Cl2, Br2, and I2 were measured in coastal Pacific air from 2 to 29 January 2006. Air was sampled at 10 m over the sea surface near the end of Scripps Pier (La Jolla, California). The measurements were made using atmospheric pressure chemical ionization with tandem mass spectrometry (APCI/MS/MS). Over the course of this study, Cl2, Br2, and I2 levels ranged from below detection limits of 1.0, 0.5, and 0.2 ppt, respectively, to maxima of 26, 19, and 8 ppt, respectively. Mean dihalogen levels for the study period were 2.3 ± 1 ppt for Cl2, 2.3 ± 0.4 ppt for Br2, and 0.7 ± 0.1 ppt for I2 (expressed as geometric mean ±1 geometric standard error). The mixed dihalogens BrCl, ICl, and IBr had geometric mean levels below 0.3 ± 1 ppt and never exceeded their detection limits of 0.5 ppt. Consistent patterns of diurnal variability were observed for Cl2 and I2, with Cl2 maxima during daytime and I2 appearing almost exclusively at night. The detection of I2 appeared to be related to the passage of air over nearby kelp beds. The observed dihalogen levels suggest that (1) chlorine atom oxidation of hydrocarbons makes a significant contribution to the formation of ozone and (2) halogen atom oxidation of airborne mercury contributes to mercury deposition in polluted coastal air.
 In the atmosphere, chlorine, bromine, and iodine atoms can participate in catalytic ozone destruction cycles involving the formation and recycling of weakly bonded halogen oxides. In polluted regions, oxidation of hydrocarbons by chlorine atoms can also lead to ozone formation through the production of peroxy radicals and resulting conversion of NO to NO2. The abundance of halogen radicals in the troposphere is not well documented and their role in tropospheric photochemistry is not well understood. The coastal boundary layer is an environment where halogen photochemistry is potentially important because of the abundance of marine aerosols and the interaction of marine air with polluted air over populated regions.
 The oxidation of atmospheric mercury (Hg0) to reactive gaseous mercury (RGM, HgII) has important implications for mercury cycling. Unlike elemental mercury, RGM is soluble in water and rapidly deposits to surfaces. Hg0 is enhanced in polluted urban air relative to the global background [Shon et al., 2008], and model studies of Hg have suggested that halogen oxidation of Hg0 could be a significant source of RGM in the free troposphere [Holmes et al., 2006], the marine boundary layer [Laurier et al., 2003; Hedgecock et al., 2003], and in polluted urban air [Shon et al., 2005].
 This paper presents measurements of the dihalogens Cl2, Br2, and I2 over the sea surface off La Jolla, California. To the best of our knowledge, these data include the first reported mixing ratios of Br2 in the MBL outside of the Arctic and the first observations of ambient I2 by atmospheric pressure chemical ionization tandem mass spectrometry.
2. Experiment Location, Meteorology, and Ancillary Measurements
 Field measurements were made at the Scripps Institute of Oceanography pier facility in La Jolla, California from 2 to 29 January 2006. The pier is located at 32.87°N, −117.26°W. The pier is 330 m long and 10 m above the sea surface. The coastline at this site is oriented northeast/southwest. The instruments were housed in a small building 20 m from the end of the pier.
 Local measurements at Scripps Pier are shown as a time series in Figure 1. Skies were clear and sunny with the exception of cloudy days on 2 and 3 January. Daytime temperatures on most days were 14–16°C (Figure 1, top). During 5–8 January, however, temperatures reached 22°C and did not cool off significantly at night. Temperatures on most nights were 10–13°C. Wind speeds were generally low (1–3 m s−1) and wind direction was highly variable (Figure 1, middle). There were four periods of elevated wind speeds (2–4, 15–17, 20–21, and 26–28 January). Maximum wind speeds were attained on the morning of 3 January, reaching 14 m s−1. Tidal height ranged from –1 to +2 m (Figure 1, bottom). Besides the diurnal and semidiurnal variations, there were significant weekly variations in tidal height.
2.3. Ancillary Measurements
 NO, NOx, and O3 were measured using ThermoElectron Corp. models 42C and 49C, respectively. NOx and O3 data were recorded at 1-min intervals and averaged over 15 min periods. The data typically exhibit strong diurnal variations with daytime ozone maxima and nighttime NOx maxima resulting from photochemical reactions, emissions, and surface losses. There are two anomalous periods (14–15 and 18–19 January) during which O3 did not decrease at night and NOx did not increase. The air masses associated with these two episodes arrived at the pier from the east after spending several days over land at an altitude ∼650 mbar.
 Sulfur dioxide was monitored qualitatively as an indicator of when the instrument sampled ambient air and scrubbed ambient air. These measurements were made with the APCI/MS/MS using the SO5–→SO3– (112→80 amu) mass transition. The instrument was calibrated using an SO2 permeation tube after the completion of the project. The sensitivity of SO2 to humidity and O3 levels was not quantified.
3.1. Detection of Cl2, Br2, I2, and Mixed Dihalogens
 The dihalogens Cl2, Br2, I2, BrCl, ICl, and IBr were monitored using atmospheric pressure chemical ionization tandem mass spectrometry (APCI/MS/MS; Thermo Corporation TSQ Quantum). Negative ions were formed at atmospheric pressure by passing ambient air through a 1/4” SS tube lined with a 63Ni foil. The foil emits 67 keV that rapidly thermalize by collisions with O2 and N2. The primary negative ions formed are likely O2– and O4–, which rapidly cluster with water, giving O2–(H2O)n and O4–(H2O)n [Siegel and Fite, 1976]. These and other weakly electronegative transients undergo a variety of ion-molecule collisions reactions in the APCI source and 1 torr declustering region. The exact sequence of charge transfer reactions leading to the formation of dihalogen ions (e.g., Cl2–, Br2–, I2–) is not known.
 The dihalogens are considerably more electronegative than most common constituents of air. Some highly electronegative species such as NO3 and CO3 are formed in the source through ion molecule reactions. NO2 has an electron affinity similar to that of Cl2, and charge transfer can occur between either anion and the other neutral molecule [Hughes et al., 1973]. These species compete for charge with the dihalogens and variations in their abundance have the potential to alter the instrumental response to dihalogens. Ambient air scans show NO3–, CO3–, and NO2– peaks (62, 60, and 46 amu, respectively) that are consistent in size, and do not vary with changes in ambient conditions such as NOx levels, humidity, or temperature (Figure 2). This indicates that the abundance of these ions is primarily a function of ion-molecule chemistry in the ion source rather than ambient levels of precursors [Finley, 2007]. The insensitivity in major ions to ambient conditions likely explains why the sensitivity of the instrument to dihalogens is fairly constant. This is verified by standard additions of dihalogens to scrubbed ambient air and calibration curves in scrubbed air, as discussed below. It is important to note that the ion spectra observed in this instrument do not necessarily reflect their abundance in the ion source itself, but are modified by ion-molecule collisions during passage through the 1 torr region.
 Although halogen radicals exist in the ionization region, radical chemistry (e.g., Br + Br→Br2) is unlikely to be a significant source of dihalogens. If radical chemistry were the source of the dihalogen signal (e.g., Br + CH3Br→Br2 + CH3), we would expect to see similar dihalogen signals in both ambient air and in scrubbed ambient air (see section 3.2). In fact, dihalogen levels in scrubbed ambient air are usually <1% of the levels observed in ambient air. This result suggests that radical-mediated production of dihalogens is not a major contributor to the overall dihalogen signal.
 Once formed, the dihalogen ions were quadrupole mass filtered, then collisionally dissociated (1 mtorr Ar, 20 eV) into their constituent atoms (Cl–, Br–, and I–), which were mass filtered in a second quadrupole. The ion transfer capillary between the APCI source and the mass spectrometer was maintained at 100°C.
 Chlorine and bromine each have two stable isotopes that yield four possible mass transitions for each molecular ion. Only one stable, naturally occurring isotope of iodine (127I) exists resulting in a single transition for I2. The dihalogen mass transitions monitored during this study are listed in Table 1. The mass spectrometer was run continuously except during periods of instrument calibration. Data was lost because of power outages on 11–12 and 14–15 January.
Table 1. Mass Transitions Used to Monitor Dihalogen Compounds in Aira
Mass Transitions Monitored
The transitions used for quantitation are marked with an asterisk.
3.2. Sampling Inlet
 The air intake was located 2 m from the side of the pier and 10 m above the sea surface. Because dihalogen gases are surface reactive, a two-stage, laminar flow inlet was used to bring air to the instrument. The first stage consisted of approximately three meters of black 2” ID ABS plastic pipe with PVC connectors. An airflow rate of 63 slpm was drawn through the pipe by a blower and monitored using a TSI 4216 flow sensor (TSI Corp.). The second stage consisted of a 1/2” OD PFA Teflon tube that sampled 2.25 slpm isokinetically from the center of the first stage airflow using a vacuum pump and mass flow controller. Air was drawn from the central portion of this flow into the 1/4” SS tube that contained the 63Ni foil in the APCI ionization source. This flow rate was 1 slpm and was controlled by the capillary ion transfer tube and the vacuum in the mass spectrometer. The flow rates were selected to maintain laminar flow and uniform linear flow rates across the inlet. The linear velocity of the gas was maintained at 0.5 m s−1, resulting in a residence time of about 6 s. The inlet had a passing efficiency of 90 ± 11% for Cl2, 91 ± 12% for Br2, and 87 ± 12% for I2. Passing efficiencies were determined using standard additions to scrubbed ambient air.
 The inlet had two air intakes, connected by a 2” ID tee. The “straight-through” side of the tee allowed sampling of ambient air, and the “right angle” side admitted ambient air through a 0.5 m length of black ABS pipe filled with a carbonate-coated glass wool scrubber. The scrubber was prepared by soaking glass wool in a 1% (w/w) solution of Na2CO3 and Milli-Q water for 1 h and allowing it to dry. The treated fiberglass wool was >99% effective at removing dihalogens from the air stream. Two pneumatically controlled gate valves were used to alternate between ambient and carbonate-scrubbed air. Aerosols were not intentionally removed from the sampled air stream. Instrument blank signals in a clean inlet were compared to blank signals in an inlet that was exposed to ambient air and sea salt aerosols for periods of 48 h and one week, respectively. The exposed inlets did not show significantly different blank levels or dihalogen response compared to the clean inlet, suggesting that accumulation of sea salt on inlet surfaces was not a significant source of interference.
 Cl2, Br2, and I2 standards were generated using permeation tubes (VICI Metronics) stored under nitrogen in a temperature-controlled environment. The Cl2 and Br2 tubes were maintained at 30 ± 1°C and the I2 tube was maintained at 100 ± 1°C. Weight records for these tubes were maintained before and after the experiment, with consistent results. As pointed out by Keene et al. , the weight loss record of Cl2 permeation tubes may not reflect the actual emission rate, presumably because of chemical reactions either in the tubes or on the outer surfaces of the tubes. To validate the permeation rates, the output of each of the permeation tubes was bubbled through a neutral KI solution (2% w/w KI, 1 mM phosphate buffer, pH = 7), in which dihalogens oxidized I– to I3–. The absorption of I3– at 352 nm was measured with an Ocean Optics SR2000 spectrophotometer and a Xe lamp. The rate of halogen release from the permeation tubes was calculated from the rate of change of I3– absorbance, and the molar absorptivity of I3– (ɛ = 23,200 L mol−1 cm−1) [Wei et al., 2005]. The emission rates were consistently lower than the rates obtained from weight loss. The correction factors for the emission rates were 0.75 for Cl2, 0.51 for Br2, and 0.91 for I2.
 Gas standards were added to the instrument inlet just after the 2” ID tee for 15 min every 4 h to check instrument sensitivity. The standards were generated at ppb levels from the permeation sources using a two-stage dilution system [Gallagher et al., 1997]. A high-purity N2 stream of 85 cm3 min−1 was passed over the permeation device. A variable percentage of the flow (40–73%) was then dumped to waste. The remaining flow was diluted with a clean nitrogen stream and a second dump was used to vary the diluted flow. The dilution system flow was then added to the 63 slpm flow in the 2” ID pipe, reducing the ppb levels generated in the dilution system down to ppt levels.
 This approach has two advantages for calibration of reactive gases. First, the dihalogen standard contacts only PFA tubing and fittings, minimizing loss through surface reactions. Second, the range of different mixing ratios in the sampled air stream is generated by varying the flow rate of the standard delivered into the air inlet, while the mixing ratio of dihalogen in the tubing used to deliver the standard does not change. This avoids having to reequilibrate the tubing walls during calibration.
 Primary standards were not available for the mixed halogens BrCl, ICl, or IBr. Laboratory experiments to determine the instrument sensitivity to BrCl were carried out following the field study, as described in more detail below. Mixing ratios of ICl and IBr were estimated using the averaged Br2 calibration curve.
3.4. Blanks and Interferences
 The inlet gate valves were programmed to periodically direct carbonate-scrubbed air through the inlet. The dihalogen levels measured during these periods are considered system “blanks,” which are subtracted from the ambient dihalogen signals. The “blank” air contains many of the unreactive constituents of air such as O2, N2, most hydrocarbons, and halocarbons such as CFCs, HCFCs, methyl halides, etc. The scrubber should remove acidic constituents of air such as HCl, HNO3, SO2, and organic acids. This method of determining the system blank is not foolproof, but it should account for many potential interferences. To successfully generate a positive artifact in the system, an interfering molecule would need to (1) be removed by the carbonate scrubber and (2) form a dihalogen in the source either through neutral chemistry or ion-molecule interactions. Possible interferents are molecules like dichloroacetic and trichloroacetic acid (breakdown products of tetrachloroethene and 1, 1, 1-trichloroethane, respectively) and NCl3 (a by-product of water treatment). We have tested these gases in the laboratory and found that they produce no Cl2 response in the instrument. HCl does not produce Cl2 signal.
3.5. Calibrations and Detection Limits
 Instrument calibrations consisted of adding varying levels of Cl2, Br2, and I2 gas standards, ranging from 4 to 61 ppt, to blank air for 5 min at each level (Figure 3). Calibration curves were obtained from a linear least squares fit to the data. For all calibrations, there was a small negative intercept which may indicate the presence of a zero-order loss in the sampling system or ion source.
 Calibration for BrCl was done using a permeation device constructed in the laboratory. The permeation cell was a cylindrical 350 cm3 PFA Teflon chamber with a radius of 5.4 cm and a height of 3.8 cm and Viton o-ring sealed flanges. A 60 cm length of 1/16” OD PTFE Teflon tubing (0.016” wall thickness) was passed into and out of the chamber via PFA compression feed-throughs. The chamber was maintained at 30°C with an 85 cm3 min−1 flow of N2. The permeation cell was flushed and filled with either 1% Cl2 in zero air (1% Cl2, AirGas), 1% Br2 (99.5% Br2 liquid, Fisher Scientific) in N2, or a mixture of both. The emission rates of Br2 and Cl2 into the tubing were determined by filling the chamber with 1% Cl2 or 1% Br2 and comparing the output with gas streams from the commercial permeation sources described earlier. BrCl was generated by filling the chamber with a mixture of 1% Cl2 and 1% Br2 and allowing it to react for 24 h.
 BrCl, Cl2, and Br2 emitted from the permeation chamber were measured with the APCI/MS/MS. The permeation rate for BrCl was determined from the decrease in Cl2 and Br2 signals relative to the signals when the chamber was filled with each compound individually. The BrCl permeation rate was used in conjunction with the observed BrCl signal to estimate the sensitivity of the instrument for BrCl. Five different mixing dilutions of BrCl from the chamber output were added to the instrument inlet to generate a calibration curve. The APCI/MS/MS sensitivity to BrCl is 25% higher than that for Cl2, intermediate between Cl2 and Br2. As an independent check on this method, the calculated mixing ratios were used to calculate the equilibrium constant for BrCl in the Cl2+Br2 mixture. The Keq calculated was 9.2 ± 0.3 which is consistent with the literature value of 9.1 [Tellinghuisen, 2003].
 Similar experiments have not yet been carried out for IBr and ICl. In this paper, the Br2 calibration curve was applied to the IBr and ICl signals. Applying the Cl2 or I2 calibration curves to the mixed halogen signals would not substantively alter the interpretation of the results.
 The practical detection limit of the dihalogen analysis in air is imposed by the variability in the blank signal. Typical blank signals correspond to <1–5 ppt of dihalogen. The detection limit was estimated using the observed variability in the blank signal, the error in the blank interpolation (see section 3.6), and the error in the calibration curve slope. Detection limits were estimated at 1 ppt (S/N = 2) for the 72→35 Cl2 transition, 0.5 ppt for the 162→81 Br2 transition, and 0.2 ppt for the 254→127 I2 transition. The detection limit is equivalent to approximately 1 ions s−1 for Cl2, 14 ions s−1 for Br2, and 47 ions s−1 for I2.
3.6. Data Processing
 Data collection with the APCI/MS/MS consisted of monitoring each mass transition in Table 1 for 1 s sequentially from lowest to highest parent mass. The total cycle time was 17 s, including the dihalogen and SO2 mass transitions. The inlet gate valves were cycled to provide a 15 min period of blank after each hour of ambient air sampling. The SO2 signal clearly indicates the blank intervals due to its high signal to noise and effective removal by the carbonate scrubber. The SO2 signal was used to validate the operation of the gate valves when the instrument was unattended. Because of minor misalignment, one of the gate valves occasionally failed to close fully. Periods of blank containing anomalously high SO2 levels were removed from the data set. The samples immediately following these periods of contaminated blanks were removed as well, since it was impossible to know whether the sample valve had properly opened or if the sample was contaminated with scrubbed air.
 The raw ambient measurements were averaged over 15 min intervals and corrected for instrument blank. The blank for each sample was calculated by linearly interpolating between the blank before and after each sampling interval. These blanks were subtracted from the ambient signals. The blank-corrected signals were converted into a mixing ratio using the calibration curve. A single calibration curve was used for the entire measurement period. This calibration was obtained by averaging the 6 individual calibrations measured during the study. No systematic changes in calibration were observed during the study period, within the estimated uncertainties. For Cl2, Br2, and I2, the 72→35, 162→81, and 254→127 mass transitions were used for quantification.
 Raw data for Cl2, Br2, and I2 is shown Figure 4. The ambient signal is considered below the instrument detection limit when the ambient signal cannot be distinguished from the blank signal. Standard additions to scrubbed ambient air are also shown.
 The uncertainties in the 15 min averaged mixing ratios were determined by propagating the lognormal variance in the raw 1 s data (as 1 geometric standard error of the mean), the uncertainty in the blank, and the uncertainty in the slope of the calibration curve. The resulting uncertainties ranged from ±0.1 to ±5 ppt. All means and uncertainties are reported as the geometric mean ± one geometric standard error of the mean. For computing mean dihalogen levels, data points below the detection limit were assumed to be 1/2 the detection limit.
4. Results and Discussion
4.1. Isotope Ratios
 Isotope ratios were examined as a test of the validity of the ambient dihalogen measurements. Because of large ambient variability and low ion counting rates, considerable time averaging was needed in order to reduce the uncertainty in the isotope ratios to useful levels. The ratios of various mass transitions were computed from the 15 min average blank-corrected ion currents. For Cl2, these ratios were then averaged to compute a weekly average ratio for ambient air, standards, and blanks (Figure 5). The uncertainties in the ratios are reported as one standard error of the mean, calculated from the observed variances of the ratios. For ambient air and gas standards, only measurements above 4 ppt were used to compute isotope ratios.
 If the mass spectrometer signal at a parent mass of 72 is entirely due to 35Cl37Cl, fragmentation should produce ion fragments at mass 35 and 37 with a ratio of 1. In the mass 72 signals from scrubbed air blanks, weekly average 72→35/72→ 37 ratios ranged from 0.23 to 0.32, with an overall mean of 0.28 ± 0.03 (n = 373) for the entire measurement period. This low isotope ratio suggests that the blank is due either to an unchlorinated compound, or to a molecule containing a single chlorine atom. This result suggests that the blank is not primarily Cl2. The blank, therefore, is not simply due to desorption of Cl2 from the inlet walls or to the production of Cl2 from aerosols that pass through the scrubber and reach the ion source. One possible candidate for the blank is HOCl·H2O–. This singly chlorinated cluster would yield a 72→37 transition but would not give a corresponding 72→35 transition. Because HOCl is not be expected survive passage through the carbonate scrubber, it would need to be formed in the ion source. Cl2 gas standards and ambient air exhibited 72→35/72→37 isotope ratios at mass 72 close to the expected value of 1.
 At the natural chlorine isotopic abundance, the ratio of 70Cl2 (35Cl35Cl) to 72Cl2 (35Cl37Cl) should be 1.5. This ratio was computed from the data by taking the ratio of the signal at the 70→35 transition to the sum of the 72→35 and 72→37 signals. The blanks have an average 70Cl2:72Cl2 ratio of 2.2 ± 0.1 (n = 373) over the entire measurement period, with a range of 2.0 to 2.3. This is consistent with the presence of an HOCl·H2O– cluster, which would give a 70→35 transition as well as a 72→37 transition, and with the conclusion above that the blank is not primarily due to Cl2. Cl2 gas standards gave weekly average 70Cl2:72Cl2 ratios ranging from 1.4 to 1.6 with an overall mean for the whole study of 1.5 ± 0.1 (n = 56). Weekly average ambient air ratios ranged from 1.55 to 1.8, with an overall mean 70Cl2:72Cl2 ratio of 1.6 ± 0.2 (n = 236). During the first week of measurements, the 70Cl2:72Cl2 ratio was slightly elevated, suggesting the presence of some interference in the at the 70→35 transition. Because the ratio of 72→35/72→37 does not show evidence of interference for standards or ambient air, the 72→35 transition was used to quantify ambient Cl2 mixing ratios.
 Bromine isotope ratios (79Br:81Br) were also examined. Only Br2 signals above 3 ppt were included in the ratio analysis for ambient air and gas standards. Ratios were computed from the 15 min average blank-corrected ion currents. The ratios were averaged over the full 27 day measurement period because of the very low ion statistics during the final week and the lack of regular standard additions during the first week. The uncertainty in the ratios is given as one standard error of the mean and is calculated from the observed variance of the ratios of the various bromine mass transitions.
 If the signal at mass 160 were composed entirely of Br2, the ratio of 160→79/160→81 would be close to 1. The mean ratio of these transitions in blanks from this study was 0.78 ± 0.02 (n = 373). This ratio is slightly low, indicating that that at most, Br2 can account for about 80% of the signal at mass 160. For the gas standards, the ratio was 1.0 ± 0.03 (n = 26) and for the ambient air the ratio was 0.94 ± 0.02 (n = 183). The results for the standards and ambient air are consistent with the signals originating primarily from Br2.
 The isotope ratios of the parent bromine ions, 158Br2:160Br2:162Br2 should be 0.5:1.0:0.5. This ratio was calculated for the field data by summing the 160→79 and 160→81 signals to calculate 160Br2. The blank signals gave ratios of 0.46:1.0:0.45. This is consistent with the presence of some contamination of the 160Br2 signal. The parent ratios for the Br2 gas standards were 0.54:1.0:0.51 and ambient air gave ratios of 0.56:1.0:0.48. The uncertainty in these ratios ranged from 0.01 to 0.02 (1 se). Overall, the parent isotope ratios are consistent with the identity of the signal as predominantly due to Br2, although the uncertainty in the ratios does not preclude minor contamination from other, possibly bromine-containing compounds. For quantifying the Br2 signal, the 162→81 mass transition was chosen because it was the most consistent of the four transitions and showed the least evidence of interference.
4.2. Dihalogens and Large-Scale Meteorology
 Five-day air mass back trajectories at 250 m and 500 m were calculated at midnight, 0600, noon, and 1800 local time every day from 2 to 29 January. After generating the back trajectories, the trajectory cluster analysis function of the HYSPLIT v4.7 model was used to divide the trajectories into five distinct groups (Figure 6). Cl2, Br2, I2, total NOx, and O3 ranges observed during each group are summarized in Table 2. The continental group consisted of air masses that spent 5 days over the U.S. and Canada. These air masses had 6 h or less of coastal influence before arriving at the site. The oceanic group was composed of air masses that spent 5 days over the Pacific Ocean with 12 h or less near the coast. The low-latitude group consisted of air masses originating over the Pacific Ocean from 30 to 33°N, the midlatitude group originated between 38 and 45°N, and the high-latitude group originated from 50 to 55°N. The low-, middle-, and high-latitude groups all originated over the Pacific Ocean, but spent at least 2–3 days over the U.S. west coast before arriving at the pier.
Table 2. Five-Day Air Mass Back Trajectory Groups With Observed Dihalogen and Pollutant Ranges
Date in January
10–13, 18, 20–22, 24
19, 23, 25
4.2.1. Dihalogens in the Continental Group
 The continental group of air masses consistently exhibited high levels of NOx and O3 due to the presence of strong anthropogenic sources. Cl2 was regularly detectable in all of these air masses (Figure 7a). Typical Cl2 levels were <5 ppt, but a 26 ppt maximum was observed around noon on 15 January. Typical Br2 levels were <1 ppt except on 15 January when they reached 5.5 ppt (Figure 7b). I2 in continental air was generally <0.5 ppt. On 27 January, I2 increased from <0.1 ppt at 1500 local time to a maximum of 4 ppt near midnight (Figure 7c). These results suggest there may be a significant continental source of Cl2 since it was regularly observed in continental air with no coastal influence. This source would most likely be anthropogenic, and related to use in swimming pools, cooling towers, and water treatment facilities. Br2 and I2 were only observed episodically, implying that continental sources may be less important for these compounds. The high I2 mixing ratios on 27 January were observed in air masses that spent considerable time (4–6 h) over the coast. The source of the I2 was most likely coastal, although anthropogenic sources cannot be ruled out.
4.2.2. Dihalogens in the Oceanic Group
 The oceanic air masses consistently showed the lowest levels of NOx and O3 (Table 2). Cl2 levels in oceanic air were usually <5 ppt but were consistently above the detection limit (1 ppt) during the day. The highest Br2 levels were observed in the oceanic group. Br2 levels were typically <1 ppt but reached a daytime maximum of 9 ppt. The highest Br2 mixing ratios in this group were observed during the day. Nighttime levels were lower (<0.5–2 ppt), although mixing ratios up to 5 ppt were observed after midnight on 16 January (Figure 7b). I2 mixing ratios were consistently low in oceanic air, reaching a maximum of 1.5 ppt just before dawn on 5 January. Results from the oceanic group suggest a possible ocean/coastal source of Cl2 and Br2. This source is likely photochemical since the dihalogens were observed during the day, and the results are consistent with sea salt aerosol production of Cl2 and Br2. The oceanic group had up to 12 h of coastal influence before reaching the site, so the influence of anthropogenic pollutants (direct emissions or precursors) cannot be ruled out. The presence of I2 in this group is consistent with a coastal source, most likely the scattered kelp beds near the coast.
4.2.3. Dihalogens in the Low-Latitude Group
 Low-latitude air masses (30–33°N) consistently showed low NOx and O3, except 5 January when NOx reached 121 ppb. The large NOx events in this group were likely caused by ship activity, since these air masses passed over the San Diego naval shipyard 16 km to the southeast. Dihalogens in this group were consistently above detection for Cl2, Br2, and I2 (Figures 7a–7c). Cl2 reached 21 ppt around 1300 local time on 6 January. Typical Cl2 levels were between 2 and 4 ppt during the day, similar to those observed in continental air. Br2 was typically 1–2 ppt, with a maximum of 9 ppt at noon on 6 January. This group showed the highest I2 mixing ratios, reaching a peak of 4 ppt at 0400 local time on 3 January. The high I2 levels in this group were likely due to the passage of the air masses over the large kelp beds in the bay to the southeast.
 Cl2 and Br2 variability was higher in the low-latitude group than any other group, with levels ranging from undetectable levels to several parts per trillion within a few hours. This suggests either a highly variable, fixed dihalogen source or a spatially heterogeneous source. The I2 was also variable in these air masses and can probably be explained by the patchiness of the kelp beds.
4.2.4. Dihalogens in the Midlatitude Group
 The midlatitude group (38–45°N) had the highest average NOx levels (usually 25–40 ppb) with O3 mixing ratios similar to the continental and high-latitude groups. The air masses in this group originated over the ocean, moved down the west coast, and spent several days recirculating over the southern California coast. I2 mixing ratios were generally <0.2 ppt. Br2 levels were typically <1 ppt, but reached 5 ppt on 10 January (Figure 7b). Cl2 mixing ratios were similar to continental and low-latitude air, ranging from <1–17 ppt.
4.2.5. Dihalogens in the High-Latitude Group
 NOx levels in the high-latitude air masses were similar to low-latitude and midlatitude air but had higher O3 levels similar to continental air (Table 2). High pollutant levels were due to anthropogenic coastal sources. All of these air masses spent 2–3 days moving down the west coast and passed over Los Angeles, Orange, and northern San Diego counties. This group had the lowest dihalogen levels of any continentally influenced air masses. Cl2 was usually <5 ppt and I2 was <0.5 ppt. Br2 levels were below 1 ppt most of the time. With minimal coastal influence in the 24–48 h prior to arriving at the site, sea salt aerosol Cl2 and Br2 sources were unlikely. The absence of large kelp beds north of the site probably explains the lack of I2. The low dihalogen variability in these air masses may suggest that they were aged air masses that had lost dihalogens to photolysis and contained dehalogenated or very few sea salt aerosols.
4.3. Dihalogens and Local Meteorology
 Dihalogen mixing ratios are plotted as a function of wind speed and direction in Figure 8. Cl2 was detectable at levels ranging from <1–28 ppt in winds from all directions (Figure 8). The highest Cl2 levels (11–28 ppt) were observed in winds with speeds of 1–3 m s−1 with no directional preference and in winds from the southeast at 6–9 m s−1. The detection of Cl2 at low wind speeds during the day suggests a local Cl2 source and is consistent with sea salt aerosol production, although local anthropogenic sources (e.g., swimming pools, sprinkler systems) could not be ruled out.
 The highest levels of Br2 (4–9 ppt) were consistently detected in daytime onshore winds from the northwest with speeds of 6–8 m s−1 (Figure 8). Elevated Br2 mixing ratios in onshore air are consistent with a sea salt aerosol source of Br2. Br2 was detected at elevated wind speeds, which also suggests a sea salt aerosol source due to higher sea salt loads [see Lewis and Schwartz, 2004, Figure 22]. This may not be the only source, however, since Br2 was detectable in winds from all directions, including offshore winds.
 Simultaneously high Br2 and Cl2 levels were observed in winds from the southeast on 6 and 15–16 January (Figure 8). On 7 and 27 January, Cl2 was observed at levels of 5–10 ppt during periods of low wind speed (1–3 m s−1) when Br2 levels only reached 3 ppt. During these two cases (simultaneous Cl2/Br2 and primarily Cl2) the observed NOx and O3 had similar ranges and mean levels. The similarity in the pollutant concentrations suggest that some additional source of variability, aerosol pH for example, is needed to explain these results.
 Maximum I2 levels were observed in southeasterly winds at 6–9 m s−1(Figure 8). That air was likely influenced by large local kelp beds, a known source of I2 [Palmer et al., 2005]. Other periods of increased wind speed either occurred during midday, when I2 photolysis would prevent its transport to the sampling site, or did not pass over the kelp beds before reaching the pier. I2 was detectable at low wind speeds (1–3 m s−1) independent of direction, suggesting the presence of a source very close to the pier. The local source may be heterogeneously distributed patches of kelp. Offshore winds also showed detectable I2, perhaps due to kelp and other seaweed washed up on the beach.
Figure 9 shows that I2 was not strongly correlated with tidal height. I2 was observed during high tide and low tide. The highest I2 mixing ratio of 4.5 ppt was observed during a period of high tide. The afternoon and evening of 3 January and the period from 2300 local time to sunrise on 4–5 January showed increasing I2 levels even as the tide was rising. The rapid I2 decreases from 0.6 to 0.9 ppt to near the detection limit on 3 and 5 January occurred at sunrise, when photolysis rapidly destroyed I2.
4.4. Mean Dihalogen Levels
 Geometric mean dihalogen mixing ratios for the 27 days of observations were calculated using the 15-min averaged blank-corrected data (Figure 7a). For Cl2, the mean level was 2.3 ± 1 ppt. Cl2 was above the detection limit (1 ppt) for 55% of the 15 min measurement periods. The daytime mean Cl2 level was 2.8 ± 1.0 ppt and the nighttime mean level was 1.5 ± 0.7 ppt.
 Br2 in ambient air had a mean mixing ratio of 2.3 ± 0.4 ppt, well above the detection limit of 0.5 ppt (Figure 7b). Br2 was detectable for 82% of the measurement intervals. Iodine ranged from below the detection limit of 0.2 ppt to a maximum of 4.5 ppt (Figure 7c). The mean I2 mixing ratio for the entire month was 0.7 ± 0.1 ppt. 73% of the total I2 measurements were above the detection limit.
4.5. Dihalogen Temporal Patterns
 Cl2 levels >5 ppt occurred more frequently during the daytime than at night, supporting a photochemical production mechanism. There were 146 Cl2 measurements greater than 5 ppt. 104 of these points were detected during the day and only 42 were detected at night. Figure 10 shows that Cl2 was above the detection limit every day for several hours during the first week of measurements, reaching a maximum between noon and 1600 local time. At night, Cl2 levels were close to the detection limit and consistently exceeded the detection limit only on 6–7 January. The data do not support a strong nighttime Cl2 production mechanism. Cl2 was regularly detectable during the first week of the study. The appearance of Cl2 during the last three weeks was irregular, usually for an hour or two at a time.
 Br2 was observed both day and night with no consistent diurnal cycle (Figure 10). Br2 was detectable during the day on 3–10 January and during the night on 4–10 January. The ambient levels of Br2 follow a pattern similar to Cl2, with higher levels during the first week, and lower, more variable levels afterward (Figure 7). During the first week of measurements, Br2 regularly reached levels of 1–4 ppt. At night, Br2 usually appeared unaccompanied by Cl2.
 The observed diurnal pattern of Br2 requires daytime and nighttime production. The only known mechanisms that can account for nighttime Br2 are surface reactions of O3, BrONO2, and BrNO2 with aerosol Br– [Schweitzer et al., 1998; Sander et al., 1999; Hunt et al., 2004]. Proposed daytime production mechanisms for Br2 include bulk aerosol/gas phase autocatalytic cycles and surface reactions of OH and Br– [Fan and Jacob, 1992; Vogt et al., 1996; Frinak and Abbatt, 2006]. The episodic nature of the Br2 signals observed in this study suggests that variations in aerosol chemistry may play a dominant role in the production of Br2. Future field studies involving simultaneous Br2 and aerosol chemistry measurements are needed to rigorously test hypotheses for Br2 formation.
 I2 exhibited strong diurnal variability. During the first week, nighttime levels were 1–3 ppt higher than daytime levels (Figure 10). For the last three weeks, the difference between night and day was usually 0.2–0.5 ppt. Daytime I2 reached a maximum of 3 ppt on the afternoon of 26 January. Overall, I2 levels were highest during the first week of the study. The final 2 days of measurements also showed I2 levels up to 4.5 ppt. When I2 was detectable during the day, it was observed between 0800 and 1000 local time and from 1400 local time until sunset. I2 was rarely detectable around noon. Daytime I2 mixing ratios around noon ranged from <0.2–0.5 ppt. The source of the daytime I2 was likely very close to the pier. The very short photolysis lifetime (∼5 s) prevents transport of I2 to the site.
 Nighttime I2 levels are strongly influenced by NO3. NO3 reacts rapidly with I2 (k = 1.5 × 10−12 cm3 molecule−1 s−1) [Chambers et al., 1992] to form IONO2 and I atoms. NO3 levels at the pier were estimated to be 12 ppt using a photochemical box model and the NOx and O3 levels measured at the pier. Twelve ppt of NO3 gives an I2 lifetime of 37 min. It would take 30–50 min for I2 from the kelp beds to the southeast to reach the pier at the typical 1–2 m s−1 wind speeds observed. This suggests that when NO3 levels were ∼10 ppt or higher, I2 may not reach the sampling site before being destroyed by NO3.
4.6. Mixed Halogens
 The mixed halogens BrCl, ICl, and IBr were never observed above their detection limits, estimated at 0.5 ppt, 0.5 ppt, and 0.4 ppt, respectively. The two transitions monitored for ICl and the two monitored for IBr were indistinguishable from the blank during the entire period. The three mass transitions monitored for BrCl were never distinguishable from the blank. These results imply that there was not a detectable source of the mixed halogens from production via sea salt aerosols. Gas phase production of mixed halogens from ClO+BrO should be insignificant at the levels of Cl2, Br2, and I2 observed. Other sources, such as Cl + Br2, were also expected to be negligible since halogen atoms were more likely to react with NOx, hydrocarbons, or O3 than with other dihalogens.
4.7. Comparison With Previous Measurements
 The Cl2 levels observed in this study are consistent with other measurements of Cl2. Table 3 compares our results with several other field observations of Cl2. While most of our results are below the detection limits of the previous work, there is overlap in the observed ranges from 15 to 30 ppt. The very high levels (>100 ppt) reported by Spicer et al.  or Keene et al.  were not observed.
Table 3. Summary of Published Measurements of Reactive Chlorines Species in Marine Aira
The tandem mist chamber measures Cl* as Cl2 plus a small contribution from HOCl. These reported ranges assume all of the Cl* is Cl2.
 The Br2 measurements here are, to the best of our knowledge, the first reported in the coastal marine boundary layer so there is no basis for comparison with previous work. These are the first I2 measurements using APCI/MS/MS in coastal urban air. At Mace Head, Ireland, I2 correlated with tidal height, suggesting that emissions were associated with exposure of nearby macroalgae [Saiz-Lopez and Plane, 2004]. The I2 signal at Scripps Pier does not show a significant relationship with tidal height (Figure 8), although the tidal range at the two sites is similar (3 m). The lack of I2 correlation with tidal height may be due to the absence of exposed macroalgae near the pier.
4.8. Implications of Observed Dihalogen Levels
 The Cl2 levels observed at Scripps Pier suggest that chlorine chemistry can have a significant influence on O3 levels in polluted coastal air. Photochemical box model simulations of similar Cl2 levels at Irvine, California suggested that up to 10% of the observed O3 during this study may be attributable to production by chlorine atoms [Finley and Saltzman, 2006; Finley, 2007]. Those results were consistent with previous model studies of halogen-mediated O3 formation in polluted air [Knipping and Dabdub, 2003].
 Halogen atoms generated from the photolysis of Br2 and Cl2 can also affect the conversion of airborne elemental mercury to reactive gaseous mercury. A photochemical box model was used to estimate the levels of Cl and Br atoms resulting from the Br2 and Cl2 measured during this study [Finley, 2007; Yvon et al., 1996]. At these levels, the atmospheric lifetime of elemental mercury (Hg0) is estimated to be 59 days with respect to Br atoms and 150 days with respect to Cl atoms (Table 4). The combined lifetime of 42 days is considerably shorter than the global mean lifetime of 6–24 months [Schroeder and Munthe, 1998]. This suggests that conversion of mercury may be enhanced in coastal urban environments, and that polluted air advected over the oceans may have an impact on the global lifetime of elemental mercury.
Table 4. Lifetimes of Elemental Mercury With Respect to Oxidation by OH, Br, and Cl, Based on the Mean Dihalogen Levels Observed During This Studya
Radical concentrations are 24 h means, estimated using a photochemical box model [Finley, 2007].
 This study demonstrates low parts per trillions detection of dihalogens in air using APCI/MS/MS. The detection limit of the technique is controlled by the variability in the blank signal in carbonate-scrubbed air. The origin of this blank signal is not yet fully understood. The detection limits of the technique may be improved through better characterization and control of the instrument blank.
 The dihalogens Cl2, Br2, and I2 were present at significant and highly variable levels in polluted coastal air. The halogen atoms resulting from the photolysis of Cl2 cause a net production of ozone, while those from Br2 and I2 destroy ozone. The observed dihalogen levels place a lower bound on the possible importance of halogen chemistry, as there may be other precursors of halogen atoms that do not involve dihalogens. An example is the reaction of OH + HCl. This reaction has been shown to be important when HCl levels are >1 ppb [Keene et al., 2007]. Measurements of halogen reservoir species like HOCl or ClONO2 are needed to assess the full extent of halogen photochemistry in tropospheric air. This study suggests that dihalogen chemistry may play a role in coastal urban air quality. Further study is needed to assess the spatial and temporal extent of dihalogens and fully understand their origin, fate, and impact.
 The authors wish to acknowledge John Greaves for analytical advice and Cyril McCormick for engineering assistance. The authors thank Brad Hart and Alan Schoen of Thermo Corporation for their support and technical advice. The authors benefited from scientific discussion with Mike Lawler, Olga Pikelnaya, Jochen Stutz, John Plane, Alfonso Saiz-Lopez, and Roland von Glasow. The authors thank the Scripps Institute of Oceanography for use of the pier, Christian McDonald for logistical support, Ed Parnell for a kelp bed map, and reviewers for helpful criticism and comments. This research was supported by the National Science Foundation under grant ATM-0614816 and is a contribution to U.S. SOLAS.