INTRODUCTION

The behavior of carbonates in the subduction zone controls a carbon redistribution between deep-seated reservoirs and the surface [1, 2]. According to modern estimates, about 80 million tons of carbon in the form of carbonate minerals is annually supplied to the mantle in subduction zones in metasediments (pelites) [1]. Only 30% of the subducted carbon returned to the surface. Nevertheless, subduction of carbonate-rich sediments provides maximum volumes of isotopically heavy CO2 emissions from island-arc volcanic systems [1].

As subduction proceeds, volatile-rich carbonate-bearing pelites located in the upper hottest part of the subducting plates can undergo melting [36]. At pressures up to the second critical end point in the volatile-rich pelite system, the solidus is controlled by the peritectic reaction of phengite + clinopyroxene + coesite + water fluid = garnet + melt, which provides for the formation of K-rich granite-like melts [46]. According to modern estimates, melting of carbonate-bearing pelites to depths of 200 km is considered unlikely [78]. Under the redox conditions typical of subduction zones, carbonates remain stable and do not react with silicates [2, 69]. However, they can be actively dissolved in water-rich fluids, which are formed due to dehydration of both pelites and the underlying serpentinized peridotite slab [8, 9]. The solubility of CO2 in supercritical fluid is at the level of 5 wt % or somewhat higher [9]. Moreover, according to the model experiment data, the CaCO3 dissolution in water fluid can essentially increase in the presence of chlorides [10, 11]. The behavior of carbonates changes fundamentally when reaching a depth of 200–250 km. At these levels, due to the intersection of slab geotherms with the solidus of carbonate-bearing pelites alkaline low-viscosity carbonate melts are formed [2, 7].

An urgent issue of the deep carbon cycle is to determine the PT conditions and fluid regime under which carbonates can remain stable in pelites in the subduction zone up to 200–250 km depths. We have experimentally studied the stability of carbonates in natural carbonate and Cl-bearing pelite at different regimes of defluidization.

MATERIALS AND METHODS

For our experiments, we used deep marine sediment (pelite) of the Maykop Shale (Taman Peninsula, Russia) (Table 1) [12]. The pelite composition (wt %) was muscovite (52), quartz (20), illite (15), albite (5), kaolinite (5), calcite (2), and siderite (1.7). According to the thermogravimetric analysis, it contains 1.87 wt % CO2 and 5.4 wt % H2O. The content of Cl ions in pelite is 0.1 wt %. The contents of chlorides were measured by the turbidimetric method. The method sensitivity is 1 µg of chloride in a sample. The total measurement error of the chloride ion with a level of confidence P = 0.95 is 15%.

Table 1. Comparison of the compositions of the Maykop pelite and average subduction sediment GLOSS-II [3]

Finely ground pelite powder was placed in Pt or Au capsules (6 or 10 mm in diameter and 0.2 mm wall thickness). The weight of the samples ranged from 120 to 500 mg. The diamond trap method was used to separate the fluid from the pelitic sample under high P–T parameters [13].

For this purpose, a layer (about 30% of the sample weight) of synthetic 14–20 µm diamond microcrystals was added to the ampoule. The amounts of gas and liquid products of dehydration and decarbonation in pelite samples (quenched fluid) after the experiments were determined by weighing the capsules before and after. After opening, the capsules before weighing were dried for 24 hours at 100°C.

Two series of experiments with two different scenarios of defluidization were carried out. The first scenario used the initial, unaltered pelite sample. The second scenario involved stage-by-stage defluidization of samples at 3.0 → 5.5 → 7.8 GPa. Beginning from the second stage, each subsequent experiment of the series simulating the subduction with increasing P–T parameters was performed using a partially defluidized sample (dried and without trapped matter) from the previous experiment. It was impossible to conduct experiments with stage-by-stage defluidization of samples without applying the diamond trap method. Since the quenching products of highly concentrated fluids would remain in the intergranular space of the samples, the use of traps made it possible to separate almost all fluid detached from the samples during dehydration and decarbonation.

Experiments lasting 40 hours were carried out on a BARS multi-anvil apparatus [14]. The pressures and temperatures were measured with an accuracy of ±0.1 GPa and ±20°C, respectively [15]. Redox reactions involving Fe3+, as well as dehydration and decarbonation reactions of pelite under P–T parameters of the experiments, provided the formation of a fluid phase containing H2O and CO2 in the samples. The excess of diamond controlled fO2 in the samples near the CCO equilibrium values. The compositions of the phases obtained were investigated using a Tescan MIRA 3 LMU scanning electron microscope equipped with an INCA Energy 450 (Oxford Instruments) microanalysis system and a Jeol JXA-8100 electron microprobe.

RESULTS AND DISCUSSION

The experiment with an initial volatile-rich pelite was carried out under 3.0 GPa. Already at 3.0 GPa and 750°C, pelite was converted into an eclogite-like mineral association including garnet, coesite, phengitic muscovite (muscovite with a high content of the seladonite), omphacite clinopyroxene, kyanite, and pyrrhotine, as well as accessory phases such as monazite, zircon, rutile, and carbonate (Fig. 1a). The solidus of the studied pelite is located between 750 and 900°С at 3.0 GPa (Table 2). Under 3.0 GPa and 900°C, the peritectic reaction of phengite + clinopyroxene + coe-site + fluid = melt + garnet leads to the formation of 38–49 wt % of the granite-like melt enriched in SiO2, Al2O3, and K2O. The formula of carbonate mineral is \({\text{C}}{{{\text{a}}}_{{0.02-0.03}}}{\text{M}}{{{\text{g}}}_{{0.49-0.63}}}{\text{F}}{{{\text{e}}}_{{0.33-0.42}}}{\text{M}}{{{\text{n}}}_{{0.01-0.06}}}{\text{C}}{{{\text{O}}}_{3}}\).

Fig. 1.
figure 1

SEM images of the samples obtained in experiments with the initial and preliminary defluidized pelite. (а) Sample Р1 (initial pelite, 3.0 GPа and 750°C); (b) sample Р5 (initial pelite, 7.8 GPa and 940°С); (c) sample Р1-2 (defluidized pelite, 5.5 GPa and 850°С); (d) sample of salt aggregate on the ampoule surface, formed during the drying of quenched fluid, experiment Р1 (initial pelite, 3.0 GPa and 750°С). For the phase abbreviations, see Table 2. EDS-SEM data on the salt compositions are given in Table 3. Scale bar in µm.

Table 2. Parameters of 40-hour experiments with pelite, the fraction of quenched fluid in samples, and phase composition of samples

The first-series experiments with the initial pelite at pressures of 5.5 and 7.8 GPa and temperatures of 850 and 1030°C led to the formation of a loose aggregate of solid phases (Fig. 1b) and 10–20 wt % of fluid, which was captured by a diamond trap. In the obtained samples, the compositions of solid phases acquire characteristic features of eclogite minerals from ultra high pressure metamorphic complexes. Garnets contain from 38 to 43 mol % of the pyrope endmember and have high contents of Na, Ti, and P admixtures. In clinopyroxenes, the jadeite endmember fraction increases from 66 to 92 mol %. Phengitic muscovite is fixed in equilibrium with both the melt and the fluid. Its temperature stability boundary in the range of 3.0–7.8 GPa is approximately 50°C higher than that suggested in [4, 6]. As the P–T parameters increase, the Si + Mg content in phengitic muscovite increases from 3.8 to 4.4 p.f.u. and the Al content decreases from 2.1 to 1.6 p.f.u. (calculated to 11 atoms of oxygen). Pyrrhotine remains stable in equilibrium with the fluid. None of the samples obtained at 5.5–7.8 GPa contained carbonate.

At the second series of experiments with preliminary defluidized pelite (5.5 and 7.8 GPa), the phase composition of the samples is similar to that obtained in the first series (Table 2, Fig. 1c). Due to the low fluid content in the samples, their structure is denser with a smaller number of voids in the intergranular space. Nevertheless, many phases show traces of recrystallization. However, the main difference is the presence of carbonate Ca0.01–0.03Mg0.51–0.60Fe0.37–0.45Mn0.01CO3 in the pelite sample, actually of the same composition as in samples obtained at 3.0 GPa.

When opening the capsules after the first-series experiments with the initial pelite, water and carbon dioxide are intensively released. The water contained a significant amount of dissolved salts, primarily potassium and sodium chlorides. Figure 1d shows the salt aggregate that formed on the surface of the capsule after drying of the quenched fluid from experiment P1 (P = 3.0 GPa and T = 750°C).

The compositions of salts from this aggregate are given in Table 3 (low sums of some analyses are due to high porosity of a part of the aggregate). In the diamond trap after the experiments, a significant amount of quenching fluid products was fixed. When opening the ampoules with defluidized samples, water and carbon dioxide were not released, but a loss of mass due to the drying the ampoules was established (Table 2).

Table 3. The composition of salts from quenched fluid (wt %), detached after opening the ampoule after the experiment Р1, Р = 3.0 GPa, and Т = 750°С (drying after opening lasts 24 hours)

The amount of H2O + CO2 released during the opening of the capsules is given in Table 2. In the initial pelite samples, the amount of dehydration and decarbonation products increased with increasing the P–T parameters from 4.9 to 6.3–6.5 wt %. It is interesting that after preliminary defluidization of pelite at 3.0 GPa and 750°С with a loss of 4.9 wt % of H2O + CO2, the experiment at 5.5 GPa and 850°С led to the formation of 1.0 wt % of H2O + CO2. Using a twice defluidized pelite at 7.8 GPa and 940°C provided 0.2  wt % of H2O + CO2. According to EDS-SEM data, the quenched fluid from the preliminarily defluidized samples did not contain chlorine.

In general, during transformation of the initial pelite under increasing P–T parameters along the middle and hot subduction geotherms, it turns into an association of mineral phases characteristic of eclogite from ultrahigh-pressure metamorphic complexes. In garnet, the pyrope endmember fraction increases; in clinopyroxene, the jadeite fraction. Muscovite was subject to progressive phengitization.

In addition, fluid or melt forms in the samples (only at 3.0 GPa and 900°C) forms in the samples. Under the minimum pressure of 3.0 GPa and temperature of 750–900°C Mg-Fe-carbonate remains stable in the pelite. This composition of carbonate is due to the low Ca content in the initial pelite.

The amount of H2O + CO2 formed due to dehydration and decarbonatization at 3.0 GPa and 750°C is less (4.9 wt %) than at 5.5 GPa and 850°C (6.3 wt %) and also at 7.8 GPa and 940°C (6.5 wt %). Thus, about 25 wt % carbonate of the initial pelite should have been preserved in the sample after the experiment at 3.0 GPa and 750°C. Since the initial pelite contained 1.87 wt % CO2 in the form of carbonates, the complete dissolution of carbonate in the fluid provides 29–30 wt % of CO2 in quenched fluid (molar ratio CO2/(CO2 + H2O) = 0.15). It means that the solubility of carbonate in aqueous chloride fluid, stable in pelite at 5.5–7.8 GPa and 840–1030°C is noticeably higher than the estimates available [8, 10, 11].

During three successive stages of defluidization, almost all volatiles were removed from the pelite. However, the carbonate was not completely dissolved in the fluid phase, as it happened in the experiments with the initial pelite at the same P–T parameters. The behavior of chlorine turned out to be an important factor of carbonate stability. Since there are no solid Cl hosts in the pelite, the fluid formed at the first stage (3.0 GPa and 750°C) dissolves and concentrates all the chlorine contained in the pelite (0.1 wt %). According to the mass balance, the chlorine content in the first portion of the quenched fluid ranged from 1 to 2 wt %. The next fluid portions during the stepwise defluidization no longer contained chlorine. Consequently, removal of chlorine from the system reduces the solubility of carbonate in the fluid phase of the pelitic system. It is possible that chlorine also significantly affects the properties of the supercritical fluid itself. Additional data are needed to evaluate the effect of the chlorine content, the composition, and the concentration of carbonate on its stability in the pelitic system. This is important because the content of the carbonate fraction in actually subducting sediments can vary greatly [3].

Chlorides play an important role in fluids of subduction zones [16, 17]. Data on fluid inclusions in olivines from island arc basalts, minerals of high-pressure metamorphic rock, and mantle xenoliths from subduction zones indicate that, in most cases, the chloride content in subduction-zone fluids exceeds 1 wt % [1720]. Based on the data obtained, it may be concluded that early defluidization of the pelitic material with the removal of all chlorine with fluids from subduction zones provides effective CO2 transport to deep reservoirs during further subduction of the slab. However, if the subducting pelitic material passes the main defluidization phase at a depth of ~150 km, this will lead to the formation of a chloride-bearing fluid, in which all the carbonate in the pelite may be dissolved. This process can provide complete decarbonation of the pelite with removal of a large amount of CO2 into the mantle wedge.