Constraints for precise and accurate fluid inclusion stable isotope analysis using water-vapour saturated CRDS techniques

Hydrogen (δ 2 H) and oxygen (δ 18 O) isotopes of water extracted from speleothem fluid inclusions are important proxies used for paleoclimate reconstruction. In our study we use a cavity ring-down laser spectroscopy system for analysis and modified the approach of Affolter et al. (2014) for sample extraction. The method is based on crushing of small sub-gram speleothem samples in a heated and continuously water-vapour purged extraction line. The following points were identified: Injection of reference water shows a precision (1σ) of 0.4-0.5 ‰ for δ 18 O values and 1.1-1.9 ‰ for δ 2 H values for water amounts of 0.1-0.5 μl, which improves with increasing water amount to 0.1-0.3 ‰ and 0.2-0.7 ‰, respectively, above 1 μl. The accuracy of measurements of water injections and water-filled glass capillaries crushed in the system is better than 0.08 ‰ for δ 18 O and 0.3 ‰ for δ 2 H values. The reproducibility (1σ) based on replicate analysis of speleothem fluid inclusion samples with water amounts >0.2 μl is 0.5 ‰ for δ 18 O and 1.2 ‰ for δ 2 H values, respectively. Isotopic differences between the water vapour background of the extraction system and the fluid inclusions have no significant impact on the measured fluid inclusion isotope values if they are within 10 ‰ for δ 18 O and 50 ‰ for δ 2 H values of the background. Tests of potential adsorption effects with inclusion free spar calcite confirm that the isotope values are unaffected by adsorption for water contents of about 1 μl (fluid inclusion) water per g of carbonate or above. Fluid inclusion analysis on three different modern to late Holocene speleothems from caves in northwest Germany resulted in δ 18 O and δ 2 H values that follow the relationship as defined by the meteoric water line and that correspond to the local drip water. Yet, due to potential isotope exchange reactions for oxygen atoms, hydrogen isotope measurements are preferentially to be used for temperature reconstructions. We demonstrate this in a case study with a Romanian stalagmite, for which we reconstruct the 20th century warming with an amplitude of approximately 1 °C, with a precision for each data point of better than ±0.5 °C.