Decarbonation Reactions Involving Ankerite and Dolomite under upper Mantle P,T-Parameters: Experimental Modeling

: An experimental study aimed at the modeling of dolomite- and ankerite-involving decarbonation reactions, resulting in the CO 2 ﬂuid release and crystallization of Ca, Mg, Fe garnets, was carried out at a wide range of pressures and temperatures of the upper mantle. Experiments were performed using a multi-anvil high-pressure apparatus of a “split-sphere” type, in CaMg(CO 3 ) 2 -Al 2 O 3 -SiO 2 and Ca(Mg,Fe)(CO 3 ) 2 -Al 2 O 3 -SiO 2 systems (pressures of 3.0, 6.3 and 7.5 GPa, temperature range of 950–1550 ◦ C, hematite bu ﬀ ered high-pressure cell). It was experimentally shown that decarbonation in the dolomite-bearing system occurred at 1100 ± 20 ◦ C (3.0 GPa), 1320 ± 20 ◦ C (6.3 GPa), and 1450 ± 20 ◦ C (7.5 GPa). As demonstrated by mass spectrometry, the ﬂuid composition was pure CO 2 . Composition of synthesized garnet was Prp 83 Grs 17 , with main Raman spectroscopic modes at 368–369, 559–562, and 912–920 cm − 1 . Decarbonation reactions in the ankerite-bearing system were realized at 1000 ± 20 ◦ C (3.0 GPa), 1250 ± 20 ◦ C (6.3 GPa), and 1400 ± 20 ◦ C (7.5 GPa). As a result, the garnet of Grs 25 Alm 40 Prp 35 composition with main Raman peaks at 349–350, 552, and 906–907 cm − 1 was crystallized. It has been experimentally shown that, in the Earth’s mantle, dolomite and ankerite enter decarbonation reactions to form Ca, Mg, Fe garnet + CO 2 assemblage at temperatures ~175–500 ◦ C lower than CaCO 3 does at constant pressures. ﬂuid and garnet solid solutions [20]; a —activity of component, Ms—magnesite, Ky—kyanite, Co—coesite, Prp—pyrope, Grs—grossular. curves, resulting in the formation of CO 2 fluid and garnet solid solutions [20]; a —activity of component, Ms—magnesite, Ky—kyanite, Co—coesite, Prp—pyrope, Grs—grossular.


Materials and Methods
Experimental modeling of decarbonation reactions involving ankerite and dolomite was performed in the Ca(Mg,Fe)(СО3)2-SiO2-Al2O3 and CaMg(СО3)2-SiO2-Al2O3 systems using a multi-anvil high-pressure split-sphere apparatus (BARS) [49]. The experiments were carried out at pressures of 3.0, 6.3, and 7.5 GPa, in the temperature range of 950-1550 °C and for durations from 15 min to 60 h. Methodological features of the assembly, the design of the high-pressure cell, as well as data on the pressure and temperature calibration have been published previously [21,[50][51][52]. As initial reagents, we used natural ankerite (Ca(Mg,Fe)(СО3)2 (Mésage Mine, Saint-Pierre-de-Mésage, France) and dolomite (CaMg(СО3)2 (Satka, Urals, Russian Federation), as well as synthetic SiO2 and Al2O3 with a purity of 99.99%. Raman spectra of these natural carbonates are shown in Figure 5. The molar proportions of the starting materials ensure garnet and CO2 fluid formation when the interaction is complete ( Table 1). The starting reagents were ground and thoroughly homogenized. Taking into account previous experience of the studies in carbonate-oxide systems under high P and T [53][54][55], Pt was chosen as the capsule material.  Thus, it seems relevant to perform experimental modeling of decarbonation reactions involving natural Mg,Ca carbonates (dolomite, ankerite), and associated with the formation of garnets and CO 2 fluid, and to determine the position of the corresponding decarbonation curves at a wide range of pressures and temperatures of the upper mantle.

Materials and Methods
Experimental modeling of decarbonation reactions involving ankerite and dolomite was performed in the Ca(Mg,Fe)(CO 3 ) 2 -SiO 2 -Al 2 O 3 and CaMg(CO 3 ) 2 -SiO 2 -Al 2 O 3 systems using a multi-anvil high-pressure split-sphere apparatus (BARS) [49]. The experiments were carried out at pressures of 3.0, 6.3, and 7.5 GPa, in the temperature range of 950-1550 • C and for durations from 15 min to 60 h. Methodological features of the assembly, the design of the high-pressure cell, as well as data on the pressure and temperature calibration have been published previously [21,[50][51][52]. As initial reagents, we used natural ankerite (Ca(Mg,Fe)(CO 3 ) 2 (Mésage Mine, Saint-Pierre-de-Mésage, France) and dolomite (CaMg(CO 3 ) 2 (Satka, Urals, Russian Federation), as well as synthetic SiO 2 and Al 2 O 3 with a purity of 99.99%. Raman spectra of these natural carbonates are shown in Figure 5. The molar proportions of the starting materials ensure garnet and CO 2 fluid formation when the interaction is complete ( Table 1). The starting reagents were ground and thoroughly homogenized. Taking into account previous experience of the studies in carbonate-oxide systems under high P and T [53][54][55], Pt was chosen as the capsule material.

Materials and Methods
Experimental modeling of decarbonation reactions involving ankerite and dolomite was performed in the Ca(Mg,Fe)(СО3)2-SiO2-Al2O3 and CaMg(СО3)2-SiO2-Al2O3 systems using a multi-anvil high-pressure split-sphere apparatus (BARS) [49]. The experiments were carried out at pressures of 3.0, 6.3, and 7.5 GPa, in the temperature range of 950-1550 °C and for durations from 15 min to 60 h. Methodological features of the assembly, the design of the high-pressure cell, as well as data on the pressure and temperature calibration have been published previously [21,[50][51][52]. As initial reagents, we used natural ankerite (Ca(Mg,Fe)(СО3)2 (Mésage Mine, Saint-Pierre-de-Mésage, France) and dolomite (CaMg(СО3)2 (Satka, Urals, Russian Federation), as well as synthetic SiO2 and Al2O3 with a purity of 99.99%. Raman spectra of these natural carbonates are shown in Figure 5. The molar proportions of the starting materials ensure garnet and CO2 fluid formation when the interaction is complete ( Table 1). The starting reagents were ground and thoroughly homogenized. Taking into account previous experience of the studies in carbonate-oxide systems under high P and T [53][54][55], Pt was chosen as the capsule material.   The volume of reaction capsules was selected to ensure that all necessary analytical methods could be employed, taking into account the size of the high-pressure cell. The internal diameter of the Pt capsules for experiments at 3.0 and 6.3 GPa was 1.5 mm at a length of 6 mm, and at 7.5 GPa, 1.5 mm at a length of 4 mm.
There is a problem of hydrogen diffusion into the reaction volume through the walls of capsules in the high-temperature high-pressure experimental studies [56,57]. Hydrogen diffusion can significantly decrease the oxygen fugacity in capsules, change the composition of the fluid in the reaction volume, and lead to a shift in the decarbonation curves in the P,T-field ( Figure 6a). To prevent the effect of hydrogen diffusion on the course of the experiment, in this study, we used a specially designed high-pressure cell with a hematite buffer container [51]. The effective time of this buffer at temperatures below 1200 • C is more than 150 h, and at 1500 • C, it is around 5 h. After the experiments, control studies of the chemical composition of the hematite buffer container were carried out. In all cases, the material of the buffer container was analyzed; it was represented by hematite and magnetite, which indicates the effectiveness of the hematite container throughout the entire experiment. The duration of the experiments for each temperature was selected based on the time of effective action of the hematite container. The experimental temperatures were selected according to the calculations ( Figure 6b) [58,59]. The volume of reaction capsules was selected to ensure that all necessary analytical methods could be employed, taking into account the size of the high-pressure cell. The internal diameter of the Pt capsules for experiments at 3.0 and 6.3 GPa was 1.5 mm at a length of 6 mm, and at 7.5 GPa, 1.5 mm at a length of 4 mm.
There is a problem of hydrogen diffusion into the reaction volume through the walls of capsules in the high-temperature high-pressure experimental studies [56,57]. Hydrogen diffusion can significantly decrease the oxygen fugacity in capsules, change the composition of the fluid in the reaction volume, and lead to a shift in the decarbonation curves in the P,T-field ( Figure 6а). To prevent the effect of hydrogen diffusion on the course of the experiment, in this study, we used a specially designed high-pressure cell with a hematite buffer container [51]. The effective time of this buffer at temperatures below 1200 °C is more than 150 h, and at 1500 °C, it is around 5 h. After the experiments, control studies of the chemical composition of the hematite buffer container were carried out. In all cases, the material of the buffer container was analyzed; it was represented by hematite and magnetite, which indicates the effectiveness of the hematite container throughout the entire experiment. The duration of the experiments for each temperature was selected based on the time of effective action of the hematite container. The experimental temperatures were selected according to the calculations ( Figure 6b) [58,59].
The phase and chemical compositions of the final samples were determined by energy dispersive spectroscopy (Tescan MIRA3 LMU scanning electron microscope, TESCAN, Brno, Czech Republic) and microprobe analysis (Camebax-micro analyzer, CAMECA, Gennevilliers, France). Standards used for the analyses of garnet were pyrope (for SiO 2 and Al 2 O 3 contents), ferrous spessartine (for FeO and MnO), and diopside (for MgO and CaO). Silicate, carbonate, and oxide mineral phases were analyzed at an accelerating voltage of 20 kV, a probe current of 20 nA, a counting time of 10 s on each analytical line, and an electron beam diameter of 2-4 µm. Phase relationships in the samples were studied by means of scanning electron microscopy (SEM).
At the preliminary stage of the experiments, we calculated the theoretical positions of the decarbonation curves (Figure 6b) according to data given in [20] on carbonation curves for grossular, pyrope, and "diluted" grossular. This technique implied ideal mixing for the garnet solid solution. The lines obtained as a result would be true for the complete decarbonation (all ankerite or dolomite is decomposed).
The structural features of the obtained garnet and carbonate were studied by Raman spectroscopy (Jobin Yvon LabRAM HR800 spectrometer equipped with an Olympus BX41 stereo microscope, Horiba Jobin Yvon S.A.S., Lonjumeau, France). An He-Cd laser with a wavelength of 325 nm (Laser Quantum, Stockport, UK) was used as an excitation source. To monitor the effectiveness of the hematite buffer, the composition of the fluid phase was identified by mass spectrometry. For this, the platinum capsule after the experiment was placed in a vacuum device connected to a sample injection system in a Delta V Advantage mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) and equipped with a special mechanism for puncturing samples. After preliminary pumping out of the device with the sample to a pressure of 2.7 × 10 −2 mbar, guaranteeing the absence of atmospheric gases in the device, the capsule was punctured and the gas released at room temperature was let into the mass spectrometer analyzer. Analytical studies were performed at the Sobolev Institute of Geology and Mineralogy, SB RAS, and at the Analytical Center for multi-elemental and isotope research, SB RAS.

Results
The experimental parameters and the results obtained are presented in Table 2. Taking into account the previously developed approach [20], the main criterion for the decarbonation reaction to occur is the appearance of garnet and CO 2 fluid in the reaction volume. The formation of these phases was accompanied by a decrease in the amount of ankerite or dolomite and oxides in the reaction volume (Figure 7a,b). The partial preservation of carbonate and oxides in the samples is a consequence of the incomplete passage of the decarbonation reaction during the effective time of the hematite buffer. At temperatures above the onset of the decarbonation reaction (Figures 7c-f and 8a), grossular-pyrope-almandine and kyanite were formed in the samples. The compositions of the obtained mineral phases are shown in Table 3. The resulting CO 2 fluid segregated and formed cavities in the entire volume of the samples; the size and shape of the cavities depends on the temperature and, accordingly, the depth of the decarbonation reactions. At the initial stage of the system decarbonation, the fluid was in the interstitial space, and with increasing temperature, it formed rounded cavities with sizes from 10 to 150 µm.
Minerals 2020, 10, x FOR PEER REVIEW 7 of 17 obtained mineral phases are shown in Table 3. The resulting CO2 fluid segregated and formed cavities in the entire volume of the samples; the size and shape of the cavities depends on the temperature and, accordingly, the depth of the decarbonation reactions. At the initial stage of the system decarbonation, the fluid was in the interstitial space, and with increasing temperature, it formed rounded cavities with sizes from 10 to 150 μm.    Table 3. The resulting CO2 fluid segregated and formed cavities in the entire volume of the samples; the size and shape of the cavities depends on the temperature and, accordingly, the depth of the decarbonation reactions. At the initial stage of the system decarbonation, the fluid was in the interstitial space, and with increasing temperature, it formed rounded cavities with sizes from 10 to 150 μm.   It was found that decarbonation in the Ca(Mg,Fe)(CO 3 ) 2 -SiO 2 -Al 2 O 3 system occurred at 1000 ± 20 • C (3.0 GPa), 1250 ± 20 • C (6.3 GPa), and 1400 ± 20 • C (7.5 GPa) (Figure 8a). When studying crystals of synthesized garnet by Raman spectroscopy, it was found that the main modes for them were at 349-350 cm −1 (librational R(SiO 4 ) 4− ), 552 cm −1 (internal bending (Si-O) bend , υ 2 ), and 906-907 cm −1 (stretching (Si-O) str , υ 1 ) (Figure 9a, Table 3 (Table 4).

Discussion
A detailed study of the phase and chemical composition of the obtained samples, as well as the characteristic structures of the zonal aggregates, makes it possible to reconstruct decarbonation processes in the carbonate-oxide systems CaMg(CO 3 ) 2 -Al 2 O 3 -SiO 2 and Ca(Mg,Fe)(CO 3 ) 2 -Al 2 O 3 -SiO 2 . It was experimentally demonstrated that at temperatures below the onset of decarbonation reactions in the reaction volume, an interaction between coesite and corundum occurs, with the formation of kyanite at the contact of these phases. A similar process was previously established in the carbonate-oxide-sulfide system (MgCO 3 -SiO 2 -Al 2 O 3 -FeS system, P = 6.3 GPa, T = 1250-1450 • C) [11]. At the same time, it must be emphasized that both ankerite and dolomite under P,T parameters below the decarbonation reactions in the samples are stable and do not undergo breakdown or phase transitions, which is confirmed by the results of Raman spectroscopy and mass spectrometry and is also consistent with modern experimental data ( Figure 2) [7].
At temperatures above the onset of decarbonation, a number of processes occur in the reaction volume: (1) crystallization of kyanite, (2) interactions of kyanite + coesite + carbonate and corundum + coesite + carbonate, leading to crystallization of garnet and CO 2 fluid release, and (3) recrystallization of the starting carbonates and change of their compositions. It was established that the proportions of Ca, Mg, and Fe in the synthesized garnets differ from the initial carbonates. It should be noted that the methodical approach with hematite buffer limits the duration of the experiments; therefore, only partial rather than complete decarbonation is realized in the samples. In the case of complete decarbonation, the molar ratio of divalent cations in the initial carbonate and newly formed garnet would be completely identical. However, it is the established features of partially realized decarbonation reactions that make it possible to reconstruct natural processes. The main regularities were as follows: a decrease in Ca# and an increase in Fe# in garnets relative to the proportions in the initial carbonates; a corresponding increase in Ca# and a decrease in Fe# in carbonates after the experiments, relative to the initial ones ( Figure 11). diluting pyrope with the almandine component causes temperature downshift of decarbonation reactions, while adding a grossular component, on the contrary, increases the reaction temperature ( Figure 4). Correspondingly, CO2 fluid and garnets with decreased Ca# and increased Fe# obtained in this work as a result of partial decarbonation of dolomite or ankerite-bearing assemblages are formed at temperatures lower than those necessary for complete decarbonation. This process is accompanied by a specific change in the composition of dolomite or ankerite (an increase in Ca# and a decrease in Fe# relative to the initial ones), expanding the stability field of these carbonates without significantly changing their structures. Despite the fact that, in the present study, we performed experiments on the elucidating of positions of decarbonation curves in dry systems, we find it useful to discuss the principal differences between our results and existing ideas on realistic mantle conditions where aqueous fluids are present. In our previous research [16,21] on the formation of diamond and graphite via the carbonate-silicate interaction in MgCO3-SiO2, CaMg(CO3)2-SiO2 and MgCO3-SiO2-Al2O3 systems under mantle P,T conditions, the source of hydrogen was used. According to gas chromatographic data, fluid composition during diamond crystallization varied from almost pure CO2 to aqueous, and it strongly affected the decarbonation parameters. For example, in the MgCO3-SiO2 system at 1500 °C and 6 GPa, in the case of H2O-CO2 fluid presence in the reaction volume, the formation of enstatite was recorded, i.e., decarbonation occurred; with the same parameters but in a "dry" system, the final assemblage was magnesite and coesite, i.e., no decarbonation processes took place. In the MgCO3-Al2O3-SiO2 system at 6 GPa and 1500 °C, the formation of garnet occurred regardless of the fluid composition. However, when H2O-component dominated in fluid, the amount of pyrope increases sharply, which can be explained both by the additional decomposition of magnesite, not only due to the thermodynamic factor but also due to the redox factor, and by an increase in kinetics due to the formation of an aqueous fluid. Another important result obtained is evidence of poor wettability of carbon dioxide under mantle P,T parameters, obtained from the specific structural features of experimental samples. As shown in Figures 7c,e,f and 10c,f, fluid segregates and forms rounded bubbles, which allows us to determine the high wetting angle; this is in good agreement According to the existing theoretical and experimental data on the decarbonation parameters of Fe, Ca, Mg carbonate-oxide systems, the lowest temperatures are required for the formation of ferrous garnets and the highest for calcium ones [20,58,59]. Specifically, it was demonstrated that diluting pyrope with the almandine component causes temperature downshift of decarbonation reactions, while adding a grossular component, on the contrary, increases the reaction temperature ( Figure 4). Correspondingly, CO 2 fluid and garnets with decreased Ca# and increased Fe# obtained in this work as a result of partial decarbonation of dolomite or ankerite-bearing assemblages are formed at temperatures lower than those necessary for complete decarbonation. This process is accompanied by a specific change in the composition of dolomite or ankerite (an increase in Ca# and a decrease in Fe# relative to the initial ones), expanding the stability field of these carbonates without significantly changing their structures.
Despite the fact that, in the present study, we performed experiments on the elucidating of positions of decarbonation curves in dry systems, we find it useful to discuss the principal differences between our results and existing ideas on realistic mantle conditions where aqueous fluids are present. In our previous research [16,21] on the formation of diamond and graphite via the carbonate-silicate interaction in MgCO 3 -SiO 2 , CaMg(CO 3 ) 2 -SiO 2 and MgCO 3 -SiO 2 -Al 2 O 3 systems under mantle P,T conditions, the source of hydrogen was used. According to gas chromatographic data, fluid composition during diamond crystallization varied from almost pure CO 2 to aqueous, and it strongly affected the decarbonation parameters. For example, in the MgCO 3 -SiO 2 system at 1500 • C and 6 GPa, in the case of H 2 O-CO 2 fluid presence in the reaction volume, the formation of enstatite was recorded, i.e., decarbonation occurred; with the same parameters but in a "dry" system, the final assemblage was magnesite and coesite, i.e., no decarbonation processes took place. In the MgCO 3 -Al 2 O 3 -SiO 2 system at 6 GPa and 1500 • C, the formation of garnet occurred regardless of the fluid composition. However, when H 2 O-component dominated in fluid, the amount of pyrope increases sharply, which can be explained both by the additional decomposition of magnesite, not only due to the thermodynamic factor but also due to the redox factor, and by an increase in kinetics due to the formation of an aqueous fluid. Another important result obtained is evidence of poor wettability of carbon dioxide under mantle P,T parameters, obtained from the specific structural features of experimental samples. As shown in Figure 7c,e,f and Figure 10c,f, fluid segregates and forms rounded bubbles, which allows us to determine the high wetting angle; this is in good agreement with studies of CO 2 fluid wetting properties under crustal conditions [63,64]. Thus, implicating the results both of our present and previous [16,21] studies on the mantle processes, we can state that the presence of water-bearing fluids will reduce the temperatures of garnet + CO 2 fluid formation and overall increase the intensity of the decarbonation process. The formation of H 2 O-CO 2 fluid under these conditions results in the better wettability of mantle rocks with fluid, its better migration, and wider distribution of this mantle metasomatic agent.
As a natural media simulated in this paper, one can consider carbonated eclogite. The stability field of garnet + clinopyroxene + kyanite + coesite + carbonate assemblage relatively to garnet + clinopyroxene + CO 2 depends on garnet and clinopyroxene carbonation/decarbonation P,T parameters. Processes of carbonation/decarbonation also influence the mantle fluid regime and global cycle of carbon. Results obtained in the present study show that an increase in the Ca component of carbonate expands its stability field, while increasing the Fe component has the opposite effect, or, in other words, Fe-rich garnets have a larger stability field than Ca-rich ones. Thus, it can be suggested that the association of group I eclogites is more stable in processes of CO 2 metasomatism than that of group II, as group I garnets are richer in iron and lower in calcium [65]. Carbonation of garnet, however, occurs under higher pressures than that of pyroxenes (reactions (2-4)), which should be taken into account (Knoche et al., 1999) [20]. Another important aspect of this work is the use of a hematite buffer. This allows us to study a non-ultrareduced dry system and observe the decarbonation reaction without formation of carbonatite-like melts, leaving, however, a possibility for further research into water-bearing systems.
The currently available information on inclusions of Ca, Mg, Fe carbonates in natural diamonds, as well as on the composition of garnets from carbonated eclogites, indicates that the data obtained in this work can be fully applicable in the reconstruction of decarbonation processes in the Earth's mantle. In particular, diamonds of pipes of Mwadui, Tansania [31], Juina, Brasil [66], and Cancan, Guinea [67] contain dolomite inclusions (Ca#~0.5) and siderite (63.5 wt. % FeO). Ankerite microinclusions are also described in a number of works [68,69]. The composition of garnets from eclogite xenoliths varies over a wide range of Mg# 0.46-0.89, Ca# 0.02-0.42, with an average content of Fe# 0.30 [66,70]. Thus, the results of the present experimental study on the estimation of CO 2 and double carbonates' stability fields in the Earth's mantle can be used in terms of constraints of diamond-forming processes, with a CO 2 as a carbon source for diamond crystallization [71][72][73].
When comparing the obtained data on the position of decarbonation curves with experimental and calculated results [20], it was found that decarbonation of the CaMg(CO 3 ) 2