Influence of the Sea Surface Microlayer on Oceanic Iodine Emissions

The influence of organic compounds on iodine (I2) emissions from the O3 + I– reaction at the sea surface was investigated in laboratory and modeling studies using artificial solutions, natural subsurface seawater (SSW), and, for the first time, samples of the surface microlayer (SML). Gas-phase I2 was measured directly above the surface of liquid samples using broadband cavity enhanced absorption spectroscopy. I2 emissions were consistently lower for artificial seawater (AS) than buffered potassium iodide (KI) solutions. Natural seawater samples showed the strongest reduction of I2 emissions compared to artificial solutions with equivalent [I–], and the reduction was more pronounced over SML than SSW. Emissions of volatile organic iodine (VOI) were highest from SML samples but remained a negligible fraction (<1%) of the total iodine flux. Therefore, reduced iodine emissions from natural seawater cannot be explained by chemical losses of I2 or hypoiodous acid (HOI), leading to VOI. An interfacial model explains this reduction by increased solubility of the I2 product in the organic-rich interfacial layer of seawater. Our results highlight the importance of using environmentally representative concentrations in studies of the O3 + I– reaction and demonstrate the influence the SML exerts on emissions of iodine and potentially other volatile species.


Chemicals:
All chemicals were used as received, without further purification. All solutions were made with HPLC grade water, purged for > 6 hrs with N2 to eliminate dissolved residual halocarbons. Buffered solutions of iodide were prepared by adding different volumes of concentrated KI stock solutions to a solution of 0.01 M NaH2PO4 ( 99%, Sigma-Aldrich) adjusted to pH 8 with NaOH solution (50 vol%, Sigma) and HCl (37 vol%, Sigma). Artificial seawater (AS) solutions were made by adding 0.5 M of KCl ( 99.5%, Sigma-Aldrich) and 8 × 10 −4 M of KBr ( 99.5%, Fisons, UK) into a 0.01 M (0.1 M for the halocarbon experiments) buffered solution of NaH2PO4 .

Sampling subsurface seawater and surface microlayer:
The natural seawater samples were obtained from the North Sea, off the coast of Bridlington, North Yorkshire, United Kingdom (see Table S2 below for details). Subsurface seawater (SSW) was sampled by dipping a capped HPDE bottle mounted on a telescopic pole to approx. 50 cm under the sea surface and then opening the cap. The surface microlayer (SML) was sampled using two identical Garrett screens (16 mesh, 0.80  0.54 m). The thickness of the SML samples, as calculated from the volume collected per dip, ranged from 155 to 225 µm, similar to reports in the literature. 1 Next, the samples were filtered through GF/F ashed quartz filters (47mm diameter, Whatman) under a gentle vacuum and stored in acid washed HDPE bottles at -20˚C. All material used for the seawater sampling, apart from the metal Garrett screen, was acid washed in a 4 vol% HCl solution (min. 4hr) and rinsed with Ultrapure water (18MΩ, Elga). The Garrett screens were cleaned by successive rinses with hot tap water, ethanol and Ultrapure water. These were initially dipped 3-5 times in the seawater to rinse them and the water collected from these dips was used to rinse 2.5 L Winchester storage flasks. The screen was lowered horizontally into the water to full submersion and then slowly and horizontally brought up through the surface, in order to collect the SML adhering to the metal mesh. The SML and subsurface samples were transferred to the brown glass Winchester storage flasks for transport to the lab. All sampling was done at the upwind side of the boat to avoid contamination.

Analysis of iodide:
Iodide in the natural seawater samples was analysed by cathodic stripping square wave voltammetry following the method in reference ( 2 ), using a μAutolab III potentiostat connected to a 665VA stand (Metrohm) with a hanging mercury drop electrode, using a Ag/AgCl reference electrode and a carbon or platinum auxiliary electrode. 12 or 15 mL of the sample was introduced in a glass cell and 90 μL of Triton X-100 (0.2%) was added. The sample was purged with oxygen-free grade N2 for 5 minutes before each measurement. The deposition potential was set at 0 V and the deposition time was typically 60 sec, scan step was 2 mV with a 75 Hz frequency and 0.02 V wave amplitude and ranged from 0 to 0.7 V. Each scan was repeated 5 or 6 times. Repeatability was equal or better than 5%. Iodide concentrations were determined using 2 or 3 standard additions of a KI solution of 2  10 5 M, aiming to cover the end of the linearity domain with the last addition.

Determination of Dissolved Organic Carbon (DOC):
To convert inorganic carbonates to carbon dioxide (CO2), approx. 2 vol% of a hydrochloric acid solution (10 vol%) was added to the seawater samples, resulting in a pH value of 2. Afterwards, particle-free oxygen was passed through the acidified sample to remove the formed CO2. The oxidation of the carbon within the sample was performed by adding 250 μL of the carbonate-free solutions to a catalyst (platinum on quartz wool) held at a temperature of 850 °C. The formed carbon dioxide was detected by NDIR (nondispersive infrared spectroscopy). DOC content was determined by using an external calibration with potassium hydrogen phthalate and a certified TOC-standard (50 mg/L, Sigma) and reported as equivalent concentration of carbon. The enrichment factor based on DOC represents the ratio of [DOC]SML/[DOC]SSW.

Surface tension measurements:
Surface tension (γ) was measured using Du Noüy Ring tensiometry 3 performed using an Attension Sigma 70 instrument with thermocouple (density resolution: 1  10 4 g cm 3 , force resolution: 0.1 μN) which was calibrated daily. The short-term precision of the instrument, as measured from ~100 repeat measurements, was typically 0.05 to 0.1 mN m 1 . Before each measurement, the sample vessel (∼30 mL) was washed with ethanol, then submerged in a 4 % HCl acid bath before being rinsed with Ultrapure water (18.2 MΩ, Elga) and dried. Before measurements, the sample was left for at least 5 mins to equilibrate. The platinum ring was cleaned with ultrapure water and burned using a butane flame before each measurement. The enrichment factor EFπ is the ratio of the surface pressure (π = γ0 -γ) of the SML over the SSW, where γ0 is the surface tension of artificial seawater determined at the same conditions as the sample (here 74.88 mN/m at 20°C).The surface pressure reflects the concentration of the total surfactants in the sample and is useful in comparisons between studies, provided γ0 is accurately determined each time. EFπ thus expresses the enrichment of surfactants in the SML compared to those the SSW.

Broadband cavity enhanced absorption spectroscopy (BBCEAS) for gas-phase molecular iodine:
A schematic of the BBCEAS apparatus is shown in Fig S1 below. The spectrometer's light source was a high intensity light emitting diode (Dragon 1 Powerstar, Intelligent LED Solutions, peak wavelength = 530 nm) mounted on a temperature-stabilised, peltier-cooled heat sink. Light from the LED was coupled into a fibre optic cable abutted up against the emitter's surface and re-collimated at the far end of the fibre by a microscope objective lens. Two turning mirrors directed the light beam into the optical cavity formed by two highly reflective mirrors (Layertec GmbH). The total length of the cavity was 83 cm; however 11 cm sections inside each mirror mount were purged with synthetic air to prevent contamination of the mirror surfaces. Thus the sample was confined to the central 61 cm portion of the cavity, and the absorption measured inside the cavity was therefore corrected for the cavity's length factor (LF = 1.36  0.06). LF was determined from a total of n = 14 replicate measurements using two different methods: measuring the O4 absorption bands in synthetic air and measuring the Rayleigh scattering due to argon in samples supplied into the central portion of the cavity. The reflectivity of the cavity mirrors, R(), was characterised versus wavelength by measuring BBCEAS spectra of NO2 samples prior to this work; the reflectivity was re-measured at the start of each experiment by flowing pure oxygen through the cavity and recording and fitting the O4 bands. The peak reflectivity of the cavity mirrors was 99.9927% at 548 nm, giving a maximum effective S4 absorption path of 11 km. The reflectivity at the extremes of the wavelength range used here (519.8 to 571.3 nm) was still sufficient to achieve a path length of 5 km.
Light transmitted through the cavity's output mirror was collected by an f = 30 mm lens, focussed into a fibre optic cable, and dispersed and detected as a function of wavelength by a fibre-coupled spectrometer (OceanOptics HR4000) interfaced to a laptop computer. To minimise any thermal drifts, the spectrometer was housed inside a thermally stabilised enclosure at 12 C. Twelve BBCEAS spectra (5 seconds each) were averaged over 1 minute, and the spectral structure was fitted with reference absorption cross sections of I2, NO2, O4, H2O and OIO to retrieve I2 concentrations. As previously described, 4 the I2 cross sections came from a calculated, high resolution I2 absorption spectrum, degraded with the spectrometer's instrument function (0.13 to 0.19 nm HWHM, variable across the wavelength range), and scaled to the I2 cross sections reported by Saiz-Lopez et al (2004). 5 The NO2 and H2O cross sections 6,7 were also smoothed with the spectrometer's instrument function, whereas the O4 (Hermans, as reported in the HITRAN database 8 ) and OIO cross sections, 9 with their broader features, were used without smoothing. NO2 concentrations retrieved from the BBCEAS spectra were used to verify that any NO2 from room air entering into the reaction vessel when adding samples had been flushed from the vessel before the spectra used to quantify I2 were acquired.
The statistical uncertainty for retrieving I2 concentrations from fitting the absorption structure in BBCEAS spectra was typically 4 pptv (1 in 1 minute), and this represents the limit of detection for I2 reported in this paper. However the total uncertainty on the I2 measurement also depends on systematic uncertainties in the reference absorption cross sections (taken as 15% for I2 and OIO, and 3% for NO2, H2O and O4) and for determining the mirror reflectivity from O4 spectra (assumed 3%, like the O4 cross sections themselves). Under most conditions encountered in this work, the systematic uncertainty (totalling 16% for I2 including the uncertainty in LF) dominates over the statistical fit uncertainty, and the latter is only important at low I2 concentrations (below about 100 pptv). The total uncertainty we quote on I2 measurements is the systematic uncertainty and the statistical fit uncertainty, combined in quadrature.
The I2 flux data points shown in plots in this paper come from averaging the I2 concentrations measured over 10 to 20 replicate BBCEAS spectra at a fixed set of conditions (i.e. the average I2 over 10 to 20 minutes sampling period). Any I2 concentrations that were changing (e.g. around the addition of iodide aliquots into the reaction vessel) were excluded. The plotted error bars are the average total uncertainties on the 10 to 20 BBCEAS measurements of I2 that were averaged to create the data point. The variability in the measured I2 concentrations within the sampling windows was always smaller than the total BBCEAS retrieval uncertainty, and thus was not included in calculating the plots' error bars.

Ozone measurements:
During the experiments the ozone concentrations downstream was monitored using 3 different ozone monitors (model 205, 2B-Tech; 400A and T400A EnviroTechnology) which were intercompared and the 2B-tech instrument was calibrated against an external reference instrument (Thermo TE49i) at the University of York. All ozone concentrations are corrected for instrumental differences and dilution differences between measurements before and after the vessel. Ozone concentrations reported are the concentrations measured downstream of the reaction vessel.
Where we used buffered KI solutions and artificial seawater solutions (AS), the nominal ozone concentration is the concentration over the solution before addition of any iodide. For the natural samples (SSW and SML), the ozone concentration reported is the concentration downstream of the cleaned, empty vessel prior to the addition of seawater at the start of an experiment. Typically, ozone losses of approx. 3.5 ppbv were seen over the aqueous solutions containing no iodide. One experiment presented, using a solution of KI, was performed in the absence of an ozone monitor but with the same settings as an experiment done the previous day at [O3] = 35.9 ppbv. Therefore, the ozone concentration for this experiment is assumed to remain the same. Due to the uncertainty on eventual wall losses we set the error on the ozone concentration to the maximum losses (4ppbv) observed over all experimental runs.

Halocarbon analysis:
Ozone was produced by exposing dry scrubbed air to a Hg UV lamp of a commercial ozone generator (185 nm excitation, UVP, U.K.). Ozone concentrations before or after the vessel were measured using a commercial dual-cell ozone monitor (model 205, 2B-Tech, U.S.A.), before and after the experiments, as shown in Fig S2. During the halocarbon sampling, ozone in the reaction vessel could not be monitored due to the high gas flow needed to supply the O3 monitor. Halocarbon sampling was done for 35 minutes by directing the flow through heated and covered PTFE tubing from the vessel towards an air server coupled to a thermal desorption unit (CIA-8, Unity-2, Markes, U.K.). There the 100 cm 3 min 1 sample flow was directed over a Peltier cooled (30 ˚C) adsorbent trap (Tenax, Markes, U.K.). After sampling, the trapped compounds were desorbed by heating the trap to 300˚C for 7 mins and injected on the column (DB5-MS-UI column, 60m x 250 μm internal diameter (I.D.), 0.25 μm film thickness, Agilent) of a gas chromatograph (Agilent 6850). Separation of the halocarbons was obtained using the following oven program: start at 40˚C for 2 min, heating at 50 ˚C min 1 to 130˚C, held for 0.5 min, heating at 10 ˚C min 1 to 165˚C, held for 2 min, ramped up at 10 ˚C min 1 to 200 ˚C, held for 2 min and finally heated at 40 ˚C min 1 to 240 ˚C for 2 min. The separated analytes were detected using a quadrupole mass spectrometer (Agilent 5975C) operated in selective ion monitoring (SIM) mode, using 4 selected ion windows to scan for a quantifier and qualifier ion per analyte (except for CH3I and C2H5I). Table S3 shows the retention times and ions used for each analyte. The MS source and quadrupole temperatures were kept at to 250˚C and 200˚C respectively.
The analytes were quantified using the Mass Hunter software and two different standards. Halocarbons were quantified using the NOAA (National Oceanic and Atmospheric Administration) gas standard SX-3581 (created in 2014) in daily calibrations with repeat injections between samples. Although the standard is relatively old, it should be noted that the stability of the halocarbons in the NOAA standard is known to not drift significantly over one year (< 2 to 5% yr 1 ). 10 Moreover, for most compounds (CH2BrCl, CH2BrI, CH2ClI, CH2I2), the mole fractions are reported to have stayed consistent since 2009 or longer (CH3I, CH2Br2, CHBr3) although some drift cannot totally be excluded. 11 The NOAA standard does not contain iodo-ethane (C2H5I), nor either isomer of iodopropane (2-C3H7I, 1-C3H7I) and therefore these were only measured qualitatively. In addition to the NOAA standard, a gravimetrically made, in-house standard was used containing higher concentrations of halocarbons for certain experiments. Errors reported reflect the uncertainty on the quantification of the halocarbons. Limits of detection (3 ) are calculated using the average background levels detected in scrubbed air blanks before and after the samples.

Modelled iodine emissions: the sea-surface model
The interfacial model described in Carpenter et al. 12 was used, with some modifications to estimate I2 (and HOI) emissions from the experiments described. Iodine production is initiated by the gasphase flux of O3 into the interfacial layer, FO3 : Assuming that the O3 + I − reaction takes place entirely as an aqueous reaction, then the change in O3 concentration in the interfacial layer, [O3(int)], is: where A/V is equal to the inverse of the depth of the interfacial layer, defined as the reacto-diffusive length δ over which the O3 + I − reaction can occur. 12 The effective ozone uptake coefficient for the lab experiments, γeff (and hence the deposition velocity νD = ω γeff / 4, where ω is the thermal velocity of O3), was calculated according to (e.g. ref 13 ): where γI-is the ozone uptake coefficient for the O3 + I − reaction and 1/Γdiff is the gas phase diffusion resistance. 14 We assume that γI-is controlled by the aqueous-phase O3 + I − reaction as described by Eq. (1) in the main text. We use the value k I− in Eq. (2) in the main text, from Liu et al. 15 16 predicts an approximately equal surface and aqueous (bulk) contribution to the heterogeneous interactions of ozone with iodide. As a sensitivity study, we included total surface and aqueous O3 uptake in our model 16 , without any changes to the iodine emissions scheme, i.e. , and that those are the experimental conditions used in work, we did not include surface reactivity for the remainder of this work.

S7
The interfacial layer was treated as a box, assuming no horizontal advection (i.e. assuming horizontal gradients to be small) but mixing vertically with underlying "bulk" water at a fixed interfacial layer turnover time of 0.12 s 1 .
According to the two-resistance model for air-water partitioning, the mass flux F (mol m -2 s -1 ) of a trace gas species is described by ( 18,19 ): where cw and cg are the respective water and gaseous concentrations, H is the dimensionless gas-overliquid form of the Henry's law constant and KT is the overall mass transfer coefficient (m s -1 ). As in our previous work, 12 we assumed that I2 was supersaturated, i.e. F = KT cw.
KT is defined by equation (6): where Kw and Ka are the respective liquid phase and air-phase mass transfer coefficients, the inverse of which represent the liquid phase and air-phase resistances.
For sparingly soluble molecules (such as I2 in water) the rate of mass transfer is dominated by the water-side resistance. If solubility is increased, then the air-side resistance can be significant, and reduce the total mass transfer.
As in our previous work, 12 we calculated Ka for our laboratory experiments using an empirical formulation relevant for laminar flow conditions/indoor environments (ref 20 ) based upon the dimensionless Sherwood number, Sh, which is a function of the temperature-dependant Schmidt number of the gas in question in air (Sca, dimensionless), and the Reynold's number Re: Here Ka is in m h -1 , L is the characteristic length (m) calculated from the square root of the source area, and Da is the diffusivity in air (m 2 h -1 ).
To calculate the liquid mass transfer coefficient Kw (in m h -1 ), we again use the empirical approach of ( 20 ) applicable to still water, where Dw is the liquid phase diffusion coefficient (m 2 h): The total mass transfer coefficient KT was calculated from Eq. 6 using the temperature-dependent Henry's law coefficients (cg/cw) for I2 in water (9.3 x 10 -3 at 17 o C; refs. 21,22 ), giving a KT of 1.04 x 10 -6 m s −1 . As described in the main text, we found that reducing the model's aqueous-air mass transfer term to 4 × 10 −7 m s −1 (i.e. ~40% of the pure water transfer term) produced good agreement between the model and the observed I2 fluxes from SSW. Assuming that this reduction in the total mass transfer term was entirely due to increased solubility, i.e. reduction in H, this equates to an enhancement of I2 solubility in seawater of about a factor of 6 (i.e. H reduced from 9.3 x 10 -3 to 1.5 x 10 -3 ) , equivalent to solubility increasing from 0.03 g/kg (value in pure water) to ~0.2 g/kg.    Note that the scale on the Y-axis is different in both panels. All experiments at 17C. The errors bars reflect the overall uncertainty on the BBCEAS measurements, as detailed in section 1.6 above.  (see table S2 for more details).