X series 2

Whole rock fragments of D’Orbigny, Sah99555 and NWA1670 weighing around 1 g were mildly crushed with an agate mortar and pestle, rinsed in acetone and distilled water and sieved. Whole rock samples of approximately 20 mg were taken from the 250–500 µm size fraction of D’Orbigny and Sah99555. Feldspar separates were generated from the remaining size fraction of these two meteorites with a Frantz isodynamic magnetic separator whereas olivine and pyroxene separates were handpicked from the crushed bulk sample. Due to the cryptocrystalline nature of the NWA1670 groundmass, only olivine and cryptocrystalline ground-mass separates of a few mg each were handpicked. Samples were digested using HF–HNO₃ acid mixtures and, after complete dissolution, Al/Mg ratios were determined on an aliquot of the digested sample to 2% accuracy using a ThermoFisher X-Series II inductively coupled plasma source mass spectrometer (ICPMS).
Magnesium was purified by ion-exchange chromatography and its isotopic composition analyzed using a ThermoFisher Neptune Plus multiple collector ICPMS (MC-ICPMS) equipped with a Sampler Jet and Skimmer X-cone and an Apex sample introduction system following protocols outlined in Bizzarro et al. (2011) . The Mg isotope composition was measured in high resolution mode using a 50 μm entrance slit (M/ΔM > 5000). At an uptake rate of ~50 μL/min, the sensitivity of the instrument was ~200 V/ppm. Depending on sample size signal intensities were typically 100 V or 25 V on mass ²⁴Mg (see Table 1). Faradays cups that received signals larger than 50 V were connected to an amplifier with a 10¹⁰ Ω feedback resistor, whereas Faradays cups collecting smaller signals were connected to an amplifier with a 10¹¹ Ω feedback resistor. Single analyses comprised 1667 s of data acquisition and each sample was bracketed by standard analyses and analyzed ten times when permissible. Mg isotope data are reported in the μ-notation as relative deviations from the DTs-2b standard (µ²⁵Mg_DSM-3 = −122 ± 17 ppm (2 sd); Bizzarro et al., 2011 ) according to the following formula: $${\mathrm{\text{μ}}}^{\text{25}}\text{Mg=}\left[\frac{{{(}^{25}{\text{MG/}}^{\text{24}}\text{Mg)}}_{sample}}{{{(}^{25}{\text{Mg/}}^{\text{24}}\text{Mg)}}_{DTS-2b}}-1\right]\times {10}^6.$$
The mass-independent component in ²⁶Mg (µ²⁶Mg*) is reported in the same fashion, but represents deviations from the internally normalized ²⁶Mg/²⁴Mg of the sample from the reference standard, normalized to ²⁵Mg/²⁴Mg = 0.126896 using the exponential mass fractionation law. We used the exponential law (β = 0.511) to correct for instrumental mass bias and potential natural mass fractionation experienced by the samples. Two lines of evidence indicate that this approach yields accurate ²⁶Mg* values. First, the rock samples analyzed here are rapidly cooled magmatic lavas and, under such conditions, equilibrium mass fractionation is believed to be negligible (Teng et al., 2007 ). Second, instrumental mass fractionation under typical analytical conditions in our laboratory is driven by kinetic processes rather than equilibrium (i.e., Bizzarro et al., 2011 ). As such, variable instrument-related mass fractionation in samples relative to the standard can be accurately corrected for using our approach. We note that, similar to earlier studies (Spivak-Birndorf et al., 2009 ; Schiller et al., 2010 ), stable isotope variability exists between individual mineral fractions, typically for samples containing little Mg. We infer that this reflects analytical artifacts in the stable isotope data during analysis by MC-ICPMS induced by the presence of residual organics accumulated in the purification process rather than true isotope variability. Thus, we caution that the reported internal uncertainties for ²⁵Mg do not reflect the true reproducibility of the data, which may be significantly larger for small samples.
All Mg data reduction was conducted off-line using Iolite (Paton et al., 2011 ) and changes in mass bias with time were interpolated using a smoothed cubic spline. For each analysis, the mean and standard error of the measured ratios were calculated using a 3 sd threshold to reject outliers. Individual analyses of a sample were combined to produce a weighted average by the propagated uncertainties of individual analyses and reported final uncertainties are the 2 se of the mean. ²⁶Al–²⁶Mg isochrons were calculated based on ²⁷Al/²⁴Mg and Mg isotope data reported in Table 1 using Isoplot applying an external reproducibility of the μ²⁶Mg* data of 2.5 ppm for samples analysed systematically ten times at signal intensities of ~80 V for ²⁴Mg (Bizzarro et al., 2011 ). Samples measured at intensities of ~25 V for ²⁴Mg yielded a larger external reproducibility of 5 ppm that is predicted by counting statistics for approximately four times lower signals.