Stability of carbonated magmas at the base of the Earth's upper mantle



[1] We measured the density of carbonated basaltic melt containing 5.0 wt.% CO2 at 2573 K and in the pressure range from 16.0 to 20.0 GPa by using the sink/float method with single crystal diamond as a density marker. We observed sinking of diamond at 19.0 GPa and flotation of diamond at 20.0 GPa and 2573 K. Using the third order Birch-Murnaghan equation of state, the calculated isothermal bulk modulus (KT) of the carbonated basaltic melt (5.0 wt.% CO2) and its pressure derivative (K′) are 16.0 ± 1.0 GPa and 5.2 ± 0.2, respectively. Our result implies that magmas can contain up to 3.0–4.0 wt.% CO2 to be denser than the surrounding mantle at the top of the 410 km discontinuity.

1. Introduction

[2] Density of magma plays an important role in fractionation, differentiation and magma genesis in the Earth and other planetary interiors [e.g., Stolper et al., 1982; Ohtani, 1983, 1984; Delano, 1990; Ohtani et al., 1995]. Several seismological and electrical conductivity studies indicate the presence of low velocity zones (LVZ) at the top of the 410-km discontinuity which reveal the possible existence of a melt at this boundary [e.g., Reveanaugh and Sipkin, 1994; Toffelmier and Tyburczy, 2007]. The detection of seismic and electrical conductivity anomalies at the 410 km discontinuity triggered suggestion of a conceptual geophysical model [Bercovici and Karato, 2003] which shows wet mantle material passing through the 410 km discontinuity and accumulates at the bottom of the upper mantle, decouples trace element circulation from bulk mantle material, and explains the geochemical observations for separate mid-ocean ridge basalt and ocean-island basalt source regions while maintaining whole mantle convection. Density measurements of anhydrous basaltic melts indicate that it is denser than the surrounding mantle near 410-km depth [Ohtani and Maeda, 2001]. However, the melting temperature of peridotite is much higher than about 1750 K, estimated at 410-km depth [e.g., Akaogi et al., 1989; Katsura et al., 2004]. Therefore, it can be suggested that possible melts at the top of the 410 km discontinuity may contain some volatiles such as H2O, CO2 or both [Ohtani et al., 2004]. It has been shown recently that hydrous peridotitic and basaltic melts are denser than peridotite at the top of the 410 km discontinuity and therefore can be accumulated at the base of the upper mantle [Matsukage et al., 2005; Sakamaki et al., 2006]. After H2O, CO2 is the second most abundant volatile component in the mantle and it could be also important to constraints the conceptual models experimentally for the explanation of LVZ near a 410 km depth. In the present study, we have measured the density of carbonated basaltic melt at pressures from 16–20 GPa and 2573 K and discussed its possible stability at the base of the upper mantle.

2. Experimental Procedure

[3] Density of carbonated basaltic melts was determined using the sink/float method, which has been widely used for dry and hydrous melt [e.g., Agee and Walker, 1993; Ohtani et al., 1993; Suzuki et al., 1995; Ohtani and Maeda, 2001; Sakamaki et al., 2006]. Single crystal diamond of ∼100 μm diameter was used as a density marker and placed at the center of a sample capsule. Depending on the movement of diamond upward or downward, the density of the surrounding melt has been determined. The starting material was a synthetic carbonated basalt (5 wt.% CO2), which is close to the average mid-oceanic ridge basalt, MORB [after Melson et al., 1976] in the silicate compositions (Table 1). The glass was synthesized from pure oxides at 1300°C under controlled oxygen fugacity to remove Fe+3 from the starting material. Carbon dioxide was added as CaCO3 adjusting the proportions of CaO. All experiments were carried out in a Kawai-type multi-anvil high-pressure apparatus (1000 ton press) installed at Tohoku University, Sendai, Japan. WC cubes with truncated edge length of 3.5 mm were used for all experiments. The furnace assembly was essentially the same as the one used by Litasov and Ohtani [2002]. Semi-sintered zirconia was used as a pressure medium and cylindrical LaCrO3 heater was used as a heating element. A powdered mixture of carbonated MORB was packed into a Pt/Re double capsule with an inner diameter of 1.5 mm and a height of 2.0 mm. The pressure calibration was carried out at high temperature using α-β [Morishima et al., 1994; Katsura et al., 2004] and β-γ [Suzuki et al., 2000] phase transformations in Mg2SiO4 at 1873–2273 K. Pressure measurements are considered to be accurate to ±0.5 GPa. Temperature was measured using a W3%Re-W25%Re thermocouple up to about 1800–2200 K, however the thermocouple was not stable at higher temperatures. Thus, temperature and power ratio was used to estimate the target temperature of the experiment. Estimated temperature uncertainty is about 100 K. The duration of the each experiment was 1 minute. These experimental conditions are above the liquidus of the dry MORB system [Yasuda et al., 1994; Litasov and Ohtani, 2005] and it was confirmed by the quenched textures of the melts in the run products. Finally, the position of the diamond marker was observed using the optical microscopic images and back scattered electron images (BSE) of the samples. The compositions of the quenched carbonatitic basaltic melts were analyzed by an electron probe micro analyzer (JEOL Superprobe JXA-8800). The amount of CO2 in the melt was estimated by the total deficit of the microprobe analyses [e.g., Dalton and Presnall, 1998] although it is not an ideal procedure but most satisfactory available at present. We found that the composition of the run products are the same as that of the starting material in both major compounds and CO2 contents, and the error in the CO2 content is less than 10 % as shown in Table 1.

Table 1. Compositions of the Starting Material and Run Products
 MORBCarbonated MORBDBC-19 19 GPa/2300°C n = 12DBC-10 20 GPa/2300°C n = 13
  1. a

    n = number of analysis.

SiO251.0948.248.07 (0.95)48.10 (0.45)
Al2O315.714.914.60 (0.75)14.80 (0.28)
FeO9.819.39.27 (0.82)9.21 (0.36)
MgO7.747.47.33 (0.54)7.40 (0.23)
CaO11.5110.910.81 (0.36)10.73 (0.13)
Na2O2.484.34.17 (0.72)4.05 (0.46)
CO2 5.0  
Total (wt.%)10010094.2594.29

3. Results and Discussion

[4] Density measurement of carbonated basaltic melt was carried out in the pressure range from 16.0 to 20.0 GPa at 2573 K. We have observed sinking of diamond at 16.0 to 19.0 and flotation of diamond at 20.0 GPa (Figures 1a and 1b). The results of the experiments are summarized in Figure 2. The density of diamond marker at high pressure and temperature is calculated using the third-order Birch-Murnaghan equation of state (EOS),

equation image

where ρ is the density, ρ0 is the zero pressure density, KT is the isothermal bulk modulus, and K′ is the pressure derivative of bulk modulus.

Figure 1.

(a) Back-scattered electron image of run DBC-10 (20 GPa, 2573 K and 1min) showing flotation of diamond marker in the carbonated basaltic melt. (b) Back-scattered electron image of run DBC-19 (19 GPa, 2573 K and 1min) showing sinking of diamond marker in the carbonated basaltic melt.

Figure 2.

Pressure versus density diagram. Solid triangle, flotation; and solid down triangle, sinking of diamond in MORB melt + 5.0 wt.% CO2. The compression curve of dry MORB is after Ohtani and Maeda [2001].

[5] Consequently, the density of carbonated basaltic melt containing 5.0 wt.% CO2 is 3.57 g/cm3 at 19.5 GPa and 2573 K, by using the thermoelastic parameters of diamond [Zouboulis et al., 1998] (Figure 2).

[6] In order to calculate the bulk modulus we have to know the zero-pressure partial molar volume of CO2 (equation image). However, the solubility of CO2 in the silicate melts is very low at 1 bar [e.g., Khitarov and Kadik, 1973] which makes accurate estimation of equation image for carbonated basaltic melt at 1 bar and temperatures, corresponding to those in the present study (2573 K), very difficult. Table 2 summaries the estimates of the partial molar volume of CO2 in carbonate liquids in the literature. Liu and Lange [2003] estimated the thermal expansivity (equation image) for carbonatite melts between 3.0 × 10−3 and 5.0 × 10−3 cm3/mol-K. We assumed that the partial molar volume of CO2 (equation image) of the carbonated magma is the same as equation image in CaCO3 melts and take an average value of equation image based on data by Dobson et al. [1996] and Genge et al. [1995]. We also used thermal expansivity 4.0 × 10−3 cm3/mol-K, which is the average of the data by Liu and Lange [2003]. Based on above data we calculated equation image cm3/mol (2573 K) at 1 bar. Combining these data with the density of dry MORB melt [Ohtani and Maeda, 2001], we can obtain the density of carbonated basaltic melt with 5.0 wt.% CO2 at 1 bar and 2573 K as 2.37 ± 0.02 g/cm3.

Table 2. Estimates of the Partial Molar Volume of CO2 in Carbonate Liquids Determined by Different Authors
Compositionequation image, cm3/molP, GPaT, Kequation image, cm3/mol-KReference
  • a

    Calculated from the densities of the dry basalt and carbonated basalt containing 5.0 wt.% CO2 (see the text in detail).

CaCO333.360.116734.0 × 10−3Genge et al. [1995]
K2Ca(CO3)235.200.119504.0 × 10−3Dobson et al. [1996]
K2Ca(CO3)229.922.519504.0 × 10−3Dobson et al. [1996]
K2Ca(CO3)231.764.019504.0 × 10−3Dobson et al. [1996]
K2CO332.350.111004.3 × 10−3Liu and Lange [2003]
Na2CO328.730.111005.4 × 10−3Liu and Lange [2003]
CaCO325.810.111003.4 × 10−3Liu and Lange [2003]
Carbonated basalta20.9819.525734.0 × 10−3Present study

[7] Assuming the pressure derivative of the isothermal bulk modulus (K′) is the same as the dry MORB melt, i.e. 5.0 [Ohtani and Maeda, 2001], we calculated the isothermal bulk modulus (KT) of the carbonated basaltic melt containing 5.0 wt.% CO2 and i.e. KT = 17.0 ± 1.0 GPa using equation (1).

[8] Using the density of carbonated basaltic melt we can calculate the partial molar volume of CO2 in the melt at 19.5 GPa and 2573 K. Density of anhydrous basaltic melt was estimated as 3.71 g/cm3 at 19.5 GPa and 2573 K [Ohtani and Maeda, 2001]. The volume increase of carbonated melt was obtained by subtracting the volume of dry melt from the volume of carbonated basaltic melt. Using that volume, the partial molar volume of CO2 in the carbonated basaltic melt is 21.0 ± 1.0 cm3/mol at 19.5 GPa and 2573 K. Using the partial molar volume of CO2 at 1.0 bar, 2.5 and 4.0 GPa [Genge et al., 1995; Dobson et al., 1996] we can calculate the EOS of CO2 species in the silicate melt. We used the Vinet EOS [Vinet et al., 1989], which is superior to the other types of EOS for soft materials, which is represented by the following equation,

equation image

where V is the volume, Vo is the zero pressure volume, KT is the isothermal bulk modulus, and K′ is the pressure derivative of bulk modulus.

[9] By using equation (2), we can calculate KT = 3.7 ± 1.8 GPa and K′ = 9.0 ± 2.0 of CO2 species in the silicate melt. Figure 3 shows the compression curves of partial molar volume of CO2 in the melt at 2573 K. Using obtained EOS we can estimate the density of carbonated melt in the Earth's interior. The corrections based on the calculated EOS for CO2 in the melt to present equation of state give KT = 16.0 ± 0.6 GPa and K′ = 5.2 ± 0.2 for MORB + 5.0 wt.% CO2 by using the third order Birch-Murnaghan EOS. Applying the EOS for CO2 in the melt for peridotite melt using data for anhydrous peridotite melt (KT = 32 GPa and K′ = 4.6) after Suzuki et al. [1998] and Suzuki and Ohtani [2003] assuming linear mixing of CO2 component into the melt, we can obtain KT = 24.9 ± 0.6 GPa and K′ = 5.1 ± 0.3 for carbonated peridotitic melt with 5.0 wt.% CO2.

Figure 3.

Compression curve of the partial molar volume of CO2 in melt at 2573 K. Reference data are after Genge et al. [1995], Dobson et al. [1996], and Liu and Lange [2003]. The error bars represent the upper and lower bounds derived from the present density measurement. The error bars for reference data represent uncertainty in determination of thermal expansion coefficient [Liu and Lange, 2003]. The compression curve can be expressed by the Vinet equation of state with equation image = 37.09 cm3/mol, K = 3.7 GPa, and K′ = 9.0 (see the text in detail).

[10] Figure 4 shows the density of carbonated, hydrous, dry peridotitic melts in comparison with ak135 and PREM density profile after Kennett et al. [1995] and Dziewonski and Anderson [1981] at 1873 K. Assuming the linear dependence of the density from CO2 content in the melt, we can suggest that the peridotite melt can contain up to ∼4.0 wt.% CO2 to be denser than the surrounding mantle at the top of the 410 km discontinuity, whereas it is less dense compared to the surrounding mantle in the transition zone below the 410 km discontinuity due to formation of a denser olivine polymorph, wadsleyite. Thus, the melt can be accumulated at the top of the discontinuity. These amounts of CO2 are comparable with the amount of H2O in the hydrous basaltic (∼3.0 wt.%) and peridotitic (∼6.7 wt.%) melts, which is stable atop of the 410 km discontinuity [Sakamaki et al., 2006]. Recently Dasgupta and Hirschmann [2006, 2007] have studied the natural carbonated peridotite system between 3 and 10 GPa and observed that solidus of natural carbonated peridotite is 450–600°C lower than the solidus of natural volatile free peridotite and near solidus melt is carbonatitic in nature containing ∼45 wt.% CO2. If we combine our data with hydrous basaltic melt [Sakamaki et al., 2006] and consider the linear mixing between H2O and CO2 then the basaltic melt with 1.5 wt.% H2O and ∼1.3 wt.% CO2 and peridotitic melt with 3.3 wt.% H2O and 2.0 wt.% CO2 could be stable at the top of 410 km discontinuity.

Figure 4.

Comparison of compression curve of carbonated melts at 1873 K and mantle density profile ak135 [Kennett et al., 1995] and PREM [Dziewonski and Anderson, 1981]. The density curve of peridotite melts with 5.0 wt.% CO2 is expressed by the Birch-Murnaghan equation of state with zero-pressure density ρ0 = 2.67 ± 0.01 g/cm3, isothermal bulk modulus KT = 24.9 ± 0.8 GPa, and its pressure derivative K′ = 5.1 ± 0.3. Data for hydrous peridotitic melt containing 2.0 and 8.0 wt.% H2O (dotted line) after Sakamaki et al. [2006] are also shown for comparison.


[11] We thank two anonymous reviewers for their helpful comments on the manuscript. We are grateful to H. Terasaki of Tohoku University for his useful discussion and experimental assistance. We also thank Y. Ito for performing EPMA analysis of our run products. S.G. gratefully acknowledges the Ministry of Education, Culture, Science, Sport and Technology, Japan for providing him the Monbukagakusho Fellowship. This work was supported by the Grants-in-aid for Scientific Research from Ministry of Education, Culture, Science, Sport and Technology of Japanese Government (15104009 and 16075202) to E. Ohtani, and conducted as a part of the 21st Century-of-Excellence program, ‘Advanced Science and Technology Center for the Dynamic Earth’ at Tohoku University.