For anchoring CO2 isotopic measurements on the δ18OVPD-CO2 scale, the primary reference material (NBS 19 calcite) needs to be digested using concentrated ortho-phosphoric acid. During this procedure, great care must be taken to ensure that the isotopic composition of the liberated gas is accurate. Apart from controlling the reaction temperature to ±0.1°C, the potential for oxygen isotope exchange between the produced CO2 and water must be kept to a minimum. The water is usually assumed to reside on the walls in the headspace of the reaction vessel. We demonstrate here that a large fraction of the exchange may also occur with water inside the acid. Our results indicate that both exchange reactions have a significant impact on the results and may have largely been responsible for scale inconsistencies between laboratories in the past.

The extent of CO2/H2O oxygen exchange depends on the concentration (amount of free water) in the acid. For acids with a nominal H3PO4 mass fraction of less than 102%, oxygen isotope exchange can create a substantial isotopic bias during high-precision measurements with the degree of the alteration being proportional to the effective isotopic contrast between the acid and the CO2 released from the calcite. Water evaporating from the acid at 25°C has a δ18O value of −34.5‰ relative to the isotopic composition of the whole acid. This large fractionation is likely to occur in two steps; by exchange with phosphate, water inside the acid is decreased in oxygen-18 relative to the bulk acid by ∼ −22‰. This water is then fractionated further during evaporation. Oxygen exchange with both water inside the acid and water condensate in the headspace can contribute to the measured isotopic signature depending on the experimental parameters.

The system employed for this study has been specifically designed to minimize oxygen exchange with water. However, the amount of altered CO2 for a 95% H3PO4 at 25°C still accounts for about 3% of the total CO2 produced from a 40 mg calcite sample, resulting in a δ18O range of about 0.8‰ when varying the δ18O value of the acid by 25‰. Least biased results for NBS19-CO2 were obtained for an acid with a δ18O value close to +23‰ vs. VSMOW. In contrast, commercial acids from several sources had an average δ18O value of +13‰, amounting to a 10‰ offset from the optimal value. This observation suggests that the well-known scale incompatibilities between laboratories could arise from this difference with measurements that may have suffered systematically from non-optimal acid-δ18O values, thus producing variable offsets, depending on the experimental details.

As a remedy, we suggest that the δ18O of phosphoric acid reacted with calcites for establishing a δ18O scale anchor be adjusted, and this should reduce the variability of the δ18O of CO2 evolved in acid digestion to less than ±0.05‰. The adjustment should be made by taking into account the difference in δ18O between the calcite-CO2 and the acid, with a target difference of 16‰. With this strategy, agreement between δ18O scales based on water, atmospheric CO2, and carbonates as well as data compatibility between laboratories may be substantially improved.