Release of gas-phase halogens from sodium halide substrates: heterogeneous oxidation of frozen solutions and desiccated salts by hydroxyl radicals

Motivated by the need to determine the mechanism of the initial release of halogens from sea ice and marine aerosol substrates, a study of the interactions of OH radicals with a variety of halide-containing surfaces has been performed in a coated-wall flow tube using chemical ionization mass spectrometry for gas-phase analysis. The salts studied were NaCl with 0.01% and 0.002% impurities of Br− and I− respectively, and NaCl/NaBr mixtures with Cl−/Br− seawater ratios. The surfaces were desiccated salts, desiccated salts exposed to elevated relative humidity, and frozen solutions. In all cases, gas-phase Br2 and BrCl were formed, with the Br2 yield (defined as a molar ratio of halogen produced to OH lost) larger than 0.2 and the BrCl yield as roughly 0.01. For the first time, an observation of heterogeneous release of iodine-containing halogens (in particular IBr) was made with yields comparable to those of BrCl. We note that pH neutral frozen solutions demonstrated halogen release, although the yields were higher for acidic frozen solutions.


Introduction
It is well established that halogen chemistry plays a significant role in selected tropospheric environments (e.g. Honninger 2003, von Glasow et al 2004). Arctic ozone depletion events with concomitant increases of BrO (Barrie et al 1988) have been observed consistently over the past two decades. Subsequent field campaigns have further investigated the role of halogens at high latitudes (Hausmann and Platt 1994, Foster et al 2001, Avallone et al 2003 and other environments such as dry lake beds (Hebestreit et al 1999 and the marine boundary layer (Saiz-Lopez et al 2006, Bloss et al 2005.
Gas-phase chemistry has proven incapable of explaining the observed extended ozone depletion events. Therefore, 1 Author to whom any correspondence should be addressed. heterogeneous reaction mechanisms, such as the autocatalytic bromine explosion (Vogt et al 1996), have been invoked to provide an additional halogen source. In the marine boundary layer aged aerosols have displayed depleted bromine to sodium ratios with respect to seawater (Sander et al 2003). This suggests that bromine has been preferentially removed from the aerosol over its lifetime. The release of halogens from dry salts (Finlayson-Pitts 2003, Rossi 2003 has also been investigated in laboratory studies. Despite these studies, the mechanisms that release halogens to initiate these cycles remain ill-defined, which limits our predictive capabilities. There are a number of possibilities, including a dark reaction involving O 3 (Oum et al 1998b), interactions with NO x (Finlayson-Pitts et al 1990), and free radical processes (Mozurkewich 1995, Matthew et al 2003. In this regard, Frinak and Abbatt (2006) investigated halogen release from NaCl/NaBr solutions by heterogeneous oxidation involving hydroxyl radicals (OH). Both Br 2 and Cl 2 were released from acidic solutions. In this letter, we extend this work by examining the release of halogens by OH interactions with desiccated salts and frozen sodium halide solutions. Of interest is whether the relationship between acidity and halogen release is maintained on ice surfaces and whether halogen release occurs from neutral substrates, as described by Finlayson-Pitts (2003). To our knowledge, this is the first investigation of the interactions of OH with halide surfaces of this type.

Experimental details
Observations of gas-phase halogens were obtained with a chemical ionization mass spectrometer (CIMS) coupled to a coated-wall flow tube. This apparatus allows for controlled exposure of hydroxyl radicals to prepared surfaces, as described in detail in Frinak and Abbatt (2006). Briefly, the operating pressure of the flow tube was 90 Torr with a total flow rate of 1.5 lpm of N 2 . Water vapor mixing ratios were controlled by passing a fraction of the carrier gas through a bubbler before entering the flow tube. Within the CIMS ionization region, a trace flow of SF 6 in 10.5 lpm of dry N 2 passed over a 210 Po ion source. SF − 6 was chosen as the reagent ion (Huey et al 1995, Bertram et al 2001 because it allowed for simultaneous detection of NO 2 as well as multiple halogencontaining species (Br 2 , BrCl, IBr, HCl, and Cl 2 ).
A 3 ml aliquot of sodium halide solution was added to a 1.5 cm i.d., 30 cm-long rotatable glass insert that had been washed first with a 5% HF solution and then with lots of 18 M water. For preparation of desiccated salts, the solution was gradually dried for a period of roughly two hours while rotating the tube under vacuum. This will have prepared surfaces that will not have liquid water present but will have significant amounts of strongly adsorbed water. In particular, the pressure within the flow tube was gradually lowered from ambient to 90 Torr with a low flow (<1 lpm) of dry N 2 passing over the substrate. For frozen solutions, the aliquot was lowered to a temperature of 248 K, with a humidified flow (<1 lpm) of N 2 . Once frozen, the pressure of the flow tube was lowered to 90 Torr. Solutions used were 3.4 M NaCl (ACP 0.01% Br − , 0.002% I − impurities by mass) i.e. Solutions that we refer to as pH neutral had a pH of 6.5-7.0.
When the desired substrate phase had been obtained, a 100 sccm flow of Ar-containing trace H 2 was flowed through a microwave generated plasma. The resulting hydrogen atoms were then introduced into a controlled flow of a known quantity of NO 2 gas to generate OH, within a glass injector positioned with its tip at the upstream end of the coated glass insert.
H + NO 2 → OH + NO. (1) The initial OH concentrations used in the experiments are high with a median number density of 5×10 11 molecules cm −3 . As described in Frinak and Abbatt (2006), these concentrations were determined by assuming that all of the NO 2 loss was due to formation of OH and we consider them to have an absolute uncertainty of roughly a factor of two.
As determined by calibrations using known flows from gas bulb standards, the limits of detection were 6 × 10 9 and 5 × 10 8 molecules cm −3 for NO 2 and Br 2 respectively, of a 30 s integration period where S/N = 1. Their sensitivities were 3.1 × 10 −10 Hz/molecules/cm 3 for NO 2 and 2.4 × 10 −9 Hz/molecules/cm 3 for Br 2 . The detection limits are considerably enhanced over Frinak and Abbatt (2006) and we believe this is due in part to lessened interference from water vapor. The sensitivities of BrCl and IBr were assumed to be the same as Br 2 .

Results
The effects of phase, acidity and bromide concentration of the bulk substrate on the release of halogen gases were investigated. In particular, experiments were run over frozen solutions (248 K), desiccated salts (298 and 253 K) and humidified salts (258 and 253 K). For the humidified salt experiments, the relative humidity in the flow tube was set between the deliquescence relative humidities for NaBr (58%) and NaCl (75%) so that thermodynamics would predict that most of the NaCl will have been in the solid state and the NaBr in solution (Koop et al 2000). However, we note that we have no evidence ruling out the possibility that the surfaces had additional liquid character arising from metastable solutions.
To best quantify the release of halogens relative to OH exposure, experiments were frequently conducted as a series of runs, where a run is a 15 min period where the OH source was turned on. The amount of halogen release was then calculated from the change in signal on and signal off (see figure 1, for an example). Initially, Br 2 and Cl 2 were the two halogen species observed. In particular, Frinak and Abbatt (2006) observed an initial release of Br 2 and then later a Cl 2 release from liquid substrates. Here, Cl 2 was not observed even after prolonged (i.e. multiple hour) exposure to OH, even though the experimental configuration is the same as that of Frinak and Abbatt (2006) and so should have been observed if formed in reasonable yields. However, small signals of other species such as BrCl and IBr were observed, in addition to a large signal from Br 2 . Initially the signals rose slowly with exposure, presumably as the downstream salt surfaces became conditioned with the halogens, and then a steady signal was observed. For each run, the OH concentration was determined by the change in NO 2 signal from when the microwave signal was off and on. As demonstrated in Frinak and Abbatt (2006), all OH radicals were lost via interactions with the coated walls of the glass insert, and there is minimal gas-phase loss.
The results of the OH initiated release of gas-phase halogen species are summarized in table 1, which shows the maximum ratio of gas-phase halogen release to OH loss for each substrate and the run at which the maximum was observed. Typically 1 to 2 experiments were conducted for each set of conditions. When multiple experiments of the same substrate were performed, the mean ratios and range of values are presented. Note that IBr was not monitored in the runs with NaCl.  1 .20 ± 0.14 8 0.020 5 0.030 ± 0.003 a 'Run' refers to the run number at which the maximum ratio was observed.
The effect of pH was initially to be investigated for both desiccated salts and frozen solutions. However, after the acidified bulk solution was dried the CIMS spectrum was dominated by gas-phase HCl. Several hours of dry N 2 were required to bring the HCl concentrations down to a level where SF − 6 was again the dominant ion. Given the significant gasphase concentrations present, and because the surfaces may have been significantly modified from the high level of HCl release, no experiments were conducted with these substrates. However, the effect of pH could be investigated for the frozen sea-salt solutions where this enhanced gas-phase HCl was not observed. Presumably the HCl was sufficiently soluble in the substrate that it was not easily volatilized. Figure 2 shows that the release of Br 2 from an acidified surface was a factor of five greater than from a neutral surface. In this figure, the signals have been normalized to a constant OH concentration to enable a meaningful comparison of product yields from one experiment to another. Acidification of the frozen sea-salt substrate sustains the release of Br 2 over the entire experiment, whereas the neutral frozen seasalt substrate achieves an early maximum and then gradually decreases over the remainder of the experiment. The release of IBr and BrCl are also enhanced at lower pH and the release of BrCl occurs later than either IBr or Br 2 .
In figure 3 the effect of Br − concentration on halogen release from frozen pH neutral solutions is explored for NaCl and sea-salt substrates. As noted in section 2, although the [Cl − ]/[Br − ] = 13 700:1 for NaCl, Br 2 is the dominant gas observed. The release of BrCl was also observed but there was no evidence of gas-phase Cl 2 . For both NaCl and seasalt substrates, the observed maximum signal of gas-phase BrCl occurs after the maximum of Br 2 indicating that BrCl formation occurs once the Br − ion concentration has been depleted. The sustained production of Br 2 from the sea-salt and the higher yield of BrCl from the NaCl also reinforce this conclusion.
In figure 4, experiments were conducted with neutral seasalt substrates where the plasma was left on for ∼2.5 h to investigate how prolonged exposure to OH would affect the release of Br 2 from different phases. The release of Br 2 is dominated by an initial burst that rapidly drops off. At experiment end the release of Br 2 is an order of magnitude lower than initially. In contrast, the release of Br 2 from the frozen solution remains fairly constant throughout albeit roughly a factor of three lower than the maximum observed for  the desiccated salt. The humidified salt behavior falls between the desiccated salt and the frozen solution. The humidified salt achieves a maximum that is somewhat less pronounced and later in arriving than with the desiccated salt. The Br 2 release from the humidified salt also drops off with time, but the attenuation from its maximum is only a factor of two lower than the observed maximum.

Discussion
To summarize the findings from the results we note that (i) Br 2 is the primary halogen product for OH heterogeneous interactions with the surfaces, with smaller yields of BrCl but no observations of Cl 2 , even with prolonged exposure; (ii) frozen acidified solutions yield more halogens than neutral substrates; (iii) enhanced concentrations of Br − lead to more sustained release of Br 2 and delayed BrCl release; (iv) frozen solutions lead to more sustained release than solid salts; and (v) an iodine-containing product (IBr) is observed. To discuss these findings, potential reaction mechanisms will be presented.
In an earlier study by Frinak and Abbatt (2006), it was established that aqueous solutions also exhibit halogen release upon exposure to gas-phase OH, and that enhanced yields are observed from more acidic substrates. In that work, the CIMS sensitivity was not nearly as high as in this study so the lack of halogen products observed under neutral conditions (and the lack of observation of BrCl and IBr) may have arisen from their concentrations being below detection limits. This work indicates that halogens can indeed be released from neutral solid substrates. To determine that there were no experimental artifacts in this regard and noting that the pH of solutions may change upon freezing (Robinson et al 2006) we tested this observation by one set of experiments on frozen solutions formed from pH 9 solution and still observed halogen release.
We focus initially on pathways to form the principal product, Br 2 , and to explain the acidity enhancement (for more details on the mechanism and rate constants of specific reactions see Finlayson-Pitts (2003), or Frinak and Abbatt (2006)). In particular, although interfacial chemistry is likely to dominate, it is not known whether OH will initially react with a Br − ion on the surface or a Cl − ion. Given that the bulk concentration of Cl − dwarfs that of Br − , it is possible that the initial interaction is with Cl − : Note that the acidity dependence appears in the competition between reactions (3) and (4), relative to the loss rate back to form OH. Also, if the Br − concentration drops, then we postulate that BrCl could form via self-reaction of BrCl − . However, it is now recognized that Br − and I − ions concentrate at the surface of humidified solid salts and deliquesced solutions. This has been shown by ambient pressure electron spectroscopy and Rutherford back scattering studies (Zangmeister et al 2001, Ghosal et al 2005, Hess et al 2007. Molecular dynamics simulations indicate that polarizable species such as Br − and I − associate with the surface more than do less polarizable species such as Cl − and F − (Jungwirth and Tobias 2006).
As a result of surface enhancements, there is also the potential that the OH could first interact with a surficial Br − species: OH + Br − ↔ HOBr − followed by reactions (8) and (9) to form Br 2 . Again, an acidity dependence would arise via the competition between reactions (11) and (12). Finally, there is also the possibility of a surface reaction mechanism involving the self-reaction of short-lived OHhalide complexes but not involving protons (Oum et al 1998a, Finlayson-Pitts 2003: OH + Br − → OH·Br − Molecular dynamics studies by Roeselova et al (2003) infer that the length of time that OH is adsorbed to a surface is sufficient for heterogeneous chemistry of this type to occur in the interfacial region.
For aqueous-phase substrates, Frinak and Abbatt (2006) observed the release of Cl 2 as the signal due to Br 2 dropped, presumably due to Br − depletion. In this work, we did not see the release of Cl 2 from any of the substrates even after running experiments for as long as 12 h. Two effects might explain this. First, the phase was different in these experiments, with the vast majority of the chloride in the solid phase as either NaCl or NaCl·2H 2 O. Thus, chloride would not be as available for reaction. Also, there would be a significant enhancement of Br − at the top of the substrate, arising from the manner of substrate preparation. In particular, exclusion of ions during solution freezing may form a highly concentrated halide layer at the surface. Depletion of this layer of Br − may not occur as readily as it does in the uppermost layers of a solution that does not experience the same degree of concentration enhancement.
We note that we observe the release of IBr from the frozen solutions, even though iodide is present at only the 0.002% level in the NaCl salt. This is consistent with iodine being more easily oxidized than other halides, and with it also being more surface active (Ghosal et al 2005).
Finally, the phase of the substrate had an effect on the temporal profile of the halogen release. For the desiccated salt, although there may be a few layers of surface-adsorbed water, it is expected that both the bromide and chloride will be predominantly in the solid phase. Br 2 release was prompt from these surfaces, followed by a rapid decline in signal. This is consistent with there being depletion of surficial bromide and protons that are not readily replenished due to slow diffusion times from the bulk. In the case of the mechanism initiated by reactions (14) or (15), the formation of NaOH may poison the surface.
For frozen solutions, we note that there will be small pockets of bromide-containing brine given that the eutectic with NaBr·5H 2 O is ∼245 K (Koop et al 2000), and perhaps even the presence of metastable solutions. Thus, the mobility of bromide in the brine may sustain halogen release by replenishment of protons and halides from below the uppermost surface layers. Finally, the behavior of the humidified salts is intermediate between the desiccated salts and frozen solutions probably because the humidity was high enough to deliquesce the sodium bromide.

Conclusions
All the pH neutral substrates display the release of gasphase halogens, primarily Br 2 , upon exposure to gas-phase OH. The release was initially most pronounced with the desiccated salts due to the high concentration of Br − ions at the surface. For frozen solution substrates and humidified salts, the product release was more sustained with time, indicative of replenishment of halide ions at the surface. An acidity dependence was observed for the halogen release from the ice substrate, with the yields of all the halogens increasing when acidic solutions were used. Lastly, an oxidized iodine product, IBr was observed. From an environmental perspective, this work demonstrates that these surfaces, whether neutral or acidified, are fully able to act as sources of atmospheric halogens, principally Br 2 , when exposed to gas-phase OH.