Mantle–slab interaction and redox mechanism of diamond formation

Significance

The primary question that we address in this study is what happens when a carbonate-bearing crust is subducted to depths where the Earth’s mantle is metal saturated. Subduction plays an important role in the evolution of the Earth’s interiors, but the mechanism of the interaction between the oxidized slab and reduced mantle remains unclear. Here we report the results of high-pressure redox-gradient experiments on the interaction between Mg-Ca-carbonate and metallic iron, modeling the processes at the mantle–slab boundary, and present mechanisms of diamond formation ahead of and behind the redox front. We demonstrate that the redox mechanism revealed in this study can explain the contrasting heterogeneity of natural diamonds on the composition of inclusions, carbon isotopic composition, and nitrogen impurity content.

Abstract

Subduction tectonics imposes an important role in the evolution of the interior of the Earth and its global carbon cycle; however, the mechanism of the mantle–slab interaction remains unclear. Here, we demonstrate the results of high-pressure redox-gradient experiments on the interactions between Mg-Ca-carbonate and metallic iron, modeling the processes at the mantle–slab boundary; thereby, we present mechanisms of diamond formation both ahead of and behind the redox front. It is determined that, at oxidized conditions, a low-temperature Ca-rich carbonate melt is generated. This melt acts as both the carbon source and crystallization medium for diamond, whereas at reduced conditions, diamond crystallizes only from the Fe-C melt. The redox mechanism revealed in this study is used to explain the contrasting heterogeneity of natural diamonds, as seen in the composition of inclusions, carbon isotopic composition, and nitrogen impurity content.

Subduction of crustal material plays an important role in the global carbon cycle (1–6). Depending on oxygen fugacity and pressure-temperature (P-T) conditions, carbon exists in the Earth's interior in the form of carbides, diamond, graphite, hydrocarbons, carbonates, and CO₂ (7–11). In the upper mantle, the oxygen fugacity (fO₂) varies from one to five log units below the fayalite-magnetite-quartz (FMQ) buffer, with a trend of a decrease with depth (6, 12–15). At a depth of ∼250 km, mantle is reported to become metal saturated (16, 17), which holds true for all mantle regions below, including the transition zone and lower mantle. The subduction of the oxidized crustal material occurs to depths greater than 600 km (4–6). The main carbon-bearing minerals of the subducted materials are carbonates, which are thermodynamically stable up to P-T conditions of the lower mantle (10, 11, 18). As evidenced by the compositions of inclusions in diamond, which vary from strongly reduced, e.g., metallic iron and carbides (19–23), to oxidized, e.g., carbonates and CO₂ (6, 20, 24–28), carbonates may be involved in the reactions with reduced deep-seated rocks, including Fe⁰-bearing species (29–31). A scale of these reactions is determined mainly by the capacity of subducted carbonate-bearing domains. An important consequence of such an interaction is that it can produce diamond. However, studies on diamond synthesis via the reactions between oxidized and reduced phases are limited (32–35).

To understand the mechanisms of the interaction of carbon-bearing oxidized- and reduced-mineral assemblages, we performed high-pressure experiments with an iron-carbonate system; an approach was used that enabled the creation of an oxygen fugacity gradient in the capsules (Materials and Methods and SI Materials and Methods).

Results and Discussion

The experimental results and the phase compositions are given in Table 1 and Table S1, respectively. At temperatures of 1,000 and 1,100 °C, the iron–carbonate interaction can be described, in general, by the reaction

Table 1.

Experimental conditions and results

Run no.	P, GPa	T, °C	t, h	Initial composition, mg		Phase association of zones* (from center to periphery of samples)
				Mg0,9Ca0,1CO3	Fe0
1567	6.5	1,350	20	440	507	[Сoh, (Fe-C)L] → [Coh, Mws, Gr] → [Mws, (Carb+Mws)L, Gr] → [Fms, (Carb+Mws)L, Dm, Dm*]
1250	6.5	1,450	20	340	700	[Сoh, (Fe-C)L] → [Coh, Mws, Gr, Dm] → [Mws, (Carb+Mws)L, Gr, Dm] →[(Carb+Mws)L, Mws, Gr, Dm]
1566	6.5	1,550	10	340	535	[Сoh, (Fe-C)L, Dm] → [Coh, Mws, Dm] → [Mws, (Carb+Mws)L, Gr, Dm] → [(Carb+Mws)L, Gr, Dm]
1571	6.5	1,600	8	440	494	[Coh, (Fe-C)L, Dm] → [Coh, Mws, Dm] → [Mws, (Carb+Mws)L, Gr,Dm] → [(Carb+Mws)L, Gr, Dm]
1541	7.5	1,000	60	110	55	Coh → [Coh, Mws, Gr] → [Mws, Marg, Gr] → [Fms, Marg]
1540	7.5	1,100	60	110	53.9	[Mws, Marg, Gr] → [Fms, Marg, Gr]
1532	7.5	1,200	60	110	58	[Mws, (Carb+Mws)L, Gr] → [Fms, (Carb+Mws)L, Mws, Gr, Dm**]
1528	7.5	1,300	60	130	80	[Mws, (Carb+Mws)L, Gr, Dm] → [Fms, (Carb+Mws)L, Mws, Gr, Dm, Dm**]
1525	7.5	1,400	60	130	78	[Mws, (Carb+Mws)L, Gr, Dm] → [Fms, (Carb+Mws)L, Mws, Gr, Dm, Dm**]
1517	7.5	1,450	30	140.1	59.1	[Mws, (Carb+Mws)L, Gr, Dm] → [(Carb+Mws)L, Mws, Fms, Gr, Dm]
1515	7.5	1,550	20	127.2	56.1	[Mws, (Carb+Mws)L, Gr, Dm] → [(Carb+Mws)L, Mws, Fms, Gr, Dm]
1521	7.5	1,650	8	140	75.8	[Mws, (Carb+Mws)L, Dm] → [(Carb+Mws)L, Fms, Dm]
				Mg0,9Ca0,1CO3	Fe3С
1561	7.5	1,000	60	110	63	Coh → [Coh, Mws] → [Coh, Mws, Gr] → [Mws, Arg, Gr] → [Fms, Arg]
1560	7.5	1,100	60	110	57	Coh → [Coh, Mws] → [Coh, Mws, Gr] → [Mws, Arg, Gr] → [Fms, Arg]
1552	7.5	1,200	60	110	67	Coh → [Coh, Mws, Gr] → [Mws, (Carb+Mws)L, Gr] → [Fms, (Carb+Mws)L, Mws, Dm**]
1548	7.5	1,300	60	120	107	Coh → [Coh, Gr, Mws] → [Mws, (Carb+Mws)L, Gr] → [(Carb+Mws)L, Mws, Fms, Gr, Dm**]
1545	7.5	1,400	60	120	110	Coh → [Coh, Gr, Mws] → [Mws, (Carb+Mws)L, Gr] → [Fms, (Carb+Mws)L, Mws, Gr, Dm, Dm**]

^*Compositions of phases from different zones are given in Table S1. Arg, aragonite; (Carb+Mws)_L, carbonate melt with dissolved magnesiowustite; Coh, cohenite (Fe₃C); Dm, diamond; Dm**, diamond growth on seeds; (Fe-C)_L, iron-carbon melt; Fms, ferromagnesite; Gr, graphite; Marg, Mg-aragonite; Mws, magnesiowustite.

Iron carbide (Fe₃C, cohenite) in the above reaction is formed on saturation of the iron with reduced carbon. This iron carbide further reacts with the initial carbonate to produce an association of magnesiowustite + metastable graphite + Ca-rich carbonate. Finally, in the central (reduced) part of the capsules, Fe₃C is produced, which is surrounded by a reaction zone (Fig. 1A), consisting of carbide, graphite, and magnesiowustite. The magnesiowustite exhibits an increase in the Mg number (Mg#) from 0.13 at the contact with carbide to 0.37 at the periphery of the zone. In the outer (oxidized) part of the capsules, an assembly of magnesiowustite (Mg# = 0.38), aragonite (Ca# = 0.89), and graphite (Fig. 2A) is formed. When cohenite was used as the starting material instead of iron, the character of the interaction did not change, but the amount of graphite significantly increased.

Fig. 1.

SEM micrographs: (A) reaction zone (cohenite + magnesiowustite + graphite) (N 1541); (B) magnesiowustite crystals in a dendritic aggregate of quenched carbonate and magnesiowustite (N 1521); (C) lens of metastable graphite in a dendritic aggregate of carbonate and magnesiowustite (N 1517); (D) dendritic aggregate of carbonate and magnesiowustite (N 1532); (E) diamond and magnesiowustite in quenched carbonate melt (N 1515); (F) diamond in polycrystalline aggregates of magnesiowustite, cohenite, and quenched iron-carbon melt (N 1566). Coh, cohenite (Fe₃C); Mws, magnesiowustite; L, carbonate melt with dissolved magnesiowustite; (Fe-C)_L, iron-carbon melt; Arg, aragonite; Gr, graphite; Dm, diamond.

Fig. 2.

SEM micrographs of representative phase assemblages, combined with compositional maps: (A) magnesiowustite + aragonite + graphite at the contact with ferromagnesite (N 1541); (B) dendritic aggregate of carbonates and magnesiowustite (N 1515); (C) ferromagnesite in carbonate melt (N 1517); (D) magnesiowustite and metastable graphite in carbonate melt (N 1250); (E) magnesiowustite, metastable graphite, and diamond in carbonate melt (N 1525); (F) reaction zone (cohenite + magnesiowustite) (N 1250). Fms, ferromagnesite.

At 1,200 °C and higher temperatures, a Ca-rich carbonate melt coexisting with almost Ca-free magnesiowustite and ferromagnesite was present in the system. On quenching, this melt crystallized to dendritic aggregates of Ca,Mg,Fe carbonates and magnesiowustite (Figs. 1 B and D and 2 B–E). A notation [CaCO₃ + (Fe,Mg)O]_L will further be used to designate the Ca-rich carbonate melt with dissolved magnesiowustite. As a result of the redox reaction, metallic iron extracts carbon from carbonate to form carbide. A reaction zone, consisting of high-Fe magnesiowustite, Fe₃C, and graphite is formed around the carbide (Fig. 2F). Cohenite reduces some carbon from the carbonate melt, according to the reaction

Thus, in the center, where carbide and the reaction zone (Mws + Coh + Gr) are present, the fO₂ values correspond approximately to that of the iron-wustite (IW) buffer (about FMQ-5 log units). In the peripheral part of the capsules, where the magnesiowustite + carbonate melt + C⁰ association is formed, fO₂ corresponds to values characteristic of magnesiowustite-poor carbonate melt in equilibrium with graphite or diamond (about FMQ-1 log unit) (14). Therefore, at the initial stages of the interaction, the oxygen fugacity gradient over the capsules is nearly 4 log units. As a consequence, a redox front arises at the iron/carbonate boundary. As the interaction progresses, the fO₂ gradient in the capsules gradually decreases. A reconstructed scheme of the iron-carbonate interaction is shown in Fig. 3. The exchange between the oxidized periphery and reduced center involves a fluid and Ca-rich carbonate melt, capable of dissolving and transporting a significant amount of magnesiowustite. In this process, the melt is partly consumed to form magnesiowustite and elemental carbon. The rate of the iron–carbonate interaction can be estimated by the width of the reaction zone containing magnesiowustite. It was determined that the fronts of the iron–carbonate and carbonate–cohenite reactions propagate at similar rates, which increase from 6.7 to 14.8 µm/h as the temperature increases from 1,000 °C to 1,400 °C, respectively, at 7.5 GPa. The redox interaction ceases when Fe₃C is completely consumed.

Fig. 3.

A scheme of the iron–carbonate interaction, illustrating the mechanism of redox front formation via fO₂ gradient: (A) distribution of phases in a sample; (B) fO₂ profiles in a sample at the initial stage (solid line) and at a certain moment of the redox interaction (dashed line). Arrow denotes direction of the redox front propagation.

The compositions of the final phases and their trends are shown in Table S1 and Fig. 4. The Mg# of magnesiowustite and ferromagnesite, coexisting with the carbonate melt, increases with temperature. An increase in temperature, and thus the degree of partial melting, leads to a decrease in Ca# of the carbonate melt. Analyses of the quench aggregate show that the carbonate melt can dissolve a significant amount of magnesiowustite, as high as 15 wt.% at 1,200 °C and up to 22 wt.% at 1,650 °C.

Fig. 4.

Concentration profiles of FeO, MgO, and CaO across the redox front from the center to periphery of samples: (A) N 1541 (1,000 °C, 7.5 GPa); (B) N 1566 (1,550 °C, 6.5 GPa).

The formation of elemental carbon (graphite and diamond) through the iron–carbonate redox interaction deserves special consideration. In most experiments, as a result of the oxidation of iron carbide, metastable graphite in association with magnesiowustite is formed in the central zone (Fig. 2 D and E). Crystallization of graphite rather than diamond can be accounted for by the absence of the solvent, as well as by the P-T conditions insufficient for the direct graphite-to-diamond transformation. Finally, in most experiments after complete exhaustion of cohenite, a graphite lens (Fig. 1C) with inclusions of carbide in association with magnesiowustite remains in the center of the samples. However, in some experiments at temperatures ≥1,450 °C (Table 1), a quenched metal melt with relatively large (up to 1 mm) diamond crystals (Fig. 5) occurred. The number of diamond nucleation centers is rather small (10⁻¹/mm³), and the growth rate is ∼40 µm/h. Crystallization of these diamonds is associated with the appearance of a metal-carbon melt, which arises due to the existence of the metastable Fe-graphite eutectic (8, 36). At the initial stages of the interaction, the iron is gradually saturated with carbon to form metal-carbon melt and carbide. If graphite is in contact with the metal melt, nucleation of diamond occurs. As iron reacts with carbonate, the metal melt saturates with carbon, and the iron is gradually consumed by oxidation. Thereby, carbon supersaturation is established in the residual melt, which facilitates further growth of the diamond. A similar mechanism possibly operates in the micropools of metal melt, located in the reaction zone containing magnesiowustite. In this case, diamond microcrystals (3–5 μm in size) and their aggregates are formed. Finally, the metal-carbon melt is consumed completely and replaced with diamonds, in association with magnesiowustite. It is important to note that in experiments where Fe₃C was used instead of Fe⁰ as the starting material, no metal-carbon melt was generated, and consequently, no diamonds were seen to form ahead of the redox front.

Fig. 5.

SEM micrographs of crystallized diamonds and their typical infrared spectra: (A) diamond crystal from the metal-carbon melt (N 1566, 6.5 GPa, 1,550 °C); (B) diamond crystal from carbonate melt (N 1521, 7.5 GPa, 1,650 °C); (C) typical infrared absorption spectra of diamond crystals from (1) metal-carbon melt and (2 and 3) carbonate melt. The spectra have been vertically displaced for clarify.

Behind the redox front, in the oxidized conditions, diamond growth on the seeds in the carbonate melt is established at temperature as low as 1,200 °C. At 1,300 °C, spontaneous growth of diamond crystals occurs on the seeds, which demonstrates heterogeneous nucleation. At temperatures ≥1,400 °C, homogeneous nucleation of diamond takes place (Fig. 5). With increasing temperature from 1,200 °C to 1,650 °C, the average diamond growth rate increases from 0.04 to 2 μm/h, respectively. The number of diamond nuclei in the carbonate melt increases from tens to hundreds per cubic millimeter as the temperature increases from 1,400 °C to 1,650 °C. The formation of diamond from the carbonate melt can generally be described by reaction 2. These diamonds grew through the carbon reduction from the carbonate melt, as long as the Fe₃C was present in the capsule. New diamonds nucleate near the interface between the interacting Fe₃C and carbonate melt. When the Fe₃C is completely consumed, redox reaction 2 ceases, but a significant amount of carbon in the form of metastable graphite remains in the central part of the capsule. Therefore, in prolonged experiments, diamond growth continues owing to the transport of carbon dissolved in the carbonate melt, along with the recrystallization of magnesiowustite. Thus, we can conclude that the carbonate melt, generated in the course of the carbonate–iron interaction at temperatures ≥1,200 °C, plays a key role in the formation of diamond, being both a crystallization medium and the source of carbon.

Infrared absorption measurements show that diamond crystals synthesized in the metal-carbon melt contain 100–200 atomic ppm of nitrogen impurity in the form of single substitutional atoms (C-centers; Fig. 5). Inclusions of the quenched Fe-C melt are typical of these crystals. Diamonds, formed in the carbonate melt, exhibit significantly higher nitrogen concentrations ranging from 1,000 to 1,500 ppm. Nitrogen in this case is present mainly as C-centers, but for diamonds crystallized at higher temperatures (>1,500 °C) with considerable run duration (20 h), a significant portion of nitrogen (30–40%) occurs in the aggregated form, as nitrogen pairs or the so-called A-centers. In addition, absorption spectra frequently showed an absorption peak at ∼1,440–1,450 cm⁻¹, which is associated with carbonate inclusions, and a weak peak at 3,107 cm⁻¹, which is due to hydrogen-related defects (Fig. 5). It is interesting to note that, whereas the nitrogen content of ∼200 ppm is typical for most synthetic diamonds produced from metal–carbon systems, concentrations of around 1,000 ppm is normal for natural type Ia diamonds.

Carbon isotope analysis reveals that the starting carbonate has δ¹³C = +0.2 ‰, relative to the Pee Dee Belemnite (PDB) standard. The iron carbide formed in the reduced part of the capsules at 6.5 GPa and temperatures of 1,350 °C, 1,450 °C, and 1,550 °C has δ¹³C of −3.7‰, −5.9‰, and −5.0‰, respectively. The carbonate melt, which is present in the oxidized part of the capsules, is enriched in the heavy carbon isotope and has δ¹³C = +2.1‰ (1,350 °C), +1.7‰ (1,450 °C), and +1.4 ‰ (1,550 °C). Thus, the isotope fractionation, accompanying the formation of iron carbide from the carbon of carbonate, has an average magnitude of 6.5‰.

The interaction of carbonate with Fe⁰ is considered a simplified model of the processes attending the interaction of Fe⁰-saturated peridotites with oxidized subducted crustal material in the deep zones of the Earth. Therefore, we believe that the basic regularities, found in this study, may reasonably be applied to natural, more complex systems. Taking into account the results of previous studies on diamond crystallization in different systems modeling natural media, we can discuss the possible effects of the individual components on the diamond-forming processes. First, it should be noted that the presence of alkalies will significantly decrease the temperature of generation of carbonate melts (37–39). A similar effect will be produced by H₂O-CO₂ fluids (40–42). It was experimentally shown that the addition of these components in the system significantly increases the diamond-forming ability of mantle fluids/melts (43, 44), acting in the oxidized conditions behind the redox front. Under reducing conditions, ahead of the redox front, where the diamond crystallized from Fe-C melt only, the presence of H₂O will lead to the inhibitory effect on the diamond-forming process (45). The addition of sulfur to the Fe-C system significantly decreases the melting temperature (46, 47). The produced melts are two immiscible liquids, with compositions similar to the Fe₃C and FeS. Sulfide melts are the least efficient diamond-forming media compared with carbonate, fluid, and carbonate-silicate-fluid environments (48, 49). The role of sulfides as reducing agents in the interaction with carbonates or CO₂ fluid, leading to the formation of elemental carbon (graphite or diamond), was experimentally demonstrated (35, 50).

The applicability of the redox mechanism is to the existence of deep zones in the Earth, containing native iron or iron carbides, and the possibility of subduction of carbonate minerals to these depths. In fact, there has been a growing body of evidence for the ultradeep subduction of carbonates and their participation in the formation of diamonds. This evidence involves the discovery of inclusions in diamonds, consisting of carbonates in association with the superdeep phases CaSiO₃ or MgO + MgSiO₃, as well as some experimental and geochemical observations (51–54). Our results suggest that the Ca-rich carbonate melt, which forms through the carbonate–iron interaction and can be generated even in the absence of alkalis and H₂O, can be considered as a transmantle interstitial melt. High solubility of magnesiowustite in this carbonate melt produces a relatively low-melting temperature. The minimum temperature of nucleation and growth of diamonds in the carbonate melt, established in this study, is lower than that previously found for dry alkaline-carbonate melts (55).

Another important implication of the proposed redox front model is that it can explain the occurrence of inclusions in natural diamonds, with contrasting fO₂, which are traditionally considered as indicators of differing redox conditions (6, 21). The specifics of nucleation and growth of diamond under the carbonate–Fe⁰ interaction may result in diamond formation at reduced conditions. Findings of central inclusions in diamond, represented by Fe⁰ + Fe₃C + graphite and FeO + graphite assemblages (21), indicate possible relevance of the studied interaction to the natural processes. On the other hand, diamonds, which are thought to originate from the lower part of the transition zone, contain primary alkali-earth carbonate inclusions (e.g., CaCO₃) (52). These diamonds could have formed from a Ca-rich carbonate melt, produced as a result of the subduction of oxidized material. We believe this to be a possible situation in nature, whereby diamonds nucleate in the metal melt and subsequently crystallize in the carbonate melt. In addition, it was determined that magnesiowustite formed in our experiments both at reduced and oxidized conditions, and its composition changes significantly as the redox front propagates. Therefore, we propose that the variations in the Mg/Fe ratio found for magnesiowustite inclusions in diamonds (53, 56) are directly related to the processes investigated in our experimental study.

A noteworthy finding in our study is highly contrasting concentrations of nitrogen impurities, exhibited by diamonds, crystallized at reduced and oxidized conditions in the course of a single redox interaction. This observation demonstrates that the partition coefficients of nitrogen are different for diamond crystallization in metal and carbonate melts and will require additional investigation. Nevertheless, even with the present results, it becomes clear that the changes in the crystallization conditions, brought about by the redox front propagation, can give rise to contrasting zonation in nitrogen and related defects distribution within a single diamond crystal. In general, the nitrogen content increases as the conditions evolve from reducing to oxidizing, with variations being on the order of ∼1,000 ppm.

Our results show that the carbide, formed via the interaction of carbonate with carbon-free iron, is significantly depleted in the heavy-carbon isotope relative to the carbonate source. It is logical that the residual carbonate is enriched with ¹³C. Thus, experimental data showed that nitrogen-poor diamonds grew in the reduced ¹³C-depleted part of capsule, whereas nitrogen-rich diamonds occurred in the oxidized ¹³C-enriched part of capsule. This general tendency of a simultaneous decrease of ¹³C value and N abundance in diamonds fits well the “limit sector” correlation of N abundance and ¹³C values in natural diamonds worldwide (57). One of the possible reasons of such observed interrelations of nitrogen and ¹³C concentrations in natural diamonds can be a segregation of carbonate melt to form a diaper and its subsequent migration (38, 58) that can lead to the formation of contrasting domains of carbon isotopes in the Earth’s low mantle. Taking into account the average magnitude of the fractionation of 6.5‰ and the nitrogen behavior (see above), the redox interaction can be considered as one of the mechanisms responsible for the complex compositional heterogeneity of natural diamonds (59–64).

Most of subducted carbonates have compositions of the MgCO₃-CaCO₃ series. Based on our results with such carbonate systems, we conclude that inclusions of Ca-rich carbonates, especially in association with magnesiowustite and other mantle minerals, are likely the products of the mechanism deduced from our experimentation.

Materials and Methods

Experiments were performed at pressures of 7.5 and 6.5 GPa, at temperatures of 1,000–1,650 °C and 1,350–1,600 °C, respectively, and with durations from 8 to 60 h, using a split-sphere multianvil apparatus (65). Details on the P-T calibration and accuracy of the measurements are given in ref. 66. Pt capsules of relatively large volume (6 and 10 mm in diameter at 7.5 and 6.5 GPa, respectively, 3.5–4.0 mm long) were used to enable a detailed study of the effects associated with the redox front propagation. The starting materials were mixtures of natural magnesite and dolomite, with bulk composition of (Mg_0.9Ca_0.1)CO₃, powdered Fe⁰ (99.999%), and presynthesized Fe₃C. A pellet of pressed iron or cohenite was placed into a carbonate container, which was then loaded into Pt capsules. This sandwich-type assembly of the reagents provided an fO₂ gradient over the samples and prevented the reaction between the metallic iron and Pt (Fig. S1). After runs, samples from different parts of capsules were studied using X-ray and electron microprobe analyses, optical and scanning electron microscopy, and Raman and infrared spectroscopy. Phase identification of run products was performed by X-ray diffraction (a DRON-3 diffractometer) and Raman spectroscopy. Raman spectra were measured using a Horiba J.Y. LabRAM HR800 spectrometer, with an Ar-ion laser (514 nm). An investigation of phase relations and measurements of energy dispersive spectra (EDS) of various phases were performed using a Tescan MIRA3 LMU SEM. The morphology of microdiamond crystals was studied using a Tescan MIRA3 LMU SEM and an Olympus BX51 optical microscope. Infrared absorption spectra of diamonds were measured using a Bruker Vertex 70 FTIR spectrometer fitted with a Hyperion 2000 IR microscope. The isotopic analysis of carbon was performed using an isotope-ratio MAT-Delta mass spectrometer. Chemical compositions of synthetic phases were investigated using a Cameca Camebax-Micro microprobe. For electron microprobe analysis, polished sections of the samples were prepared. Mineral phases were analyzed with a focused electron beam of 1 μm diameter. Compositions of the quenched melt were defined using a defocused beam of 20–30 μm in diameter.