Iron(III)-induced activation of chloride from artificial sea-salt aerosol

Iron-containing aerosols have natural (e.g. mineral dust, glacial flour) or anthropogenic (combustion of fossil fuel, fuel-oil fly ash, metal processing industry, etc.) sources and represent (together with upwelling) the source of dissolved iron in offshore waters.^[1] The deposited dissolved iron, in the range of 0.26 Tg year^–1,^[2] has been suggested as being coupled to atmospheric carbon dioxide (CO₂) concentration by activation of the oceanic food chain (‘Iron hypothesis’).^[3–7] Apart from fertilisation of the oceans, iron is known to act as a catalyst in the photo-Fenton cycle in ferrous (Fe²⁺) and ferric (Fe³⁺) form, and it is utilised in wastewater treatment. Concerning the role of iron in the atmosphere, besides the proposed indirect CO₂ reduction, the interactions between gaseous and aqueous phases containing iron and other organic and inorganic compounds are of interest.

Reactive halogen species (RHS) have a large effect on the budget of ozone (O₃) and nitrogen oxides (NO_x), affecting the oxidation capacity of the atmosphere^[8,9] and interacting with secondary organic aerosols.^[10] In particular, the atomic chlorine radical (Cl) is an important atmospheric oxidant that can considerably influence the lifetime of methane (CH₄).^[11] The main mechanisms responsible for the activation of halides to reactive forms are well investigated, including heterogeneous activation by NO_x and O₃ and autocatalytic halogen activation by HOBr and HOCl.^[12,13] However, the potential role of iron in the release mechanisms of RHS remains unclear and is of special interest in the current work. Whereas a global tropospheric influence is questionable, the photochemistry of iron can significantly contribute to atomic Cl production in the gas phase, at least in some regions. Relevant iron–halide mixtures are present e.g. in coastal regions, intertidal zones, the combustion plumes of ships, volcanic plumes or brine-containing soils. In particular, hypersaline environments, such as the Dead Sea or Australian salt lakes,^[14] offer conditions that promote iron-induced halogen release, as recently investigated by Wittmer et al. for modelled salt surfaces.^[15] The present study, however, concentrates on the effect of Fe³⁺ in saline aerosols that can also be found in these regions and when mineral aerosol particles mix with sea salt. In particular, chloride (Cl^–) enhances the dissolution of iron.^[16] The dominant presence of dissolved Fe³⁺ in saline aerosols and the photoreduction to Fe²⁺ were observed by several studies,^[17–19] but mainly with a focus on the role of the photochemistry of iron in the aqueous phase.^[20] Whereas the photolysis of ferric ions in saline media has been well investigated,^[21–23] less effort has been directed at the implications for the gas phase.

The photolysis of FeCl²⁺ and FeCl₂⁺ directly yields chlorine atoms (Cl^•) in the liquid phase of the humid saline aerosol, and these react very fast with Cl^– to form Cl₂^•– (R1, reaction rate constant k = 2 × 10¹⁰ M^–1 s^–1[21]). For convenience, only aqueous-phase radicals will be marked with a dot in the following. The combination of Cl^• with another Cl₂^•– (R2), or alternatively the combination of two Cl₂^•– (R3) and the dissociation of the produced Cl₃^– thereby (R4) leads to a degassing of Cl₂.^[24]

An alternative, indirect pathway for chloride activation is the photolysis of the slightly less photoactive species FeOH²⁺, producing OH^• radicals in the liquid that again can form Cl^• by ClOH^•–.

A more detailed summary and description of Fe^III-induced Cl₂ formation can be found in Lim et al.^[24] or Wittmer et al.^[15]

In addition to the photosensitive iron-induced halogen release mechanism, further trace gas species, such as O₃, NO_x and sulfur dioxide (SO₂), possibly influence Cl₂ release, mainly when dissolved in the aqueous phase where they can change the pH,^[25] coordinate with Fe²⁺ or Fe³⁺,^[26] scavenge Cl^•[22] or additionally activate halides by the known heterogeneous mechanisms.^[13] Concerning O₃, Sadanaga et al.^[27] observed an enhancement of the O₃ uptake rate and Cl₂ release in the presence of water-soluble Fe³⁺ and O₃ in the dark. When SO₂ is involved, the absorbed amount is highly pH-dependent.^[28] The dissolved SO₂ is mainly in bisulfite and sulfite forms, which are oxidised to sulfate depending on the pH and the availability of catalysts such as Fe³⁺.^[29,30] The presence of sulfate can strongly inhibit the chloride activation process by scavenging Cl^• and OH^• or forming stable complexes with ferrous ions.^[15,22] Apart from iron-influenced systems, polluted air masses considerably influence halogen activation in sea salt,^[8,12,31,32] as well as the environmental oxidation of Cl^– and Br^– by the triplet states of chromophoric dissolved organic matter.^[33] The present study aims at a deeper insight into the effects of Fe^III in saline media with a focus on a better quantification of surface-related gaseous Cl production and the effects of the trace gases mentioned above.

The experiments were conducted in a cylindrical Teflon smog chamber (fluorinated ethylene propylene, FEP 200A, DuPont, Wilmington, DE, USA) with a volume of more than 3500 L (diameter: 1.33 m, height: 2.5 m, surface/volume ratio: 3.8 m^–1). Here, the chamber and its analytical instrumentation are only briefly introduced. More detailed specifications can be found elsewhere.^[15,31,34]

The chamber was suspended above a solar simulator that generates a mean actinic flux comparable with the summer sun at 50° latitude. The chamber was equipped with a differential pressure sensor (Kalinsky Elektronik DS1, Kalinsky Sensor Elektronik GmbH & Co. KG, Erfurt, Germany) to monitor the slight overpressure of 0.6–1 Pa, which was controlled by a continuous flow of 5–6 L min^–1 of hydrocarbon-free, humidified zero air (zero-air generator, cmc instruments, <1 ppb of O₃, <500 ppt NO_x, <100 ppb of CH₄) for all experiments. The temperature was adjusted to 20 °C and monitored by two light-shielded sensors (Rotronic, HC2-IC102) at different heights (bottom and top) to observe and avoid possible thermic layering. A Teflon fan inside the chamber assured constant mixing and kept temperature gradients below 1 °C. NO, NO_x and O₃ were monitored by chemiluminescence gas analysers (EcoPhysics, CLD 88p (EcoPhysics GmbH, Dürnten, Switzerland) coupled with a photolytic converter, PLC 860, for NO and NO_x, and UPK 8001 (Bendix, Bad Nauheim, Germany) for O₃).

Gas-phase Cl, Br and OH were indirectly quantified by the radical clock method,^[35] monitoring the consumption of selected hydrocarbons (2,2-dimethylpropane, DMP, Linde AG, Munich, Germany, ≥99 %; 2,2-dimethylbutane, DMB, Sigma–Aldrich Chemie GmbH, Taufkirchen, Germany, ≥99 %; 2,2,4-trimethylpentane, TMP, Janssen-Cilag GmbH, Neuss, Germany, ≥99 %; toluene, Tol, Aldrich ≥99.9 %) and n-perfluorohexane (PFH, Aldrich, ≥99 %) as inert standard. The concentrations of the hydrocarbons was measured at time intervals of 15 min by gas chromatography (GC, Siemens Sichromat 2, 50-m Al₂O₃-PLOT column, 0.25 mL min^–1 He as carrier gas) with a flame ionisation detector (FID), custom-built liquid nitrogen cryo-trap enrichment in glass-lined stainless-steel tubing and a Nafion tube counterflushed from outside with zero air to dry the sampling flux.^[15]

An artificial seawater stock solution with a molar Cl^–/Br^– ratio of 997 was prepared according to Kester et al.^[36] The mixture included NaCl (23.9 g L^–1), Na₂SO₄·10H₂O (9.1 g L^–1), KCl (0.68 g L^–1), NaHCO₃ (0.2 g L^–1), KBr (0.098 g L^–1), H₃BO₃ (0.026 g L^–1), NaF (0.003 g L^–1), MgCl₂ (5.1 g L^–1), and CaCl₂·2H₂O (1.5 g L^–1). Depending on the experiment, a specific amount of FeCl₃ was dissolved in the stock solution and diluted afterwards (1 : 28) to obtain a Cl^– concentration of 15–35 mmol L^–1 (Table 1) that provides optimal aerosol size distributions for long residence times of the suspended droplets. The aerosol production from the prepared solutions and the subsequent measurement are described in the following section. For the reference experiment with Fe^III-doped NaCl, the FeCl₃ was directly added to an 18 mmol L^–1 NaCl solution. The added amount of FeCl₃ and the resulting Cl^–/Fe^III ratios are listed in Table 1. More details of the prepared artificial seawater are included in the Supplementary material (Table S1).

Table 1.  Compositions of the stock solutions obtained by adding FeCl₃ to the artificial seawater (prepared according to Kester et al.^[36]) or to a 1 g L^–1 NaCl solution and resulting molar Cl^–/Fe^III ratios

On the basis of the different sea-salt mixtures, the effect of various gas species was tested by adding O₃ from a silent ozoniser (Sorbios GSG 12, Sorbios GmbH, Berlin, Germany) with an electrical discharge applied to pure oxygen (Rießner-Gase GmbH, Lichtenfels, Germany, >99.995 %), NO₂ from a gas cylinder (Rießner Gase, 104 vpm (volume parts per million) NO₂ with a purity of 98 % in synthetic air), or SO₂ from a gas cylinder (Rießner Gase, 0.99 % SO₂ with a purity of 99.98 % in N₂ with a purity of 99.999 %).

The sea-salt particles were generated by nebulising the prepared solutions (Table 1) with an ultrasonic nebuliser (Quick Ohm QUV-HEV FT25/16-A, 35 W, 1.63 MHz, Quick-Ohm Küpper & Co. GmbH, Wuppertal, Germany) at a starting relative humidity (RH) of ≥40 % to avoid crystallisation.^[12] During injection (taking typically 30–50 min), the RH increased to 70–90 %. Thus, we can assume that the droplets equilibrate to form a saturated solution, leading to most abundant particle diameters between 350 and 460 nm based on the concentration of the stock solution and the final RH after injection.^[37] The size distributions were determined by an electrostatic classifier (TSI, 3071, TSI GmbH, Aachen, Germany) in combination with a neutraliser (⁸⁵Kr) and a condensation nucleus counter (TSI, 3020). The scanning method and the multiple charge correction were applied with custom-written software.^[38] Typical size distributions, determined after 30–50 min of injection, are shown in Fig. 1 for different stock compositions. To avoid the condensation of the sea salt within the transfer line (~6.4-mm tube made of copper) and thus the dripping of solution inside the chamber, the tube was additionally heated by a heating wire to ~60–70 °C. During aerosol generation, the NO_x signal rose typically up to 3–5 ppb, probably originating from the sonochemical formation of nitrate and nitrite^[39] and including a possible cross-sensitivity of the NO_x analyser to nitrous acid (HONO).^[40]

Fig. 1.  Typical initial (multiple-charge-corrected) size distributions for aerosols generated by the ultrasonic nebuliser (30–50-min injection time) from various diluted stock solutions containing NaCl, FeCl₃ or artificial seawater. N, number density; d_P, particle diameter.

The chamber walls were either cleaned with deionised water (Seralpur Pro 90 CN, <0.055 µS cm^–1, Seral, Ransbach-Baumbach, Germany), or the total FEP-Teflon film was exchanged after the experiment (depending on the risk of contamination expected for subsequent runs). To minimise the outgassing of gas-phase products,^[41] every new chamber was conditioned by adding O₃ at an RH of 50–80 % and irradiating for at least 4 h with the solar simulator and an additional UV lamp (Phillips TUV 55 W, λ = 253.7 nm).

After every experiment, the particles were collected by a Sioutas cascade impactor (SKC) in the aerodynamic diameter ranges of >2.5, 1–2.5, 0.5–1, 0.25–0.5 and <0.25 µm^[42] on aluminium foil as a collection substrate. The impactor samples were analysed using an FEI Quanta 200 scanning-electron microscope (SEM), equipped with an energy-dispersive X-ray (EDX) detector for imaging (Octane Pro Silicon Drift (SDD) EDX detector from AMETEK GmbH, Wiesbaden, Germany). The hyperspectral data-cube from the SEM-EDX measurements was analysed using the software package Imagelab (Epina Software Laboratories, www.imagelab.at, accessed April 2015).^[43]

The gas chromatograms were evaluated the same way as described before^[15] to quantify the quasi-stationary concentrations of Cl_qs, OH_qs, and Br_qs in the gas phase, based on the measured, dilution-corrected and smoothed hydrocarbon (HC_i) profiles, resulting in a system of four differential equations with three unknown variables according to:

where k_X,i is the reaction rate constant of HC_i towards the radical X (X = Cl, Br, and OH; Table S2, Supplementary material). The corresponding total production (Q_X) and production rates (dQ_X/dt) are obtained by equalising sources and sinks in a photostationary steady-state:

Considering a constant total reactivity of the chamber contents towards [X] on the one hand (Σ_ik_X,i[HC_i]₀) and only the reactivity of the measured hydrocarbons on the other hand (Σ_ik_X,i[HC_i]_t), which means either assuming that reaction products have the same reactivities as the HC_i or neglecting the reactivities of the reaction products totally, we obtain an estimate of the minimum and maximum Q_X and dQ_X/dt. These assumptions are reasonably valid for Cl atoms but less reliable for Br atoms and OH radicals, where the reactivities of reaction products may exceed the reactivities of the original hydrocarbons by far. However, they are valid at the very beginning of the irradiation, where the consumption of the hydrocarbons is still low.

The measurements of the particle size distributions (number density N, cm^–3) allow us to determine the available reaction surface area (A, m² m^–3) and the particle volume (V_particle, m³ m^–3), which equals the liquid water content (LWC) plus the tare volume of the ions (radius, r_Cl^– = 181 pm, r_SO4^2– = 230 pm, r_Na⁺ = 102 pm, r_K⁺ = 138 pm, r_Mg²⁺ = 72 pm, r_Ca²⁺ = 100 pm, r_Fe³⁺ = 65 pm)^[44] at the adjusted RH. The measured LWC was corrected by considering the main ions as spherical. The contribution of deposited particles on the chamber walls to the activated Cl was considered by measuring the Cl activation of a totally wall-deposited aerosol load (Chamber wall effects section). The Cl^– production terms are normalised to obtain absolute production rates, dQ_abs/dt, and absolute total production Q_abs cm^–2 of the aerosol surface. Therefore, the production rate (atoms cm^–3 s^–1) is multiplied by the chamber volume (V_chamber) to obtain the total number of atoms produced per second. The result is divided by the actual active aerosol surface, which we define as the sum of (i) the measured aerosol surface when the lights were turned on (A_0,light) and its deposition rate, exp(–t_light/τ_s), with the aerosol surface lifetime τ_S; (ii) the active, deposited surface during the time of injection; and (iii) the active, deposited surface formed after the injection is finished (Eqn 3). Whereas (ii) is determined by assuming an approximately linear increase of the aerosol surface during injection and calculating the respective deposition until the injection is stopped (t_inj,end), (iii) is based on the measured aerosol surface area directly after injection (A_0,inj) and its deposition during the time t_inj, which starts when the injection ends. Both terms are multiplied by a factor of 0.2, which is the fraction of deposited surface area that contributes to the halogen activation (Chamber wall effects section).

Table 2 presents an overview of the experiments performed, including Cl^–/Fe^III ratios, pH of the diluted stock solution, gas-phase composition, initial reactivity of HCs against Cl, OH, and Br, initial aerosol surface, resulting quasi-stationary concentrations and absolute production. Based on the two-to-five orders of magnitude lower reactivity of the applied HCs against Br and OH compared with Cl, depletion was dominated by the reaction with Cl, and no significant interpretation for Br and OH was possible in most cases. Before adding FeCl₃ to the samples, several Fe^III-free blank experiments were performed, including pure NaCl aerosol, artificial sea-salt aerosol and artificial sea-salt aerosol with O₃ and NO₂ addition respectively (see Table 2). Except for the artificial sea-salt aerosol with addition of 700 ppb O₃, Cl production was hardly detectable. Generally, no indication of Cl, Br or OH production was observed during the dark phases (aerosol, NO₂, O₃, SO₂ injections and waiting periods), based on the constant HC time profiles.

Table 2.  Molar Cl^–/Fe^III ratio, pH of the diluted, nebulised stock solution, gas-phase composition within the smog chamber, total initial reactivity of the hydrocarbons in the chamber against Cl, OH and B, total aerosol surface measured directly after injection, quasi-stationary Cl_qs, Br_qs and OH_qs concentrations during irradiation during the first hour, and resulting total Cl and Br production per square centimetre aerosol surface
n.d., not detected

Owing to flashovers in the classifier throughout some experiments caused by the high RH, only the measurements at the beginning of those experiments were evaluable. Therefore, the measurements after injection and when the lights were turned on were used to calculate the dilution-corrected loss by deposition, based on the mean dilution-corrected lifetimes τ_N,S (N, number density; S, total aerosol surface area) over the experiments without flashovers and neglecting the loss by coagulation. The deposition velocity depends on the salt concentration in the nebulised solution and thus on the mean particle diameter after injection. For instance, the artificial sea-salt solution and the pure NaCl solution gave lifetimes of τ_N = 24 800 s (τ_S = 30 100 s) and τ_N = 25 000 s (τ_S = 31 100 s). For the higher-concentration Fe^III-doped artificial seawater (29–35 mmol Cl^– L^–1), the most abundant particle diameters were between 430 and 460 nm, resulting in a faster sedimentation with τ_N = 4590 ± 240 s and τ_S = 6070 ± 520 s compared with τ_N = 6845 s and τ_S = 8820 s for the lower-concentration pure NaCl + FeCl₃ solution (24 mmol Cl^– L^–1) with most abundant particle diameters between 390 and 420 nm. The respective contour plots can be found in the Supplementary material (Fig. S1).

To quantify the particle deposition and its contribution to the active surface area (and thus to Cl activation), a test measurement was performed to determine the fraction of Cl release by the active wall surface compared with the active aerosol surface: the iron-doped artificial seawater sample (Cl^–/Fe^III = 13) was injected and allowed to deposit totally for 17 h (<0.005 % of the surface area should have remained suspended) while keeping the RH at 80 %. Then the ‘aerosol-free’ chamber was irradiated, resulting in Cl production that was 20 ± 4 % compared with the actual production measured for the same sample in an aerosol experiment (Iron(III)-catalysed Cl atom production section). This production was evaluated by taking the mean of the quotient of each total production (deposited and not deposited) normalised by the respective initial LWC directly after injection. In Eqn 3, the contribution of deposited, active aerosol surface area is accounted for by adding 20 % of the deposited surface area since the time of injection to the surface area when the lights were turned on (corrected for deposition). The smaller active fraction when deposited on the wall can be explained by a physical surface reduction during the adhesion process of the droplets (possibly coagulating to larger droplets on the hydrophobic surface) and by an inhomogeneous irradiation of the chamber walls (especially the large fraction of wall surface perpendicular to the solar simulator), but also by a drying effect due to the heating of the Teflon walls during irradiation. Fig. 2 demonstrates the contribution of active, deposited surface on the wall to the total active surface area during the experiment with Fe^III-doped, pH-adjusted (pH 2.1–2.3) artificial seawater (Cl^–/Fe^III = 101). The figure additionally includes dQ_abs/dt and Q_abs, which are discussed in the Effect of pH and Fe^III speciation chemistry section.

Fig. 2.  Time profile of the sum of the total measured aerosol surface area and the active deposited surface during the experiment with Fe^III-doped, pH-adjusted (pH = 2.1–2.3) artificial seawater (Cl^–/Fe^III = 101) and the corresponding absolute gaseous mean Cl production rate (dQ_abs/dt) (atoms cm^–2 s^–1) and time-integrated total Cl production Q_abs (atoms cm^–2) of active aerosol surface. The light was switched on at 0 min.

Significant Cl activation was detected for every sample with Cl^–/Fe^III < 101 (Table 2), whereas Cl production was close to the detection limit during the blank experiments including Fe^III-free NaCl and artificial sea-salt aerosol in zero air. The effect of the constituents of the Fe^III-containing artificial seawater mixture was investigated by comparing the Cl release with a pure NaCl + FeCl₃ mixture. Both samples were prepared with a molar ratio of Cl^–/Fe^III = 13. Both experiments were conducted in the same manner and under the same conditions (21 ± 1 °C and 72 ± 2 % RH). After 45 min of aerosol injection and a 15-min waiting period, the light was turned on. The aerosol size distributions (Fig. 1) showed small differences, with a lower diameter for the most abundant particles but only a slightly lower surface-to-volume ratio for the artificial sea salt/Fe^III mixture (1.29 × 10⁷ m^–1) compared with the NaCl/Fe^III mixture (1.31 × 10⁷ m^–1) caused by the similar hydration behaviour at the given RH (Köhler theory).^[45] The Cl production rates and corresponding total production are shown in Fig. 3. Slightly higher Cl production of the pure NaCl/Fe^III mixture was observed, which was probably caused by the speciation chemistry (as photolabile Fe^III-Cl complexes can form more easily in the absence of competing ions such as SO₄^2–).^[46] Considering the diluted stock solutions, the pH was between 1.9 and 2.2 for both experiments. The buffer effect of the added bicarbonate (0.2 g L^–1) in the artificial seawater is very low, considering the high amount of FeCl₃ added (13 g L^–1). Although the radical clock method is also able to quantify Br and OH, the depletion of HCs was dominated by Cl with its much higher reaction rate constants (Table S2, Supplementary material), which prevents significant interpretation of Br and OH (in most cases), which were close to the detection limit (Br ~10⁹ molecules cm^–3; OH ~10⁶ molecules cm^–3).

Fig. 3.  Absolute gaseous mean Cl production rate (dQ_abs/dt) (atoms cm^–2 s^–1) and time-integrated total minimum and maximum Cl production Q_abs (atoms cm^–2) of aerosol surface during the experiments with Fe^III-doped artificial seawater and NaCl with Cl^–/Fe^III = 13. The error bars for dQ_abs/dt include the minimum and maximum production rates and the respective statistical uncertainty. The light was switched on at 0 min.

In order to determine the dependence of the chlorine activation on the Cl^–/Fe^III ratio in the aerosol, a series of experiments with a Cl^–/Fe^III ratio of (i) 13, (ii) 53, (iii) 101, (iv) 955 and (v) a blank without iron addition was conducted. Whereas no significant Cl concentration was detected for run (iv) and (v), (iii) resulted in a total Cl production of (0.9–1.0) × 10²¹ atoms cm^–2, run (ii) of (2.3–2.6) × 10²¹ atoms cm^–2, and run (i) of (1.1–1.8) × 10²² atoms cm^–2 in the first 100 min (Fig. 4). This demonstrates a continuous increase in produced Cl with increasing fraction of Fe^III in the salt. A 1.9-times higher fraction of Fe^III results in a 2–3-times higher Cl production (comparing (iii) and (ii)), and 4.1 times more Fe^III results in 5–7-times higher Cl production (comparing (ii) and (i)). The higher amount of added FeCl₃ cannot only explain the disproportionately higher Cl activation. In fact, the shift in pH from adding 10 times more FeCl₃ (pH 4 → 2) is in addition responsible for a higher fraction of photolabile Fe^III-Cl complexes (discussed in the Effect of pH and Fe^III speciation chemistry section). A slight decrease was observed for (iii), indicating exhaustion of the Cl source, whereas samples (i) and (ii) resulted in a more stable Cl production rate.

Fig. 4.  Absolute gaseous mean Cl production rate (dQ_abs/dt) (atoms cm^–2 s^–1) and time-integrated total minimum and maximum Cl production Q_abs (atoms cm^–2) of aerosol surface during the experiments with Fe^III-doped artificial seawater at various Fe^III concentrations: Cl^–/Fe^III = 13 (black), 53 (magenta), 101 (red). The light was switched on at 0 min.

Experiments on the typical tropospheric traces and pollutants NO₂, O₃ and SO₂ were performed by adding 20 ppb of NO₂, 630 ppb of O₃, and 20 and 200 ppb of SO₂. These levels were chosen with the aim to better observe their effects and to obtain aerosol to pollutant ratios approximately comparable with natural conditions (discussed in the Environmental significance section). Figs 5 and 6 show the time profiles of the trace gases (corrected for dilution) combined with the results on Cl_qs, dQ_abs/dt and Q_abs obtained by the radical clock for the experiments with Fe^III-doped artificial seawater (Cl^–/Fe^III = 13) and addition of NO₂ and O₃ respectively. After the aerosol injection, the addition of NO₂ and a waiting period of 55 min (12 min in case of the O₃ experiment), the solar simulator was turned on (indicated by 0 min in the figures). For both experiments, significantly elevated Cl production was observed compared with the experiments in zero air. The higher Cl source is related to the additional activation mechanisms in the presence of O₃ and NO_x,^[13] which are discussed in the following text. Fast formation of O₃ was observed in the NO_x experiment when the light was turned on, caused by the well-known photochemical cycle of NO, NO₂ and O₃ (photolysis rate coefficients: J_NO2 = 6.7 × 10^–3 s^–1, J_O3 = 2 × 10^–4 s^–1) on the one hand and the re-oxidation of NO to NO₂ and OH by HO₂ (formed by the depletion of the injected HC_i) on the other hand. Therefore, the O₃ concentration exceeded the NO_x concentration in the course of the experiment, and the O₃ formation stopped when NO was depleted. Because O₃ is not present during the dark period, only negligible amounts of dark-phase nitrogen oxides like dinitrogen pentoxide (N₂O₅) could be present at the beginning of the irradiation and potentially explain the additional Cl source by the proposed heterogeneous activation mechanisms.^{[12,13,32,47]} During irradiation, the presence of O atoms and O₃ formation can cause the formation of NO₃ and thus N₂O₅, although NO₃ is extremely short-lived under our conditions (J_NO3 = 0.11 s^–1). At an NO_x concentration of 20 ppb, the proposed formation and photolysis of nitrosyl chloride (ClNO) by NO₂ uptake of the sea-salt aerosol^[48–50] is too slow to contribute significantly to gaseous Cl production. For instance, Karlsson and Ljungström^[51] estimated a total ClNO production of the order of 1.1 × 10⁷ atoms cm^–3 in a flow reactor at aerosol concentrations of 3 × 10⁵ cm^–3 (surface/volume = 2 × 10⁷ m^–1) when 50 ppb NO₂ was present. Owing to a photolysis rate of J_ClNO = 1.8 × 10^–3 s^–1 in our chamber, this would result in comparable Cl production, whereas the actual total production in our experiments is of the order of 10¹²–10¹³ atoms cm^–3 and thus exceeds the production caused by ClNO formation by far, even if Karlsson and Ljungström^[51] assume a sea-salt surface three orders of magnitude lower. The negligible influence of NO₂- or N₂O₅-induced Cl activation was confirmed by a blank experiment (iron-free artificial sea salt with 20 ppb NO₂ addition), where the artificial sea-salt aerosol without Fe^III was irradiated in the presence of 20 ppb NO_x and the quasi-stationary Cl concentration remained below 10⁴ molecules cm^–3 combined with a slower NO_x depletion (Supplementary material, Fig. S2). Rather, the formation of XONO₂ (X = Cl, Br) by XO (R7, R8), or HNO₃ by OH (R9) and the subsequent uptake by the aerosol may explain the NO_x loss,^[12,31,52] which predominantly ends up as nitrate (with a possible back-reaction pathway to N₂O₅ or XNO₂ by HNO₃ at low pH).^[13]

Fig. 5.  Time profiles of NO_x, NO and O₃ (upper box) and of the quasi-stationary Cl concentration (Cl_qs), the absolute gaseous mean Cl production rate (dQ_abs/dt) (atoms cm^–2 s^–1) and the time-integrated total minimum and maximum Cl production Q_abs (atoms cm^–2) of aerosol surface (lower box) during the experiment with Fe^III-doped artificial seawater (Cl^–/Fe^III = 13) and 20 ppb NO_x. The light was switched on at 0 min.

Fig. 6.  Time profiles of NO_x, NO, and O₃ (upper box) and of the quasi-stationary Cl concentrations (Cl_qs), the absolute gaseous mean Cl production rate (dQ_abs/dt) (atoms cm^–2 s^–1) and of the time-integrated total minimum and maximum Cl production Q_abs (atoms cm^–2) of aerosol surface (lower box) during the experiment with (a) Fe^III-doped artificial seawater (Cl^–/Fe^III = 13) and 630 ppb O₃; and (b) a corresponding blank experiment with Fe^III-free artificial seawater and 700 ppb O₃. The light was switched on at 0 min.

The uptake of XONO₂ can result in additional X₂ or BrCl release, as e.g. summarised in the review by Rossi.^[13] The uptake of HNO₃ leads to an acidification of the aerosol and contributes to the reoxidation of Fe^II to Fe^III through photolysis and OH^• formation in the aqueous phase^[20] and thus is potentially responsible for the enhanced gaseous Cl production, according to the mechanisms described in the Introduction. Additionally, the uptake of HNO₃ leads to acid displacement with a subsequent release of HCl,^[53,54] whose reaction with OH is supposed to be a main Cl–atom source in the marine troposphere.^[55] In general, a low pH facilitates the release of HCl and other gaseous Cl precursors.^{[25,31,56,57]} For example, Keene and Savoie^[56] determined a gaseous HCl mixing ratio in the range of 0.1 ppb, and Keene et al.^[25] modelled HCl outgassing of 1 ppb day^–1 for acidified sea salt (pH 3). Transferred to our conditions with typical OH concentrations of 10⁶–10⁷ molecules cm^–3 and a much higher LWC, this leads to gaseous Cl production rates in the range of 10⁹–10¹⁰ atoms cm^–3 h^–1 and thus potentially contributes less than 10 % to the observed Cl production. This accounts also for the HCl formation caused by hydrogen abstraction during the consumption of the injected HC_i by Cl atoms. Furthermore, Zetzsch and Behnke^[58] investigated photochemical Cl production rates from 200 to 500 ppb O₃ and 300 ppb HCl in the presence of NaCl, Fe₂O₃ and SiO₂ aerosol. They concluded that the heterogeneous Cl^– activation exceeds the Cl source from the gas-phase reaction of OH and HCl by far.

Several effects were observed in the experiment where 630 ppb O₃ was added (Fig. 6). The total Cl production per hour is three to five times higher than the value obtained from the Fe^III-containing sample in zero air and is similar to the NO_x experiment (Table 2). However, the reactivity of O₃ towards Cl (~180 s^–1) is comparable with the total reactivity of the injected HC_i (~200 s^–1) and not considered in the calculation of the total production Q, as it does not represent a final sink but rather initiates a reaction cycle by HO₂ and HOCl where finally Cl₂ is produced.^[59,60] In addition, Sadanaga et al.^[27] observed enhanced O₃ uptake rates in the presence of water-soluble Fe^III in synthetic sea salt without irradiation. During the short dark period, O₃ depletion was hardly detectable, whereas we observed an approximately four times lower O₃ lifetime for the experiment with added Fe^III (~10⁴ s) compared with the pure artificial sea-salt sample (~4 × 10⁴). In general, the O₃ destruction is related to autocatalytic halogen activation where Br especially plays a dominant role.^[34] This difference possibly explains the much higher Cl and Br production (Fig. 6a, b) at the same level of quasi-stationary OH concentrations (~10⁷ molecules cm^–3) for both experiments. Concerning Br production, the quasi-stationary Br concentration was again of the order of 10⁹ atoms cm^–3 (close to the detection limit) and resulted in Q_abs values of (1–3) × 10²¹ atoms cm^–2 h^–1 for the iron-free and (3–8) × 10²¹ atoms cm^–2 h^–1 for the iron-doped sea salt. Adding O₃ to the iron-free aerosol, the significant enhancement in Cl production (~3–4.5 × 10²¹ atoms cm^–2 h^–1 from the iron-free sample in zero air to the iron-free sample with O₃ addition) is relatively low compared with the much stronger increase in Cl production when adding O₃ to the iron-containing sample (1–4 × 10²² atoms cm^–2 h^–1 from the iron-containing sample in zero air to the iron-containing sample with O₃ addition). This indicates that we observed an O₃-induced Cl and Br production that increases with addition of Fe^III (or decreasing aerosol pH) and contributes markedly to the enhanced iron-induced activation. Several mechanisms come into consideration for the observed effects. At these O₃ levels, a main mechanism (which is responsible for the high Cl and Br production and fast O₃ depletion) is the so-called ‘Bromine explosion’^[61] with the net reaction:

and the associated formation of OClO and ClO.^[34] However, it does not sufficiently explain the increase in production rates from ~2 × 10¹⁸ to ~10¹⁹ atoms cm^–2 s^–1 with O₃ addition when Fe^III is involved (Figs 3 and 6a), because the pH (when <5) is supposed to have no large effect on the mechanism,^[25,62] and the additional Cl production rate (when estimated from the iron-free experiment) is ~10¹⁸ atoms cm^–2 s^–1 (Fig. 6b).

A further approach is the enhanced formation of H₂O₂ in the gas phase (R11, R12):

which enters the aqueous phase rapidly and oxidises Fe^II back to Fe^III, and is able to form HOCl^– or HOBr^–, which further dissociate and finally form Cl₂ or Br₂.^[63]

Knipping et al.^[64] even report a direct uptake of OH and subsequent Cl₂ release from NaCl aerosols. Furthermore, NO_x (3–5 ppb is present after aerosol injection) forms NO₃ and N₂O₅ by O₃ during the dark phase, indicated by the loss of NO_x when O₃ is injected (Fig. 6a, b), and thus leads to HNO₃ formation in the aqueous phase with similar consequences to those described above (R9). NO₃ is able to efficiently produce atomic X from solid and humidified salts.^[65,66] Moreover, the uptake of N₂O₅ activates Cl^– by releasing ClNO₂^[47,67] (or even Cl₂ in a multistage process by ClNO₂ at acidic pH)^[68] and Br^– by releasing Br₂ or BrNO₂,^[69,70] which is photolysed to atomic X (J_ClNO2 = 0.2 × 10^–3 s^–1, J_BrNO2 = 3.7 × 10^–3 s^–1, J_Br2 = 17 × 10^–3 s^–1). However, considering the low NO_x concentrations, the dark-phase activation mechanisms by NO₃ and N₂O₅ probably play an only minor role compared with autocatalytic light-induced activation. All given mechanisms favour the activation of bromide and could explain the high Br production rates in both O₃ experiments.

In contrast to the reinforcing effects of NO_x and O₃, the situation changes with SO₂. Two experiments were conducted with SO₂ concentrations of ~20 and ~200 ppb and a waiting period of 45 min before turning the solar simulator on. Similarly to the effect of sulfate in a salt pan,^[15] SO₂ inhibited Cl production, because slightly lower dQ_abs/dt and Q_abs were observed (Fig. 7). Several studies have already examined the role and the uptake of SO₂ in combination with Fe^III and Fe^II for seawater. For example, Hoppel et al.^[28] report an uptake of 0.21–1.2 mmol L^–1 of nebulised seawater, which is in a similar range to the sulfate concentration of 1 mmol L^–1 in the artificial seawater stock solution (which is nebulised in our case). However, we used a 29-times diluted stock solution to obtain size distributions with a maximum diameter of 400–450 nm. Additionally, the dissolution of SO₂ depends on further dissociation reactions in the aqueous phase that depend on pH, temperature and ion content.^[71] Therefore, the uptake rate of Hoppel et al. is only transferable with caution, but considering the less-concentrated stock solution, it should only be a small amount of SO₂ that dissolves in the aqueous phase of the aerosol, mainly in the form of bisulfite (HSO₃^–) given the pH range (2–6) of our samples. Especially at a low pH, where a high portion of dissolved Fe³⁺ is available, the oxidation of sulfite to sulfate is catalysed by Fe³⁺.^[29] Moreover, a significant oxidation path is the reaction of sulfite with HOCl and HOBr, both of which can intervene in the autocatalytic halogen release.^[72,73] Although sulfate can strongly inhibit halogen activation, the effect of the freshly dissolved and oxidised SO₂ is probably very low with respect to the amount of sulfur that is already present in the artificial sea salt (Sample preparation and chemicals used section). A further influence of SO₂ is observed, concerning the quasi-stationary OH concentrations that are close to the detection limit of 10⁶ molecules cm^–3 for both experiments, due to the oxidation of SO₂ in the gas phase (SO₂ + OH → HOSO₂; k = 1.3 × 10^–12 cm³ molecules^–1 s^–1),^[74] which competes with Cl production by HCl + OH (k = 7.8 × 10^–13 cm³ molecules^–1 s^–1).^[75] Therefore, slightly lower Cl production is observed for the SO₂ experiments. In the extreme case of 200 ppb, Cl production decreases to ~70 % of that from the SO₂-free experiment.

Fig. 7.  Absolute gaseous mean Cl production rate (dQ_abs/dt) (atoms cm^–2 s^–1) and time-integrated total minimum and maximum Cl production Q_abs (atoms cm^–2) of aerosol surface during experiments with Fe^III-doped artificial seawater (Cl^–/Fe^III = 13) and different SO₂ mixing ratios in the chamber (~20 ppb, red; ~200 ppb, black). The light was switched on at 0 min.

To estimate the effect of the aerosol pH on Cl production, the pH of the diluted stock solution (Cl^–/Fe^III = 101) was adjusted to pH 2.1–2.3 by gently adding 32 % HCl. The results demonstrate the immense effect of the pH (Fig. 2). The higher absolute production rate during the low-pH experiment led to eight times higher total Cl production per hour compared with the unadjusted sample (pH 3.9–4.2, Fig. 4; see also Table 2). There are multiple reasons for this effect. As evaluated in several studies for Fe^III-doped salts, the speciation chemistry of Fe^III strongly depends on the pH and ionic strength.^[15,24,76] Fig. 8a, b show the portions of the total Fe^III as Fe^III complexes as a function of the pH, calculated with an equilibrium model in PHREEQC.^[77] According to Wittmer et al., the model is based on the MINTEQ database,^[78] and the activity coefficients are corrected by the Pitzer ion interaction approach^[79] with the parameters listed in Tosca et al.^[80] The main equilibrium constants involving Fe^III are listed in the Supplementary material (Table S3). For some ions (e.g. F^–), there are no Pitzer parameters available and the extended Debye–Hueckel equation^[81,82] was applied to calculate the respective activity coefficients. To get an insight into the processes during nebulisation and evaporation, the speciation was calculated for the diluted stock solution (Fig. 8a), which the nebuliser was filled with, as well as for the final aerosol (Fig. 8b), assuming a saturation in Cl^– (6.1 mol L^–1) and unchanged molar ratios towards the other constituents. This assumption is extremely simplified based on the complexity of a multicomponent salt system,^[45] and implies a similar hydration and solubility behaviour of NaCl and artificial sea salt but accounts for the fact that Na⁺ and Cl^– represent by far the main ions. For example, McCaffrey et al.^[83] measured similar Cl^– saturation concentrations in evaporating seawater whereas the concentration may change with increasing Fe^III content.

Fig. 8.  Molar fraction of the formed Fe^III species relative to the total Fe^III content in Fe^III-doped artificial seawater (Cl^–/Fe^III = 101) as a function of pH (according to the PHREEQC model) of the low-concentration diluted stock solution before nebulising (a); and of the final, highly concentrated aqueous phase of the aerosol (b). The dashed line indicates a molar fraction of 1 (100 %).

Whereas high fractions of Fe³⁺ ions and Fe^III-hydroxy complexes are present in the stock solution, the decisive photosensitive Fe^III-Cl complexes start to form as a considerable fraction at pH 1–4.5 with increasing ionic strength (solvent concentrations). The pH of 4.5 is a turning point for the aerosol speciation, where mainly Fe^III-hydroxy complexes are present. Because Fe^III-Cl complexes comprise a fraction of less than 5 % in the stock solution at the given pH of 3.9–4.2 (and it takes some time until speciation equilibrium is reached during the transition from the low-concentration stock solution to the high-concentration liquid aerosol), the smaller amount of Fe^III-Cl complexes could explain the lower gaseous Cl production for the untreated sample. This situation changes when the pH is adjusted to 1.9–2.2 already in the diluted stock solution. At such low pH values, Fe^III-Cl complexes are formed much more easily. Furthermore, the solubility of Fe^III increases with decreasing pH, whereas more dissolved Fe^III becomes available on an absolute scale.^[46,84] Also, the formation of the highly soluble hypochlorous acid (HOCl; Henry’s law constant at 298 K: 2.6–9.3 × 10² M atm^–1)^[85] predominates at a pH between 4 and 7 over Cl₂, which is favoured at a pH lower than 4 and is much more easily released into the gas phase (Henry’s law constant: 6.2–9.1 × 10^–2 M atm^–1).^[85] Similar observations were also made by Lim et al.,^[24] who ascribed the decreasing Cl₂ source with decreasing pH to the speciation chemistry and solubility of the various chlorine species.

Aerosol particles originating from the pure artificial sea water are mainly composed of NaCl, CaSO₄, MgCl₂ and KCl (Fig. 9a). Whereas NaCl and CaSO₄ particles remain as single crystals, KCl and MgCl₂ exhibit some phase-mixing and a more amorphous structure. In the case of the FeCl₃-containing artificial seawater, single NaCl and CaSO₄ crystals are also visible. Additionally, FeCl₃ is visible in the KCl–MgCl₂ phase (Fig. 9b). FeCl₃ appears to cover nearly all other particles like NaCl and CaSO₄, thus indicating that all aerosol particles originating from the FeCl₃-enriched artificial seawater are covered by an active Fe^III chloride–salt layer.

Fig. 9.  Scanning electron microscope­–energy dispersive X-ray (SEM-EDX) images of the aerosol particles of the pure sea-water (a), and FeCl₃-containing sea-water (b) with NaCl (blue), CaSO₄ (green), MgCl₂ (yellow), KCl (red) and FeCl₃ (turquoise).

Multiplying the initial LWC₀ (dm³ cm^–3) with the assumed saturation concentration of Cl^– (6.1 mol L^–1) in the aerosols produced and with Avogadro’s constant (N_A = 6 × 10²³ mol^–1) and dividing the molar Cl^–/Fe^III ratio (R_Cl/Fe) in the stock solution yields the concentration of Fe^III in the chamber. The ratio of the minimum concentration of Cl atoms produced in the first hour of the experiments (Q_min) provides an estimation of the minimum fraction of active iron λ_Fe involved in Cl production:

assuming that each Fe^III produces 1/2Cl₂ by excluding recycling effects and secondary activation mechanisms. Owing to these major assumptions and unknown uncertainties (R_Cl/Fe in the aerosol can differ from the bulk, or the concentration of Cl^– can vary depending on the composition and thus deliquescence of the aerosol), λ_Fe represents a rather qualitative value for relative comparisons between the experiments. The contribution of the tare volume of the ions to the measured LWC is 0.11 m³ m^–3 (considered in the calculation), whereas the influence of Fe³⁺ ions is negligible. Fig. 10 shows the results for all Fe^III-containing samples investigated, combined with the measured total minimum and maximum Cl production Q_abs during the first hour. Additionally, the corresponding initial minimum production rates are displayed in the Supplementary material (Fig. S3). Again, the increase in Q_abs for the artificial seawater samples with increasing Fe^III concentration (Cl^–/Fe^III = 13, 53, 101) is highlighted, although, λ_Fe is comparable, which indicates that similar Cl activation mechanisms are going on. This is also the case for the pure NaCl + Fe^III sample where Q_abs and λ_Fe are higher, caused by the absence of competing ligands for Fe^III and thus a higher fraction of photolabile Fe^III–Cl complexes.

Fig. 10.  Overview of the minimum and maximum absolute Cl production Q_abs (atoms cm^–2 h^–1) over the first hour and the corresponding fraction of active Fe^III (λ_Fe) for each experiment with Fe^III-containing samples and various air contaminants (NO_x, O₃, SO₂).

The presence of 20 ppb NO_x in the gas phase led to a considerable increase by a factor of 3–7 for Q_abs from the Cl^–/Fe^III = 101 and the Cl^–/Fe^III = 13 samples. The parallel increase of λ_Fe (even more than 100 %) is an artefact that actually represents the additional Cl activation mechanisms induced by NO_x,^[13] as discussed in the Effects of NO₂, O₃ and SO₂ section. The same applies for the experiment with 630 ppb O₃. Decreasing the pH enhanced the activity of Fe^III, and more than 100 % of Fe^III seemed to be involved, which either is a consequence of recycling effects (re-oxidation of Fe^II that was formed by photolysis of Fe^III complexes), or is caused by the uncertainty of λ_Fe. For the low-pH sample as well as for the untreated samples (Cl^–/Fe^III = 13 and 101), an increase in Q_abs and λ_Fe was observed with the addition of 20 ppb NO_x, which clearly demonstrates the NO_x-induced activation of chloride. The effect is not as pronounced at low aerosol pH.

SO₂ was able to slightly inhibit Cl production. In particular, high SO₂ mixing ratios decreased the Q_abs and thus the fraction of active iron. One can compare λ_Fe with the salt-pan measurements, where a proportion of 0.05–0.07 % of active Fe^III was determined.^[15] Considering a salt pan containing 0.5 g FeCl₃·6H₂O and 99.5 g NaCl results in a total molar Fe^III content of 1.8 × 10^–3. This is quite high compared with the aerosol experiments with only ~3 × 10^–6 mol Fe^III at an LWC of ~2.7 × 10^–9 but total Cl sources comparable with the salt-pan experiments. These large differences prove the small active surface area of the salt pans compared with the homogeneously distributed and irradiated aerosol in the chamber.

The ratio of the concentrations of added pollutants and the generated aerosol concentrations (2–6 mg m^–3) are approximately in agreement with the conditions for sea-salt aerosol over the ocean. Although the smog chamber helps to understand the mechanisms, it is difficult to transfer the results to large-scale atmospheric processes because the aerosol surface-to-volume ratio in the smog chamber is some orders of magnitude higher than in the atmosphere.

We investigated a very simplified system with no organic contaminants in the salt and under hydrocarbon-free conditions. Note that Fe complexes with oxygen-containing organics (such as phenolic compounds) can significantly increase the solubility of iron in seawater.^[86] However, complexation can strongly inhibit iron-induced chlorine activation,^[15] until the phenolic constituents are oxidised by OH^• and Cl^• from the photo-Fenton mechanism to the less-inhibiting but solubility-promoting oxalate. The soluble fraction of iron over the ocean ranges from 0 to 95 % as the bulk marine aerosol type reflects a mixing of multiple aerosol types, and solubility varies with the origin of the iron, aerosol size and composition.^[87–89] Moreover, the solubility increases owing to the photoreduction of Fe^III, which is responsible for an Fe^II fraction in the aerosol of up to 50 % in remote marine areas.^[90] An additional crucial factor for the role of Fe^III photochemistry is the aerosol pH, which varies from 1–9, mainly depending on the origin and age of the aerosol and the corresponding altering processes.^[25,56]

Accounting for these effects in order to quantify iron-induced Cl production in the troposphere, a sophisticated large-scale model would be required. Based on a mean molar Cl^–/Fe ratio of 100–200 in the marine aerosol (central Atlantic),^[91] on the lower fraction of dissolved Fe^III in the aerosol that can form photolabile complexes, and on the 20–200-times larger aerosol surface area in the chamber compared with the Atlantic aerosol,^[92] the mean natural contribution of Fe^III-induced Cl activation cannot compete with alternative mechanisms on a global scale (also indicated by Cl below the detection limit in our experiment with Cl^–/Fe^III = 997). However, it may become important on a local scale with a larger Fe^III burden exposed to saline media, e.g. ship plumes or other iron-containing combustion aerosols^[93] or where soil dust comes into contact with sea salt, but also in brine-containing soils^[94] or salt lakes such as the Dead Sea or Australian salt lakes.^[14]

Furthermore, iron-doped sea-salt aerosols have been proposed as a method for climate engineering, aiming to enhance CH₄ depletion with higher Cl levels in the marine boundary layer and to simultaneously fertilise the oceans.^[95] Based on our results, one may try to estimate the feasibility of such a project. For our conditions, increasing the mean molar Cl^–/Fe^III ratio to, say, 50 by adding Fe^III would have significant effects on Cl production and thus reduce the lifetime of CH₄. Transferred to the global production rate of sea salt Cl^– of 1785 Tg year^–1,^[96] an annual contribution of 56 Tg year^–1 of pure Fe^III (that dissolves totally in the sea-salt aerosol) would be needed. In addition, the effect of enhanced Cl production has to be scaled down by the much lower typical aerosol surface in the marine boundary layer (~60–200 µm² cm^–3[91]) compared with our experiments (~10 000–30 000 µm² cm^–3). Thus, an increase of Cl surface production by a factor of 4 in the chamber (which is the case for Cl^–/Fe^III = 51 compared with the blank) would result in an increase of ~2.5 % transferred to the marine boundary layer, neglecting any effect of dissolved organics or gas-phase species. A mixture with Cl^–/Fe^III = 13 would need 220 Tg year^–1 of Fe^III and result in ~17–19 % increase in Cl production. Note that the increase of atomic Cl does not directly correlate with a reduction of the CH₄ lifetime owing to the flat vertical profile of CH₄ (almost evenly distributed over the total troposphere)^[91] compared with the Cl-producing sea salt (confined to <1 km above the oceans, the marine sea-salt layer)^[91] and thus more inefficient depletion, apart from technical challenges and other potential implications for the ecosystem.

In the present work, we confirmed and quantified Cl formation processes induced by the photochemistry of Fe^III dissolved in the liquid aerosol phase. We found that depending on the Fe^III loading in the artificial sea-salt aerosol, an enormous amount of chlorine can be activated. The initial production rates varied from ~1.9 × 10¹⁸ atoms cm^–2 s^–1 (Cl^–/Fe^III = 13) to ~2.8 × 10¹⁷ atoms cm^–2 s^–1 (Cl^–/Fe^III = 101), whereas no significantly enhanced Cl production was detected for Cl^–/Fe^III = 955 compared with the Fe^III-free salt. The differences can mainly be explained by the amount of Fe^III and the corresponding speciation of Fe^III complexes that depend on the pH.

The presence of 20 ppb NO_x or 630 ppb O₃ considerably increased the initial Cl production rate to ~7 × 10¹⁸ atoms cm^–2 s^–1 (Cl^–/Fe^III = 13) and ~9 × 10¹⁸ atoms cm^–2 s^–1 (Cl^–/Fe^III = 13) respectively (as compared with the blank value of 2 × 10¹⁸ atoms cm^–2 s^–1), owing to the well-known halogen activation processes and potential pH effects. Gas-phase SO₂ instead slightly inhibited Cl production to ~1.7 and ~1.1 × 10¹⁸ atoms cm^–2 s^–1, by adding ~20 ppb and >200 ppb SO₂ respectively. The strong effect of the aerosol pH on Fe^III speciation and subsequent Cl formation was demonstrated by comparing two samples with an equal Cl^–/Fe^III ratio of 101 but a pH adjusted to 2.1–2.3 and 3.9–4.2. Here, the production rate increased by almost an order of magnitude for the more acidic pH. An approach to calculate the active fraction of Fe^III involved in Cl production confirms a higher λ_Fe with lower pH, highlights recycling effects and identifies additional activation mechanisms where NO_x and O₃ are involved. In natural environments, these processes may occur in natural salty media at high dissolved Fe^III concentrations and locally contribute to photochemical Cl formation.

The supplementary material (available on the journal’s website at http://www.publish.csiro.au/?act=view_file&file_id=EN14279_AC.pdf) includes more details on the artificial seawater composition (Table S1), the rate constants of the applied HCs towards the radicals (Table S2), the main equilibrium constants for important Fe^III complexes, the contour plots of selected experiments (Fig. S1), the NO_x, NO, and O₃ time profiles for the Fe^III-free experiment with 20 ppb NO₂ addition (Fig. S2) and an overview of the initial minimum absolute Cl production rate (Fig. S3).