Natural moissanite (SiC) – a low temperature mineral formed from highly fractionated ultra-reducing COH-fluids

High pressure experiments

A second set of experiments intended to synthesize SiC first at low temperature from tetrakis-silane (((CH₃)₃Si)₄Si) or stearic acid (C₁₈H₃₆O₂) and olivine + serpentine or talc + MgO at 500°C, 0.2 GPa. These experiments did not result in SiC or equilibrium assemblages. We then aimed at synthesizing SiC + magnesio-silicates at 1000-1600°C, 2-10 GPa employing talc + MgO and tetrakis-silane as starting materials. At 6 GPa, 1300°C we obtained SiC + olivine + opx, but the assemblage was unequilibrated with strong compositional zoning in the silicates and further experiments in this direction were abandoned.

Thermodynamic data

Thermodynamic databases suited for calculating phase equilibria in Earth sciences mostly contain only silicates and oxides. In order to perform calculations in mantle materials at ultra-reduced conditions, alloys, carbides and silicides were combined with silicate data from Holland and Powell ([2011]). The data sources employed were: Cementite (Fe₃C): calorimetric data from Gustafson ([1985]), volume data from Li et al. ([2002]), Wood et al. ([2004]); Fe₇C₃: Djurovic et al. ([2011]), Mookherjee et al. ([2011]), Nakajima et al. ([2011]), Moissanite (SiC): Lacaze and Sundman ([1991]), Miettinen ([1998]), Aleksandrov et al ([1989]), Li and Bradt ([1987]). The FeSiC-alloy and the iron silicides are based on the 1-atm phase diagram by Lacaze and Sundman ([1991]) and on the volume EoS of Brosh et al. ([2009]). Details of the adoption of the ternary metal, carbides and silicides are given in a forthcoming paper (Golubkova et al. [2014]). Phase relations were computed by free energy minimization with the Perple_X software package (Connolly JAD [2009]).

Phase equilibria and phase compositions

Experiments at 2 GPa, 1500°C (G#2) and 10 GPa, 1700°C (G#6) resulted in equilibrated olivine + orthopyroxene + SiC plus iron silicides (Table 1, Figure 1). The experiment at 10 GPa, 1500°C (G#3) resulted in coarse opx + SiC plus three iron silicides (FeSi, FeSi₂, and FeSi₄) and olivine that did not develop equilibrium rims large enough for measurement. These charges did not contain any residual silicates from the starting material, but in both the 1500°C and 1700°C experiments at 10 GPa, most of the San Carlos olivine disintegrated to an almost Fe-free olivine containing many submicron iron silicide inclusions (Figure 1b). These inclusions are too small to determine their stoichiometry. The equilibrium X _Mg of olivine and opx in these experiments are 0.993-0.998 (Table 1). Initially, we added Ir metal as internal oxygen fugacity monitor (Woodland and O’Neill, [1997]; Stagno and Frost, [2010]), this experiment (G#1) at 1300°C, 2 GPa yielded opx (X _Mg = 0.982) and quartz. However, abundant residual starting material San Carlos olivine testifies for incomplete equilibration questioning whether the X _Mg of the opx of this experiment corresponds to equilibrium with SiC. We have thus only used experiments #2, #3 and #6 for oxygen fugacity calculations.

Oxygen fugacity (f _O2 ) in the experiments

SiC and graphite or diamond being pure phases. The resulting oxygen fugacities are log f _{O 2} = − 15.1. (6) or $\Delta \mathit{\text{log}}{f}_{O2}^{IW}=-6.4$. at 2 GPa, 1500°C (G#2), and log f _{O 2} = − 10.6 (4) or $\Delta \mathit{\text{log}}{f}_{O2}^{IW}=-5.6$ at 10 GPa, 1700°C (G#2).

Calculation of T-f _O2 sections

Reactions governing reduced phases in the mantle

To understand the succession of phase assemblages with decreasing oxygen fugacity, we calculated f _{O 2}-temperature diagrams (Figure 2) and phase compositions (Figures 3 and 4) for a model harzburgite at 2 and 10 GPa. For reference, we also give the quartz-fayalite-magnetite (QFM), the graphite/diamond-CO₂-CO equilibrium (CCO), and the iron-wustite buffer (IW), even if these do not occur as reactions in the harzburgite phase diagram. The position of the iron-wustite buffer is calculated from Campbell et al. ([2009]), which yields about half a log-unit lower values than the formulation of O'Neill ([1988]).

In the carbon-bearing upper mantle, the iron-wustite equilibrium does not occur, its equivalent leading to the appearance of reduced Fe⁰ are the iron carbide forming reactions

$6\mathrm{olivine}+2\mathrm{graphite}=3\mathrm{o}\mathrm{p}\mathrm{x}+2\mathrm{cementite}\left({\mathrm{F}\mathrm{e}}_3\mathrm{C}\right)+3{\mathrm{O}}_2$

(10a)

and

$14\mathrm{olivine}+3\mathrm{diamond}=7\mathrm{o}\mathrm{p}\mathrm{x}+2{\mathrm{F}\mathrm{e}}_7{\mathrm{C}}_3+7{\mathrm{O}}_2$

(10b)

which are only stable at relatively low temperatures, i.e. 1140 to 1260°C at 2 to 10 GPa. At higher temperatures, carbides do not form but a metal alloy results from

$2\mathrm{olivine}+2\mathrm{x}\mathrm{graphite}/\mathrm{diamond}=1\mathrm{o}\mathrm{p}\mathrm{x}+\left(2+2\mathrm{x}\right){\mathrm{FeSiC}}^{\mathrm{alloy}}+1{\mathrm{O}}_2$

(11)

where FeSiC^alloy is a ternary metal solid solution. The stoichiometry of reaction (11) depends on the alloy composition. The molar fraction of C in the alloy ranges from 0.13 to 0.33 at 2-10 GPa, 1150-1700°C. At oxygen fugacities of initial metal formation, Si-fractions are <0.1 mol% in the alloy. For a graphite/diamond saturated mantle with a bulk X _Mg of 0.9, reactions (10a, 10b) and (11) occur at 2 GPa about one log-unit below the iron-wustite buffer and about one log-unit above IW at 10 GPa (Figure 2a,b). About 5-9 log-units below the carbide-forming reactions, increasing chemical potential of Si leads to the replacement of Fe₃C or Fe₇C₃ by the FeSiC-alloy then containing significant Si (Figure 5b,d). Silicon contents in the alloy increase with decreasing oxygen fugacity leading to a phase transition from a face centered to a body centered metal structure (fcc and bcc in Figure 5).

Silica carbide forms principally through a reaction involving moissanite, olivine, orthopyroxene and elemental carbon as graphite or diamond (moissanite-olivine-opx-carbon: MOOC)

$1\mathrm{o}\mathrm{p}\mathrm{x}+1\mathrm{graphite}/\mathrm{diamond}=1\mathrm{olivine}+1\mathrm{SiC}+1{\mathrm{O}}_2$

(12)

at 2 GPa at 7.7-4.8 log-units (1100-1700°C) below equilibria (10a, 10b) and (11). At 10 GPa this difference is 9.9-6.5 log units (1100-1700°C). This corresponds to a Δlogf _O2 of IW-9 to IW-6, the difference decreasing with temperature. Along a mantle adiabat, moissanite forms at IW-8.0 at 2 GPa and at IW-6.7 at 10 GPa. Reaction (12) was previously identified as being responsible for the appearance of moissanite in the upper mantle (Mathez et al. [1995]; Ulmer et al. [1998]).

Further reduction causes the FeSiC-alloy to be replaced by stoichiometric iron silicide FeSi. At 2 GPa this happens about ½ a log-unit below the SiC-buffer (12), while at 10 GPa, the FeSi-forming reaction intersects reaction (12) leading to an invariant point. At about 3 and 5 log-units below the SiC-forming reaction (12), opx and then olivine (at these conditions pure enstatite and forsterite, respectively) become unstable and a ternary metal alloy of almost pure Si forms through

$1\mathrm{o}\mathrm{p}\mathrm{x}+0.01\mathrm{SiC}=1\mathrm{olivine}+1.01{\mathrm{S}\mathrm{i}}^{\mathrm{alloy}}+1{\mathrm{O}}_2$

(13)

$1\mathrm{olivine}+0.01\mathrm{SiC}=1\mathrm{periclase}+1.01{\mathrm{S}\mathrm{i}}^{\mathrm{alloy}}+1{\mathrm{O}}_2$

(14)

where the Si-alloy has 1 mol% C. Completion of reactions (13) and (14) would leave no oxidized Si. The calculations do not indicate any Fe in the Si metal and only 0.2 to 1.3% C. Reaction (13) is relevant in nature as pure Si-metal attached to SiC has been reported by Trumbull et al. ([2009]).

Summarizing, a harzburgitic mantle would have the following succession of reduced phases: graphite/diamond → (Fe₃C/Fe₇C₃) → Fe-rich FeSiC^alloy → SiC → FeSi → Si^alloy. At pressures ≥10 GPa, FeSi may form before moissanite.

Note that our calculations do not include liquids and that at high temperatures FeSiC^alloy and FeSi become metastable with respect to liquid. This is of particular relevance at oxygen fugacities just below the metal-forming reaction (11) where a de facto binary FeC-alloy is almost Si-free. The Fe-C eutectic is located at 1150°C, 1 atm (Chipman [1972]) and 1210°C, 10 GPa (Hirayama et al. [1993]; Rohrbach et al. [2014]), although this location is not entirely unambiguous as Lord et al. ([2009]) report the eutectic at 10 GPa at 1420°C. In a graphite/diamond-saturated system, the ternary FeSiC-alloy would then be replaced by a metallic melt phase. FeSi has a temperature of congruent melting of 1410°C at 1 atm (Lacaze and Sundman [1991]), but to our knowledge its high pressure melting is unknown.

Mineral compositions at reduced conditions

The above reactions are discontinuous reactions that delineate the various stability fields. For the silicate mantle the continuous reactions consuming the fayalite and ferrosilite components in olivine and orthopyroxene are more dramatic. As illustrated in Figures 3 and 4, X _Mg in olivine is almost invariant in each field of coexistence with magnesite or graphite/diamond. Immediately below the metal forming reaction, Fe²⁺ becomes strongly reduced, within only 2 log-units, the X _Mg of olivine (and similarly opx) in equilibrium with metal or Fe₃C/Fe₇C₃ and graphite/diamond increases to ~0.99 (Figures 3 and 4). When oxygen contents corresponding to the oxygen fugacities of SiC are reached, the equilibrium X _Mg in the silicates is >0.999, almost invariant with pressure and temperature (at least for 1100-1700°C, 2-10 GPa). This confirms the experimental findings. Concomitant with the increase in X _Mg of the silicates, an increase of the Si-content of the FeSiC-alloy is predicted (Figures 4 and 5), mirrored by a decrease of X _Fe ^metal from ~0.90 to 0.70 at 2 GPa and from 0.76 to 0.66 at 10 GPa.

Silica carbide stability with temperature and pressure

A major question concerning the stability of SiC in the mantle regards the evolution of the difference in oxygen fugacity between the metal or Fe-carbide forming (10a, 10b, 11) and the SiC-forming reactions (12). This difference, Δlogf _{O 2} ^{(12)-(10a,10b,11)}, decreases with increasing temperature but at near-adiabatic temperatures only by -0.1 log-units per 100°C. With increasing pressure there is a slight increase of Δlogf _{O 2} between the moissanite forming reaction and the equivalent of IW in the mantle such that along a mantle adiabat Δlogf _{O 2} ^{(12)-(10a,10b,11)} remains almost constant at ~ -4.5.

Oxygen fugacities calculated from the experiments

In Figure 2a and b we plot the oxygen fugacities calculated for the three experiments with olivine and/or orthopyroxene equilibrated with SiC. The oxygen fugacities calculated from the experiments through the moissanite-forming equilibrium (7) are in perfect agreement with the calculated phase diagram. They are also in good agreement with the values derived from Ulmer et al. ([1998]) from experiments on a Mg-Si-C-O system that cross-compared equilibrium (12) at 1500°C to other common oxygen buffers.

Oxygen fugacities from the iron silicide involving equilibria (red symbols) were calculated without involving SiC and may hence be taken as independent confirmation of the f _O2 -value of the SiC-forming reaction. Nevertheless, the interpretation of the silicides in the two experiments at 1500°C is not straightforward. Congruent melting of Fe₂Si occurs at 1212°C at 1 atm, but the dP/dT slope of this reaction is unknown. It is thus well possible that the Fe₂Si in the 2 GPa experiment G#2 crystallized upon quench from a FeSiC-liquid. At 10 GPa, the silicides should be stable as solids. Nevertheless, the experiment at 1500°C (G#3) contains three silicides, FeSi, FeSi₂ and FeSi₄. Together with opx, diamond and SiC, these are six phases, one phase too much for a divariant field of a five component system (Mg-Fe-Si-C-O), but possible if Ni is also accounted for as a component. Nevertheless, one of the silicides may well form in a transient stage, when the FeO of the fayalite component is reduced. We interpret the f _O2 -value calculated from equilibrium (1) as reflecting a transient stage at the beginning of the experiment.

X _Mg of silicates coexisting with SiC

The results from the experiments in terms of X _Mg of silicate minerals coexisting with SiC agree very well with the calculated phase compositions (Figures 3 and 4). Together with the constraints on oxygen fugacity, a coherent picture forms: moissanite can only be expected in a mantle or any rock with silicates essentially devoid of Fe²⁺. This unsurprising finding has strong implications for the origin of SiC in mantle rocks. It demonstrates that any significantly large domain of mantle at oxygen fugacities of SiC-stability (i.e. void of Fe²⁺ in silicates and oxides) has not yet been identified.

Both the experiments and the calculations demonstrate that for plausible mantle compositions with SiC stable, the X _Mg of coexisting olivine or orthopyroxene is >0.993. Hence SiC cannot be in equilibrium with typical mantle that has olivine and pyroxenes with X _Mg of 0.88-0.92. The experiments only constrain the X _Mg of olivine and orthopyroxene in equilibrium with SiC. Nevertheless, in nature, X _Mg of serpentine is generally higher than that of coexisting orthopyroxene, which is similar to that of clinopyroxene and X _Mg of olivine is lowest among these phases. One can thus safely assume that for all silicates characteristic for peridotites (except of garnet) similarly high X _Mg of >0.993 are to be expected.

As SiC is found in mantle rocks and kimberlites from which anomalous X _Mg values for the silicate phases are not reported, the formation mechanism of SiC should explain how average X _Mg mantle silicates and SiC persist in the same rock volume.

Diffusion limits on temperatures of SiC preservation

Considering a SiC grain within a matrix of FeMg-silicates, one can calculate the time scale for diffusive equilibration between SiC and the FeMg-silicates. With time, SiC would react with the fayalite component of olivine to form opx and metal or carbide until the SiC grain is completely consumed. Taking a putative SiC grain in a matrix of otherwise undisturbed mantle silicates, Fe-Mg diffusion coefficients of the silicates allow for calculation of the time-temperature conditions for which a given grain size of SiC would react with olivine until its complete destruction. Assuming spherical symmetry, the amount of Fe²⁺ needed to oxidize all Si of a SiC grain in an olivine (Fo₉₀) matrix occurs within 2.3 radii. Fe-Mg volume diffusion in olivine and orthopyroxene are characterized by diffusivities of D₀ = 10^-7 m²/s and an activation energy E _A of 250 kJ/mol (Brady and Cherniak, [2010]). We have calculated the relation between time and temperature for a given grain size of radius r employing the characteristic diffusion length 2.3r. This calculation yields the time necessary to diffuse a sufficient amount of Fe²⁺ towards a SiC grain, then being consumed and forming orthopyroxene (and metal), for a given temperature and grain size (Figure 6). These calculations show that mm-sized grains would disappear through diffusive equilibration within a million years at temperatures above 800-900°C. Preservation of SiC in a mantle mineral matrix with X_Mg = 0.9 would not be possible at adiabatic mantle temperatures, even SiC grains of 3-10 cm would only last for 10’000 years. This conclusion remains robust when employing the experimentally determined range of D ₀ – E _A pairs (for summary of diffusion data see Brady and Cherniak, [2010]).

These time spans could be prolonged if SiC forms in veins of essentially Fe-free silicates, but such mineral compositions are not reported for peridotites or kimberlites. Regarding the putative high-temperature origin of SiC, at temperatures of the asthenospheric mantle, SiC grains of initially >50 cm would be necessary to persist for >1 million years. It is thus unlikely that SiC would survive any period of sustained high temperatures in the mantle.

A model for SiC formation in nature

Terrestrial environments are typically too oxidizing to permit SiC formation, thus exceptional circumstances need to be explored. One plausible scenario starts with a slightly reduced graphite/diamond-saturated COH-fluid that evolves towards an ultra-reduced composition through the formation of first carbonates or spinel solid solution and then hydroxide minerals, all from anhydrous silicates. In this model, the fluids react with its immediate surroundings while percolating upwards, leaving newly formed carbonates, spinel or hydroxides behind. The fluid-mediated formation of magnetite component in spinel from mantle silicates drives any fluid to more reducing compositions (away from the O₂-corner in Figure 7). The fluid-mediated formation of carbonates (from mantle silicates) leads to more reduced residual fluids when the starting COH-fluid has a composition to the reduced side of the CO₂-H₂O tie-line (Figure 7). Fluids from sediments rich in organic material are expected to already have H:O-ratios slightly to the H-rich side of the H₂O-C tie-line (Connolly JAD [1995]; Figure 7c). Nevertheless, as long as they are saturated in graphite, also fluids slightly to the O-rich side of H₂O could be suitable precursors (Figure 7b). The following steps would then lead to a fluid that may mediate SiC formation:

1a) Starting from fluids with some CO₂, the precipitation of carbonates through reaction of the fluid with silicates (e.g. olivine + CO₂ = opx + magnesite; or cpx + olivine + CO₂ = calcite + opx) removes oxygen from the fluid, causing the fluid composition to evolve away from CO₂, the star in Figure 7b representing a hypothetical starting composition.

The simplest case enabling SiC formation is to start with an aqueous fluid that contains more methane than CO₂ (molar), i.e. with a bulk composition more reducing then the H₂O-maximum. Sequestration of oxygen from the fluid by hydroxide minerals would then directly lead to extremely reducing conditions. In mantle compositions, serpentine or brucite are only stable at relatively low temperatures (<700°C) as could be expected for fluid-percolation in supra-subduction mantle or during orogenic metamorphism. In fact, SiC has been reported from serpentinites (Xu et al. [2008]). Moissanite accounting for 8% of a rock with a bulk X _Mg  = 0.998 has been reported from a beach pebble which matrix is brucite-dominated (Di Pierro et al. [2003]). This latter rock contains minor (remnant) chromite and orthopyroxene of typical mantle composition and may represent a completely metasomatized peridotite. Otherwise, moissanite is rare in most rocks and mostly described from heavy mineral concentrates. Phases grown with or onto moissanite are Fe-silicides, Si- and other metals. Unfortunately, the textural context and relations to silicates remain mostly unknown providing little evidence for the type of reactions leading to moissanite.

Step 3 is likely to occur through grain scale equilibria where new hydrous minerals form in equilibrium with the fluid but where the bulk rock is in disequilibrium with this fluid. Alternatively, discrete metasomatic zones could host SiC and hydrous phases as in the rock from Turkey described by Di Pierro et al. ([2003]). As SiC is a rare mineral, it appears likely that in most cases either reduced fluids are consumed before extreme fractionation or fluid:rock ratios in percolation zones are too high to change fluid-compositions much. However, when such prolonged fractionation occurs, C-O-H fluids may fractionate to H:O ratios sufficiently high for SiC-precipitation.

Evidence for reduced zones in the mantle wedge

Although the mantle wedge is generally thought to be rather oxidized, evidence for a fairly reduced mantle wedge has been reported for xenoliths from the Avacha volcano, Kamtchatka (Ishimaru et al. [2009]). The bulk of minerals in these peridotite xenoliths equilibrated at 870-1040°C and oxygen fugacities of ΔQFM = -0.2 to +1.5, and most silicates have X_Mg values characteristic for the mantle. Nevertheless, some xenoliths contain small amounts of Ni-, Fe- and possibly Ti-metal and Ti-rich Fe silicides mostly along grain boundaries. Serpentine is absent in these xenoliths. The sporadic occurrence of metal and silicides are ascribed to a reducing fluid phase with oxygen fugacities ≤ IW that causes precipitation of the reduced phases under disequilibrium with the host rock. Exact temperatures could not be determined but are suggested to lie below the bulk equilibration temperature and serpentine stability (Ishimaru et al. [2009]), probably between 700 and 900°C.

Carbon isotope composition

Natural moissanites have a light carbon isotopic composition of δ¹³C = -20 to -37‰ (Mathez et al. [1995], Trumbull et al. [2009]). Generally, the main source material for such light compositions is organic matter typically reaching down to -30‰. Upon burial and heating through subduction or orogenesis such organics will generate fluids with low δ¹³C. However, the relation between the original fluid composition and the δ¹³C of precipitated SiC is not straightforward: The fluid evolves to ultra-reduced compositions through removal of H₂O by hydrous phases, this process is necessarily accompanied by graphite or diamond precipitation (see Figure 7). Poulsen ([1996]) calculated a fractionation factor Δ¹³C_graphite-CH4 of +1 to +5‰ (750 to 500°C) and hence graphite fractionation renders an already reduced fluid lighter. On the other hand mixing of the fluid with mantle carbon of δ¹³C ~ -5‰ (Deines [2002]) would lead to heavier compositions. Nevertheless, Deines ([2002]) points out that the light carbon mode of mantle samples of δ¹³C ~ -20 to -25‰ is not necessarily solely related to surface carbon, also the fractionation factor Δ¹³C_SiC-CH4 remains unknown. Hence a quantitative model is difficult, but the light isotopic carbon composition of moissanite supports an involvement of initially reduced fluids probably originating from organic material.

Kimberlites

Moissanite is reported from kimberlites (e.g. Shiryaev et al. [2011]) which contain light ¹³δ C diamonds (Trumbull et al. [2009]), but their direct paragenetic association remains uncertain (Gurney [1989]). In one case, SiC was described as part of a complex inclusion in diamond (Leung [1990]), the latter effectively halting diffusive equilibration with the surroundings. Moissanite could survive millions of years in the sub-lithospheric cratonic mantle at low temperatures (<800°C). In principle, the same fluid or melt that leads to diamond formation could also precipitate SiC in a later more fractionated stage, which should occur at shallower pressures and lower temperatures. Regardless of these observations, as kimberlites are CO₂-rich magmas whose eruptions are propelled by CO₂, it is not possible that the kimberlite magma would be equilibrated with SiC. Hence, moissanites in kimberlites represent xenocrysts, the kimberlites sampling cold and reduced parts of lithospheric mantle keels.

Oxygen-content and mass balance