Processes in upper-crustal magma reservoirs such as recharge, magma mixing, recycling of previously crystallized material, and eruption affect both the physical state and the chemical composition of magmas. A growing body of evidence shows that crystals in intermediate or silicic volcanic rocks preserve records of these processes that may be obscured due to mixing in the liquid fraction of magmas. Fewer studies have focused on crystals in basaltic lavas, but these show evidence for a more subtle, but still rich record of magmatic processes. We present new 238U–230Th–226Ra data for plagioclase, combined with δ18O and trace-element measurements of the same crystal populations, from basalts erupted at Krafla Volcanic Center, Iceland. These data document the presence of multiple crystal populations within each sample, with chemical and oxygen isotope heterogeneity at a variety of scales: within individual crystals, between crystals in a given population, between crystal populations within the same sample, and between crystals in lavas erupted from different vents during the same eruption. Comparison to whole-rock or groundmass data shows that the majority of macroscopic crystals are not in trace-element or oxygen isotope equilibrium with their host liquids. The most likely explanation for these data is that the macroscopic crystals originated within a highly heterogeneous crystal mush in the shallow magma reservoir system. U-series and diffusion data indicate that the crystals (and therefore the mush) formed recently (likely within a few thousand years of eruption, and with a maximum age of 8–9 ka), and that the crystals resided in their host magma prior to eruption for decades to a few centuries at most. These data, in conjunction with other recent studies, suggest a model where erupted Icelandic magmas are the result of diverse magmas entering the crust, followed by complex interactions between melts and previously crystallized material at all crustal levels.

High Mg# andesites and dacites (Mg# = molar Mg/Mg + Fe) from western Aleutian seafloor volcanoes carry high concentrations of Sr (>1000 ppm) that is unradiogenic (87Sr/86Sr < 0.7029) compared to lavas from emergent volcanoes throughout the arc (200–800 ppm Sr, 87Sr/86Sr >0.7030). Data patterns in plots of 87Sr/86Sr vs Y/Sr and Nd/Sr imply the existence of an eclogite-melt source component – formed by partial melting of MORB eclogite in the subducting Pacific Plate – which is most clearly expressed in the compositions of western Aleutian andesites and dacites (Nd/Sr and Y/Sr < 0.02) and which dominates the source budget for Sr in volcanic rocks throughout the arc. When viewed in combination with inversely correlated εNd and 87Sr/86Sr, these patterns rule out aqueous fluids as an important source of Sr because mixtures of fluids from altered oceanic crust with depleted mantle and sediment produce compositions with 87Sr/86Sr higher than in common Aleutian rocks. The unradiogenic nature of Sr in the western Aleutian andesite–dacite end-member may be understood if H2O required to drive melting of the subducting oceanic crust is transported in fluids containing little Sr. Mass balance demonstrates that such fluids may be produced by dewatering of serpentinite in the mantle section of the subducting plate. If the eclogite-melt source component is present throughout the Aleutian arc, melting of the subducting plate must extend into minimally altered parts of the sheeted dike section or upper gabbros, at depths >2 km below the paleo-seafloor. Oxygen isotopes in western Aleutian seafloor lavas, which fall within a narrow range of MORB-like values (δ18O=5.1–5.7), are also consistent with this model. These results indicate that the subducting Pacific lithosphere beneath the Aleutian arc is significantly hotter than indicated my most thermal models.