Experimentally determined stability of alkali amphibole in metasomatised dunite at sub-arc pressures

Abstract

The phase and melting relationships of olivine mixed with 25 % of hydrous felsic slab melt have been determined in piston-cylinder experiments between 2.5 and 4.5 GPa and 800 to 1,050 °C to constrain metasomatic processes in the mantle wedge above subduction zones. At sub-solidus conditions, olivine, orthopyroxene, phlogopite, a Na-rich amphibole and an aqueous fluid are present. Na-rich amphibole is still observed at 950 °C at 4.5 GPa, providing evidence that this hydrous phase might be stable at sub-arc depths in an alkali-rich, Ca-poor mantle wedge. The maximum temperature stability is reached at 1,000 °C at 3.5 GPa, where amphibole coexists with hydrous melt. A sodium-rich phlogopite is stable over the whole range of P–T conditions investigated. At 2.5 GPa, 850 °C, aspidolite (Na analogue of phlogopite) has been observed as a sodium-bearing phase in the peridotite. The wet solidus in the metasomatised dunite lies between 850 and 900 °C at 2.5 GPa and between 950 and 975 °C at 3.5 GPa. At 4.5 GPa, melting relations are ambiguous and no clear solidus was found. The consumption of amphibole and minor phlogopite at the wet solidus produced Na- and H₂O-rich phonolitic melts. The presence of phlogopite and sodic amphibole in the metasomatised dunite has implications on alkali and water storage in the part of the mantle wedge that is coupled to the down-going slab and might play a role on alkali and trace element recycling through arc magmatism.

Appendix

Aqueous fluid composition

The composition of aqueous fluids cannot be analysed directly from quenched products in the crystalline part of the experimental charge, and quench material in carbon spheres is too heterogeneous and scarce to be analysed with an electron beam. Laser ablation ICP-MS of carbon sphere traps provided more constant and reliable results of element ratios that are used for the calculation of the composition of aqueous fluids. In order to fully quantify the concentration of each element in the aqueous fluid, an internal standard is required. Due to the presence of abundant mica in our experiments, LILE could not be selected as fully incompatible elements. However, La, Ce, U, Th, Nb and Ta were almost completely incompatible in the residual solid assemblage and thus can be used as internal standards for LA-ICP-MS analysis of the quench material.

An initial estimation of the modal content of the different phases present in each experiment is based on SEM imaging. An initial estimation of the mass of water available for the fluids can be calculated based on the amount of hydrous phases in the solid residue of the experiment. Given that U and La are entirely incompatible in the residue, the H₂O/U and H₂O/La can be calculated. In the next step, the major element composition of the fluid is obtained from the measured ratio with respect to U and La coupled with the additional constrain that all solutes and H₂O must total 100 %. These newly calculated fluid compositions are then again inserted into mass balance calculations, providing a more accurate calculation of modal proportions. Through a cycle of two iterations, fluid and modal compositions become robust and converge to values that provide an estimation of phases present in the experiments (Electronic Supplement Table 5) as well as the composition of fluids (Table 6). Errors calculated for quantitative composition of aqueous fluids are 1σ of the average estimation from six trace elements used as internal standards. The same method applied to experiments above the solidus has shown a good match between calculated values for melts and direct analyses obtained from quenched hydrous melt ponds, confirming the validity of this method.

Due to the large sampling volume, analyses can be mixtures of quenched fluids and minerals. The presence of mineral phases from dissolution–precipitation was easily detected, as trace element pattern during ablation would be clearly modified by the presence of olivine and orthopyroxene (Ni, Sc, Mn and Cr) and mica (K, Rb, Cs, Ba and Ti). Analyses containing large contamination from one of these phases (generally mica) were rejected. Amphibole formed by dissolution–precipitation is more difficult to notice, as it does not have a distinctive trace element pattern when compared with aqueous fluids. SEM observation shows that magnesiokatophorite is the most common crystalline phase that can contaminate the fluid traps. The effect of amphibole contamination on aqueous fluids is given for an estimated value of 10 % contamination, providing final MgO contents (4–7 wt% at 3.5 GPa) that are in agreement with analysis of aqueous fluids in equilibrium with peridotites (Stalder et al. 2001; Mibe et al. 2002; Dvir et al. 2010) (Table 6). In Fig. 7, ~20 % amphibole contamination is the maximum value possible without removing completely a component (MgO and/or CaO) from the aqueous fluid at 3.5 GPa. Tick marks in Fig. 7 are also displayed for 10 and 15 % removal. At 2.5 GPa, the volume of observed magnesiokatophorite was very low and a value of 2 % contaminant (CaO removal) was selected. For each experiment, magnesiokatophorite composition from the main assemblage was used.

Mass balance calculations

The mass balance calculations and associated errors are done through a weighted least squares method to avoid large errors created by initial analytical errors for small concentrations. Major element data are used for all minerals and fluids. Olivine modal content is more accurately quantified on the basis of its transition metals content (TM: Ni, Cu, Zn and Mn), garnet on REE, TM and HFSE, pyroxene on REE and TM, mica on H₂O, LILE and TM and amphibole on H₂O, LILE, TM and REE. Aqueous fluids and hydrous melts are not subject to any preferential weight, and all major and trace elements are used to estimate the modal abundance of fluids. While subtraction of amphibole from the aqueous fluid composition has a significant effect on the chemical composition of the fluid (Table 6), it only marginally impacts the mass balance because the overall abundance of aqueous fluid in the experiment is rather small. The difference in modal abundances due to amphibole subtraction is approximately 1 %.

Calculation of reaction equations

Equations (1) and (2) are derived from phase compositions in experiments, expressed in mass ratio. Olivine, orthopyroxene and amphibole compositions are considered as unmodified between experiments (changes are below the estimated mass balance error (Electronic Supplement Table 5) in these equations). Mica has small variations in chemical composition that are taken into account in these equations as phl₁ and phl₂. Fluids have major variations in chemical composition between 3.5 and 4.5 GPa (fluid₁ and fluid₂, Table 6) and between aqueous fluids and hydrous melts at 2.5 and 3.5 GPa (Tables 5, 6).

Equation (1) expresses changes occurring between C3085 (3.5 GPa—950°C) and C3361 (4.5 GPa—950°C) where amphibole breakdown forms garnet-bearing assemblage. The equation is factorised to the lowest common denominator.

Equation (2) expresses changes occurring upon melting between C3085 (3.5 GPa—950°C) and D1078 (3.5 GPa—975°C). The equation is based on a factor of 100 for melt production.