1. Supplementary Information: Materials and Methods
High pressure experiments
The aluminous pyroxene starting material enriched in 54FeO (31wt% MgO, 49wt% SiO2, 14wt% FeO, 7wt% Al2O3; cations per 3 oxygens: Mg 0.8, Si 0.85, Fe 0.2, Al 0.15) was prepared using sintered Mg, Si and Al oxides (Alpha Aesar) with Fe added as a 100:1 mixture of natural isotopic abundance FeO to 54FeO (> 96 % 54Fe enrichment, Oakridge National Laboratories, Tennessee), following the method and degree of isotopic enrichment of a previous study (Shahar et al., 2008). The oxide mixture was homogenised in several stages: firstly, the spike-natural FeO mixture was ground for 20 minutes under ethanol with a dedicated agate pestle and mortar and secondly, all oxides including FeO were mixed in stoichiometric proportions, and ground for 20 minutes under ethanol and dried three times. The mixture was reacted in graphite capsules (graphite furnace, BaCO3 cell with silica inner sleeve, W-Re thermocouple) at 1200°C and 2GPa for 2 hours in an end-loaded piston cylinder apparatus at the University of Oxford. Graphite capsules were employed to ensure all Fe was present as Fe2+ and that no oxidation took place during the runs. In order to obtain enough starting material, the products of several identical runs were combined. Any residual carbon derived from the graphite capsules was separated from the run products by a heavy liquid (tribromomethane) extraction, after which the sample was washed with ethanol. This step was repeated twice. The final pyroxene starting material was then ground with an additional 20 wt% powdered metallic Fe under ethanol and the mixture was reacted in a series of experiments at 1850°C and 24GPa for times up to 170 minutes to generate perovskite-metal assemblages. Experiments were quenched by turning off power to the furnaces. The experiments were run at these relatively low temperatures to prevent melting and segregation of the metal phase and to ensure that metal and silicate remained as discrete reservoirs for isotopic exchange. This approach also ensured that metal and silicate remained finely disseminated during the course of the experimental runs; the goal being to facilitate isotopic equilibrium between Fe metal and solid polycrystalline silicate. Longer run times were avoided due to the risk of experiment failure. For the 24GPa experiments we employed the multi-anvil apparatuses (1200 tonne and 5000 tonne) at the Bayerisches Geoinstitüt, Bayreuth, Germany. We used different size assemblies,10/4 (S4344, S4346) and 18/8 (Z651, Z652), with Cr-doped MgO octahedra, ZrO2 sleeves, LaCrO3 furnaces, W-Re D-type thermocouples and polycrystalline MgO sample capsules.
Characterisation of experimental run products
The polished sample capsules, mounted end-on in epoxy blocks, were initially imaged on the SEM at the University of Oxford (Jeol JSM-840A, images collected at 20kV). A typical run product is shown in Fig. 1. The experiments were subsequently analysed on the JEOL-JXA-8800R electron microprobe at the Begbroke centre, University of Oxford. Samples and appropriate silicate and metal standards were analysed using an accelerating voltage of 20 kV and a sample current of 20 nA. Perovskite and metal were the only phases present in the run products. We obtained approximate Fe3+/SFe ratios of the product perovskites by assuming a stoichiometry of 2 cations to 3 oxygens (Table 1). This yields results in good agreement with measured ferric/ferrous ratios from aluminous perovskites synthesized using similar methods (Frost and Langenhorst, 2002) and, as the orthopyroxene starting material contained (from stoichiometry) almost exclusively Fe2+, provides strong evidence for the formation of Fe3+-bearing silicate perovskite during the experimental runs. Due to the small size of the samples and the significant amounts of material required for isotopic analysis, we chose not to analyse the samples by Mössbauer spectroscopy as the potential exists for significant material to be lost during sample preparation.
Phase separation and Fe isotope analysis
Following characterisation, the sample epoxy blocks were cleaned by lightly polishing and by treatment with isopropyl alcohol in an ultrasonic bath for 40 minutes. Due to the highly disseminated nature of the metallic iron in the samples and the small size of the silicates produced, it was not possible to selectively drill out individual metal and silicate phases. Instead, the entire experiment was drilled out using a vertically mounted dentists’ drill and microdiamond-coated drill bits. The procedural blank for drilling was estimated by drilling pure MgO (capsule material) and is <20ng Fe. Metal and silicate were separated from each other using a selective leaching procedure. This procedure was applied to the entire experiment, thus negating any spatial variations in metal and silicate isotopic composition and hence fractionation factor within the sample capsules. Leaching was carried out in Teflon (Savillex®) using ultra-pure distilled acids where at each stage the sample was decanted into a cleaned polypropylene micro-centrifuge tube and the supernatant removed. Previous tests of this leaching procedure on both metallic Fe (Alpha Aesar 200-400 mesh) and silicates (the USGS standard BCR-1) demonstrate that it does not produce analytical artifacts (Table 2). The leaching stages were: i) 1M HCl, cold, 20minutes; ii) 2M HCl, cold, 20minutes, ultrasonic 2 minutes; iii) 2M HCl, ultrasonic 60°C, 25 minutes; iv) HF-HNO3 ultrasonic 25 minutes; hotplate 110°C, 2 hours; v) HF-HNO3 ultrasonic 25 minutes; hotplate 130°C, 10 hours, followed by aqua regia, 12 hours hotplate 130°C. In general, minimal Fe was released during from the first leach (1M HCl) the “metal fraction” was obtained during the second leach (2M HCl); a metal-silicate mixed fraction was obtained from the third leach (2M HCl); a clean “silicate fraction” was obtained from the third leach (HF-HNO3) and the final, residual leach was dominated primarily by previously undecomposed furnace and capsule components. The estimated blanks from leaching were 4 and 6 ng Fe for the “metal” and “silicate” fractions, respectively.
The supernatant fractions of the individual leaches were evaporated down and oxidised with aqua regia (130°C, 3 days) prior to two evaporation and reflux cycles with 6M HCl (140°C, 2 days). Prior to column chemistry, small aliquots of the 6M HCl fractions were reserved for concentration measurements. Iron was quantitatively purified by anion exchange (AG1X4 200-400 mesh, HCl form) column chromatography using established methods (Williams et al., 2004). Following evaporation and treatment with concentrated H2O2 and HNO3 to decompose any resin particles from the columns, the samples were dissolved in 0.01 M HNO3. As documented in (Williams et al., 2004), Fe yields were quantitative and no measurable Fe isotope fractionation took place on the columns following these procedures. The combined blank for sample dissolution and column chromatography was <10 ng Fe; this is negligible relative to the amounts of Fe extracted from the samples: 180-300 μg Fe for metals and 120 to 240 μg Fe for silicates.
Iron isotope analyses were carried out on two different mass spectrometers (Thermo Neptune and Nu Instruments Nu Plasma HR) at two different institutions (Neptune – Durham University; Nu – Oxford University) in order to ensure that there were no inter-laboratory biases in isotopic determinations. For both instruments, samples were analysed in 0.01 M HNO3 and mass bias was corrected for by sample-standard bracketing. On both the Nu and Neptune, sample solutions were 3.5ppm Fe; sample and standard Fe beam intensities were matched to 5%. Summarised isotope ratios for samples (Table 1), chemistry tests and experimental starting materials (Table 2) incorporate measurements carried out at both at Oxford and Durham; measurements on the individual instruments including internal standards are given in Table S.1.
Isotopic analyses on the Nu Plasma were carried out at a resolving power of ~9000 where the source and alpha slits were narrowed to allow for the partial resolution of ArO+ and ArN+ from 56 and 54 Fe, respectively. Samples were introduced using a PFA nebuliser (75 μl/minute; ESA Scientific) and a Nu Instruments desolvator (DSN) and type “B” Ni skimmer and sampler cones were used. The instrument was optimised to permit collection of 57, 56 and 54 Fe and 53 Cr, to allow for correction of any interference of 54Cr on 54Fe. There are no significant differences between Cr-corrected and uncorrected ratios (Table S.1); due to the purification of the samples prior to analysis and effective separation of Cr from Fe. Analyses were carried out on the low-mass shoulder of the aligned peaks, and consisted of a 30 second electronic background reading (ESA deflection) followed by 25 8 second integrations and two 90 second washes, the first in 0.1 M HNO3, the second in 0.01 HNO3. In order to conserve sample, the magnet was only centred on the bracketing standards, i.e. before and after the sample run. Magnet drift was generally <20ppm (DAC units) between bracketing standards, which is negligible relative to the size of the interference-free peak shoulder (~180 ppm). Mass dependence as well as reproducibility and accuracy were evaluated by analysis of an in-house Fe standard, FeCl, obtained from ETH-Zürich (ETH Fe salt standard). The values obtained for this standard on both the Oxford Nu and the Durham Neptune (Table S.1) are in excellent agreement with both each other and with published values (Mikutta et al., 2009).
Isotopic analyses on the Durham Thermo Neptune were carried out in medium resolution mode (resolving power 8000-9000). Samples were introduced using a PFA nebuliser (50 μl/minute with adjustable “ST” capillary; ESA Scientific) and an Apex heated quartz spray chamber without desolvation membrane. A Ni “H”-cone was used. Analyses were carried out on the low-mass shoulder of the aligned peaks, and consisted of a 30 second electronic background reading (ESA deflection) followed by 30 4 second integrations and two 90 second washes, both in 0.01M HNO3. Masses 57, 56 and 54 (Fe) and 53 (Cr) were simultaneously collected, the later to allow for correction of any interference of 54Cr on 54Fe. As for the Nu plasma, no significant differences were observed for Cr-corrected and Cr-uncorrected ratios (Table S.1). Measurements were carried out without centring the magnet; the magnet was found to drift <5ppm over the course of 5 days.
Aliquots of the leach solutions taken at the 6M HCl stage prior to column chemistry (see above) were evaporated and converted to 0.01 M HNO3 for concentration determinations on the Thermo Element2. Thermo employing a 100µl/minute microflow nebuliser fitted to a cyclone Scott double pass spray chamber. Samples were analysed at low resolution for the isotopes 24, 25, 26Mg, 27Al, 50, 52, 53, 54Cr, 54, 56, 57, 58Fe, 58, 60, 61, 62Ni. Each sample was analysed three times. Standards were made from Romil 1000ppm stock solutions and diluted to cover the expected concentration ranges of the samples. Standards were analysed at the start, throughout and at the end of the analysis session. Instrument sensitivity was optimised to give low oxide generation whilst maintaining high overall sensitivity. Typically CeO/Ce was <2.5% with 1ppb In giving signals >1x106cps. In order to monitor sample memory effects a wash solution was analysed and recorded after each analysis. As the leached samples could not be weighed, we cannot report direct concentrations, only ratios such as Al/Fe (calculated using 57Fe, as this isotope is free of isobaric interferences on the Element2). The ratio Al/Fe can be used as a rough measure of the relative contributions of silicate and metal phases to the leaches as Al behaves as a lithophile element under all experimental conditions, although it does not permit precise quantification of metal:silicate ratios as Fe and Al may be released at different rates during the leaching procedure. Although Mg concentrations were measured in addition to Al and Fe, we did not use the ratio Mg/Fe to quantify the relative contributions of the silicate and metal phases to the leach fractions due to the potential contribution of capsule MgO to the Mg budget of the leaches.
Frost, D.J., Langenhorst, F., 2002. The effect of Al2O3 on Fe-Mg partitioning between magnesiowustite and magnesium silicate perovskite. Earth and Planetary Science Letters, 199(1-2): 227-241.
Mikutta, C. et al., 2009. Iron isotope fractionation and atom exchange during sorption of ferrous iron to mineral surfaces. Geochimica et Cosmochimica Acta, 73: 1795-1812.
Shahar, A., Young, E.D., Manning, C.E., 2008. Equilibrium high-temperature Fe isotope fractionation between fayalite and magnetite: An experimental calibration. Earth and Planetary Science Letters, 268(3-4): 330-338.
Williams, H.M. et al., 2004. Iron isotope fractionation and the oxygen fugacity of the mantle. Science, 304(5677): 1656-1659.
Table S1: Isotope data for phase separation leaches and internal standards
sample / treatment / measurement / d 57/54Fe Cr / 2SD / d 56/54Fe Cr / 2SD / d 57/56Fe / 2SD / d 57/54Fe raw / 2SD / Al/Fe / 2SDStandards
FeCl (ETH salt) / full dissolution / Ox (n=48) / -1.06 / 0.16 / -0.73 / 0.10 / -0.32 / 0.12 / -1.08 / 0.16
FeCl (ETH salt) / full dissolution / Dur (n=12) / -0.99 / 0.11 / -0.67 / 0.12 / -0.31 / 0.07 / -0.98 / 0.12
published value / ETH (n=210) / -1.08 / 0.13 / -0.72 / 0.13
Experiments
Z651 metal / 2M HCl / Ox (n=5) / -5.02 / 0.11 / -5.40 / 0.08 / 0.37 / 0.12 / -5.04 / 0.11 / 0.000 / 0.048
Z651 metal / 2M HCl / Dur (n=2) / -5.04 / 0.07 / -5.38 / 0.03 / 0.32 / 0.02 / -5.05 / 0.06 / 0.000 / 0.048
Z651 metal / 2M HCl / average / -5.03 / 0.02 / -5.39 / 0.03 / 0.35 / 0.06 / -5.04 / 0.02 / 0.000 / 0.048
Z651 silicate / HNO3-HF / Dur (n=4) / -31.06 / 0.08 / -32.97 / 0.03 / 1.97 / 0.07 / -30.99 / 0.21 / 0.989 / 0.024