1. Introduction
Establishing the specific features of the redox profile throughout the Earth’s mantle with corresponding changes of structure, phase and chemical rocks compositions is one of the primary goals of modern mantle mineralogy and petrology. The oxidation state of mantle rocks may be defined by heterogeneous reactions between oxygen and iron- or carbon-bearing minerals [
1,
2]. Under redox conditions of the upper mantle, formation of metallic iron (or Fe–Ni alloy) occurs below the metal precipitation curve, the position of which is defined by the oxygen fugacity (ƒO
2) at which Fe precipitates from mantle silicates, and the depth or pressure of metal precipitation line depends on bulk rock composition. Pressures of some metal precipitation reactions were established experimentally and theoretically [
3,
4,
5,
6,
7,
8], and it was predicted that, at depths greater than 250 km (P ≥ 7 GPa), the Earth’s mantle is reduced enough to become metal-saturated. More specifically, it was experimentally demonstrated [
6] that pure metallic iron could be formed as a result of disproportionation reaction, where majoritic garnet acted as a sink for ferric iron, with Fe
3+/ΣFe ratios varying from 0.02–0.12 (~7 GPa) to 0.4 (14 GPa). Besides the stabilization of Fe
3+-bearing garnet species, presence of highly reduced metal-saturated mantle rocks at great depths has a number of petrologic consequences, affecting the behavior of redox-sensitive elements, e.g., carbon. Under these conditions, such carbon-bearing phases as carbides, diamond and graphite, as well as fluid [
4,
9,
10,
11], are stable. According to the data in [
4,
9,
10,
12], under metal-saturated conditions in the depleted mantle (20–120 ppm C), carbon presents as a solution in metal and in the undepleted mantle (300–800 ppm C) it presents as carbides (Fe3C and Fe7C3), diamond and graphite. Presence of iron carbides at great depths is evidenced by inclusions in diamonds [
13,
14,
15,
16,
17]. Recent experimental studies [
18,
19,
20] show that iron carbide participates in the deep carbon cycle processes, involving redox graphite- and diamond-forming reactions, metasomatic interactions, generation of Fe–C-based melts and formation of ferrous and ferric silicates.
Studies on the deep carbon cycle [
9,
21,
22,
23] demonstrate that subduction is one of the main mechanisms of the transport of carbonate-bearing rocks, oxidized melts and fluids into the Earth’s mantle. Discovery of xenoliths of strongly oxidized rocks with ƒO
2~FMQ + 1 log unit [
5,
24] evidences that part of the lithospheric mantle was subjected to the action of mobile metasomatic agents—carbonate melts or oxidized components of C–O–H fluid [
2,
25,
26,
27,
28]. Subduction of crustal material into the mantle also enables redox interactions between oxidized (carbonate–oxide–silicate) species and reduced metal- or carbide-bearing rocks [
7]. Possibility of this interaction is confirmed by data on inclusions in diamonds, varying in composition from highly reduced [
13,
29,
30,
31], to oxidized, including silicates and oxides [
32,
33,
34,
35], carbonates and pure CO
2 [
36,
37]. Recently, the model of diamond-forming mixing process was proposed based on the complex study of the carbon isotopic compositions of sub-lithospheric diamonds and oxygen isotopic compositions of inclusions therein [
38]. The relationship between δ
13C and δ
18O for Brazil diamonds suggests the diamond formation via interaction of slab-derived carbonate melt with reduced (carbide- or metal-bearing) mantle and confirms the presence of iron carbide phase in the deep mantle.
To date, carbide behavior under mantle pressures and temperatures has been experimentally investigated in Fe–C, Fe–Ni–C, Fe–C–S and Fe–Ni–C–S systems [
12,
19,
39,
40,
41,
42], as well as in more complex systems, modeling carbide interaction with mantle minerals–oxides [
43], sulfides [
20] and carbonates [
18]. However, there is a lack of studies devoted to the modeling of the iron carbide behavior in the presence of fluid, under silicate mantle P,T,x-conditions, which can be valuable sources of new data on mantle oxygen fugacity (ƒO
2) patterns, potential graphite/diamond-producing processes and characteristic features of the formation of Fe-bearing silicates as well as carbonate-silicate melts.
The main goals of the present study, which can be achieved by high-pressure experiments, were: (1) to model carbide–fluid reactions in the carbide–oxide–carbonate systems at the pressures close to metal-precipitation line (6.3 and 7.5 GPa); (2) to reconstruct the graphite- and diamond-forming reactions, in which carbide and carbonate are both sources of carbon; and (3) to study the conditions of formation of ferric and ferrous silicates and Fe-rich carbonate-silicate melts under lithospheric mantle P,T-parameters.
2. Materials and Methods
Experiments were carried out using a multi-anvil high-pressure apparatus of a “split-sphere” type (BARS) [
44] in the Fe
3C–SiO
2–Al
2O
3–(Mg,Ca)CO
3 (carbide–oxide–carbonate) systems at the pressures of 6.3 and 7.5 GPa, in the temperature range of 1100∓1650 °C, and a run duration from 8 to 40 h. Details on high pressure cell design as well as the calibration data were published previously [
45,
46]. To model carbide–fluid interactions, oxide–carbonate mixtures were used as a CO
2-fluid source due to the decarbonation reactions in high pressure-high temperature experiments. Silicates, formed via decarbonation, acted as a sink for oxidized iron. Initial reagents were synthetic Fe
3C (cohenite), SiO
2, Al
2O
3 (99.99% pure) as well as natural specimens of magnesite and dolomite (<0.05 wt % of impurities, Satka deposit, Urals, Russia). Cohenite was preliminarily synthesized in a Fe–C system at P = 5.7 GPa and T = 1450 °C, and thoroughly analyzed by X-ray powder diffraction (Stoe IPDS-2T diffractometer, STOE, Darmstadt, Germany). The initial mixture of the carbonates had a bulk composition of Mg
0.9Ca
0.1CO
3. Proportions of initial materials (
Table 1 and
Table 2) were chosen to produce CO
2-dominated fluid and garnet via decarbonation reaction (Reaction (1)) and to realize further interaction of these phases with cohenite (Reaction (2)):
Two different methodical approaches were used for experiments at 6.3 and 7.5 GPa. The experimental technique and methodical approach for the series at 7.5 GPa (so-called “sandwich”-type experiments) was developed and described in detail in [
47]. The elaborated technique made it possible to perform experimental modeling of fluid- and melt-generating processes as well as of redox interaction of these fluids and melts with Fe-rich mantle minerals (ilmenite, chromite, wüstite, Fe
0, pyrrhotite, and cohenite) under P,T-conditions of the lithospheric mantle [
18,
47,
48,
49,
50]. Using this technique ensures the formation of a redox gradient by separation of the reaction volume of ampoules into peripheral (carbonate–oxide, i.e., oxidized) and central (containing an iron concentrator, i.e., reduced) parts (
Figure 1a). This enables decarbonation reactions and generation of Fe-enriched carbonate-bearing melts as well as C
0-producing redox interaction of the newly formed fluid or melt with the Fe-concentrating phases under mantle P,T-parameters. It should be emphasized that “sandwich”-type experiments were developed to study specific reactions on a base of reaction zoning of samples, without reaching equilibrium conditions.
When assembling, an ampoule made of the pressed mixture of oxides and carbonates was placed into a Pt capsule (6 mm in outer diameter). A pellet made of pressed Fe
3C was mounted in the center of this ampoule. Therefore, the iron-bearing phase was separated from the Pt-capsule. This methodical approach guarantees minimal iron loss, when even runs at 1650 °C and 7.5 GPa are characterized by Fe loss about 1.0–1.2 wt %, which is confirmed by analysis of Pt-capsule material after experiments and by the mass balance calculations. When assembling, synthetic diamond seed crystals of the cuboctahedral shape (~0.5 mm in size) were placed in the carbonate–oxide ampoule (
Figure 1a) to estimate whether the redox conditions in the reaction ampoules correspond to the stability field of diamond and to evaluate the possibility of diamond growth.
The methodical approach used for the series of experiments in carbide–oxide–carbonate system at 6.3 GPa (so-called “mixture”-type experiments) was described in detail in [
43]. Even though Pt is the optimal capsule material for experiments under high pressures and high temperatures, it cannot be used for Fe-bearing systems if there is a direct contact of Pt with phases of iron. Considering our previous experience in a field of carbide-involving experiments under mantle P,T-conditions [
39,
43], we used graphite as appropriate capsule material. These graphite capsules set the upper limit of the oxygen fugacity values in samples close to the C–CO buffer equilibrium [
47]. Starting materials, placed in graphite capsules, were powdered (particle size ~10–20 µm) and thoroughly homogenized. It should be noted that some part of the iron carbide was added into charge as relatively large crystals (300∓400 µm) to enable the processes of the carbide–oxide, carbide–carbonate and carbide–silicate interactions. For this reason, a small excess of iron carbide relative to the other reagents of Reaction (1) was created. Diamond seed crystals of cuboctahedral habit (~500 µm in size) were placed into the ampoules, to evaluate the possibility of diamond growth under conditions of the experiments.
We should separately emphasize the problem of the oxygen fugacity buffering of the samples in the present research. Iron carbide and CO2 fluid are extremely fO2-contrast, with ∆fO2 of about 5 log units. When modeling carbide–fluid reactions, which can occur as a result of subduction of oxidized material in the deep reduced mantle, it is inappropriate to use outer buffer that will set ƒO2 values at IW level. If this were done, CO2 fluid, generated as a result of decarbonation reactions, would be immediately reduced to form elemental carbon without participation of iron carbide. Thus, in the “sandwich”-type experiments, we used excess of cohenite to model reduced conditions in the central part of the sample and, in “mixture”-type experiments, we used outer buffer—graphite capsules, setting upper limit of fO2 values close to C–CO (and FMQ) buffer, which are believed to be average upper mantle values.
The chemical composition of final phases was analyzed by energy dispersive spectroscopy (Tescan MIRA3 LMU scanning electron microscope, TESCAN, Brno, Czech Republic) and microprobe (WDS) analysis (Camebax-micro analyzer, CAMECA, Gennevilliers, France).
Microprobe analyses of silicates, oxides, and carbides were performed at an accelerating voltage of 20 kV, a beam current of 20 nA, a counting time of 10 s for each analytical line, and electron beam diameter of 2∓4 µm. The electron beam diameter of 20∓100 µm was used when studying the composition of microdendrites of quenched metal–carbon melt. Quantitative analysis (both microprobe and energy dispersive spectroscopy) of carbon performed using mineral standards of cohenite, calcite and magnesite. Polished samples and standards for microprobe were coated with carbon; samples for energy dispersive spectroscopy were coated with carbon for first series of analyses, and with chromium for second series of ones. The phase relationships were studied by scanning electron microscopy. Analytical studies mentioned above were performed at the Center for Collective Use of Multi-element and Isotopic Analysis of the Siberian Branch of the Russian Academy of Sciences. Analysis of the composition of iron-bearing phases and valence state of iron therein was performed using Mössbauer spectroscopy. Measurements were carried out at room temperature on a MS-1104Em spectrometer (Cordon, Rostov on Don, Russia) with a Co57(Cr) source and a powdered absorber with a thickness of 1∓5 mg/cm
2. The methodology of measurements and analysis of Mössbauer spectra was described in detail in [
51] (and Supplementary Information therein).
5. Conclusions
(1) It was experimentally demonstrated that carbide–oxide–carbonate interaction led to decarbonation, with the formation of Fe,Mg,Ca-garnet and CO2 fluid, and graphite or diamond-producing redox reactions, involving iron carbide, mantle oxidized fluids, melts, carbonates, Mg,Fe silicates and oxides.
(2) Carbide–oxid–carbonate interaction realized in the experiments with ƒO2-gradient at 7.5 GPa and 1250–1350 °C resulted in the crystallization of magnesiowüstite + garnet + graphite ± cohenite assemblage. Graphite under reduced conditions was produced through the redox interactions of carbide with carbonate or CO2, and, under the oxidized conditions, as a result of the redox reaction of magnesiowüstite and CO2. At 1450–1650 °C, carbide–oxide–carbonate interactions resulted in the formation of Fe3+-magnesiowüstite + Fe3+-garnet + graphite assemblage as well as generation of Fe2+,3+-rich carbonate–silicate melt. This melt, saturated with carbon, acted as a medium of graphite crystallization and diamond growth on seeds.
(3) The experiments without ƒO2-gradient at 6.3 GPa demonstrated that carbide–oxide–carbonate interaction was accompanied by decarbonation reactions with the formation of Fe,Mg,Ca-silicates and CO2 fluid as well as by carbon-producing redox reactions between the CO2 fluid and cohenite. These processes resulted in the formation of Fe2+,Fe3+,Mg-silicate + magnetite + graphite assemblages (1100∓1200 °C) and Fe3+-bearing garnet + orthopyroxene + graphite (± diamond growth) (1300∓1500 °C).
(4) Potential mechanisms for crystallization of graphite or diamond during the carbide–oxide–carbonate interaction are oxidation of cohenite by CO2 fluid or carbonate–silicate melt to FeO and Fe2O3, which is accompanied by carbon extraction from Fe3C and corresponding reduction of CO2 or carbonate component of the melt to C0.