Measurement of ¹⁸O¹⁸O and ¹⁷O¹⁸O in atmospheric O₂ using the 253 Ultra mass spectrometer and applications to stratospheric and tropospheric air samples

Abstract

Rationale

The doubly substituted isotopologues (e.g., ¹⁸O¹⁸O, ¹⁷O¹⁸O) in atmospheric O₂ are potential tracers for ozone photochemistry and atmospheric temperatures. Their low abundances and isobaric interference are the major analytical challenges. The 253 Ultra high‐resolution stable isotope ratio mass spectrometer is suitable for resolving isobaric interferences.

Methods

O₂ from air is purified using gas chromatography on a packed column filled with molecular sieve 5 Å and cooled to −78°C. The δ¹⁷O, δ¹⁸O, Δ¹⁷O, Δ₃₅ and Δ₃₆ values are measured on the extracted O₂ with the 253 Ultra at medium mass resolution (M/ΔM ~10000) using Faraday detectors for the singly substituted isotopologues and ion counters for the doubly substituted isotopologues.

Results

Interferences from isobars, mainly ³⁵Cl for ¹⁷O¹⁸O and H³⁵Cl and ³⁶Ar for ¹⁸O¹⁸O, are sufficiently resolved to enable high‐precision determination of Δ₃₅ and Δ₃₆. The Δ₃₅ and Δ₃₆ values of O₂ after photochemical isotope equilibration at −63°C and heating to 850°C agree with the theoretical prediction. The stratospheric Δ₃₅ and Δ₃₆ values are close to isotopic equilibrium at the ambient temperatures. However, the values for tropospheric O₂ differ from those expected at equilibrium.

Conclusions

The 253 Ultra allows interference‐free clumped isotope measurements of O₂ at medium mass resolution. The Δ₃₅ and Δ₃₆ signatures in atmospheric O₂ are mainly governed by O₃ photochemistry, temperature and atmospheric transport. Tropospheric O₂ is isotopically well mixed and retains a significant stratospheric signature.

1. INTRODUCTION

Molecular oxygen (O₂) is an important constituent of the Earth's atmosphere and it is intricately linked to most life forms on earth. The atmospheric O₂ reservoir exchanges with the biosphere on a timescale of roughly 1200 years.1 Its isotopic composition is affected by biological, hydrological and photochemical processes.1, 2, 3, 4, 5, 6, 7 However, only the bulk isotopic ratios of O₂ (i.e., ¹⁸O/¹⁶O and ¹⁷O/¹⁶O) have been studied extensively. In nature, these ratios vary primarily in response to biological oxygen cycling, although the influence of stratospheric photochemistry has also been observed.2, 6, 7, 8, 9, 10 The doubly substituted isotopologues, ¹⁸O¹⁸O, ¹⁷O¹⁸O and potentially also ¹⁷O¹⁷O, also called clumped isotopes, can be useful to independently constrain the cycling of oxygen in the Earth's atmosphere.7

Clumped isotope signatures in O₂, expressed as Δ₃₅ and Δ₃₆ (see definitions below), are quantified as deviations of the abundances of ¹⁷O¹⁸O and ¹⁸O¹⁸O, respectively, from the abundances that are expected stochastically from the bulk isotopic composition. Under thermodynamic equilibrium conditions, the magnitude of the deviations purely depends on the temperature at which isotopic equilibration of O₂ takes place. The main challenges to measuring clumped isotope signatures in atmospheric O₂ are their low abundances (1.6 ppm for ¹⁷O¹⁸O and 4 ppm for ¹⁸O¹⁸O) and the isobaric interferences, for example, the interference of ³⁶Ar on ¹⁸O¹⁸O. Using careful separation and appropriate corrections for this interference, clumped isotope compositions in O₂ were first measured with low‐resolution mass spectrometers.6, 11 Yeung et al7 later showed with a medium‐resolution isotope ratio mass spectrometer (modified Nu Perspective IS, CAMECA, Gennevilliers, France, mass resolving power ~3600) that separation of ¹⁸O¹⁸O from ³⁶Ar is possible. Here we demonstrate that the 253 Ultra high‐resolution stable isotope ratio mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) can measure the Δ₃₅ and Δ₃₆ values along with the traditional isotope ratios (δ¹⁷O and δ¹⁸O values) at a precision close to the counting statistics limit. We show that traces of ³⁶Ar and other interferences can be well separated from ¹⁷O¹⁸O and ¹⁸O¹⁸O at medium mass resolution (~10000).

The first clumped isotope studies showed that Δ₃₅ and Δ₃₆ values are mainly governed by O₂ + O(³ P) photochemistry in the atmosphere.6, 7, 11 Significantly higher values of Δ₃₅ and Δ₃₆ in the Earth's atmosphere than expected from a stochastic distribution of isotopes led Yeung et al6 to hypothesize that O(³ P) + O₂ isotope exchange reactions reorder the isotopes in O₂ toward an isotopic equilibrium that depends on the ambient temperature. At lower temperatures, the ¹⁷O¹⁸O and ¹⁸O¹⁸O clumping should be relatively high, i.e., higher values of Δ₃₆ and Δ₃₅, and vice versa.12 Therefore, in thermodynamic equilibrium the clumped isotope signatures of O₂ should vary throughout the atmosphere because of variations in temperature. Yeung et al6 observed that the clumped isotopes of stratospheric O₂, above 22 km, are indeed in thermodynamic equilibrium at the low ambient stratospheric temperatures. In contrast, there was a clear deviation from isotopic equilibrium at the higher temperatures in the troposphere. This was explained in terms of the time scales of isotopic re‐equilibration, which are long in the troposphere because of a low abundance of O atoms and short in the stratosphere because of much higher O atom levels there. Thus, in different regions of the atmosphere the isotopic re‐equilibration can be faster or slower than the transport timescales. Therefore, the atmospheric Δ₃₅ and Δ₃₆ values reflect a dynamic balance between isotope exchange (temperature and oxidant dependent) and transport. Interestingly, the biological recycling of O₂, which is of primary importance for its bulk isotopic composition, has a negligible effect on the clumped isotopic composition because of the slow resetting time scale compared with photochemical isotope exchange.1, 7

Here we report a method for measuring clumped and bulk stable isotope ratios in atmospheric O₂ using the 253 Ultra mass spectrometer. After establishing and validating the purification and the mass spectrometric methodology, an empirical transfer function was developed to convert the measured isotope compositions into the absolute reference frame (ARF) by isotopically equilibrating O₂ at low and high temperatures. We also report Δ₃₅ and Δ₃₆ values measured in stratospheric and tropospheric air O₂ samples. In addition, we discuss the challenges and possibility of measuring the rarest isotopologue, ¹⁷O¹⁷O, of O₂ using the 253 Ultra.

2. CLUMPED ISOTOPES IN O₂ AND THEIR MEASUREMENTS

2.1. Conventional and clumped isotopes in O₂

The conventional singly substituted isotopic composition of a gas is characterized by the delta value, i.e., δ¹⁷O = (¹⁷R_sam/¹⁷R_std – 1) and δ¹⁸O = (¹⁸R_sam/¹⁸R_std – 1), where ¹⁷R = ¹⁷O/¹⁶O, ¹⁸R = ¹⁸O/¹⁶O, and subscripts ‘sam’ and ‘std’ stand for sample and standard, respectively. The δ values are expressed in per mill (‰) or parts per thousand. Values of δ¹⁷O and δ¹⁸O are reported with respect to Vienna‐Standard Mean Ocean Water (V‐SMOW).13

The doubly substituted isotopologues of O₂, also called clumped isotopes, are ¹⁸O¹⁸O, ¹⁷O¹⁸O and ¹⁷O¹⁷O. Their excess abundances compared with a stochastic distribution of the three oxygen isotopes in O₂ (i.e., the isotopes are randomly distributed over all isotopologues) are expressed using the relationship Δ_i = (ⁱR_measured/ⁱR_random − 1), where i refers to mass 36, 35 and 34 (e.g., ³⁶ R _measured = [¹⁸O¹⁸O]_measured/[¹⁶O¹⁶O]_measured) and

$$\begin{array}{l}{}^{36}{R}_{\text{random}}=\frac{{\left[{}^{18}\mathrm{O}{}^{18}\mathrm{O}\right]}_{\text{random}}}{{\left[{}^{16}\mathrm{O}{}^{16}\mathrm{O}\right]}_{\text{random}}}=\frac{\left[{}^{18}\mathrm{O}\right]\left[{}^{18}\mathrm{O}\right]}{\left[{}^{16}\mathrm{O}\right]\left[{}^{16}\mathrm{O}\right]}={\left({}^{18}R\right)}^2,\\ {}^{35}{R}_{\text{random}}=\frac{{\left[{}^{18}\mathrm{O}{}^{17}\mathrm{O}\right]}_{\text{random}}}{{\left[{}^{16}\mathrm{O}{}^{16}\mathrm{O}\right]}_{\text{random}}}=2\frac{\left[{}^{18}\mathrm{O}\right]\left[{}^{17}\mathrm{O}\right]}{\left[{}^{16}\mathrm{O}\right]\left[{}^{16}\mathrm{O}\right]}=2^{18}{R}^{17}R,\text{and}\\ {}^{34}{R}_{\text{random}}=\frac{{\left[{}^{17}\mathrm{O}{}^{17}\mathrm{O}\right]}_{\text{random}}}{{\left[{}^{16}\mathrm{O}{}^{16}\mathrm{O}\right]}_{\text{random}}}=\frac{\left[{}^{17}\mathrm{O}\right]\left[{}^{17}\mathrm{O}\right]}{\left[{}^{16}\mathrm{O}\right]\left[{}^{16}\mathrm{O}\right]}={\left({}^{17}R\right)}^2\end{array}$$

Δ_i values are also reported in ‰. For a stochastic distribution of isotopes, Δ_i = 0. A nonzero Δ_i value for a sample represents an excess or deficit of multiply substituted isotopes (e.g., ¹⁸O¹⁸O) compared with what is expected based on their bulk isotopic composition. We calculate the clumped isotope ratios following the previously adopted established conventions and data reduction procedures for O₂ (Δ₃₅ and Δ₃₆) and CO₂ (Δ₄₇)11, 14, 15 (see supporting information for detailed calculation).

2.2. Measurement of O₂ clumped isotopes with the 253 Ultra

Isotopic measurements including the clumped isotopes (Δ₃₅ and Δ₃₆) were carried out with the 253 Ultra isotope ratio mass spectrometer at Utrecht University, which is one of the first instruments produced by Thermo Fisher Scientific on the basis of the prototype described in Eiler et al.16 It is a double‐focusing mass spectrometer with an electrostatic analyzer followed by a magnetic sector. It can be operated at three different mass resolutions, set by one of the three slits between the source and the electrostatic analyzer. The widths of the slits are 250, 16 and 5 μm for low, medium and high mass resolutions, respectively. Mass resolving power (M/ΔM) is defined as the ratio between the nominal mass and the width, measured in mass units, that corresponds to an intensity increase from 5% to 95% of the maximum peak signal. The maximum mass resolving power achievable with the present instrument is ~40 000 with the narrowest slit. The typical mass resolving powers at low and medium resolution are ~2000 and ~10 000, respectively. The collector assembly has nine different collector positions, where the central collector is fixed and the other eight are movable. The central collector can be switched between a Faraday collector and a Secondary Electron Multiplier (SEM), and the three collector blocks at the high mass end contain both a Faraday collector and a Compact Discrete Dynode (CDD) SEM next to each other. The Faraday collectors can be operated using a choice of resistors from 3*10⁸ Ω to 10¹³ Ω. Operation of the instrument is controlled by the Qtegra software package (Thermo Fisher Scientific).

In the present work, the medium‐resolution slit with a mass resolving power of ~10 000 (M/ΔM) for O₂ is used. The major isotopologues (¹⁶O¹⁶O, ¹⁶O¹⁷O, ¹⁶O¹⁸O) are measured in Faraday cups with resistors 10¹⁰ Ω, 10¹² Ω and 10¹¹ Ω, respectively, while the minor ion beams (¹⁷O¹⁸O, ¹⁸O¹⁸O) are measured using two CDDs.

Isotope ratio measurements are generally made at a source pressure of about 2.0 × 10⁻⁷ mb. Under these conditions, in medium‐resolution mode the ion signal is 1 × 10¹⁰ counts per second (cps) at mass 32; the corresponding signals for masses 35 and 36 are ~20 × 10³ and ~50 × 10³ cps, respectively. A sample is measured for 30 to 60 sample‐reference cycles, divided into 5 to 10 acquisitions, each with 6 cycles. Peak centering, background scanning (60 s) and pressure‐balancing (tolerance 0.3%) are performed before each acquisition. The equilibration time (idle time) and integration time are set as 60 s and 67 s, respectively. Measurement of a sample takes ~7 h which consists of an actual measurement time of ~4.5 h, roughly half of which is required for equilibration after valve switching and the other half for signal acquisitions of the sample and the working gas. The remaining ~2.5 h is used for background scanning, reference and sample signal balancing, peak centering and peak positioning (all of these are performed at the beginning of each acquisition).

2.3. Preparation of equilibrated gases for clumped isotope calibration

We isotopically equilibrate aliquots of O₂ at different temperatures in order to express our measured clumped isotope ratios against a scrambled reference. At high temperatures (850°C), isotopic equilibration of O₂ is carried out in the presence of a platinum catalyst (~120 mg of Pt mesh; Goodfellow, Huntingdon, UK) in a quartz tube of length 42 cm and diameter 1.5 cm. Heating is carried out by inserting ~30 cm of the tube into a furnace set at 850°C. After 2–3 h of heating, the tube is directly connected to one of the bellows of the dual‐inlet system of the 253 Ultra to transfer hot O₂ for measurement.

At low temperatures, it is not straightforward to equilibrate O₂ isotopically as it does not readily exchange isotopes. Therefore, UV‐induced photochemical isotope exchange11 is used to equilibrate a number of O₂ gases at lower temperatures. UV photolysis experiments are carried out in a spherical ~500‐mL volume Pyrex reactor (diameter 10 cm) fitted with a Suprasil finger (outer diameter 2 cm, length 10 cm; LSP Quartz, Wijchen, The Netherlands) at the center. The Suprasil finger transmits UV radiation (Pyrex is opaque to UV), and it is connected to the reactor using a KF clamp with a Teflon O ring. A Hg lamp (NTE 5 W.210; Radium, Wipperfürth, Germany) is introduced into the Suprasil finger as the UV source (major wavelengths: 184.9 and 253.6 nm).

The reactor is connected to a gas mixing system17 with a 6 mm ultra‐torr fitting and through a vacuum stopcock with a Teflon seal. The reaction vessel is first evacuated to a high vacuum of 1*10⁻⁵ mbar and then pure O₂ working gas (IMAU O₂; Linde Gas, Schiedam, The Netherlands) is introduced into the vessel to a pressure of 40–50 mbar. The flask is immersed in an ethanol or water bath and cooled to various temperatures between +26°C and −78°C. While the outside of the reactor is at the temperature of the cooling fluid, determination of the temperature of the gas inside the reactor is challenging, because the temperature of the Suprasil finger with the glowing UV lamp is much higher than the outside temperature of the reactor. For example, when the outer surface is at dry‐ice temperature (−77.8°C), the temperature inside the Suprasil finger is found to be 10 ± 5°C at steady state. The effective temperature of the gas inside the flask is calculated from the change in pressure using Gay‐Lussac's pressure law P₁/T₁ = P₂/T₂, where P₁ is the pressure inside the chamber at room temperature (T₁), i.e. before putting the chamber inside the bath and without UV lamp, and P₂ and T₂ are the pressure and temperature with the chamber inside the dry‐ice bath and the glowing UV lamp inside the finger after stabilization. For the lowest bath temperature of −77.8°C the effective gas temperature is −63 ± 5°C. This effective temperature is then used to assign the expected isotopic ordering of O₂ inside the flask based on the theoretical calculations of Wang et al.12 After irradiation for 2–3 h, the equilibrated O₂ samples are separated from the O₃ formed during the UV treatment by passing the gas through a spiral tube immersed in liquid nitrogen. It is observed that 3–4% of the O₂ inside the chamber is converted into O₃ at steady state. Therefore, immediate separation of O₃ is necessary to prevent further O₂ formation from decomposition of O₃, which would probably change the isotopic signatures. The isotopically equilibrated O₂ samples are measured without any further purification.

All samples are measured against a working gas (IMAU O₂) with nominal δ¹⁷O and δ¹⁸O values of 9.254‰ and 18.542‰, respectively, versus V‐SMOW, assigned by E. Barkan at the Institute of Earth Sciences, Hebrew University of Jerusalem (Jerusalem, Israel).

2.4. Sample O₂ purification

O₂ is separated from air using a gas chromatography (GC) column (column length: 3.05 m, OD: 1/8 inch, ID: 2 mm, Molsieve 5A; Agilent Technologies, Santa Clara, CA, USA). Air is filled into a sample loop of 10 mL volume (Figure 1) either by flushing air from a high‐pressure container or by using an air‐tight syringe of 50 mL capacity. The syringe is loaded with 50 mL of the air sample and is used to flush and fill the sample loop before injecting into the GC column. High‐purity He is used as the carrier gas at a flow rate of 25 mL/min. The GC column is kept at dry ice temperature (−77.8°C). The eluted gases are monitored using a Thermal Conductivity Detector (TCD). The Ar peak appears at ~32 min and O₂ at ~40 min with a separation of ~5 min between the two peaks (Figure 2). We do not observe the N₂ peak in the 80‐min time window set in the present case unless the column temperature is increased. To show the N₂ peak (Figure 2), the column is removed from the dry‐ice bath and is kept at room temperature after O₂ has eluted. The purified O₂ sample is collected at liquid nitrogen temperature in two stainless steel U traps (OD: 6 mm) filled with silica gel. O₂ is released from the U traps by removing the liquid nitrogen and then transferred to a glass ampule of ~2 mL volume filled with a few pellets of preconditioned silica gel. O₂ from the ampule is directly introduced into the bellow of the mass spectrometer to measure its isotopic composition. After each sample extraction, the GC column is baked at 200°C for 1 h at a He flow rate of 25 mL/min and longer baking (overnight) is performed at the end of the day. The two silica gel U traps for sample collection are also baked at 200°C for 1 h under high vacuum (1*10⁻⁵ mb) after each sample collection.

Figure 1.

O₂ purification system. An air sample or gas mixture is admitted to a sample loop (10‐mL volume) via a six‐port valve. O₂ is separated from the other air constituents using a packed molecular sieve 5 Å (length: 3.05 m, OD: 1/8 inch, ID: 2 mm, Molsieve 5 Å) column cooled to dry‐ice temperature (−77.8°C). The effluent from the GC column is monitored using a thermal conductivity detector (TCD, see Figure 2) and the sample is collected on silica gel at liquid nitrogen temperature (−196°C)

Figure 2.

Gas chromatogram showing the separation of O₂ from Ar and N₂ in the GC column at dry‐ice temperature (−77.8°C). The separation between the Ar and O₂ peaks is ~5 min when the GC column is kept at −77.8°C. Under these conditions N₂ is trapped on the GC column. To release the N₂ and observe the N₂ peak on the chromatogram, the GC column was taken out of the dry ice/ethanol bath when O₂ collection was finished, and kept at 25°C (indicated by the arrows below the detector trace)

2.5. Preparation of O₂ for testing isotopic reordering

There is a possibility of isotopic reordering in O₂ during sample storage, purification (e.g. GC column) and analysis in the mass spectrometer.7 To quantify potential reordering effects, we prepared enriched O₂ (¹⁸O₂ ~700‰) in a 2‐L volume canister by diluting 97.2 atom % ¹⁸O₂ (Euriso‐top, Saint‐Aubin, France) with our IMAU working gas O₂. The enriched O₂ was then measured before and after passing through the GC column to quantify the isotopic reordering in the GC column and the source of the mass spectrometer.

2.6. Tropospheric and stratospheric air sampling and measurements

To determine the clumped isotope composition in lower tropospheric air O₂, and to check the reproducibility of measurements, we have used ambient air compressed to ~140 bar in a ~40‐L cylinder. This air was collected outside the Center for Isotope Research on the campus of Groningen University (Groningen, The Netherlands), in August 2017. Air is compressed using a Rix oil‐free air and gas compressor (RIX Industries, Benicia, CA, USA). O₂ is extracted from the compressed air cylinder as described in section 2.4 and the isotopic composition is measured following the procedure described in section 2.2. We also collected an air sample from the Utrecht University Campus (Utrecht, The Netherlands) on 18 September 2018 and measured its isotopic composition.

Stratospheric and upper tropospheric air samples were obtained with a whole air sampler operated on the high‐altitude aircraft M55 GEOPHYSICA on two flights carried out within the project StratoClim (stratospheric and upper tropospheric processes for better climate predictions). StratoClim is a collaborative research project under the European Commission's 7th Framework program (H2020) to improve the understanding of the key processes in the Upper Troposphere and Stratosphere with the goal of producing more reliable climate projections. We analyzed 13 samples collected on 1 and 6 September, 2016 during the StratoClim campaign over the Mediterranean (base airport Kalamata, Greece) at heights between 10 000 and 20 000 m and covering latitudes between 33 and 41°N and longitudes between 22 and 29°E. The temperature was −58°C near the highest Geophysica sampling altitudes (~ 20 km) and −42°C at the lowest sampling altitudes (~10 km). Stainless steel canisters of ~2 L volume were filled and compressed to ~4 bar with ambient air at different heights. O₂ was extracted as described in section 2.4. Stable isotope ratios including Δ₃₅ and Δ₃₆ values were measured on purified O₂ in the 253 Ultra as discussed in section 2.2.

The N₂O mole fractions are measured together with the N₂O isotopic composition in a Delta Plus XP continuous flow stable isotope ratio mass spectrometer (Thermo Fisher Scientific) following the method developed by Röckmann et al.18 Briefly, the air samples are passed through an Ascarite (sodium hydroxide coated silica; Sigma‐Aldrich, Inc., St Louis, MO, USA) column to remove CO₂, and the N₂O (together with remaining traces of CO₂ and other condensable species) is preconcentrated cryogenically. The cryo‐focused sample is purified on a capillary GC column (PoraPlot Q, 0.32 mm; Agilent Technologies, GA, USA) and N₂O is introduced into the mass spectrometer via an open split interface. From the area of the peaks, the concentration of N₂O is calculated.

3. RESULTS AND DISCUSSION

3.1. Experimental results

3.1.1. Zero enrichment and counting statistics

Table 1 presents the results from the zero‐enrichment measurements. The zero‐enrichment values vary in the ranges of −0.001 to 0.03‰ for δ³³ values, −0.004 to 0.047‰ for δ³⁴ values, −0.176 to 0.112‰ for δ³⁵ values and 0.008 to 0.132‰ for δ³⁶ values (Table 1). For the δ³³ and δ³⁴ values, the external errors of the six measurements are greater than the individual internal errors, which indicates that the mass spectrometer is not limiting the precision but an extra error of the order of 0.01‰ is introduced from other sources such as sample handling. For the clumped isotopes, the internal errors are larger and similar to the external errors of the six measurements, showing that the mass spectrometer precision (in principle the counting statistics) is the limiting factor here. A slightly positive value is observed for δ³⁶, but it is within the error associated with the individual measurements. Therefore, a zero‐enrichment correction is not applied for the sample measurements.

Table 1.

Results of zero‐enrichment measurements with bellows filled with IMAU‐O₂. The errors are based on 30–60 cycles of sample and working gas measurements. Expected errors based on the counting statistics (EECS) are also presented. All δ values in this table are expressed in ‰ with respect to the working gas

Sl. No.	δ33 ± 1SE*	EECS	δ34 ± 1SE	EECS	δ35 ± 1SE	EECS	δ36 ± 1SE	EECS
1	0.030 ± 0.014	0.010	−0.006 ± 0.004	0.004	0.050 ± 0.180	0.224	0.068 ± 0.111	0.138
2	−0.018 ± 0.009	0.009	0.008 ± 0.004	0.004	−0.108 ± 0.167	0.223	0.087 ± 0.142	0.138
3	−0.001 ± 0.007	0.008	−0.004 ± 0.004	0.003	0.112 ± 0.143	0.177	0.039 ± 0.111	0.119
4	0.021 ± 0.014	0.013	0.004 ± 0.006	0.005	−0.176 ± 0.197	0.280	0.091 ± 0.187	0.189
5	−0.008 ± 0.009	0.009	−0.010 ± 0.004	0.004	−0.071 ± 0.165	0.198	0.008 ± 0.119	0.134
6	0.013 ± 0.010	0.010	0.047 ± 0.005	0.004	−0.157 ± 0.196	0.215	0.132 ± 0.097	0.138
Average ± 1σ	0.006 ± 0.018		0.006 ± 0.020		−0.058 ± 0.115		0.071 ± 0.043

^* 1 Standard Error.

The measured errors for all the isotopic ratios including the δ³⁵ and δ³⁶ values are compared with the errors expected from counting statistics and found to be similar (EECS, see Table 1 and supporting information). This proves that the 253 Ultra is very stable over the 7‐h duration of these measurements. The variation in the EECS is due to different signal strength and measurement duration. As the measurement uncertainty closely follows the counting statistics, the precision of measurements can be improved by increasing the signal intensity or the measurement time. For the present measurements we targeted an individual precision of 0.2 and 0.1‰ for Δ₃₅ and Δ₃₆, respectively. Higher precisions (e.g. Yeung et al7 reported an external reproducibility of 0.038‰) can probably be achieved by increasing measurement time and/or intensity.

We regularly observe when using the CDDs that the measurement errors are better than the counting statistics limit. We tentatively attribute this to the fact that the electron multipliers may not properly sample the Poisson distribution of count rates, because of the detector dead times. The Poisson distribution is the underlying statistics for calculating the counting statistics limit, and, if the high count rate tail is not detected, part of one tail of the distribution is cut off, which may lead to lower calculated errors for the CDDs. In the case of Faraday cups, all ions are counted as they pass through a resistor and the errors are similar to the counting statistics limit as expected.

3.1.2. Ar and other isobaric interference

It is difficult to achieve full separation of Ar from O₂ by passing the gases through a GC column. Some traces of Ar from the tail of the Ar peak always remain in the extracted O₂ with the current GC separation setup. The influence of ³⁶Ar (mass 35.9675 u) on ¹⁸O¹⁸O (mass 35.9983 u) is demonstrated using an Ar‐O₂ mixture gas that is prepared by mixing 0.16% pure Ar with our pure O₂ working gas. The mass difference between these two isobars is 0.0308 u and the resolving power required to separate their peaks is ~1160, which is easily achieved with the 253 Ultra at medium mass resolution. Figure 3 shows peak scans for pure O₂ and the Ar‐O₂ mixture. The O₂ isotopologue peaks are about 0.03 u wide and they are in general very stable in terms of signal intensity, width of the plateau and collector positions. In the Ar‐O₂ mixture, the ³⁶Ar peak appears about 0.03 u before the ¹⁸O¹⁸O. There is a region of 0.009 u where the peaks overlap. The tilted peak for ³⁶Ar is probably due to inaccurate adjustment of the CDD operating voltage. With further tuning of the operating voltage we later also achieved a flat peak for ³⁶Ar (not shown here).

Figure 3.

Peak scan for (A) pure O₂ and (B) 0.16% Ar mixed with O₂. All the five detector traces for measuring masses 32, 33, 34, 35 and 36 are shown. Although there is a partial peak overlap, the ³⁶Ar peak does not interfere with the measurement of ¹⁸O¹⁸O when the measurement is made at the right half of the ¹⁸O¹⁸O peak (right side of position 2, indicated by arrows on the top of the ¹⁸O¹⁸O mass spectrum in B). Interference of ³⁶Ar was observed when measured at position 3 (see text and Table 2). All the measurements for the air samples presented here were made at position 1. Magnified views of the top and bottom 10% of the peaks for masses 35 and 36 are shown in (C) and (D). The noise on the top (C) (~2%) could be reduced with longer integration. Significantly higher baselines for masses 35 and 36 on the left of the peaks (D) are due to contribution from ³⁵Cl and H³⁵Cl (see text). The small peak (~5% of the maximum) for mass 35 on the far left of the actual peak, observed occasionally, could be due to contamination but does not influence the ¹⁷O¹⁸O measurement [Color figure can be viewed at wileyonlinelibrary.com]

In addition to ³⁶Ar, other interferences can be identified by the mass differences relative to the O₂ isotopologue peaks. Chlorinated species are particularly important, especially in new mass spectrometers, as chlorinated solvents are used to degrease the parts, and we have regularly observed increases in these interferences after heating the ion source. ³⁵Cl (mass 34.9688 u) causes the apparent increase in the background (~2% of the ¹⁷O¹⁸O peak height in Figure 3D) on the low‐mass side of the ¹⁷O¹⁸O peaks. It is detected 0.02 u before the ¹⁷O¹⁸O peak, and leaves the detector again at 0.01 u into the ¹⁷O¹⁸O peak. The effect of H³⁵Cl (mass 35.9767 u) on ¹⁸O¹⁸O is visible on the left side of the ¹⁸O¹⁸O peak in the case of pure O₂ (Figure 3D). However, for the Ar‐O₂ mixture (Figure 3B), it is barely visible because it is superimposed on (and partly contributes to) the tilted slope of the ³⁶Ar peak. Note that mass interferences conceptually cause “step changes” when they enter and exit the detector, and not tilted peak tops. The interferences from these isobars can easily be avoided in the 253 Ultra by measuring the isotopologue ratios at the right position of the peaks as the mass resolving power used for the present configuration (~10 000) is much higher than that required to resolve most of these isobars (e.g., 1664 for resolving ¹⁸O¹⁸O and H³⁵Cl, and 1188 for resolving ¹⁷O¹⁸O and ³⁵Cl). We note that the background for mass 33 is ~1.7% of the maximum peak height (Figures 3A and 3B). It increases about 0.02 u before the ¹⁶O¹⁷O peak but does not return to zero after the peak, as checked by extending mass scanning on the right side of the ¹⁶O¹⁷O peak. If it were due to contamination from interfering species (e.g., ³³S and H¹⁶O¹⁶O), the background signal should return to zero after the width of an interfering peak of 0.03 u, as the width of any species falling on any cup other than the fixed Centre cup is ~0.03 u (Figure 3). We could not identify the cause of this offset, but it is similar for both sample and working gas and expected to have little effect on the final δ³³ values.

Table 2 shows the isotopic ratios measured on pure O₂ and the Ar‐O₂ mixture at different positions of the peaks, as indicated in Figure 3B. A clear ³⁶Ar interference to ¹⁸O¹⁸O is observed when measurements are made near the ³⁶Ar peak, i.e., position 3 in Figure 3B and last line in Table 2. In addition, the internal error is highest for this measurement, indicating that small shifts in the mass scale change the contribution of ³⁶Ar during the measurement period. Away from the ³⁶Ar peak (positions 2 and 1), no significant ³⁶Ar interference is observed (Table 2). We conclude that with the 253 Ultra, ¹⁸O¹⁸O can be easily resolved from ³⁶Ar and other interferences even at medium resolution (resolving power: ~10 000 in this case). Therefore, interference‐free measurements of the clumped isotopic composition of O₂ are possible even if the samples are not fully free from Ar. All the measurements presented here are made near the right edge of the peak plateauxs (position 1 in Figure 3B).

Table 2.

Isotopic composition (δ¹⁷O and δ¹⁸O values in ‰ vs VSMOW) for the IMAU O₂ working gas and a mixture of IMAU O₂ and Ar to study the isobaric interference of ³⁶Ar on the measurement of ¹⁸O¹⁸O. Measurements are made at different peak positions as shown in Figure 3. The ³⁶Ar isobaric interference for Δ₃₆ is very prominent when the measurement is made near the right edge of the Ar peak (position 3 in Figure 3)

Sample and peak positioning	δ17O (VSMOW)	δ18O (VSMOW)	Δ35 (ARF)	Δ36 (ARF)
Pure O2	9.260 ± 0.007	18.548 ± 0.008	1.262 ± 0.115	2.475 ± 0.043
O2–Ar mixture at peak position 1	9.173 ± 0.004	18.412 ± 0.008	1.236 ± 0.153	2.412 ± 0.086
O2–Ar mixture at peak position 2	9.178 ± 0.005	18.391 ± 0.006	1.190 ± 0.115	2.331 ± 0.101
O2–Ar mixture at peak position 3	9.174 ± 0.012	18.391 ± 0.008	1.303 ± 0.151	21.419 ± 0.257

The small but significant differences in δ¹⁷O and δ¹⁸O values between pure O₂ and the Ar‐O₂ mixture (Table 2) are attributed to isotope fractionation during sample handling in the gas mixing process. No significant differences are observed for Δ₃₅ or Δ₃₆ values, indicating that fractionation during sample handling, if any, is within the measurement uncertainty (discussed in sections 3.1.3 and 3.1.5).

3.1.3. Isotopic reordering in the GC column and the source of the mass spectrometer

To test isotopic reordering in the GC column and in the source of the mass spectrometer, isotopically spiked O₂ (Δ₃₆ ~684‰ and Δ₃₅ ~31‰) is used (Table 3). The spiked O₂ sample is analyzed after preparation and then again after 4 days of storage in a stainless‐steel canister of the same type as used for the stratospheric air samples presented below. The results agree within the analytical errors, suggesting that isotopic reordering in this storage canister is insignificant (Table 3).

Table 3.

Changes in the isotopic composition (δ and Δ values are given in ‰) of enriched O₂ due to isotopic reordering in the GC column and storage in stainless‐steel canister

Sample	δ33 (sam‐wg)	δ34 (sam‐wg)	δ35 (sam‐wg)	δ36 (sam‐wg)	δ17O (VSMOW)	δ18O (VSMOW)	Δ35 (ARF)	Δ36 (ARF)	% change in Δ36
IMAU O2	0.006 ± 0.018	0.006 ± 0.020	−0.058 ± 0.115	0.071 ± 0.043	9.260 ± 0.007	18.548 ± 0.008	1.262 ± 0.115	2.475 ± 0.043
Spiked O2	0.015 ± 0.010	0.030 ± 0.005	31.093 ± 0.150	700.107 ± 0.201	9.270 ± 0.010	18.573 ± 0.005	31.398 ± 0.115	684.252 ± 0.284
Spiked O2 stored 4 days in a canister	0.006 ± 0.027	0.027 ± 0.019	30.476 ± 0.377	699.935 ± 0.368	9.260 ± 0.027	18.570 ± 0.020	30.808 ± 0.205	684.004 ± 0.228	0.04
GC separation at −75°C	0.066 ± 0.009	0.151 ± 0.003	31.258 ± 0.223	689.285 ± 0.0159	9.321 ± 0.010	18.696 ± 0.004	31.384 ± 0.321	673.297 ± 0.143	1.6
GC separation at −80°C	0.074 ± 0.020	0.168 ± 0.008	31.042 ± 0.453	688.600 ± 0.290	9.329 ± 0.021	18.713 ± 0.008	31.149 ± 0.377	672.569 ± 0.012	1.7

To investigate reordering in the purification system (in particular the GC column), an aliquot of the isotopically spiked O₂ is mixed with pure helium to prepare an O₂ concentration of ~20%, similar to atmospheric O₂, and the mixture is passed through the purification system following the same procedure as for the samples. The measured isotope ratios before and after passing through the purification system are presented in Table 3. From the changes in Δ₃₅ and Δ₃₆ we conclude that the isotopic reordering in the purification system (GC column) is 1.6% when He‐O₂ is passed through the column cooled to −75°C and 1.7% at −80°C. For our measurements of atmospheric O₂ with a typical Δ₃₆ value of 2.4‰, 1.7% isotopic reordering towards a scrambled isotopic distribution would induce a systematic error of ~0.04‰ (assuming the same relative reordering of 1.7%), which is below our reported uncertainty. The observed values for isotopic reordering are similar to the 1.1% reported by Yeung et al11 for their system.

The measurement of the spiked gas itself can also provide information about possible isotope reordering in the ion source of the mass spectrometer.11 Significant isotopic reordering of such a highly enriched gas due to fragmentation‐recombination reactions inside the ion source would change the δ³⁴ value of the spiked O₂ compared with the unspiked O₂. This may occur because in the fragmentation‐recombination process, some ¹⁸O¹⁸O will be converted into ¹⁸O¹⁶O. The natural abundance of ¹⁸O¹⁸O is 4.2*10⁻⁶; therefore, the spiked gas with Δ₃₆ ~684‰ contains 2*684‰*4.2*10⁻⁶ ≈ 5.7*10⁻⁶ extra ¹⁸O atoms_. If 1% of the added ¹⁸O¹⁸O is converted into ¹⁸O¹⁶O, this would produce 5.7*10^{−8 18}O¹⁶O molecules, which would cause a change in the δ³⁴ value of 0.014‰. The actually observed δ³⁴ difference between the spiked and unspiked O₂ is 0.02‰ (Table 3) which is similar to the analytical error of 0.02‰ (Table 1). This indicates that the isotopic reordering in the source of the mass spectrometer is less than 2%. The spiked O₂ has 97.2 atom % ¹⁸O¹⁸O. If the remaining 2.8% is ¹⁸O¹⁶O, this would change the ¹⁸O¹⁶O of the IMAU O₂ by ~0.007‰. We conclude that isotopic reordering effects in storage canisters, in the purification system and in the source of the mass spectrometer are below our reported measurement errors and no corresponding corrections are applied.

3.1.4. Calibration and reporting the clumped isotope ratios on the absolute reference frame

The Δ₃₅ and Δ₃₆ values of the O₂ that is isotopically equilibrated at different temperatures (−63°C, 4°C, 8°C, 26°C and 850°C) are plotted against the corresponding values calculated for O₂ in thermodynamic isotope equilibrium12, 19 in Figure 4 (the numerical values are provided in Table S1, supporting information). The linear fit to the data presented in Figure 4 provides an empirical transfer function to convert the measured Δ₃₅ and Δ₃₆ values for O₂ into the Absolute Reference Frame (ARF).15 The transfer functions (weighted regression linear fit lines) for Δ₃₅ and Δ₃₆ in the present case are Δ_35(ARF) = (0.970 ± 0.035), Δ_35(meas) + (1.281 ± 0.018) and Δ_36(ARF) = (0.974 ± 0.026), Δ_36(meas) + (2.429 ± 0.035), respectively (Figure 4). The slopes of the best‐fit lines are close to 1, indicating close agreement between the theoretical prediction and measurements. This is additional evidence that isotopic reordering in the source of the mass spectrometer is insignificant, since isotopic reordering would lead to scale compression and result in slopes greater than 1 for the empirical transfer functions. The intercepts of the transfer functions indicate the Δ₃₅ and Δ₃₆ values (Δ₃₅ = 1.281 and Δ₃₆ = 2.429‰) of our working gas (IMAU‐O₂).

Figure 4.

Construction of an empirical transfer function for conversion of measured Δ₃₅ and Δ₃₆ values (expressed against our working gas) into the absolute reference frame (ARF). Theoretical values (y axis) are those predicted by equilibrium thermodynamics.13 The equilibration of O₂ induced by UV radiation is carried out at temperatures of −63, 4, 8 and 26°C (see text) and heated O₂ is prepared at 850°C. The errors are 1 SE for 30–60 cycles of measurement. The errors in the theoretical values due to uncertainty in temperature estimates are discussed in the text [Color figure can be viewed at wileyonlinelibrary.com]

For the samples analyzed, the Δ₃₅ and Δ₃₆ values before and after application of the transfer function are provided in Table S2 (supporting information). The measured Δ₃₅ and Δ₃₆ values before converting to the absolute scale are the measured values of the samples versus those of the heated gas assuming that the heated gas is isotopically scrambled. No significant difference is observed in the Δ₃₅ and Δ₃₆ values before and after conversion to the ARF (Table S2, supporting information). Our scientific interpretations below are based on the Δ₃₅ and Δ₃₆ values after conversion to the ARF, which facilitates inter‐laboratory comparison.

3.1.5. Reproducibility and accuracy

We measured the isotopic composition including Δ₃₅ and Δ₃₆ values of O₂ from a compressed air cylinder throughout the duration of sample measurements to monitor the reproducibility as well as possible long‐term drift or variability, if present. The results of the compressed air measurements are presented in Table 4. Errors in the δ¹⁷O and δ¹⁸O values were larger during the initial phase of measurements, possibly due to instability of the 253 Ultra, which caused some drifts in the mass scale and changes in signal sensitivity. With the passage of time, the δ¹⁷O and δ¹⁸O values became more stable and the measurement errors in the later phase are similar to the errors expected from counting statistics (Table 4). As for the zero enrichment (Table 1), the variations in the δ¹⁷O and δ¹⁸O values between measurements are larger than the measurement errors, which probably originate from fractionation in sample purification and handling. The values of Δ¹⁷O, Δ₃₅ and Δ₃₆ and their uncertainties are close to the counting statistics limit throughout the measurement period. The smaller variability in Δ¹⁷O, Δ₃₅ and Δ₃₆ is probably because the fractionations associated with handling are mass dependent and do not affect the second‐order isotope signatures. The long‐term (3 months) external reproducibility (1σ standard deviation) of the measured values is 0.115‰ for δ¹⁷O, 0.215‰ for δ¹⁸O, 0.021‰ for Δ¹⁷O, 0.101‰ for Δ₃₅ and 0.098‰ for Δ₃₆ (Table 4). Table 4 (bottom row) also shows results for the air sample collected at the Utrecht University Campus. The isotopic results are similar to those of the air from the reference compressed air cylinder.

Table 4.

Isotopic composition including Δ₃₅ and Δ₃₆ values of tropospheric air in ‰ vs VSMOW (from reference compressed air cylinder and air from Utrecht University campus, The Netherlands) measured at different times during the measurement period. The measured standard errors and the errors expected based on the counting statistics (EECS) are also presented

Sl no	Date of measurement	δ17O (VSMOW)	EECS	δ18O (VSMOW)	EECS	Δ17O	EECS	Δ35 (ARF)	EECS	Δ36 (ARF)	EECS
1	19‐3‐18	11.893 ± 0.010	0.010	23.654 ± 0.014	0.003	−0.240 ± 0.012	0.010	1.391 ± 0.152	0.167	2.547 ± 0.090	0.103
2	19‐3‐18	11.965 ± 0.013	0.008	23.742 ± 0.019	0.003	−0.214 ± 0.016	0.008	1.447 ± 0.155	0.167	2.517 ± 0.097	0.104
3	21‐3‐18	11.977 ± 0.063	0.008	23.855 ± 0.035	0.003	−0.259 ± 0.064	0.008	1.414 ± 0.153	0.170	2.524 ± 0.092	0.106
4	03‐4‐18	12.182 ± 0.010	0.010	24.125 ± 0.008	0.004	−0.192 ± 0.010	0.010	1.213 ± 0.225	0.224	2.391 ± 0.146	0.139
5	04‐4‐18	12.088 ± 0.007	0.008	24.016 ± 0.004	0.004	−0.230 ± 0.007	0.008	1.212 ± 0.142	0.189	2.381 ± 0.101	0.120
6	04‐4‐18	12.146 ± 0.010	0.010	24.131 ± 0.007	0.005	−0.230 ± 0.010	0.010	1.360 ± 0.203	0.240	2.526 ± 0.178	0.156
7	04‐4‐18	11.912 ± 0.008	0.008	23.621 ± 0.004	0.004	−0.205 ± 0.008	0.008	1.136 ± 0.126	0.190	2.306 ± 0.093	0.119
8	12‐4‐18	11.912 ± 0.007	0.008	23.724 ± 0.003	0.004	−0.258 ± 0.007	0.008	1.289 ± 0.150	0.184	2.386 ± 0.094	0.120
9	24‐4‐18	11.958 ± 0.008	0.008	23.746 ± 0.008	0.003	−0.223 ± 0.009	0.008	1.281 ± 0.119	0.156	2.320 ± 0.088	0.101
10	02‐5‐18	11.832 ± 0.008	0.008	23.499 ± 0.004	0.003	−0.223 ± 0.008	0.008	1.360 ± 0.155	0.177	2.564 ± 0.102	0.110
Average ± 1σ		11.986 ± 0.115		23.811 ± 0.216		−0.227 ± 0.021		1.310 ± 0.101		2.447 ± 0.098
Tropospheric air from Utrecht, The Netherlands
1	18–9‐18	11.979 ± 0.013	0.014	23.752 ± 0.006	0.006	−0.205 ± 0.013	0.014	1.279 ± 0.248	0.298	2.356 ± 0.164	0.186

3.1.6. Dependence of clumped isotope ratio (Δ) on bulk isotope ratio (δ value)

Clumped isotope measurements with isotope ratio mass spectrometers may exhibit a dependence of the clumped isotope ratio Δ on the bulk isotopic signature δ. A corresponding correction is found to be necessary for most of the CO₂ clumped isotope measurements using low‐resolution isotope ratio mass spectrometers.15, 19, 20, 21 In order to examine whether such an effect is present for our clumped isotope measurements of O₂ with the 253 Ultra, we isotopically scrambled two gases with widely different bulk isotopic compositions at high temperature (850°C). The isotopic ratios of the two gases, viz. IMAU O₂ and GEO O₂ before and after heating, are presented in Table 5. IMAU O₂ is our working gas and GEO O₂ is a commercial O₂ gas supplied by Air Liquide (Eindhoven, The Netherlands). GEO O₂ has very low bulk isotope ratios and high Δ₃₅ and Δ₃₆ values compared with IMAU‐O₂, but after isotopic equilibration at 850°C (for 2–3 h) in the presence of platinum, the Δ₃₅ and Δ₃₆ values of the two O₂ gases are identical, indicating full isotopic scrambling at this temperature. This also indicates the absence of any dependence of the Δ₃₅ and Δ₃₆ values on the bulk isotopic composition within our measurement precision. We note that heating does affect the δ¹⁷O and δ¹⁸O values, which is possibly due to some exchange with oxygen in the quartz material of the reactor at high temperature. However, this exchange does not affect the Δ₃₅ and Δ₃₆ values as these signatures are purely driven by exchange temperatures.

Table 5.

Stable isotopic composition (in ‰) including Δ₃₅ and Δ₃₆ values of the laboratory standards (IMAU O₂ and GEO O₂) before and after heating. Heating is carried out at 850°C for more than 2 h in presence of platinum sponge. The δ³⁵ and δ³⁵ values are expressed against IMAU O₂

O2 gas	δ17O (VSMOW)	δ18O (VSMOW)	δ35 (sam‐wg)	δ36 (sam‐wg)	Δ35 (ARF)	Δ36 (ARF)
Before heating
IMAU O2	9.260 ± 0.007	18.548 ± 0.008	−0.060 ± 0.049	0.068 ± 0.020	1.262 ± 0.115	2.475 ± 0.043
GEO O2	−20.729 ± 0.002	−37.921 ± 0.018	−81.962 ± 0.179	−105.984 ± 0.104	2.821 ± 0.128	4.423 ± 0.114
After heating at 850°C
IMAU O2	10.919 ± 0.014	21.899 ± 0.004	3.671 ± 0.109	4.178 ± 0.146	0.005 ± 0.116	0.027 ± 0.085
GEO O2	−18.549 ± 0.137	−33.647 ± 0.226	−78.327 ± 0.243	−102.022 ± 0.413	0.108 ± 0.105	0.080 ± 0.080

3.1.7. Possibility of measuring ¹⁷O¹⁷O in the 253 Ultra

We also investigated the possibility of measuring the rarest isotopologue ¹⁷O¹⁷O (mass 33.99826; 0.032 ppm in the atmosphere). The biggest challenge is the interference from ¹⁶O¹⁸O (mass 33.99408; abundance 4100 ppm). The required mass resolution to separate these two masses is 8132, so they can in principle be resolved, in the medium‐ or high‐resolution mode. However, the detection is complicated by the very low abundance of ¹⁷O¹⁷O compared with¹⁶O¹⁸O. Under the present measurement conditions (at medium resolution) the expected signal strength for ¹⁷O¹⁷O is 1450 cps when the signal for its isobar ¹⁶O¹⁸O is 4.11*10⁷ cps, so ¹⁷O¹⁷O needs to be resolved on the tail of the very large ¹⁶O¹⁸O peak. When going to high‐resolution mode the signals decrease correspondingly, and the small signal of ¹⁷O¹⁷O (a few hundred cps) is only observable with a CDD. High‐precision determination of ¹⁷O¹⁷O will therefore take even longer than that of the more abundant clumped isotopes. The ¹⁷O¹⁷O isotopologue is expected to carry similar information to the other two clumped isotopologues and its measurement is therefore not yet being targeted rigorously.

3.2. Stratospheric and tropospheric clumped isotope ratios

The Δ₃₅ and Δ₃₆ values along with the conventional isotope ratios measured in the stratospheric and upper tropospheric air O₂ are presented in Table 6. The δ¹⁷O and δ¹⁸O values show only a small spread, between 11.89 and 11.98‰ for δ¹⁷O values and between 23.60 and 23.77‰ for δ¹⁸O values (Table 6), and are indistinguishable from those observed in the lower troposphere (Table 4). The δ¹⁷O and δ¹⁸O values are in agreement with numerous previous investigations that span the range of 11.98 to 12.26‰ for δ¹⁷O values and 23.53 to 24.15‰ for δ¹⁸O values.22, 23, 24 The average ¹⁷O anomaly, defined by Δ¹⁷O = ln(δ¹⁷O + 1) − 0.516*ln(δ¹⁸O + 1) is found to be −0.23 ± 0.03‰. Note that the three‐isotope slope (λ) value used here is 0.516. When using the same slope, the data from Luz and Barkan25 yield a Δ¹⁷O value of −0.17, close to our result. Laskar et al26 reported the Δ¹⁷O value of atmospheric O₂ to be −0.22 ± 0.02‰ based on indirect measurement of car exhaust CO₂, as the source of the O₂ in the exhaust's CO₂ is atmospheric O₂, consumed during the combustion of fuel. The agreement with previous values for the bulk isotopes including Δ¹⁷O provides confidence that the extraction of O₂ from air is made without strong fractionation.

Table 6.

Sampling information and stable isotopic composition (in ‰) including Δ₃₅ and Δ₃₆ values of stratospheric and upper tropospheric O₂ samples collected on the GEOPHYSICA M55 aircraft over the Mediterranean. The samples were collected on 1 and 6 September 2016

Alt (km)*	Latitude (oN)#	Longitude (°E) #	Temp (°C)†	δ17O (VSMOW)	δ18O (VSMOW)	Δ17O	Δ35 (ARF)	Δ36 (ARF)
20.0–20.2	35.52	25.13	−58.7	11.920 ± 0.007	23.681 ± 0.003	−0.227 ± 0.007	1.439 ± 0.132	3.078 ± 0.073
20.0	34.52	27.79	−59.5	11.947 ± 0.006	23.711 ± 0.003	−0.216 ± 0.006	1.630 ± 0.129	2.950 ± 0.120
16.5–20.0	40.34	24.95	−61.1	11.918 ± 0.008	23.634 ± 0.003	−0.206 ± 0.008	1.579 ± 0.086	2.978 ± 0.088
20.0	39.06	25.57	−59.8	11.930 ± 0.009	23.736 ± 0.003	−0.245 ± 0.009	1.572 ± 0.093	3.021 ± 0.086
19.5–19.6	33.60	30.42	−64.1	11.975 ± 0.008	23.709 ± 0.003	−0.187 ± 0.008	1.657 ± 0.103	3.097 ± 0.121
19.4–19.7	36.70	27.07	−63.3	11.920 ± 0.007	23.603 ± 0.004	−0.188 ± 0.007	1.579 ± 0.088	2.903 ± 0.044
18.4–19.1	34.54	25.68	−66.4	11.892 ± 0.008	23.679 ± 0.003	−0.254 ± 0.008	1.653 ± 0.082	2.914 ± 0.073
17.6	34.30	28.57	−67.4	11.941 ± 0.007	23.652 ± 0.003	−0.192 ± 0.007	1.551 ± 0.118	2.865 ± 0.051
17.5–17.6	35.11	26.12	−63.3	11.945 ± 0.007	23.662 ± 0.003	−0.193 ± 0.007	1.605 ± 0.093	2.889 ± 0.060
17.5	35.58	24.89	−63.4	11.901 ± 0.008	23.770 ± 0.003	−0.291 ± 0.008	1.535 ± 0.163	2.923 ± 0.089
14.8–15.4	36.56	22.58	−60.8	11.951 ± 0.007	23.640 ± 0.003	−0.176 ± 0.007	1.092 ± 0.143	2.436 ± 0.092
9.7–10.8	36.24	23.18	−42.5	11.912 ± 0.008	23.670 ± 0.004	−0.229 ± 0.008	1.301 ± 0.156	2.518 ± 0.110
10.7–11.2	40.91	23.67	−48.7	11.906 ± 0.007	23.680 ± 0.003	−0.241 ± 0.007	1.137 ± 0.147	2.225 ± 0.077

^* The air samples were collected during a period of ~ 5 min, the range of the altitude covered during sampling is given.

^# Latitude and longitude at the starting point of sampling.

^† Temperature is the average temperature over the collection duration.

Figure 5 shows the Δ₃₅ and Δ₃₆ values measured in stratospheric and upper tropospheric O₂ obtained from the GEOPHYSICA air samples. The values lie in the range of 1.1 to 1.6‰ for Δ₃₅ and 2.2 to 3.1‰ for Δ₃₆ (Table 6). The Δ₃₅ and Δ₃₆ values are higher in the stratospheric samples than in those from the troposphere. The Δ₃₅ and Δ₃₆ values of the lower tropospheric air O₂ from the compressed air cylinder filled at Groningen and an air sample collected at the Utrecht University campus are also plotted in Figure 5 for comparison. The equilibrium Δ₃₅ and Δ₃₆ values calculated for the temperatures measured on the GEOPHYSICA aircraft during the sampling flights are plotted for comparison (lines in Figure 5A). The equilibrium Δ₃₅ and Δ₃₆ values at the time of sampling at Groningen are shown by the black square. Figure 5B shows the difference between the measured and the calculated equilibrium values. Stratospheric O₂ is close to thermodynamic equilibrium at the ambient temperatures, as the measured values closely follow the equilibrium line. Previous data7 showed that the rapid isotopic exchange due to the O(³ P) + O₂ reaction in the stratosphere caused full thermodynamic equilibration above ~22 km, whereas equilibration was not complete below 22 km. For our samples in the lower stratosphere, due to larger errors, it is not possible to decide from the clumped isotope values alone whether the isotopic equilibration is complete. However, the correlation with N₂O (see below) indicates that the air in the lower stratosphere still has a partially tropospheric character. This demonstrates that full isotope equilibrium has not yet been achieved in the lower stratosphere.

Figure 5.

(A) Vertical profiles of Δ₃₅ and Δ₃₆ for the StratoClim (GEOPHYSICA) samples. The Δ₃₅ and Δ₃₆ values plotted at 1000 m altitude represent multiple measurements of the tropospheric reference air cylinder that was filled in August 2017 at Groningen, The Netherlands, and a sample from Utrecht, The Netherlands. They are plotted at an altitude of 1000 m to distinguish them from the data of Yeung et al7; the real altitude is ~0 m a.s.l. The solid lines are the equilibrium Δ₃₅ and Δ₃₆ values calculated for the ambient temperatures measured on the GEOPHYSICA aircraft. The black squares represent the equilibrium Δ₃₅ and Δ₃₆ values, estimated based on the local temperatures in The Netherlands during sampling. The Δ₃₅ and Δ₃₆ values in the stratosphere and troposphere reported by Yeung et al7 are also presented for comparison (in the altitude range covered by our data). Error bars indicate 1 SE for 30–60 cycles of measurement for each sample. (B) Deviation of the measured Δ₃₅ and Δ₃₆ values from the equilibrium values expected at the ambient temperatures [Color figure can be viewed at wileyonlinelibrary.com]

The upper tropospheric Δ₃₅ and Δ₃₆ values are lower than the equilibrium values, whereas the lower tropospheric values are higher (Figure 5B). This is due to the lower temperatures in the upper troposphere which correspond to higher Δ₃₅ and Δ₃₆ values predicted from thermodynamic equilibrium than in the lower troposphere. No significant difference in the Δ₃₅ and Δ₃₆ values between the upper and lower troposphere is observed, indicating that tropospheric O₂ is vertically well mixed in terms of the clumped isotopic composition. In other words, the O₂ isotopic reordering time in the troposphere is longer than typical mixing time scales. The isotopic reordering times have been estimated to be a year in the upper troposphere and several years near the surface using O atom distributions calculated from the GEOS‐Chem atmospheric chemistry model.7

The Δ₃₅ and Δ₃₆ values of stratospheric and tropospheric air O₂ reported by Yeung et al7 up to the altitude of ~22 km are also presented in Figure 5A for comparison. In general, a clear distinction between the stratospheric and tropospheric Δ₃₅ and Δ₃₆ values is observed in both data sets. However, absolute differences of ~0.3 in Δ₃₅ and ~0.4‰ in Δ₃₆ between the two data sets are evident, which are beyond the analytical uncertainty of ~0.1‰ for both Δ₃₅ and Δ₃₆ (Table 4). The tropospheric Δ₃₅ values measured by Yeung et al7 vary between 0.6 and 1.2‰ with an average of 1.0‰ and the Δ₃₆ values between 1.7 and 2.3‰ with an average of ~2.0‰. The present values of tropospheric Δ₃₅ vary between 1.1 and 1.4‰ with an average of 1.3‰ and of Δ₃₆ between 2.3 and 2.6‰ with an average of ~2.4‰.

The two sets of samples have different geographical origins, but a difference of ~0.4‰ is not expected as the mixing times in the troposphere are fast (weeks within one hemisphere and 1 year for inter‐hemispheric exchange) compared with the isotope reordering time scale of the order of a decade. In addition, the samples from Yeung et al7 cover a wide range in time and space. We suggest that the differences are due to differences in isotope scales between the two laboratories, in particular the assignment of effective temperatures for exchange experiments which is used for calibration. In our laboratory we use the change in pressure as an indication of the effective gas temperature (see section 2.3), whereas Yeung et al7 used the surface‐weighted radiative power (Stefan‐Boltzmann law) of different cold and warm parts of the reactor to estimate the gas temperature. If we use the same approach for our experiments, the effective temperature for the lowest temperature experiment (cold bath at −77.8°C and Suprasil finger at +10°C) would be −55°C inside the chamber for this experiment, compared with −63 K for the pressure‐based approach. This could explain differences of 0.1‰ and 0.2‰ in the Δ₃₅ and Δ₃₆ values, respectively for the coldest reactor temperatures (which are similar to stratospheric temperatures). The discrepancy in estimating the effective temperature of equilibration of O₂ inside the chamber is probably a major reason for the differences in the Δ₃₅ and Δ₃₆ values between the present measurements and those reported by Yeung et al.7 The effect on the tropospheric Δ₃₅ and Δ₃₆ values, however, would be smaller. Another potential point of concern is that the effective temperature, however determined, is an average over very different temperature conditions in the reactor, changing from hot near the illuminated Suprasil finger to cold at the outer surface. Thus, reactor geometry and illumination conditions may also play a role. Although the differences between stratospheric and tropospheric results reported by Yeung et al7 and our new results are similar and lead to similar conclusions, the significant discrepancy of the absolute values highlights the need for an inter‐laboratory comparison and careful assessment of the calibration scale.

Figure 6 shows the correlation between Δ₃₆ in O₂ and the N₂O mole fractions measured on the same samples. N₂O is almost exclusively produced at the Earth's surface from natural and anthropogenic sources and destroyed almost completely in the stratosphere by UV photolysis and reaction with O(¹ D).27, 28 Therefore, N₂O is a useful tracer for stratospheric photochemistry and is often used as “pseudo vertical coordinate” for stratospheric studies, where lower N₂O values indicate higher altitudes. A negative correlation is observed between Δ₃₆ and N₂O values for stratospheric samples (note that the N₂O scale is reversed). The photochemical N₂O removal is a slow process with timescales of months to years. The correlation between N₂O and Δ₃₆ therefore probably reflects mixing between fresh tropospheric air with higher N₂O and lower Δ₃₆ values and older stratospheric air with lower N₂O and higher Δ₃₆ values. This also implies that the lower stratospheric air is not in full isotopic equilibrium and retains some signatures of the troposphere, since isotope equilibration would eradicate the tropospheric signal and the correlation with N₂O. The correlation is not observed in the troposphere because the troposphere is well mixed without any significant vertical gradient in N₂O and Δ₃₆.

Figure 6.

Cross plot of Δ₃₆ and N₂O. Color codes (scale on the right) indicate altitude at which samples were collected. The best‐fit line is drawn for the stratospheric samples only. Errors in N₂O mole fractions are smaller than the symbol sizes [Color figure can be viewed at wileyonlinelibrary.com]

4. CONCLUSIONS

We demonstrated that the 253 Ultra high‐resolution stable isotope ratio mass spectrometer allows clumped isotope analysis of O₂ in atmospheric air. Isobaric interferences of ³⁶Ar and H³⁵Cl for ¹⁸O¹⁸O and ³⁵Cl for ¹⁷O¹⁸O are easily resolved at medium mass resolution. Isotopic equilibration experiments using O₂ heated to 850°C and photochemical isotope exchange via O(³ P) + O₂ at lower temperatures are used to convert the measured values into the Absolute Reference Frame for clumped isotope analysis through an empirical transfer function. A critical parameter for this calibration is the effective isotope exchange temperature in the photolysis reactor, and uncertainties in assigning the correct temperature could be the cause of an absolute difference of ~0.4‰ between the data presented here and a previously published dataset.7 Analyzed stratospheric air samples collected with the M55 GEOPHYSICA aircraft show that the clumped isotopic composition of O₂ is close to thermodynamic equilibrium at the ambient temperature, because of rapid equilibration due to the O(³ P) + O₂ isotopic exchange reaction. However, the correlation with N₂O suggests that isotopic equilibration is not complete in the lower stratosphere. The clumped isotope composition of tropospheric O₂ deviates significantly from the thermodynamic equilibrium at the respective local temperatures and there is no significant vertical clumped isotope gradient in the troposphere. This shows that the isotopic resetting time (several years)7 is much smaller than the tropospheric vertical mixing time scale (weeks). These observations indicate that either isotopic reordering is only partial in the troposphere and tropospheric O₂ retains part of the isotopic signature acquired in the stratosphere or that tropospheric O₂ is isotopically reordered in the upper troposphere at an effective temperature of ~ −40°C. This is being investigated with measurements and model data.

Supporting information

Table S1. Δ₃₅ and Δ₃₆ values of the equilibrated and heated O₂ samples at different temperatures expressed in ‰ against IMAU O₂, the working gas. These values are plotted against the corresponding thermodynamically predicted values to construct the empirical transfer functions (Figure 4 in the main text).

Table S2. Δ₃₅ and Δ₃₆ values before (measured) and after conversion to the absolute scale using the empirical transfer function (See Figure 4 in the main text). The measured clumped isotope data are the difference between the Δ₃₅ and Δ₃₆ values of the samples and the heated O₂. Tropospheric air from the surface level was collected from Groningen and Utrecht, The Netherlands and stratospheric and upper tropospheric O₂ samples were collected using the GEOPHYSICA M55 aircraft (see main text for details and discussion).

Data S1. Supporting information

Laskar AH, Peethambaran R, Adnew GA, Röckmann T. Measurement of ¹⁸O¹⁸O and ¹⁷O¹⁸O in atmospheric O₂ using the 253 Ultra mass spectrometer and applications to stratospheric and tropospheric air samples. Rapid Commun Mass Spectrom. 2019;33:981–994. 10.1002/rcm.8434