The production of HONO from illuminated aerosol surfaces is studied using an aerosol flow tube system coupled to a photo-fragmentation laser-induced fluorescence (PF-LIF) cell which allows the highly sensitive detection of the OH radical formed through photo-fragmentation of HONO into OH and NO followed by laser-induced fluorescence (LIF) detection at low pressure. The experimental setup used in this investigation is described in detail in Boustead (2019), as well as similar systems having been used to measure HONO in the field (Liao et al., 2006; Wang et al., 2020), and therefore only a brief description of the setup is given here. A schematic of the experimental setup is shown in Fig. 1.

All experiments were conducted at room temperature (295 ± 3 K) using nitrogen (BOC, 99.998 %) or air (BOC, 21 ± 0.5 % O2) as the carrier gas. A humidified flow of aerosols, ∼ 6 L min-1 (total residence time of 104 s in the flow tube), was introduced through an inlet at the rear of the aerosol flow tube (Quartz, 100 cm long, 11.5 cm i.d.), which was covered by a black box to eliminate the presence of room light during experiments. A 15 W UV lamp (XX-15LW bench lamp, λpeak=365 nm) was situated on the outside of the flow tube to illuminate aerosols and promote the production of HONO (half the length of the flow tube was illuminated, leading to an illumination time of 52 s). The concentration of HONO is measured by PF-LIF with sampling from the end of the flow tube via a protruding turret containing a 1 mm diameter pinhole, through which the gas exiting the flow tube was drawn into the detection cell at 5 L min-1. The detection cell was kept at low pressure, ∼ 1.5 Torr, using a rotary pump (Edwards, E1M80) in combination with a roots blower (Edwards, EH1200). All gas flows in the experiment were controlled using mass flow controllers (MKS and Brooks). The relative humidity (RH) and temperature of the aerosol flow was measured using a probe (Rotronics HC2-S, accuracy ±1 % RH) the former calibrated against the H2O vapour concentration measured by a chilled mirror hygrometer (General Eastern Optica) in the exhaust from the flow tube.

Solutions for the generation of TiO2 aerosol solutions were prepared by dissolving 5 g of titanium dioxide (Aldrich Chemistry 718467, 99.5 % Degussa, 80 % anatase : 20 % rutile) into 500 mL of Milli-Q water. Polydisperse aerosols were then generated from this solution using an atomiser (TSI model 3076), creating a 1 L min-1 flow of TiO2 aerosol particles in nitrogen hereafter referred to as the aerosol flow. This aerosol flow was then passed through a silica drying tube (TSI 3062, capable of reducing 60 % RH incoming flow to 20 % RH) to remove water vapour and then passed through a neutraliser to apply a known charge distribution and reduce loss of aerosols to the walls. After the neutraliser the aerosol flow was mixed with both a dry and a humidified N2 flow (controlled by mass flow controllers) to regulate the relative humidity of the system by changing the ratio of dry to humid nitrogen flows. A conditioning tube was then used to allow for equilibration of water vapour adsorption and re-evaporation to and from the aerosol surfaces for the chosen RH, which was controlled within the range ∼ 10 %–70 % RH. A portion of the aerosol flow was then passed through a high-efficiency particle filter (HEPA) fitted with a bypass loop and bellows valve allowing control of the aerosol number concentration entering the aerosol flow tube. Previous studies (George et al., 2013; Boustead, 2019) have shown the loss of aerosol to the walls of the flow tube to be negligible. Aerosol size distributions were measured for aerosols exiting the flow tube using a scanning mobility particle sizer (SMPS, TSI 3081) and a condensation particle counter (CPC, TSI 3775) which was calibrated using latex beads. Any aerosol surface area not counted due to the upper diameter range of the combined SMPS/CPC (14.6–661.2 nm, sheath flow of 3 L min-1, instrumental particle counting error of 10 %–20 %) was corrected for during analysis by assuming a log-normal distribution, which was verified for TiO2 aerosols generated in this manner (Matthews et al., 2014). However, the majority of aerosols, > 90 %, had diameters in the range that could be directly detected. In addition to the experiments with single-component TiO2, mixed ammonium nitrate/TiO2 and single-component ammonium nitrate aerosols were also generated using the atomiser for investigations of HONO production from nitrate aerosols without NO2 present. An example of an aerosol size distribution from this work for single-component ammonium nitrate aerosols, mixed ammonium nitrate/TiO2 and single-component TiO2 aerosols is shown in Fig. 2.

As HONO is not directly detectable via LIF, it was necessary to fragment the HONO produced into OH and NO (Liao et al., 2007), with detection of OH via LIF. A 355 nm photolysis laser (Spectron Laser Systems, SL803) with a pulse repetition frequency (PRF) of 10 Hz and pulse duration ∼ 10 ns was used to fragment HONO into OH. This fragmentation wavelength was chosen as HONO has a strong absorption peak at ∼ 355 nm, leading to the breakage of the HO–NO bond to form NO and OH in their electronic ground states (Shan et al., 1989). A Nd:YAG pumped dye probe laser (JDSU Q201-HD, Q series, Sirah Cobra Stretch) with a PRF of 5000 Hz was used for the detection of OH via the fluorescence assay by gas expansion (FAGE) technique, which employs the expansion of gas through a small pinhole into the detection cell. The OH radical was measured using on-resonance detection by LIF via the excitation of the A2Σ+ (ν′=0) ←X2Πi (ν′′=0) Q1(2) transition at 308 nm (Heard, 2006). A multi-channel plate (MCP) photomultiplier (Photek, MCP 325) equipped with an interference filter at 308 nm (Barr Associates, 308 nm, FWHM – 8 nm, ∼ 50 % transmission) was used to measure the fluorescence signal. A reference OH cell in which a large LIF signal could be generated was utilised to ensure the wavelength of the probe laser remained tuned to the peak of the OH transition at 308 nm. OH measurements are taken both before and after each photolysis laser pulse, allowing measurement of any OH already present in the gas flow to be determined as a background signal for subtraction. The OH generated from HONO photolysis was measured promptly (∼ 800 ns) after the 355 nm pulse to maximise sensitivity to OH before it was spatially diluted away from the measurement region (Boustead, 2019). Offline measurements, with the probe laser wavelength moved away from the OH transition (by 0.02 nm), were taken to allow the signal generated from detector dark counts and scattered laser light to be measured and subtracted from the online signal. To determine an absolute value of the HONO concentration, [HONO], a calibration was performed, in order to convert from the HONO signal, SHONO, using SHONO=CHONO [HONO], as described fully in Boustead (2019). A glass calibration wand was used to produce OH and HO2 in equal concentrations from the photolysis of water vapour at 185 nm: R7H2O+hν⟶λ=185nmOH+H,R8H+O2+M→HO2+M.

An excess flow of NO was then added to generate HONO, which was then detected as OH in the cell. The excess flow of NO (BOC, 99.5 %) ensures rapid and complete conversion of OH and HO2 to HONO. The concentration of OH and HO2 produced, and therefore the amount of HONO produced in the wand, is calculated using 1OH=HO2=H2OσH2OϕOHFt, where [H2O] is the concentration of water vapour in the humidified gas flow, σH2O is the absorption cross section of H2O at 185 nm 7.14 × 10-20 cm2 molec.-1 (Cantrell et al., 1997), ϕOH is the quantum yield of OH for the photo-dissociation of H2O at 185 nm (=1), F is the lamp flux and t is the irradiation time (the product of which is determined using ozone actinometry (Boustead, 2019).

A typical value of the calibration factor was CHONO=(3.63±0.51)×10-9 counts mW-1 for N2, leading to a calculated limit of detection of 12 ppt for a 50 s averaging period and a signal-to-noise ratio (SNR) of 1 (Boustead, 2019). The typical error in the HONO concentration was 15 % at 1σ, determined by the error in the calibration.

The experiments were performed with a minimum flow of 6 L min-1 through the aerosol flow tube, giving a Reynolds number of ∼ 150, which ensured a laminar flow regime. The HONO signal, converted to an absolute concentration using a calibration factor, was measured over a range of aerosol surface area densities, both in the presence and absence of illumination, and background measurements without aerosols present were also performed.

The HONO signal originates from several sources: the illuminated aerosol surface, the illuminated quartz flow tube walls, dark reactions on aerosol surfaces, dark reactions on the flow tube surface, and finally from impurities in the NO2 (Sigma Aldrich, > 99.5 %, freeze pump thawed to further remove any remaining NO or O2) and N2 flows (either HONO itself or a species which photolyses at 355 nm to give OH). Of interest here is the HONO production from both dark and illuminated aerosol surfaces which is atmospherically relevant. Following transit through the flow tube, and in the presence of NO2, the total concentration of HONO measured by the PF-LIF detector is given by 2HONO=HONOilluminated  aerosols+[HONO]illuminated walls+[HONO]dark aerosols+[HONO]dark walls+[HONO]impurities.

Any HONO seen without the presence of aerosol was therefore due to HONO impurities in the N2 or NO2 gas, the dark production of HONO from the flow tube walls or from the production of HONO from the illuminated reactor walls, which may include production from TiO2 aerosols coating the flow tube in the presence of NO2. This background HONO concentration depended on the experimental conditions and on how recently the flow tube and PF-LIF cell had been cleaned to remove any build-up of TiO2 deposits. However, the build-up of TiO2 on the flow tube walls was relatively slow, and back-to-back measurements were made in the presence and absence of aerosols to obtain an accurate background. Additional experiments showed no significant production of HONO on TiO2 aerosol surfaces without the presence of NO2. Even though the aerosol surface area density (∼ 0.02 m2 m-3) was small compared to the surface area density of the reactor walls (35 m2 m-3), very little HONO signal was produced without the presence of aerosols and was always subtracted from the signal in the presence of aerosols. The HONO signal was measured both with the lamp on and off for each aerosol surface area density to investigate the production of HONO from illuminated aerosol surfaces. The HONO signal was averaged over 50 s (average of 500 of the 355 nm photolysis laser pulses with a PRF of 10 Hz). Once aerosols were introduced into the flow tube system, a period of ∼ 30 min was allowed for equilibration and the measured aerosol surface area density to stabilise. In general, the relative humidity of the system was kept constant at RH ∼ 15 % for all experiments investigating HONO production as a function of NO2 mixing ratio over the range 34–400 ppb. In a number of experiments, however, RH was varied in the range ∼ 12 %–37 %.

The mixing ratio of NO2 entering the flow tube was calculated using the concentration of the NO2 in the cylinder and the degree of dilution. The NO2 mixing ratio within the cylinder was determined using a commercial instrument based on UV–Vis absorption spectroscopy (Thermo Fisher 42TL, limit of detection 50 pptv, precision 25 pptv). For each individual experiment, the mixing ratio of NO2 was kept constant (within the range 34–400 ppb), and the aerosol surface area density was varied from zero up to a maximum of 0.04 m2 m-3, in order to obtain the HONO produced from illuminated aerosol surfaces in the flow tube for a given mixing ratio of NO2. As well as subtraction of any background HONO, a correction must be made for any loss of HONO owing to its photolysis occurring within the flow tube.

In order to determine the rate of photolysis of HONO, the rate of photolysis of NO2 was first determined using chemical actinometry, and the known spectral output of the lamp and the literature values of the absorption cross sections and photo-dissociation quantum yields for NO2 and HONO were used to determine the rate of photolysis of HONO. When just flowing NO2 in the flow tube, the loss of NO2 within the illuminated region is determined only by photolysis and is given by 3-dNO2dt=j(NO2)[NO2], where j(NO2) is the photolysis frequency of NO2 for the lamp used in these experiments. From the measured loss of NO2 in the illuminated region, and with knowledge of the residence time, the photolysis frequency, j(NO2), was determined to be (6.43 ± 0.30) × 10-3 s-1 for the set of experiments using one lamp to illuminate the flow tube. j(NO2) is given by 4jNO2=∫λ1λ2σλϕλFλdλ, where λ1 and λ2 represent the range of wavelengths over which the lamp emits, and σλ and ϕλ are the wavelength-dependent absorption cross section and photo-dissociation quantum yield of NO2, respectively, and Fλ is the flux of the lamp at a given wavelength. The flux of the lamp, the spectral intensity of which was measured using a spectral radiometer (Ocean Optics QE65000) as a function of wavelength, is shown in Fig. 3.

From the measured j(NO2), and with knowledge of σλ and φλ for NO2, the flux of the lamp was determined to be (1.63 ± 0.09) × 1016 photons cm-2 s-1 integrated over the 290–400 nm wavelength range of the lamp. Using this flux, and the known σλ and φλ for HONO over the same wavelength range, j(HONO) was determined to be (1.66 ± 0.10) × 10-3 s-1.

In the presence of aerosols under illuminated conditions, the rate of heterogeneous removal of NO2 at the aerosol surface to generate HONO is given by 5-dNO2dt=k[NO2], where k is the pseudo-first-order rate coefficient for loss of NO2 at the aerosol surface, which leads to the generation of HONO. The postulated mechanism for HONO production from NO2 is discussed in Sect. 3.3.2 below, but for the definition of k it is assumed to be a first-order process for NO2. Integration of Eq. (5) gives 6k=-lnNO20-HONOtNO20t, where NO20-[HONO]t is the concentration of NO2 at time t, assuming that each NO2 molecule is quantitatively converted to a HONO molecule following surface uptake (see Sect. 3.3.2 for the proposed mechanism), and NO20 is the initial concentration of NO2. Hence k can be determined from Eq. (6) using the measurement of the concentration of HONO, [HONO], that has been generated from TiO2 aerosol surfaces for an illumination time of t (and after subtraction of any background HONO produced from other sources and after correction for loss via photolysis; see above) and with knowledge of [NO2]0.

The reactive uptake coefficient of NO2 to generate HONO, γNO2→HONO, defined as the probability that upon collision of NO2 with the TiO2 aerosol surface a gas-phase HONO molecule is generated, is given by 7γNO2→HONO=4×kν×SA, where ν is the mean thermal velocity of NO2, given by ν=((8RT)/(πM)) with R, T and M as the gas constant, the absolute temperature and the molar mass of NO2, respectively; SA is the aerosol surface area density (m2 m-3); and k is defined as above. Rearrangement of Eq. (7) gives 8k=γNO2→HONO×SA×ν4.

Figure 4 shows the variation of k,determined from Eq. (6) above with t=52 s (illumination time in the flow tube), against aerosol surface area density, SA, for [NO2]0=200 ppb and RH = 15 %, from which the gradient using Eq. (8) yields γNO2→HONO=(2.17±0.09)×10-5.

The uncertainty in k (∼ 20 %) shown in Fig. 4 and determined by Eq. (6) is mainly controlled by the uncertainty in the HONO concentration (the HONO signal typically varies between repeated runs for a given SA by ∼ 10 % coupled with the 15 % error in the calibration factor), the initial NO2 mixing ratio (10 %), and the photolysis time, t (∼ 3 %). The uncertainty in SA is determined by the uncertainty in the SMPS (15 %). The error in the value of γNO2→HONO (typically 20 %) is calculated from the 1σ statistical error of the weighted fit shown in Fig. 4. An experiment performed using air yielded an uptake coefficient value within 7 % of the equivalent experiment done in N2, which is well within the experimental error.

A kinetic scheme within the framework of a box model was used together with the differential equation solver Facsimile 4.3.53 (MCPA Software Ltd., 2020) to investigate the mechanism of NO2 adsorption on TiO2 in the presence of light to produce HONO. The models were only semi-explicit, focusing on determining the stoichiometric amounts of NO2 needed to produce a single HONO molecule in the gas phase for comparison with the experimental dependence of HONO production upon NO2 mixing ratio and to provide a predictive framework for parameterising the HONO production rate with NO2 mixing ratio in the atmosphere. Three model scenarios were designed. The simplest model (Model 1) considered only the adsorption of a single molecule of NO2 to the TiO2 surface, the surface conversion to HONO in the presence of light and subsequent desorption of HONO – the latter was assumed to occur rapidly. The two further model scenarios investigated the effect of a 2:1 stoichiometric relationship between the NO2 adsorbed to the surface of TiO2 and the HONO produced, via the formation of an NO2 dimer. Model 2 incorporated an Eley–Rideal mechanism reliant on the adsorption of one NO2 molecule to the surface followed by the subsequent adsorption of a second NO2 molecule directly onto the first (Fig. 5). Model 3, however, features a Langmuir–Hinshelwood mechanism of adsorption in which two NO2 molecules adsorb to the surface and then diffuse to one another before colliding on the surface and forming the cis-ONO-NO2 dimer (Finlayson-Pitts et al., 2003; de Jesus Madeiros and Pimentel, 2011; Liu and Goddard, 2012; Varner et al., 2014). The formation of the asymmetric cis-ONO-NO2 dimer followed by isomerisation to form the asymmetric trans-ONO-NO2 dimer has been suggested to have an enthalpic barrier that is ∼ 170 kJ mol-1 lower than for direct isomerisation to trans-ONO-NO2 from the symmetric N2O4 dimer (Liu and Goddard, 2012). The dimerisation of NO2 and subsequent isomerisation to form trans-ONO-NO2 has been suggested under dark conditions to lead to the formation of both HONO and HNO3 in the presence of water vapour (Finlayson-Pitts et al., 2003; de Jesus Madeiros and Pimentel, 2011; Liu and Goddard, 2012; Varner et al., 2014). Although the interaction of light with TiO2 with the concomitant production of electron–hole pairs (Reaction R1) is central to HONO formation, we do not specify here the exact mechanism by which the electron–hole pairs interact with surface-bound species to generate HONO. We propose that the interaction with light speeds up the autoionisation of trans-ONO-NO2 to form (NO+)(NO3-), which is represented by Reactions (R13) and (R15) in Model 2 and Model 3 respectively. (NO+)(NO3-) can then react rapidly with surface-adsorbed water, leading to HONO formation (Varner et al., 2014).

A schematic of the proposed mechanism investigated with Model 2 and Model 3 is shown in Fig. 5 and consists of (i) the adsorption of NO2 onto a surface site, (ii) the conversion of NO2 to form HONO via the formation of an NO2 dimer intermediate on the surface via either a Eley–Rideal or Langmuir–Hinshelwood-type mechanism, (iii) subsequent desorption of HONO from the surface, and finally (iv) competitive removal processes for HONO both on the surface and in the gas phase that are either dependent or independent on the NO2 mixing ratio. The model includes the gas-phase photolysis of NO2 and HONO and the gas-phase reactions of both HONO and NO2 with OH and O(3P) atoms.

To the best of our knowledge neither the enthalpy of adsorption of NO2 onto a TiO2 surface nor the bimolecular rate coefficients for the chemical steps on the surface shown in Fig. 5 have been determined. Hence, for each of the steps a rate coefficient (s-1 or cm3 molec.-1 s-1) was assigned, as given in Table 1, and with the exception of the experimentally determined j(NO2) and the calculated j(HONO), and the gas-phase rate coefficients which are known, the rate coefficients were estimated, with the aim of reproducing the experimental NO2 dependence of the HONO production and NO2 reactive uptake coefficient; justification of chosen values is given below.

The modelled Gibbs free energy barrier for the isomerisation of N2O4 to form the asymmetric ONO-NO2 isomer (cis or trans conformation not specified) was estimated by Pimental et al. (2007) to be 87 kJ mol-1 with a rate coefficient as large as 2 × 10-3 s-1 in the aqueous phase at 298 K, stated in the study to confirm the Finlayson-Pitts model for the hydrolysis of NO2 on surfaces via the asymmetric trans-ONO-NO2 dimer (Finlayson-Pitts et al., 2003). Using this study as a guide, we estimated kR13 and kR15 as 5 × 10-3 s-1, slightly larger than that estimated by Pimental et al. (2007) due to the presence of light. A study into the decomposition of HONO on borosilicate glass surfaces suggested a rate coefficient for the loss HONO on the non-conditioned chamber walls to be (1.0 ± 0.2) × 10-4 s-1 increasing to (3.9 ± 1.1) × 10-4 s-1 when HNO3 was present on the walls (Syomin and Finlayson-Pitts, 2003). From this we estimated a light-accelerated loss rate coefficient of 1 × 10-3 s-1 for the loss of HONO(ads) by reaction with itself, kR18, and through reaction with HNO3(ads), kR17. Both of these reactions will occur on the surface of the aerosol. We make the assumption that the rate of loss of HONO to the walls of the chamber for this experiment is less than that of the heterogeneous loss reactions on the photocatalytic aerosol surface, leading to a kR22 of 1 × 10-4 s-1 as reported by Syomin and Finlayson-Pitts (2003). For kR12-kR15, initial values were adopted and were then adjusted to fit the shape of the trend in experimental results of [HONO] and γNO2→HONO versus [NO2], discussed fully in Sect. 3.3.2. For completeness, gas-phase loss reactions of HONO and NO2 with OH and the reactions of O(3P) with NO, NO2 and O2 were also included in the model, Reactions (R23)–(R27), though their inclusion had no effect on the HONO concentration. The rates of Reactions (R23)–(R27) within the model are much smaller than HONO loss reactions on the surface, Reactions (R17)–(R19), and the photolysis Reaction (R21). For both Model 2 and Model 3, the adsorption of an NO2 molecule to the surface, kR9, was assumed to be rapid and not the rate-determining step. Likewise, the desorption of HONO was also assumed to be rapid – faster than the loss rates of adsorbed HONO but slower than the adsorption of NO2; this was necessary for the model to reproduce the trend in the experimental results of [HONO] versus [NO2], discussed fully in Sect. 3.3.2.

The production of HONO on TiO2 aerosol surfaces was measured as a function of the initial NO2 mixing ratio. Figure 6 shows the dependence of the HONO concentration, measured at the end of the flow tube, on the initial NO2 mixing ratio for an aerosol surface area of (1.6 ± 0.8) × 10-2 m2 m-3. A sharp increase in HONO production at a low mixing ratio of NO2 was seen, followed by a more gradual reduction in HONO production after a peak production at ∼ 54 ± 5 ppb NO2.

Figure 7 shows the HONO concentration measured at the end of the flow tube over a range of RH values for a fixed aerosol surface area density of (1.59 ± 0.16 × 10-2 m2 m-3) and at two NO2 mixing ratios, displaying a peak in HONO production between 25 %–30 % RH. Above ∼ 37 % RH, for experiments including single-component TiO2 aerosols, it was found that significant aerosols were lost from the system before entering the flow tube, speculated to be due to loss to the walls of the Teflon lines. As such the RH dependence was only studied up to 37 % RH; however, a clear drop-off in HONO production was seen for both NO2 mixing ratios studied after ∼ 30 % RH.

A dependence of HONO production upon RH was expected due to the potential role of water as a proton donor in the production mechanism of HONO on TiO2 surfaces (Reactions R2 and R5, as shown in Fig. 5) (Dupart et al., 2014). The fractional surface coverage of water on the TiO2 aerosol core, V/Vm, at 15 % RH and above was calculated using the parameterisation below, which was determined using transmission IR spectroscopy (Goodman et al., 2001): 9VVm=cPP01-PP01-n+1PP0n+nPP0n+11+(c-1)PP0-cPP0n+1, where V is the volume of water vapour adsorbed at equilibrium pressure P, Vm is the volume of gas necessary to cover the surface of TiO2 particles with a complete monolayer, P0 is the saturation vapour pressure, c is the temperature-dependent constant related to the enthalpies of adsorption of the first and higher layers (taken as 74.8 kJ mol-1 for TiO2, Goodman et al., 2001), and n is the asymptotic limit of monolayers (eight for TiO2, Goodman et al., 2001) at large values of P/P0.

At 15 % RH, a fractional water coverage of 1.09 was calculated to be present on the surface, increasing to 1.50 at 35 % RH. It has been shown in previous work that HONO can be displaced from a surface by water, leading to an increase in gas-phase HONO with RH (Syomin and Finlayson-Pitts, 2003). The increase in HONO with RH to ∼ 25 %–30 % RH could therefore be attributed to both an increase in the concentration of the water reactant leading to more HONO formation and the increase in displacement of HONO from the surface due to preferential adsorption of water. A decrease in HONO production seems to occur above ∼ 30 % RH, which could be due to the increased water adsorption inhibiting either NO2 adsorption or the electron–hole transfer process (Gustafsson et al., 2006). H2O vapour adsorption is likely enhanced by the superhydrophilic properties of TiO2 surfaces under UV radiation, meaning that water monolayers form more quickly on the surface of TiO2 owing to light-induced changes in surface tension (Takeuchi et al., 2005; Gustafsson et al., 2006).

At the higher initial concentration of NO2=170 ppb, the RH dependence showed a similar peak in HONO production between ∼ 25 %–30 % RH, but less HONO was produced overall, as expected from Fig. 6 given the higher NO2. Previous work on the production of HONO from suspended TiO2 aerosols reported a strong RH dependence of the uptake coefficient, γ, of NO2 to form HONO with a peak at ∼ 15 % RH and decreasing at larger RH (Gustafsson et al., 2006). The same trend for the NO2 uptake coefficient was observed by Dupart et al. (2014) on Arizona test dust (ATD) aerosols with a peak in γ at ∼ 25 % RH. This increase in the RH at which the uptake coefficient for NO2 in going from TiO2 to ATD aerosols was ascribed to the lower concentration of TiO2 present in ATD aerosols as opposed to single-component TiO2 aerosols used by Gustafsson et al. (2006) as well as by differences in particle size distribution. Gustafsson et al. (2006) reported a larger aerosol size distribution with a bimodal trend with mode diameters of ∼ 80 and ∼ 350 nm for single-component TiO2 aerosols, whereas Dupart et al. (2014) reported a smaller unimodal aerosol size distribution for ATD aerosols with a mode diameter of ∼ 110 nm. In this work we also see a larger aerosol size distribution, with a lower mode diameter of ∼ 180 nm similar to Dupart et al. (2014) but for pure TiO2 aerosols (aerosol size distribution shown in Fig. 2). Similar to the results of Dupart et al. (2014) we observe a trend inversion in [HONO] vs. RH at higher RH, between 25 %–30 %. An increase in HONO as a function of RH has also been observed on TiO2-containing surfaces (Langridge et al., 2009; Gandolfo et al., 2015; Gandolfo et al., 2017) with a similar profile for the observed RH dependence of HONO being observed by Gandolfo et al. (2015) from photocatalytic paint surfaces with a maximum in HONO mixing ratio found at 30 % RH. In comparison, a study focusing on the products of the uptake of NO2 on TiO2 surfaces showed a maximum in the gas-phase HONO yield at 5 % RH, with the yield of HONO plateauing off with further increase in humidity (Bedjanian and El Zein, 2012).

The reactive uptake coefficient, γNO2→HONO for NO2→HONO, on TiO2 aerosol particles was determined experimentally for 18 different initial NO2 mixing ratios and is shown in Fig. 8. For each initial NO2 mixing ratio, the gradient of the first-order rate coefficient for HONO production, k, as a function of aerosol surface area density (e.g. Fig. 4) and in conjunction with Eq. (8) was used to obtain γNO2→HONO. The uptake coefficient initially increases with NO2, reaching a peak at γNO2→HONO=(1.26±0.17)×10-4 for an initial NO2 mixing ratio of 51±5 ppb, before sharply decreasing as the NO2 mixing ratio continues to increase above this value.

The increase in uptake coefficient with NO2 at low NO2 (< 51 ppb) has not been seen previously in studies of HONO production from TiO2-containing aerosols with similar [NO2] ranges (Gustafsson et al., 2006; Ndour et al., 2008; Dupart et al., 2014) or with other aerosol surfaces (Bröske et al., 2003; Stemmler et al., 2007) or TiO2 surfaces (El Zein and Bedjanian, 2012b). It is worth noting that several of these studies reported the overall uptake of NO2 onto aerosol surfaces and not specifically the uptake to form HONO, although HONO was indirectly measured in all studies noted here (Gustafsson et al., 2006; Ndour et al., 2008; Dupart et al., 2014). For single-component TiO2 aerosols, Gustafsson et al. (2006) reported a uptake coefficient, γNO2, of 9.6 × 10-4 at 15 % RH and 100 ppb NO2. Taking into account the HONO yield of 0.75 given by Gustafsson et al. (2006), an estimated γNO2→HONO=7.2×10-4 is determined and can be compared to the value observed in this work at 15 % RH and 100 ppb NO2, (γNO2→HONO= (2.68 ± 0.23) × 10-5). The γNO2→HONO we determine is 27 times smaller than reported by Gustafsson et al. (2006). This difference is mostly due to the lower experimental photon flux in our setup, ∼ 19 times less at λmax⁡=365 nm owing to the use of one 15 W UV lamp to irradiate the flow tube (Boustead, 2019) compared to Gustafsson et al. (2006), who utilised four 18 W UV lamps.

The origins of the increase in γNO2→HONO, together with reaching a maximum, and the subsequent decrease at larger NO2 mixing ratios were investigated using the kinetic box model and postulated mechanism for HONO production described in Sect. 2.5. The aim was to compare the observed production of HONO and γNO2→HONO with the modelled values, as a function of NO2 mixing ratio. The skill of the model to reproduce the observed behaviour enables a validation of the postulated mechanism for HONO production, and variation of the kinetic parameters enables the controlling influence of different steps in the mechanism on HONO production to be evaluated.

The HONO production on illuminated TiO2 aerosol surfaces was investigated for each of the mechanisms outlined in Table 1.

The photolysis of particulate nitrate has been postulated as a source of HONO under ambient sunlit conditions during several field campaigns, from both aircraft- and ground-based measurements (Reed et al., 2017; Ye et al., 2017a, b). Here, experiments were carried out to investigate the formation of HONO from particulate nitrate photolysis, with and without the addition of a photocatalyst. This is of significant interest for marine environments downwind of arid desert regions due to the availability of TiO2 or other photocatalytic materials within aerosols in dust plumes that are transported from these regions (Hanisch and Crowley, 2003).

Using the aerosol flow tube setup described in Sect. 2.1–2.4, an aqueous solution of ammonium nitrate (5 g NH4NO3 in 500 mL Milli-Q water) was used to generate nitrate aerosols. At the RH used in this experiment, ∼ 50 %, the aerosols were still deliquesced. For these experiments the residence time of the aerosols in the illuminated region of the flow tube was 30 s (flow rate ∼ 6 L min-1), with the production of HONO following illumination measured as a function of aerosol surface area density. The number of lamps was increased from one to four, increasing the photon flux from (1.63 ± 0.09) × 1016 to (8.21 ± 2.39) × 1016 photons cm-2 s-1 and j(NO2) from (6.43 ± 0.30) × 10-3 to (3.23 ± 0.92) × 10-2 s-1. The j(NO2), j(HONO) and flux values for four lamps were more than 4 times that of one lamp only due to the lamp casings being mirrored, and so with four lamps, with two lamps on either side of the flow tube, the casings reflected the light back into the flow tube, increasing the effective light intensity. For these experiments, no gaseous NO2 was added to the gas entering the flow tube. As shown in Fig. 11, for the illumination of pure nitrate aerosols, although a small amount of HONO was observed at higher aerosol loadings, no statistically significant production of HONO was seen.

A second set of experiments were performed with an aqueous solution of titanium dioxide and ammonium nitrate combined in a 1:1 mass ratio to give a TiO2 / NH4NO3 aerosol mixture (5 g NH4NO3 and 5 g TiO2 in 500 mL Milli-Q water) to investigate if the photocatalytic properties of TiO2 facilitate the production of HONO in the presence of nitrate. The RH was decreased to ensure the maximum TiO2 photocatalytic activity (Jeong et al., 2013). A recent study using Raman micro-spectroscopy to observe phase changes in salt particles reported an efflorescence point of pure ammonium nitrate to be between 13.7 %–23.9 % RH (Wu et al., 2019). It is possible therefore that at the RH used in this experiment, ∼ 20 %, the aerosols were still deliquesced. As shown in Fig. 11, the presence of TiO2 in the aerosol mixture showed a significant production of HONO without the presence of NO2, a potentially significant result for the production of HONO in low-NOxenvironments in the presence of mixed dust/nitrate aerosols, for example in oceanic regions off the coast of West Africa or in continental regions impacted by outflow from the Gobi desert. Using the Aerosol Inorganic Model (AIM) (Clegg et al., 1998; Wexler and Clegg, 2002), the nitrate content of the aerosol at both 20 % and 50 % RH was calculated, in accordance with the experimental RH conditions. From this and the aerosol volume distribution given by the SMPS, the [NO3-] within the aerosols could be calculated. The formation of HONO by photolysis of particulate nitrate is given by 10dHONOdt=jpNO3[NO3-] and hence 11HONO=jpNO3NO3-t, where j(pNO3) is the photolysis frequency of nitrate for the lamps used in these experiments and t is the illumination time of the experiment. With knowledge of [HONO], [NO3-] and t=30 s, j(pNO3) can be calculated from a measurement of [HONO] as a function of [NO3-], as shown in Fig. 12, for the mixed nitrate/TiO2 experiment.

When using the four lamps together, the experimental particulate nitrate photolysis rate, j(pNO3), was determined to be (3.29 ± 0.89) × 10-4 s-1 for the mixed nitrate/TiO2 aerosol. From this, it is possible to estimate j(pNO3) for ambient conditions typical of the tropical marine boundary layer. Taking the ratio of the experimental j(HONO) for four lamps ((8.35 ± 0.18) × 10-3 s-1) and the measured j(HONO) from the RHaMBLe campaign held at the Cape Verde Atmospheric Observatory (May–June 2007) (1.2 × 10-3 s-1) (Carpenter et al., 2010; Whalley et al., 2010; Reed et al., 2017) and assuming that j(pNO3) and j(HONO) scale in the same way, ambient j(pNO3) can be determined from 12jpNO3N=jpNO3×1.2×10-3j(HONO), where jpNO3N is the photolysis rate coefficient of particulate nitrate at Cabo Verde, jpNO3 is the experimentally determined photolysis rate coefficient of particulate nitrate to form HONO and j(HONO) is the HONO photolysis rate coefficient calculated from the experimentally determined j(NO2).

Using j(pNO3)= (3.29 ± 0.89) × 10-4 s-1, the rate of HONO production from nitrate photolysis at Cabo Verde was calculated to be jpNO3N=(4.73±1.01)×10-5 s-1 from the mixed nitrate/TiO2 aerosol experiment. Although for pure nitrate aerosol in the absence of TiO2 the data were scattered and the HONO production was small (Fig. 11), an upper limit estimate of jpNO3N=(1.06 ± 1.15) × 10-6 s-1 under conditions at Cabo Verde could be made using Eq. (11), as done for rate of HONO production from mixed nitrate/TiO2 aerosols. The atmospheric implications of this will be considered below.

For the reactive uptake of NO2 onto illuminated TiO2 particles as a function of the initial NO2 mixing ratio, as shown in Fig. 8, a maximum value of γNO2→HONO= (1.26 ± 0.17) × 10-4 was determined at 51 ± 5 ppb NO2 for a photon flux from the lamp of (1.63 ± 0.09) × 1016 photons cm-2 s-1. These experiments were for single-component TiO2 particles, and so for dust aerosols a value of γNO2→HONO= (1.26 ± 0.17) × 10-5 is appropriate, assuming a 10 % fraction of TiO2 and/or other photoactive materials (which behave similarly for HONO production) in mineral dust (Hanisch and Crowley, 2003). Dust aerosols are transported from the Gobi desert to urban areas of China where high NOx and nitrate aerosol concentrations have been observed, and in these areas HONO production facilitated by photocatalysts may be important (Saliba et al., 2014).

Using an average daytime maximum for [NO2], j(NO2) and aerosol surface area measurements for a non-haze period in May–June in 2018 in Beijing, of 50 ppb, 1 × 10-2 s-1 and 2.5 × 10-3 m2 m-3 (of which a maximum of 0.3 % was assumed to be TiO2, though this could be higher in dust impacted events, Schleicher et al., 2010) respectively, a production rate of HONO of 1.70 × 105 molec. cm-3 s-1 (∼ 24.8 ppt h-1) has been estimated using the maximum reactive uptake coefficient measured in this work, γNO2→HONO= (1.26 ± 0.17) × 10-4. The average RH in Beijing during summertime is significantly higher than the range of RH used in the TiO2 aerosol experiments. In previous work (Gustafsson et al., 2006), the NO2 reactive uptake coefficient decreased for relative humidities above those studied here, and hence the HONO production calculated under the conditions in Beijing may represent an upper limit. The lamp used to illuminate the TiO2 aerosols in these experiments gives rise to j(NO2)= (6.43 ± 0.3) × 10-3 s-1, and so γNO2→HONO has been scaled by a factor of 1.55 to match the noon j(NO2) measured in May–June 2018 in Beijing (10-2 s-1), to take into account the relatively small difference in experimental and atmospheric photon flux for Beijing. The HONO production rate estimated here for noontime summer (May–June 2018) in Beijing (∼ 25 ppt h-1) is similar to the value for the maximum production of HONO from urban humic acid aerosol surfaces in Europe, 17 ppt h-1 at 20 ppb NO2 reported by Stemmler et al. (2007). For comparison, the net gaseous production rate of HONO at noon in May–June (2018) Beijing was determined from the measured rate of gas-phase production and losses: 13PHONO=kOH+NOOHNO-(jHONO×HONO+kOH+HONOOHHONO), where kOH+NO= 3.3× 10-11 cm3 molec.-1 s-1 (Atkinson et al., 2004), kOH+HONO = 6 × 10-12 cm3 molec.-1 s-1 (Atkinson et al., 2004) and j(HONO) = 1 × 10-2 s-1 for an average maximum noontime OH concentration of 8 × 106 molec. cm-3 (Whalley et al., 2021), NO concentration of 1.45 ppb (Whalley et al., 2021) and HONO concentration of 0.8 ppb (Whalley et al., 2021).

The net gas-phase production of HONO from Eq. (13) was calculated to be -3.8 ppb h-1 (a net loss) as expected due to HONO loss by photolysis peaking at solar noon, suggesting the production of HONO heterogeneously from TiO2 and NO2 (∼ 25 ppt h-1) would have little effect on the overall HONO budget for Beijing summertime at noon.

Particulate nitrate photolysis could be an important source of HONO in oceanic environments, for example the Atlantic Ocean, where both dust aerosols from the Sahara and high concentrations of mixed nitrate aerosols from sea spray are present, despite low NO2 concentrations (Hanisch and Crowley, 2003; Ye et al., 2017b). From the particulate nitrate photolysis experiments in the absence of NO2 conducted here, a jpNO3N= (4.73 ± 1.01) × 10-5 s-1 was determined in the presence of the TiO2 photocatalysts (Sect. 3.4). Using the experimental j(pNO3), scaled to typical ambient light levels, and a mean noon concentration of nitrate aerosols of 400 ppt measured at Cabo Verde (Reed et al., 2017), taken as an example marine boundary layer environment with a high concentration of mineral dust aerosols, a rate of HONO production from particulate nitrate at Cabo Verde was calculated as 4.65 × 105 molec. cm-3 s-1 (68 ppt h-1). We note that this value would be ∼ 50 times smaller for pure nitrate aerosols. The missing rate of HONO production, i.e. not taken into account by the gas-phase production and loss, Pother, from the Cabo Verde RHaMBLe campaign, can be calculated using the observed HONO concentration, [HONO], and the known gas-phase routes for HONO production and loss: 14Pother=([HONO](jHONO+kOH+HONOOH))-(kOH+NOOHNO)), where kOH+NO= 3.3 × 10-11 cm3 molec.-1 s-1 (Atkinson et al., 2004), kOH+HONO=6×10-12 cm3 molec.-1 s-1 (Atkinson et al., 2004) and j(HONO) = 2 × 10-3 s-1 for average maximum measured concentrations of 1 × 107 molec. cm-3 for OH (Whalley et al., 2010), 5.41 × 107 molec. cm-3 for NO (Whalley et al., 2010) and 1.23 × 108 molec. cm-3 for HONO (Whalley et al., 2010).

Using Eq. (14) this missing HONO production rate for Cabo Verde was 34.6 ppt h-1, which is within a factor of 2 of the rate of HONO production (68 ppt h-1) calculated from nitrate photolysis using our experimental HONO production data for mixed nitrate/TiO2 aerosols. These results provide further evidence that particulate nitrate photolysis in the presence of photocatalytic compounds such as TiO2 found in dust could be significant in closing the HONO budget for this environment (Whalley et al., 2010; Reed et al., 2017; Ye et al., 2017a).