Fenton chemistry at aqueous interfaces

Significance

The Fenton reaction, Fe²⁺ + H₂O₂, plays fundamental roles in vivo and in advanced oxidation processes. Its mechanism and the identity of the intermediates involved, however, remain controversial. Here we present direct, mass-specific evidence of the prompt formation of mono- and poly-iron Fe^IV=O (ferryl) species on the surface of aqueous FeCl₂ microjets exposed to gaseous H₂O₂ or O₃ beams. Remarkably, Fe²⁺ ions at the aqueous surface react with H₂O₂ and O₃ >10³ times faster than Fe(H₂O)₆²⁺ in bulk water. Our results suggest that interfacial Fenton and Fenton-like chemistries could play a more significant role than hitherto envisioned.

Abstract

In a fundamental process throughout nature, reduced iron unleashes the oxidative power of hydrogen peroxide into reactive intermediates. However, notwithstanding much work, the mechanism by which Fe²⁺ catalyzes H₂O₂ oxidations and the identity of the participating intermediates remain controversial. Here we report the prompt formation of O=Fe^IVCl₃⁻ and chloride-bridged di-iron O=Fe^IV·Cl·Fe^IICl₄⁻ and O=Fe^IV·Cl·Fe^IIICl₅⁻ ferryl species, in addition to Fe^IIICl₄⁻, on the surface of aqueous FeCl₂ microjets exposed to gaseous H₂O₂ or O₃ beams for <50 μs. The unambiguous identification of such species in situ via online electrospray mass spectrometry let us investigate their individual dependences on Fe²⁺, H₂O₂, O₃, and H⁺ concentrations, and their responses to tert-butanol (an ·OH scavenger) and DMSO (an O-atom acceptor) cosolutes. We found that (i) mass spectra are not affected by excess tert-butanol, i.e., the detected species are primary products whose formation does not involve ·OH radicals, and (ii) the di-iron ferryls, but not O=Fe^IVCl₃⁻, can be fully quenched by DMSO under present conditions. We infer that interfacial Fe(H₂O)_n²⁺ ions react with H₂O₂ and O₃ >10³ times faster than Fe(H₂O)₆²⁺ in bulk water via a process that favors inner-sphere two-electron O-atom over outer-sphere one-electron transfers. The higher reactivity of di-iron ferryls vs. O=Fe^IVCl₃⁻ as O-atom donors implicates the electronic coupling of mixed-valence iron centers in the weakening of the Fe^IV–O bond in poly-iron ferryl species.

High-valent Fe^IV=O (ferryl) species participate in a wide range of key chemical and biological oxidations (1–4). Such species, along with ·OH radicals, have long been deemed putative intermediates in the oxidation of Fe^II by H₂O₂ (Fenton’s reaction) (5, 6), O₃, or HOCl (7, 8). The widespread availability of Fe^II and peroxides in vivo (9–12), in natural waters and soils (13), and in the atmosphere (14–18) makes Fenton chemistry and Fe^IV=O groups ubiquitous features in diverse systems (19). A lingering issue regarding Fenton’s reaction is how the relative yields of ferryls vs. ·OH radicals depend on the medium. For example, by assuming unitary ·OH radical yields, some estimates suggest that Fenton’s reaction might account for ∼30% of the ·OH radical production in fog droplets (20). Conversely, if Fenton’s reaction mostly led to Fe^IV=O species, atmospheric chemistry models predict that their steady-state concentrations would be ∼10⁴ times larger than [·OH], thereby drastically affecting the rates and course of oxidative chemistry in such media (20). Fe^IV=O centers are responsible for the versatility of the family of cytochrome P450 enzymes in catalyzing the oxidative degradation of a vast range of xenobiotics in vivo (21–28), and the selective functionalization of saturated hydrocarbons (29). The bactericidal action of antibiotics has been linked to their ability to induce Fenton chemistry in vivo (9, 30–34). Oxidative damage from exogenous Fenton chemistry likely is responsible for acute and chronic pathologies of the respiratory tract (35–38).

Despite its obvious importance, the mechanism of Fenton’s reaction is not fully understood. What is at stake is how the coordination sphere of Fe²⁺ (39–46) under specific conditions affects the competition between the one-electron transfer producing ·OH radicals (the Haber–Weiss mechanism) (47), reaction R1, and the two-electron oxidation via O-atom transfer (the Bray–Gorin mechanism) into Fe^IVO²⁺, reaction R2 (6, 23, 26, 27, 45, 48–51):

Ozone reacts with Fe²⁺ via analogous pathways leading to (formally) the same intermediates, reactions R3a, R3b, and R4 (8, 49, 52, 53):

At present, experimental evidence about these reactions is indirect, being largely based on the analysis of reaction products in bulk water in conjunction with various assumptions. Given the complex speciation of aqueous Fe²⁺/Fe³⁺ solutions, which includes diverse poly-iron species both as reagents and products, it is not surprising that classical studies based on the identification of reaction intermediates and products via UV-absorption spectra and the use of specific scavengers have fallen short of fully unraveling the mechanism of Fenton’s reaction. Herein we address these issues, focusing particularly on the critically important interfacial Fenton chemistry that takes place at boundaries between aqueous and hydrophobic media, such as those present in atmospheric clouds (16), living tissues, biomembranes, bio-microenvironments (38, 54, 55), and nanoparticles (56, 57).

We exploited the high sensitivity, surface selectivity, and unambiguous identification capabilities of a newly developed instrument based on online electrospray mass spectrometry (ES-MS) (58–62) to identify the primary products of reactions R1–R4 on aqueous FeCl₂ microjets exposed to gaseous H₂O₂ and O₃ beams under ambient conditions [in N₂(g) at 1 atm at 293 ± 2 K]. Our experiments are conducted by intersecting the continuously refreshed, uncontaminated surfaces of free-flowing aqueous microjets with reactive gas beams for τ ∼10–50 μs, immediately followed (within 100 μs; see below) by in situ detection of primary interfacial anionic products and intermediates via ES-MS (Methods, SI Text, and Figs. S1 and S2). We have previously demonstrated that online mass spectrometric sampling of liquid microjets under ambient conditions is a surface-sensitive technique (58, 62–67).

Results and Discussion

Fig. 1 A and B shows negative ion ES mass spectra of the surface of 10 μM and 100 μM FeCl₂ aqueous microjets, respectively, while being exposed to O₂(g) and O₃(g)/O₂(g) mixtures for contact times τ of the order of a few tens of microseconds. Gas–liquid contact times τ correspond to the estimated lifetimes of the microjets, i.e., before they are broken up by the nebulizer gas into submicrometer-sized droplets carrying net excess charges (58) (SI Text). A hard upper bound to τ can be derived from the fact that microdroplets carried by the nebulizer gas issuing from the injector nozzle at typical v ∼2 × 10⁴ cm⋅s⁻¹ velocities would reach the inlet to the detection section of the mass spectrometer ∼2 cm away from the tip of the nozzle in <100 μs. The net charges produced during the aerodynamic breakup of the liquid jet represent the raw information acquired by the mass spectrometer (58). We confirmed experimentally that in our instrument charge separation among the anions and cations present in interfacial layers is largely driven pneumatically (rather than electrostatically/electrochemically) (68) by showing that signal intensities and anion fractionation increase at higher nebulizer gas velocities v and extrapolate to zero as v → 0 (58). In Fig. 1A, the ES mass spectral multiplets at m/z = 161, 163, and 165 correspond to Fe^IICl₃⁻. The characteristic multiplet patterns arising from natural abundance ³⁵Cl (75%) and ³⁷Cl (25%) chlorine isotopes let us establish the number of Cl⁻ contained in each detected species and, hence, the molecular composition of these singly charged ions (Fig. S3). In the presence of O₃(g)/O₂(g) mixtures, new ES mass spectral clusters appear at m/z = 177, 179, and 181 and m/z = 196, 198, and 200, which can be readily assigned to O=Fe^IVCl₃⁻ and Fe^IIICl₄⁻, respectively. We verified that Cl⁻ is inert toward O₃(g) and does not participate in the oxidation process, in accord with the small value of k(Cl⁻ + O₃) = 0.1 M⁻¹⋅s⁻¹ in bulk water (69). Thus, ∼50% Fe^II is oxidized by 6 parts per million by volume (ppmv) (6 × 10⁻⁶ atm) O₃(g) into Fe^III and Fe^IV at the air–water interface within τ ∼10–50 μs. This is a remarkable result because from (i) the (maximum) equilibrium concentration of dissolved O₃ in the experiments of Fig. 1—[O₃(aq)] = 6 × 10⁻⁸ M [from Henry’s law constant for O₃(g) in bulk water at ambient temperature H = 0.01 M atm⁻¹] (70)—and (ii) the rate coefficients of reactions R3a—k_3a = (1.7 ± 0.4) × 10⁵ M⁻¹⋅s⁻¹—and R4—k₄ = (8.2 ± 0.3) × 10⁵ M⁻¹⋅s⁻¹—in bulk water (7), we estimate that less than 0.1% Fe²⁺ should have been consumed under present conditions. In other words, reactions R3a and R4 proceed ∼10³–10⁴ times faster at the gas–water interface than in bulk water. The modest concentration enhancements of many gases at the air–water interface predicted by theoretical simulation (70) and demonstrated experimentally (71–74) would not substantially alter the above statement. We tentatively ascribe the significant acceleration of reaction R4 at the gas–water interface to the enhanced lability and/or distorted geometry (75) of the hydration shell of Fe²⁺ at the interface relative to bulk water, a condition that would facilitate the substitution of O₃ for hydration waters and, hence, the direct interaction with the metal center required by O-atom transfer during subnanosecond gas–liquid encounters (76) (see below).

Fig. 1.

Negative ion ES mass spectra of 10 μM (A) and 100 μM (B) FeCl₂ aqueous microjets exposed to O₂(g) or O₃(g)/O₂(g) for ∼10–50 μs. X and Y correspond to O=Fe^IV·Cl·Fe^IICl₄⁻ and O=Fe^IV·Cl·Fe^IIICl₅⁻, respectively. See text for details.

Fig. 1B shows additional peaks at higher masses. The ES mass signals at m/z = 287, 289, and 291 correspond to Fe^II₂ Cl₅⁻. The group at m/z = 303, 305, and 307, hereafter labeled X, can be assigned to O=Fe^IV·Cl·Fe^IICl₄⁻ on the basis of peak masses and the characteristic Cl₅-multiplet pattern. Similarly, we assign the group at m/z = 339, 341, 343, 345, and 347, hereafter labeled Y, to O=Fe^IV·Cl·Fe^IIICl₅⁻. Our results are qualitatively consistent with previous reports based on the UV-absorption detection of O=Fe^IV species during the bulk ozonolysis of acidic Fe²⁺ (8, 23, 49, 52).

Fig. 2 A–D shows the evolution of reactant and products as functions of [O₃(g)] at FeCl₂ concentrations spanning the 1–1,000-μM range. It is apparent that although all signal intensities increase with [FeCl₂], the ratio α = O=Fe^IVCl₃⁻/Fe^IIICl₄⁻ is not constant, as expected for the products of concurrent reactions R3a and R4, but both depend on O₃(g) and FeCl₂ concentrations (8, 51). It should be pointed out that mass signal intensities are not linear functions of bulk concentrations throughout, because the interfacial concentrations detected herein will plateau as the interface becomes saturated. Also, reactant signals may bottom out rather than vanish at sufficiently large O₃(g) [or H₂O₂(g)] concentrations because interfacial layers are continuously replenished by diffusion from the bulk liquid (66). Because the O=Fe^IVCl₃⁻ intermediate reacts further with Fe²⁺ via reaction R5,

the observed increase of α at higher O₃(g) and FeCl₂ concentrations is not the result of secondary chemistry (see below). Note that (i) the higher-mass products X, Y, and Fe^III₂Cl₇⁻ appear at [FeCl₂] ≥10 μM, and (ii) α depends weakly on pH (Fig. 3), in contrast to previous reports that O=Fe^IV is formed only under very acidic (pH ≤2) bulk conditions (6, 8, 77). The more extensive hydrolysis of Fe²⁺ in more basic solutions prevented us from performing experiments above pH ∼6.5. It should be realized, however, that the acid–base properties of the air–water interface are quite different from those of bulk water. Whereas bulk water is neutral at pH 7, the aerial surface is neutral on bulk water at pH ∼3.5 (63, 65, 78). This caveat prevents direct comparisons from being made between the pH-dependences observed herein and those previously reported for similar experiments in bulk solution. We wish to emphasize that the concentration dependences observed in our experiments strongly support our assumption that the detected species are produced on the surface of the intact jet (whose composition is identical to that of the injected solution) rather than on the ensemble of daughter droplets (whose compositions will span the broad distributions generated by random solvent evaporation) (63, 65) (see also SI Text).

Fig. 2.

ES mass spectral signal intensities of reactant and products at the surface of 1 μM (A), 10 μM (B), 100 μM (C), and 1,000 μM (D) FeCl₂ aqueous microjets as functions of the O₃(g) mixing ratio (1 ppmv = 2.5 × 10¹³ molecules per centimeter⁻³ at 1 atm, 293 K). All experiments in 1 atm N₂(g) at 293 K. Background signal was subtracted for Fe(III)Cl₄⁻. X and Y correspond to O=Fe^IV·Cl·Fe^IICl₄⁻ and O=Fe^IV·Cl·Fe^IIICl₅⁻, respectively.

Fig. 3.

ES mass spectral signal intensities of reactant and products from 100 μM FeCl₂ aqueous microjets at pH 2.3 (A), pH 2.9 (B), pH 4.4 (C), and pH 5.6 (D) as functions of the O₃(g) mixing ratio (1 ppmv = 2.5 × 10¹³ molecules per centimeter⁻³ at 1 atm, 293 K). All experiments in 1 atm N₂(g) at 293 K. Background signal was subtracted for Fe(III)Cl₄⁻. X and Y correspond to O=Fe^IV·Cl·Fe^IICl₄⁻ and O=Fe^IV·Cl·Fe^IIICl₅⁻, respectively.

We tested the effects of tert-butanol (t-BuOH) and DMSO additions to FeCl₂ microjets exposed to O₃(g). t-BuOH is an efficient scavenger of ·OH radicals (k_·OH+t-BuOH = 5 × 10⁸ M⁻¹⋅s⁻¹ in bulk water) (79), whereas DMSO functions as both an ·OH scavenger and an O-atom acceptor (6, 49, 52). Fig. 4 shows negative ion mass spectra of aqueous 100 μM FeCl₂ microjets containing large excesses (100 × [FeCl₂]) of t-BuOH or DMSO upon exposure to O₃(g). Notably, the addition of t-BuOH has no effect whatsoever on mass spectra, thereby implying that ·OH radicals do not participate in the formation of the observed products. Because ·OH radicals, if present, also would react rapidly with Fe²⁺ [k_·OH+Fe(II) = 3.2 × 10⁸ M⁻¹⋅s⁻¹] (80) to produce more Fe³⁺, we infer that the decomposition of the their O₃^·- precursor, reaction R3b, is too slow under present conditions.

Fig. 4.

Negative ion ES mass spectra of aqueous microjet containing 100 μM FeCl₂ in the absence (A) and presence (B) of 10 mM t-BuOH and in the presence of 10 mM DMSO (C), exposed to O₃(g) for ∼10–50 μs. X and Y correspond to O=Fe^IV·Cl·Fe^IICl₄⁻ and O=Fe^IV·Cl·Fe^IIICl₅⁻, respectively.

The addition of DMSO as a cosolute, in contrast, has marked effects on product distribution. DMSO quenches most (but not all) ozonation products, such as O=Fe^IV·Cl·Fe^IICl₄⁻ and O=Fe^IV·Cl·Fe^IIICl₅⁻, and all higher-mass poly-iron species. Remarkably, the mono-iron ferryl O=Fe^IVCl₃⁻ (and, as expected, Fe^IIICl₄⁻) is not affected.

Similar experiments in which aqueous microjets containing 10 μM and 1 mM FeCl₂, in the absence and presence of t-BuOH as a cosolute, were exposed to H₂O₂(g)/N₂(g) mixtures led to the products shown in Fig. 5. A comparison of Figs. 1 and 5 confirms that H₂O₂ and O₃ react with interfacial Fe²⁺ along reactions R1–R2 and R3a,b–R4, respectively, leading to (formally) the same products, albeit in different proportions. Note again that from Henry’s law constant for H₂O₂(g) in bulk water—H = 10⁵ M⋅atm⁻¹ (∼10⁷ times larger than for O₃) (81)—the maximum equilibrium concentration of dissolved H₂O₂ in the experiments of Fig. 5A is [H₂O₂(aq)] = 0.03 M. Therefore, from k(Fe^II + H₂O₂) ∼ 50 M⁻¹⋅s⁻¹, we estimate that Fe²⁺ half-lives would be on the order of seconds if reactions R1–R2 took place in bulk water, rather than a few tens of microseconds, as observed in our experiments (81). Note that in contrast to the O₃ case, O=Fe^IVCl₃⁻ is undetectable (compare Fig. 5 vis-a-vis Fig. 1), an event we ascribe to the slowness of R2 relative to R4 under present conditions, because O=Fe^IVCl₃⁻ would be consumed at the same rates via R5 in both systems. We confirmed that the addition of DMSO as a cosolute in these experiments (Fig. S4) has an effect similar to those displayed in Fig. 4C.

Fig. 5.

Negative ion ES mass spectra of 10 μM (A and B) and 1 mM (C and D) FeCl₂ aqueous microjets in the absence/presence of excess t-BuOH as a cosolute exposed to H₂O₂(g) for ∼10–50 μs. X and Y correspond to O=Fe^IV·Cl·Fe^IICl₄⁻ and O=Fe^IV·Cl·Fe^IIICl₅⁻, respectively.

The prompt formation of products in our experiments, at rates several orders of magnitude larger than those estimated for the same reactants dissolved in bulk water, and the peculiar variation in the ratio of the products of parallel reactions R3a and R4 (Fe^IIICl₄⁻ and O=Fe^IVCl₃⁻) as a function of ozone and Fe²⁺ concentrations reveal the exceptional characteristics of Fenton-like chemistry at the air–water interface. These phenomena, which could be typical of water–hydrophobe interfaces in general, reveal (i) the enhanced reactivity of interfacial Fe²⁺ as an O-atom acceptor from O₃ and H₂O₂ and (ii) the progressive emergence of such species to the outermost interfacial layers at higher Fe²⁺ bulk concentrations. Extensive ab initio molecular dynamics (Car–Parrinello) calculations that incorporate the water solvent explicitly would be required to elucidate the molecular details of these unique events at the gas–water interface (26, 48, 82). The possibility that doubly charged Fe²⁺ cations might be present in shallow interfacial layers is supported by molecular dynamics calculations, which predict that the concentration of doubly charged Mg²⁺ peaks at ∼4 Å below the Gibbs dividing surface at values twice as large as its concentration in bulk water (83). Recent X-ray reflectivity studies confirmed the existence of nonmonotonic cation density profiles within ∼1-nm interfacial layers of aqueous electrolyte solutions (84). We cannot rule out the possibility that impinging gases are hydrated before colliding with the liquid surface (85, 86), but we deem it inconsequential because O₃–(H₂O)_n and H₂O₂–(H₂O)_n complexes represent a very small fraction of O₃ and H₂O₂ gas flows under present conditions.

By considering that O-atom transfer, in contrast to electron transfer (reactions R1 and R3a), requires direct contact of the O-atom donors (H₂O₂ and O₃) with the metal ion, and that the hydration waters of Fe(H₂O)₆²⁺ in bulk water are exchanged approximately every 0.5 μs (87), whereas O₃(g) remains trapped on the water surface for only 0.1 ns (76), our results imply that either (i) dissociative ligand substitution in interfacial Fe(H₂O)₆²⁺ (IF) is much faster than in Fe(H₂O)₆²⁺ in bulk water (B) or (ii) interfacial Fe(H₂O)₆²⁺ has a distorted octahedral geometry, on account of the broken symmetry, that lets O₃ approach the Fe²⁺ center via low-energy associative interchange pathways (88). Against this backdrop, our findings reveal that the dynamics and thermodynamics of ion hydration at aqueous interfaces are quite different from those in bulk water (75, 89). Thus, the roles and behavior of ions in many physical, chemical, and biological interfacial processes may not be predicted (or analyzed) from the properties of the corresponding ions in bulk water.

The dissimilar evolutions of O=Fe^IVCl₃⁻ and Fe^IIICl₄⁻ in Figs. 2 and 3 are consistent with IF(z) depth profiles at the interface that depend on total [Fe²⁺]. Everything happens as though the IF species involved in O-atom transfer are preferentially pushed to the surface of more concentrated Fe²⁺ solutions. In this context, it is relevant to point out that we recently found that hydronium (H₃O⁺) emerges at the surface of water less than pH 4 as a “superacid” that protonates impinging gases having proton affinities larger than water (62, 64, 90). Thermodynamics dictates that this is possible only if interfacial H₃O⁺ is weakly hydrated. If Fe²⁺ behaves similarly, the enhanced reactivity of IF relative to B and its emergence at the surface of more concentrated solutions could be alternatively ascribed to an incomplete hydration shell of IF.

We associate the significantly enhanced reactivity of poly-iron ferryls relative to O=Fe^IVCl₃⁻ as O-atom donors to DMSO with the weakening of the O=Fe^IV bond, and ascribe such weakening to electronic rather than inductive effects. Our view is based on the fact that strong electron-donating ligands, such as thiolate (24), in the axial position are known to weaken and elongate the O=Fe^IV bond in low-spin (S = 1) complexes by increasing the population of its σ- and π-antibonding molecular orbitals (45). Because a Cl⁻ ligand coordinatively bound to Fe^II (as in X) or Fe^III (as in Y) should be less nucleophilic than unbound Cl⁻, the enhanced reactivity of poly-iron ferryls appears to be a result of the electronic coupling of the iron centers via Cl⁻ bridges. Further experimental work and high-level quantum chemistry calculations are needed to fully elucidate the molecular basis of the O-donating power of poly-iron ferryls.

In summary, we present compelling evidence of the prompt formation of mono- and poly-iron Fe^IV species on the surface of aqueous Fe^IICl₂ microjets exposed to gaseous H₂O₂ or O₃ beams. The exceedingly fast reactions of interfacial Fe²⁺ with gas-phase H₂O₂ and O₃ [10³–10⁴ times faster than similar reactions of Fe(H₂O)₆²⁺ in bulk aqueous media] are ascribed to a labile/incomplete hydration shell that favors inner-sphere O-atom transfers over outer-sphere one-electron transfers. The finding that di-iron ferryls O=Fe^IV·Cl·Fe^IICl₄⁻ and O=Fe^IV·Cl·Fe^IIICl₅⁻ are fully scavenged by the O-atom acceptor DMSO, whereas O=Fe^IVCl₃⁻ is not, implicates the electronic coupling of mixed-valence iron centers in weakening the Fe^IV=O bond. Present results suggest a more significant role than hitherto envisioned for the Fe^IV species produced in Fenton and Fenton-like chemistries at aqueous interfaces opposite hydrophobic media, such as air in atmospheric aerosols and clouds, proteins in living tissues, biomembranes, and bio-microenvironments.

Methods

Our experiments involve the injection of aqueous FeCl₂ jets into the spraying chamber of an ES mass spectrometer (Agilent 6130 Quadrupole LC/MS Electrospray System) flushed with N₂(g) at 1 atm, 293 K. Jets are exposed therein to orthogonal gas-phase O₃ or H₂O₂ beams. The species produced on the surface of such jets are analyzed in situ via online ES-MS. The present experimental setup essentially is the same as the one reported elsewhere (62, 64, 65). Solutions are pumped (100 μL⋅min⁻¹) into the spraying chamber through a grounded stainless steel needle (100-μm bore) coaxial with a sheath-issuing nebulizer N₂(g) at a high gas velocity v_g (∼160 m/s). The species detected by ES-MS are assumed to be produced in collisions of gaseous H₂O₂ or O₃ with the surface of the intact aqueous jets containing microdroplets (D₀ > 1 μm) as they emerge from the nozzle, i.e., before they are broken up into submicrometer-sized droplets (58) (see also SI Text). These smaller droplets already carry net charge of either sign. It should be emphasized that charge separation is a one-time event driven by the conversion of kinetic energy of the nebulizer gas into surface and electrostatic energies of submicrometer-sized droplets (58). We have demonstrated the surface specificity of our experiments by showing that (i) anion signal intensities in the mass spectra of equimolar salt solutions adhere to a normal Hofmeister series (rather than being identical) (91, 92), (ii) the depth of the interfacial layers sampled is controllable as a function of nebulizer gas velocity v (58), and (iii) they allow the detection of products of gas–liquid reactions that could be formed only at the air–water interface (62, 64–66, 71).

Gaseous hydrogen peroxide, H₂O₂(g), was injected into the spraying chamber carried by ultrapure (>99.999%) N₂(g) sparging hydrogen peroxide solution [extra pure reagent, 30% (wt/wt) in water; Nacalai Tesque] kept in a trap held at 293 K in a temperature-controlled bath (TRC-4C; Thomas). Carrier gas flow rates were regulated by calibrated digital mass flow controllers (SEC-400 Mark 3; Horiba STEC) up to 1 standard liter per minute (Fig. S1). [H₂O₂(g)] was derived from the reported H₂O₂ vapor pressures of H₂O₂:H₂O mixture at the temperatures (93). We verified that Cl⁻ is inert toward H₂O₂(g) under present conditions (Fig. S5). Teflon gas lines were cleaned and dried daily with ultrapure nitrogen gas. Ozone was generated by flowing ultrapure O₂(g) (>99.998%; Kyoto Teisan) through a silent discharge ozonizer (KSQ-050; Kotohira) and quantified via online UV-visible absorption spectrophotometry (Agilent 8453; Agilent Technologies) at 250 and 300 nm [absorption cross sections σ(250 nm) = 1.1 × 10⁻¹⁷, σ(300 nm) = 3.9 × 10⁻¹⁹ cm²⋅molecule⁻¹ at 298 K] before entering the reaction chamber (Fig. S2). Throughout, the reported [O₃(g)] values, which correspond to the concentrations actually sensed by the microjets in the reaction chamber, are estimated to be ∼13 times smaller than the values determined from UV absorbance because of further dilution by the drying gas. The gas molecule hitting the surface of the pH-adjusted (by concentrated HCl/NaOH, and the pH was already measured by a calibrated pH meter, Horiba LAQUA F-74, before the experiments) aqueous microjet can stick to it by accommodation, reacting therein, or rebound (94, 95). See Supporting Information for more details.