Multiphase petrography of volcanic rocks using element maps: a method applied to Mount St. Helens, 1980–2005

Abstract

Quantitative textural analyses including crystal size distributions (CSDs) provide insights into crystallisation kinetics of magmatic systems. Investigations of volcanic crystal textures often rely on greyscale variations on backscattered electron images to identify crystal phases, which must then be thresholded and/or traced manually, a laborious task, and investigations are typically restricted to a single crystal phase. A method is presented that uses energy-dispersive X-ray element maps to generate textural data. Each pixel is identified as a crystal phase, glass or vesicle according to relative chemical composition enabling concurrent acquisition of multiphase CSD, crystallinity and mineral mode data. Data processing is less time intensive for the operator but considerable instrument time is required to generate element maps. The method is applied to 17 dacite samples from the 1980–1986 and 3 from the 2004–2005 eruptive periods of Mount St. Helens volcano (USA) to provide quantitative insights into multiphase textural evolution. All of the CSDs are curved and concave-up in the standard CSD plot with curvature increasing with plagioclase content. To facilitate comparisons with previous studies, CSDs for microlites (<50 μm length crystals) are approximated as straight lines. The line intercepts and slopes provide information on n ⁰ (nucleation density) and characteristic length or Gτ (the product of growth rate (G) and residence time (τ)), respectively. These parameters, as well as the total groundmass crystallinity, show distinct differences between explosive deposits from summer 1980 and post-summer 1980 domes. Post-summer 1980 microlite n ⁰ values are mostly at the lower end of the range of those measured for summer 1980 samples. Total groundmass crystallinities during summer 1980 are between 10 and 30 vol.%, whereas post-summer 1980 crystallinity increases to between 39 and 51 vol.%. The range of n ⁰ values is similar to those previously published for Mount St. Helens but Gτ is consistently higher. Gτ of a May 1985 sample analysed in this study is approximately 2 μm higher compared with previously published data for the same sample when processed using similar methodologies. Groundmass crystallinity data show similar trends to those previously published for the 1980 to 1986 eruption, increasing sharply after summer 1980 then increasing more gradually during the dome-building phase of the eruption. The effects of varying L, the apparent crystal size, and crystal aspect ratio on resultant CSDs are also investigated. Whilst relative temporal variations in CSDs for a given dataset can be investigated, absolute values from different studies cannot be compared unless methods of data acquisition and processing are exactly the same.

Appendices

Appendix 1

MATLAB algorithm

EDS element maps were converted into phase maps using simple IF/AND/OR statements in MATLAB. An example isolating plagioclase using Al and Ca element map data is shown below:

Appendix 2

Effects of data processing

In CSD studies it is imperative that user-defined parameters are explicitly detailed to enable comparison with other datasets. Various definitions of apparent crystal size (Fig. 15a) and selection of appropriate crystal aspect ratios (Fig. 15b) have a profound effect on CSD data, as demonstrated using data from SH201, a Mount St. Helens spine sample erupted 24 May 1985. To facilitate future comparisons with our data, plots and tables of each CSD included in this study are provided in the Electronic supplementary material.

Fig. 15

Artifacts of CSD data processing using SH201 as an example. Different definitions of: a L and b crystal aspect ratio affect the CSDs form and maximum crystal size

Definition of apparent crystal size

For a given crystal population, using lengths of ellipses of equal area to crystal outlines rather than widths serves to increase the population density of the smallest defined size bin. For SH201, the population density of crystals is consistently lower for any given size bin when crystal widths are used instead of lengths, apart from at very small crystal sizes (<0.03 mm, Fig. 15a). Using L = square root of crystal area reduces the largest bin size to 160 μm compared with 600 μm when using length or width of rectangular fits (Fig. 15a).

Shape effects

Crystal shape (i.e. short/intermediate/long dimension) must be estimated to best-fit crystal populations when calculating CSDs using CSDCorrections. Freely available programs such as CSDSlice (Morgan and Jerram 2006) enable estimations of average aspect ratios for crystal populations. Here, several crystal shapes (1:1:1, 1:1:5, 1:3:5, 1:3:10 and 1:5:10) have been used to construct CSDs for SH201 to exhibit how varying aspect ratios affect the form of a CSD, particularly on the slope and the calculated largest bin size (Fig. 15b). Increasing the long dimension has the effect of decreasing n ⁰ and produces a flatter CSD slope, i.e. Gτ increases. n ⁰ increases from approximately 8.9 × 10⁶ mm⁻⁴ for a 1:5:10 aspect ratio to 1.8 × 10⁸ mm⁻⁴ for an aspect ratio of 1:1:1. Increasing the intermediate dimension increases the population density of crystals in the smallest size bin and steepens the gradient of the CSD, whereas the maximum crystal size bin decreases to smaller L.

Systematic changes in crystal shape with size from prismatic microlites to tabular or equant microphenocrysts may also contribute to the curved shape of CSDs; there is a low probability of sectioning the full length of an acicular shape whereas the probability of sectioning tabular crystals in orientations that will give true short/intermediate and short/long axis sections is much higher (Morgan and Jerram 2006). This phenomenon can contribute to an overestimate of small crystals resulting in curvature of the CSD.