Volatile organoiodine compounds (VOIs) are the main carrier of iodine from the oceans to the atmosphere. We have identified a novel, sea-surface source of the short-lived VOIs CH2I2, CHClI2 and CHI3 in a series of laboratory experiments. These compounds were formed when seawater, collected during winter in the North Sea, was exposed to ambient levels of ozone. The VOIs are produced from the reaction of marine dissolved organic matter with hypoiodous acid/molecular iodine, which are formed at the sea surface when ozone reacts with dissolved iodide. The same three VOIs were formed when we incubated seawater of different productivity levels with molecular iodine during a cruise in the tropical Atlantic Ocean. We suggest that the presence of dissolved iodide, dissolved organic matter and ozone can lead to the sea-surface production of CH2I2, CHClI2 and CHI3. As such, this process could provide a ubiquitous source of iodine to the marine atmosphere.
 It is known that hypoiodous acid (HOI) and molecular iodine (I2), produced at the sea surface by the reaction of ozone with dissolved iodide [Garland et al., 1980; Thompson and Zafiriou, 1983], react with marine dissolved organic matter (DOM) to produce dissolved organic iodine (DOI) [Truesdale and Luther, 1995]. However, already in 1980, it was suggested that the methyl iodide observed at the ocean surface could result, at least in part, from the oxidation of iodide by ozone and the subsequent reaction of HOI with organic molecules [Garland et al., 1980]. Here, we show that a small fraction of the organic iodine formed via the above reaction is composed of the very reactive VOIs CH2I2, CHClI2 and CHI3. Because they are volatile and formed at the sea surface, these compounds can readily escape to the atmosphere. Moreover, their very short atmospheric lifetimes account for their not being generally detected in the marine atmosphere. This is particularly the case for CHI3 for which we are not aware of any previous report of its existence in the marine environment, either in water or air. Thus, we have identified a new, and potentially ubiquitous, surface source of VOIs.
 Surface seawater was collected in the Central North Sea in January 2008, and had a concentration of dissolved organic carbon of 110 ± 10 μM. Typical primary production values for this area in January are of 260 mg C m−2 day−1 [Howarth et al., 1994]. The water was gently filtered through 0.2 μm Millipore nylon filters using a combusted glass filtration unit, stored in glass bottles in the dark at 12°C and used within 2 weeks of collection.
 All experiments were carried out in the dark. A 2-L round glass cell was modified to accommodate inlets and outlets for gases and a stopcock for water sampling. The water and air in the cell were mixed with a home-made PTFE stirrer, which consisted of a motor-operated rod inserted in a PTFE stopper. Two o-rings inserted in the stopper ensured that the vessel was air-tight. Hydrocarbon-free nitrogen and ultra-pure oxygen were mixed in a ratio of approximately 4:1 before passing through a home-built quartz cell, where ozone was generated with a UVP Pen-Ray mercury lamp. The ozone concentration was controlled by adjusting the oxygen flow rate, and was measured by a Thermoelectron 49C analyser, connected at the head of the cell. Ozone was generated at a concentration of 66–67 ppb, which resulted in a residual ozone concentration in the 2-L cell of 35 ppb, the drop in concentration being caused by ozone reaction on the cell walls. The total gas flow rate of 1.8 L/min was largely determined by the requirements of the ozone analyser. The gas flow in the cell was 500 mL/min. For the experiment, the cell was flushed with gas for 4–5 hours before the addition of 1-L purged (halocarbon-free) seawater. The surface of water in contact with the gas phase was 200 cm2. The system was left to equilibrate overnight, after which the ozone lamp was switched on for 48 hours.
 The gas-phase VOIs were sampled, once per hour, at the exit of the cell via a 1/8” silcosteel line at a flow rate of 50 mL min−1 for 10 minutes. Samples were preconcentrated at −10°C on a Markes UNITY/Air Server system using a carboxen/carbograph (TBC) trap and were dried prior to preconcentration with a 183 cm counter-flow Permapure Nafion dryer. The sample was injected onto a 105 m, 0.32mm rtx-502.2 Restek capillary column by heating the sample trap to 250°C for 10 minutes at a helium carrier gas flow of 2 mL min−1. Detection was performed by an Agilent 5973N mass spectrometer operated in negative ion mode, with methane used as the reagent gas. The instrument, and its significant sensitivity enhancements for organoiodine species, are described in more detail by Worton et al. . Calibrations were performed every 6 samples by introduction of a gas standard containing CH3I into the sampling manifold and Nafion dryer under identical conditions to the air samples. The gas standard had previously been calibrated against a certified standard from the National Oceanic and Atmospheric Administration. The sensitivity of the MSD to CH2I2 during these experiments was estimated using the observed CH3I sensitivity (obtained directly from the calibrations) and the relative sensitivity of the MSD to CH2I2 compared to that of CH3I. This relative sensitivity has been measured on three occasions (along with those of several alkyl iodides) in the past using permeation tubes generating 5–10 pptv mixtures. The mixing ratios of CHClI2 and CHI3 were estimated assuming the same sensitivity as CH2I2. The lower sensitivity of higher alkyl iodides and CH2I2 compared to CH3I suggests that the sensitivities of CHClI2 and CHI3 may be lower than for CH2I2, resulting in mixing ratios that could be up to a factor of two higher than estimated here.
 Seawater was sampled at the beginning (ozone switch-on) and end of the experiment, and once during the experiment (24 h). The reason for sampling only once mid-experiment was in order to minimise variations in the volume of seawater in the flask. However, two additional experiments were carried out where the seawater was sampled for VOIs every ∼30 minutes for 5–6 hours, in order to determine the initial iodocarbon production rates in seawater. For all experiments, 40 mL of seawater were sampled from the cell into gas-tight glass syringes. After the addition of two deuterated internal standards for monitoring any GC-MS sensitivity drift [Martino et al., 2005], the water was transferred to a home-built glass stripper and purged at 40 mL min−1 for 20 minutes with hydrocarbon-free nitrogen. The gas stream passed through glass wool for removal of aerosol particles, and was dried by passing through one static and two counter-flow Permapure Nafion driers (122 and 183 cm). The analytes were cryo-focussed in an empty stainless steel loop over liquid nitrogen (−150°C), and then injected into an Agilent 6890N-5973 GC-MS by heating the loop with boiling water. Compounds were separated on a DB-VRX capillary column (0.25 mm × 30 m, 1.4 μm film thickness) with a He flow rate of 1 mL min−1. The detection limit for CH2I2 was 0.2 pM. CH2I2 standards were prepared gravimetrically by dilution of the neat compound (Sigma-Aldrich) in HPLC-grade methanol (Merck). There are no commercial standards available for CHClI2, and we could not prepare reproducible CHI3 standards, presumably because of the “sticky” nature of this compound. Therefore, for CHClI2 and CHI3 we assumed a GC-MS response factor equal to that of CH2I2.
 Samples were also taken during the experiment for analysis of dissolved iodine species. Iodate was determined spectrophotometrically at 350 nm [Jickells et al., 1988]. Iodide was determined by adsorptive cathodic stripping voltammetry [Campos, 1997]. Dissolved organic iodine (DOI) was first converted into inorganic iodine by UV-irradiation for 3.5 hours with a 400 W high-pressure Hg-vapour lamp, after addition of hydrogen peroxide [Wong and Cheng, 1998]. The resulting inorganic iodine species were then reduced to iodide with ascorbic acid and determined by voltammetry [Campos, 1997]. Dissolved organic iodine was determined by the difference in concentration of total inorganic iodine after and before UV-irradiation. Typical precisions were 1–4% for iodide and iodate, and 4–13% for DOI.
3. Results and Discussion
 The mass balance for dissolved iodine species before and after each of the four ozonation experiments is shown in Figure 1. The reaction of ozone with dissolved iodide at the water surface resulted in the decrease of iodide from 112 ± 2 nM to 33 ± 5 nM during 48 hours (means of four experiments). At the same time, the concentration of dissolved organic iodine (DOI) increased from 138 ± 7 nM to 208 ± 9 nM. These results can be explained by the very reactive intermediates HOI and I2, formed when ozone reacts with iodide at the sea surface [Garland et al., 1980; Thompson and Zafiriou, 1983]:
Both HOI and I2 tend to react rapidly with the marine DOM present in solution, leading to the observed increase in dissolved organic iodine (DOI, Figure 1), although a small fraction also disproportionates into iodide and iodate [Truesdale et al., 1995].
 The novelty of this study is that we also observed the formation of CH2I2, CHClI2 and CHI3 both in the liquid and in the gas phase, as direct response to ozonation of the seawater. Figure 2 shows one of the four replicate experiments carried out over 48 hours. Plots of the other replicate experiments are shown in the auxiliary material (Figure S1). The four experiments show that all three compounds transfer to the gas phase. Concentrations in the liquid phase after 48 hours reached 5–20 pM pM CH2I2, 4–11 pM CHClI2 and 50–500 pM CHI3. Because of the picomolar levels of these compounds, their contribution to the dissolved iodine mass balance is negligible (Figure 1). 48-h controls carried out (i) without ozonation or (ii) with ozonation but without any seawater in the cell, showed no production of the three iodocarbons under either condition.
 In general, the gas phase concentrations increased steadily for the first ∼30 hours, reaching 0.3–0.4 ppt CH2I2, 0.6–1 ppt CHClI2, and ∼0.15 ppt CHI3, and started to drop after 40 hours. This is probably due to the decrease in iodide and/or DOM concentrations. It should be noted that, in the natural environment, the concentrations of iodide and DOM in the microlayer do not decrease, as they are continuously supplied from the underlying water: therefore the drop observed here at 40 hours probably does not take place in the environment.
 In the liquid phase, we started to detect the three compounds between 1.5 and 3 hours after the beginning of the ozonation (Figure 3). This “lag phase” is generally not observed in the gas phase concentrations during the 48-h long experiments (Figures 2 and S1), and therefore it could be an experimental artefact due to the relatively high detection limits in the water phase. The apparent crash in CHI3 concentration in Figure 3a is probably an experimental artefact, resulting from the sticky nature of this compound. The average rates of production calculated from these two replicate experiments were 0.66 ± 0.05 pM h−1 for CH2I2 and 0.50 ± 0.07 pM h−1 for CHClI2. We calculated a CHI3 production rate of 10 ± 1 pM h−1 using the data from experiment 3B only. The errors on the production rates were calculated from the uncertainties of the linear fit to the slopes. The iodoform production rate was more than an order of magnitude higher than that of CH2I2 and CHClI2, suggesting that iodoform is the primary compound formed by this process.
 The production of CH2I2, CHClI2 and CHI3 can be explained by the haloform or a haloform-type reaction, in which HOI (or I2) reacts with organic compounds containing hydrogens α- to keto-groups. This reaction is known to occur during the disinfection of natural waters, leading to a wide range of polyhalomethanes, including CHI3 and CHClI2 [Bichsel and von Gunten, 1999; Hansson et al., 1987]. Moreover, Carpenter et al.  suggest that this reaction also occurred at the quasi-liquid layer of sea-ice and was responsible for the observed atmospheric CH2I2, CH2BrI and CH2ClI in the sub-Arctic atmosphere at Hudson Bay. Those authors did not monitor CHI3 or CHClI2.
 The abundance of VOIs produced by the process shown here might depend on the abundance and perhaps on the nature of the organic substrate. These can vary widely both temporally and spatially [Aluwihare et al., 1997; Carlson et al., 1994]. However, we found that production of CH2I2, CHClI2 and CHI3 is not limited to seawater from the North Sea, since we observed the formation of the same three iodocarbons in the dark, in on-deck incubations of filtered seawater with added molecular iodine (100 nM) during the D325 SOLAS INSPIRE cruise in the Eastern Tropical Atlantic Ocean (see Figures S2–S4 and Table S1). The incubations were carried out at five stations with a wide range of primary productivities and chromophopric dissolved organic matter (CDOM) levels. The rates of production varied between 0.3 and 2.4 pM h−1 for CH2I2, 0.01 and 0.09 pM h−1 for CHClI2 and between 1 and 2.6 pM h−1 for CHI3. The production rates were not related to the primary production or the CDOM levels, suggesting that production of the iodocarbons might not be related to the presence and/or abundance of labile DOM freshly-produced by phytoplankton.
 Our results show that 1% (±0.3%) of the iodide oxidised by ozone reacts with organic matter to form reactive VOIs. If all of the reactive VOIs formed this way escaped to the atmosphere, an ozone concentration of 20 ppb and deposition velocity of 0.05 cm s−1 would result in a flux of 5 × 107 cm−2 s−1 reactive iodine from the sea surface to the marine boundary layer (assuming 20% of the ozone deposited to the sea surface reacts with iodide [Garland et al., 1980]). This is comparable to an estimated downward flux of iodine to oceanic waters of 3 × 107 atoms cm−2 s−1, based on a mean aerosol iodine concentration of 50 pmol m−3 [Arimoto et al., 1995; Duce et al., 1983] and a deposition velocity of 1 cm s−1 [Jickells and Spokes, 2001]. However, not all of the VOIs formed by the mechanism identified here will evade from the sea surface: a fraction will undergo (i) transport from the ocean surface to the underlying mixed layer and (ii) photolysis at the surface and in the water column [Martino et al., 2006]. This fraction will depend on solar irradiance, wind speed and state of the sea, and cannot be quantified easily, but is likely to be significant. Therefore, even if only half of the VOIs produced at the sea surface are released into the atmosphere, it would result in an upward flux of 2.5 × 107 cm−2 s−1 reactive iodine. This is enough to roughly balance the depositional flux of iodine to the sea surface, and is a potentially significant iodine source to the marine atmosphere.
 Our results show that CH2I2, CHClI2 and CHI3 are produced when seawater is exposed to ambient levels of ozone. This is the first time these compounds have been found to be formed this way. Production of these iodocarbons is the result of a purely chemical process depending on the presence of ozone, dissolved iodide, and dissolved organic matter. Our results seem to suggest that this process might occur independently of the biological generation of “fresh” DOM, but instead might depend on the presence of the background DOM present ubiquitously at the sea surface [Buffle, 1988]. As such, this process could provide a ubiquitous source of iodine to the marine atmosphere. Further investigation is necessary to assess the likely extent of this process in the natural environment, and the resulting fluxes of iodine to the atmosphere.
 We thank Keith Weston for collection of seawater for our experiments and Rosie Chance for assistance with the iodine analyses during the INSPIRE cruise. We are grateful to Alex Baker, Claire Hughes, Phil Nightingale, John Plane and Roland von Glasow for helpful discussion. This manuscript benefited from the constructive criticism and suggestions of Eric Saltzman and an anonymous reviewer. Financial support for this work was provided by the Natural Environment Research Council (NERC) grant NE/C511572/1. JW was funded by the EU Marine Aerosol Production grant 018332.