Measurements of Cl2 in coastal urban air were made in Irvine, California from Aug.-Nov., 2005, using negative ion atmospheric pressure chemical ionization with tandem mass spectrometry. Cl2 ranged from <2.5 to 20 ppt with a two month mean of 3.5 ± 3.5 ppt (1σ) and an average noon photolysis rate of 44 ppt hr−1 Cl2. Model calculations suggest that Cl2 photochemistry results in an increase of 5–8 ppb in daily maximum O3 levels, relative to a base case scenario without Cl2.
 The OH-initiated oxidation of volatile organic compounds (VOCs) is the primary pathway leading to O3 formation in urban air. Another pathway to O3 formation is oxidation of organic compounds by Cl atoms, generated from photolysis of Cl2 or other precursors, or from the reaction of OH with HCl [Singh and Kasting, 1988]. The reaction rate constants of Cl atoms with many VOCs are 1–2 orders of magnitude faster than the corresponding OH reactions [DeMore et al., 1997]. As a result, Cl competes effectively with OH at much lower concentrations. To date, there have been few direct measurements of reactive halogens in air, and halogen chemistry is not included in standard air quality models.
 A variety of indirect methods have been used to infer the levels of reactive chlorine in air. These include observations of chlorine isotope ratios [Volpe et al., 1998], aerosol chloride deficits [Johansen et al., 2000], tandem mist chamber measurements of volatile inorganic halogen species [Keene et al., 1993; Pszenny et al., 2004], and the ratios of various nonmethane hydrocarbons [Jobson et al., 1994; Singh et al., 1996; Wingenter et al., 1999]. These studies suggest that reactive chlorine precursors may be present in marine air at ppt levels, and that oxidation of organics by Cl atoms has a significant role in tropospheric photochemistry. Spicer et al.  reported the direct detection of Cl2 over Long Island by API/MS/MS, at levels of up to 150 pptv at night in air with marine air mass back trajectories.
 Molecular chlorine (Cl2) levels were measured in this study using negative ion atmospheric pressure chemical ionization with tandem mass spectrometry [Spicer et al., 1998]. The instrument was run nearly continuously from Aug. to Nov., 2005. The sampling site was at Croul Hall on the UCI campus, approximately 4 km inland from Newport Bay. The air sample inlet was situated 3 m above ground level near a parking lot. Concurrent measurements of O3 (10–70 ppb) and NOx (1–100 ppb) were also made using Thermoelectron models 49C and 42C, respectively. Regional meteorological data for this period was obtained from NCEP Reanalysis and local conditions were obtained from measurements recorded at John Wayne International Airport in Santa Ana, about 3 km from the measurement site (National Climate Data Center, Unedited Local Climatological Data, http://cdo.ncdc.noaa.gov/ulcd/ULCD). Air mass back trajectories were computed using the Hysplit v4.7 model [Draxler and Hess, 1997, 1998].
 Cl2 is highly surface active, and a laminar flow, multi-stage inlet was used to minimize wall interactions during transit of the air to the instrument. The sample inlet consisted of a 6 m long 2″ ID plastic pipe (ABS and PVC) operated with an air flow rate of 0.5 m/s (64 SLPM). A ″ ID Teflon tube sampled isokinetically from the center of the larger pipe (2.25 SLPM) to deliver air to the atmospheric pressure ionization source. The ion source was a ″ ID SS tube lined with a 63Ni beta-emitting foil. The source tube sampled air isokinetically from the center of the ″ tube at 1 L/min. This inlet arrangement resulted in a nearly wall-less sampling line with a residence time of 3 seconds and a passing efficiency for Cl2 of 85 ± 10%.
 Cl2 was measured by tandem quadrupole mass spectrometry of Cl2− in single reaction mode (ThermoFinnigan Quantum). The Cl2− parent ions were mass filtered in the first quadrupole, then dissociated by collision with Ar at a pressure of 1 mtorr and 20 eV to form daughter Cl- ions which were mass filtered in the final quadrupole. The two natural isotopes of chlorine (35Cl and 37Cl) result in four possible mass transitions (70/35, 72/35, 72/37, 74/37) for Cl2-. The ratios of the various transitions are determined by the unique isotopic distribution of chlorine (35Cl:37Cl = 3:1).
 Instrument blanks were taken every hour by diverting the inlet air stream through a section of ABS pipe filled with fiberglass wool treated with a 1% (w/w) NaCO3 solution. Every four hours, a Cl2 standard was added to blank air to measure instrument sensitivity changes. A VICI Metronics permeation device was used as a source of Cl2 in N2. A dilution system [Gallagher et al., 1997] was used to generate low-level (ppt) standards that were then introduced to the sample flow. Full instrument calibrations were carried out weekly. These consisted of sequentially adding Cl2 gas standards to scrubbed ambient air at four mixing ratios from 10 to 60 ppt. Several months after this study, the permeation source was calibrated by bubbling through neutral potassium iodide solution and monitoring the formation of triiodide ion (I3−) by UV absorption at 352 nm. The emission of oxidant (expressed as Cl2) was about half of that inferred from the gravimetric calibration. This is consistent with previous observations by Keene et al. . As a result of this calibration, a correction factor was retroactively applied to the absolute calibration of the field measurements. The gravimetric calibration was constant throughout the measurement and post-calibration period. This correction introduces an uncertainty of ±25% into the absolute calibration of these measurements.
Figure 1 illustrates the typical instrument response for a 24 hour period, including blanks and 25 ppt standards added to scrubbed air. The instrument blank signal follows a reproducible diurnal pattern, with lower blank signals in the early morning, increasing around mid-day and reaching a maximum during the late afternoon. The blank remains high until well after sunset, and returns slowly to morning levels. When determining ambient Cl2 levels, these blanks are subtracted from the observed Cl2 signal. The identity of the compound(s) causing these blanks is unknown. The blank appears to be caused by a photochemically produced substance with a lifetime of several hours. We speculate that the blank may be due to a polyhalogenated hydrocarbon which decomposes to form Cl2- following ionization in the mass spectrometer source. The nocturnal decrease in the blank suggests a heterogeneous loss mechanism for the compound(s).
 Instrument sensitivity at the 70Cl/35Cl transition is approximately 5 cps ppt−1 and the instrument blank ranges from 0.5–60 cps. The detection limit for Cl2 is estimated at 2.5 ppt (S/N = 2), imposed by the variability in the blank. All four Cl2 mass transitions are detected in air. The ratio of the signal intensities of the four Cl2 transitions observed in ambient air is consistent with that observed in Cl2 gas standards and with the natural abundance of chlorine isotopes.
 It is important to note that the possibility of positive interference in these measurements cannot be completely ruled out. An interferent would need to both: 1) generate Cl2− in the ion source either by decomposition or ion-molecule reactions, and 2) be effectively removed by the carbonate scrubber. Laboratory tests investigated and eliminated dichloroacetic and trichloroacetic acids (breakdown products of tetrachloroethene and 1,1,1-trichloroethane) and NCl3 (a by-product of wastewater treatment). Aerosols were not removed from the air stream, and remain a potential interferent.
3. Results and Discussion
 Cl2 mixing ratios during this study ranged from <2.5 ppt (below detection limit) to 20 ppt (Figure 2). Daily average Cl2 mixing ratios ranged from <2.5 to 15 ppt. Over the entire measurement period, the average mixing ratio was 3.5 ± 3.5 ppt (1σ). Early in the study (August 24–30), a consistent diurnal pattern was observed, with low Cl2 mixing ratios during the morning (<2.5–5 ppt) and an increase beginning around noon up to 5–10 ppt by the evening. The higher evening Cl2 levels decreased overnight or during the next morning. For the remainder of the study, this diurnal pattern disappeared, and there was little difference between daytime and nighttime Cl2 mixing ratios. There were two periods (August 19–22 and September 1–4) which had elevated Cl2 levels (up to 15 and 7.5 ppt, respectively) that persisted over 24 hours. During the final two weeks of the study, Cl2 was consistently below the detection limit.
 There is no obvious correlation between Cl2 levels and meteorological conditions (wind speed/direction, temperature, relative humidity) or local pollutant levels (NOX, O3). Five day air mass back trajectories show a predominantly marine origin for air masses during the study. Most of the air mass trajectories were northwesterly, originating over the Pacific ocean at northern mid-latitudes. Aug 24–27 was a period of southwesterly trajectories originating over the subtropics. The last two weeks of the study (October 13–27) were characterized by cloudy or overcast skies with regular drizzle and rainfall.
 These are the first extensive measurements of Cl2 over full diurnal periods. Daytime values typically ranged from below detection to 5 ppt, but on occasion reached levels as high as 15 ppt. The observation of such high daytime levels is surprising, given the short photolysis lifetime of Cl2.
4. Comparison With Previous Measurements
 The only previously published mass spectrometric measurements of Cl2 are those of Spicer et al. , who reported Cl2 levels ranging from <10–150 ppt in marine air over Long Island from sundown to shortly after sunrise. Keene et al.  reported 13–127 ppt of Cl* (Cl2 + HOCl, as Cl2) in Miami using a tandem mist chamber. Pszenny et al.  reported <6 to 38 ppt Cl* in Oahu with the same technique. There is considerable overlap between the ranges of these published studies and the Cl2 levels reported in this study.
5. Implications for Ozone Production in Coastal Urban Air
 The photolysis rate of Cl2 was estimated based on the mixing ratios observed in this study, absorption coefficients from Maric et al. , and a two-stream radiative transfer model [Madronich and Flocke, 1999]. These calculations yield a noontime photolysis lifetime of 3.7 minutes for 33.6° latitude, −117.8° longitude, Julian day 244, 287 DU total ozone column, and an albedo of 0.5. The short lifetime of Cl2 implies that it must have a daytime source, presumably photochemical in nature. At night, one expects surface deposition to deplete Cl2, and there must be some nighttime production to balance this relatively slow loss.
 Based on the observed ambient mixing ratios, Cl2 photolysis represents a source of Cl atoms on the order of 88 ppt/hr. In urban air, Cl atoms react primarily with VOC's to form peroxy radicals. These peroxy radicals ultimately result in the oxidation of NO to NO2, which photolyzes to produce O3. A photochemical box model incorporating the gas phase chemistry of Cl2, Cl, ClO, HCl, HOCl, and ClONO2 was used to simulate the study conditions. Simulations suggest that chlorine-mediated oxidation of VOC's can account for several ppb per hour of additional ozone production during daylight hours. Assuming a constant Cl2 mixing ratio of 3.5 ppt, the model yielded a noon Cl atom concentration of 1.2 × 105 cm−3 and a noon OH concentration of 8.5 × 106 cm−3. The Cl2 increases peak O3 levels by 5–8 ppb over a base case scenario with no chlorine chemistry.
 These model results are consistent with previously published simulations using a 3-dimensional regional air quality model of the California South Coast Air Basin with detailed chlorine chemistry [Knipping and Dabdub, 2003]. In those simulations, 12 pptv of Cl2 at sunrise led to an O3 increase of 12 ppb, compared with no-chlorine conditions. If the observations in this study of detectable Cl2 levels throughout the day are correct, then Cl-mediated ozone production constitutes a substantial fraction of the total ozone production.
6. Implications for Cl2 Production Mechanisms
 The observed Cl2 levels may reflect either gas phase/aerosol interactions or direct emissions from anthopogenic sources such as industrial and water treatment plants, cooling towers, chlorinated domestic water supplies, fuel combustion, or incineration. The field observations place some constraints on the mechanism of Cl2 formation. For Cl2 to be present during daylight hours, when the photolysis lifetime is only minutes, requires a large production rate. The fact that Cl2 does not exhibit a large and consistent diurnal cycle suggests that the rates of production and photolysis covary. This implies a production mechanism involving photochemical precursors. The OH-mediated surface reaction mechanism [Knipping et al., 2000] is a potential source of daytime Cl2, but it does not appear sufficient to explain the observed levels. Box model calculations [Yvon et al., 1996] were carried out using the parameterization of Knipping and Dabdub , with a seasalt aerosol surface area of 145 um2 cm−3, estimated from Lewis and Schwartz , for 10 m height, 5 m s−1 wind speed, 80% RH, and an aerosol Cl- concentration of 5 M [Herrmann et al., 2003]. Under these conditions, the noontime concentration of Cl2 reached only 0.5 ppt, a factor of 7 less than observed. By itself, the OH surface mechanism also leads to a diurnal pattern with a midday maximum. The rate of production of Cl2 involving the seasalt aerosol surface reaction of ClONO2 was also estimated using a similar approach [Caloz et al., 1996; Knipping and Dabdub, 2003]. Inclusion of this mechanism leads to a Cl2 level of about 1 ppt. These two mechanisms together account for 44% of the observed average Cl2. Inclusion of a ClONO2 mechanism better models the observed temporal behavior by smoothing out the diurnal variations compared to the OH-only surface reaction. Clearly, future studies incorporating both gas phase and aerosol measurements are needed in order to rigorously test various production mechanisms.
 The ambient levels of Cl2 observed in this study suggest that current air quality models in coastal regions may either underestimate the overall daily ozone formation rate or misattribute a significant amount of ozone production to OH-mediated, rather than Cl-mediated, oxidation of VOC's. Further research is needed to: 1) verify these observations and explore the ambient distribution of Cl2 and other reactive halogen species, and 2) fully understand the mechanisms by which reactive halogen compounds are formed in the troposphere.
 The authors wish to thank John Greaves of the UCI mass spectrometer facility and Brad Hart of ThermoFinnigan Corp. for analytical advice. We also wish to thank an anonymous reviewer for raising the important issue of Cl2 permeation tube artifacts. This research was supported by the NSF Atmospheric Chemistry Program (ATM-0403421).