To measure surface resistance, ozone-containing gas ([O3]starting = 40 ppb) was passed over the buffered blank solution or iodide solutions that were pumped through the reactor at liquid flow rates of between 9–35 mLmin-1. Gaseous flow rates were set at 1600, 1900, 2400, 3200 and 4500 sccm, giving possible ozone–solution reaction times ranging from 20–66 s (reaction time = headspace volume/flow rate). The buffered phosphate solution was pre-ozonised by passing a high concentration of ozone (approx. 200 ppb) over the solution in the flow reactor. The total volume of buffer solution required for the experiment was circulated through the reservoir and flow reactor during this time to ensure the entire solution was pre-ozonised. This was continued until a stable ozone concentration was obtained, which typically took around 1 h. At this point the glass, tubing, fittings and buffer were considered “conditioned”. The ozone concentration was lowered to the desired experimental concentration measured while flowing through the bypass line, which had also been pre-conditioned with ozone. The blank measurement was then carried out over the buffer solution, which was pumping through the flow reactor. Blanks were performed in duplicate or triplicate. Directly following the blank measurement, the solution was spiked with iodide into the reservoir, and the solution was mixed and circulated through the flow reactor for approximately 10 min to ensure homogeneity. During mixing, the ozone was passed through the bypass line to avoid reaction. The iodide measurement was performed directly after mixing in triplicate. An example experiment output is shown in Fig. .

Residual ozone was measured after each reaction time, and a mean ozone concentration for each reaction time was obtained ([O3]). A plot of ln[O3] against reaction time (Fig. ) yielded a linear trend, the gradient of which was calculated for each repeat of both the blank (mblank) and the iodide-containing samples (msample), which were each averaged. A blank-corrected gradient (mcorrected, Eq. ) was used to calculate vd by Eq. (), where V is the headspace volume and SA is the liquid surface area (values for all physical constants and metadata provided in Appendix  and calculations described in Appendix ).

Total resistance, rt, is the inverse of gas-phase corrected vd, from which rs can be calculated (Eq. ). The value of ra is variable in the environment but constant in the controlled environment of the flow reactor. Measured over a high iodide concentration (0.02 M), it is assumed that there is negligible surface resistance; therefore, 1/rt ≈ 1/ra . When measured for this flow reactor, an ra of 6.6 ± 0.14 scm-1 was obtained. This value was subtracted from total resistance measured for each sample. 11mcorrected=msample-mblank12vd=-mcorrectedVSA

The relationship between the second-order rate constant, kI--O3, and rs was defined by Eq. (), where H is the temperature-dependent dimensionless Henry's law coefficient of ozone (Hgas/aqueous = 3.63 at 298 K), calculated using Eq. () ; T is the solution temperature (K); and μ is the molar ionic strength (≈ molar concentration, at the concentrations used in this work). To apply Eq. () to our measurements, for each temperature applied, a plot of 1/rs against [I-] gave a linear relationship (see “Results and discussion”; Fig. ), from which the gradient, m, was used to calculate kI--O3, according to Eq. (). 13lnKh=-2297T-1+2.659μ-688.0μT-1+12.1914kI--O3=(mH)2Daq

Conditions within the flow reactor were chosen to emulate the remote marine surface ocean and atmospheric boundary layer. The mixing ratio of ozone in air was not expected to affect ozone uptake to iodide solution as, under these conditions, a bulk aqueous reaction between ozone and iodide is anticipated, for which a lack of ozone dependence is characteristic. Dependence of uptake on the mixing ratio of ozone would be expected if the reaction were proceeding via a surface-mediated Langmuir–Hinshelwood reaction. This is due to surface saturation of ozone at higher mixing ratios limiting potential reactivity on the surface. Therefore, a greater ozone uptake would be expected at lower ozone mixing ratios. Under the conditions of our experiments, there was no significant dependence of deposition velocity on the ozone mixing ratio (Fig. ; p=0.74), confirming the reaction occurs in the bulk phase under the conditions employed. Despite the lack of dependence of ozone mixing ratio on reactivity, consistent conditions of 40 ppb ozone were used in each experiment to mimic a typical mixing ratio of ozone in the troposphere. Similarly, although there is evidence that pH has no impact on ozone deposition to iodide solutions , the solutions were buffered to pH 8 to mimic typical oceanic alkalinity.

Iodide was the reagent in excess in the pseudo-first-order conditions sought for kinetic analysis. At the low iodide concentrations required to emulate marine conditions, iodide was expected to be depleted during the ∼ 90 min experiment time ; therefore, the liquid phase in the flow reactor was continuously replenished from a reservoir, passing through the flow reactor and out to waste. To verify that the chosen pump rate was sufficient to keep iodide depletion below 10 % during the residence time of the liquid in the reactor, liquid samples were collected before and after being exposed to ozone in the flow reactor. The sample after the flow reactor was taken at the midpoint of the experiment (ozone flow rate 3 of 5 and during run 2 of 3; indicated by the red circle in Fig. ). For all reported experiments, the iodide concentration after the experiment was verified by liquid chromatography. It was confirmed that the concentration during the midpoint of the experiment, taken to represent average iodide loss across all ozone exposure times, had less than a 10 % difference from the starting concentration (Appendix ). For all further analysis, the iodide concentrations reported are the midpoint values, rather than the starting values, to best represent the average conditions of the experiments.

The second-order rate constant for ozone with iodide was measured for iodide concentrations between 102 nM and 9.88 µM and for water temperatures between 288 and 303 K (Fig. ). Results were compiled to an Arrhenius plot (Fig. ), leading to a calculated pre-exponential factor, A, of 5.4 ± 23.0 × 1011 M-1s-1 and activation energy, Ea, of 7.0 ± 10.5 kJmol-1. The Pearson correlation coefficient of the Arrhenius plot indicated that the trend was not statistically significant (p=0.53). Therefore, the null hypothesis cannot be rejected; i.e. we cannot conclude that the reaction between ozone and iodide is dependent on temperature.

The blank measurement was subtracted from the iodide-containing measurement (Eq. ), to account for ozone loss to the phosphate buffer and to the glass walls and fittings. It should be noted that if ozone loss during the blank measurement were to be occurring due to reactions outside the iodide-containing reacto-diffusive length, this could cause our measurements to be an underestimation of the true reaction rate.

Several studies have measured the second-order rate constant at around room temperature, as compiled in Table . The rate constants obtained from these studies are plotted as a function of temperature in Fig. . Based on experimental conditions and relevance to marine environments, the studies which are most comparable to ours are those which employ iodide concentrations below 10-4 M ; these are , , and . The study by was carried out with iodide concentrations approaching the upper limit of where aqueous reactivity dominates, but ozone was applied in solution, removing the possibility of surface reactivity. , , and report a comparable but slightly lower kI--O3 than this work. A possible reason for this could be the lack of replenishment of iodide in their studies. Iodide depletion could occur within the timeframe of their experiments, which would have resulted in a rate of ozone loss lower than anticipated for their expected iodide concentration. Iodide was not explicitly measured in those studies, so any depletion would not be known. For all other reported rate constants, the conditions employed could promote surface reactivity; therefore, the measurements are not comparable to the results reported in this work.

Of the various reported rate constants for the ozone iodide reaction (Table  and Fig. ), only one other study has explicitly investigated the temperature dependence and obtained A = 1.4 × 1022 M-1s-1 and Ea = 73.08 kJmol-1, with an estimated error of 40 % . Our work contradicts the strong positive temperature dependence reported in their work. The difference in conclusion could be due to the differences in conditions. At the concentrations used in our study, bulk reactivity is expected to occur, whereas the conditions employed by (iodide concentrations of up to 3 M) are in a range which could display surface reactivity. The surface reactivity is also dependent on the gaseous ozone concentration, which was not reported. Further, while interfacial reactivity is not yet fully understood, the pre-exponential factor reported by is approximately 12 orders of magnitude greater than a diffusion-controlled reaction. In contrast, the pre-exponential factor reported in our work could feasibly be attributed to a diffusion-controlled reaction, within error bounds.

Amalgamating results of single-temperature studies by , at 277 K, and the room temperature measurements of , , , and , as well as this study, yields a negligible or slightly negative temperature dependence, within the associated experimental errors, for the bulk-phase reaction between ozone and iodide (Fig. ). Comparing only those experiments which are environmentally applicable, there is no clear trend in temperature dependence. Both conclusions are consistent with the results of our study.

Theoretical studies of this reaction have previously been performed. One study simulating the aqueous phase concluded a strong positive activation energy (20 kcalmol-1 or approx. 84 kJmol-1) for the formation of intermediate complex [OOOI]- (Eq. ), acting as the rate-limiting step . A subsequent gas-phase simulation concluded a weaker but still positive activation energy of 7.5 kcalmol-1 or approx. 32 kJmol-1 for the same adduct formation . These values are in contradiction with the results of this experimental study, and if the energetic barrier to this reaction is indeed adduct formation, our results indicate this barrier is overestimated by computational studies. Therefore, we propose that further work is required to reconcile mechanistic theory with observations.

A previous laboratory study of inorganic iodine (HOI and I2) emissions from ozonised iodide solutions (0.1–5 µM iodide; 222–3600 ppb ozone; temperature of 276–298 K) yielded effective activation energies of 17 ± 50 kJmol-1 for HOI emissions and -7 ± 18 kJmol-1 for I2 emissions . The emissions of HOI and I2 depend on several chemical reactions, each with individual dependencies on temperature, and several physical factors including the solubility and diffusivity of ozone. Thus, the temperature dependence can be negative or positive depending on the combination of these factors. The study was carried out in conditions favouring bulk-phase reactions between ozone and iodide, making their results experimentally comparable to our work. When modelled the emissions of HOI and I2, they demonstrated that their results could only be accurately replicated when assuming that Ea ∼ 0 kJmol-1 for the ozone–iodide reaction and not when using the temperature dependence from . The model employed by was the sea surface microlayer (SML) model described by , except with the inclusion of temperature-dependent processes. The model with temperature dependence did not account for iodine depletion following its fast reaction with ozone and did not account for iodide replenishment from the waters below . To evaluate whether the rate coefficient obtained in our experimental work is consistent with the measured temperature dependence of fluxes of gaseous iodine compounds from iodide solution under ozone, we applied our rate coefficients to an updated SML model . The updated SML model (details in Appendix ) includes the mixing of iodide between the SML and the underlying water and simulates surface iodide depletion, especially at low wind speeds and/or reduced turbulence. The depth of the SML in this model is defined as the reacto-diffusive length of ozone, which is unique to each combination of conditions.

The SML model was constrained to the range of conditions reported by . The effective activation energy was calculated (Ea = - gradient ×R) from an Arrhenius-type plot of the natural log of the calculated emissions of HOI and I2 (in units of molec.cm-2s-1) against the inverse of the temperature (in K). It was not possible to accurately calculate an equivalent wind speed for the MacDonald laboratory experiments; therefore, two low wind speeds (0.005 and 0.03 ms-1) were assumed. When assessing the impact of different wind speeds, we applied conditions of 1500 ppb ozone and 2.5 µM iodide.

The HOI and I2 emissions obtained from the updated SML model are displayed as an effective Arrhenius plot in Fig. . The modelled emissions indicate there is a dependence of the effective activation energy on the wind speed of the experiments for I2 emissions. At the higher wind speed, the model overestimates the activation energy, outside the error range of the experimental measurements. The lower wind speed predicts an Ea within the experimental error range. For HOI emissions, a slightly negative Ea is predicted by the model, which lies within the experimental error range quoted by MacDonald et al. (2014), and does not show a dependence on wind speed.

Knowing that there is negligible temperature dependence on the reaction between iodine and ozone, the relative changes in each step of the production and emission of HOI and I2 in the SML model were interrogated to explain the predicted activation energies. Iodide in the surface layer is depleted if the replenishment from the bulk solution occurs at a slower rate than the reaction of iodide with ozone. Iodide depletion was modelled over the range of ozone, iodide and temperature reported by , and depletion was predicted to increase with increasing ozone concentration and wind speed and with decreasing iodide concentration and slightly decrease with temperature (Fig. ). The SML model shows that iodide depletion increases with increasing ozone concentration and with decreasing iodide concentrations due to the chemical consumption of the available iodide. Greater depletion is seen at the higher wind speed because of the increase in ozone deposition as aerodynamic resistance (ra) is reduced with higher air-side turbulence (vd = 0.0013 cms-1 at 0.03 ms-1 and vd = 0.0003 cms-1 at 0.005 ms-1). Waterside transfer velocity, kw, which replenishes iodide from the bulk to the SML, is still low at these wind speeds and does not offset the increase in ozone–iodide reactivity; therefore, iodide is seen to be more depleted at 0.03 ms-1 compared to 0.005 ms-1. This is the opposite to what we expect in the environment, where greater wind speeds would be associated with a greater degree of mixing from the bulk and less iodide depletion. This is discussed in detail in . The trend with wind speed we describe in the current work is specific to low-wind-speed laboratory conditions. This should therefore be considered in future experimental design if environmentally applicable emission data are sought.

The SML model predicted iodide depletion to slightly decrease with temperature; that is, we predict a higher iodide concentration in the SML at higher temperature given equal wind speed and ozone concentration. This is due to the small positive temperature dependence in the ozone–iodide reaction being offset by the increase in kw with temperature, which increases the mixing of iodide from the bulk to the SML. The effect of temperature is more pronounced at higher wind speed because, though small, kw is higher at greater wind speed and therefore has greater impact at 0.03 ms-1 compared to 0.005 ms-1. This impact is minor compared to the effects of ozone, iodide and wind speed. In light of this discussion, it should be noted that there are very few observations of iodide concentrations at or proximal to the surface layer , and it is this iodide which is available for reaction with ozone. Concentrations of iodide at or proximal to the surface may frequently be different from the reported bulk concentration.

For both wind speeds investigated, ozone deposition velocity was effectively constant over the modelled temperature range (Fig. ). This is because at low wind speeds, the deposition of ozone is limited by air-side resistance, ra, which has no temperature dependence. Therefore, factors which decrease rs (e.g. changes in ozone solubility, aqueous diffusivity of ozone, and reaction rate between ozone and iodide with temperature) do not influence deposition significantly. Hence, the availability of ozone in solution was constant over the prescribed temperature range. Temperature trends in HOI emissions are therefore controlled by drivers in mixing to the bulk or drivers in Reaction (), which could include air-side transfer velocity, ka, and the solubility of HOI, expressed as the dimensionless Henry's law coefficient, HHOI, which is calculated using equations in and using physical constants from . The relative changes in selected chemical and physical drivers across the temperature range are displayed in Fig. . Flux into the air, Fa, is controlled by Eq. (), where CSML and Ca are concentrations of the species in question in the SML and in the air, respectively. In this model, Ca is assumed to be zero. Across this temperature range, we expect a small decrease in the product of HHOI×ka(HOI), resulting in a lower flux of HOI into the atmosphere as T increases and hence the slight negative Ea for HOI emissions.

Due to its high solubility, the majority of HOI produced in the SML is mixed into the bulk liquid phase, Fb (Eq. ), rather than being emitted. Here, kw is the waterside transfer velocity, and Cb is the bulk concentration, set to zero in this model. Chemical production of HOI is the only source of HOI in the SML, and the second-order rate constant of the chemical formation of HOI, kI--O3, increases with temperature. At 0.03 ms-1 wind speed, this is augmented by the reduced iodide depletion with higher temperature, leading to production of more HOI. At 0.005 ms-1 wind speed, this effect is reduced by the effectively unchanging iodide depletion across the temperature range. In both instances, increased HOI production with T counteracts some of the increased loss to the bulk and more so at 0.03 ms-1 wind speed. This results in a slightly higher [HOI] in the SML at the higher temperature range, which increases the Ea slightly at the higher wind speed (-4.65 kJmol-1 at 0.005 ms-1 compared to -4.08 kJmol-1 at 0.03 ms-1). 15Fa=ka(H×CSML-Ca)16Fb=kw(Cb-CSML)

For I2, emissions are largely controlled by its low solubility. Temperature impacts on chemical and physical drivers are displayed in Fig. . We calculate a strong increase in ka(I2)×HI2, which leads to a higher Fa at greater temperatures, and a positive Ea at both wind speeds. While the rate constant of mixing to the bulk, kb, also increases with temperature, its effect is counteracted by the increased chemical production of I2 with increased temperature, driven by increased kI--O3 and decreased iodide depletion. This results in a more positive Ea at the higher wind speed.

The effects of varying iodide concentrations on HOI and I2 emissions were also investigated at a wind speed of 0.005 ms-1 due to the model's better comparability to experimental results at this wind speed. The calculated activation energies are displayed in Fig. . For the range of iodide concentrations used in MacDonald et al. (2014), we calculate that the Ea of HOI emissions has no dependence on iodide and remains slightly negative for all modelled conditions (driven by increased mixing to the bulk at higher temperature, as above). For I2 emissions, the modelled Ea was more strongly positive at the lowest iodide concentration due to greater iodide depletion in the SML. The iodide limitation is counteracted at higher temperatures by increased mixing from the bulk. In Fig. , it can be observed that for 100 nM iodide, the iodide enrichment factor increases with temperature, which is not observed for the other iodide concentrations. This results in a more strongly positive Ea at the lowest modelled iodide concentration of 100 nM, which is comparable to conditions found in the environment. We therefore believe that the activation energy of I2 emissions reported in MacDonald et al. (2014) may underestimate the temperature dependence of oceanic I2 emissions.

The model was also used to test for ozone sensitivity; however, across the range studied here, the ozone mixing ratio was not determined to influence activation energy for HOI or I2 emissions.

It is important to note that there are large uncertainties present in calculating gas and water transfer coefficients (ka and kw), quoted as up to a factor of 2 . Furthermore, there are also large uncertainties in the solubilities of HOI and I2; for example, the Henry's law coefficient of HOI is assumed to be within the range of 0.44 to 440 molm-3Pa-1 . Uncertainties in kw, ka and H are not included in this analysis as they are not accurately quantified, nor is it clear how they relate to iodine emissions. An additional caveat in comparing the SML model to laboratory studies is that the model was designed for application to environmental settings and therefore is not optimised for the low wind speeds and high ozone and iodide concentrations explored in this work.