Subduction zone Hf-anomalies: Mantle messenger, melting artefact or crustal process?

Abstract

The origin of Hf elemental depletions in subduction zone magmas is investigated using new major- and trace-element data for cumulate xenoliths from the Mariana arc, and deep sea sediments recovered by the DSDP and ODP drilling programmes. Results indicate that most of the rare earth element (REE) and Hf inventory in the xenoliths is contained within two minerals—clinopyroxene and titanomagnetite—and that removal of a typical gabbroic fractionating assemblage reduces the depletion in Hf relative to neighbouring REE on a mantle normalised trace element diagram (commonly denoted Hf/Hf*) in the evolving magmas. Confirmation of this observation is provided by a variety of literature data from different subduction zones in which bulk-rock samples also define a positive correlation between Hf/Hf* and the silica content of the magmas. In agreement with experimental studies on REE–HFSE partitioning, we observe that the ability of clinopyroxene to influence the Hf/Hf* of fractionating magmas is associated with its aluminium content. This decoupling of Hf from the REE in differentiating arc magmas suggests that bulk rock Hf/Hf* values, when used in isolation, are unlikely to provide a robust measure of source REE–Hf characteristics, even when suites are filtered to exclude all but the most mafic samples. It may be possible to normalise data to a constant degree of fractionation, and in this way distinguish subtle changes in source Hf/Hf* but most existing datasets are of neither the size nor quality to attempt such calculations.

Modification of Hf/Hf* is also seen when modelling mantle melting processes and there are strong suggestions that source variations are influenced by not only subducted sediment, which exhibits a remarkably wide range in Hf/Hf*, but also subduction zone fluids. These observations remove some of the constraints imposed on recent models that attempt to reconcile Hf isotope data with Hf–REE abundance data in some arc suites. Although a case may be made for the involvement of residual, minor, phases in the downgoing slab, fluid and sediment addition, and the role of major phases during partial melting, in particular clinopyroxene, in the mantle wedge can also exert a strong influence on Hf–REE relationships.

Introduction

The high field strength elements (HFSE)—specifically Nb, Ta, Zr, and Hf—have long held a special fascination for arc geochemists. The widely held assumption of relative immobility in fluids released from the downgoing plate has led researchers to the conclusion that HFSE abundances observed in arc magmas may provide the best opportunity to ‘see through’ the contaminating effects of the subduction component and thus investigate the nature the supra-subduction zone mantle wedge itself (e.g., McCulloch and Gamble, 1991). Debate continues however, surrounding the origin of the characteristic HFSE depletions preserved in arc magmas when compared to the concentration of trace elements of similar compatibility during mantle melting. Thus, for example, HFSE depletions have been attributed to variations in mantle fertility (e.g., Woodhead et al., 1993), retention of residual phases in the slab (e.g., Brenan et al., 1994), or reactions between the magma and mantle wedge during melt percolation (e.g. Baier et al., 2008, Kelemen et al., 1990).

The advent of MC–ICPMS technologies, allowing relatively straightforward Hf-isotope analysis offered an opportunity to further promote such studies by providing a means of tracking the actual sources and sinks of elemental Hf in arc magmas. In an intriguing twist, and contrary to the accepted view, Woodhead et al. (2001) presented evidence suggesting that slab-derived fluids themselves might be capable of transporting Hf (and by inference, other HFSE) into the mantle wedge. Subsequent work appears to support the idea of Hf mobility at least in sediment-derived melts (e.g., Hanyu et al., 2002, Tollstrup et al., 2010) although other studies document little evidence for Hf transfer in fluids (e.g., Jicha et al., 2004) and some show mixed results within an individual system (Barry et al., 2006).

In the last decade, attention has focussed on the occurrence of Hf elemental anomalies (relative to the REE) in arc lavas and their potential link to Hf isotope compositions and/or HFSE depletion. Two papers exemplify these discussions. Pearce et al. (1999) use negative Hf anomalies¹ to track the addition of a REE-bearing subduction component to the sub-arc mantle. They argue that, although the REE, including Nd, are added to the mantle wedge, they consider Hf to be immobile, or effectively so, with the consequence that Nd isotope ratios are modified in the mantle source region but Hf isotope ratios are not. In contrast, Tollstrup and Gill (2005) suggest that negative elemental anomalies of Hf in arc magmas require trace quantities of residual rutile, zircon and monazite to be stabilised in the subducting slab. Although these two studies are mutually consistent (i.e., requiring that REE are mobilised in a slab-derived flux, whereas Hf is immobile) it has proven difficult to reconcile the detailed trace element fractionations observed in some arc suites within isotope constraints. For example, in their investigation of volcanism at Anatahan Volcano in the Mariana Islands, Wade et al. (2005) note “…bulk sediment mixing explains the isotopes, but not the concentration anomalies” whereas “…zircon saturated sediment melts may explain concentration anomalies but not the isotopes”.

In our own studies of arc datasets we have observed that, where high precision (predominantly ICPMS) trace element data are available, and data are filtered to include only coeval suites of samples from individual volcanoes or volcanic centres, a correlation is often apparent between Hf elemental anomalies and indices of differentiation such as silica (e.g., Hergt and Woodhead, 2005). This first-order observation suggests that magmatic differentiation may have at least some role to play in the generation of such anomalies. Furthermore, the nature of the correlations (less negative anomalies occurring in more differentiated samples) is at odds with simple models involving crystallisation and removal of minor phases such as zircon in highly silicic samples. These discrepancies initiated the current study—to investigate whether it is possible that Hf anomalies (expressed herein as Hf/Hf*¹), even when measured in primitive samples, may track sources and/or processes distinct from Hf isotope compositions preserved in subduction zone lavas?