Elsevier

Geochimica et Cosmochimica Acta

Volume 91, 15 August 2012, Pages 32-39
Geochimica et Cosmochimica Acta

Lithium defects and diffusivity in forsterite

https://doi.org/10.1016/j.gca.2012.05.034Get rights and content

Abstract

Lithium is an important geochemical tracer used to infer the thermal and chemical evolution of minerals in the Earth’s upper mantle. Knowledge of point defect chemistry and diffusion is critical for the interpretation of Li distribution in minerals. Using quantum mechanical methods we show that in forsterite Li will be incorporated as bound interstitial–substitutional pairs. Furthermore, there will be temperature dependent fractionation of its two isotopes between the different sites. The fractionation decreases dramatically from 87.1‰ at 300 K to 1.0‰ at 3000 K. Diffusion is predicted to occur via two inter-related mechanisms: Mg–Li exchange, and a second, vacancy assisted interstitial mechanism. This behaviour is complex, facilitates migration of the heavier isotope and offers insights into observations of Li mobility and zoning in olivine, the most volumetrically important upper mantle mineral.

Introduction

The distribution of impurities in minerals can provide important clues to thermal and chemical processes occurring in the Earth’s mantle over geological time. Lithium (Li) is increasingly used as a geochemical tracer as its two isotopes (6Li and 7Li) have a large relative mass difference and thus fractionation can lead to compositional variations at low temperatures. At higher temperatures, isotopic fractionation is less pronounced, although high temperature diffusion can lead to variations in observed isotopic signatures in mantle minerals. The importance of this process is increasingly recognized as shown by recent modeling of Li partitioning between co-existing phases (Gallagher and Elliott, 2009). The results of this work show that fractionation between melt and crystals of clinopyroxene is driven by cooling. Thus there is considerable scope for the use of Li isotopes in geospeedometry.

Within the Earth’s mantle, the largely incompatible Li is expected to be hosted by olivine [(MgFe)2SiO4] (Seitz and Woodland, 2000, Ottolini et al., 2004) and can provide insights into changing conditions during crustal recycling processes (Elliott et al., 2004). The mechanism by which Li and other monovalent cations are incorporated into olivine and their diffusivities however, are poorly understood. Recent research has shown that the distribution of Li isotopes in co-existing phases such as olivine and pyroxene, can vary greatly (e.g. Magna et al., 2006, Jeffcoate et al., 2007) and, in addition, there can be major variations within a single grain. Further complications arise in response to “matrix effects” described by Bell et al. (2009), resulting from small chemical variations that change local environments in the crystal lattice. Li is known to be highly mobile in plagioclase (Richter et al., 2003) and clinopyroxene (Richter et al., 2003, Coogan et al., 2005) with diffusivities that are orders of magnitude faster than cations such as Mg in the same systems. Recent experiments (Dohmen et al., 2010) provided the first real clues as to the controls on Li diffusion rates in natural olivines, although the mechanisms responsible for diffusion are hard to interpret unambiguously.

In order to fully interpret chemical signatures and model processes occurring over long timescales, a chemical understanding of defect and diffusion processes in minerals such as olivine is needed. The aim of the current work is to provide an insight at the atomic level into diffusive mechanisms operating in olivine and their influence on Li isotopic variability in mantle rocks. Thus we have carried out calculations, primarily using first principles methods, to model Li defects and migration in forsterite, the Mg end member olivine. Our results complement those from a recent experiment (Dohmen et al., 2010) and provide an insight into processes occurring at the atomic level, not accessible by other means. The present work provides quantitative estimates of the stability of Li in forsterite, and the corresponding substitution and migration mechanisms. Additionally, we investigate the possibility of Li isotopic site fractionation. The results enhance our knowledge of the behaviour of chemical exchange, material transport and electrical conductivity in the Earth’s upper mantle, and to interpret observed isotopic variations in real rocks.

Section snippets

Methodology

Based on density functional theory (DFT), our ab initio simulations were performed with the CASTEP code (Segall et al., 2002, Clark et al., 2005). The Perdew Burke Ernzerhof (PBE) functional was used with the generalized gradient approximation (GGA) for all calculations (Perdew et al., 1996). Ultrasoft pseudopotentials that we used have the valence-electron configurations of 1s22s1(core radius 1.86 a.u.) for Li, 2p63s2 (core radius 2.06 a.u.) for Mg, 2s22p4 (core radius 1.0 a.u.) for O, 3s23p2

Defect species

Li can be incorporated into the pure forsterite lattice as either an interstitial or as a substitutional defect at an Mg site, both of which require charge compensation. In Kröger–Vink defect notation these are denoted (Lii,) and (LiMg), for the interstitial and substitutional configurations respectively. Charge neutrality of Li incorporated in the pure forsterite lattice involves combinations of interstitials and substitutions that are described as:

  • (a)

    Lii+LiMg, where the interstitial is

Conclusions

Lithium isotopes migrate and fractionate through forsterite via a complex vacancy assisted interstitial mechanism that has been simulated using computational methods. We have shown that the most stable route for incorporation of Li in the forsterite lattice is via formation of Li interstitial/substitutional pairs. The differential partitioning of Li isotopes between these two sites is the prime mechanism for Li isotopic fractionation in the major near-surface mineralogical Li reservoirs, which

Acknowledgements

We thank the two anonymous reviewers, Dr. Ralf Dohmen and the associate editor, Dr. Michael Toplis, for their constructive comments. This work was carried out with the support of the Australian Research Council (DP0878453), with computational resources from iVEC and the NCI National Facility in Canberra, Australia, which are supported by the Australian Commonwealth Government. F.Z gratefully acknowledges Julian Gale for the fruitful discussion and help.

References (28)

Cited by (0)

View full text