The Origin of Rhyolitic Magmas at Krafla Central Volcano (Iceland)

Abstract

Abstract We present a detailed petrologic study of rhyolites from seven eruptions spanning the full (∼190 ky) history of rhyolitic volcanism at Krafla volcano, northeast Iceland. The eruptions vary widely in size and style, but all rhyolites are crystal-poor (<6 modal%: plagioclase + augite ± pigeonite ± orthopyroxene ± titanomagnetite ± fayalite) and have similar evolved compositions (73.7–75.8 wt% normalized whole-rock SiO2) and trace element patterns. Macrocryst rim compositions from each eruption cluster within a narrow range and are appropriate for equilibrium with their carrier melt. Crystal cores and interiors display complex growth patterns and commonly host resorption surfaces, but compositional variations are slight (e.g. typically <10 mol% An for plagioclase, Mg# <10 for pyroxene), and consistent with an overall trend of cooling and differentiation by crystal fractionation. Although most crystal core and interior compositions are broadly appropriate for equilibrium with melts similar to their host whole-rock, variable growth histories, juxtaposition of grains with distinct trace element compositions, and scatter in melt inclusion compositions indicate mixing of antecrysts from compositionally similar evolved melts and/or assimilated felsic mush or intrusions before final rim growth. Evidence for mafic recharge (e.g. coupled increases in An and Fe in plagioclase) is absent in most crystals; rhyolite storage and fractionation thus occurred largely in isolation from the underlying mafic system. Comparison of observed matrix glass compositions with published experimental work on melting of altered (meta)basalts casts doubt on previous models favouring rhyolite generation by partial melting of hydrothermally altered basalts, instead supporting recent isotopic and modelling arguments for a crystallization-driven process [Hampton, R. L. et al. (2021). Journal of Volcanology and Geothermal Research 414, 107229]. MELTS fractional crystallization and assimilation-fractional crystallization (AFC) models at 1 kbar predict liquid major and trace element compositions similar to Krafla rhyolites after ∼60–70 vol% crystallization of a quartz tholeiite melt representative of the evolved crystal-poor basalts commonly erupted within Krafla caldera. We thus suggest that stalling and crystallization of these evolved basalts at shallow depth forms crystal mushes from which evolved (broadly dacitic to rhyolitic) melts are extracted. These melts ascend and mix with other compositionally similar melt bodies and/or assimilate felsic intrusive material in the uppermost crust. The Daly gap between ∼57 and 71 wt% SiO2 at Krafla is consistent with preferential extraction of evolved melts from quartz tholeiite mushes in the ∼50–70% crystallinity window. Residual solid (cumulate) compositions predicted by MELTS are exclusively mafic, hence efficient silicic melt extraction from quartz tholeiite mushes may also explain the apparent compositional bimodality in some Icelandic plutonic suites.