Iron K-edge X-ray absorption near-edge structure spectroscopy of aerodynamically levitated silicate melts and glasses

The local structure about Fe(II) and Fe(III) in silicate melts was investigated in-situ using iron K-edge X-ray absorption near-edge structure (XANES) spectroscopy. An aerodynamic levitation and laser heating system was used to allow access to high temperatures without contamination, and was combined with a chamber and gas mixing system to allow the iron oxidation state, Fe3+/ΣFe, to be varied by systematic control of the atmospheric oxygen fugacity. Eleven alkali-free, mostly iron-rich and depolymerized base compositions were chosen for the experiments, including pure oxide FeO, olivines (Fe,Mg)2SiO4, pyroxenes (Fe,Mg)SiO3, calcic FeO-CaSiO3, and a calcium aluminosilicate composition, where total iron content is denoted by FeO for convenience. Melt temperatures varied between 1410 and 2160 K and oxygen fugacities between FMQ – 2.3(3) to FMQ + 9.1(3) log units (uncertainties in parentheses) relative to the fayalite-magnetite-β-quartz (FMQ) buffer. Remarkably, XANES pre-edge peak areas imply mean Fe-O coordination numbers (nFeO) close to 5 in all cases, with only a slight tendency toward higher values in the most iron rich melts, suggesting an intermediate role for both Fe(II) and Fe(III) in terms of network formation. End member coordination numbers for Fe(II)-O and Fe(III)-O are estimated to be similar, having means (and standard deviations) of 5.0(2) and 4.9(1), respectively. As such, the preference for ferric iron to occupy lower coordination sites than ferrous is weak, in contrast to published behavior in some alkali-rich systems, which may explain the larger published viscosity variations with Fe3+/ΣFe in alkali-, compared to alkaline earth-iron silicates. Temperature effects on nFeO are inferred to be small based on the melt data, as well as by comparison to glasses formed on quenching. Positive shifts of the pre-edge peak centroids observed in many cases on quenching are attributed to rapid oxidation enabled by the stirring of the melt droplets by the levitation gas jet. Fe3+/ΣFe values were estimated from XANES pre-edge peaks using published calibrations and compared to semi-empirical thermodynamic model calculations and Mössbauer measurements on quench products. Whilst showing positive correlation, the comparisons highlight the limitations involved in applying XANES calibrations and models for Fe3+/ΣFe derived from measurements on glasses, to high temperature basic melts. Fe3+/ΣFe varies from approximately zero up to about 65% in the high temperature melts and 75% in the glasses.