Reactive aldehyde chemistry explains the missing source of hydroxyl radicals

Hydroxyl radicals (OH) are the primary atmospheric oxidant controlling the removal of most trace gases and the troposphere’s self-cleaning capacity, influencing secondary pollution, air quality, health, ecosystems, and climate. Despite their central role, models systematically underestimate OH in environments with low NO and high VOCs, even when including updated isoprene chemistry. Prior attempts (e.g., the Leuven Isoprene Mechanism and RO2 self-reactions) reduce but do not eliminate the discrepancy, and measurement interferences have been shown negligible, indicating missing radical chemistry. This study hypothesizes that reactive aldehyde chemistry—autoxidation of carbonyl peroxy radicals derived from higher aldehydes producing hydroperoxyl-carbonyls (HPC) that photolyze to regenerate OH—constitutes a major missing OH source. The authors combine a meta-analysis of surface radical observations across major Chinese city clusters in warm seasons (2006–2019), quantum chemical calculations, and observation-constrained modeling to characterize and evaluate this mechanism and its broader atmospheric impact.

Multiple studies have documented underestimation of OH in low-NO, high-VOC settings (forests, suburban regions). Proposed explanations include the Leuven Isoprene Mechanism (LIM), involving intramolecular H-shifts in isoprene-derived RO2, and RO2 self-reactions, which help in forested areas but remain insufficient. Parameterized schemes of RO2 + X and HO2 + X pathways have also been used to match OH without co-producing HO2, highlighting the need for more detailed chemistry. Prior quantum and experimental work showed H-migration rates in carbonyl peroxy radicals increase with carbon number, becoming competitive with NO reactions for C5+ aldehydes; autoxidation potentials for aldehydes and other VOC classes (alkanes, aromatics, ketones, ethers, alcohols, alkenes) have been reported, forming highly oxidized products and potentially radicals via multi-step H-shifts. HPC photolysis producing OH has been demonstrated. These foundations motivate examining higher-aldehyde-driven RO2 autoxidation as a widespread OH source beyond isoprene-specific chemistry.

• Observational meta-analysis: Seven warm-season campaigns in China (2006–2019) across major regions (North China Plain: Yufa, Wangdu; Pearl River Delta: Backgarden, Heshan, Shenzhen; Yangtze River Delta: Taizhou; Sichuan Basin: Chengdu). OH and HO2 measured by LIF-FAGE; RO2 measured at select sites. Comprehensive meteorology, photolysis frequencies, NOx, O3, VOCs, and total OH reactivity (kOH) collected.
• Experimental budget analysis: Compared OH production (primary: HONO photolysis, O3 photolysis, alkene ozonolysis; secondary: HO2 + NO and HO2 + O3) with OH destruction (observed OH × kOH) to diagnose missing sources, focusing on 10:00–15:00.
• Quantum chemical calculations: Explored H-migration in R(CO)O2 from higher aldehydes. Structures and energies obtained using WMS//M06-2X/MG3S, benchmarked against high-level methods; IRC confirmed transition states. Reaction rate constants computed via multi-structural transition state theory with Eckart tunneling. For hexanal-derived peroxy radicals, the favored 1,7 H-shift has k = 0.321 s−1 at 298 K; prior work indicates iso-pentanal up to 0.58 s−1. Conformer considerations and conservative rate selection applied.
• Mechanism development (HAM): Higher Aldehyde Mechanism formulated: R(CO)O2 undergoes intramolecular H-shift to OOR(CO)OOH; reaction with NO forms OR(CO)OOH; two subsequent rapid H-shifts and O2 addition yield HO2 and hydroperoxyl-carbonyls (HPC). HPC photolyzes rapidly to produce OH; for conservative modeling, HPC photolysis frequency set to 10× the methacrolein photolysis frequency (~(0.6–0.9) × 10−11 around noontime). Rate constants and reactions summarized in supplementary materials.
• Box modeling and sensitivity tests: Zero-D box model constrained by observations (5-min averages), 2-day spin-up, dry deposition per Wesely et al. Base chemistry: RACM2. Sensitivity Test 1 adds LIM1; Sensitivity Test 2 adds HAM. Model uncertainty ~40%.
• Generalized parameterization (RAM): Reactive Aldehyde Mechanism extended to include other VOC classes (alkanes, alkenes, aromatics, OVOCs) capable of undergoing successive H-shifts to produce HPC-like species. Introduced an index φ representing generalized molar OH yield from VOC oxidation via HPC chemistry, derived by inverse modeling to match observed OH; φ determined for each campaign.
• OH recycling probability: Computed metric y from primary (P) and gross (G) OH formation (y = 1 − P/G) to assess system stability and evaluate impacts of LIM1 and RAM across campaigns and NO regimes. Global assessment included seven additional international campaigns with radical observations.

• Systematic OH underestimation by models at low NO (<1 ppb) across seven warm-season Chinese campaigns; the underestimation intensifies as NO decreases. Measurement interferences were found negligible.
• Missing OH sources correlate strongly with OVOC chemistry: a strong positive correlation between missing OH and the ratio k_OVOCs/k_NO, outperforming correlations based on total anthropogenic or biogenic VOC reactivity. OVOCs contribute about half of total VOC reactivity and peak around noon/afternoon, matching periods of largest OH underestimation.
• Quantum chemistry shows higher-aldehyde RO2 autoxidation is fast: for hexanal-derived RO2, the most favorable 1,7 H-migration has k = 0.321 s−1 at 298 K; prior estimates indicate iso-pentanal RO2 up to 0.58 s−1, allowing autoxidation to outcompete RO2 + NO. Conservative rates were used for modeling.
• Mechanistic pathway (HAM): RO2 autoxidation forms OR(CO)OOH intermediates that, after two fast H-shifts and O2 addition, produce HO2 and HPC; HPC photolyzes rapidly to yield OH. A conservative HPC photolysis frequency was set to 10× methacrolein (~(0.6–0.9) × 10−11 around noontime).
• Experimental budgets indicate unclassical OH sources contribute 15–68% of total OH production (10:00–15:00) under low-NO conditions.
• HAM explains 10–90% of the unclassical OH sources across campaigns; HAM contributions exceed LIM1 by factors of 1.7–8.2. In lower [VOCs]/[NOx] settings (Taizhou, Wangdu, Shenzhen), HAM explains nearly all missing OH within uncertainties.
• Residual missing OH (68–86%) in certain campaigns (Backgarden, Yufa, Heshan, Chengdu) likely due to unaccounted high-carbon aldehyde emissions (≥C8/C9), underestimation of HPC-like products from other VOCs (alkanes, ethers, alcohols, alkenes), and underestimation of HPC photolysis for different functional groups; recently proposed aromatic autoxidation contributed negligibly to OH in these campaigns.
• Generalized RAM with fitted φ improves agreement between observed and modeled OH and HO2. Derived φ ranges 0.4–2.2 across the seven Chinese campaigns; in the broader global analysis φ values of 0.2–2.93 reconcile OH recycling probability.
• Observed OH recycling probability is typically 0.8–1.0 and shows weak dependence on NO. The base model underestimates this stability, especially at low NO; adding LIM1 and RAM closes the gap across 14 campaigns on three continents.
• Regime dependence: LIM1 impacts are strongest under ultra-low NO (forested areas), while RAM operates effectively across low to high NO (suburban to urban) regimes; both mechanisms are insignificant under ultra-high NO where RO2 + NO dominates.

The study demonstrates that reactive aldehyde autoxidation, particularly from higher aldehydes, constitutes a major, previously underrepresented OH regeneration pathway in low-NO, high-VOC environments. By linking the missing OH source to OVOC reactivity and quantifying rapid intramolecular H-shifts in carbonyl peroxy radicals, the work provides a mechanistic basis for generating HPC species whose photolysis efficiently regenerates OH. Incorporating this chemistry (HAM) substantially closes experimental OH budget gaps and increases modeled OH recycling probability, outperforming isoprene-specific LIM1 across suburban and urban settings. The generalized RAM framework captures additional VOC classes capable of multistep autoxidation, further improving agreement with observed OH and HO2. These results address the long-standing OH underestimation in diverse environments, indicating that OVOCs are not solely OH sinks but can be net sources via photolabile intermediates. The findings imply a broader and more stable tropospheric oxidation capacity than previously modeled, especially as NOx emissions decline, and underscore the need to update atmospheric chemical mechanisms to include reactive aldehyde autoxidation and HPC photolysis.

Reactive aldehyde chemistry—autoxidation of higher-aldehyde-derived RO2 forming photolabile hydroperoxyl-carbonyls—explains much of the missing OH source in low-NO, high-VOC regimes. The Higher Aldehyde Mechanism (HAM) accounts for a large fraction (10–90%) of unclassified OH production across multiple campaigns and generally exceeds isoprene-based LIM1. Extending to a generalized Reactive Aldehyde Mechanism (RAM) that encompasses additional VOC classes closes discrepancies in OH recycling probability across global datasets, demonstrating significance from suburban to urban regimes. Given projected declines in NOx under carbon neutrality policies, RAM’s role in sustaining OH and atmospheric self-cleaning will likely grow. Future research should refine kinetic parameters (H-shift rates, product yields) for higher aldehydes and other VOCs, quantify HPC formation spectra and photolysis frequencies across functional groups, and integrate these updates into mainstream mechanisms (e.g., RACM2, MCM) to improve air quality and climate assessments.

• Primary emissions of higher-carbon aldehydes (≥C8/C9), e.g., from biomass burning and cooking, were not explicitly included; their oxidation could yield RO2 with even higher H-shift rates than modeled.
• Potential underestimation of total HPC-like species, as many VOCs beyond higher aldehydes (long-chain alkanes, ethers, alcohols, alkenes) can form such products.
• HPC photolysis parameters likely vary with functionalization and may have been underestimated; reported photolysis frequencies may differ by over a factor of four.
• The RAM approach employs a simplified, campaign-fitted index (φ) to represent generalized OH yields, introducing parameterization uncertainty and limiting mechanistic specificity.
• Model uncertainty (~40%) arises from measurement constraints and kinetic parameters; remaining OH underestimation in some campaigns indicates additional missing processes.
• Recently proposed aromatic autoxidation pathways were found negligible for OH generation in the studied campaigns, but their broader relevance may vary by environment and requires further evaluation.