Managing argon interference during measurements of 18O/16O ratios in O2 by continuous-flow isotope ratio mass spectrometry

Abstract Monitoring changes in stable oxygen isotope ratios in molecular oxygen allows for studying many fundamental processes in bio(geo)chemistry and environmental sciences. While the measurement of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^{18}$$\end{document}18O/\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^{16}$$\end{document}16O ratios of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm {O}_{2}$$\end{document}O2 in gaseous samples can be carried out conveniently and from extracting moderately small aqueous samples for analyses by continuous-flow isotope ratio mass spectrometry (CF-IRMS), oxygen isotope signatures, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\updelta ^{18}$$\end{document}δ18O, could be overestimated by more than 6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\permille$$\end{document}‱ because of interferences from argon in air. Here, we systematically evaluated the extent of such Ar interferences on \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^{18}$$\end{document}18O/\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^{16}$$\end{document}16O ratios of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm {O}_{2}$$\end{document}O2 for measurements by gas chromatography/IRMS and GasBench/IRMS and propose simple instrumental modifications for improved Ar and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm {O}_{2}$$\end{document}O2 separation as well as post-measurement correction procedures for obtaining accurate \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\updelta ^{18}$$\end{document}δ18O. We subsequently evaluated the consequences of Ar interferences for the quantification of O isotope fractionation in terms of isotope enrichment factors, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\upepsilon _{\mathrm {O}}$$\end{document}ϵO, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^{18}$$\end{document}18O kinetic isotope effects (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^{18}$$\end{document}18O KIEs) in samples where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm {O}_{2}$$\end{document}O2 is consumed and Ar:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm {O}_{2}$$\end{document}O2 ratios increase steadily and substantially over the course of a reaction. We show that the extent of O isotope fractionation is overestimated only slightly and that this effect is typically smaller than uncertainties originating from the precision of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\updelta ^{18}$$\end{document}δ18O measurements and experimental variability. Ar interferences can become more relevant and bias \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\upepsilon _{\mathrm {O}}$$\end{document}ϵO values by more than \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$2\permille$$\end{document}2‱ in aqueous samples where fractional \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm {O}_{2}$$\end{document}O2 conversion exceeds 90%. Practically, however, such samples would typically contain less than 25 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\upmu$$\end{document}μM of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm {O}_{2}$$\end{document}O2 at ambient temperature, an amount that is close to the method detection limit of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^{18}$$\end{document}18O/\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^{16}$$\end{document}16O ratio measurement by CF-IRMS. Graphical abstract Supplementary Information The online version contains supplementary material available at 10.1007/s00216-022-04184-3.

Argon (99.999%), N 2 (99.999%), O 2 (99.995%), He (99.999%), and synthetic air (20% O 2 , 80% N 2 ) were obtained from Carbagas (Gümligen, Switzerland). All chemicals and enzymes were purchased from Sigma-Aldrich (Buchs, Switzerland) and used as received. Sodium acetate buffer was prepared with sodium acetate (99%) and hydrochloric acid (HCl, 32%). For O 2 consumption experiments, we used D-(+)-glucose (99.5%), and glucose oxidase from Aspergillus niger (type VII, 224 890 units g −1 ). All solutions were prepared in ultrapurified water (18.2 MΩ·cm, NANOpure Diamond water purification system; Barnstead). O 2 -free solutions were obtained by heating water to 90°C for 30 min while purging with N 2 gas. Purging continued thereafter for at least 2h. Samples containing O 2 -free water were prepared in an anaerobic glove box (< 1 ppm O 2 ) with a N 2 atmosphere (Unilab 2010; MBraun GmbH, Germany). With the parameters described in the method section, this corresponds to a MDL of 280 µM O 2 in the headspace and 100 µM dissolved O 2 in aqueous samples, respectively, for triplicate injections of 1000 µM . The MDL was thus approximately 6 times higher than reported for for the singe-injection setup (16 µM final aqueous O 2 concentration). 1 In the following, we show that by introducing N 2 gas to the sample vials to a pressure of 2 bars, the number of possible injections from sample headspace into a GC/IRMS could, theoretically, be increased to up to 16.

S2.1 Derivation of the maximum number of gas injections from the sample head space
Multiple injections from a single vial are possible if there is sufficient overpressure inside the vials. This can be rationalised based on the assumption of ideal gases and the ideal gas law (eq. S3 S1).
where p is the pressure (bar), V is the volume (L), n is the amount of substance (µmol), R is the ideal gas constant, 8.314 J K −1 mol −1 , and T is the absolute temperature. We assume that sample vials are tight and retain a constant pressure even after multiple piercings of the septa.
Each injection, x, reduces the amount of O 2 , n O 2 (x), and the pressure inside the vial, p vial (x), by the amount removed by the syringe according to equations S2 to S5.
where V inj is the injection volume, V vial is the effective gaseous volume of the vial, and n 0 O 2 and p 0 vial are the initial values of n O 2 and p vial , respectively. Figure S1(a) illustrates how both n O 2 and p vial decay by the same rate. The amount of O 2 withdrawn in the syringe is reduced accordingly. When the syringe is removed from the vial, however, pressure equalises to ambient pressure, p amb , with a loss of sample proportional to the decreasing difference in pressure. As long as there is overpressure inside the vial, i.e. p vial (x) > p amb , the amount of O 2 injected into the IRMS, n inj O 2 , is given by equation S6.
Thus, n inj O 2 is constant for as long as sample is escaping the syringe for pressure equalisation ( Fig  S1b). O 2 contamination from ambient air is limited to diffusive contamination into the syringe during sample transfer to the injector. Once p vial (x) approaches p amb , the amount of sample O 2 in the syringe is limited by the residual O 2 concentration in the vial and contaminated by ambient air containing χ vol% O 2 entering the syringe during pressure equalisation according to equation S7.
Consequently, the maximum number of injections, x max , for reproducible measurements can be estimated by equation S8.  Figure S1 Example for predicted dynamics of total pressure and total O 2 inside sample vials (a) and of O 2 injected (b). Both graphs are based on parameters as described in the methods section. The initial amount of O 2 in the vial, n 0 O 2 , is 8.31 µmol, which corresponds to 1 mL of artificial air in a 11.8 mL vial (V vial ), the initial pressure inside the vial, p 0 vial , is 2000 hPa, and each injection, x, is 500 µL (V inj ). Ambient pressure, p amb is 1013 hPa and 20.9 vol% O 2 (χ). The maximum number of injections without significant air contamination, x max , is indicated by an arrow.
The maximum injection volume or minimal vial pressure can be determined accordingly. To account for setup specific contamination, the minimal vial pressure can be tested by repeated injections from the same vial with a defined pressure. Once the amplitude of the O 2 peak increases, the maximum number of injections is reached.          Table S3. Error bars correspond to standard deviations of seven replicate measurements. Empty symbols were not included in the linear fit.

S10
Because automatic peak integration did not distinguish between the Ar and O 2 peaks, we performed manual peak integration. Figure S4 shows an example of a typical chromatogram of a gaseous sample with an Ar:O 2 ratio of 0.55 on a 30 m GC/IRMS setup. While automatic peak integration includes both peaks (red bar), manual integration started at the minimum between the two peaks (green bar).

S5 Quantification of Ar:O 2 ratios
The concentrations of Ar and O 2 and thus the Ar:O 2 ratio in the gaseous 3 mL headspace (subscript g) created in aqueous samples according to procedures described in Pati et al. 2 was calculated on the basis of equilibrium air-water partitioning. 7 The gaseous and aqueous concentrations of both Ar and O 2 in this two-phase system were determined by the initial aqueous concentrations of each species i, c i,w , in the sample before creating a N 2 headspace.
The gas phase concentrations of species i in the sample headspace (superscript ⊟ ), c ⊟ i,g , follow from the mass fraction in gaseous phase, f g , which are determined by the gas-water volume ratios and the dimensionless Henry's law constant, K i,H , as in eq. S9. The mass fractions are multiplied by a volume ratio (v w /v g ) to account for the fact that the total mass of O 2 in the sample vial (12 mL) originates from the aqueous phase sample only (v w = 9 mL).
where v w and v g are the volumes of aqueous and gas phase in the sample vial. The Ar:O 2 ratio follows from the ratio of gas phase concentrations (eq. S10).
where P in eq. S11 is a constant defined by the aqueous to gaseous volume ratio (9 mL/3 mL), Note that for the sake of simplicity, subscript g and superscripts ⊟ for the Ar:O 2 ratio in the headspace are omitted throughout the main manuscript and the SI for simplicity, as indicated in eq. S12.

S6.1 Derivation of theoretical relationship
Changes in δ 18 O of O 2 due to O 2 reduction follow from the general Rayleigh equation (eq. 3 from the main manuscript). The δ 18 O of O 2 in an aqueous sample is determined by its initial O isotope signature, δ 18 O 0 , the fraction of remaining dissolved O 2 , c O 2 ,w /c 0 O 2 ,w , and the enrichment factor, ε O , of the O 2 reduction reaction according to eq. S13.
Without correction, the measured value, δ 18 O ⋆ , is increased by Ar interference according to the ratio of Ar to O 2 , c Ar /c O 2 , and the instrument specific correction factor, b, as defined in eq. 1 S12 of the main manuscript.
In order to express the Ar enrichment as a function of the fraction of remaining dissolved O 2 , equation S12 is transformed to include the initial dissolved oxygen concentration in the t 0 Thus, the measured δ 18 O can be expressed as a function of ε O and c O 2 ,w /c 0 O 2 ,w (eq. 6 in the main manuscript).
When uncorrected δ 18 O ⋆ values are used to derive the ε O ⋆ of the reaction, they introduce an error, ∆ε O , described by the difference to the "true" value of ε O without Ar interferences. Here, we determine ε ⋆ O from a linear form of eq. 3 of the main manuscript based on δ 18 O ⋆ 0 and δ 18 O ⋆ max as in eq. S18.
Insertion of eq. S16 for δ 18 O ⋆ into eq. S20 leads to eq. S21 for ∆ε O . ∆ε O now depends on the two variables ε O and c O 2 ,w /c 0 O 2 ,w . Note that both the initial and the final δ 18 O measured is affected by Ar interference.

S6.2 Illustrative calculations
We illustrate the consequence of Ar interferences onto the quantification of O isotope fraction-  µM would have been 93% with an Ar:O 2 of 0.848 (eq. S12). In reality, the maximum turnover was about 80% for both experiments resulting in an Ar:O 2 of about 0.295 (eq. S12). While we cannot retrospectively determine the exact correction factor, b, of Ar interference on δ 18 O ⋆ , it is reasonable to assume a value comparable to the one we assessed for the same instrument setup (b of 8.57 ± 0.16, Table S2). Based on equation S14, we would expect deviations of δ 18 O ⋆ from the real value in the maximum turnover sample of 2.5 and 7.3 for the real sample (c/c 0 = 0.2) and the sample at detection limit (c/c 0 = 0.07), respectively.
Each O 2 consumption experiment consisted of seven samples at different fractions of conversion which reduces the extent of Ar interference in the determination of ε O . As discussed S16 in the main manuscript, the error is most distinct in reactions with a small ε O .