Optimization of lithium diffusion modelling in plagioclase: implications for the assessment of pre-eruptive timescales

Abstract

Lithium diffusion chronometry in plagioclase is used to retrieve timescales of pre-eruptive processes that occur over seconds to hours. The random orientation of 2D crystal sections exposed for microanalysis can strongly affect modelling results, leading to flawed interpretations of pre-eruptive timescales. To evaluate the effect of random crystal orientation, we systematically applied and compared the results of one-dimensional (1D) and multidimensional (analytical 3D correction) Li diffusion modelling to multiple oriented sections obtained from a set of eight plagioclase crystals from Stromboli volcano (Italy). For each crystal, Li concentration profiles were acquired using laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS) on the section oriented parallel to (010) and on a sequence of three to four sections oriented perpendicular to (010), starting from the core-intersecting section. The diffusion model accounts for the analytical uncertainties in Li measurements and estimates the optimal timescale within a set of 1000 possible values through bootstrapping. Timescales obtained from Li concentration profiles acquired on sections perpendicular to (010) are maximum in the sections intersecting the crystal cores and decrease markedly within each crystal (by up to 93% for simple 1D modelling and up to 85% after 3D correction) with increasing distance from this section. In turn, timescales derived from sections parallel to (010) are systematically lower than those obtained for the theoretically equivalent section perpendicular to (010). This means that large underestimates of the timescales are introduced by the analysis of 2D crystal sections with relatively small offset from the core-intersecting section, also in the case of plagioclase exposing sections parallel to (010), where the offset from the core-intersecting section cannot be well evaluated on a morphological basis. The application of the analytical 3D correction to diffusion profiles in part compensates for the underestimates derived from the offset relative to the core-intersecting section. Yet, the choice of an adequate plagioclase crystal in 2D sections of natural rocks remains a major source of uncertainty, which adds to other uncertainties related to the choice of diffusion coefficients and Arrhenius parameters. The collective impact of these uncertainties is discussed based on the comparison between Li diffusion timescales in plagioclase obtained in this and previous studies on Stromboli volcano. The implemented methodology for Li diffusion modelling presented in this study allows a more robust assessment of pre-eruptive timescales and is applicable to other volcanic systems.