Improving calibrations of in-house water isotope standards using CRDS and OA-CRDS: memory effects versus drift

<p>The isotope composition of rainfall provides information on the initial isotope composition of the moisture source, conditions during evaporation and condensation of water vapor, and the rain-out history of an air-parcel. A standard method to analyze the rainfall isotope composition is by using Cavity Ring Down Spectrometry (CRDS). The accuracy of the analysis highly depends on the water isotope standards used, which determines the degree to which absolute values from different labs can be compared. The amount of international water isotope standards like VSMOW2 and SLAP2 primary water standards is extremely limited; therefore the International Atomic Energy Agency recommends calibrating in-house water isotope standards once a year by using VSMOW2 and SLAP2. The isotope range between VSMOW2 and SLAP2 is extreme, with 55.5‰ for d<sup>18</sup>O and 427.5‰ for d<sup>2</sup>H. The isotope range used in a sequence poses a problem for CRDS techniques that are characterized by significant memory effects.</p><p>In this study, we compare the behaviors of two different CRDS systems: a Picarro L2140i and a LGR WIA 35EP. We evaluate the relation between isotope differences of subsequent samples and the memory effect. We show that after 100 injections, memory effects may still be visible in hydrogen. Even when the isotope composition of subsequent injections of the same standard or sample does not show a trend anymore, the raw isotope data seems biased towards the isotope composition of multiple different samples or standards run prior. Running long sequences of for example 1100 injections in high precision <sup>17</sup>O mode, also requires several vaporizer septa changes. The timing of a septa change is important, because opening the vaporizer allows water vapor from the atmosphere to enter the otherwise closed system, from which it takes approx. 20 injections to recover to the prior absolute values. Here we aim to provide a more practicle approach to a calibration sequence architecture and number of injections per primary and in-house standards, taking into account the potential drift of the analyzers.</p>