When Is a Diffusion Profile Not a Diffusion Profile? the Importance of Initial State Assumptions in Diffusion Modelling

Abstract:

Over the past 20 years, diffusion chronometers have evolved from a niche tool into one of routine application, with more practitioners, new tools and increasingly large datasets. As we expand the horizons of diffusional geochronometry, it is worth taking stock of developments in methodologies and data acquisition, and taking time to revisit the underpinnings of the technique. Data collected as part of recent projects on Campi Flegrei, the Bishop Tuff and Fimmvörðuháls–Eyjafjallajökull are here used to investigate the initial state assumption, an absolutely vital aspect underpinning most diffusional work and one that is rarely evaluated despite its fundamental importance. To illustrate the nature of the problem we consider two widely-used element-mineral systems for felsic and mafic systems, respectively.

First, barium and strontium profiles within sanidine crystals, modelled independently, can give strongly contrasting timescales from the same crystal zone. We can reconcile the datasets only for a situation where the initial boundary within the crystal was not a sharp step function, but relatively fuzzy before diffusion onset. This fuzziness effectively starts both chronometers off with an apparent, and false, pre-existing timescale, impacting the slower-diffusing barium much more strongly than the faster-diffusing strontium, yielding thousands of years of non-existent diffusion history. By combining both elements, a starting width of tens of microns can be shown, shortening the true diffusive timescales from tens of thousands of years to hundreds.

Second, in olivine, we encounter different growth-related problems. Here, Fe-Mg interdiffusion occurs at a rate comparable to growth, with the compound nature of zonation making it difficult to extract the diffusion component. This requires a treatment of changing boundary conditions and sequential growth to generate the curvature seen in natural data, in order to recover timescales for anything but the outermost crystal zones with a high degree of confidence.

Diffusion methods continue to drive our understanding of short-timescale processes in volcanic systems, with faster and better techniques, but as this study indicates, the initial state assumption for each timescale must be well-constrained for the value of those advances to be fully realised.