The role of TiO₂ phases during melting of subduction-modified crust: Implications for deep mantle melting

Abstract

Partitioning of Nb, Ta, Hf and Zr between rutile, its high-pressure polymorph TiO₂(II) and silicate melt has been experimentally determined at 2 GPa/1200 °C, 6 GPa/1600 °C, 8 GPa/1800 °C and 10 GPa/1900 °C. Results show that characteristic depletion of Nb and Ta in partial melts due to the presence of rutile in solid residues (for example during melting of subducting oceanic crust) is strongly dependent on depth of partial melting. With increasing pressure, changes in melt structure result in marked reduction in D_min/melt for Nb (14.8, 5.4, 2.5, and 2.4 with increasing P/T) and Ta (28.0, 17.0, 6.9, and 5.5). A strong pressure effect is also noted in D_min/melt for Zr (2.1, 0.6, 0.9, and 1.2) and Hf (4.1, 0.9, 1.3, and 1.3), although for these elements the rutile to TiO₂(II) transition also influences partitioning behaviour. Results have important implications for melting of oceanic crust in Earth's deep mantle. Ancient subduction-modified crust cannot be a direct source for ocean-island basalts (OIB) unless depth of melting is greater than 300 km, or degree of partial melting is much higher than suggested on the basis of previous trace element modeling work (and sufficient to remove TiO₂ phases from solid residues). Likewise, the absence of strong depletion of Nb and Ta in OIB also provides constraints on degree of partial melting vs depth of partial melting for models where melting of ancient crust acts as an indirect source for OIB by metasomatic interaction with the mantle. The controlling influence of melt structure on partitioning behaviour of high-field strength elements (HFSE) implies that relative enrichment of Nb and Ta and reduction in Zr/Nb in high-pressure partial melts should occur even when TiO₂ phases are not present in solid residues. As such, depth of partial melting may be as important a factor as mineral and melt chemistry and degree of partial melting in constraining the composition of partial melts from Earth's deep interior.

Introduction

Subduction of oceanic lithosphere provides the main mechanism for recycling material back into Earth's deep interior, although the role of subduction in terms of geochemical recycling and melt generation in the mantle is poorly understood. It is a paradigm of mantle geochemistry that melting of ancient, recycled oceanic crust in the Earth's mantle is a key process in the genesis of ocean-island basalts (OIB) (Hofmann, 1997). Melting of ancient crust could provide a direct source for OIB through large-scale melting of subducted material accumulated in the lower mantle (Ohtani and Maeda, 2001), melting of entrained nodules in upwelling mantle plumes (Letich and Davies, 2001), or shallower melting of oceanic crust recycled in the upper mantle (Kogiso et al., 2003). Alternatively, deep melting of recycled ancient crust may act to metasomatise the mantle. It has recently been demonstrated from high-pressure experiments (Kogiso et al., 2003, Kogiso and Hirschmann, 2006) that partial melting of ancient, subduction-modified crust (garnet–pyroxenite) can produce nepheline-normative partial melts, and that in terms of major element chemistry, melts from ancient crust are a potential source for one of the major components necessary for the genesis of parental OIB magmas. In contrast, it has proven challenging to explain the trace element chemistry of OIB using existing data on trace element partitioning. Melting of ancient crust alone is unable to explain trace element arrays in OIB, particularly in terms of concentrations of high-field strength elements (HFSE) (Stracke et al., 2003). Currently, HIMU-type sources (one of the major components of OIB, possibly representing melts from ancient crust) can only be modeled by melting of recycled oceanic crust if it has undergone substantial modification during subduction. Modeling OIB arrays requires very small degrees of partial melting of substantially modified crust (around 1% by weight), varying degrees of aging, and mixing of various components (e.g. crust + sediment) (Stracke et al., 2003). This is in contrast with larger degrees of partial melting inferred from major element analysis. Alternatively, it has recently been suggested that melting of ancient crust may act as a secondary source for OIB primitive magmas by providing fluids which metasomatise the surrounding peridotite mantle, producing a pyroxenite lithology (Sobolev et al., 2007). Melting of this pyroxenite in, for example, upwelling plumes could then act as the direct source for OIB. This model is dependent upon the fact that melting of oceanic crust will result in formation of quartz-normative (silica-rich) partial melts.

A major problem in testing any model of OIB genesis is that data on trace element behaviour during partial melting of various possible protoliths under very high-pressure conditions is limited. Some insight may be gained by studying examples where direct melting of oceanic lithosphere occurs at shallower depths. Melting of oceanic crust may occur during subduction in geologically limiting circumstances where there is rapid subduction of relatively young, hot lithosphere, or early in Earth's history when mantle geotherms were considerably higher. Direct melting of subducting crust produces magmas that characteristically silica-rich and exhibit marked depletion in HFSE such as Nb and Ta. This depletion can be ascribed to the presence of the titania mineral rutile during partial melting (Foley et al., 2000). Rutile is unusual among rock-forming minerals in that it can contain large amounts of Nb, Ta and other HFSE, and its presence during melting controls the entire budget of such elements. Depletion in HFSE is not a common feature in OIB, which either implies that rutile is not stable in ancient crust during partial melting, perhaps due to the increased solubility of TiO₂ in garnet at high pressure (Zhang et al., 2003) or complete solubility of rutile in basaltic melts during high-degree partial melting (Ryerson and Watson, 1987), or implies that melts from ancient crust are not a direct source, or a major component, for OIB. Because rutile is a minor phase in eclogite, determining its stability is not straight-forward. Limited experimental data suggest that rutile should be stable in eclogite to at least 15 GPa under sub-solidus conditions, even in relatively Ti-poor systems (Okamoto and Maruyama, 2004). In fact, TiO₂ should be enriched in subducting crust during dehydration and alteration due to low solubility in fluids at high pressures (Audetat and Keppler, 2005), and may be further enriched in solid residues during partial melting (Klemme et al., 2005). There is also field evidence to suggest that rutile is an important minor phase in eclogite at very high pressures. The common occurrence of rutile in high-pressure eclogites and in kimberlite nodules has actually prompted the suggestion that rutile in ancient crust entrained deep in the mantle could have a major effect on partitioning of Nb between different geochemical reservoirs, and may explain the marked deficiency of Nb in the bulk silicate earth (BSE) (Kamber and Collerson, 2000a, Rudnick et al., 2000). However, it remains uncertain whether rutile will remain in residual material during partial melting of subduction-modified crust. TiO₂ solubility in melts varies significantly as a function of composition, especially silica content (Ryerson and Watson, 1987), and TiO₂ solubilities in basaltic melts (silica-undersaturated) are characteristically high. The reason why a rutile signature is sometimes produced by shallow hydrous melting of hot subducting slabs is that these melts are relatively silicic and exhibit much lower TiO₂ solubilities, meaning that rutile is more likely to remain in the solid residue for reasonable degrees of partial melting. OIB are typically silica-undersaturated and experimental data on phase relations during partial melting of Si-poor eclogite suggest that the stability of TiOphases in solid residues is constrained to very low degree partial melts at best (Kogiso et al., 2003, Kogiso and Hirschmann, 2006). This implies that if melting of ancient crust is a direct source for OIB, absence of a ‘rutile signature’ can simply be ascribed to an absence of rutile during partial melting of crust in the mantle source region. However, trace element modeling based on OIB chemistry suggests very low degrees of partial melting of ancient crust in the mantle, typically below 3 wt.% (Stracke et al., 2003, Willbold and Stracke, 2006). For such low degrees of partial melting, TiO₂ phases are expected to be present in solid residues. Furthermore, several additional factors favour the stability of TiO₂ during low degree partial melting at high pressures. The presence of CO₂ during partial melting results in a decrease in TiO₂ solubility in melts (Ryerson amd Watson, 1987), and rutile may be stable during melting of carbonated eclogite at high pressure up to at least 10% partial melting (Dasgupta et al., 2004). Rutile stability during melting also depends largely on the silica content of the melt, which in turn will depend on the bulk composition of subduction-modified crust. Models where melting of ancient crust results in mantle metasomatism are dependent upon the fact that melts are highly Si-enriched, which implies that rutile is also likely to be present during significant degrees of partial melting. The effect of pressure on TiO₂ solubility in melts is only constrained up to 3 GPa, with a slight negative correlation suggested (Ryerson amd Watson, 1987). At higher pressures, especially above 5 GPa, changes in TiO₂ solubility in silica melts will occur due to systematic changes in melt structure, especially in terms of the coordination of Si and degree of melt polymerisation. This should result in a decrease in TiO₂ solubility, because high-pressure melts are characteristically more polymerised than low-pressure melts, and extent of polymerisation of silicate melts is very much connected with a decrease in TiO₂ solubility. In fact, preliminary work conducted by the authors (Bromiley, unpublished results) suggests that rutile is stable during melting of model (volatile-free) subduction-modified crust to at least 10–15 wt.% melt at pressures from 5 to 15 GPa.

Although the influence of rutile on melt geochemistry is well constrained at pressures up to 3 GPa (i.e. of direct relevance to slab melting), there is an absence of data on higher pressure partitioning of HFSE between rutile and melt. Of particular importance in this regard is the fact that the rutile structure has a limited stability range. At pressures above 6 GPa, rutile undergoes a phase transition to an orthorhombic structure known as TiO₂(II). This transition is likely to influence mineral chemistry and trace element partitioning because it has a marked effect on the geometry of the main cation site, which undergoes significant deformation and ‘sees’ an increase in average Ti–O bond distance. Here we present results of an experimental investigation to determine the effects of pressure and the rutile–TiO₂(II) phase transition on partitioning of HFSE (Nb, Ta, Hf, Zr) between TiO₂ phases and basaltic melt under conditions relevant to OIB genesis.