Oxygen fugacity determination

In these experiments the fo2 was buffered by the presence of graphite and a carbonate-bearing phase coexisting nominally with cpx, garnet and coesite in an analogous equilibrium to (6.2). The fo2 of this equilibrium can be measured, however, using a second equilibrium employing the phases present plus Ir-Fe alloy. In experiments with kyanite (START 2) the oxygen fugacity can be measured using equilibrium,

Fe3Al2Si3O12 = Al2SiO5 + 2SiO2 + 3Fe + 1.5O2 (6.3) garnet kyanite coesite alloy

For this equilibrium only the activities of Fe3Al2Si3O12 and Fe must be considered and both are well known. End member thermodynamic data used in addition to that used in previous chapters is listed in Table 6.3. The fo2 is then calculated from the expression,

[ ] Garnet

The determination of , the Fe activity in Ir-Fe alloy, has been previously described. The activity of the Fe3Al2Si3O12 garnet component is calculated from,

Where XFe is the molar Fe/(Fe+Mg+Ca) in garnet and γFe is the activity coefficient determined using the model of O’Neill and Wood (1979) from,

RTlnγFe=XMg2WMg-Fe+XCa2WCa-Fe+XCaXMg(WMg-Fe+WCa-Fe-WCa-Mg) (6.6)

Where WMg-Fe, WCa-Fe, WCa-Mg are Margules interaction parameters with the values 800, 0 and 11000 J mol^-1 respectively.

The oxygen fugacity in the second series of Ti-bearing experiments can be calculated using the equilibria

FeTiO3 = TiO2 + Fe + 0.5O2 (6.7) ilmenite rutile alloy

Unfortunately, ilmenite was not actually found in the experiments, however if we assume its presence the fo2 can be calculated from the relation,

[ ] Ilmenite

Rutile in the recovered experiments is relatively pure. The activity of FeTiO3 in ilmenite is unknown as no analyses could be made. However, in the similar experiments of Dasgupta et al. (2005) ilmenite was observed and estimates of the likely activity can be made. Variations in ilmenite activity do not affect the fo2 determined from (6.8) significantly.

Table 6.3 End member thermodynamic data used for calculating oxygen fugacity

Ref.

Notes: [1] O’Neill and Wood (1979) [2] O’Neill (1987) [3] Robie and Hemingway (1995) [4] Holland and Powell (1990) [5] Fei (1995) [6] Fabrichnaya et al. (2004) [7] Komabayashi and Fei (2010).

6.5 Discussion

In figure 6.7 oxygen fugacities calculated from both kyanite and rutile bearing experiments using equilibria (6.4) and (6.8) are shown. In one kyanite bearing experiment no metal alloy could be analysed. The DCDG curve is equilibrium (6.2) calculated from the pure end-member thermodynamic data. The EMOG curve, which describes the fo2 where graphite and carbonate coexist in peridotitic assemblages, is shown calculated from the equation given in chapter 3. Melting of carbonated peridotite commences at approximately 1300 K but no account has been made of this in figure 6.7 as the mole fraction of CO2 in the EDDOG equation from chapter 3 has been fixed at 0.5.

The agreement between Fe-Mg partitioning between cpx and garnet shown in figure 6.6 with previous models indicates a close approach to equilibrium for the kyanite-bearing experiments, while those from rutile-bearing experiments are in poorer agreement. No ilmenite could be found in the recovered rutile-bearing samples, so the fo2 calculation for these samples must be considered less reliable. In addition, the carbonate phase in the rutile-bearing experiments was very Fe rich. It is possible that the Fe-enrichment in the carbonate phase resulted in higher determined oxygen fugacities but this melt is anyway inconsistent with previous carbonate phases measured in eclogites at similar conditions (Dasgupta et al. 2005). In the kyanite bearing experiments carbonate melt was analysed, which was consistent with previous studies. In addition as Ir metal is Fe free in the starting compositions, more reduced results have been displaced further from the starting material and are therefore likely to be closer to equilibrium. For these reasons the kyanite results are considered to be more reliable than the rutile-bearing experiments.

-3

Figure 6.7 Oxygen fugacities calculated from kyanite (blue circles) and rutile (brown squares) bearing experiments using equilibria (6.4) and (6.8) respectively. Values are compared with calculations of the DCDG equilibrium (6.2) from thermodynamic data and the determination of the EDDOG buffer from chapter 3. Uncertainties propagated from compositional measurements are of the order of 0.2 log units.

Luth (1993) argued that the DCDG buffer was at a fo2 above that of EDDDOG and that for a fixed oxygen fugacity eclogitic rocks could be in the graphite/diamond stability field while peridotites could be in the carbonate field. Similarly carbonate melts infiltrating from peridotites might therefore be reduced to diamond/graphite on entering eclogites, thus supporting the observation of diamonds being hosted disproportionately by eclogitic rocks in some localities. The results from kyanite-bearing experiments do not support this hypothesis, however, because the fo2 buffered by the coexistence of carbonate melt and graphite in these experiments is below the EDDOG buffer and thus carbonates would be stable at even lower oxygen fugacities in eclogites compared to peridotites. An additional issue not considered by Luth (1993) is that as the carbonated solidus in eclogitic rocks is at least 200 °C lower than in peridotites carbonate liquids infiltrating from a peridotite would likely flux a much greater proportion of melting on entering eclogitic rocks. At similar temperatures carbonate melts in

eclogites contain a larger silicate component than in peridotitic rocks and, therefore, as seen in chapter 3, the fo2 for graphite/diamond formation will be driven to even lower values by the dilution of the CO2

component in the melt.

For the previously described reasons it is, therefore, unlikely that the diamond-eclogite association in some localities arises from a difference in the graphite-carbonate buffering reaction between peridotites and eclogites. One possibility for the association is that carbonates, initially associated with eclogites during subduction, remain at low enough temperatures for a solid-state transformation to diamond to occur. However, this is implausible as diamond is a metasomatic mineral and growth rings imply growth from a liquid or fluid. A further possibility is that carbonate liquids entering eclogites are forced to reduce due to redox reaction with the silicate minerals. Garnet in eclogites, for example, is more Ca-rich than in peridotites and a plausible carbonate reduction reaction might consider the andradite component i.e.

CaCO3 + 2CaSiO3 + 2FeSiO3 = Ca3Fe2Si3O12 + SiO2 + C (6.9) melt cpx cpx garnet coesite graphite/diamond

In order to determine whether such a reaction could take place the fo2 of the oxidation reaction involving andradite would need to be determined. The most straight forward method to achieve this would be to measure the andradite component in garnets produced in similar experiments to those described in this chapter.

6.6 Conclusions

The oxygen fugacity at which graphite coexists with solid carbonate or melt in eclogitic assemblages has been determined at 3 GPa and temperatures in the range between 950-1000 °C using iridium as redox sensor. Preliminary results are presented for two different eclogitic assemblages, which are thought to originate by metamorphism of MORB basalt (kyanite and TiO2-bearing). This assemblage might represent a potential carrier for carbon down into the mantle in subduction zones. However, conditions at which carbonate or graphite/diamond may persist at depth need to be investigated based on the proposed study. Preliminary results show also compositions of coexisting silicate phases, which are in agreement with previous studies where MORB-like bulk compositions were used. On the other hand, results are in contrast with studies that considered alterated MORB basalts as starting composition. Finally, experiments carried out using a kyanite-bearing assemblage show fo2, which are more reliable than TiO2-bearing and appear in contrast with previous predictions on the equilibrium between diamonds and carbonate in these rocks. However, results from TiO2-bearing system can be improved by determining the concentration of ferric iron in garnet at a given fo2 and fixing the oxygen fugacity with the coexistence of ilmenite.

Further results will provide the basis for determining whether with increasing pressure during subduction there will be a tendency for carbonates to reduce to graphite (diamond) in eclogitic rocks as a result of pressure favouring the formation of Fe³⁺ components in garnet and clinopyroxene.

The investigation of the Earth’s mantle is crucial to our understanding of the geochemical and petrological processes generating oxidation of the mantle, which in turn can affect volcanic eruptions and loss of volatiles to the atmosphere. To date, knowledge concerning the mantle composition and structure has come mainly from the study of seismic wave velocities through the Earth’s interior in addition to the interpretation of mantle xenoliths and erupted magmas. Xenoliths, unfortunately, arise from depths <200 km into the mantle and can be influenced by melting and metasomatic processes.

Magmas are often altered and influenced by processes of magmatic differentiation, assimilation and mixing. A great deal more information can be provided by natural mineral and melt inclusions in erupted xenocrysts and diamonds. Some diamond inclusions, for example, are considered to be direct samples of the mantle from depths greater than 600 km and show the presence of solid carbonate.

These inclusions provide important information not only about the mineralogy of the region of the mantle where they formed but rather unique information on the most oxidized conditions at which diamond can be stable along with mantle silicates.

In the course of this thesis, the conditions at which carbonate minerals and melts form from either graphite or diamond were studied as a function of pressure, temperature and oxygen fugacity in typical mantle peridotite assemblage. These conditions not only define those that are likely prevalent during diamond growth but also identify the conditions under which reduced carbon in up welling mantle would become oxidized to produce carbonate minerals and melts. In addition the redox state of carbon/carbonate equilibria was determined with respect to the proportion of ferric iron-bearing mineral components of the upper mantle, transition zone and lower mantle, which in turn can be used to refine oxy-thermobarometry equilibria for determining the redox state of the mantle.