Atmospheric autoxidation is increasingly important in urban and suburban North America

Significance

Unimolecular hydrogen shift reactions to peroxy radicals have been shown to be important in the atmospheric photooxidation of isoprene and α-pinene. These studies also report the efficient generation of highly oxidized organic molecules known to contribute to particle formation and growth. Here, we quantify the rate of this oxidation pathway for peroxy radicals produced in the oxidation of n-hexane under conditions relevant to the atmosphere. The results suggest that autoxidation pathways are competitive against bimolecular reactions for a broad range of substrates, including many that result from urban emissions. The formation of organic hydroperoxides from atmospheric autoxidation has unknown implications for air quality.

Abstract

Gas-phase autoxidation—regenerative peroxy radical formation following intramolecular hydrogen shifts—is known to be important in the combustion of organic materials. The relevance of this chemistry in the oxidation of organics in the atmosphere has received less attention due, in part, to the lack of kinetic data at relevant temperatures. Here, we combine computational and experimental approaches to investigate the rate of autoxidation for organic peroxy radicals (RO₂) produced in the oxidation of a prototypical atmospheric pollutant, n-hexane. We find that the reaction rate depends critically on the molecular configuration of the RO₂ radical undergoing hydrogen transfer (H-shift). RO₂ H-shift rate coefficients via transition states involving six- and seven-membered rings (1,5 and 1,6 H-shifts, respectively) of α-OH hydrogens (HOC-H) formed in this system are of order 0.1 s⁻¹ at 296 K, while the 1,4 H-shift is calculated to be orders of magnitude slower. Consistent with H-shift reactions over a substantial energetic barrier, we find that the rate coefficients of these reactions increase rapidly with temperature and exhibit a large, primary, kinetic isotope effect. The observed H-shift rate coefficients are sufficiently fast that, as a result of ongoing NO_x emission reductions, autoxidation is now competing with bimolecular chemistry even in the most polluted North American cities, particularly during summer afternoons when NO levels are low and temperatures are elevated.

The gas-phase oxidation of organic compounds in the atmosphere proceeds through a number of reactive pathways. It is well established that reactions are initiated by oxidants including OH, NO₃, and O₃, and that, in the presence of oxygen, organic peroxy radicals (RO₂) are usually formed (1). The subsequent chemistry of RO₂ is diverse and depends on the chemical state of the atmosphere. Traditionally, bimolecular reaction with NO, HO₂, or other RO₂ has been assumed to dominate the fate of these radicals. Gas-phase autoxidation previously received significant attention only in combustion chemistry where high temperatures promote the process by permitting the reactants to overcome sizeable energetic barriers. The critical reaction in autoxidation, which generally governs the overall kinetics, is an intramolecular hydrogen shift to the RO₂. This reaction produces hydroperoxyalkyl radicals (often denoted QOOH), which are known intermediates in autoignition (2, 3). QOOH have only recently been observed (4). While several studies conducted at elevated temperatures have suggested that autoxidation is important in tropospheric chemistry (5–9), experimental kinetic measurements at atmospherically relevant temperatures have been lacking.

Several studies now report the atmospheric significance of RO₂ H-shift chemistry (10–24). Crounse et al. (15) suggested that this mechanism may explain the rapid oxygenation of hydrocarbons that contribute to particle growth. Subsequently, autoxidation was implicated in the generation of low-volatility molecules resulting from a single addition of OH or O₃ to monoterpenes (25–30). In these systems, autoxidation reactions proceed through successive isomerizations and O₂ additions, resulting in the formation of molecules with high O/C ratios and, often, multiple hydroperoxide groups (31). Such compounds have recently been observed to undergo gas–particle transfer (32) and have been shown to be important in particle nucleation (33–50).

While appreciation for the importance of autoxidation is increasing, significant shortcomings exist that preclude adequate characterization of its impact. Research to date has failed to fully describe the autoxidation mechanism of monoterpenes. Attempts have been made to explore autoxidation using cyclohexene as a model system (51, 52). Even in such simplified systems, however, multiple QOOH are formed, each of which can proceed through a large number of possible pathways to produce closed-shell products. Thus, elucidation of the mechanisms has proved challenging.

Here, we use both experimental and computational methods to determine the room temperature rate constants for autoxidation in a simple system—peroxy radicals produced via photooxidation of n-hexane in the presence of NO. The existence of an autoxidation pathway in this system has previously been demonstrated at elevated temperature (5). In the atmosphere, oxidation of hexane by the hydroxyl radical (OH) in the presence of NO produces alkoxy radicals, an example of which is shown in Scheme 1. Some of these alkoxy radicals can isomerize and react with O₂ to yield hydroxyperoxy radicals. For simplicity, our experiments use 2-hexanol as the precursor to produce a suite of these hydroxyperoxy radicals. In urban regions, the expectation is that these RO₂ react further with NO, ultimately producing hydroxy carbonyls, HO₂, and NO₂. Autoxidation, on the other hand, competes with the second NO reaction as shown in Scheme 1. In contrast to the mechanism proposed to explain the highly oxidized products observed in the OH and O₃ initiated oxidation of monoterpenes (25–30), the RO₂ in the hexane system primarily proceed through only a single H-shift yielding QOOH, which react with O₂ to produce ketohydroperoxides and HO₂. This simplification enables unambiguous experimental constraints for the specific RO₂ H-shift rate coefficients.

Scheme 1.

The mechanism to produce the 2,5 RO₂ (orange box) from n-hexane in the atmosphere is shown. This RO₂ reacts with NO to produce a hydroxy ketone (RONO₂, as shown in Scheme 3, are also produced in a minor channel). Competing with this chemistry is a unimolecular 1,6 RO₂ H-shift (autoxidation), which produces a ketohydroperoxide and HO₂, after further reaction with O₂. The first-order rate constants are provided at 300 K and 1 atm of pressure.

Results and Discussion

Computational Approach.

We calculate the rate constants of the H-shift reactions using multiconformer transition state theory (MC-TST) (53–56). The MC-TST expression for a rate constant is given by the following:

$$k_{\mathrm{MC-TST}}=\kappa \frac{k_{\mathrm{B}}T}{h}\frac{{\displaystyle {\sum }_i^{\mathrm{all}\hspace{0.17em}\mathrm{TS}\hspace{0.17em}\mathrm{conf}\hspace{0.17em}.}\exp \left(\frac{-\mathrm{\Delta }{\mathrm{E}}_i}{k_{\mathrm{B}}T}\right)Q_{\mathrm{TS},i}}}{{\displaystyle {\sum }_j^{\mathrm{all}\hspace{0.17em}\mathrm{R}\hspace{0.17em}\mathrm{conf}.}\text{exp}\left(\frac{-\mathrm{\Delta }{\mathrm{E}}_j}{k_{\mathrm{B}}T}\right)Q_{\mathrm{R},j}}}\exp \left(-\frac{E_{\mathrm{TS},0}-E_{\mathrm{R},0}}{k_{\mathrm{B}}T}\right),$$

where k_B is the Boltzmann constant, h is Planck’s constant, T is the temperature, Q_TS,i is the partition function for the ith transition state (TS) conformer, and ΔE_i is the difference in zero-point corrected energy between the ith TS conformer and the lowest energy TS conformer. E_TS,0 is the zero-point corrected energy of the lowest energy TS conformer. The analogous symbols apply for the reactant conformers. $\kappa$ is the tunneling correction factor. Here, we use the 1D Eckart tunneling approximation, which takes the forward and reverse barrier height and the imaginary frequency of the TS as input (57). The partition functions, energies, barrier heights, and imaginary frequencies needed to calculate k_MC-TST are obtained following the approach described by Møller et al. (56). Briefly, ωB97X-D/aug-cc-pVTZ was used for the geometries, frequencies, partition functions, zero-point energy corrections, and relative energies between unique conformers. The conformers were located by a systematic conformer search using molecular mechanics methods (58–60). CCSD(T)-F12a/VDZ-F12 single-point energy calculations were performed for more accurate electronic energies in the barrier heights (61–65). See SI Appendix for a detailed description.

Experimental Approach.

The RO₂ studied in this work were prepared via oxidation of 2-hexanol by OH (Scheme 2) in a ∼1-m³ environmental chamber made of Teflon. To determine the rate constants of the H-shifts, we studied the competition between bimolecular and unimolecular chemistry in a suite of experiments with differing concentrations of NO and HO₂, thereby producing a range of RO₂ bimolecular lifetimes $\left({\tau }_{\text{bimolecular}}\right):$

$${\tau }_{\text{bimolecular}}=\frac{1}{k_{{\text{RO}}_2+\text{NO}}\left[\text{NO}\right]+k_{{\text{RO}}_2+{\text{HO}}_2}\left[{\text{HO}}_2\right]}.$$

Determination of the concentrations of NO and HO₂ in our experiments is described in SI Appendix. The rate constants ($k_{{\text{RO}}_2+\text{NO}}$ and $k_{{\text{RO}}_2+{\text{HO}}_2}$) are taken from the literature (66) and described further in SI Appendix. We assume the ratio of the rate constants $\left(k_{{\text{RO}}_2+\text{NO}}/k_{{\text{RO}}_2+{\text{HO}}_2}\right)$ is isomer independent.

Scheme 2.

We oxidize 2-hexanol with OH in air to produce a suite of RO₂ radicals including the 2,5 RO₂ (orange box) shown in Scheme 1.

We use measurements of the organonitrates (RONO₂), produced as minor products in the RO₂ + NO channel, to probe the bimolecular chemistry. Following oxidation, the RONO₂ isomers were separated by gas chromatography and detected with chemical ionization mass spectrometry (GC-CIMS) using CF₃O⁻ (m/z = 85) as a reagent ion (Fig. 1). GC-CIMS has been documented extensively (14, 15, 24, 67–70). A deconvolution algorithm described in SI Appendix was required to analyze the chromatograms as it was not possible to increase the column length to fully separate the RONO₂ isomers without suffering significant isomer-specific transmission losses. Quantification became more difficult (with resulting higher uncertainty) for experiments at elevated temperature and/or long bimolecular lifetimes due to the higher water vapor concentration (∼300 ppmv) that results from diffusion of water through the Teflon chamber walls.

Fig. 1.

A chromatogram of five first-generation organonitrate (RONO₂) isomers formed in the OH oxidation of 2-hexanol. We measure the isomer distribution relative to the 2,3 isomer at different bimolecular lifetimes (τ_bimolecular). Loss of the 2,4 and 2,5 RONO₂ is evident in the experiment at τ_bimolecular ∼ 4 s (black) compared with the experiment at τ_bimolecular < 0.03 s (red). Details on structural assignment can be found in SI Appendix.

The H-shifts from the 2,3 RO₂ are calculated to be orders of magnitude slower than bimolecular chemistry for all experiments reported here (Table 1 and SI Appendix, Table S5). We therefore assume that ${\tau }_{\mathrm{2,3}\hspace{0.17em}{\text{RO}}_2}={\tau }_{\text{bimolecular}}$. For the 2,4 and 2,5 RO₂ isomers, however, we find that the amount of time available to react with NO is shortened by unimolecular chemistry:

$${\tau }_{{\text{RO}}_2}=\frac{1}{k_{{\text{RO}}_2+\text{NO}}\left[\text{NO}\right]+k_{{\text{RO}}_2+{\text{HO}}_2}\left[{\text{HO}}_2\right]+k_{\text{unimolecular}}}.$$

As a result, the yields of the 2,4 and 2,5 RONO₂ relative to that of the 2,3 RONO₂ serve as a sensitive probe of unimolecular chemistry. A depiction of the method is shown in Scheme 3.

Table 1.

H-shift rate coefficients (seconds⁻¹) and factors derived from theory and experiment

^*For the temperature-dependent rate expressions, refer to SI Appendix. Uncertainty in the calculated rate coefficients is estimated to be less than a factor of 10.

^†Reported values are scaling factors that afford the best fit to the experimental data, assuming a fixed ratio between the calculated diastereomer rate constants and an initial racemic mixture of isomers. Uncertainty is estimated as described in SI Appendix. The experiments are unable to differentiate the S,R and S,S isomers.

^‡Data for 318 K are available in SI Appendix.

Scheme 3.

Peroxy radicals produced in the reaction of OH with 2-hexanol produce organonitrates (RONO₂, blue boxes) after reaction with NO, in a branching ratio measured here to be ∼25% at 296 K. Reaction with HO₂ produces hydroperoxides (ROOH). In competition with bimolecular chemistry, the 2,4 RO₂ undergoes a unimolecular H-shift to produce QOOH. By measuring the changing yield of the 2,4 and 2,5 RONO₂ isomers relative to that of the 2,3 RONO₂, an isomer that isomerizes negligibly under the conditions of these experiments, we experimentally constrain the H-shift rate constant. The chemistry of the 2,5 RO₂ radical (not shown) parallels that of the 2,4 isomer.

H-Shift Rate Coefficients.

As shown in Fig. 2, the observed ratio of the 2,4 and 2,5 RONO₂ to the 2,3 RONO₂ decreases as bimolecular lifetimes extend beyond 1 s. At 296 K, the equivalence point in the ratio of the 2,4 and 2,5 RONO₂ relative to the 2,3 RONO₂ observed at ${\tau }_{\text{bimolecular}}\approx 10\hspace{0.17em}\text{s}$ is consistent with unimolecular chemistry occurring at a rate of ∼0.1 s⁻¹.

Fig. 2.

Comparison of experimental (black dots) and computational (red lines) results at 296 K for the 1,5 H-shift (Top) and 1,6 H-shift (Bottom). We demonstrate the difference in calculated 2,5 RO₂ H-shift rate coefficients between S,R and S,S diastereomers with the two dashed red lines that, when combined assuming an initial racemic mixture, produce the solid red line. The gray shaded region represents the range of uncertainty in the calculated rate coefficients.

Consistent with a large energetic barrier encountered along the H-shift reaction coordinate, the falloff occurs at a shorter bimolecular lifetime at elevated temperature (71) (Table 1 and SI Appendix). Providing further evidence for our assignment of the mechanism, the falloff occurs at a bimolecular lifetime more than 20 times greater when deuterium is substituted α to the OH group (HOC-D; SI Appendix). This is consistent with the expected primary kinetic isotope effect for RO₂ H-shifts (25, 52, 72).

In Fig. 2, comparisons are shown between the measurements and a model using the calculated rate coefficients for the H-shift reactions (Table 1). The 1,4 H-shift from the 2,3 isomer is calculated to be very slow (<10⁻⁴ s⁻¹ at T = 296 K) and, as discussed in SI Appendix, a 1,5 H-shift from this isomer was considered but is also expected to be slow. In contrast, the calculated rate coefficients of the 1,5 and 1,6 H-shifts from the 2,4 and 2,5 RO₂ are orders of magnitude faster. Surprisingly, the calculated H-shift rate coefficients of the S,S and S,R isomers of the 2,5 RO₂ differ significantly. As seen in Fig. 3, the origin of this difference is a change in the hydrogen bond-like interaction between the hydroxy group and the carbon-bonded oxygen atom of the peroxy radical moiety between the reactant and the TS in the S,R diastereomer. This produces a ∼1 kcal/mol difference in barrier height and enhances the rate constant of the S,R diastereomer by a factor of ∼5. However, illustrating the difficulty in generalizing such behavior, opposing effects related to the barrier height, partition functions, and tunneling result in similar calculated rate coefficients for the 1,5 H-shift from S,R and S,S isomers of the 2,4 RO₂ (SI Appendix).

Fig. 3.

ωB97X-D/aug-cc-pVTZ optimized structures of the lowest-energy conformers of the reactant and TS. The S,R diastereomer of the 2,5 RO₂ exhibits a hydrogen bond-like interaction which stabilizes the TS. Consequently, the H-shift rate of this isomer is enhanced by a factor of ∼5. Green halos highlight atoms involved in the hydrogen bond-like interaction. Blue halos are used when no such interaction exists.

In the model used in Fig. 2 (solid red line), we assume that a racemic mixture is produced when O₂ adds to the alkyl radical and that the rate coefficients of RO₂ H-shifts for enantiomeric pairs are identical (Table 1). Simulations where the peroxy radicals are either entirely (R,R; S,S) or (R,S; S,R) are shown as dashed lines.

The experimental factors in Table 1 are the multiplicative scaling of the calculated rate constants required to best fit the experimental data (SI Appendix, Figs. S8–S10). For instance, a factor of 1.0 would indicate that a best fit was achieved without scaling the calculated rate constants. The best fit to the experimental data at 296 K require scaling of the calculated values by less than a factor of 3 for both the 1,5 and 1,6 H-shifts, well within the combined uncertainties. The reasonable agreement between the experimental and computational results for these H-shifts suggests that the computational approach used here, and described fully by Møller et al. (56), provides an efficient method applicable to a much broader range of substrates than can plausibly be investigated in the laboratory.

Autoxidation Products.

We observe a CIMS signal at m/z 217 (a cluster of CF₃O⁻ with a compound of molecular weight 132 amu), corresponding to the mass of expected autoxidation products, ketohydroperoxides. We assign this signal to the two RO₂ α-OH H-shift reactions shown in Scheme 4. Consistent with an autoxidation mechanism, the absolute yield of m/z 217 $(\mathrm{\Delta }m/z\hspace{0.17em}217/\mathrm{\Delta }2\text{-}\mathrm{hexanol})$ at similar bimolecular lifetimes increases with temperature.

Scheme 4.

Formation mechanism of ketohydroperoxides (blue highlighted). Hydrogen atoms involved in the H-shifts are orange highlighted. Also shown is the route to hydroxy ketohydroperoxides, suspected products of a 1,5 H-shift from the 2,5 RO₂. Approximate, calculated rate coefficients are provided where available at 296 K.

An additional signal at m/z 233 (m/z 234 with D substitution) was observed, consistent with formation of a hydroxy ketohydroperoxide that arises from the 2,5 RO₂ through successive isomerizations as indicated in Scheme 4. The formation of this compound was enhanced by deuterium substitution at the α-OH center, which slows the 1,6 H-shift channel. The signal is much smaller in the nondeuterated experiments as the 1,6 H-shift outruns this chemistry. See SI Appendix for further details.

Similar ketohydroperoxides were previously detected in the low NO oxidation of C₁₂ alkanes (73–75). Although autoxidation was not discussed, second-generation alkoxy radicals form hydroxyperoxy radicals that almost certainly undergo H-shifts at rates comparable to those reported here. The ketohydroperoxides were observed to partition to the particle phase with simultaneous conversion to peroxyhemiacetals. Additionally, a cyclization pathway from γ-ketohydroperoxides to form endoperoxides was suggested, and is similar to a pathway discussed in SI Appendix and observed elsewhere (9, 76).

Atmospheric Implications.

In the atmosphere, the bimolecular lifetimes of peroxy radicals typically range from 1 s to more than 100 s. The lowest radical abundances (and therefore long bimolecular RO₂ lifetimes) are characteristic of attenuated UV environments (e.g., at night or in shaded regions below thick cloud or tree canopies) or regions remote from anthropogenic NO_x emissions. Due to emission reductions from power generation and transportation, however, NO_x levels are declining rapidly across North America, reaching levels unprecedented in the past several decades (77, 78). The 2013 SENEX and SEAC4RS aircraft campaigns, for example, sampled large swaths of the southeastern United States including areas significantly influenced by urban emissions. NO mixing ratios were often <100 pptv and nearly always <1 ppbv (79, 80). Even more impressive, NO_x levels in Pasadena, California, declined by more than a factor of 2 between 2010 and 2017 (81). In August 2017, we measured [NO] below 500 pptv on several weekend afternoons, corresponding to RO₂ lifetimes longer than 10 s. With afternoon temperatures typically exceeding 305 K, the autoxidation chemistry described here is now competing with reactions between peroxy radicals and NO in the middle of one of North America’s most polluted cities.

While the rate of autoxidation is highly dependent on the substrate, this chemistry is undoubtedly important for many of the organic compounds emitted into the urban atmosphere. A recent review of vehicle emissions and urban aerosol formation speculated that autoxidation might play a role in the degradation of certain unsaturated compounds (82), but its role in alkane oxidation was not appreciated. Alkanes constitute a substantial fraction of urban nonmethane hydrocarbon emissions (83–87), and n-alkanes with greater than five carbons are known emissions from gasoline- and diesel-powered vehicles arising from both incomplete combustion and fuel evaporation (88, 89). These compounds will undergo oxidation pathways that are nearly identical to those reported here for n-hexane.

The importance of organic hydroperoxide formation via autoxidation will depend on the fate and toxicity of these compounds. Peroxides are reactive oxygen species that are known to produce oxidative stress in plants and animals (90–92). Additionally, the rate of oxidant (e.g., ozone) and aerosol formation is almost certainly sensitive to autoxidation. Because autoxidation leads to the degradation of volatile organic compounds without converting NO to NO₂, ozone formation may be slower when this chemistry is important. While the impact on aerosol formation is less clear, it is likely to be enhanced because autoxidation adds oxygen with minimal fragmentation. For example, Zhao et al. (93) recently showed that while NO_x emissions are lower with the latest vehicle emissions control technology, the organic aerosol yield is greater than from the emissions using older-generation technology. Although the authors do not provide a mechanistic explanation, the higher yields are likely attributable, in part, to hydroperoxide formation via an autoxidation mechanism.

As a result of highly successful policies to reduce emissions of NO_x, our results suggest that autoxidation is now becoming an important pathway for urban photochemistry across North America. However, the photochemical models that have been used to inform these policies have little if any validation in the low-NO regimes we are now experiencing. Thus, there is a risk that attainment of ozone compliance may occur at the expense of other air quality goals because of more efficient hydroperoxide and aerosol formation. It is thus imperative that our understanding of the low-NO chemistry for the suite of organic compounds typically found in the atmosphere advance quickly and that monitoring efforts to quantify low-NO processes, such as autoxidation, be undertaken with haste.

Methods

The experimental apparatus, including the GC-CIMS technique, has been previously described and is outlined in SI Appendix for these experiments (14, 15, 67–70). Experiments are performed in a ∼1-m³ Teflon environmental chamber. In nearly all experiments, the precursor used was 2-hexanol, while CH₃ONO was used as a photolytic source of HO₂ and NO (and thus OH). NO was added before oxidation for experiments focusing on short RO₂ lifetimes and was quantified using a Teledyne 200EU chemiluminescence NO_x analyzer. For experiments without additional NO added, we use established methods to estimate its abundance (and that of HO₂) as detailed in SI Appendix. We attempted to replicate the method of Jorand et al. (5) near room temperature, but, for reasons described in SI Appendix, these experiments were not successful.

Experimental Uncertainty.

Considerable (>50%) experimental uncertainty arises in our estimate of ${\tau }_{\text{bimolecular}}$ for $\tau >10\hspace{0.17em}\text{s}$, due to imprecision in the interpretation of chromatographic peaks, and temperature fluctuations in our chamber. Details are provided in SI Appendix.

Computational Uncertainty.

We estimate the uncertainty of the calculated rate constants to be less than a factor of 10. The uncertainties arise primarily from the barrier height, tunneling correction, and the partition functions. Due to error cancelation, the ratio of the theoretical rate constants for these different H-shifts are likely more accurate than the absolute rate coefficients. This is especially true for reactions of the same or very similar molecules. See SI Appendix for details.