Choice of 17O Abundance Correction Affects Δ47 and Thus Calibrations for Paleothermometry

The clumped isotope composition of CO2 derived from carbonate (Δ47) varies with temperature, making it a valuable geothermometer with broad applications. However, its accuracy is limited by inter-laboratory discrepancies of carbonate reference materials and disagreement among Δ47-temperature calibrations. Here we use a suite of CO2-H2O equilibrations at known temperatures with a wide range in 13C and 18O compositions to show how the correction for the abundance of 17O impacts Δ47 and potentially explains these discrepancies. When a traditional value of 0.5164 is used for the fractionation between 17O and 18O (λ), corrected Δ47 in 23 °C CO2-H2O equilibrations exhibits a dependence on 13C composition that is equivalent to 20 °C (Δ47 range of 0.06 ‰). In contrast, these discrepancies are effectively removed when λ=0.528, as in global meteoric waters. Furthermore, carbonate standards with identical formation temperatures have significantly different Δ47 when corrected using λ=0.5164, but agree within error when λ=0.528. The choice of λ affects the accuracy of all sample Δ47 values, unless the sample CO2, mass spectrometer reference gas, and equilibrated gases share the same 13C composition. The sensitivity of Δ47 to the choice of λ, and the apparent dependence on 13C when 0.5164 is used, are relevant to the abiogenic experiments used in Δ47-temperature calibrations given that precipitation methods involving CO2 bubbling produce carbonates depleted in 13C by tens of permil relative to methods that mix salts. We evaluate the influence of 17O correction on Δ47-temperature calibrations using a suite of 58 abiogenic carbonates precipitated at 4-85 °C using CO2 bubbling and the mixing of salts. Aliquots of precipitated carbonates were digested at 25 and 90°C, but all other preparatory and analytical variables were held constant. When corrected using λ=0.5164, various precipitation methods yield sub-parallel Δ47-temperature relationships with slopes of 0.034-0.044 (x 106/T2), but offset intercepts. Conversely, Δ47-temperature relationships overlap within error when λ=0.528. This suggests that the method used to correct for 17O abundance likely contributes to observed calibration discrepancies and that adopting λ=0.528 may reduce the uncertainty in Δ47 temperature reconstructions.