Refining crystal size distributions and kinetic histories using automated scanning electron microscopy and manual methods: a hybrid approach

Abstract

Igneous rock textures are directly tied to pre-, syn- and post-eruptive processes that occur throughout the evolution of a melt. One aspect of texture concerns crystal size distributions (CSDs), which are volume-based, individual counts of crystals from a defined volume of rock. CSDs provide insight into the kinetic evolution of igneous rocks. Direct measurements of crystal volumes from rock samples are often impractical, forcing the use of crystal lengths from thin sections as proxies for crystal volumes. We use a combination of manual tracing and automated SEM-based methods to delineate crystal borders while software (ImageJ, CSDSlice and CSDCorrections) addresses direct length acquisition and stereological conversions. Using texturally diverse lava samples, we demonstrate that this hybrid approach 1) preserves the crystal-length accuracy of manually traced phenocrysts while 2) refining groundmass crystal lengths through automated SEM-based analysis. High-resolution SEM-based imaging produces groundmass crystal length that refine nucleation densities when compared to back-extrapolations from manually-derived CSD curves. Using energy-dispersive X-ray spectroscopy to produce crystal images permits population size characterization by geochemistry while electron backscatter diffraction images allow detection of discrete crystals within complexly twinned plagioclase crystal clusters using lattice orientation. CSD curves from lava samples suggest that manual methods often produce reduced and misleading nucleation densities and shallower CSD slopes, which in turn decrease apparent average residence times. The hybrid method preserves the integrity of the portions of CSDs represented by manually traced crystals while introducing improved resolution of CSDs at the finer size fraction.