Experimental method

For all experiments an eclogite composition was employed similar to the G2 composition of Pertermann and Hirschmann (2003) in the system SiO2-Al2O3-MgO-FeO-CaO-Na2O. The composition was assembled from oxides with a deficit of MgO, CaO and FeO in order that 10 % carbonate phases could be later added. The composition was then melted and quenched to glass and reduced in a gas

mixing furnace. This carbonated eclogite composition was finally mixed with 5 wt. % Ir metal powder and 10 % graphite powder to make a starting composition called START 1.

The first starting composition used in the experiments was produced by mixing START 1 with natural kyanite (~10 wt. %) from Minas Gerais (Brazil) to make START 2. A second composition was produced by mixing START 1 with, 10 wt. % TiO2 and FeTiO3 to make START 3 in order that both rutile and ilmenite are stabilized as observed in some experiments on natural eclogite compositions (Dasgupta et al. 2004; Zhang et al. 2003). Starting compositions are given in Table 6.1. Experiments were performed at 3 GPa in a multianvil press at 950 and 1000 °C employing an 18/11 assembly. The large assembly allowed two capsules to be employed in each experiment so both compositions (START 2 and 3) were run simultaneously. Experimental and analytical techniques employed were the same than those described in the previous chapters.

Table 6.1 Starting materials and comparison with compositions used in previous studies.

This study Previous studies

Table 6.1 Experimental run conditions and products.

Run no. P (GPa) T (°C) Time (hr) phases

V636a 3 950 ~42 Omph Gnt Coe G Ky Cc Liq V637a 3 950 120 Omph Gnt Coe G Ky Cc Liq V624a 3 1000 ~31 Omph Gnt Coe G Ky Liq V636b 3 950 ~42 Omph Gnt Coe G Rut Cc V637b 3 950 120 Omph Gnt Coe G Rut Cc V624b 3 1000 ~31 Omph Gnt Coe G Rut Liq

Notes: Omph, omphacite; Gnt, garnet ; Coe, coesite; G, graphite; Ky, kyanite; Rut, rutile; Cc, carbonate s.s.; Ilm, ilmenite;

Liq, melt.

6.3 Results

Experimental run conditions and resulting assemblages are shown in Table 6.2. Phase compositions including melt and alloy are reported in Table S6.1 and S6.2 (Appendix). Recovered samples from both compositions showed crystallization of the main silicate phases, such as omphacitic clinopyroxene and garnet. Grain sizes ranged between 5 and 20 μm. Small grains of coesite were also observed in addition to Ir-Fe alloy and graphite dispersed in a matrix or occurring as inclusions in garnet inline with previous studies. In experiments employing the START 2 composition kyanite was present. At 950 °C a carbonate melt was produced in equilibrium with cpx, garnet, kyanite and coesite. The melt contained approximately 6 wt. % SiO2 and had Ca# of 0.64, which is similar to the eutectic carbonate composition described by Dasgupta et al. (2005) and shown in figure 6.3. Melt composition was estimated also at 950 °C but after 120hr (V637a) to test the attainment of equilibrium between solid phases, liquid and alloy composition. The Ca# of this melt remained constant with a slight increase of MgO from 8.40 wt. % after 42 hr to 10.32 wt. % after 120hr accompanied by a simultaneous decrease of CaO (~ 2 wt. %) while FeO remained constant. Although melt was observed at 1000 °C run, its composition was not measurable.

Rounded garnet crystals occurred always strictly in contact with euhedral omphacitic cpx. The Mg number of garnet, calculated as Mg/(Mg+Fe), increased with temperature from 0.40 to 0.50 in agreement with experiments recovered both from carbonated systems (Yaxley and Green, 1994) and dry systems (Pertermann et al., 2003) with CaO content being stable and likely dependent on the bulk composition. The alumina content in clinopyroxene was approximately 15 wt. %, which is higher than the more typical 5 wt. % observed in previous carbonated eclogite experiments at the same conditions (Dasgupta et al. 2005), due to kyanite saturation but in agreement with cpx crystallized from the same MORB-like bulk composition (Pertermann et al., 2003) with a Mg number of ~0.78. Further, the Na2O content of cpx is observed to decrease from 7.17 to 5.75 wt. %, while the CaO content increases from 11.72 to 14.43 wt. % mainly as consequence of the increased melt fraction.

In experiments performed using the START 3 composition, rutile was identified along with cpx, garnet, coesite and carbonate phases, while ilmenite was not found as expected from results of previous studies on a similar bulk composition (Yaxley and Green, 1994; Pertermann et al., 2003). Carbonate phases observed in the short run at 950 °C are siderite and siderite-magnesite-calcite solid solutions, which show evidence of disequilibrium, as they appear dissociated each other and dispersed within other phases. After 120 hr siderite still occurs along with other carbonate, such as a siderite-magnesite and a calcite-siderite-magnesite solid solution. At 1000 °C no solid carbonates are found and a glass is produced which is not analysable but might have a composition reported by the already mentioned authors. The Al2O3 content of the omphacitic cpx is observed to increase with temperature from 9.80 to 13.22 wt. % while the other oxides do not change. Much more Fe is contained in cpx, about 8 wt. %, compared to the kyanite-bearing assemblage. Finally, garnet has an Mg number ranging from 0.36 to 0.42 as temperature increases from 950 to 1000 °C. However, garnet has more Fe than START 2, while the TiO2 content is about 1.0 wt. %, a bit higher when compared with experiments where garnet

coexists with ilmenite (Dasgupta et al., 2005). Neither TiO2 is observed to dissolve in the iridium-iron alloy or iridium into the rutile phase.

The partitioning of Fe-Mg between garnet and clinopyroxene in both series of experiments is compared with the equation calibrated by Ellis and Green (1979) calculated for the same compositions in figure 6.6. As shown in this figure the value is usually observed to decrease as a function of temperature and it is strongly dependent on the composition. Kyanite-bearing runs are for the main part in better agreement than those containing rutile.

Figure 6.5. Left) recovered sectioned run product V636b (3 GPa/950 °C) showing a Ti-bearing assemblage with rutile after 42 hr. Right) V637a run (3 GPa/950 °C) showing a heterogeneous mineral assemblage after 120 hr with kyanite. Notes:

coesite (coe), garnet (gnt), graphite (G), liquid (Lq) and the iron-iridium alloy (bright phase).

Figure 6.6 The Fe-Mg distribution coefficient calculated between garnet and cpx as function of temperature. Dotted curves are calculated using Ellis and Green (1979) with XCa representative of the appropriate experiment (see legend). Green curve with a XCa of 0.2; yellow curve is 0.23 and black curve is 0.27. Small numbers indicate XCa in garnet.