Iodine chemistry influences the oxidizing capacity of the atmosphere, principally by depleting ozone, and induces the formation of new particles in the marine boundary layer. The photochemistry of iodine dioxide (OIO) plays a key role in both these processes. Here we report that OIO photolyses in the visible (480–650 nm), yielding iodine atoms with a quantum efficiency of unity (1.07 ± 0.15). As a result, much smaller sources of iodine precursors are required to cause significant ozone depletion, which has important implications for the marine boundary layer ozone budget.
 Iodine monoxide (IO) and dioxide (OIO) have been observed in the mid-latitude and tropical marine boundary layer (MBL) and in Antarctica [Alicke et al., 1999; Read et al., 2008; Saiz-Lopez et al., 2007]. Global maps of the IO distribution obtained with the satellite-borne instrument SCIAMACHY have also been published recently [Schoenhardt et al., 2008]. Besides the well-documented impacts of iodine chemistry on the oxidizing capacity of the MBL [Carpenter, 2003], there is strong evidence that the bursts of ultra-fine particles observed in coastal regions are produced by the biogenic emission of I-bearing precursors, followed by photochemical production and condensation of iodine oxide vapours [O'Dowd and Hoffmann, 2005].
 In the troposphere, I atoms are generated from the photolysis of I2 or iodocarbons, and oxidized by O3 to form IO, which subsequently self reacts. At 1 bar pressure and 290 K, OIO and the IO asymmetric dimer (IOIO) are produced with branching ratios of ∼40% and 60%, respectively [Gómez Martín et al., 2007]. Theoretical calculations indicate that IOIO is relatively unstable under tropospheric conditions, and decomposes rapidly to OIO + I [Kaltsoyannis and Plane, 2008]. Thus, OIO is produced by the IO self reaction with an effective yield close to 100%, and is very likely the major precursor to the higher iodine oxides (I2Ox, x = 3 − 5) [Burkholder et al., 2004]. This raises the question of its photochemical stability, since OIO has strong absorption bands between 480 and 650 nm, where photolysis to yield I + O2 is possible thermochemically [Ashworth et al., 2002; Cox et al., 1999; Spietz et al., 2005]. If this did occur then the IO self-reaction and the IO + BrO reaction would overall yield two halogen atoms, which would then react with O3 to reform IO or BrO, thus completing a very effective O3-depleting cycle:
where X = I or Br. Moreover, the photolytic lifetime of OIO would be very short, explaining most field observations [Allan et al., 2001; Saiz-Lopez and Plane, 2004]. Although three recent studies report upper limits to the I atom quantum yield, the lowest of which is 0.05 [Ingham et al., 2000; Joseph et al., 2005; Tucceri et al., 2006], they are in apparent contradiction to the spectroscopic evidence: the absence of fluorescence following photo-absorption, and the lack of rotational structure in the vibrational bands of the absorption spectrum, implying a lifetime of only 200 fs for electronically excited OIO [Ashworth et al., 2002].
 Here we report a study of OIO photochemistry by combining pulsed laser photolysis, cavity ring-down spectroscopy (CRDS) and atomic resonance fluorescence (RF) (see Figure S1 of the auxiliary material). A maximum concentration of 2 × 1012 molecules cm−3 of OIO was produced by excimer laser photolysis at 193 nm of CF3I + N2O or CF3I + O3 mixtures, and probed by CRDS between 526 and 574 nm. Transient absorption from OIO and fluorescence from I atoms generated from the IO self reaction were monitored simultaneously as a function of time by varying the delay between the excimer and the probing pulses (Figure 1). The photolysis of OIO was investigated by directing a third pulsed laser, which was tunable over the wavelength range 450–650 nm, through the photochemically active volume. This photolysis laser was usually delayed with respect to the excimer laser pulse to coincide approximately with the maximum OIO concentration. The removal of OIO (Δ[OIO]), and the production of a secondary I atom signal (SOIOI), were obtained from the difference between the corresponding signals obtained with and without the photolysis laser being fired (Figure 1). The objectives of these experiments were: (a) to calculate OIO ‘photobleaching’ cross sections from the fraction of OIO removed (Δ[OIO]/[OIO]) and the absolute laser fluence, and (b) to see if there was a correlation between Δ[OIO] and SOIOI. In order to determine the iodine atom quantum yield from OIO photolysis (ΦOIOI) from these quantities, a reference photochemical process with a known quantum yield was required. Therefore, calibration experiments were performed immediately after the OIO photolysis experiments at selected wavelengths by flowing I2 diluted in N2 through the reactor and measuring [I2] and the post-photolysis signal, SI2I (Figure 1, bottom). I2 was the obvious choice as the reference because of its strong absorption and well known photolysis quantum yield in the same spectral range [Brewer and Tellinghuisen, 1972; Tellinghuisen, 1973]. The OIO photolysis iodine quantum yield was then calculated using the expression:
where σX (X = OIO or I2) is the absorption cross section, SXI is the secondary I atom signal after photolysis, and ΦXI is the photolysis iodine atom quantum yield [Tucceri et al., 2006]. The product [X] × σX(λ) was derived from the ring-down time recorded at the same wavelength as the photolysis laser.
3. Results and Discussion
 Absorption versus time OIO curves at different wavelengths, pressures, radical concentrations and laser fluences were recorded. Immediately after interrogating OIO with the photolysis laser between 480 and 650 nm, its absorption dropped to an extent depending on the absorption cross section at the photolysis wavelength. Prompt return of OIO to the pre-pulse concentration level was not observed. The behavior of the post-pulse OIO concentration depended on the laser fluence, the concentration of radicals and the chemical scheme employed, and could be explained when using either N2O or O3 using numerical simulations of the kinetics based on standard chemistry (Figures 1 and S2). Following photolysis at 532 nm or 568 nm, no pressure dependence of the subsequent OIO concentration was observed when adding between 4 and 70 Torr N2 (Figure S3). These results render it unlikely that electronic excitation of OIO is followed by internal conversion into vibrationally excited ground state OIO and subsequent collisional relaxation, as previously suggested [Joseph et al., 2005; Tucceri et al., 2006]. From the relative change in OIO signal and the measured laser fluence, photolysis cross sections were calculated (Figures S3 and S6) which are in excellent agreement with the OIO absorption cross sections reported in the literature [Bloss et al., 2001; Spietz et al., 2005]. This strongly suggests a quantum yield of unity for OIO dissociation.
 The remaining question then is which photofragments are generated. Iodine atom production, simultaneous with OIO removal, was observed unambiguously as shown in Figures 1 and S2. Variation of the delay of the photolysis laser resulted in the I atom signal following the temporal behavior of [OIO] (Figure 2). Moreover, the relative yield of I atoms was proportional to the OIO cross sections (Figure 3). Depending on the conditions of the experiment and the delay between the 193 nm and visible laser pulses, non-zero intercepts were found in plots of I atom signal against OIO absorption cross sections (Figure S4), corresponding to I photofragments from photolysis of species other than OIO (mostly I2). This intercept was used, if necessary, to correct the value of SOIOI entering equation (1).ΦOIOI at selected wavelengths between 525 and 575 nm was calculated from equation (1), using values of [X] × σX(λ) and SXI obtained from back-to-back OIO and I2 experiments, and was found to be unity within error: 1.07 ± 0.15, with 1σ uncertainty (Figure 3 and Table S1), over the wavelength range 513–640 nm (Figure S5). At wavelengths below 513 nm, the uncertainty in the iodine atom signal was larger than 50% because of the significant decrease in σOIO.
 An interesting question from the fundamental point of view is the electronic state in which the I photofragment is released, since excited I(2P1/2) is also energetically accessible. To answer this, sufficient O2 (5 Torr) was added to the reaction mixture in order to quench any I(2P1/2) through rapid resonance energy transfer [Young and Houston, 1983]. However, the value of ΦOIOI obtained agreed within error with that obtained in the absence of O2. Since the I atoms generated from the reference I2 photolysis at 70 Torr N2 are all in the ground state (as a result of dissociation of I2[B3Π (0u+)] by collisional dissociative quenching [Brewer and Tellinghuisen, 1972]), it can be concluded that photolysis of OIO also produces I(2P3/2). If that were not the case, rapid quenching would have changed the ratio SOIOI/SI2I in equation (1), and the resulting ΦOIOI would have changed when O2 was added. Production of I(2P3/2) means that both O2(X3Σg+) and O2(a1Δg) are energetically accessible as co-products of OIO photolysis [Ashworth et al., 2002].
 The new results reported here disagree with three previous studies. Joseph et al.  reported prompt recovery of OIO within 3 μs after photo-absorption, which we have not seen under any conditions. Our OIO detection limit is better by one order of magnitude and we use a rigorous approach encompassing sequential monitoring of empty cavity, “photolysis off” and “photolysis on” ring-down times, on a point-to-point basis. Tucceri et al. , whose work superseded the previous study by Ingham et al. , monitored both I and OIO concentrations, but failed to see an iodine atom signal related to OIO. However, we note that Ingham et al.  did observe secondary iodine atom fluorescence correlated to the OIO concentration time dependence at ∼100 mJ cm−2. The more recent and now well established lower absorption cross sections of OIO [Joseph et al., 2005; Rowley et al., 2001; Spietz et al., 2005; Tucceri et al., 2006], and the extremely short lifetime of electronically excited OIO [Ashworth et al., 2002] seem to rule out 2-photon photolysis at that fluence. Tucceri et al.  could have underestimated their detection limit and/or worked with too low a fluence (2.2 mJ cm−2), resulting in a marginal fraction of OIO dissociation. We have also observed that using O3 as the photolytic precursor, as Tucceri et al. did, leads to substantial formation of aerosol and deposition on windows and optics, causing a sharp decrease in the resonance fluorescence sensitivity and poor reproducibility.
4. Atmospheric Implications
 In order to examine the atmospheric relevance of this new result for ΦOIOI we now employ a photochemical box model of halogen chemistry in the MBL [Saiz-Lopez et al., 2007]. A list of the reactions in the model is included in the auxiliary material. The model is used to reproduce a recent set of IO measurements at Cape Verde, an equatorial mid-ocean site which is representative of remote open ocean conditions [Read et al., 2008]. The model was constrained with typical observed peak midday values of NO = 5.0 ppt; OH = 0.3 ppt; HO2 = 12.0 ppt and BrO = 2.5 ppt [Read et al., 2008], while the concentrations of all the iodine species and O3 were allowed to vary ([O3] was initialized at 35.2 ppb [Read et al., 2008]). In the model, the irreversible loss of iodine arises from formation of I2O5 [Saunders and Plane, 2005], which then polymerizes to form ultra-fine particles [O'Dowd and Hoffmann, 2005]. A maximum daytime I atom injection rate of 1.7 × 104 molecule cm−3 s−1 was required to reproduce the observed diurnal IO concentration profile (Figure 4a) with the new value of ΦOIOI = 1. Figure 4b shows that 2.2 ppb of O3 (i.e., 6%) was destroyed by iodine photochemistry over 12 hours. Figure 4c illustrates the predicted peak IO concentration and diurnal O3 depletion as a function of I atom injection rate (all other parameters in the model are the same). The non-linear response of peak IO to the I atom injection rate is mostly caused by the IO self reaction. It should be noted that IO concentrations up to 8 ppt are observed in the coastal MBL when certain types of macroalgae are exposed at low tide [Alicke et al., 1999; Saiz-Lopez and Plane, 2004], and even higher IO levels have been observed in coastal Antarctica [Saiz-Lopez et al., 2007].
 In order to demonstrate the importance of the present study, Figure 4 also contains model predictions where ΦOIOI is set to 0.05, consistent with the upper limit reported previously [Tucceri et al., 2006]. Figure 4a shows that the predicted IO is only 50% of the observed concentrations, because now OIO does not recycle efficiently back to I and IO, and the resulting rate of O3 depletion is smaller by a factor of ∼2 (Figure 4b). In fact, the observed IO can only be modeled with an I atom injection rate that is ∼5 times larger than when ΦOIOI = 1, as shown in Figure 4c.
 A final point to consider is the effect of OIO photolysis on the production of the higher iodine oxides, and hence on the formation of new ultrafine particles. For the scenario in Figure 4 the model predicts that the build-up of I2O5 is delayed during the day by about 2 hrs when ΦOIOI = 1, but there is then a pulse of particle production around sunset when OIO photolysis ceases. This prediction should be tested in future field campaigns.
 This new laboratory study on the controversial photochemical stability of iodine dioxide demonstrates that this radical falls apart after absorption of visible light between 500 and 650 nm to yield iodine atoms with unit quantum efficiency. Photochemical box modeling of halogen chemistry in the equatorial, mid-ocean MBL shows that the photochemistry of OIO should make a significant contribution to O3 depletion over much of the world's oceans, mainly because much smaller iodine source fluxes are required to cause significant diurnal O3 loss. Since there is a strong positive correlation between sea-surface temperature and the emission of iodocarbons [Yokouchi et al., 2008], iodine-catalyzed O3 depletion should now be included in global chemistry-climate models.
 This work was supported by the Natural Environment Research Council (grant NE/E005659/1). We are grateful to T. Ingham, M. Blitz and A. Goddard for helpful discussions and technical support.