Optimising TC/EA–IRMS hydrogen isotope measurements of volcanic glass: a correction for low-H2 signal bias and application to SIMS calibration

<jats:p> In TC/EA–IRMS hydrogen isotope (δD) analysis of volcanic glass, sample mass is commonly adjusted according to H <jats:sub>2</jats:sub> O content so that all samples yield similar H <jats:sub>2</jats:sub> amounts, a practice that can obscure any dependence of measured δD on H <jats:sub>2</jats:sub> amount. Using a wide-bore reactor with a large graphite crucible that accommodated up to 17 mg of sample and gave an approximately four-fold greater peak area than the standard configuration, we measured two reference materials, IAEA-CH-7 polyethylene (δD = −100.3‰) and an in-house chlorite (δD = −41.1‰), over a wide range of H <jats:sub>2</jats:sub> amounts. Measured δD depended logarithmically on H <jats:sub>2</jats:sub> amount, with low-H <jats:sub>2</jats:sub> analyses biased toward higher δD by up to approximately 20‰ for polyethylene, whereas chlorite showed only a weak dependence. Six submarine basaltic glasses showed the same δD-dependent pattern, the bias being largest for low-δD glasses. Evaluation and correction of this H <jats:sub>2</jats:sub> -amount effect are therefore critical for accurate δD determination. We propose a two-point scaling correction based on the natural logarithm of H <jats:sub>2</jats:sub> amount and calibrated with the two reference materials; for one low-δD glass it shifted the mean δD by 7.9‰. Using the corrected values as reference data, normalised SIMS instrumental mass fractionation correlated linearly with anhydrous mole weight across nine basaltic glasses, showing that a composition-dependent matrix correction is required even within basaltic compositions. These results improve both bulk TC/EA–IRMS and in situ SIMS δD analyses and indicate that SIMS datasets calibrated with earlier reference values should be re-evaluated. </jats:p>