Halogen species record Antarctic sea ice extent over glacial – interglacial periods

Sea ice is an integral part of the earth’s climate system because it affects planetary albedo, sea-surface salinity, and the atmosphere–ocean exchange of reactive gases and aerosols. Bromine and iodine chemistry is active at polar sea ice margins with the occurrence of bromine explosions and the biological production of organoiodine from sea ice algae. Satellite measurements demonstrate that concentrations of bromine oxide (BrO) and iodine oxide (IO) decrease over sea ice toward the Antarctic interior. Here we present speciation measurements of bromine and iodine in the TALDICE (TALos Dome Ice CorE) ice core (159 11 E, 7249 S; 2315 m a.s.l.) spanning the last 215 ky. The Talos Dome ice core is located 250 km inland and is sensitive to marine air masses intruding onto the Antarctic Plateau. Talos Dome bromide (Br) is positively correlated with temperature and negatively correlated with sodium (Na). Based on the Br /Na seawater ratio, bromide is depleted in the ice during glacial periods and enriched during interglacial periods. Total iodine, consisting of iodide (I ) and iodate (IO−3 ), peaks during glacials with lower values during interglacial periods. Although IO−3 is considered the most stable iodine species in the atmosphere it was only observed in the TALDICE record during glacial maxima. Sea ice dynamics are arguably the primary driver of halogen fluxes over glacial–interglacial timescales, by altering the distance between the sea ice edge and the Antarctic plateau and by altering the surface area of sea ice available to algal colonization. Based on our results we propose the use of both halogens for examining Antarctic variability of past sea ice extent.


Introduction
The atmospheric chemistry of iodine (I) and bromine (Br) are receiving increasing attention on account of their importance for photochemical atmospheric processes.Recent developments in atmospheric iodine chemistry include gas phase reaction kinetics, breakthroughs in the understanding of iodine reaction cycles in ambient air and ensuing CCN (cloud condensation nuclei) production (Saiz-Lopez et al., 2012).Bromine is important for its ozone depletion capabilities (Simpson et al., 2007b) and associated applications in atmospheric research (Wofsy et al., 1975).
Though the oceans are considered the main reservoirs and emission sources of I and Br (Greenwood, 2005), satellite observations (Schönhardt et al., 2008(Schönhardt et al., , 2012;;Kaleschke et al., 2004;Frieß et al., 2004;Nghiem et al., 2012;Saiz The main source of atmospheric iodine is oceanic biogenic production (Wong, 1991;Saiz-Lopez et al., 2012), in which iodine is injected into the atmosphere as organic iodine compound (R-I) that is subsequently oxidized into inorganic species in the presence of ozone and light (Vogt et al., 1999).However, recent laboratory data suggest additional abiotic production lend a non-negligible contribution of I 2 to atmospheric iodine (Saiz-Lopez et al., 2012;Carpenter et al., 2013).Recently it has been postulated that algae growing under sea ice also heavily contribute to the total iodine concentration in the atmosphere (Saiz-Lopez et al., 2007b;Atkinson et al., 2012) especially in the Antarctic region.
Iodine is a highly reactive element and undergoes rapid photochemical reactions in the polar atmosphere.In Antarctica, thin seasonal sea ice acts as a substrate for algal growth and permits the percolation of iodine compounds up to the atmosphere.Satellite images clearly demonstrate enhanced concentrations of IO above circum-Antarctic sea ice during springtime (Schönhardt et al., 2008).The production of IO is attributable to the photochemical degradation of compounds such as CH 2 I 2 and I 2 ; the resulting iodine atoms then react with O 3 .Iodine monoxide can also promote autocatalyzing reactions.For example IO can react with another IO molecule producing I 2 O 2 and OIO, a precursor of IOPs (iodine oxide particles) which then accumulate into CCN.IO can also react with other halogens such as BrO and ClO, which enhances their ozone-depleting potential (Saiz-Lopez et al., 2012).
Photochemical atmospheric reactions transform iodine and organic iodine to iodide (I − ), iodate (IO − 3 ), all of which have been found in atmospheric aerosols (Vogt, 1999).Current models of atmospheric chemistry (McFiggans et al., 2004) indicate that the aerosol I − concentration is negligible due to its transformation into species that revert to the gas phase by reaction with ozone.Instead IO − 3 is relatively inert and therefore readily accumulates in particles (Pechtl et al., 2006;Saunders et al., 2012b).Iodine speciation in water varies: in anoxic water most of iodine exists as I − , while in oxygenated waters, such as surface (300 m) ocean water, the dominant species is IO − 3 .Bluhm et al. (2010) demonstrated that phytoplankton can reduce IO − 3 to I − , while UV light can also drive the reduction of IO − 3 to I − in seawater (Saunders et al., 2012b;Wong, 1991).Doubts remain regarding iodine inorganic speciation in aerosols and which factors are able to control this speciation (Saiz-Lopez and Plane, 2004); UV light and ozone have been proposed to play important roles (Saunders et al., 2012a).
In polar regions bromine is mostly emitted as sea spray aerosol by wave breaking but a connection between bromine and ozone depletion was identified in the 1980s.Oltmans (1981) was the first to observe surface ozone variations on timescales of hours to years.Connections between atmospheric bromine excess and ozone depletions were suggested later when Barrie et al. (1988) measured an "excess of filterable bromine" that could not be explained by windblown sea salt.A strong negative correlation between ozone and bromine was detected suggesting a linear correlation between the two compounds.Bromine-induced ozone depletion events (ODEs) are started by photochemical activation of bromide present in the sea salt causing a cascade of chemical reactions able to drastically reduce the ozone concentration in the boundary layer (Simpson et al., 2007b).It has been suggested that sea-salt bromide is present in young seaice surfaces, frost flowers, snow pack, sea water, and marine aerosols due to its photochemistry which converts Br − into reactive halogen species (e.g.Br atoms and BrO) that are able to deplete ozone in the remote atmosphere (Simpson et al., 2007a).ODEs are caused by so-called "bromine explosions", a series of photochemical and heterogeneous acid catalyzed reactions.Vogt et al. (1996) proposed a mechanism for bromine activation where sulfur and nitrogen acid species, such as MSA, SO 2− 4 and NO x , are able to activate the reactions between ozone and bromide producing the gas phase species BrO, Br 2 and HOBr.The bromine explosion is strongly dependent on pH because the key reaction of HOBr with Cl − or Br − is pH dependent (Ayers et al., 1999;Simpson et al., 2007a).Satellite measurements show enhanced atmospheric bromine oxide (BrO) in the polar spring, coinciding with an ozone minimum measured by ground-based instruments (Simpson et al., 2007a).The bromine explosion may be promoted by the presence of frost flowers that grow above the young sea ice; they are ice crystals of extremely high salinity and are an efficient substrate for the atmospheric emission of chemicals due to their high surface area (Kaleschke et al., 2004).Recently it has also been proposed that fresh snow covering first year sea ice could be a source for gas phase bromine (Pratt et al., 2013).
The current utility of polar ice cores for quantifying sea ice variability is minimal, due to the lack of a sufficiently reliable and specific proxy.Methanesulphonic acid (MSA) has been used to estimate sea ice variations over the last century at Law Dome in coastal Antarctica (Curran et al., 2003); however, this compound is unstable and easily remobilised in the ice matrix after deposition (Smith et al., 2004) limiting its utility for long-term sea ice reconstructions.Further, it has been demonstrated that MSA is not a reliable indicator for sea ice extent at some other coastal sectors of Antarctica (Abram et al., 2011).Sea-salt sodium has also been used to quantify sea ice variations (Wolff et al., 2006) based on the formation of high salinity frost flowers and brine (Rankin et al., 2000) on sea ice surfaces.Such influences are difficult to distinguish from the dominating sodium input from sea spray aerosol (Petit et al., 1999) and open waters have been shown to be the dominant source of sodium deposited at many Antarctic sites, e.g.Law Dome (Curran et al., 1998) and James Ross Island (Abram et al., 2011).
Even though iodine and bromine species have already been measured in polar snow, to our knowledge no measurement of this species has yet been carried out in Antarctic ice cores.Based on the method developed by Spolaor et al. (2012), we present here the first extended (200 ky) record of iodine and bromine species from the Talos Dome ice core and evaluate the implications of these halogen species for reconstruction of sea ice variability on millennial timescales.

Ice core samples
The Talos Dome ice core drilling site (159 • 4 21 E, 72 • 47 14 S; 2318.5 m a.s.l.) is located over approximately 1880 m of ice, the deepest of which has been dated to 250 ky before present (BP) using the TALDICE-1a (Buiron et al., 2011;Schupbach et al., 2011).Featuring a mean annual temperature of −41 • C and an average annual accumulation rate of 80 mm water equivalent, the site is an excellent location for the long-term preservation of climatic signals of southern hemispheric climate change.There are currently only a few ice cores from coastal sites (EPICA-DML, Berkner Island, Taylor Dome, Siple Dome) that record Holocene and post-Holocene climate change or the retreat of the ice sheet since the Last Glacial Maximum (LGM).The coastal position of the Talos Dome (TD) ice core makes this core important for studying marine-related processes during the Antarctic deglaciation.Analyses were conducted on discrete samples of Talos Dome ice sampled for continuous flow analysis (Kaufmann et al., 2008).For each 1 m long ice sample, 15 mL was collected in coulter counter accuvettes at a rate of 0.5 mL min −1 .The samples were kept frozen and shielded from light until analysis.

Determination of halogens, sodium and iron
Iodine (I − and IO − 3 ) and bromine (Br − ) species were measured at picogram-per-gram levels by coupling highperformance liquid chromatography (HPLC), ion chromatography (IC) and inductively coupled plasma mass spectrometry (ICP-MS).All reagent solutions and samples were prepared in a shipping container converted into a particlefree laboratory.The air qualities ranged from classes 10 to 100 (US Fed.Std 209E, equivalent to ISO 4 to 5).Chromatographic separation was achieved using an IONPAC ® AS16 Analytical Column (Dionex, US) with 35 mM NaOH 99.99 %, Sigma-Aldrich (Missouri, USA) eluent.The flow was set to 0.280 mL per min with a run time of 9 min and 70 bar operating pressure.The samples were melted without acidification to avoid any Br or I volatilization and a volume of 0.3 mL was used for analysis.The detector was an Agilent 7500 series ICP-MS (Agilent, CA, USA) with mono-isotopic I measured at mass 127 and Br measured at mass 81.Detection limits for I and Br species were 5 to 9 pg g −1 with an uncertainty of less than 2.5 % for all considered species.Iodide had a recovery of 80 % while Br − and IO − 3 had higher recoveries around 90 %.For the determination of sodium and iron concentrations, samples were melted and acidified to pH 1 using sub-boiling distilled HNO 3 (Romil, Cambridge, UK) and analysed at least 24 h later by Inductively Coupled Plasma Sector Field Mass Spectrometry (ICP-SFMS; Finnigan ™ ELEMENT2, Thermo Fisher Scientific Inc., Bremen, Germany) coupled to an APEX Q desolvating introduction unit (Elemental Scientific, Omaha, NE, USA).The detection limit of these two elements, defined as three times the standard deviation of the instrument blank, was 0.029 ng g −1 , with precisions of 16 % for Holocene samples and 10 % for Last Glacial Maximum (LGM) samples.

Iodate stability experiment
A number of experiments (Table 1 lists the experimental parameters), were conducted to determine if IO − 3 is unstable in ice under near-UV/visible light irradiation.These experiments were conducted in a photo-chemical flow reactor (PCFR) under conditions described in detail by Saiz-Lopez et al. (2004) and Saunders and Plane (2005).Solar irradiance was simulated between wavelengths 300 to 750 nm (λ max = 583 nm) using a 1000W Oriel ® Xenon Arc Lamp Solar Simulator (Oriel Instruments, Model 91191) with a water filter for removal of infrared light.The PCFR featured a flow cell with an inner volume of 0.2 L and equipped with an external cooling jacket filled with methanol (Fig. 1).The flow cell was half filled with 0.1 L of 10 −5 M IO − 3 solution (produced from sodium salt) and subsequently cooled to −20 • C until the solution was frozen.The temperature was kept at −20 ± 2 • C for the duration of the experiment (4 h).The gasphase iodine species produced in the flow cell, resulting from the initial experiment conditions, were then converted into iodine oxide particles (IOPs) in a downstream photolysis cell by the presence of visible light from a tungsten lamp, and O 3 (produced by illuminating a flow of O 2 with a low pressure Hg pen ray lamp at 180 nm).Briefly, production of IOPs first require the formation of gas phase iodine as IO from the oxidation of I by O 3 ; IO then self-reacts to form OIO, which can then go on to form higher oxides which polymerize to generate IOPs (Saunders et al., 2010;Saiz-Lopez et al., 2012).The IOP size distribution, and hence total mass, was measured using an electrical mobility spectrometer (EMS VIE10: Tapcon GmbH), consisting of a nano-differential mobility analyzer (nano-DMA) and a Faraday cup electrometer (FCE).Particles were counted in the range from 2 nm to 40 nm in diameter.
A 650 sccm (standard cubic centimeter per minute) flow of pure N 2 gas was used as the carrier gas for the PCFR system with an O 3 concentration of 78 ppbv.The gas flow in the flow cell was maintained at 50 sccm of pure N 2 except for experiment 4 in which the gas flow was composed of a mixture of 40 sccm of N 2 and 10 sccm O 3 , giving a total flux www.atmos-chem-phys.net/13/6623/2013/Atmos.Chem.Phys., 13, 6623-6635, 2013 Table 1.IOP (iodine oxide particle) production experiment conditions and results.IOP mass productions are expressed as g cm −3 measured for 180 min (3 h).Abbreviations: Light -solar simulator, CG -carrier gas, BF -blue filter, O 3 -ozone in the flow cell, RC -reaction cell, Sol -solutions (W: water, HA: humic acid), an "x" represents an active parameter while "-" represents an inactive parameter. of 50 sccm.To compare the IOP mass production for each experiment, reported in Table 1, we used the sum of IOPs produced in 180 min (expressed in g cm −3 ) while the blank was assessed by repeating the experiment three times using only N 2 as carrier gas in the absence of light, thus obtaining a value of 4.3 ± 1.0 × 10 −13 g cm −3 (Table 1).
To avoid potential interference from the presence of other iodine species, all glassware was carefully cleaned with deionized water and a spectrophotometer analysis in the 200-600 nm range was carried out to ensure the absence of any such species (Fig. 2).The PCFR system was cleaned by fluxing N 2 and O 3 for 2 h (the time necessary to reach a stable blank value) thus removing any iodine compounds that may have resided on the inner tube surface.The blank signal was checked before each experiment to ensure that no contaminants were present in the system.Although it has been demonstrated that IO − 3 is stable in water at least for 30 h in the absence of humic acid or hydrogen peroxide (Saunders et

Statistical data analysis
Confidence intervals for Pearson correlations can be computed with the Fisher z transformation approach.However, this statistical method requires a pair of variables with a bivariate normal distribution.Such an assumption is not met by TD elemental and species data whose sample distributions differ substantially from the normal shape.A more robust approach for confidence intervals of correlations that does not rely on the normality assumption is the nonparametric bootstrap percentile method (Efron, 1987).The nonparametric bootstrap is a simulation method which consists of random sampling with replacement from the original data in order to quantify uncertainty of a sample estimate, such as the Pearson correlation index.The R package "boot" (Davison and Hinkley, 1997) provides a convenient implementation of this simulation approach.In this paper, we employ the nonparametric bootstrap percentile method using 1000 bootstrap replicates.

Bromide
Talos Dome (TD) bromide concentrations range from below the detection limit (9 pg g −1 ) to 200 pg g −1 , corresponding to fluxes from < 0.3 to 17.8 µg m −2 yr −1 (Fig. 3).The greatest Br − fluxes occurred during interglacial periods with an average value of 9.5 µg m −2 yr −1 , while during glacials the average flux decreased to 2.9 µg m −2 yr −1 (Fig. 3).The sodium (Na) content was used to evaluate the influence of maritime air masses on Antarctic deposition (Wolff et al., 2006).Na is a conservative element that is not transformed after sea spray emission (Simpson et al., 2005;Ayers et al., 1999), whereas Br − is not stable in sea-salt aerosols and is depleted with respect to seawater (Ayers et al., 1999).
Bromide concentrations in the TD ice core negatively correlate with Na concentrations (Fig. 3 We use Na as a sea-salt proxy (Wolff et al., 2010) to calculate the bromide fractionation index (%Br fi ) with respect to sea-salt concentrations using the formula %Br fi =([Na]•0.006−Brmeasured )•100/([Na]•0.006), which assumes a constant Br/Na mass ratio of 0.006 (Millero, 1974) for seawater and fresh sea spray through time (Ayers et al., 1999).The glacial %Br fi is positive, (mean 68 %, up to 100 %) while interglacial %Br fi values are negative (up to −440 %) (Fig. 4).Talos Dome %Br fi and sea ice presence (SIP) (Crosta et al., 2004)  BP to MIS 7, 240 ky BP) show similar %Br fi values.Factors that may influence %Br fi include the acidity of the aerosols catalyzing the bromide depletion (Ayers et al., 1999;Vogt et al., 1996), and atmospheric residence time which is linked to the distance from source to sink, i.e. sea ice extent.Bromine explosions cause an excess of total (gas and aerosol phase) bromine compared to sea salt in the atmospheric boundary layer (Simpson et al., 2005).During spring, bromine explosions are favoured at the seasonal sea ice surface by the presence of salts from the freezing seawater process and light mediated activation.The reduced extent of multi-year sea ice during interglacials (present times) implies that bromine explosions are located nearer to the Antarctic coast during such climatic phases.Bromine explosions result in an increased atmospheric Br − /Na ratio with respect to seawater (Simpson et al., 2005), and these Br − enriched air masses are then transported by the prevailing winds to TD (Fig. 5).On this basis, we hypothesize that an extended multi-year sea ice surface during glacial periods increased the distance between bromine explosions and the TD ice core site.HBr is the end product of the bromine atmospheric chemistry chain (Vogt et al., 1996).We assume that the deposition velocities of gas-phase HBr and sea-salt aerosols to the snowpack lead to differential deposition rates as these species are carried inland by cyclonic winds.When seasonal sea ice is located farther from the Antarctic landmass, the increased atmospheric travel time allows the quickly deposited HBr to deplete relative to atmospheric sea-salt aerosol.Thus the Br − enhancement observed during interglacial periods changes to a Br − depletion in colder climates because of the reduced proximity to multi-year sea ice.This hypothesis is supported by Br − and Na measurements in inland interglacial EPICA Dome C (75 • 06 S, 123 • 21 E; m a.s.l.3233) ice core samples showing small depletions of Br (%Br fi = 14 %) compared to an enrichment of −63 % detected in TD ice.
If Br − /Na ratios and %Br fi are linked with air mass travel times it is important to consider changes in atmospheric circulation between cold and warm periods.Hemispheric-scale models, as well as observations of dust particle sizes, indicate only modest changes in surface winds, atmospheric circulation, and meridional transport over Antarctica between glacial and interglacial climates (Petit and Delmonte, 2009;Fischer et al., 2007).Although the deglacial retreat of the Ross Ice Shelf has been related to activation of local dust deflation zones (Albani et al., 2012), it has been observed that only 25 % of air mass trajectories arriving at TD originate from this sector: the majority of air masses at TD originate from the westerly Atlantic-Indian Ocean sector (Scarchilli et al., 2011).The constant Na fluxes observed throughout the Holocene at TD (Fig. 3) are indicative of a constant flux of marine air masses during this period, suggesting that any deglacial changes to regional atmospheric circulation did not considerably affect the influence of marine air masses on TD.

Bromide model calculation and outcome
The purpose of the modelling described in this section is to provide a plausible explanation for the variation in %Br fi measured at Talos Dome over several glacial cycles.It is not possible to demonstrate unambiguously that the variation in sea-ice extent caused these variations by using a simple 1-D transport model where several parameters are poorly known (as discussed below).Rather, our objective is to show that by choosing sensible values of these parameters it is possible to explain the variations in %Br fi levels, and thus to estimate the change in the sea-ice extent between glacial and interglacial periods.
The Tropospheric HAlogen chemistry MOdel (THAMO) is a 1-D chemistry transport model (Saiz-Lopez et al., 2008) with 200 stacked boxes at a vertical resolution of 1 m (total height 200 m).The model was developed to interpret measurements made at Halley Bay (Antarctica) of IO and BrO in the coastal boundary layer (Saiz-Lopez et al., 2007b).The model treats iodine, bromine, O 3 , NO x and HO x chemistry using over 210 reactions.The chemical scheme is from Saiz-Lopez et al. (2008) and has been updated by Mahajan et al. (2010).The model is used here to follow the evolution of the halogens in an air parcel which crosses the multi-year sea-ice edge and then travels across the sea ice and coastal margin to Talos Dome.If the average wind speed is 9 m s −1 (Scarchilli et al., 2011), then it would take about 7 h to transport air from the sea-ice edge to Talos Dome (∼ 230 km).This corresponds to the interglacial situation.During glacial times, the air would travel about 780 km in 24 h, which is similar to the distance from the sea-ice edge to Dome C at present (Scarchilli et al., 2011).
The following conditions were used, mostly based on the extensive CHABLIS field campaign at Halley Bay (Jones et al., 2008).This coastal Antarctic station is roughly 20 km from the sea ice edge, and so the large suite of atmospheric chemistry and physics measurements made during CHABLIS should be appropriate for the present modeling exercise.The temperature was set to 250 K at the base of the boundary layer, and the vertical eddy diffusion coefficient (K z ) was set to be constant (3 × 10 4 cm 2 s −1 ), based on the wind speed vertical profile at Halley (Saiz-Lopez et al., 2008) (Read et al., 2007).The modelled HO x concentrations ([OH] = 1.3 × 10 6 cm −3 [HO 2 ] = 1.7 × 10 8 cm −3 at midday) are in sensible accord with measured values at Halley Bay (Saiz-Lopez et al., 2008).
The initial sea-salt aerosol loading was set to 40 particles cm −3 of 0.5 µm diameter, providing a volumetric particle surface area of 3.0 × 10 −7 cm 2 cm −3 and a volumetric particle volume of 2.6 × 10 −15 dm 3 cm −3 .The corresponding mass concentration is 2600 ng m −3 .This concentration is somewhat higher than the 4 yr average of 1400 ng m −3 measured at the coastal Antarctic station of Dumont D'Urville, which was stated to be a lower limit because of sampling size cut-offs (Wagenbach et al., 1998).The prevailing wind at Dumont D'Urville is directly onshore (average 9.5 m s −1 ) and the station is on a small island, and so provides a reasonably close match to the model scenario.The initial Br − concentration was assumed to be that of seawater, 65 mg L −1 , so the Br − /Na + elemental ratio was 1.5 × 10 −3 .The deposition velocities of HBr and HI to the snowpack were set to 1 cm s −1 .This would be typical for a hydrophilic acidic gas (e.g.HNO 3 ) at an average wind speed of 9 m s −1 over a smooth surface (Ma and Daggupaty, 2000).The deposition velocity of the sea-salt aerosol was set to 0.2 cm s −1 .The deposition velocity is quite size dependent (being larger than 0.2 cm s −1 for particles smaller than 0.1 µm or larger than 4 µm, and reaching a minimum of about 0.1 cm s −1 for 2 µm diameter particles (Ma and Daggupaty, 2000)).This deposition velocity therefore represents an average over the sea-salt size distribution, and is treated as a fitted parameter in the model.This combination of deposition velocity and K z profile in the model gives an e-folding lifetime for the removal of sea-salt aerosol from the BL of around 30 h.
The ice core record contains the integrated annual accumulation of Br − and Na.The annual BrO record in the boundary layer at Halley Bay exhibits active springtime bromine explosion chemistry averaging around 10 ppt of BrO, less bromine in the summer and autumn (around 4 ppt), and essentially nothing during the polar night (Saiz-Lopez et al., 2007b).In order to account for this seasonal variation, 1 ppt of Br 2 was injected into the first 40 m of the boundary layer at time zero, when the air parcel crossed the sea-ice edge.This injection simulated the effect of a time-averaged "bromine explosion".An I 2 flux of 7 × 10 7 cm −2 s −1 was required to generate an IO concentration of 1-2 ppt while the air parcel was over the sea ice (cf. a flux of 5 × 10 8 cm −2 s −1 is required to generate the average of 6 ppt measured at Halley during summer (Saiz-Lopez and Boxe, 2008).The uptake of iodine species  such as HOI and IONO 2 on sea-salt aerosol converts aerosol bromide into gas-phase BrO, etc.
Figure 6 compares the modelled deposition fluxes with the measurements at Talos Dome and Dome C, using the travel times from the sea ice edge (Talos Dome = 7 h during interglacials, 22 hrs during glacial; Dome C = 24 h during interglacials).Using the assumptions listed above, good agreement is achieved between this simple model and the measured fluxes.An important conclusion is that, for the model to predict the change in Br dep with travel time from the seaice edge, the deposition velocity of HBr needs to be at least 3 times larger than the average sea-salt aerosol deposition velocity, which is consistent with dry deposition theory and measurement (Ma and Daggupaty, 2000).While there are too many assumed parameters in the model to constrain them rigorously, the purpose of this exercise is to show that the observed changes in fluxes are plausibly explained by a change in the distance of the sea-ice edge from the continent.
Our model calculations demonstrate the dependence of Br − /Na ratios on air mass transport times where the transition from Br enrichment to depletion is approximately 8-9 h (Fig. 7).Considering an average air-mass speed of 850 km day −1 (Scarchilli et al., 2011) and the average Holocene %Br fi value (−60 %), we calculate that the air masses travelled approximately 300 km, the distance between TD and the coast.The 70 % depletion observed during glacial times would require a travel time of 700-800 km suggesting that the young sea ice was likely located 500 km farther north than the interglacial extent, which is consistent with other estimates of the Antarctic sea ice in the literature (Gersonde et al., 2005).
While the iodine record is consistent with the Antarctic spring sea ice extension (Schönhardt et al., 2012), other factors may have contributed to the glacial-interglacial changes in I tot .Such factors include dust-induced ocean fertilization (Martin, 1990) and potential algal population changes.The iodine fluxes doubled between the two investigated glacial periods: 3.6 µg m −2 yr −1 (MIS 2 to MIS 5) and 1.6 µg m −2 yr −1 (MIS 6 to MIS 7), and correspond to dust flux changes and therefore the potential aeolian nutrient supply.We evaluated correlations between I tot flux and Fe (aeo-  19,0.10]).Aeolian ocean fertilization appears to impact atmospheric iodine concentrations less than the influence of sea ice extent.The differences between the two glacial stages may also be due to the slightly warmer glacial between MIS 5 to 7 (mean δ 18 O -40.1 ‰) compared to MIS 2 to 4 (mean δ 18 O -40.6 ‰) (Stenni et al., 2011).These local temperature data agree with the LR04 global benthic marine stack (Lisiecki and Raymo, 2005) as a proxy for global ice volume.Assuming a linear relation between sea ice and I 2 production, we calculate that the LGM summer sea ice extent was approximately double that of the penultimate glacial maximum.

Iodine speciation and iodate reactivity
Iodate is the most stable iodine species in the atmosphere (Saiz-Lopez et al., 2012;Pechtl et al., 2006) although the ) from the frozen IO − 3 solutions under different experimental conditions as summarised in Table 1.The blank is shown at the beginning of each experiment.Abbreviations: Light -solar simulator, CG -carrier gas, BF -blue filter, O 3 -ozone in the flow cell, RC -reaction cell, Sol -solutions (W: water, HA: humic acid).An "x" represents an active parameter while "-" represents an inactive parameter.atmospheric iodate/iodide ratio is variable (Baker, 2004;Gilfedder et al., 2008).Iodide is present throughout the TD record, from interglacial fluxes of 1.1 µg m −2 yr −1 to glacial fluxes of 2.9 µg m −2 yr −1 , while IO − 3 is only present above the detection limit during the glacial maxima (Fig. 8).Two iodate horizons between 16.8-33.8ky BP and 134.2-143.2ky BP contain concentrations up to 30 times more than the detection limit.Observational evidence (Baker et al., 2001) and model calculations (Pechtl et al., 2006;Saunders et al., 2012b) suggest that IO − 3 should accumulate in marine aerosol while iodide should be considered negligible due to its transformation into species that revert to the gas phase after reaction with hypohalous acids (HOX, X = Cl, Br, I) to form volatile IX (Baker, 2004) by reaction cycles proposed by Vogt et al. (1999).However, disagreement exists between observations of the atmospheric ratio of I − /IO − 3 , where a study from the tropical Atlantic Ocean detected a majority of IO − 3 and negligible I − (Saiz-Lopez and Plane, 2004) whereas another study found non-negligible I − concentrations in aerosols (Baker, 2004).Although IO − 3 has been considered an inert inorganic iodine species our results suggest substantial reactivity.In Talos Dome ice samples, IO − 3 is present only during glacial maxima when dust fluxes were at their maximum and temperatures were at their minimum, while iodide is present in all samples.These observations suggest that in polar ice I − is the more stable species, while IO − 3 is present only under particular conditions.Considering the results of the IO − 3 stability experiments, it was found that IOPs were produced from aqueous solutions (experiments 2, 3, 4 and 6), but not from humic acid solu-tions (experiment 5), in the presence of light (Fig. 9).IOP formation results from the release of iodine gas compounds (likely HOI, IO and I 2 ) from the frozen solution surface indicating reduction of IO − 3 has occurred.It is important to note that only IO − 3 was dissolved in deionized water, so the IO − 3 reduction must have been photochemically driven.Though IO − 3 in aqueous media absorbs light only in the UV range (Awtrey and Connick, 1951) at wavelengths below 270 nm (Fig. 2), theoretical calculations indicate the photochemical threshold for the reaction IO − 3 + νv → IO(g) + O − 2 in ice is 315 nm which is consistent with the large IO concentration which has been observed spectroscopically within the polar snow pack (Frieß et al., 2010).In the presence of light, our experiments demonstrate the greatest IOP production, 1.0 × 10 −12 g cm −3 , with O 3 present in the N 2 carrier gas and the RC activated.The lowest IOP production occurred when the RC was not activated (7.1 × 10 −13 g cm −3 ).Similar intermediate levels of mass production, 8.5 × 10 −12 g cm −3 , were obtained in experiments 2 and 3 where the only difference was the absence of light of λ > 310 nm in experiment 3.This demonstrates that, under the experimental conditions, visible light did not contribute to IOP production.
It is evident from these experiments that the release of gasphase iodine from frozen IO − 3 solutions can occur, however it is difficult to distinguish which type of inorganic iodine was formed.Considering the UV-VIS spectra (Fig. 2) before and after irradiation it is likely that I 2 was present, as suggested by the increased absorbance in the range 270-400 nm where I − 3 and I 2 have absorbance maxima.The presence of other iodine gas species such as IO and OIO is suggested by a www.atmos-chem-phys.net/13/6623/2013/Atmos.Chem.Phys., 13, 6623-6635, 2013 comparison of the results of experiments 2 and 3 with experiment 6, in which the RC was deactivated.If only I 2 was produced no IOP formation would be possible since both ozone and visible light are necessary to convert I 2 to IOPs.As previously noted, IO and OIO are able to produce IOPs in the absence of O 3 and visible light.The presence of IO and OIO is suggested by the difference in mass production between the blank (4.3 × 10 −13 g cm −3 ) and experiment 6 (7.1 × 10 −13 g cm −3 ), in which light was present but the reaction cell was deactivated.The possibility exists that IO − 3 could be reduced to these forms and then dissolve in the liquid water present in ice crystal boundaries (Sazaki et al., 2012).This could explain the fact that the production of IOPs and hence the release of gas-phase iodine from ice is not uniform but presents an increase after at least 100 min (Fig. 9) suggesting a complex chemistry and diffusion of I 2 and likely IO through the matrix.
In light of these experiments we propose that IO − 3 , though it has been found stable in liquid water, is not stable in ice and is photoreduced to other iodine species, principally IO and I 2 but with the possibility of also other species such as I − , I − 3 and OIO.Regarding the presence of IO − 3 only in the dusty and cold glacial climate periods, it seems that increased dust fluxes (from 1-2 mg m −2 yr −1 during interglacials to 15-20 mg m −2 yr −1 during glacials, Lambert et al., 2008) favour the formation of metal iodates by interactions of gas phase iodine oxide molecules and dust and, together with decreasing temperature (10 • C lower from interglacial to glacial, Petit et al., 1999) could promote greater IO − 3 stability.The correlation (R = 0.77) obtained between IO − 3 and fine dust (0.7-5 µm) (Delmonte et al., 2004) suggests that during glacial periods gas-phase molecules such as I 2 O 4 attach to dust particles during transport.

Conclusions
We report the first bromine and iodine concentrations in ice cores as well their speciation.Based on the presented data and our simple Br depletion model, we estimate that glacial multi-year sea ice extended ∼ 500 km north of the Antarctic coast, with maximum extension from 16 to 80 ky BP and 134 to 205 ky BP.This interpretation is also consistent with the increased iodine fluxes observed during glacial periods, which imply an approximate doubling of LGM winter sea ice extent compared to the penultimate glacial maximum.In addition, our results seem to suggest a reduction of iodate after deposition that could lead to iodide formation.Further studies should be carried out at other coastal and inland stations of Antarctica to compare with the results presented here.
It is now well known that bromine and iodine chemistry are particularly active in the polar atmosphere and have been linked by satellite observations to Antarctic sea ice, as opposed to the interior of Antarctica where background levels are observed (Schönhardt et al., 2012).This may be due to the temperature dependence of bromine activation (Oldridge and Abbatt, 2011), in which the lower temperatures of the Antarctic plateau (TD mean temperature −41 • ) (Frezzotti et al., 2004) reduce the reactivity of bromine and iodine.It is important to further study the depositional and postdepositional chemical processes of halogens in polar ice to confirm the validity of these species as long-term proxies of sea ice location and extent.
-Lopez et Published by Copernicus Publications on behalf of the European Geosciences Union. A. Spolaor et al.: Halogen species record Antarctic sea ice extent al., 2007a) have linked I and Br activity to sea ice in the polar regions.

Fig. 1 .
Fig. 1.Schematic diagram of the photochemical flow reactor set up to study the stability/reactivity of IO − 3 in ice.The solid line represents the gas flow pattern, the dashed line shows the chilled methanol circulation.

Fig. 2 .
Fig. 2. Results of two iodine oxide particle (IOP) production experiments, showing the spectrophotometric spectra of 10 −5 M IO − 3 solutions in the presence of O 3 (experiment 4) and without O 3 (experiment 2).Blue lines represent the UV-VIS spectra at the end of the experiment while the red lines show initial conditions (fresh solution).IOPs were produced in both experiments, indicated by the increased absorbance in the range 270-400 nm where I − 3 and I 2 have absorbance maxima.

Fig. 3 .
Fig. 3. Plot of Br − fluxes (red) as a function of time in the TD ice core.Br − fluxes correlate with δ 18 O (a proxy for the temperature record -black curve) and anti-correlate with Na fluxes (blue curve).The inset box represents the sodium flux between 6 and 12 ky BP.Note that Sodium and δ 18 O are recorded as a high-resolution data set while bromide was measured at lower resolution.

Fig. 4 .
Fig.4.%Br fi is plotted with sea ice presence (SIP, green area, expressed as month/year) and February sea-surface temperature (red line)(Crosta et al., 2004).Talos Dome δ 18 O is also shown.The red areas indicate Br − enrichment (i.e.negative %Br fi ) while blue areas indicate bromide depletion.

Fig. 5 .
Fig. 5. Simplified scheme for bromide aerosol deposition in coastal Antarctica.Blue lines indicate the aerosol phase while red lines indicate the gas phase.Green line the HBr deposition.

Fig. 6 .
Fig.6.Comparison between the modelled deposition fluxes for sodium (black) and bromide (blue) with measurements at Talos Dome (Circles, 7 h from sea ice edge) and Dome C (Circles, 22 and 24 h from sea ice edge).

Fig. 7 .
Fig. 7. Variability in the observed Br/Na ratios expressed as a function of the air mass travel time due to the different deposition velocities of HBr and sea salt in the model.After 8 h of transport, bromide begins to be depleted relative to sea salt.

Fig. 8 .
Fig. 8. Iodate (IO −3 ,brown), iodide (I − , light blue) and total iodine (I tot , blue) in Talos Dome ice core samples.While I − is present in all samples, IO − 3 is closely linked with the fine dust load (grey shaded area,Delmonte et al., 2004) and is present only from 16.8 to 33.8 ky BP and 134.2 to 143.2 ky BP.

Fig. 9 .
Fig. 9.Rates of IOP particle production (in g cm −3 ) from the frozen IO − 3 solutions under different experimental conditions as summarised in Table1.The blank is shown at the beginning of each experiment.Abbreviations: Light -solar simulator, CG -carrier gas, BF -blue filter, O 3 -ozone in the flow cell, RC -reaction cell, Sol -solutions (W: water, HA: humic acid).An "x" represents an active parameter while "-" represents an inactive parameter.