Large, Negative Atmospheric Carbon Monoxide Clumped Isotope Values Result From Kinetic Isotope Fractionation, Tracing OH• Reactivity

Abstract

Because of its global abundance and reactivity with hydroxyl radicals (OH•), tropospheric carbon monoxide indirectly impacts the lifetimes of other OH•‐reactive gases, in particular methane and reactive hydrocarbons. The origin and chemistry of atmospheric CO have been studied using stable isotopes. Both ¹³CO and C¹⁸O undergo isotopic fractionation during its main chemical loss reaction, CO + OH•. The kinetic isotope effect (KIE) for ¹³CO is mass dependent, with a value of ∼5‰; ¹²CO reacts faster than ¹³CO with OH. Whereas C¹⁸O + OH• exhibits an inversely mass dependent KIE ∼−10‰. We hypothesize these KIEs result in a relative depletion of ¹³C¹⁸O, a CO clumped isotope. To test this, we collected CO from air samples on Long Island, NY, and discovered a −3 to −8‰ difference in the clumped isotope ratio, Δ₃₁, relative to a random distribution of ¹³C and ¹⁸O in CO. A clear negative trend between [CO] and Δ₃₁ is driven by two factors: (a) the atmospheric addition of CO from either a primary or secondary source with a Δ₃₁ of ∼0‰ and (b) the continuing reaction of CO with OH•, leaving the remaining CO pool relatively depleted in ¹³C¹⁸O. This is analogous to the mechanism that determines CO Δ¹⁷O values. This study is among the first to show clumped isotope fractionation resulting from atmospheric chemistry and not thermal equilibration, which may inform the identification of clumped isotope KIEs in other atmospheric trace gases. These first Δ₃₁ observations motivate future experimental and observational studies targeted at characterizing the clumped isotopes of CO sources, background CO, and experimentally fractionated CO.

Introduction

Atmospheric carbon monoxide (CO) is a reactive, abundant trace gas that has a significant impact on tropospheric chemistry largely as a result of its reaction with hydroxyl radicals (OH•). Because of its reactivity, CO is one of the largest global sinks for OH•. And since OH• is the most important oxidant for almost all reactive trace gases, CO indirectly impacts the chemical lifetimes of OH• reactive gases, including methane (CH₄). Together with the nonmethane hydrocarbons (NMHCs), CO is implicated in ∼0.25°C of warming from 2010 to 2019 relative to 1850–1900, which is larger than the warming potential of N₂O or halogenated gases (Masson-Delmotte et al., 2021). Analysis of historical CO concentrations from polar ice records reveals a peak during the early to mid-1900s (Petrenko et al., 2013; Wang et al., 2010, 2012). This upward trend was followed by a recent decline, attributed to motor vehicle regulations in North America and Europe (Masson-Delmotte et al., 2021). Significant sources of CO include fossil fuel combustion, atmospheric oxidation of methane and other hydrocarbons, direct emission via biomass burning, and other anthropogenic combustion processes. Atmospheric concentrations of CO can range from as low as 35 ppbv (Mak et al., 1992; NOAA Earth System Research Laboratories, 2023; Novelli et al., 1994; Petron et al., 2021) to well over 500 ppbv in urban settings. Typical concentrations in the remote Northern Hemisphere atmosphere range from about 60 to 150 ppbv, with a seasonal minimum in the summer and seasonal maximum during the winter. This seasonality is largely driven by the seasonal variation in OH• (and therefore the chemical lifetime of CO). The global seasonally averaged lifetime of CO is on the order of 3 months, but CO lifetime ranges from days to many months (e.g., Park, Emmons, et al., 2015).

The most significant sources of atmospheric CO are known to be fossil fuel combustion, biomass burning, NMHC oxidation, and CH₄ oxidation. While the exact contribution from each source will vary with time and space, some sources, such as CH₄-derived CO, are quite well constrained. The abundance of ¹³CO and C¹⁸O (indicated by δ¹³C and δ¹⁸O of CO) are impacted by the isotopic signatures from specific sources. For example, combustion-derived CO is enriched in ¹⁸O, and methane-derived CO is depleted in ¹³C, with respect to atmospheric δ¹³C and δ¹⁸O values (Figure 1). Model studies, combined with new observations of atmospheric CO δ¹³C and δ¹⁸O, have been used to help constrain relative source strengths (Park, Wang, et al., 2015). These differences offer valuable insights into assessing the relative contributions of different sources, thereby aiding in constraining the global atmospheric CO budget. However, the isotopic composition of atmospheric CO is also impacted by its reaction with OH•.

[IMAGE OMITTED. SEE PDF]

The reaction CO + OH• is the most important loss mechanism for atmospheric CO and thus determines its atmospheric lifetime. It has been known for decades that the isotopic fractionation during this reaction for the single substituted isotopologues ¹³CO, C¹⁸O, and C¹⁷O, are both varied and unusual. In the case of the most abundant minor isotopologue, ¹³C¹⁶O, the KIE of the reaction CO + OH• → CO₂+H• reflects a “normal” mass dependence, with a measured ¹³k/¹²k (where *k is the rate constant for *CO or C*O isotopologues) of ∼0.995, depending on pressure (Stevens et al., 1980). Therefore, the rate constant is larger for the lighter isotopologue by about 0.5%, or 5‰, where ¹³KIE, (‰) = (1−¹³k/¹²k) × 1,000. ¹²C¹⁸O, on the other hand, exhibits an inverse mass dependence, with a ¹⁸k/¹⁶k of ∼1.01, or −10‰, depending on pressure and temperature (Stevens et al., 1980). Thus, C¹⁸O reacts faster with OH• than does ¹³CO. Furthermore, the least abundant, singly substituted isotopologue of CO, ¹²C¹⁷O, has a mass independent KIE, leading to a ¹⁷O excess (Δ¹⁷O) of ∼5‰ (Röckmann, Brenninkmeijer, Neeb, & Crutzen, 1998, Röckmann, Brenninkmeijer, Saueressig, et al., 1998). While it has been postulated that the inconsistent observed KIEs are related to the formation and stability of the HOCO• intermediate as well as electron tunneling (Johnson et al., 2011, 2014), a detailed theoretical explanation of the sum of these observations has yet to be realized. Because these previously observed reaction rates are only for the singly substituted isotopologues of CO, we hypothesized there exists an unusual (i.e., not controlled by equilibrium thermodynamics) fractionation of the clumped isotopologue, ¹³C¹⁸O, in atmospheric CO that results from the same reaction mechanism. Isotope ratio measurements of this CO isotopologue may therefore offer new and consequential insight into the CO + OH• → CO₂ + H• reaction mechanism and the abundance of reactants.

Clumped isotope measurements of atmospheric gases have largely focused on the major constituents: N₂, O₂ and CO₂, due to the relative ease of their sampling and pre-existing purification methods. In air, the clumped isotopologues of N₂ (¹⁵N¹⁵N) and O₂ (both ¹⁷O¹⁸O and ¹⁸O¹⁸O) have been observed to be out of equilibrium with respect to ambient atmospheric temperatures, sometimes by several per mil for O₂ (Yeung, 2016; Yeung et al., 2012) and exceptionally by 10's of per mil for N₂ (Yeung et al., 2017). In N₂, KIEs associated with destruction of molecular nitrogen are hypothesized to drive this disequilibrium. In O₂, the photosynthetic production of O₂ results in unequal pairing of ¹⁸O from differential fractionation in the water-splitting reaction (2H₂O → O₂ + 4H⁺; Yeung et al., 2015). These effects are notably different than the various natural processes that have been shown to equilibrate multiply substituted molecules at ambient temperatures, for example, oxygen isotope exchange between CO₂ and H₂O (Affek, 2013; Clog et al., 2015) and by the O(³P) + O₂ isotope exchange reaction (Yeung et al., 2012, 2014). The only atmospheric trace gas whose clumped isotopes have been characterized is methane, CH₄ (Haghnegahdar et al., 2023). Tropospheric methane clumped isotopes reflect a KIE associated with atmospheric oxidation, as well as physical mixing with methane sources that have exotic clumped isotope values associated with KIEs from methanogenesis. Clumped isotope measurements of N₂O, H₂, and C₂H₆, all trace gases in the atmosphere, are reported, but there are no data from natural samples and thus their atmospheric fractionations are unknown (Clog et al., 2018; Magyar et al., 2016; Popa et al., 2019; Taguchi et al., 2022). Nevertheless, a useful paradigm can be constructed from these gaseous clumped isotope measurements. Chemical reactions that equilibrate atmospheric gas isotopologue abundances have isotope ratio values that can be predicted from thermodynamic calculations (Wang et al., 2004) and the ambient temperatures at the time of collection (e.g., Laskar & Liang, 2016). Whereas those reactions with known kinetic isotope effects are likely to deviate from thermal equilibrium predictions with a sign and magnitude that may be useful for refining atmospheric budgets that are underconstrained using singly substituted isotopologues alone (e.g., Haghnegahdar et al., 2023; Yeung et al., 2015).

Here we present the first clumped isotope measurements of atmospheric carbon monoxide, which was oxidized to and analyzed as CO₂. We also show that laboratory heating or generation of CO at temperatures of 1000°C for sufficiently long timescales equilibrates ¹³C¹⁸O isotopologues in a manner consistent with ab initio isotopologue abundance calculations (Wang et al., 2004). The clumped isotope fractionation associated with CO oxidation by the Schütze reagent appears to be stochastic, thus halving the ¹³C¹⁸O abundance transferred to the CO₂ analyte. Next, we show several years of measurements of tropospheric CO collected from the campus of Stony Brook University (SBU), Stony Brook, NY, USA, that are several per mil lower than a stochastic distribution of ¹³C and ¹⁸O. We hypothesize that kinetic effects associated with atmospheric CO oxidation by OH• yield negative clumped isotope values and that the observed correlation between CO concentrations and Δ₃₁, the isotope ratio that quantifies ¹³C¹⁸O abundance in CO after oxidation to CO₂, results from differential expression of this fractionation in natural samples.

Materials and Methods

Equilibrium Gases and Atmospheric Collection

In order to assess possible isotope fractionation associated with CO oxidation to CO₂ for gas-source isotope ratio mass spectrometry, and to provide a thermodynamic reference point for atmospheric samples (e.g., as was done by Huntington et al., 2009 for CO₂; Stolper et al., 2014 for CH₄, Yeung et al., 2017 for N₂, etc.), we sought to equilibrate CO isotopologues by heating aliquots of gas at elevated temperatures. To do this, we experimented with three methods: (a) heating sealed aliquots of pure CO gas in quartz glass tubing with several mm of 99.9% platinum (Pt) wire (0.25 mm dia, 99.9% (metals basis), Alfa Aesar), (b) heating calcium carbonate minerals with zinc metal powder (<10 μm, >98%, Sigma Aldrich), which yields CO by carbonate decomposition above ∼750°C according to 1 $${\text{CaCO}}_{3}+\text{Zn}\to \text{CaO}+\text{ZnO}+\text{CO}$$ and (c) the Boudouard reaction, which is the disproportionation of CO into CO₂ and graphite, or the reverse, at temperatures >700°C: 2 $$2\text{CO}\leftrightarrow {\text{CO}}_{2}+\mathrm{C}$$

In all cases reaction aliquots were sealed in quartz glass tubes and heated to 350, 375, 450, 650, or 1000°C in calibrated tube furnaces for 187 to 60,620 min, that is ∼3 hr to 42 days, respectively (Table S1; Figures S1–S3 in Supporting Information S1). These heated clumped isotope equilibration experiments were quenched using a flow of compressed air or submersion into a beaker of water in order to reach room temperature in <30 s.

Twenty four atmospheric samples were processed from 3 December 2019 to 4 March 2022 from an intake vent on the roof at the SBU School of Marine and Atmospheric Sciences (Stony Brook, NY, USA; 40.9062° N, −73.1192° E), which is connected to the extraction line described below. Given the large sample requirements for clumped isotope measurements (10's of μmols), ∼6 hr of collection per day was needed to process enough air at ambient CO concentrations typical of the northeastern United States. Sample collection and combination details are in Table S2.

The CO extraction system used to process both thermally equilibrated CO and atmospheric samples is described in detail by Mak and Kra (1999), but briefly, air is metered into the extraction line using a mass flow controller drawn through a rotary vane pump. Two cryogenic traps immersed in liquid nitrogen remove condensable gases from the air stream. The air is then passed through a bed of Schütze reagent, consisting of iodine pentoxide (I₂O₅) on an acidified silica gel. This oxidant quantitatively converts CO to CO₂ at room temperature (∼25°C). The resulting CO₂ is then cryogenically trapped and transferred to a calibrated manometer. From this pressure measurement we calculate the CO mixing ratio of the original air sample. Oxygen isotope ratios were corrected for Schütze oxidation using the equation (Brenninkmeijer, 1993) 3 $${\delta }^{18}{\mathrm{O}}_{\text{CO}}=2{\delta }^{18}{\mathrm{O}}_{\text{CO}2}-\left(2{\delta }^{18}{\mathrm{O}}_{\text{calibCO}2}-{\delta }^{18}{\mathrm{O}}_{\text{calibCO}}\right)$$ where δ¹⁸O_CO2 is the measured value, δ¹⁸O_calibCO2 is the measured value of the CO calibration gas-derived CO₂, and δ¹⁸O_calibCO is that of the CO calibration gas. To obtain enough CO₂ for clumped isotope analysis, several days' worth of CO collections were combined. Samples were combined on the basis of their measured CO mixing ratio, not chronologically, and daily aliquots of CO-derived CO₂ were combined cryogenically on a standalone vacuum extraction line. As many as 6 but as few as 3 days were combined (Table S2), depending on the CO mixing ratio; lower daily mixing ratios required more sample integration. Sample volume weighted average CO mixing ratios (±1 SD) are indicated in Table S1. Infrequently, but when CO mixing ratios were high, all CO was collected during a single day (e.g., Julian day 21208; Table S1).

Isotope Ratio Mass Spectrometry

CO₂ samples were analyzed on a Thermo Scientific MAT 253 Plus isotope ratio mass spectrometer in the SBU Department of Geosciences after cryogenic and chromatographic purification on a custom CO₂ vacuum line, similar to those described by Passey et al. (2010) and Henkes et al. (2013). Because clumped isotope analysis of CO₂ is routine in many laboratories worldwide, our methods could be adopted by other groups wishing to study CO clumped isotopes. The CO-derived CO₂ was run along with clumped isotope equilibrium reference gases, generated by CO₂–H₂O oxygen isotope exchange at 30°C or heating of pure CO₂ to 1000°C for 2 hr, and the isotopologue data was corrected according to Dennis et al. (2011) using IUPAC ¹³R, ¹⁷R, and ¹⁸R values (Brand et al., 2014). This constitutes the “carbon dioxide equilibrium scale” or CDES reference frame. We also analyzed CO₂ that had been heated to 1000°C and that had δ¹³C and δ¹⁸O values similar to the atmospheric and experimental CO samples (i.e., −24.5‰ VPDB and 23.8‰ VSMOW, respectively) and treated these samples as “unknowns” during the processing of clumped isotope data. These measurements yielded a corrected average clumped isotope value Δ₄₇ of 0.051 ± 0.010‰ (1 SD, n = 6), which is close to, but slightly higher than the theoretical Δ₄₇ value of 0.027‰ for random distribution of CO₂ isotopologues at 1000°C (Wang et al., 2004). Δ₄₇ is calculated as: 4 $${{\Delta }}_{47}=\left[\left({R}^{47}/{R}^{47\ast }-1\right)-\left({R}^{46}/{R}^{46\ast }-1\right)-\left({R}^{45}/{R}^{45\ast }-1\right)\right]\times 1000$$ where Rⁱ = mass i/mass 44, and the asterisk indicates calculated ratios for isotopologues at abundance levels predicted from a random distribution of ¹³C and ¹⁸O using the measured δ¹³C and δ¹⁸O values and R values for C and O isotope standards VPDB and VSMOW (Affek & Eiler, 2006). Rⁱ* is different for each sample measurement because they vary in their δ¹³C and δ¹⁸O. Since R⁴⁶/R⁴⁶* and R⁴⁵/R⁴⁵* ratios are near 0, Δ₄₇ can be redefined as: 5 $${{\Delta }}_{47}\cong \left({R}^{47}/{R}^{47\ast }-1\right)\times 1000$$

And following the results of the CO clumped isotope equilibration experiments reported on in Section 3.1, we define the abundance of measured ¹³C¹⁸O isotopologues as: 6 $${{\Delta }}_{31}=0.5\times {{\Delta }}_{47}$$

The oxidation of CO to CO₂ appears to add ¹⁶O or ¹⁸O to ¹³C randomly, thereby halving the measured Δ₄₇ value. Reported CO δ¹³C and δ¹⁸O values are calculated relative to a reference CO₂ from Oztech Trading Corporation (δ¹³C = −3.59‰ VPDB, δ¹⁸O = 24.91‰ VSMOW). In this study we ignore the impact of potentially anomalous CO Δ¹⁷O values (Röckmann, Brenninkmeijer, Neeb, & Crutzen, 1998, Röckmann, Brenninkmeijer, Saueressig, et al., 1998) because it's impact on measured Δ₄₇ values is <0.1‰, which is substantially lower in magnitude than the Δ₃₁ range we observed in natural air samples. Furthermore, we do not expect ¹⁷O anomalous CO for our experimental samples (e.g., Equations 1 and 2).

Results and Discussion

Thermal Equilibration Experiments

The products of CO heating or generation at temperatures at 1000°C yielded Δ₄₇ values that conform with predictions from equilibrium clumped isotope distribution calculations (Figure 2 and Figure S1 in Supporting Information S1; Henkes et al., 2024; Wang et al., 2004). Aliquots of CO heated to 1000°C in the presence of Pt wire, generated by calcite decomposition and Zn metal oxidation at 1000°C (Equation 1), and produced by the Boudouard reaction at 1000°C (Equation 2) for sufficiently long (Figures S2 and S3 in Supporting Information S1) had an average Δ₄₇ value of 0.065 ± 0.009‰ (1 SE, n = 18), which is statistically indistinguishable from aliquots of pure CO₂ with similarly light δ¹³C and δ¹⁸O values that were also heated to 1000°C in the same tube furnace (see Section 2.2; Δ₄₇ = 0.051 ± 0.010; unpaired t-test, p = 0.35). Δ₄₇ values of both CO heated to 650°C with Pt wire and calcite decomposition-derived CO at 775°C were slightly greater than the all-reaction 1000°C average Δ₄₇, as well as the 1000°C average for each reaction type, but low replication for both (n = 3 and n = 2, respectively) precludes statistically meaningful comparisons (Figures S1b–S1d in Supporting Information S1). Two attempts were made to thermally equilibrate CO at 350 and 375°C with Pt wire for several weeks each; their average Δ₄₇ values were 0.234‰ (n = 2) and 0.025‰ (n = 4), respectively (Figure S1b in Supporting Information S1). This is both higher (375°C) and lower (350°C) than the thermodynamic predictions. The CO triple bond is among the strongest in nature (=1,072 kJ/mol; for comparison, the N₂ triple bond strength is 942 kJ/mol) and the kinetics of bond reordering at such relatively low temperatures are unknown. We have no data at timescales <38 or >42 days at 375°C and only ∼63 days at 350°C and therefore cannot systematically evaluate the kinetics of this bond reordering. Furthermore, equilibration experiments conducted at 1000°C appeared to only achieve a Δ₄₇ value that matched thermodynamic predictions after 1,000 min for CO heating in the presence of Pt wire (Figure S2 in Supporting Information S1) and after 140 min for CO generation by carbonate decomposition (Figure S3 in Supporting Information S1).

[IMAGE OMITTED. SEE PDF]

These thermal equilibrium experiments allow us to evaluate the clumped isotope fractionation associated with O addition to CO by Schütze reagent during the CO extraction and collection process. Two possible scenarios for Schütze oxidation are either a stochastic, random addition of an oxygen atom to CO or a uniform, that is, ¹³C¹⁸O abundance-independent, ¹⁸O or ¹⁶O oxidation preference which would shift the measured Δ₄₇ to values with respect to equilibrium CO isotopologue abundances (Wang et al., 2004). The result of the former scenario would be a halving of ¹³C¹⁸O abundances as any non-stochastic equilibrium isotopologue would be diluted by ½ by the random addition of an additional oxygen (Equation 6). Because combinatorial effects are predictable for clumped isotopes (e.g., Yeung et al., 2015), it is possible to simply consider these effects during CO conversion to CO₂. There is no carbon isotope fractionation associated with the Schütze reagent, therefore we do not expect the oxygen isotope fractionation (Brenninkmeijer, 1993) to yield a combinatorial clumped isotope effect ⪆0‰, thus supporting a stochastic O isotope addition. The latter of the scenarios described above is imperfect because addition reactions differ in their isotope fractionation from bond breaking ones, but it would be analogous to the clumped isotope fractionation associated with phosphoric acid digestion of carbonate minerals (Defliese & Lohmann, 2015; Guo et al., 2009; Swart et al., 2019). These scenarios are represented graphically in Figure S1a in Supporting Information S1, with one curve showing an unfractionated conversion of ¹³C¹⁸O (Δ₃₁; Wang et al., 2004) to CO₂ and another showing the halving of those predictions upon conversation to CO₂. At high temperature the two curves converge, and the results of our high temperature thermal equilibrium experiments are not precise enough to support one mechanism over the other (Figure 2). At lower temperatures these curves diverge, and collectively our observations are more consistent with the stochastic addition of oxygen isotope by the Schütze reagent. On the contrary, predicted Δ₄₇ values following the CO Schütze reagent fractionation that instead approached equilibrium at laboratory room temperatures (∼25°C) would presumably yield higher measured Δ₄₇ values because thermal equilibrium values for CO₂ are ∼0.8‰ (Wang et al., 2004). A uniform, that is, across all equilibration temperatures and reaction types, positive Δ₄₇ offset is not observed in the heating experiment data (Figure 1 and Figure S1 in Supporting Information S1).

Atmospheric Measurements

Twenty four air samples were collected from the Stony Brook University campus on Long Island, NY between 3 December 2019 and 3 March 2022 and subsequently measured (Henkes et al., 2024). Observations presented here are generally consistent with the seasonal record of CO, δ¹³C, and δ¹⁸O previously determined from Montauk Point, Long Island, which is approximately 100 km east of Stony Brook (Figure 1; Mak & Kra, 1999). Variations in both δ¹³C and δ¹⁸O in CO over time are driven by both the temporally (and thus spatially) varying relative source strengths of CO with different isotopic signatures, as well as the temporally varying magnitude of the chemical loss rate via reaction with OH•. For a discussion of this readers are referred to other references (e.g., Mak & Kra, 1999; Mak et al., 2003, Park, Wang, et al., 2015; Park, Emmons, et al., 2015).

Figure 3 shows our Δ₃₁ measurements plotted versus manometrically determined CO concentration for each sample. The measured Δ₃₁ values range from −7.75 to −1.76‰, which is much lower than equilibrium values expected for the source Δ₃₁ of atmospheric CO. Even though atmospheric CO isotopes are known to be dominated by KIEs, a simple basis for evaluating these new data is to compare them with clumped isotope values predicted for thermal equilibrium. Three of the four principle sources of CO to the atmosphere, the exception being CH₄ and NMHC oxidation, generate CO at high temperatures and therefore with expected Δ₃₁ values near 0‰. Mak and Kra (1999) conclude that one of these, “upwind” fossil fuel combustion, is the dominant δ¹³C and δ¹⁸O determinant to CO collected on Long Island and likely over the entire northeastern United States (Figure 1). With some exceptions, our sample δ¹³C and δ¹⁸O largely overlap with their data. While we have not yet measured Δ₃₁ of possible atmospheric sources (e.g., tail pipe or industrial plant emissions), we do report Δ₃₁ for three industrial CO tanks, with differing δ¹³C and δ¹⁸O. All have Δ₃₁ values near 0‰ (Table S1). The source of CO of these purified gases is unknown, but CO purified from industrial hydrocarbon or coal combustion is common.

[IMAGE OMITTED. SEE PDF]

The relationship between Δ₃₁ and CO concentration, specifically that Δ₃₁ becomes continuously more negative with decreasing CO, requires that it is the CO removal process that is responsible for the continuing increase in magnitude of the Δ₃₁ signal. Because of this, we assert that this is evidence of the CO + OH• reaction fractionating clumped isotopes, just as it is known to do for δ¹³C, δ¹⁸O, and Δ¹⁷O. Thus, assuming that the dominant CO sources to tropospheric air that ended up over Long Island during our collection period have Δ₃₁ near 0‰, as supported by our own measurements of experimentally generated or thermally equilibrated CO, then the magnitude of Δ₃₁ is determined by the extent of chemical reactivity of CO via the CO + OH• reaction. This predicts that as the average chemical age of the CO increases, Δ₃₁ continues to become more negative. Precendent for this as a driver of the observed clumped isotope signal is the fractionation of Δ¹⁷O from the CO + OH• reaction, where the Δ¹⁷O of CO sources are also shown to be ∼0‰ (Röckmann, Brenninkmeijer, Neeb, & Crutzen, 1998, Röckmann, Brenninkmeijer, Saueressig, et al., 1998).

To demonstrate that the CO and hydroxyl reaction readily explains the observed range clumped isotope fractionation, we use the known kinetic isotope effects for CO δ¹³C and δ¹⁸O (¹³KIE and ¹⁸KIE; Röckmann, Brenninkmeijer, Neeb, & Crutzen, 1998) to predict that the clumped isotope kinetic isotope effect (“^13–18KIE”) for Δ₃₁ follows the product rule (Wang et al., 2016; Whitehill et al., 2017). This rule implies that the primary and secondary fractionation factors associated with CO oxidation (i.e., ¹³KIE × ¹⁸KIE) together yield ^13–18KIE. A ¹³KIE = 0.994 and an ¹⁸KIE = 1.01, which are expected values at near-surface temperatures and pressures, together predict a ^13–18KIE = 1.00344. It is possible to model the evolution of an assumed source CO δ¹³C, δ¹⁸O, and Δ₃₁ as the CO + OH• proceeds according to the equations: 7 $${\text{CO}}_{t}={\text{CO}}_{\mathrm{i}}\mbox{--}\left({}^{12}k\times \text{OH}\times {\text{CO}}_{\mathrm{t}-1}\times \mathrm{d}t\right)$$ where CO_t and CO_i are the number of CO molecules per cm³ at time t and initially (i), respectively. OH is the amount of OH (also molecules per cm³), which is constant, ¹²k is the CO + OH• reaction rate constant, because ∼99% of CO is ¹²C¹⁶O, and dt is the time step, 60 s. The reaction rate is typically expressed as cm³ molecule⁻¹ s⁻¹ so once the initial model values are selected, Equation 7 can be used to iteratively forward model the reaction. Differing reaction rates for CO isotopologues (e.g., *k) are thus expressed as 8 $${}^{\mathrm{x}}\mathrm{C}{\mathrm{O}}_{\mathrm{i}}={\text{CO}}_{\mathrm{i}}{\times }^{\ast }k$$ where x is the isotopologue mass associated with a heavy isotope substitution (e.g., x = 29 for * = 13) and *k = *KIE × k defines *k. Combining Equations 7 and 8 yields 9 $${}^{\mathrm{x}}\mathrm{C}{\mathrm{O}}_{t}={\text{CO}}_{\mathrm{i}}\mbox{--}\left({}^{\ast }k\times \text{OH}{\times }^{\mathrm{x}}{\text{CO}}_{t-1}\times \mathrm{d}t\right)$$

The δ values at any time t can thus be calculated as 10 $${\delta }^{13}{\mathrm{C}}_{t}=\left[\left(\left({}^{29}{\text{CO}}_{t}{/}^{28}{\text{CO}}_{t}\right){/}^{13}{\mathrm{R}}_{\text{VPDB}}\right)-1\right]$$ 11 $${\delta }^{18}{\mathrm{O}}_{t}=\left[\left(\left({}^{30}{\text{CO}}_{t}{/}^{28}{\text{CO}}_{t}\right){/}^{18}{\mathrm{R}}_{\text{VSMOW}}\right)-1\right]$$ where ¹³R_VPDB and ¹⁸R_VSMOW and the defined ratios for VPDB and VSMOW respectively. Relatedly, the accompanying change in Δ₃₁ is 12 $${{\Delta }}_{31t}=\left[\left(\left({}^{31}{\text{CO}}_{t}{/}^{28}{\text{CO}}_{t}\right)/\left({}^{31}{\text{CO}}_{\mathrm{i}}{/}^{28}{\text{CO}}_{\mathrm{i}}\right)\right)-1\right]$$ which can be added to the source Δ₃₁ depending on the modeled system.

This curve, shown in Figure 3, is identical to a Rayleigh fractionation curve of the form: 13 $${\delta }_{t}\text{or}{{\Delta }}_{t}={f}^{(\ast \text{KIE}-1)}\times \left(1000+{\delta }_{\mathrm{i}}\text{or}{{\Delta }}_{\mathrm{i}}\right)\mbox{--}1000$$ where the residual CO isotopic composition is calculated using the appropriate KIE (e.g., ¹⁸KIE, ¹⁸KIE, or ^13–18KIE from above), the fraction remaining, f, and the starting delta values (δ_i or Δ_i).

The solid and dashed line curves in Figure 3 are model solutions that assume a CO source Δ₃₁ = 0‰ and 700 ppb. The root mean squared error between our data and the model predictions (solid line) is low, just 0.63‰, and a logarithmic curve fit to the data (i.e., not Rayleigh model derived) predicts a slope, or ^13–18KIE, of 3.62‰ (r² = 0.82) that is virtually identical to the product rule prediction of 3.44‰. The dashed lines shown in Figure 3 are ^13–18KIE error envelopes from the propagation of error through the product rule assuming ±0.00075 for ¹³KIE and ±0.0005 ¹⁸KIE. All but three data points from 2020 fall within these conservative error bounds.

Atmospheric Transport Versus Chemical Aging

Two of the samples with the highest CO concentrations are from both the summer of 2020. These are unusually high CO concentrations compared to normal background conditions at this time on Long Island (Figure 1a). These samples were likely influenced by large wildfires in northern California, Oregon, mainly the Dixie and Bootleg fires, and Ontario, Canada that all started in early to-mid July and resulted in smoke visible by satellite in the northeastern US (2021 NICC Annual Report; Chow et al., 2022; NOAA Earth Global Systems Laboratory, 2021). What is the chemical age of the CO in those air parcels? Based on our hypothesis, if the elevated CO were instantly emitted from a biomass burning event, then that newly formed CO would have Δ₃₁ of ∼0‰. However, for those samples, Δ₃₁ is 2–3 ‰ more negative. Thus the CO has been “aged” as a result of oxidation via OH• following wildfire emission. That the Δ₃₁ of those high concentration samples also fall on the same trendline as all the other data points indicates that the effective age of those parcels is not significantly younger than background CO (<100 ppb), even though the CO concentration is elevated (Figure 3). HYSPLIT back trajectories show northern California and Pacific northwest or central Canadian origins for airmass on these collection days (Figures S4 and S5 in Supporting Information S1). Thus, the CO from those summertime high concentration samples originated from a distant source, and those parcels had chemically aged during the time it took to be transported. Distinctly, it is not the distance transported; it is the chemical age, which is dependent on OH• and time. Likewise, atmospheric mixing of one parcel with another would not alter the relationship of CO to Δ₃₁; it would merely result in the mass averaged chemical age of the two respective air parcels.

The Impact of Atmospheric Mixing on Δ₃₁

The clumped isotope effects of mixing gases of different δ¹³C and δ¹⁸O are predictable, well-known for other gases (Defliese & Lohmann, 2015; White & Defliese, 2023), and counter to conventional stable isotopes, non-linear (Eiler, 2007; Wang et al., 2004). Thus, it is possible to calculate the maximum impact of a clumped isotope effect on Δ₃₁ from physical mixing of atmospheric sources (or source and background CO levels) and evaluate it against our results. For example, a mixture of CO derived from fossil fuel combustion contributing 200 ppb CO, with δ¹³C = −27.5‰ (VPDB), δ¹⁸O = 22‰ (VSMOW), and Δ₃₁ = 0‰ to background CO at ∼100 ppb, with δ¹³C = −30‰, δ¹⁸O = 5‰, and Δ₃₁ = −6.5‰, results in a distinct curvilinear relationship shown in Figure 3, predicting Δ₃₁ values between 100 and 300 ppb that are approximately 1‰ higher than measured values, but that mostly fall inside of the error envelopes shown for the CO + OH• reaction curve. This same mixing scenario produces a small, +0.06‰ “mixing effect” anomaly for Δ₃₁. This clumped isotope effect is insensitive to the end-member Δ₃₁ values (i.e., what is described is the maximum deviation from algebraic two end-member mixing), and the magnitude is indistinguishable compared to the reported Δ₃₁ values. If there were large variations in Δ₃₁ from different sources in the atmospheric CO samples we measured, then we would not expect such apparent coherence between our atmospheric data and the CO + OH• reaction curve. Instead, we would expect a greater range in Δ₃₁ for a given atmospheric CO concentration. As discussed in the next section, future measurements will be targeted at better parsing fractionation from OH• reaction from the effects of source-background mixing.

In our initial atmospheric data set, we cannot preclude that mixing effects on the order of ∼1‰ are superimposed on the effects of CO + OH• Δ₃₁ fractionation. For example, in future sampling campaigns, more observations from stations with clearer seasonal CO variation and better understood CO source partitioning will help us understand the magnitude of the mixing on measured atmospheric Δ₃₁. Park, Wang, et al. (2015) used an Icelandic CO δ¹³C and δ¹⁸O time series and a chemical tracer model (MOZART-4) to show seasonal differences in contributions from fossil and biofuel CO emissions, which peak in March–April, biomass burning, which peaks in June–July, CH₄ and NMHC oxidation, which peaks in August-September. Despite this source variability, CO concentrations cyclically vary by only ∼100 ppb annually, which was consistent across at least 6 of the study years. In contrast, our Long Island samples spans several hundred ppb (Figures 1 and 3) and are less consistent over the several years of collection. A dominant CO + OH• signal in an Icelandic Δ₃₁ data set would be expected to produce a tight relationship with [CO] (as in Figure 3), whereas mixing of different CO sources during different times of the year may introduce greater Δ₃₁ variability from multiple different CO sources than the reaction curve alone would predict.

Future Work

The clumped isotope trends identified in atmospheric samples from the northeastern United States are compelling and allow us to formulate hypotheses about the fractionation accompanying the CO oxidation reaction, but validation of these requires additional research. Foremost, it is possible to derive clumped isotope KIE fractionation factors for the CO + OH• reaction by observing the change in Δ₃₁ with decreasing CO concertation in a experimental chamber with known source CO isotopic values, concentrations, and controlled OH• generation (e.g., Röckmann, Brenninkmeijer, Neeb, & Crutzen, 1998, Röckmann, Brenninkmeijer, Saueressig, et al., 1998). This will be necessary to confirm the ^13–18KIE predicted by the product rule in Section 3.2 and shown in Figure 3. Next, the Δ₃₁ from various atmospheric CO sources should reflect the high temperatures of their formation, with values close to 0‰. Future efforts to characterize sources will confirm this hypothesis, and add additional data to the known δ¹³C and δ¹⁸O ranges for CO sources. Finally, the Rayleigh distillation curve in Figure 3 predicts increasingly low Δ₃₁ values as CO concentrations approach the minimum observed values in the atmosphere. To confirm the maximum extent of fractionation Δ₃₁, measurements from clean air are needed, preferably from the southern hemisphere where year-round concentrations are <100 ppb. This would be in contrast to published values from Long Island (Figure 1a) and elsewhere in the northern hemisphere, which are typically >100 ppb. The curve in Figure 3 would predict geographically widespread Δ₃₁ values < −8‰ in the southern hemisphere.

A combination of these initial northern hemisphere observations, and future work to validate ^13–18KIE and the source and background air Δ₃₁ may allow for interpretations of new atmospheric Δ₃₁ values as a proxy for the extent of CO + OH• reaction, and therefore the average chemical age of that parcel of air. This could improve box model simulations of δ¹³C and δ¹⁸O as reflecting the relative CO contributions from isotopically distinct sources (Park, Wang, et al., 2015). And eventually, improvements in CO clumped isotope analysis that reduce sample size and throughput may allow for the analysis of historical or ancient air preserved in polar firn or ice cores samples (Wang et al., 2010, 2012). Presently, clumped isotope measurements require 10's of micromoles of CO₂ gas to measure Δ₃₁. This requires continuous air collection and CO oxidation for hours. An order of magnitude reduction in clumped isotope sample size requirements would make the challenge of making glacial ice Δ₃₁ measurements on par with ice core CH₄ radiocarbon analysis (e.g., Petrenko et al., 2009, 2013), which requires micromoles of sample from 1000s of kilograms of ice. Records of Δ₃₁ that extend back even hundreds of years would enable atmospheric comparisons across known climatic changes, such as the Little Ice Age, Medieval Warming Period, and present-day warming.

Conclusions

We have presented the first clumped isotope results, along with δ¹³C, δ¹⁸O, and concentration data, for carbon monoxide (CO) from the atmosphere. These are among the first trace gas atmospheric clumped isotope measurements of any kind, and show that sufficient quantities of ppb-level gas concentrations can be collected for clumped isotope analysis. The observed CO δ¹³C, δ¹⁸O, and concentration values from 2019 to 2022 are very similar to measurements made from Montauk Point, Long Island from 1996 to 1997, though some outliers during 2021 appear coincident with air masses originating from the western United States and central Canada when wildfires were likely an important CO source. The results from atmospheric CO clumped isotope measurements yield large, negative, “anti-clumped” Δ₃₁ values that are far out of equilibrium with ambient temperatures, but decrease with decreasing CO concentration. We conclude that this relationship results from carbon monoxide's reaction with hydroxyl radicals, and an indeterminant amount of mixing between source and background CO, but observe that unlike CO δ¹³C and δ¹⁸O it is unlikely there are meaningful Δ₃₁ differences between CO sources. We also separately evaluated thermal resetting CO clumped isotopes, and the effects of CO oxidation to CO₂ for isotope analysis, using three techniques to equilibrate CO clumped isotopes at elevated temperatures. These experimental results suggest that the oxidation reaction halves the Δ value of CO by randomly adding ¹⁶O or ¹⁸O. By demonstrating a reliance of CO clumped isotopes on known carbon and oxygen kinetic isotope effects we present a possible new isotopologue tracer for the CO + OH• reaction in the atmosphere, which is promising for developing revised atmospheric budgets for both reactants.

Acknowledgments

The authors declare no conflicts of interest. This research was supported by the U.S. National Science Foundation Division of Atmospheric Sciences (award #1927950). We thank Jabrane Labidi, Daniel Stolper, an anonymous reviewer, and editor-in-chief Susan Trumbore for very helpful, careful, and critical reviews of the original manuscript.

Conflict of Interest

The authors declare no conflicts of interest relevant to this study.

Data Availability Statement

New stable isotope data presented in this paper can be found in Supplementary Information Table S1 and consist primarily of CO concentrations and isotope values (Table S1). Data on the methods used to calculate the average atmospheric CO concentrations for our large clumped isotope samples is provided in Table S2. These supplemental data files can also be downloaded at .

References

Affek, H. P. (2013). Clumped isotopic equilibrium and the rate of isotope exchange between CO₂ and water. American Journal of Science, 313(4), 309–325. [DOI: https://dx.doi.org/10.2475/04.2013.02]