Technical note: Gas-phase nitrate radical generation via irradiation of aerated ceric ammonium nitrate mixtures

We present a novel photolytic source of gas-phase NO₃ suitable for use in atmospheric chemistry studies that has several advantages over traditional sources that utilize NO₂ + O₃ reactions and/or thermal dissociation of dinitrogen pentoxide (N₂O₅). The method generates NO₃ via irradiation of aerated aqueous solutions of ceric ammonium nitrate (CAN, (NH₄)₂Ce(NO₃)₆) and nitric acid (HNO₃) or sodium nitrate (NaNO₃). We present experimental and model characterization of the NO₃ formation potential of irradiated CAN $/$ HNO₃ and CAN $/$ NaNO₃ mixtures containing [CAN] = 10⁻³ to 1.0 M, [HNO₃] = 1.0 to 6.0 M, [NaNO₃] = 1.0 to 4.8 M, photon fluxes (I) ranging from 6.9 × 10¹⁴ to 1.0 × 10¹⁶ photons cm⁻² s⁻¹, and irradiation wavelengths ranging from 254 to 421 nm. NO₃ mixing ratios ranging from parts per billion to parts per million by volume were achieved using this method. At the CAN solubility limit, maximum [NO₃] was achieved using [HNO₃] ≈ 3.0 to 6.0 M and UVA radiation (λ_max = 369 nm) in CAN $/$ HNO₃ mixtures or [NaNO₃] ≥ 1.0 M and UVC radiation (λ_max = 254 nm) in CAN $/$ NaNO₃ mixtures. Other reactive nitrogen (NO₂, N₂O₄, N₂O₅, N₂O₆, HNO₂, HNO₃, HNO₄) and reactive oxygen (HO₂, H₂O₂) species obtained from the irradiation of ceric nitrate mixtures were measured using a NO_x analyzer and an iodide-adduct high-resolution time-of-flight chemical ionization mass spectrometer (HR-ToF-CIMS). To assess the applicability of the method for studies of NO₃-initiated oxidative aging processes, we generated and measured the chemical composition of oxygenated volatile organic compounds (OVOCs) and secondary organic aerosol (SOA) from the β-pinene + NO₃ reaction using a Filter Inlet for Gases and AEROsols (FIGAERO) coupled to the HR-ToF-CIMS.

1 Introduction

The importance of NO₃ as a nighttime atmospheric oxidant is well established (Wayne et al., 1991; Brown and Stutz, 2012; Ng et al., 2017; Wang et al., 2023). NO₃ is generated via the reaction NO₂ + O₃ → NO₃ + O₂, followed by achievement of temperature-dependent equilibrium between NO₃, NO₂, and dinitrogen pentoxide (N₂O₅). N₂O₅ also hydrolyzes efficiently to HNO₃ on aqueous surfaces (Brown et al., 2004). Thus, any investigation of the influence of NO₃ chemistry in a specific source region necessarily must account for the local temperature, humidity, and particle surface area along with other factors. Despite these complications, for decades, laboratory studies investigating gas-phase NO₃ chemistry have utilized the same NO₂ + O₃ reactions and/or N₂O₅ thermal decomposition to produce NO₃ as it occurs in the atmosphere and accommodated the inherent limitations associated with N₂O₅, namely, that it must be stored under cold and dry conditions until use. Few viable alternative methods for the generation of gas-phase NO₃ have been identified. Reactions between fluorine atoms and nitric acid (F + HNO₃ → HF + NO₃) or chlorine atoms and chlorine nitrate (Cl + ClNO₃ → Cl₂ + NO₃) require handling and/or synthesizing hazardous halogen-containing compounds (Burrows et al., 1985; Bedjanian, 2019). F and Cl can also compete with NO₃ for the oxidation of target analytes, as can O₃ if its reaction with NO₂ is used as the NO₃ source.

In the 1960s and 1970s, following earlier research into the properties of ceric solutions (Meyer and Jacoby, 1901; Wylie, 1951; Hinsvark and Stone, 1956; Blaustein and Gryder, 1957), Thomas Martin and coworkers discovered that irradiating solutions containing ceric ammonium nitrate (CAN, (NH₄)₂Ce(NO₃)₆) generates aqueous NO₃ (Henshall, 1963; Martin et al., 1963, 1964; Glass and Martin, 1970; Martin and Glass, 1970; Martin and Stevens, 1978). In ≳ 6 M nitric acid (HNO₃), CAN is thought to dissociate primarily into ${\mathrm{NH}}_{\mathrm{4}}^{+}$ cations and hexanitratocerate ($\mathrm{Ce}({\mathrm{NO}}_{\mathrm{3}}{)}_{\mathrm{6}}^{\mathrm{2}-}$) anions (Henshall, 1963). The $\mathrm{Ce}({\mathrm{NO}}_{\mathrm{3}}{)}_{\mathrm{6}}^{\mathrm{2}-}$ is subsequently reduced to $\mathrm{Ce}({\mathrm{NO}}_{\mathrm{3}}{)}_{\mathrm{5}}^{\mathrm{2}-}$ upon irradiation by ultraviolet light, and NO₃ is generated as a primary photolysis product. A similar process occurs in other solvents, although the ensuing ceric composition in solution is complex and influenced by several factors. For example, in glacial acetic acid (CH₃COOH), CAN dissociates into primarily Ce(NO₃)₄ (Henshall, 1963). Additionally, ceric ions containing complexed hydroxyl (OH) or H₂O, CH₃COOH, or acetonitrile (CH₃CN) molecules are formed in aqueous, acetic acid, or CH₃CN media, respectively (Henshall, 1963; Glebov et al., 2021). Higher solution acidity and/or CAN concentration appears to promote the formation of $\mathrm{Ce}({\mathrm{NO}}_{\mathrm{3}}{)}_{\mathrm{6}}^{\mathrm{2}-}$(Wylie, 1951) and ceric nitrate dimers (Blaustein and Gryder, 1957; Demars et al., 2015). The following generalized mechanism was proposed by Glass and Martin (1970) to describe ceric nitrate photochemistry:

$$\begin{array}{}\text{(R1)}& {\displaystyle }{\mathrm{Ce}}^{\left(\mathrm{IV}\right)}+h\mathit{\nu }\to {\mathrm{Ce}}^{\left(\mathrm{III}\right)}+{\mathrm{NO}}_{\mathrm{3}},\text{(R2)}& {\displaystyle }{\mathrm{Ce}}^{\left(\mathrm{III}\right)}+{\mathrm{NO}}_{\mathrm{3}}\to {\mathrm{Ce}}^{\left(\mathrm{IV}\right)},\text{(R3)}& {\displaystyle }{\mathrm{NO}}_{\mathrm{3}}+{\mathrm{NO}}_{\mathrm{3}}\to {\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{6}},\text{(R4)}& {\displaystyle }{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{6}}+\mathrm{2}{\mathrm{Ce}}^{\left(\mathrm{IV}\right)}\to \mathrm{2}{\mathrm{NO}}_{\mathrm{2}}+{\mathrm{O}}_{\mathrm{2}}+\mathrm{2}{\mathrm{Ce}}^{\left(\mathrm{III}\right)},\text{(R5)}& {\displaystyle }{\mathrm{NO}}_{\mathrm{3}}+{\mathrm{NO}}_{\mathrm{2}}+{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\to \mathrm{2}{\mathrm{HNO}}_{\mathrm{3}},\end{array}$$

where Ce^(IV) represents ceric nitrates as diverse as Ce(NO₃)₄, $\mathrm{Ce}({\mathrm{NO}}_{\mathrm{3}}{)}_{\mathrm{6}}^{\mathrm{2}-}$, $\left({\mathrm{NO}}_{\mathrm{3}}{)}_{\mathrm{5}}\mathrm{CeOCe}\right({\mathrm{NO}}_{\mathrm{3}}{)}_{\mathrm{5}}^{\mathrm{4}-}$, and (H₂O)₃(NO₃)₃CeOCe(NO₃)₃(H₂O)₃ that are potentially formed in solution (Blaustein and Gryder, 1957; Henshall, 1963; Demars et al., 2015). Similarly, Ce^(III) represents cerous nitrates such as Ce(NO₃)₃ and $\mathrm{Ce}({\mathrm{NO}}_{\mathrm{3}}{)}_{\mathrm{5}}^{\mathrm{2}-}$. The rate of Reaction (R2) is [HNO₃]-dependent (Martin and Glass, 1970), and the dinitrogen hexaoxide (N₂O₆) intermediate was proposed on the basis of supporting observations without direct measurements (Glass and Martin, 1970).

CAN is used routinely as an oxidizing agent in organic synthesis due to its widespread availability, low cost, high oxidative potential, and low toxicity (Nair and Deepthi, 2007). However, its usage in atmospheric chemistry to date is limited to studies of NO₃-initiated oxidative aging processes in solution; e.g., Alexander (2004). Given the potential simplicity of irradiating Ce^(IV) mixtures relative to synthesizing and storing N₂O₅ under cold and dry conditions or reacting NO₂ + O₃ under carefully controlled conditions, Ce^(IV) irradiation could in principle enable more widespread studies of NO₃ oxidation chemistry, which is understudied compared to OH chemistry (Ng et al., 2017). Here, for the first time, we investigated the use of Ce^(IV) irradiation as a source of gas-phase NO₃. First, we designed a photoreactor that generates gas-phase NO₃ from irradiated CAN $/$ HNO₃ and CAN $/$ NaNO₃ mixtures. Second, we characterized NO₃ concentrations achieved over a range of reactor operating conditions and mixture composition. Third, we characterized gas-phase reactive nitrogen and reactive oxygen species generated following Ce^(IV) irradiation. Fourth, we demonstrated application of the method to generate and characterize oxygenated volatile organic compounds (OVOCs) and secondary organic aerosol (SOA) from the β-pinene + NO₃ reaction.

2 Methods

2.1 Photoreactor design and operation

Figure 1 shows a schematic of the experimental setup used in this study. A zero-air carrier gas flow of 0.5 L min⁻¹ was bubbled through a gas dispersion line consisting of 6.35 mm o.d. × 4.8 mm i.d. fluorinated ethylene propylene (FEP) tubing into approximately 10 mL of aqueous CAN $/$ HNO₃ or CAN $/$ NaNO₃ mixtures placed at the bottom of a 12.7 mm o.d. × 11.1 cm i.d. FEP tube. The FEP tube was surrounded by low-pressure mercury fluorescent lamps installed vertically in a custom enclosure. These lamps had a 35.6 cm illuminated length. At these operating conditions, the calculated gas transit time in the illuminated portion of the reactor was approximately 3 s. After exiting the photoreactor, the carrier gas flow was passed through a filter holder (Savillex, 401-21-47-10-21-2) containing a 47 mm PTFE membrane filter (Pall Gelman, R2PJ047) to transmit NO₃ (Wagner et al., 2011) while removing stray droplets from the sample flow. At the end of each experiment, the lamps were turned off, the gas dispersion line was removed from the top of the reactor, and the FEP tubing and filter holder were flushed with distilled H₂O to remove residual Ce^(III) precipitate. Initial studies were conducted using a cavity-attenuated phase-shift (CAPS) NO₂ monitor operating at λ = 405 nm (Kebabian et al., 2005) and a second retrofitted CAPS monitor operating at λ = 630 nm, which established that NO₂ and NO₃ were produced from irradiated Ce^(IV). Subsequent studies described in the next section used a 2B Technologies model 405 analyzer to measure NO and NO₂ (Birks et al., 2018).

Figure 1 Overview of experiments conducted in this study. Aqueous mixtures of ceric ammonium nitrate (CAN) and nitric acid (HNO₃) or sodium nitrate (NaNO₃) were irradiated in a photoreactor to generate nitrate radicals (NO₃) in solution. Air was bubbled through the solution to evaporate NO₃ and other volatile photolysis products into the gas phase. The photoreactor effluent was then (i) injected into a dark oxidation flow reactor (OFR) along with a VOC mixture to characterize [NO₃] via tracer decay measurements using a Vocus proton transfer reaction time-of-flight mass spectrometer (PTR-ToF) (ii) sampled with an iodide-adduct high-resolution time-of-flight chemical ionization mass spectrometer (HR-ToF-CIMS) (iii) injected into a dark OFR to characterize β-pinene $/$ NO₃ oxidation products with a Filter Inlet for Gases and AEROsols (FIGAERO) coupled to the HR-ToF-CIMS. Supporting measurements were obtained using a NO_x analyzer.

Depending on the specific experiment, lamps with peak emission output centered at λ = 254, 313, 369, or 421 nm, respectively (GPH436TL/4P, Light Sources, Inc.; F436T5/NBUVB/4P-313, F436T5/BLC/4P-369, F436T5/SDI/4P-421, LCD Lighting, Inc.) were used. Emission spectra from the manufacturer are shown in Fig. S1 in the Supplement. A fluorescent dimming ballast (IZT-2S28-D, Advance Transformer Co.) was used to regulate current applied to the lamps. To quantify the photon flux I_λ in the photoreactor for studies that used λ = 254, 313, or 369 nm radiation, we measured the rate of externally added O₃ (λ = 254 nm) or NO₂ photolysis (λ = 313 or 369 nm) as a function of lamp voltage under dry conditions (RH < 5 %). The photon flux was not quantified in studies that used λ = 421 nm radiation. NO₂ photolysis measurements were conducted in the absence of oxygen to avoid O₃ formation. Photon flux values were then calculated using methods described in Lambe et al. (2019); maximum values of I₂₅₄ = 1.0 × 10¹⁶ photons cm⁻² s⁻¹, I₃₁₃ = 6.0 × 10¹⁵ photons cm⁻² s⁻¹, and I₃₆₉ = 7.0 × 10¹⁵ photons cm⁻² s⁻¹ were obtained.

2.2 Characterization studies

In one set of experiments, the 0.5 L min⁻¹ photoreactor effluent was mixed with a 6.5 L min⁻¹ zero-air carrier gas and injected into a dark Potential Aerosol Mass oxidation flow reactor (OFR; Aerodyne Research, Inc.), which is a horizontal 13 L conductive Teflon-coated aluminum cylindrical chamber operated in continuous flow mode. Approximately 6.5 L min⁻¹ of sample flow was pulled from the reactor, resulting in a calculated mean residence time in the OFR (τ_OFR) of approximately 120 s. To constrain NO₃ mixing ratios, a mixture of 10 VOC tracers with NO₃ reaction rate coefficients ($k_{{\mathrm{NO}}_{\mathrm{3}}}$) ranging from 3.01 × 10⁻¹⁹ to 2.69 × 10⁻¹¹ cm³ molec.⁻¹ s⁻¹ at T = 298 K (Table S1) was injected through a 10.2 cm length of 0.0152 cm i.d. Teflon tubing at a liquid flow rate of 0.94 µL h⁻¹ using a syringe pump. The tracer mixture was then evaporated into a 1 L min⁻¹ zero-air carrier gas prior to injection into the OFR. The total external NO₃ reactivity (NO₃R_ext), which is the summed product of each tracer mixing ratio and its $k_{{\mathrm{NO}}_{\mathrm{3}}}$, was approximately 5 s⁻¹. VOCs with proton affinities greater than that of H₂O were chosen to enable their measurement with a Tofwerk–Aerodyne Vocus proton transfer reaction time-of-flight mass spectrometer (hereafter referred to as “Vocus PTR”) operated using H₃O⁺ reagent ion chemistry (Krechmer et al., 2018) and ∼ 8000 (Th $/$ Th) resolving power. NO₃ mixing ratios were calculated from the measured decrease in VOC mixing ratios using the Vocus PTR. Here, we assumed that the total concentration of reacted VOCs was equal to the concentration of NO₃ injected into the OFR. Because NO₃ may additionally react with organic peroxy radicals (RO₂) generated from VOC + NO₃ reactions as well as OVOCs, these calculated NO₃ concentrations represent lower limits. Modeling calculations suggest that the fractional consumption of NO₃ by RO₂ ranged from < 0.01 to 0.17 over the range of conditions that were studied (Fig. S2). A subset of OVOCs generated from VOC + NO₃ reactions that had proton affinities greater than that of H₂O were also detected with the Vocus PTR.

In a separate set of experiments, the photoreactor effluent was diluted into 4 L min⁻¹ zero-air carrier gas and sampled with an Aerodyne iodide-adduct high-resolution time-of-flight chemical ionization mass spectrometer (HR-ToF-CIMS; hereafter referred to as “CIMS”; Bertram et al., 2011) and the NO_x analyzer. The CIMS was operated at a ∼ 4000 (Th $/$ Th) resolving power. Iodide-adduct reagent ion chemistry was used due to its high sensitivity and selectivity towards nitrogen oxides and multifunctional organic nitrates (Lee et al., 2014). To demonstrate application of the method to study NO₃-initiated oxidative aging processes, the chemical composition of β-pinene + NO₃ gas- and condensed-phase oxidation products was measured with a Filter Inlet for Gases and AEROsols (FIGAERO) coupled to the CIMS (Lopez-Hilfiker et al., 2014). Gas sampling and simultaneous particle collection were performed for 1 min intervals, followed by thermal desorption of the particle sample from a PTFE filter membrane (15 min ramp from room temperature to 200 ^∘C, 10 min holding time, 8 min cooldown to room temperature).

2.3 Photochemical model

To supplement our measurements, and to characterize aqueous-phase concentrations of species produced in the photoreactor that were not measured, we developed a photochemical box model that was implemented in the KinSim chemical kinetic solver (Peng and Jimenez, 2019). The KinSim mechanism shown in Table S2 contains reactions to model concentrations of Ce^(IV), Ce^(III), NO, NO₂, NO₃, N₂O₃, N₂O₄, N₂O₅, HNO₂, HNO₃, HNO₄, H, O, OH, HO₂, and H₂O₂. We assumed that HNO₃ that was present in solution prior to irradiation completely dissociated into H⁺ and ${\mathrm{NO}}_{\mathrm{3}}^{-}$. When possible, we used condensed-phase rate coefficients in the mechanism. For reactions that we assumed occurred but did not have published condensed-phase rate coefficients (e.g., NO₃ + OH → NO₂ + HO₂), we used published gas-phase rate coefficients instead with no modifications aside from unit conversion. Gas-phase wall loss rates of NO_x, NO_y, and HO_x species were not explicitly considered in the mechanism. UV–Vis extinction cross sections (σ_ext) of CAN $/$ HNO₃ and CAN $/$ NaNO₃ mixtures were separately obtained between λ = 200 and 600 nm using an Agilent Cary 5000 UV–Vis–NIR spectrophotometer. Because of the high absorptivity and concentrations of the mixtures, samples were prepared in a 0.01 mm short-path-length cuvette (20/C-Q-0.01, Starna) to minimize saturation of the photodetector relative to a cuvette with a standard 10 mm path length. Even with the cuvette that was used, CAN dilution was necessary in some cases in order to obtain σ_ext without photodetector saturation at shorter wavelengths. Spectra were obtained as a function of [CAN] (0.047 to 0.526 M), [HNO₃] (0 to 6.0 M), and [NaNO₃] (0 to 4.0 M) to cover the approximate range of mixture compositions that were characterized in Sect. 2.2. The σ_ext values of the mixture were then calculated using the Beer–Lambert law and applied in the KinSim mechanism. Model outputs were obtained over a total experimental time of 14400 s at 1 s intervals.

3 Results and discussion

The maximum NO₃ quantum yield (${\mathit{\varphi }}_{{\mathrm{NO}}_{\mathrm{3}}}$) of UVA-irradiated CAN $/$ HNO₃ mixtures is obtained at 6.0 M HNO₃ (Martin and Stevens, 1978); thus, this mixture composition served as the basis from which additional characterization studies were conducted. We found that 0.5 M CAN was the approximate solubility limit in 6.0 M HNO₃ at 25 ^∘C. Because 1.1 M CAN is the solubility limit in H₂O and CAN is nearly insoluble in HNO₃ (Martin and Glass, 1970), 0.7 M CAN is the estimated solubility limit in 6.0 M HNO₃ in the absence of changes in ceric nitrate composition in solution. Thus, the reduction in CAN solubility (0.7 M → 0.5 M) observed in our studies was presumably associated with significant conversion of CAN to dimeric ceric nitrates in 6.0 M HNO₃ (Blaustein and Gryder, 1957; Demars et al., 2015).

3.1 NO₃ characterization studies

Figure 2a shows time series of thiophene (C₄H₄S), 2,3-dihydrobenzofuran (C₈H₈O), cis-3-hexenyl acetate (C₈H₁₄O₂), isoprene (C₅H₈), dimethyl sulfide (C₂H₆S), 2,5-dimethylthiophene (C₆H₈S), α-pinene (C₁₀H₁₆), and guaiacol (C₇H₈O₂) concentrations following injection into the OFR and exposure to NO₃ generated in the photoreactor from irradiation of a mixture of 0.5 M CAN and 6.0 M HNO₃ at I₃₆₉ = 7 × 10¹⁵ photons cm⁻² s⁻¹. Here, concentrations of each VOC were first normalized to the acetonitrile concentration to correct for changes in the syringe pump output over time and then normalized to the VOC concentration prior to NO₃ exposure. Aside from C₆H₈S, whose relative decay was less pronounced than expected (Table S1), and butanal (C₄H₈O, not shown), whose signal decreased by approximately 30 % and did not recover for reasons that are unclear, the oxidative loss of each tracer increased with increasing $k_{{\mathrm{NO}}_{\mathrm{3}}}$. Maximum tracer consumption was observed at the beginning of the experiment due to maximum NO₃ production from Ce^(IV) irradiation. As the experiment progressed and Ce^(IV) was reduced to Ce^(III), the NO₃ concentration and corresponding VOC oxidative loss decreased. Compared to the other VOCs, the initial increase in C₁₀H₁₆ and C₇H₈O₂ concentrations over the first 2 h was delayed because of their higher $k_{{\mathrm{NO}}_{\mathrm{3}}}$ values that resulted in > 95 % consumption and lower sensitivity to changes in [NO₃] in the initial stage of the experiment. To confirm that VOC degradation shown in Fig. 2a was due to reaction with NO₃, Fig. S3 shows the relative NO₃ rate coefficients obtained from the decay of C₄H₄S, C₈H₈O, and C₈H₁₄O₂ measured with the Vocus PTR. We measured relative rate coefficients of 3.59 between C₈H₈O and C₄H₄S and 6.92 between C₈H₁₄O₂ and C₄H₄S, which are in agreement with relative rate coefficient values of 3.44 ± 1.20 and 7.68 ± 2.84 calculated from their absolute NO₃ rate coefficients (Atkinson, 1991; D'Anna et al., 2001). Time series of ions corresponding to nitrothiophene (C₄H₃NO₂S), C₅H₇NO_4–6 and C₁₀H₁₅NO_5,6 organic nitrates, and nitroguaiacol (C₇H₇NO₄), which are known NO₃ oxidation products of C₄H₄S, C₅H₈, C₁₀H₁₆, and C₇H₈O₂ (Atkinson et al., 1990; Jenkin et al., 2003; Saunders et al., 2003; Cabañas et al., 2005), along with C₈H_5,7NO_4–6 and C₈H₁₃NO_5–6 ions that may be associated with NO₃ oxidation products of C₈H₈O and C₈H₁₄O₂, respectively, were anticorrelated with those of their respective VOC precursors (Fig. S4). Tracer decay experiments, similar to the one shown in Fig. S3, were used to obtain results that are discussed in more detail in Sect. 3.2, 3.3, and 3.4.

Figure 2 Example results from an experiment in which a mixture of 0.5 M CAN and 6.0 M HNO₃ was irradiated to generate NO₃ (λ_max = 369 nm, I₃₆₉ = 7 × 10¹⁵ photons cm⁻² s⁻¹) that was injected into the OFR along with a reactive VOC tracer mixture. (a) Time series of the fractional consumption of VOC tracers measured with the Vocus following irradiation: thiophene (C₄H₄S), 2,3-dihydrobenzofuran (C₈H₈O), cis-3-hexenyl-1-acetate (C₈H₁₄O₂), isoprene (C₅H₈), dimethyl sulfide (C₂H₆S), 2,5-dimethylthiophene (C₆H₈S), α-pinene (C₁₀H₁₆), and guaiacol (C₇H₈O₂). Signals of each tracer were normalized to their initial concentrations prior to NO₃ exposure and to acetonitrile concentrations to account for changes in the syringe pump output. (b) Time series of [NO₃] calculated from (a) and Table S1.

3.2 Effect of irradiation wavelength

Figure 3a shows normalized [NO₃] values obtained following irradiation of mixtures containing CAN and 6.0 M HNO₃ or 4.8 M NaNO₃ as a function of irradiation wavelength. In CAN $/$ HNO₃ mixtures, [NO₃] was a factor of 2.4–3.5 higher following irradiation at λ = 369 compared to the other wavelengths. On the other hand, [NO₃] decreased with increasing irradiation wavelength following irradiation of CAN $/$ NaNO₃ mixtures; at λ = 254 nm, [NO₃] was 3.2–42 times higher than at the other irradiation wavelengths that were used. These differences in [NO₃] were larger than the differences in calibrated photon flux values at the maximum output of each lamp type (±40 %; Sect. 2.1). Different Ce^(IV) in CAN $/$ HNO₃ and CAN $/$ NaNO₃ mixtures may have influenced these trends, as suggested by their UV–Vis spectra (Fig. 3b). The σ_ext curves of CAN $/$ HNO₃ mixtures were generally larger, broader, and redshifted relative to those of CAN $/$ NaNO₃ mixtures, with the extent of red shifting increasing with larger [HNO₃], possibly due to higher yields of $\mathrm{Ce}({\mathrm{NO}}_{\mathrm{3}}{)}_{\mathrm{6}}^{\mathrm{2}-}$ and/or ceric nitrate dimers (Blaustein and Gryder, 1957; Henshall, 1963; Demars et al., 2015). For λ > 250 nm, CAN $/$ HNO₃ mixtures had σ_ext,max values between λ = 306–311 nm, whereas CAN $/$ NaNO₃ solutions had σ_ext,max values at λ = 296 nm. However, if [NO₃] was simply proportional to σ_ext, irradiation of CAN $/$ HNO₃ mixtures at λ = 313 nm should have produced the highest [NO₃]; this was not the case. Instead, model calculations suggest that higher [NO₂] obtained from significantly faster photolysis of HNO₃ at λ = 254 and 313 nm relative to λ > 350 nm suppressed NO₃ downstream of the photoreactor when shorter irradiation wavelengths were used (Sander et al., 2011, Table S2). At a photon flux of 10¹⁶ photons cm⁻² s⁻¹, model-calculated [NO₃] values were within ±13 % of each other for irradiation wavelengths ranging from λ = 254 to 369 nm. However, higher [NO₂] values obtained following Ce^(IV) irradiation at λ = 254 and 313 nm suppressed NO₃ by > 96 % relative to the λ = 369 nm case during 120 s of simulated NO₂ + NO₃ reactions in the OFR. Thus, although the measured NO₃ suppression at these other irradiation wavelengths was less substantial than the model output, the measurement and model trends, along with achievement of maximum [NO₃] following λ = 254 nm irradiation of CAN $/$ NaNO₃ mixtures that had lower [HNO₃], qualitatively support this explanation for the wavelength-dependent NO₃ yields observed in CAN $/$ HNO₃ mixtures.

Figure 3 (a) [NO₃] values obtained from irradiated CAN and 6.0 M HNO₃ and CAN and 4.8 M NaNO₃ mixtures as a function of irradiation wavelengths. Results were normalized to [NO₃] achieved with irradiation of CAN $/$ HNO₃ mixtures at λ = 369 nm or CAN $/$ NaNO₃ mixtures at λ = 254 nm. Error bars represent ±1σ uncertainty in binned [NO₃] values. (b) Extinction cross sections (σ_ext) of CAN $/$ HNO₃ and CAN $/$ NaNO₃ mixtures (for details see Sect. 2.3). The black dot (note the superscripted 1 in the legend) corresponds to data from Wine et al. (1988).

3.3 Effect of mixture composition

To characterize the influence of individual reagents on NO₃ formation, tracer decay experiments similar to the measurements shown in Fig. 2 were repeated as a function of [CAN], [HNO₃], and [NaNO₃]. Figure 4a shows [NO₃] obtained from irradiated 6.0 M HNO₃ solutions containing 0.001 to 0.5 M CAN (I₃₆₉ = 7 × 10¹⁵ photons cm⁻² s⁻¹) and irradiated 1.0 M NaNO₃ solutions containing 0.5 to 1.0 M CAN (I₂₅₄ = 1 × 10¹⁶ photons cm⁻² s⁻¹). Results were normalized to [NO₃] achieved with solutions containing 0.5 M CAN and 6.0 M HNO₃. Control experiments conducted with irradiated 6.0 M HNO₃ or 1.0 M NaNO₃ solutions at I₂₅₄ = 1 × 10¹⁶ photons cm⁻² s⁻¹ in the absence of CAN suggest that a fraction of the NO₃ obtained in CAN mixtures was generated via the reactions HNO₃ + hν → OH + NO₂ and HNO₃ + OH → NO₃ + H₂O. The remaining NO₃ was clearly obtained from CAN irradiation because [NO₃] increased with increasing [CAN], as expected from Reaction (R1). Overall, [NO₃] increased by approximately a factor of 3 as [CAN] was increased from 0.001 to 0.5 M in 6.0 M HNO₃.

Figure 4 [NO₃] obtained from (a) irradiated 6.0 M HNO₃ solutions containing 0.001 to 0.5 M CAN (I₃₆₉ = 7 × 10¹⁵ photons cm⁻² s⁻¹), and irradiated 1.0 M NaNO₃ solutions containing 0.5 to 1.0 M CAN (I₂₅₄ = 1 × 10¹⁶ photons cm⁻² s⁻¹). (b) Irradiated 0.5 M CAN solutions containing 1.0 to 6.0 M [HNO₃] or 1.0 to 4.8 M [NaNO₃] at the same I₃₆₉ and I₂₅₄ values used to obtain results shown in (a). Results were normalized to [NO₃] achieved with mixtures of 0.5 M CAN and 6.0 M HNO₃. Error bars represent estimated ±35 % uncertainty in [NO₃] values obtained from CAN $/$ HNO₃ mixtures; ±15 % uncertainty in [NO₃] values obtained from CAN $/$ NaNO₃ mixtures; and ±10 % uncertainty in [CAN], [HNO₃], and [NaNO₃] values.

Figure 4b shows [NO₃] obtained in irradiated solutions containing 0.5 M CAN as a function of [HNO₃] ranging from 1.0 to 6.0 M or [NaNO₃] ranging from 1.0 to 4.8 M at the same I₃₆₉ and I₂₅₄ values used to obtain results shown in Fig. 4a. Irradiated CAN solutions containing 3.0 and 6.0 M HNO₃ generated the same [NO₃] concentrations within measurement uncertainties, presumably because the NO₃ quantum yield (${\mathit{\varphi }}_{{\mathrm{NO}}_{\mathrm{3}}}$) ranged from 0.92–1.00 over this range of acidity (Martin and Stevens, 1978; Wine et al., 1988). [NO₃] decreased by a factor of 2 as [HNO₃] was decreased from 3.0 to 1.0 M, consistent with a reduction in ${\mathit{\varphi }}_{{\mathrm{NO}}_{\mathrm{3}}}$ from 0.92 to 0.46 (Martin and Stevens, 1978). On the other hand, in irradiated CAN $/$ NaNO₃ mixtures with uncharacterized ${\mathit{\varphi }}_{{\mathrm{NO}}_{\mathrm{3}}}$, [NO₃] was constant within measurement uncertainties between 1.0 and 4.8 M NaNO₃.

Other mixture components that were tested or considered included substitution of CH₃CN in place of H₂O and HNO₃, ammonium nitrate (NH₄NO₃) instead of NaNO₃, ceric potassium nitrate (K₂Ce(NO₃)₆) instead of CAN, and addition of sodium persulfate (Na₂S₂O₈) to generate additional NO₃ via ${\mathrm{S}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{8}}^{\mathrm{2}-}$ + hν → 2${\mathrm{SO}}_{\mathrm{4}}^{-}$ followed by ${\mathrm{SO}}_{\mathrm{4}}^{-}$ + ${\mathrm{NO}}_{\mathrm{3}}^{-}$ → NO₃ + ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$ (Gaillard de Sémainville et al., 2007). CAN $/$ CH₃CN mixtures are commonly used in organic synthesis applications, perhaps even more so than CAN $/$ HNO₃ (Baciocchi et al., 1988; Choidini et al., 1993; Alexander, 2004). In limited testing, CAN $/$ CH₃CN appeared to generate significantly less NO₃ than CAN $/$ HNO₃ or CAN $/$ NaNO₃, possibly due to lower ${\mathit{\varphi }}_{{\mathrm{NO}}_{\mathrm{3}}}$ of irradiated Ce^(IV)–CH₃CN complexes (Glebov et al., 2021) and/or suppression of NO₃ due to its reaction with CH₃CN in solution. K₂Ce(NO₃)₆ is less widely available and less water soluble than CAN and therefore was not considered further. Irradiation of CAN $/$ NH₄NO₃ and CAN $/$ NaNO₃ mixtures generated similar [NO₃], but we prefer NaNO₃ due to its lower volatility. Finally, ternary mixtures containing 0.5 M CAN + 2.0 M NaNO₃ + 0.5 M Na₂S₂O₈ irradiated at λ = 254 nm generated negligible additional NO₃ compared to binary CAN $/$ NaNO₃ mixtures.

3.4 Effect of photon flux

Figure 5 shows normalized [NO₃] values obtained from irradiated mixtures of (1) 0.5 M CAN and 6.0 M HNO₃ (λ = 369 nm) and (2) 0.5 M CAN and 1.0 M NaNO₃ (λ = 254 nm) as a function of photon flux ranging from 6.9 × 10¹⁴ to 7.5 × 10¹⁵ photons cm⁻² s⁻¹. Results for both CAN $/$ HNO₃ and CAN $/$ NaNO₃ mixtures were normalized to [NO₃] achieved with 0.5 M CAN, 6.0 M HNO₃, and I₃₆₉ = 6.8 × 10¹⁵ photons cm⁻² s⁻¹. Symbols are colored by the NO₃ lifetime (${\mathit{\tau }}_{{\mathrm{NO}}_{\mathrm{3}}}$), defined here as the time it took for [NO₃] to experience one e-folding decay relative to the maximum [NO₃] that was measured. Figure 5 shows that [NO₃] increased with increasing photon flux, consistent with the fact that it is a primary photolysis product, along with a concurrent decrease in ${\mathit{\tau }}_{{\mathrm{NO}}_{\mathrm{3}}}$ due to faster reduction of Ce^(IV) to Ce^(III). For the CAN $/$ HNO₃ system, [NO₃] increased by a factor of 1.5 as I₃₆₉ was increased from 6.9 × 10¹⁴ to 6.8 × 10¹⁵ photons cm⁻² s⁻¹, in agreement with the model-calculated increase in [NO₃] within measurement uncertainty. ${\mathit{\tau }}_{{\mathrm{NO}}_{\mathrm{3}}}$ decreased from 9 to 5 h. For the CAN $/$ NaNO₃ system, [NO₃] increased by a factor of 1.9 as I₂₅₄ was increased from 1.0 × 10¹⁵ to 7.5 × 10¹⁵ photons cm⁻² s⁻¹, and ${\mathit{\tau }}_{{\mathrm{NO}}_{\mathrm{3}}}$ decreased from 10 to 3 h.

Figure 5 Normalized [NO₃] values obtained from irradiated mixtures of 0.5 M CAN and 6.0 M HNO₃ (λ = 369 nm) or 0.5 M CAN and 1.0 M NaNO₃ (λ = 254 nm) as a function of photon flux ranging from 6.9 × 10¹⁴ to 7.5 × 10¹⁵ photons cm⁻² s⁻¹. Results were normalized to [NO₃] achieved with 0.5 M CAN, 6.0 M HNO₃, and I₃₆₉ = 6.8 × 10¹⁵ photons cm⁻² s⁻¹. Symbols are colored by the time it took for [NO₃] to experience one e-folding decay relative to the maximum [NO₃] that was measured (${\mathit{\tau }}_{{\mathrm{NO}}_{\mathrm{3}}}$). Error bars represent estimated ±35 % uncertainty in [NO₃] values obtained from CAN $/$ HNO₃ mixtures, ±15 % uncertainty in [NO₃] values obtained from CAN $/$ NaNO₃ mixtures, and ±30 % uncertainty in photon flux values.

To examine concentrations of NO₃ and a subset of additional gas-phase photolysis products obtained over a wider range of conditions, Fig. 6 plots model-calculated [NO₃], NO₂ : NO₃, HO₂ : NO₃, and N₂O₅ : NO₃ values as a function of photon flux ranging from 1 × 10¹⁴ to 1 × 10¹⁷ photons cm⁻² s⁻¹ following λ = 254, 313, 369, and 421 nm irradiation of a mixture of 0.5 M CAN and 6.0 M HNO₃. Figure 6a also plots the measured [NO₃] obtained from irradiation of a mixture of 0.5 M CAN and 6.0 M HNO₃ at I₃₆₉ = 7 × 10¹⁵ photons cm⁻² s⁻¹ (Fig. 2) after correcting for dilution between the photoreactor and the OFR (Sect. 2.2) as well as application of a NO₃ wall loss rate coefficient of 0.2 s⁻¹ within the photoreactor (Dubé et al., 2006). At this photon flux value, the model-calculated [NO₃] = 1.4 ppmv agrees with [NO₃] = 1.7 ± 0.6 ppmv obtained from measurements. When considering only the primary photochemical process (Reactions R1–R5), maximum [NO₃] values within ±10 % of each other were achieved at photon fluxes ranging from 5 × 10¹⁵ (λ = 313 nm) to 4 × 10¹⁶ photons cm⁻² s⁻¹ (λ = 421 nm). [NO₃] values decreased at higher photon flux values due to conversion of NO₃ to NO₂ via photolysis. As shown in Fig. 6b, significant additional NO₂ production was obtained via HNO₃ photolysis at shorter irradiation wavelengths above I ≈ 10¹⁵ photons cm⁻² s⁻¹, resulting in NO₂ : NO₃ > 10 (λ = 254 nm) and 1 (λ = 313 nm). Given additional reaction time downstream of the photoreactor, high NO₂ may suppress NO₃ (Sect. 3.2) and increase N₂O₅ : NO₃ beyond the range of values shown in Fig. 6c. We also calculated OH : NO₃ and HO₂ : NO₃ following irradiation of CAN $/$ HNO₃ mixtures over the range of conditions shown in Fig. 6. Aqueous OH : NO₃ ≈ 0.1 and did not change significantly as a function of photon flux or irradiation wavelength, and aqueous HO₂ : NO₃ values ranged from 0.05 (λ = 254 nm) to 0.25 (λ ≥ 369 nm). While OH influenced aqueous-phase chemistry inside the photoreactor via formation of reactive oxygen species (Sect. 3.5), OH probably did not influence downstream gas-phase chemistry due to significant wall losses inside the photoreactor: assuming a lower-limit OH wall loss rate coefficient of 5 s⁻¹ (Schwab et al., 1989), the estimated OH penetration efficiency through the reactor was less than 10⁻⁶.

Figure 6 Model-calculated (a) [NO₃], (b) NO₂ : NO₃, (c) HO₂ : NO₃, and (d) N₂O₅ : NO₃ values in solution as a function of photon flux ranging from 1 × 10¹⁴ to 1 × 10¹⁷ photons cm⁻² s⁻¹ following λ = 254, 313, 369, and 421 nm irradiation of a mixture containing 0.5 M CAN and 6.0 M HNO₃. [NO₃] obtained from measurements shown in Fig. 2 is plotted in (a). For details see Sect. 2.3 and Table S2.

3.5 Characterization of reactive nitrogen and reactive oxygen photolysis products

Figure 7 shows time series of reactive nitrogen and reactive oxygen species detected following irradiation of the same mixture of 0.5 M CAN and 1.0 M NaNO₃ (I₂₅₄ ≈ 10¹⁶ photons cm⁻² s⁻¹), shown here because the signal-to-noise ratio in CIMS measurements of irradiated CAN $/$ NaNO₃ mixtures was generally better than in measurements of irradiated CAN $/$ HNO₃ mixtures due to reagent ion depletion by HNO₃. A time series of [NO₃] obtained separately from VOC tracer decay measurements under similar irradiation conditions is also shown. The NO₂ and NO₃ mixing ratios reached maximum values of 26 and 58 ppbv shortly after the lights were turned on (Fig. 7a), suggesting an initial NO₂ : NO₃ ≈ 0.45 (Fig. 4). Multiple reactions may generate NO₂, including Reaction (R3), HNO₃ photolysis, and/or NO₃ photolysis as well as other reactions listed in Table S2. While NO₂ and/or HNO₂ photolysis generated NO, its concentration was negligible in these experiments.

Figure 7 Time series of (a) NO₂ and NO₃, (b) N₂O₅ and N₂O₆, (c) HNO₂ and HNO₄, and (d) HO₂ and H₂O₂ detected following irradiation of a mixture containing 0.5 M CAN and 1.0 M NaNO₃. N₂O₅, N₂O₆, HO₂, and H₂O₂ were detected as I⁻ adducts, and HNO₂ and HNO₄ were detected as both I⁻ and ${\mathrm{NO}}_{\mathrm{3}}^{-}$ adducts with HR-ToF-CIMS. CIMS signals detected as iodide adducts were normalized to the I⁻ signal prior to the start of the experiment, and CIMS signals detected as nitrate adducts were normalized to the maximum ${\mathrm{NO}}_{\mathrm{3}}^{-}$ obtained during the experiment (see Fig. S5).

Figure 7b shows time series of ${\mathrm{IN}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}^{-}$ and ${\mathrm{IN}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{6}}^{-}$ signals measured with the CIMS. ${\mathrm{IN}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}^{-}$ was formed from NO₂ + NO₃ → N₂O₅ reactions in the photoreactor and N₂O₅ + I⁻ → ${\mathrm{IN}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}^{-}$ reactions in the CIMS ion molecule reactor (IMR). As expected, ${\mathrm{IN}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}^{-}$ followed a similar profile as NO₂ and NO₃. ${\mathrm{IN}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{6}}^{-}$ was either generated from NO₃ + NO₃ → N₂O₆ reactions in the photoreactor (Glass and Martin, 1970) followed by N₂O₆ + I⁻ → ${\mathrm{IN}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{6}}^{-}$ reactions in the IMR or from the following series of reactions in the IMR: HNO₃ + IO⁻ → ${\mathrm{NO}}_{\mathrm{3}}^{-}$ + HOI, HOI + ${\mathrm{NO}}_{\mathrm{3}}^{-}$ → INO₃ + OH⁻, and INO₃ + ${\mathrm{NO}}_{\mathrm{3}}^{-}$ → ${\mathrm{IN}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{6}}^{-}$ (Ganske et al., 2019). To further explore the plausibility of N₂O₆ formation in this system, we conducted a theoretical investigation of the gas-phase NO₃ + NO₃ → N₂O₆ reaction and found that this reaction is exothermic, even more so than NO₃ + NO₃ → N₂O₅. Additional details regarding this analysis are provided in Sect. S1.

Figure 7c shows time series of ${\mathrm{IHNO}}_{\mathrm{2}}^{-}$, ${\mathrm{HNO}}_{\mathrm{2}}{\mathrm{NO}}_{\mathrm{3}}^{-}$, ${\mathrm{IHNO}}_{\mathrm{4}}^{-}$, and ${\mathrm{HNO}}_{\mathrm{4}}{\mathrm{NO}}_{\mathrm{3}}^{-}$. These ions are associated with nitrous acid (HNO₂) and peroxynitric acid (HNO₄), respectively (Veres et al., 2015). Because rapid formation of ${\mathrm{HNO}}_{\mathrm{2}-\mathrm{4}}{\mathrm{NO}}_{\mathrm{3}}^{-}$ ions was observed following Ce^(IV) irradiation, and because ${\mathrm{IO}}_x^{-}$ signals were relatively low (Sect. S2.1), we hypothesize that I⁻ +  NO₃ and/or I⁻ + HNO₃ reactions were the main source of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ (Lee et al., 2014; Dörich et al., 2021) and that subsequent competitive ${\mathrm{NO}}_{\mathrm{3}}^{-}$ + HNO₂₋₄ and I⁻ + HNO₂₋₄ reactions in the IMR generated both ${\mathrm{IHNO}}_{\mathrm{2}-\mathrm{4}}^{-}$ and ${\mathrm{HNO}}_{\mathrm{2}-\mathrm{4}}{\mathrm{NO}}_{\mathrm{3}}^{-}$. HNO₄ was generated following the reactions HNO₃ + hν → OH + NO₂, OH + NO₃ → HO₂ + NO₂, and HO₂ + NO₂ → HNO₄. This hypothesis is supported by the similarity between NO₂ and ${\mathrm{IHNO}}_{\mathrm{4}}^{-}$ time series coupled with the relatively constant concentrations of HO₂ generated via OH + OH → H₂O₂ and OH + H₂O₂ → HO₂ + H₂O reactions. H₂O₂, detected as ${\mathrm{IH}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{2}}^{-}$, also behaved similarly as ${\mathrm{IHO}}_{\mathrm{2}}^{-}$ (Fig. 7d). HNO₂ had a different temporal profile than the other nitrogen oxides: ${\mathrm{IHNO}}_{\mathrm{2}}^{-}$ increased throughout the experiment, and ${\mathrm{HNO}}_{\mathrm{2}}{\mathrm{NO}}_{\mathrm{3}}^{-}$ increased and then decreased. We hypothesize that NO₂ + NO₂ → N₂O₄ and N₂O₄ + H₂O → HNO₂ + HNO₃ reactions were the main source of HNO₂ (Sect. S2.2).

Figure S13 shows time series of the same ions plotted in Fig. 7 following irradiation of a solution containing 0.5 M CAN and 3.0 M HNO₃ (I₃₆₉ ≈ 7 × 10¹⁵ photons cm⁻² s⁻¹). Here, 3.0 M HNO₃ was used because 6.0 M HNO₃ depleted the CIMS reagent ion too much (${\mathrm{IHNO}}_{\mathrm{3}}^{-}$ : I⁻ ≈ 15) to achieve signal-to-noise ratio that was sufficient for comparison to CAN $/$ NaNO₃ mixtures (${\mathrm{IHNO}}_{\mathrm{3}}^{-}$ : I⁻ ≈ 3). The same gas-phase nitrogen oxides and reactive oxygen species were observed in this reaction system as with the irradiated CAN $/$ NaNO₃ mixture. The relative yields of each compound plotted in Figs. 7 and S13 were within a factor of 3 of each other, although signals of nitrogen oxides and reactive oxygen species obtained from irradiated CAN $/$ HNO₃ mixtures decreased at a slower rate than the same compounds obtained from irradiated CAN $/$ NaNO₃ mixtures. These trends may be due to different Ce^(IV) composition (Fig. 3 and Sect. 3.2) and/or an enhanced rate of Ce^(III) + NO₃ → Ce^(IV) reactions in HNO₃ relative to NaNO₃ (Reaction R2).

3.6 OVOC–SOA generation from β-pinene + NO₃

To demonstrate proof of principle for NO₃-initiated oxidative aging studies, we generated NO₃ via irradiation of a mixture of 0.5 M CAN and 3.0 M HNO₃ (I₃₆₉ = 7 × 10¹⁵ photons cm⁻² s⁻¹), reacted it with β-pinene in a dark OFR, and obtained FIGAERO-CIMS spectra of gas- and condensed-phase β-pinene + NO₃ oxidation products (Sect. 2.2). Figure 8a shows a spectrum of gas-phase β-pinene $/$ NO₃ oxidation products detected between $m/Q$ = 320 and 420, where the majority of the signal was observed; signals shown are unmodified (M + I)⁻ formulas. The largest ion detected was at $m/Q$ = 356 (${\mathrm{IC}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{15}}{\mathrm{NO}}_{\mathrm{5}}^{-}$), which represents a major first-generation dicarbonyl nitrate oxidation product with a relative abundance of 0.31 and a calculated saturation vapor pressure of 2 × 10⁻⁷ atm (C^∗ = 1900 µg m⁻³; Claflin, 2018). Other ions corresponding to first-generation hydroxycarbonyl nitrate (${\mathrm{IC}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{17}}{\mathrm{NO}}_{\mathrm{5}}^{-}$, C^∗ = 95 µg m⁻³), tricarbonyl nitrate (${\mathrm{IC}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{15}}{\mathrm{NO}}_{\mathrm{6}}^{-}$, C^∗ = 35 µg m⁻³), hydroxydicarbonyl nitrate (${\mathrm{IC}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{17}}{\mathrm{NO}}_{\mathrm{6}}^{-}$, C^∗ = 4.7 µg m⁻³), and hydroxycarbonyl nitrate acid (${\mathrm{IC}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{17}}{\mathrm{NO}}_{\mathrm{7}}^{-}$, C^∗ = 0.29 µg m⁻³) products were detected in addition to ${\mathrm{IC}}_{\mathrm{9}}{\mathrm{H}}_{\mathrm{13}}{\mathrm{NO}}_{\mathrm{5}}^{-}$ and a suite of additional previously characterized C₈ and C₉ organic nitrates (Nah et al., 2016; Takeuchi and Ng, 2019; Shen et al., 2021). The ${\mathrm{IC}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{7}}^{-}$ hydroxy dinitrate, which was also previously observed in FIGAERO-CIMS spectra of α-pinene $/$ NO₃ SOA (Nah et al., 2016), was generated via an unknown reaction pathway. Overall, the high molar yield and vapor pressure of C₁₀H₁₅NO₅ (Claflin, 2018) are consistent with it having the highest relative abundance in the gas phase (Fig. 8a), whereas the other C₁₀ β-pinene oxidation products were semivolatile under our experimental conditions.

Figure 8 HR-ToF-CIMS spectra of gas-phase β-pinene $/$ NO₃ oxidation products obtained following β-pinene reaction with NO₃ generated via (a) irradiation of a mixture of 0.5 M CAN and 3.0 M HNO₃ and subsequent injection into the OFR (b) thermal decomposition of N₂O₅ injected into the Georgia Tech environmental chamber. Signals shown are unmodified (M + I)⁻ formulas.

Figure 9a shows a spectrum of condensed-phase β-pinene $/$ NO₃ oxidation products obtained with the FIGAERO-CIMS; signals were averaged over the entire thermal desorption cycle and are plotted on logarithmic scale and represent unmodified (M + I)⁻ formulas. To aid interpretation of the major features of the spectrum, bands of ion signals corresponding to ${\mathrm{IC}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{15}}{\mathrm{NO}}_x^{-}$, ${\mathrm{IC}}_{\mathrm{20}}{\mathrm{H}}_{\mathrm{32}}{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_x^{-}$, and ${\mathrm{IC}}_{\mathrm{30}}{\mathrm{H}}_{\mathrm{47}}{\mathrm{N}}_{\mathrm{3}}{\mathrm{O}}_x^{-}$ oxidation products were highlighted and colored by the number of oxygen atoms in their chemical formulas. Here, the largest ion detected was at $m/Q$ = 372 (${\mathrm{IC}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{15}}{\mathrm{NO}}_{\mathrm{6}}^{-}$), which is the condensed-phase component of the same tricarbonyl nitrate detected in the gas phase (Fig. 8a). ${\mathrm{IC}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{15}}{\mathrm{NO}}_{\mathrm{5}}^{-}$ and ${\mathrm{IC}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{15}}{\mathrm{NO}}_{\mathrm{7}-\mathrm{9}}^{-}$ signals were also detected. The second-largest ion signal was measured at $m/Q$ = 571 (${\mathrm{IC}}_{\mathrm{20}}{\mathrm{H}}_{\mathrm{32}}{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{9}}^{-}$), an acetal dimer obtained from the condensed-phase reaction of two C₁₀H₁₇NO₅ monomers followed by H₂O elimination (Claflin and Ziemann, 2018). Similar accretion reactions between other C₁₀ organic nitrates yielded ${\mathrm{IC}}_{\mathrm{20}}{\mathrm{H}}_{\mathrm{32}}{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{8}}^{-}$ and ${\mathrm{IC}}_{\mathrm{20}}{\mathrm{H}}_{\mathrm{32}}{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{10}-\mathrm{13}}^{-}$ signals. Likewise, reactions between C₁₀ monomers and C₂₀ dimers generated C₃₀ trimers detected between $m/Q$ = 768–864 (${\mathrm{IC}}_{\mathrm{30}}{\mathrm{H}}_{\mathrm{47}}{\mathrm{N}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{12}-\mathrm{18}}^{-}$). The largest trimer-related ion, ${\mathrm{IC}}_{\mathrm{30}}{\mathrm{H}}_{\mathrm{47}}{\mathrm{N}}_{\mathrm{3}}{\mathrm{O}}_{\mathrm{12}}^{-}$, was generated from C₁₀H₁₇NO₄ + C₂₀H₃₂NO₉ − H₂O or C₁₀H₁₇NO₅ + C₂₀H₃₂NO₈ − H₂O reactions (Claflin and Ziemann, 2018). A fourth cluster of ion signals at $m/Q$ > 984 was also observed. Unambiguous assignment of chemical formulae to these signals was challenging due to the limited range of the CIMS $m/z$ calibration and lack of available information about C_>30 β-pinene $/$ NO₃ oxidation products. However, it seems plausible that these signals are associated with tetramers.

Figure 9 FIGAERO–HR-ToF-CIMS spectra of condensed-phase β-pinene $/$ NO₃ oxidation products obtained following β-pinene reaction with NO₃ generated via (a) irradiation of a mixture of 0.5 M CAN and 3.0 M HNO₃ and subsequent injection into an OFR and (b) thermal decomposition of N₂O₅ injected into the Georgia Tech environmental chamber. Signals shown are unmodified (M + I)⁻ formulas. Bands of ion signals corresponding to C₁₀H₁₅NO_x, C₂₀H₃₂N₂O_x, and C₃₀H₄₇N₃O_x oxidation products are highlighted and colored by the number of oxygen atoms in their chemical formulas.

To compare our results with those obtained using a conventional NO₃ generation method (room-temperature N₂O₅ thermal decomposition) in an environmental chamber study, Figs. 8b and 9b show reference gas- and condensed-phase FIGAERO-I⁻-CIMS spectra of OVOCs and SOA generated from NO₃ oxidation of β-pinene in the Georgia Tech environmental chamber (Takeuchi and Ng, 2019). The spectra obtained here and by Takeuchi and Ng (2019) exhibit an overall high degree of similarity, with linear correlation coefficients of 0.87 and 0.96 between the respective gas- and condensed-phase spectra. Clusters of ${\mathrm{IC}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{15}}{\mathrm{NO}}_x^{-}$, ${\mathrm{IC}}_{\mathrm{20}}{\mathrm{H}}_{\mathrm{32}}{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_x^{-}$, and ${\mathrm{IC}}_{\mathrm{30}}{\mathrm{H}}_{\mathrm{47}}{\mathrm{N}}_{\mathrm{3}}{\mathrm{O}}_x^{-}$ ion signals are present in both Fig. 9a and b. The main differences between the gas-phase spectra shown in Figs. 8a and 9a were the different abundances of ${\mathrm{IC}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{17}}{\mathrm{NO}}_{\mathrm{4}}^{-}$, a first-generation hydroxynitrate product (Claflin and Ziemann, 2018), and ${\mathrm{IC}}_{\mathrm{10}}{\mathrm{H}}_{\mathrm{16}}{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{7}}^{-}$. Because C₁₀H₁₇NO₄ is formed from RO₂ + RO₂ reactions (DeVault et al., 2022) and is sufficiently volatile (C^∗ = 750 µg m⁻³) to partition into the gas phase (Claflin, 2018), differences in gas-phase C₁₀H₁₇NO₄ and C₁₀H₁₆N₂O₇ yields were likely related to differences in the relative importance of RO₂ + RO₂ versus RO₂ + NO₃ reaction pathways in the study by Takeuchi and Ng (2019) compared to this work.

To further investigate the fate of RO₂ generated from VOC + NO₃ reactions as a function of CAN irradiation conditions, we calculated the fractional oxidative loss of generic alkyl and acyl RO₂ species due to reactions with HO₂, NO₃, and NO₂ ($F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{HO}}_{\mathrm{2}}}$, $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{3}}}$, $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}}$) using Eqs. (1)–(3):

$$\begin{array}{}\text{(1)}& {\displaystyle }\begin{array}{rl}& F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{HO}}_{\mathrm{2}}}=\\ & {\displaystyle \frac{k_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{HO}}_{\mathrm{2}}}\left[{\mathrm{HO}}_{\mathrm{2}}\right]}{k_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{HO}}_{\mathrm{2}}}\left[{\mathrm{HO}}_{\mathrm{2}}\right]+k_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{3}}}\left[{\mathrm{NO}}_{\mathrm{3}}\right]+k_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}}\left[{\mathrm{NO}}_{\mathrm{2}}\right]}},\end{array}\text{(2)}& {\displaystyle }\begin{array}{rl}& F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{3}}}=\\ & {\displaystyle \frac{k_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{3}}}\left[{\mathrm{NO}}_{\mathrm{3}}\right]}{k_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{HO}}_{\mathrm{2}}}\left[{\mathrm{HO}}_{\mathrm{2}}\right]+k_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{3}}}\left[{\mathrm{NO}}_{\mathrm{3}}\right]+k_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}}\left[{\mathrm{NO}}_{\mathrm{2}}\right]}},\end{array}\text{(3)}& {\displaystyle }\begin{array}{rl}& F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}}=\\ & {\displaystyle \frac{k_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}}\left[{\mathrm{NO}}_{\mathrm{2}}\right]}{k_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{HO}}_{\mathrm{2}}}\left[{\mathrm{HO}}_{\mathrm{2}}\right]+k_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{3}}}\left[{\mathrm{NO}}_{\mathrm{3}}\right]+k_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}}\left[{\mathrm{NO}}_{\mathrm{2}}\right]}}.\end{array}\end{array}$$

Here, $k_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{HO}}_{\mathrm{2}}}$, $k_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{3}}}$, and $k_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}}$ are reaction rate coefficients for the corresponding RO₂ + HO₂, RO₂ + NO₃, and RO₂ + NO₂ forward reactions whose values are summarized in Table S3. Several simplifying assumptions were made. First, we assumed that RO₂ + NO reactions were negligible. Second, we did not consider RO₂ isomerization–autoxidation and RO₂ + RO₂ reactions that are influenced by external factors. Third, we set $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}}$ = 0 for alkyl-RO₂-generated RO₂NO₂, which thermally decomposes on timescales of seconds or less (Orlando and Tyndall, 2012). Fourth, we assumed that vapor wall losses of acyl-RO₂-generated RO₂NO₂ were a minor RO₂ sink because the OFR residence time (τ_OFR ≈ 120 s, Sect. 2.2) was significantly shorter than their estimated wall loss timescale (τ_wall ≈ 400 s; Palm et al., 2016). Figure 10 shows calculated $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{HO}}_{\mathrm{2}}}$, $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{3}}}$, and $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}}$ values for alkyl-RO₂ and acyl-RO₂ as a function of photon flux over the range of NO₃ generation conditions presented in Fig. 6. For alkyl-RO₂, $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{HO}}_{\mathrm{2}}}$ decreased and $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{3}}}$ increased with increasing photon flux and decreasing irradiation wavelength. On the other hand, for acyl-RO₂, $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}}$ increased, while $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{HO}}_{\mathrm{2}}}$ and $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{3}}}$ decreased over the same irradiation conditions. Overall, at the optimal NO₃ generation conditions (e.g., λ=369 nm and I₃₆₉ ≈ 10¹⁶ photons cm⁻² s⁻¹), our calculations suggest that $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{HO}}_{\mathrm{2}}}$ ≈ $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{3}}}$ for alkyl-RO₂ (Fig. 10c) and that $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{HO}}_{\mathrm{2}}}$ ≈ $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{3}}}$ ≈ $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}}$ for acyl-RO₂ (Fig. 10g).

Figure 10 Fractional oxidative loss of alkyl and acyl organic peroxy radicals (RO₂) due to reaction with HO₂, NO₃, and NO₂ ($F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{HO}}_{\mathrm{2}}}$, $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{3}}}$, and $F_{{\mathrm{RO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{2}}}$) generated following (a, e) λ = 254, (b, f) λ = 313, (c, g) λ = 369, and (d, h) λ = 421 nm irradiation of a mixture containing 0.5 M CAN and 6.0 M HNO₃ as a function of photon flux ranging from 1 × 10¹⁴ to 1 × 10¹⁷ photons cm⁻² s⁻¹.

4 Conclusions

Ce^(IV) irradiation complements NO₂ + O₃ reactions and N₂O₅ thermal dissociation as a customizable photolytic NO₃ source. Important method parameters were [CAN], [HNO₃], or [NaNO₃]; UV intensity; and irradiation wavelength. By contrast, important parameters for NO₂ + O₃ and N₂O₅ based methods are [O₃], [NO₂], temperature, and humidity. Because Ce^(IV) irradiation already generates NO₃ in aqueous solution, its performance is not hindered by humidity to the same extent (if at all) as N₂O₅-based methods, where hydrolysis of N₂O₅ to HNO₃ decreases the efficacy of the source. Additionally, the NO₃ + H₂O reaction rate in solution or on surfaces is slow in relation to other NO₃ loss pathways (Jiang et al., 1992). Another advantage of Ce^(IV) irradiation is that it does not involve the use of O₃ as a reagent; therefore it eliminates the possibility of competing O₃ and NO₃ oxidation of compounds that are reactive towards both oxidants if NO₂ + O₃ reactions and/or online N₂O₅ synthesis are used as the NO₃ source (Lambe et al., 2020). To identify optimal operating conditions for maximizing [NO₃], we characterized concentrations of NO₃ at [CAN] = 10⁻³ to 1 M, [HNO₃] = 1.0 to 6.0 M, [NaNO₃] = 1.0 to 4.8 M, a photon flux of  6.9 × 10¹⁴ to 1.0 × 10¹⁶ photons cm⁻² s⁻¹, and irradiation wavelengths of λ = 254, 313, 369, or 421 nm. With CAN $/$ HNO₃ mixtures, maximum [NO₃] was achieved with [CAN] ≈ 0.5 M, [HNO₃] ≈ 3.0 to 6.0 M, and I₃₆₉ = 8 × 10¹⁵ photons cm⁻² s⁻¹ (4.3 mW cm⁻²). With CAN $/$ NaNO₃ mixtures, maximum [NO₃] was achieved with [CAN] ≈ 1.0 M, [NaNO₃] ≥ 1.0 M, and I₂₅₄ ≈ 1 × 10¹⁶ photons cm⁻² s⁻¹ (7.8 mW cm⁻²). Thus, for applications such as environmental chambers or OFR studies of NO₃-initiated oxidative aging processes, where significant NO₃ production over relatively short time periods is beneficial, irradiation of concentrated Ce^(IV) solutions at high photon flux is advantageous. Other applications that require sustained NO₃ production at lower concentrations and/or over longer time periods may benefit from using lower [Ce^(IV)] and photon flux. Overall, because Ce^(IV) irradiation generates NO₃ at room temperature using widely available, low-cost reagents and light sources (including high-power light-emitting diodes in addition to, or instead of, UV fluorescent lamps), it is easier to apply than other NO₃ generation techniques – especially in field studies – and it may therefore enable more widespread studies of NO₃ oxidation chemistry. Adapting a photoreactor to operate with continuous injection of fresh Ce^(IV) or alternative photolytic NO₃ precursors (e.g., Hering et al., 2015) rather than in batch mode, as was done here, may further enhance its performance and will be investigated in future work.