrary, are oxidized, and at mantle depths, the oxygen fugacity in fluids and melts formed in the slab can reach very high values, up to FMQ + 5 log units (Scambelluri and Philippot, 2001; Frezzotti et al., 2011; Walters et al., 2020; Zhang et al., 2021; Ague et al., 2022; Girnis et al., 2022). Studies of the interactions between reduced mantle rocks and oxidized slab, as key rock-forming processes that can be accompanied by diamond crystallization in the presence of carbon (Rohrbach and Schmidt, 2011), are of considerable interest for mantle petrology. The possibility of diamond formation as a result of these interactions is determined by the existence of deep zones in the Earth containing Fe,Ni metal and by the possibility of subduction of carbonate minerals to these depths. In recent decades, more and more information has appeared based on a comprehensive study of natural diamonds and inclusions therein, confirming the genetic relationship of some diamonds with subducted carbonates. It is interesting to note that the “subduction isotopic signature” is also characteristic of many diamonds with inclusions of Fe,Ni metal or carbides (Bulanova et al., 2010; Smith et al., 2018; Shatsky et al., 2020).

We have previously studied a relatively simple model of the mantle–crust interaction using the Mg,Ca carbonate–iron system as an example (Palyanov et al., 2013) and developed a technique that made it possible to create a significant oxygen fugacity gradient in the ampoule as well as to provide conditions for the occurrence and propagation of the redox front. As a result of these studies, the redox mechanism of diamond formation was experimentally studied both under reducing (metal melt) and oxidizing (carbonate melt) conditions from a single carbon source – initial carbonate. In further experimental studies on the carbonate–metal interaction, magnesite, calcite, dolomite, Na₂CO₃, and K₂CO₃ were used as carbonates, and native iron was used as the metal (Bataleva et al., 2015; Martirosyan et al., 2015, 2016; Shatskiy et al., 2023). In all high-pressure, high-temperature studies using iron as a reducing agent for carbonate, the preferential crystallization of iron carbide, rather than diamond, was established. An analysis of the available experimental and theoretical data (Rohrbach et al., 2014) shows that the use of an Fe,Ni alloy instead of native iron as a reducing agent for carbonate can completely eliminate the problem of carbide formation and create conditions for the predominant crystallization of elemental carbon phases: diamond and graphite. The nickel impurity in iron carbides reduces the temperature of their stability significantly, by hundreds of degrees. In particular, at pressures of 5–6 GPa, the temperatures of the Fe–Ni–C ternary eutectic are, according to various estimates, from 1077 to 1227 °C, and the temperature of the Fe–C double eutectic is between 1250 and 1345 °C (Strong and Chrenko, 1971; Turkevich and Kulik, 1995; Kocherzhinski and Kulik, 1996; Solozhenko et al., 2002). It should also be emphasized that the iron–nickel alloy was found in inclusions (including central ones) in diamonds and minerals of mantle diamond-bearing rocks, in assemblage with wüstite, graphite, and ±sulfides (Bulanova, 1995; Davies et al., 1999; Jacob et al., 2004; Bulanova et al., 2010; Smith et al., 2016, 2018).

Considering these data, we present the results of experiments on the (Fe,Ni)–(Mg,Ca)CO₃ interaction, aimed primarily at elucidating the conditions for the formation of elemental carbon phases (diamond and graphite) and identifying their indicator characteristics as well as patterns of phase formation during metal–carbonate interaction in a wide temperature range from 800 to 1550 °C, including the conditions of both “cold” and “hot” subduction (Syracuse et al., 2010; Perchuk et al., 2019).

Starting materials. Natural magnesite and dolomite (Satka deposit, Chelyabinsk region) with purity >99.5%, as well as powders of chemically pure Fe⁰ and Ni⁰ (99.999%), were used as initial reagents. The proportions of the initial magnesite and dolomite were 4:1 (the bulk composition of carbonate is Mg_0.9Ca_0.1CO₃), and the proportions of iron and nickel were 9:1 (Fe₉₀Ni₁₀). Information on the weights of the starting materials is given in Table 1.

High-pressure experiments. The experiments were carried out on the multianvil high-pressure apparatus of a split-sphere type (BARS) (Palyanov et al., 2017), using a high-pressure cell in the form of a tetragonal prism 21.1 × 21.1 × 25.4 mm in size. The cell configuration and heater geometry (length 18.8 mm, diameter 12 mm) ensured the presence of a low-gradient zone (Pal’yanov et al., 2002) in the central part of the heater, where the ampoule with the starting materials was located. Details on pressure and temperature calibrations in the high-pressure cell are given in (Pal’yanov et al., 2002; Sokol et al., 2015).

Experimental studies were carried out using an approach developed to model the processes of diamond formation during the interaction of oxidized carbonate-bearing subducted material with reduced mantle rocks in the alkalineearth carbonate–iron system (Palyanov et al., 2013). In the series of experiments carried out in this work, we used an alloy of iron and nickel, since Ni has an inhibitory effect on the crystallization of carbide and, thus, the potential P–T field of diamond crystallization expands as a result of redox interaction. A series of experiments was carried out in the system (Fe,Ni)–(Mg,Ca)CO₃, at a pressure of 6.3 GPa, in the temperature range of 800–1550 °С, and durations from 35 to 105 h (Table 1). During assembly, a container of carbonates (bulk Mg_0.9Ca_0.1CO₃) was placed in a platinum ampoule (Ø 10 mm, height 8 mm), in the center of which there was a cylinder pressed from powders of Fe and Ni (Fe₉₀Ni₁₀). The “sandwich” layout of the reagents allows creating a gradient of oxygen fugacity in the samples, excludes the interaction of metallic iron with Pt, and provides complete airtightness, thereby creating the possibility of experimental study of the mechanism of formation of elemental carbon (graphite and diamond) due to the redox interaction of Fe,Ni metal and carbonate. The use of large-volume Pt ampoules and a long duration of experiments (up to 105 h) has a number of advantages in experiments with an oxygen fugacity gradient and the propagation of a redox front inside the ampoule; in particular, it makes it possible to carry out a comprehensive study of each section of the obtained samples using a set of modern analytical methods.

Analytical research. Analytical studies were carried out at Institute of Geology and Mineralogy SB RAS (Novosibirsk), and the Analytical Center for Multi-elemental and Isotope Research SB RAS. After the completion of the experiments, the samples were sawn and initially examined on Stemi 508 and Axio Imager Z2m (Carl Zeiss Microscopy) optical stereomicroscopes. The phase and chemical compositions of the samples, as well as phase relationships, were studied using scanning electron microscopy, energy-dispersive spectroscopy (Tescan MIRA3 LMU), and microprobe analysis (JEOL JXA-8100). For analytical studies with the use of microprobe analysis and energy-dispersive spectroscopy, the samples were permeated with epoxy resin and polished. When analyzed by scanning electron microscopy and energy-dispersive spectroscopy, the samples were covered with carbon or chromium, depending on the problem being solved. For microprobe analysis, polished samples were covered with carbon. Carbonate, oxide, and metal phases were analyzed at an accelerating voltage of 15 to 20 kV, a probe current of 20 nA, a counting time of 10–20 s on each analytical line, and an electron beam probe diameter of 3–4 μm. To study the composition of quenched melts represented by an aggregate of microdendrites, the size of the electron beam was increased to 20–50 μm. In the case of covering of samples with carbon, the measurement of the CO₂ content in carbonates/carbonate melts was carried out based on the deficit of the sums. With chromium covering, CO₂ concentrations in carbonates were measured directly. All the platinum ampoules after the experiments were analyzed by energy-dispersive spectroscopy for the diffusion of iron from the central part of the sample. In most cases, the iron concentrations in the Pt ampoule were below the detection limit, and at the highest temperatures they did not exceed 0.2 wt.%. The morphology of diamond crystals was studied using a Tescan MIRA3 LMU SEM and an Axio Imager Z2m optical microscope, including the differential interference contrast (DIC) method. The study of the structural features of graphite was carried out by Raman spectroscopy. The Raman spectra were recorded using a Horiba J.Y. LabRAM HR800 spectrometer equipped with an Olympus BX41 microscope. The excitation source was a diodepumped solid-state laser emitting at a wavelength of 532 nm (Torus, Laser Quantum). An Olympus 50X lens (NA = 0.50) was used to focus the laser beam on the sample and collect the Raman signal. The power density of the laser beam on the samples was approximately 5 × 10⁴ W/cm². The spectra were measured with a spectral resolution of 2 cm^–1. The spectrometer was calibrated against the emission lines of 540.06 and 585.25 nm of a neon discharge lamp. The Raman spectra were measured with an accumulation time of 10 s (per one spectral window) and averaged over seven measurements. The positioning accuracy of the bands in the Raman spectra was approximately ±1 cm^–1. The defect–impurity composition of diamond crystals obtained in experiments on the redox interaction (Fe,Ni)– (Mg,Ca)CO₃ was studied by optical spectroscopy, including infrared (IR) absorption and photoluminescence (PL). Infrared absorption spectra were measured using a Bruker Vertex 70 Fourier transform IR spectrophotometer equipped with a Hyperion 2000 IR microscope. Photoluminescence spectra were measured using a laboratory stand based on a Horiba J.Y. iHR320 monochromator with a Syncerity CCD detector. Lasers with radiation wavelengths of 395 and 532 nm were used as the excitation source.

Structure of the final samples and chemical composition of the phases. The results and parameters of the experiments, along with the chemical compositions of the final phases, are presented in Tables 1–3. Data on the isotopic composition of carbon in the obtained elemental carbon and carbon-bearing phases, as well as the features of carbon fractionation in processes occurring in the (Fe,Ni)–(Mg,Ca) CO₃ system, are considered in detail in (Reutsky et al., 2023). It has been established that the characteristic features of the structure of all the obtained samples are the general patterns of alternation of reaction zones as well as trends in the change in phase and chemical compositions from the reduced center to the oxidized periphery. The structure of relatively low-temperature samples (from 800 to 1000 °C) is shown in Fig. 1a. In this temperature range, the following alternation of reaction zones was established (hereinafter, from center to periphery) (Fig. 1c): metal → metal + wüstite/magnesiowüstite → magnesiowüstite + graphite ± Mg,Fe,Ca carbonates → magnesite + aragonite. In the reduced part of the samples, where the Fe₉₀Ni₁₀ metal was initially placed, a Fe,Ni metal phase of variable composition is formed (the profiles are shown in Fig. 2a, b).

In the central part of the metal cylinder, the Ni concentrations are the lowest (0–12 wt.%), and, as they approach the initial contact with carbonate, they increase and reach 65–70 wt.% directly at the contact, where Ni-wüstite (Fe_0.48–0.83 Ni_0.17–0.52O, 1000 °C, 35 h) or magnesiowüstite (Fe_0.83–0.87 Mg_0.13–0.16O, 800–900 °C, 70–105 h) crystals are located in the metal.

Thin reaction zones that form around the metal in the range of 800–1000 °C consist of finely crystalline aggregates of magnesiowüstite and graphite (Fig. 1c). These zones are characterized by an increase in thickness with an increase in temperature (from 70 to 100 μm, Table 4), as well as a regular change in the composition of magnesiowüstite, namely, decreased MgO at the contact with the metal (Mg# = 0.2) and its gradual increase (Mg# = 0.38–0.47) toward the contact with the carbonate part of the samples (Table 2, profiles are shown in Fig. 3a–c). Carbonates in the peripheral (oxidized) zone of the samples in this temperature range are mainly represented by magnesite and aragonite; in rare cases, a reaction zone of dolomite + aragonite is distinguished (Fig. 1c). For all the carbonates, small admixtures of FeO (0.2–1.7 wt.%) are noted, and for magnesite, an admixture of CaO from 0.5 to 2.8 wt.% is noted (Table 2).

At 1200 °С, the following alternation of reaction zones in the sample was established: metal (quenched Fe,Ni–C melt) → metal (quenched Fe,Ni–C melt) + magnesiowüstite → magnesiowüstite + graphite → magnesiowüstite + graphite + aragonite → aragonite → magnesite + dolomite + aragonite. As at lower temperatures, in the central part of the metal cylinder, the Ni concentrations are low (3–7 wt.%, relative to the initial 10 wt.%), and, as they approach the initial contact with carbonate, they increase sharply and reach 72 wt.% directly at the contact, where the quenched Fe,Ni–C melt coexists with wüstite (Fe_0.91–0.95Mg_0.04–0.08 Ni_0.01)O (Figs. 2c, 3d; Table 2). A wide reaction zone (700 μm, Table 4) formed around the metal cylinder is represented by a polycrystalline aggregate of magnesiowüstite and graphite (Fig. 1b, d, e), and in the direction from the reduced to the oxidized part of the sample, it is characterized by (a) an increase in the grain sizes of both phases, (b) a change in the morphology of graphite from submicron plates to spherical formations 15 μm in size, and (c) an increase in the MgO concentration in magnesiowüstite from 12 to 49 wt.% (Fig. 3d). In the more oxidized zone of the sample, in association with graphite and aragonite, the Mg# of magnesiowüstite/ferropericlase follows the same trend, and MgO concentrations increase from 50 to 67 wt.%. Carbonates formed in the peripheral part of the sample (Fig. 1f) are represented by magnesite (~4 wt.% CaO, 1.1 wt.% FeO), dolomite (1.3 wt.% FeO), and aragonite (0.5 wt.% FeO) (Table 2).

At a higher temperature of 1330 °С, the reaction zones from the reduced to the oxidized part of the sample are represented by the following associations: metal (quenched Fe,Ni–C melt) → magnesiowüstite + metal (quenched Fe,Ni–C melt) + graphite → magnesiowüstite + graphite → magnesiowüstite/ferropericlase + graphite + carbonate melt → Mg,Ca,Fe carbonates (Table 1). At the same time, the central part of the metal cylinder is represented by a quenched melt containing 1–3 wt.% Ni, and near the initial contact with carbonate, where magnesiowüstite (Fe_0.91–0.96 Mg_0.04–0.08Ni_0.01)O and graphite coexist with the melt (Fig. 4a, b), the Ni concentration in the melt increases to 61 wt.%. The reaction zone of magnesiowüstite and graphite surrounding the metal (Fig. 4c) is very wide (1.0–1.3 mm), and the magnesiowüstite composition in it varies from (Fe_0.85Mg_0.13Ni_0.01)O (in the reduced zone, in association with graphite) to (Fe_0.55Mg_0.43Ca_0.02)O (in the oxidized zone, at the contact with the carbonate melt) (Fig. 3e). The quenched carbonate melt coexisting with magnesiowüstite is calcium-enriched (Table 3). The peripheral part of the sample is represented by dolomite, aragonite, and magnesioaragonite (Table 3).

The structure of the samples obtained in higher temperature experiments is shown in Fig. 5. At 1400 °C, the following alternation of reaction zones was established: Fe,Ni metal melt + diamond → Fe,Ni metal melt + magnesiowüstite + diamond + graphite → magnesiowüstite + graphite (Fig. 4d) → calcium carbonate melt + magnesiowüstite + graphite (Fig. 4e, f) → aragonite + magnesiowüstite + graphite. Iron–nickel metal in association with diamond, graphite, and magnesiowüstite (Fe_0.72–0.76Mg_0.23–0.26Ni_0.01)O contains 45 wt.% Ni. As in lower temperature experiments, in the magnesiowüstite + graphite reaction zone, the composition of magnesiowüstite is variable, from (Fe_0.72Mg_0.27Ni_0.01) O (reduced part of the sample) to (Fe_0.52Mg_0.46Сa_0.02)O (oxidized part) (Fig. 3f). In addition, in this zone, the crystal size increases from 10–15 μm (reduced part) to 200 μm (oxidized part, contact with carbonate melt) (Fig. 4d). The quenched, predominantly calcium carbonate melt contains crystals of magnesiowüstite (Fe_0.51Mg_0.47Ca_0.02)O and graphite (Fig. 4e, f; Table 3). Magnesiowüstite from the peripheral reaction zone (aragonite + magnesiowüstite + graphite) varies in composition from (Fe_0.48Mg_0.51Ca_0.02)O (at contact with the carbonate melt) to (Mg_0.82Fe_0.17Ca_0.01)O (near the platinum ampoule) (Fig. 3f).

Experiments carried out at 1470 and 1550 °C demonstrate similar results. In the final samples, reaction zones alternate from center to periphery (Fig. 5): Fe,Ni metal melt + diamond → Fe,Ni metal melt + magnesiowüstite + diamond → magnesiowüstite + graphite → calcium carbonate melt + magnesiowüstite + diamond + graphite → aragonite + magnesiowüstite/ferropericlase + graphite. In the reduced part of the samples, spontaneous formation of large diamond crystals (1.0–1.7 mm) occurs in association with the Fe₅₆Ni₄₄ metal phase (Fig. 7a) as well as the metal + magnesiowüstite aggregate (Figs. 6a, 7b, c).

In the thick reaction zone composed of magnesiowüstite + graphite, magnesiowüstite changes its composition from (Fe_0.73Mg_0.24Ca_0.01Ni_0.01)O to (Fe_0.29Mg_0.69Ca_0.02)O (Table 3). The carbonate melt formed in the oxidized part of the samples is characterized by a constant composition corresponding to almost pure CaCO₃ with FeO and MgO impurities, the concentration of which does not exceed 2 wt.%. Spontaneous crystallization of diamond occurs at the contact of the carbonate melt and the polycrystalline magnesiowüstite + graphite aggregate as well as directly in the carbonate melt in the oxidized part of the samples. Under these conditions, cuboctahedral and octahedral diamond crystals (Figs. 6b, 7d, e), large cuboctahedral crystals of magnesiowüstite (Fe_0.31Mg_0.66Ca_0.02)O, and plates of metastable graphite (Figs. 6c, 7f) form in the carbonate melt.

Results of the study of graphite by Raman spectroscopy. The graphite obtained during the experiments was studied using Raman spectroscopy—an effective method for characterizing graphite materials (Pimenta et al., 2007). The fundamental characteristics of the Raman spectrum of all graphite materials are the first-order Raman peak with a frequency of ~1580 cm^–1 (the so-called G band), corresponding to E_2g phonons in the center of the Brillouin zone, and the second-order Raman band (Gʹ) in the range of 2500– 2800 cm^–1. In addition, the Raman spectra of graphite samples usually contain the D band at ~1350 cm^–1 and the Dʹ band at ~1620 cm^–1. Both these modes are forbidden in highly crystalline graphite and only become active in the presence of structural disorder (Tuinstra and Koenig, 1970; Reich and Thomsen, 2004). The appearance of the D and Dʹ bands is associated with the process of double resonant (DR) Raman scattering, which includes acts of elastic scattering on crystal defects (Reich and Thomsen, 2004; Ferrari, 2007). The Gʹ band is an overtone of the D band and is allowed by the wavevector selection rules. The ratio of the integrated intensities of the D and G bands (I_D/I_G) depends on the defectiveness of graphite materials and is widely used to characterize the degree of their structural perfection (Tuinstra and Koenig, 1970; Mernagh et al., 1984; Sadezky et al., 2005; Pimenta et al., 2007; Pawlyta et al., 2015).

Figure 8 shows representative Raman spectra of graphite obtained during the experiments. The graphite samples taken from the experiments carried out at relatively low temperatures (800 and 900 °C) have intense D (1350 cm^–1), G (1580 cm^–1), and Dʹ (1620 cm^–1) bands in their spectra, which indicates a low structural quality and/or very small sizes of graphite crystallites. With an increase in the crystallization temperature, the intensity of the bands due to disorder decreases compared to the G band, and all the bands become narrower. Raman spectra measured in different regions of the same graphite sample can show somewhat different relative intensities of the D, G, and Dʹ bands, indicating that the samples are composed of graphite crystallites with varying degrees of structural disorder. These variations, however, do not have a significant effect on the general trend of spectral changes with the crystallization temperature.

To quantify the degree of structural perfection of graphite synthesized at different temperatures, the integrated intensities of the D and G bands were determined for each Raman spectrum, and the I_D/I_G ratios were calculated. To do this, the spectra were decomposed in the range of 1000–1750 cm^–1 into D, G, and Dʹ components. It should be noted that for the graphite samples obtained at 800 and 900 °C, bands with maxima at ~1200 cm^–1 and ~1500 cm^–1 were used as additional components for correct fitting. Both these bands are usually observed in the Raman spectra of low-crystalline carbonaceous materials; however, their origin is still debatable (Sadezky et al., 2005; Pawlyta et al., 2015). The dependence of the I_D/I_G ratio on the crystallization temperature of graphite is shown in Fig. 9.

The I_D/I_G values calculated for graphite samples obtained in the same experiment can differ within fairly wide limits, which indicates the structural inhomogeneity of graphite crystallites. Nevertheless, a clear pattern is traced, which consists in a decrease in the value of the I_D/I_G ratio with an increase in the crystallization temperature. Thus, we can conclude that the degree of ordering of graphite formed during the metal–carbonate interaction increases with increasing temperature.

Diamond crystallization. Comprehensive studies of diamond crystals synthesized by the redox interaction (Fe,Ni)– (Mg,Ca)CO₃ have been carried out. Crystals were obtained in experiments at temperatures of 1400, 1470, and 1550 °C, both in the central parts of the samples (Fe,Ni melt, reduced conditions) and in the peripheral parts (carbonate melt, oxidized conditions) owing to a single carbon source—the initial carbonate.

The morphology of diamonds from the metal melt is determined by freely growing {111} faces and a complex surface formed as a result of the joint growth of magnesiowüstite and diamond (Figs. 10, 11a, b). We established that diamonds formed from the metal–carbon melt in the temperature range of 1400–1550 °C are characterized by an octahedral growth pattern. Inclusions of taenite and magnesiowüstite were found in these crystals. For diamonds formed in the metal–carbon melt, the average concentration of nitrogen impurity in crystals naturally increases with temperature and is <50 ppm at 1400 °C, 100–150 ppm at 1470 °C, and 350–400 ppm at 1550 °C. The main form of nitrogen impurity is single substituting atoms (C centers). In the central areas of the crystals obtained at 1550 °C, nitrogen pairs (A centers) are additionally noted; the degree of aggregation is 20–25%. The main characteristics of the photoluminescence of diamonds from the metal melt are determined. Depending on the crystallization temperature, the PL spectra are dominated by luminescence bands due to either nitrogen–vacancy and nickel (884 nm) centers at 1400 °C or nickel–nitrogen centers (S3, 598 nm, 727 nm, 746 nm, etc.) at 1550 °C (Fig. 12).

Diamond crystals from the carbonate melt are 20–40 μm in size; they have mirror-smooth {100} and {111} faces, with no micromorphologic features revealed using the DIC method. To obtain information about the features of the microrelief of diamond faces under these conditions, an additional experiment was carried out at a temperature of 1400 °C using cuboctahedral seed crystals 0.5 mm in size.

We found that the growth of diamond in the carbonate melt is due to gentle vicinal hillocks on the {111} and {100} faces (Fig. 11e, f). The morphology of spontaneous diamond crystals from the carbonate melt is determined by the faces of a cube and an octahedron (Fig. 7e, f). The shape of the crystals changes from a cuboctahedron to an octahedron with a decrease in temperature from 1550 to 1400 °C. Microinclusions of carbonates were detected in such diamonds by FTIR spectroscopy. The micromorphology of the {100} and {111} faces with specific elements of vicinal growth is shown in Fig. 11c–f.

The reconstruction of the phase formation processes in the (Fe,Ni)–(Mg,Ca)CO₃ system showed that at relatively low temperatures (800–1000 °C) the interaction of the initial Fe,Ni metal and carbonate occurs by a redox reaction (given schematically):

This reaction includes a number of inextricably linked processes: the oxidation of iron to wüstite, the reduction of carbonate carbon to C⁰ (graphite), the redistribution of MgO from the initial Mg,Ca carbonate to wüstite with the formation of magnesiowüstite and calcium carbonate, and the redistribution of iron and nickel in the initial metal cylinder, with a trend of depletion of its central part with nickel and a sharp enrichment of the peripheral zone with Ni, owing to the selective oxidation of Fe during the formation of magnesiowüstite. In addition, a part of the reduced carbon is dissolved in the Fe,Ni metal. As a result of reaction (1), a reaction zone of magnesiowüstite and graphite is formed in the reduced part of the samples, around the metal cylinder, and in the oxidized part, a polycrystalline aggregate of newly formed aragonite and recrystallized magnesite (±dolomite) is formed. The fact that the carbonates undergo recrystallization in the course of the experiments is evidenced by both the structure of polycrystalline aggregates in the peripheral part of the samples and the change in the chemical compositions of magnesite and dolomite relative to the initial ones (FeO and CaO impurities).

Our experiments demonstrate that the formation of the only elemental carbon phase (graphite) is localized in the reaction zone containing the magnesiowüstite + graphite association and occurs from a single carbon source—carbonate. The width of this reaction zone increases from 70–80 (800 °C) to 350–450 μm (1000 °C) with increasing temperature (Table 4). At the same time, the composition of magnesiowüstite exhibits a dependence on phase association, which consists in a decreased MgO content at the contact with the metal (in the reduced part of the sample) and its gradual increase as it approaches the carbonate (oxidized) part of the sample.

At higher temperatures, the redox interaction of Fe,Ni metal and carbonate is accompanied by partial melting processes, with the formation of a Fe–Ni–C metal–carbon melt in the reduced part of the samples (at T ≥ 1200 °C) and a high-calcium carbonate melt in the oxidized part (at T ≥ 1330 °С). The formation of carbon phases (metastable graphite and diamond (T ≥ 1400 °C)) under these conditions occurs during the redox interaction of two melts, according to the schematic reaction:

We established that during this interaction, the spontaneous diamond formation both under reducing and oxidizing conditions occurs according to the mechanism that we previously demonstrated in the carbonate–iron system (Palyanov et al., 2013). According to this mechanism, diamond nucleation under reducing conditions takes place in the case of contact of graphite with a metal–carbon melt, and further crystallization of diamond is a result of supersaturation of the melt with carbon with gradual consumption of Fe⁰ due to its oxidation and crystallization of wüstite (the so-called “solvent consumption” mechanism). It should be emphasized that in the case of an iron–carbon melt, the gradual saturation of the metal with carbon inevitably leads to the formation of carbide (cohenite), upon further interaction of which with carbonate, an association of metastable graphite and magnesiowüstite is formed, and the crystallization of diamond stops in the absence of a solvent (metal melt).

In the present study, under reducing conditions, the Fe– Ni–C melt is present in the temperature range of 1200– 1550 °C, while carbide is not formed. During the evolution of the Fe–Ni–C melt composition, the iron concentration in it decreases from 90% (in the initial metal) to 55–56% (1400–1550 °C), as a result of the oxidation of Fe⁰ to wüstite/magnesiowüstite. At the same time, almost all Ni is concentrated exclusively in the melt, and only small amounts of nickel are consumed in redox reactions and enter the magnesiowüstite (up to 2.1 wt.%), which crystallizes under reducing conditions. Thus, owing to the natural decrease in the amount of metal–carbon melt in the samples during the experiments as a result of iron oxidation, the nickel concentration in the melt sharply increases, and the Fe–Ni–C melt remains until the end of the experiments, enabling stable diamond crystallization under reducing conditions.

Diamond crystallization according to the redox mechanism established earlier (Palyanov et al., 2013) occurred in the temperature range of 1400–1550 °C in both the metal–carbon and carbonate melts. However, it should be noted that the boundary conditions for the formation of diamond did not coincide with the temperature of the metal–carbon eutectic. Thus, melting in the reduced part of the samples was recorded at 1200 and 1330 °C, while crystallization of diamond was recorded only at 1400 °C. This phenomenon requires discussion, since it can also take place in the processes of natural diamond formation associated with metallic media. Such natural diamonds contain inclusions of iron, carbides, wüstite, and magnesiowüstite (Sharp, 1966; Sobolev et al., 1981; Meyer and McCallum, 1986; Bulanova, 1995; Stachel et al., 1998; Jacob et al., 2004; Jones et al., 2008; Kaminsky and Wirth, 2011; Smith and Kopylova, 2014; Smith et al., 2016, 2018; Shatsky et al., 2020). As it is shown in a number of experimental works, the synthesis of diamond in metal–carbon media is an impurity-induced process. For example, the addition of a nitrogen impurity at a concentration above 0.4 at.% (Palyanov et al., 2010) or water above 0.43 wt.% (Palyanov et al., 2012) completely blocks diamond crystallization in the Fe–Ni–C system, which leads to the formation of metastable graphite instead of diamond. In the series of experiments we performed, the role of an impurity that inhibits diamond crystallization can be played by oxygen. As shown in (Palyanov et al., 2020), an increase in the oxygen content in the Fe–Ni–O–C system also leads to the appearance of metastable graphite, although the inhibitory effect of oxygen is less significant compared to that of water and nitrogen. Undoubtedly, inhibitory impurities can significantly affect the crystallization of diamond in natural processes. However, taking into account the general trend toward an increase in the diamond-forming capacity of various media with an increase in temperature and pressure (Luth et al., 2022), as well as the possibility of the appearance of elemental Fe in the mantle at a sufficiently high pressure of 7.5 GPa and higher, the influence of impurities will not be so significant for natural diamonds crystallized from a solution of carbon in metal melts.

During the crystallization of diamonds from the carbonate melt, their growth takes place owing to the reduction of carbonate by the metal–carbon melt. Near the interface between the Fe,Ni–C melt and the carbonate melt, nucleation of new diamonds occurs and continues until the end of the experiments. In the presence of an insignificant temperature gradient in the ampoule, the growth of diamond occurs, along with other causes, owing to the transport of carbon dissolved in the carbonate melt in parallel with the process of crystallization of magnesiowüstite. Thus, during the formation of diamond as a result of the (Fe,Ni)–Mg,Ca carbonate interaction under oxidizing conditions, the carbonate melt is both a crystallization medium and a source of diamond carbon. It should also be emphasized that, in our previous studies in the iron–carbonate system (Palyanov et al., 2013), the diamond-forming redox reactions in the carbonate melt did not continue until the end of the experiment but stopped when the reducing agent (Fe₃C) was completely consumed, while in the reduced part the samples formed a significant amount of carbon in the form of metastable graphite.

When discussing the spontaneous diamond formation and crystallization of graphite as a result of redox interaction, one should also consider the issue of the appearance of a redox front in the course of experiments. Based on the obtained results, it was established that owing to the oxygen fugacity gradient in the ampoule (Δƒ_O2 about 4 log units (Palyanov et al., 2013)) at the beginning of the interaction, at the Fe,Ni metal–carbonate interface, a redox front is formed and begins to propagate. The interaction between the reduced center and the oxidized periphery is due to a fluid (T < 1330 °C) and/or high-calcium carbonate melt (T ≥ 1330 °C). At the same time, the speed of the redox front propagation depends on the temperature and increases in the studied T interval by almost 100 times, from 1.3 (800 °C) to ~118 μm/h (1550 °C) (Table 1). On the graph reflecting this dependence, an inflection is noted (Fig. 13), apparently, due to the appearance of a melt at ~1200 °C. For comparison, in the nickel-free system (Palyanov et al., 2013), the redox front propagates at velocities 2–5 times lower, and the greatest difference from the results of this study was noted at high temperatures (>1300 °C). We believe that this phenomenon is also associated with carbide formation.

The results obtained in this work are of interest in terms of reconstructing the conditions and mechanisms for the formation of carbon phases (graphite and diamond) during redox interaction. As a result of reactions between Fe,Ni metal and Mg,Ca carbonate, the formation of metastable graphite was established in the entire studied temperature range from 800 to 1550 °C. At a pressure of 6.3 GPa, this interval is larger than the temperature range between the “cold” and “hot” subduction zones (Syracuse et al., 2010), which determined the relevance of quantitative assessment of the degree of structural perfection of graphite by Raman spectroscopy. Our results show that with an increase in the crystallization temperature, the values of the I_D/I_G ratio, which characterize the degree of structural perfection of graphite, naturally decrease. Given that natural graphite from rocks of different geneses exhibits significant variations in structural perfection (Reich and Thomsen, 2004), our data on specific I_D/I_G values in graphite spectra obtained with known P–T–t parameters can be used as a relative geothermometer of some natural graphite-producing processes.

The specificity of diamond nucleation and growth during the (Fe,Ni)–(Mg,Ca)CO₃ interaction, which consists in the possibility of forming diamonds with contrasting characteristics (morphology, composition of microimpurities, spectroscopic properties, and composition of inclusions) in a single process, made it possible to identify potential indicators of metal–carbonate interaction. In particular, the inclusions of Fe,Ni metal melt in synthesized diamonds are similar to the metallic inclusions found in natural diamonds and other minerals of diamond-bearing mantle xenoliths (Bulanova, 1995; Davies et al., 1999; Jacob et al., 2004; Bulanova et al., 2010; Smith et al., 2016, 2018). It is noteworthy that, in addition to metal inclusions, there is other evidence of the participation of Fe,Ni metal–carbon melt in diamond crystallization processes—the incorporation of Ni impurity into diamonds established in this study. Defect–impurity centers associated with the incorporation of nickel into the diamond crystal lattice are characteristic of synthetic diamonds grown using nickel-containing catalyst solvents (Collins, 2000; Yelisseyev and Kanda, 2007). Similar impurities in the form of nickel–nitrogen complexes have also been found in natural diamonds from various deposits (Zaitsev, 2001; Lang et al., 2004, 2007; Yelisseyev et al., 2004; Skuzovatov et al., 2015). The possibility of Ni incorporation into the structure of diamond from sulfide melts and the corresponding formation of specific optical centers associated with Ni was confirmed by us earlier (Palyanov et al., 2006; Bataleva et al., 2016). The data obtained in this study shows that Ni-containing defect– impurity centers in diamond (S3, 598 nm, 727 nm, 746 nm, 884 nm, etc.) can also form during diamond crystallization as a result of carbon reduction of Fe, Ni metal melt.

The experimentally established fact of spontaneous formation of diamond from a calcium carbonate melt (practically pure CaCO₃), as well as the results on the distribution of Mg, Ca, and Fe between the obtained phases, suggests that inclusions of high-calcium carbonates might be indicators of the considered redox mechanism of diamond formation (Brenker et al., 2007), especially in association with magnesiowüstite (±graphite). Given that during the interaction of Fe,Ni metal and alkaline-earth carbonate, magnesiowüstite forms both under reducing and oxidizing conditions, and its composition changes significantly with the advance of the redox front, then significant variations in the Mg/Fe ratios, as well as impurities of Ca or Ni in magnesiowüstite from inclusions in natural diamonds (Svicero, 1995; Harte et al., 1999; McCammon, 2001; McCammon et al., 2004; Hayman et al., 2005; Bulanova et al., 2010; Harte, 2010; Kaminsky, 2012; Wirth et al., 2014; Zedgenizov et al., 2014), can also be considered indicators of the studied process in nature.

• (a) It has been experimentally shown for the first time that the process of carbonate–metal interaction is possible at low temperatures (such as 800 °С at 6.3 GPa) corresponding to the P–T path of cold subduction; conditions of cold subduction;

• (b) The interaction of Fe,Ni metal and carbonate in the range of 800–1200 °С leads to the formation of the following reaction zones (from the reduced center to the oxidized periphery): metal → metal + wüstite/magnesiowüstite → magnesiowüstite + graphite ± Mg,Fe,Ca carbonates → magnesite + aragonite; at 1400–1550 °С, Fe,Ni metal melt + diamond → Fe,Ni metal melt + magnesiowüstite + diamond → magnesiowüstite + graphite → calcium carbonate melt + magnesiowüstite + diamond + graphite → aragonite + ferropericlase + graphite;

• (c) At T ≥ 1200 °С, the formation of a Fe,Ni–C melt occurs under reducing conditions; at T ≥ 1330 °С, the formation of a carbonate melt under oxidizing conditions; and at T ≥ 1400 °С, spontaneous diamond formation. An increase in temperature is accompanied by an increase in the rate of advance of the redox front from 1.3 μm/h (800 °C) to 118 μm/h (1550 °C). For the compositions of the final phases, with an increase in temperature, an increase in the concentration of Ni in the metal at the contact with the initial carbonate (from 65 wt.% at 800 °C to 72 wt.% at 1200 °C) and an increase in the contents of MgO in magnesiowüstite in both reducing (from 10 wt.% at 800 °C to 16–17 wt.% at 1550 °C) and oxidizing (from 22 wt.% at 900 °C to 71.8 wt.% at 1400 °C) conditions are observed;

• (d) The carbonate–metal interaction process is characterized by a directed distribution of siderophile and lithophile elements. As a result, the Fe,Ni melt becomes enriched in Ni, and the carbonate melt acquires a predominantly calcium composition;

• (e) Raman spectroscopy was used to quantitatively assess the degree of structural perfection of graphite synthesized at 800–1550 °C. With an increase in the crystallization temperature, a clear pattern of a decrease in the value of the I_D/I_G ratio was established, which can be used as a semiquantitative geothermometer of natural graphite-producing processes;

• (f) Patterns of diamond crystallization established in experiments on the redox interaction (Fe,Ni)–(Mg,Ca)CO₃ have shown that the addition of nickel eliminates carbide formation in the reduced part of the sample and ensures stable diamond crystallization at 1400–1550 °C both in metal– carbon and carbonate melts;

• (g) Diamonds synthesized in a single mineral-forming process with (Fe,Ni)–(Mg,Ca)CO₃ redox interaction have significant differences. In the metal–carbon melt, the morphology of diamond is determined by the layer-by-layer growing faces of the octahedron. The indicator characteristics of these diamonds are nitrogen–vacancy and nickel (884 nm) centers at 1400 °C or nickel–nitrogen centers (S3, 598 nm, 727 nm, 746 nm, etc.) at 1550 °C. The inclusions are represented by taenite and magnesiowüstite. The morphology of diamonds formed in the carbonate melt is determined by the faces of a cube and an octahedron, which are characterized by vicinal growth. The photoluminescence spectra of such diamonds contain nitrogen–vacancy centers H3, NV⁰, and NV^–. Carbonates were identified as inclusions.

This work was supported by the Russian Science Foundation under grant No. 19-17-00075, https://rscf.ru/project/19-17-00075/.