Ultraviolet absorption cross sections of carbonyl sulfide isotopologues

S. Hattori, S. O. Danielache, M. S. Johnson, J. A. Schmidt, H. G. Kjaergaard, S. Toyoda, Y. Ueno, and N. Yoshida Department of Environmental Science and Technology, Interdisciplinary Graduate School of Science and Engineering, Tokyo Institute of Technology, Yokohama, 226-8502, Japan Copenhagen Center for Atmospheric Research, Department of Chemistry, University of Copenhagen, Universitetsparken 5, 2100 Copenhagen, Denmark Department of Environmental Chemistry and Engineering, Interdisciplinary Graduate School of Science and Engineering, Tokyo Institute of Technology, Yokohama, 226-8502, Japan Department of Earth and Planetary Science, Tokyo Institute of Technology, Meguro-ku, Tokyo, 152-8551, Japan


Introduction
The stratospheric sulfate aerosol (SSA) layer plays an important role for Earth's radiation budget and modulates the concentration of stratospheric ozone due to surface heterogeneous reactions (Junge, 1966;Myhre et al., 2004;Mills et al., 2005).Sulfur compounds from volcanic eruptions such as Mt.Agung (1963) and Mt. Pinatubo (1992) produce large amounts of SSA.However, SSA persists even in the absence of volcanic input indicating a background source (Crutzen, 1976;Hofmann, 1990).Carbonyl sulfide (OCS) is the most abundant sulfur containing gas in the atmosphere, with an average mole fraction of 0.5 × 10 −9 throughout the troposphere (Watts, 2000;Notholt et al., 2003).Since the discovery of SSA (Junge and Manson, 1961), the photolysis of OCS was thought to be a significant source because it is transported into the stratosphere where it is broken down by photolysis and radical reactions with OH and O( 3 P).Chin and Davis (1995) used a 1-D model to show that atmospheric OCS is inadequate to maintain the background SSA concentration; other significant of SSA sources including deep S. Hattori et al.: Ultraviolet absorption cross sections of carbonyl sulfide isotopologues convection of SO 2 from the lower troposphere (Weisenstein et al., 1997;Kjellström, 1998;Pitari et al., 2002;Stevenson et al., 2003) and an in situ sulfuric acid photolysis and sulfur dioxide oxidation process (Mills et al., 2005) have been proposed.In contrast Barkley et al. (2008) found the stratospheric lifetime of OCS to be 64 ± 21 a from ACE satellite measurements, corresponding to a stratospheric sink of 63-124 Gg S a −1 , which matches the source needed to sustain the SSA production of 30-170 Gg S a −1 (Chin and Davis, 1995).The background source of SSA is not known with any certainty.
Isotopic analysis is used to trace the sources and transformations of atmospheric trace gases (e.g.Krouse and Grinenko, 1991;Johnson et al., 2002;Brenninkmeijer et al., 2003).The observed isotopic composition of a species and chemical isotopic fractionations are linked to its sources by the mass balance equation.In addition to 34 S/ 32 S ratios, non-mass-dependent sulfur isotopic compositions have been found in the Archean rock record, Martian meteorites, and sulfate aerosols (Farquhar et al., 2000;Baroni et al., 2007Baroni et al., , 2008)).The δ 34 S value of tropospheric OCS has not been measured but was estimated to be 11 ‰ (Krouse and Grinenko, 1991).Concentration profiles of OC 34 S and OC 32 S retrieved from infrared limb transmittance spectra acquired by the JPL MkIV instrument indicated an extremely large photochemical fractionation ((73.8 ± 8.6) ‰; Leung et al., 2002).This isotopic fractionation is not consistent with the one known measurement of δ 34 S in SSA (2.6 ‰; Castleman et al., 1974).In contrast, Lin et al. (2011) recently reported relatively small fractionation in the range of −10.5 to +5.3 ‰ in 34 S in laboratory photolysis of OCS suggesting the possibility that OCS may be a viable source of SSA.Modeling shows that while reactions with OH and O( 3 P) are important for removing OCS in the lower stratosphere, photolysis is the dominant loss process above 14 km, accounting for over 80 % of the chemical sink above 20 km (Danielache et al., 2008b).Therefore, we have focused on sulfur isotopic fractionation in OCS photolysis in this study.
Like carbon dioxide (CO 2 ) and nitrous oxide (N 2 O), OCS is a linear 16 valence electron atmospheric trace gas with forbidden electronic transitions that become allowed as the molecule bends (Suzuki et al., 1998): The excited state potential energy surface imparts a strong torque on the system producing a highly rotationally excited CO fragment and trapping trajectories on the repulsive excited state potential energy surface long enough to give weak vibrational structure.Since the photolysis quantum yield is unity from 200 to 300 nm (Rudolph and Inn, 1981), the absorption spectrum of OCS can be used to calculate its photolysis rate.Isotopic substitution results in shifts in the absorption spectra of the substituted species; in the atmosphere this results in fractionation because the photolysis rate is dependent on the convolution of the actinic flux with the absorp-tion cross section (Jørgensen et al., 2008).For example, the absorption cross sections of N 2 O isotopomers and isotopologues (von Hessberg et al., 2004) and SO 2 isotopologues (Danielache et al., 2008a) have been investigated and demonstrate that photodissociation results in non-mass-dependent isotopic fractionation.In previous studies the absorption cross sections of commercial OC 34 S and natural abundance OCS were measured to determine isotopic fractionation as a function of wavelength (Colussi et al., 2004), and theoretical calculations of OCS isotopologue cross sections have been made (Danielache et al., 2009).The former study reported that OC 34 S has a stronger absorption than natural abundance OCS resulting in a large positive isotopic fractionation.The latter work reported significantly smaller isotopic fractionations than that of former study.Similarly, in a recent laboratory photolysis study of OCS, Lin et al. (2011) observed relatively small sulfur isotopic fractionation.The ultimate goal of the different approaches is the same: to investigate the sulfur isotope effect on OCS photolysis.Current theoretical methods (Danielache et al., 2009), however, are not precise enough to reconstruct the weak vibrational structures and magnitude of the absorption cross section.For laboratory experiments of OCS photolysis (Lin et al., 2011), the precision of the measurement is better than other approaches but the light source does not duplicate the stratospheric actinic flux.Therefore, one objective of this study is to make sense of sulfur isotopic fractionation using laboratory absorption cross section measurements.For this purpose, we synthesized the sulfur isotopologues of OCS using the same carbon monoxide material to isolate the effect of 13 C, 18 O and 17 O isotope substitution.
Here, we present the UV absorption cross sections of isotopically substituted OCS (OC 32 S, OC 33 S, OC 34 S, and O 13 CS) in the 195-260 nm region.Isotopic fractionation constants in photolysis as a function of wavelength for sulfur and carbon isotopes are calculated, and implications for atmospheric chemistry are discussed.

Synthesis of OCS isotopologues and purification
The OCS isotopologues OC 32 S, OC 33 S, and OC 34 S were synthesized by chemical conversion according to Ferm (1957).This procedure involved the reaction of carbon monoxide (CO) (Japan Fine Products, 99.99 % purity) with 1-5 mg of 32 S (99.99 % isotopic enrichment), 33 S (99.8 % isotopic enrichment) and 34 S (99.9 % isotopic enrichment) powders (Isoflex, USA).The elemental sulfur powders were put in Pyrex glass tubes (i.d. 9 mm) and the air was evacuated.After evacuation, a stoichiometric excess of CO (2 to 3 times) was added and the tubes were sealed.The samples were heated at 573 K for 24 h, converting S into OCS.
The product was expected to include impurities such as CO, CO 2 , hydrogen sulfide (H 2 S) and carbon disulfide (CS 2 ).In particular, CS 2 has an intense overlapping UV absorption in the region 190-212 nm that is ca.20 000 times larger than that of OCS (Molina et al., 1981).The samples were purified using a gas chromatograph (GC) equipped with a thermal conductivity detector (TCD) (GC-14B; Shimadzu, Kyoto, Japan) and packed column (Pora Pack Q, 2 mm i.d., 2.4 m) maintained at 333 K. Ultra-pure He (Japan Air Gases, >99.99995 % purity) was used as the carrier gas at a flow rate of 25 ml min −1 .The GC oven temperature program was 333 K for 8 min, increased from 333 to 473 K at a rate of 25 K min −1 , and then held at 473 K for 2 min.OCS with a retention time of ca.5.2 min was trapped at liquid nitrogen temperature (77 K) after the GC.The GC column was carefully baked between the samples to avoid contamination from other isotopologues and other components.The retention times of impurities were 3.0 min, 4.0 min and 12.0 min for CO 2 , H 2 S and CS 2 , respectively.Reanalysis of the purified OCS sample confirmed that these impurities were below the detection limit (<0.1 %).Purified OCS was transferred into evacuated glass tubes and stored in the dark at room temperature before experiments.The recovery of OCS was 70-80 % and the same CO sample was used to synthesize OC 32 S, OC 33 S and OC 34 S. Since the method does not involve sulfur other than the reagent, isotopic contamination of the samples was negligible as verified with infrared spectroscopy (Bruker IFS 66 V/s).
For13 C-enriched OCS, commercial O 13 CS (Icon stable isotopes, 13 C 99 % purity) was used.Since the O 13 CS sample was assumed to include impurities such as CS 2 and CO 2 , the commercial sample was also purified according to procedure described above.

Spectroscopy
Like N 2 O, OCS is photolyzed in the stratospheric UV window.This window coincides with the low energy shoulder of the N 2 O absorption cross section, which is sensitive to temperature.However, in the case of OCS, the stratospheric UV window coincides with the high energy side of the absorption cross section which is only weakly sensitive to temperature.Room temperature measurements are therefore adequate as a first approximation.
Spectra were recorded at (295 ± 1) K using a UV/Vis dualbeam photo spectrometer (Lambda 1050 NB; Perkin Elmer Inc., USA) equipped with deuterium lamp, using a 10.00 cm long cell with CaF 2 windows.The spectral resolution was set to 0.1 nm and data were recorded at 0.02 nm (2.96 to 5.26 cm −1 ) intervals.Sample pressures were measured using a 133 mbar range capacitance manometer (0.15 % precision; Edwards Inc. UK) calibrated to the Danish national standard.Absorptivities were derived using the measured spectrum of the empty cell and of the cell including sample.Background, sample and a second background spectrum were measured in sequence to correct for possible drifts in the system, e.g.lamp intensity and detector and electronics sensitivity.One measurement took approximately 50 min and three to four repetitions of measurements of a single OCS isotopologues were carried out in a day.After evacuation of the gas line and the cell during the night, the next isotopologue measurement was carried out.The measurements for the OCS isotopologues were carried out over a continuous series of days.Wavelength was calibrated internally using the Shumann-Runge bands of O 2 present in the spectrometer.Sample pressures ranged from 8 to 23 mbar.In this range, we confirmed that the cross section values were independent of gas pressure.Measurement sequences were performed 10 times, 6 times, 6 times, and 6 times for OC 32 S, OC 33 S, OC 34 S, and O 13 CS, respectively.
The measured intensities were converted to absorption cross sections, σ (λ), using the Beer-Lambert law: where N is the OCS number density, l is the path length of the cell, I (λ) is the intensity as a function of wavelength through cell with sample, and I 0 (λ) is the intensity through the evacuated cell.I 0 (λ) values were taken as the average of intensities measured before and after sample intensities.

Wavelength dependent of isotopic fractionations
In order to describe the change of cross section for heavier isotopologues the isotopic fractionation factor (α) as a function of wavelength (λ/nm) was investigated.However, the samples used in this study contained small isotopic impurities.The effect of the impurities was corrected using the isotopic purities provided by the manufacturer (see Appendix).Since the isotopic fractionation factors (α) are generally close to unity, it is common to quantify the isotope effect in terms of the fractionation constant (ε) and ε is often given in per mil notation (‰).In this study ε for sulfur and carbon isotopes are described using Eqs.( 2) and (3),

Isotopic fractionation constants of OCS photolysis at 20 km altitude
From the measured absorption cross sections of the different isotopologues, it is possible to compute an effective fractionation constant for solar photolysis in the lower stratosphere.
The photolysis rates constants (J ) can be evaluated using Eq. ( 5).
x J = where ϕ is the quantum yield of OCS photolysis which is assumed to be 1 for all isotopologues, and the index x represents the isotopologue.In order to investigate the fractionation of each isotopologue relative to OC 32 S in the OCS photolysis region, the actinic flux at 20 km altitude was used as I from 195-260 nm.This actinic flux was provided by Chris McLinden of Meteorological Service of Canada (McLinden et al., 2002).Actual fractionation constants can be calculated using Eqs.( 6) and ( 7).
The mass-dependent relationship describing the equilibrium distribution of three sulfur isotopes between phases has been established by Hulston and Thode (1965) as following Eq.( 8).

Sulfur isotopic compositions
Although carbon isotope ratios in OCS have not been reported, sulfur isotope ratios and fractionations in OCS and its related compounds have been considered in several studies.Here, sulfur isotopic compositions are reported as where x represents 33 or 34 and R denotes the isotope ratios ( 33/34 S/ 32 S) of samples and standards.Isotope ratios of sulfur are reported relative to the Vienna Canyon Diablo Troilite (VCDT) standard and often denoted in per mil (‰).
In addition to δ 34 S values, non-mass-dependent fractionation in sulfur is an important quantity.The capital delta notation ( 33 S) is used to describe and distinguish between mass-dependent and non-mass-dependent isotopic compositions, where 33 S = δ 33 S − δ 34 S + 1 0.515 − 1 . (11) This notation describes the excess or deficiency of 33 S relative to a reference "mass-dependent" fractionation array.

Error budget
Errors in the absorption cross sections come from several sources.Some error is due to random variation in the spectrometer (detector and amplifier sensitivity, lamp stability etc.).Error in sample density arises from the precision of the pressure gauges (±0.15 % according to the manufacturer).Measurements used the same cell (path length 10.00 cm) and the temperature of the laboratory was (295 ± 1) K; overall, error due to measurement of pressure was less than 0.4 %.These errors were propagated and included standard deviations as shown in Fig. 2a-c.The average relative standard deviations (RSD; %) in the 195-260 nm region for OC 32 S, OC 33 S, OC 34 S and O 13 CS are 2.4 %, 2.3 %, 2.2 % and 3.1 %, respectively.In the high absorbance region (210 to 235 nm) RSDs are less than 0.5 %.The isotopologues RSDs as a function of wavelength are shown in Fig. A1 of the Supplement.Sample purity may result in systematic error independent of the number of measurements.After the purification detailed above we evaluated sample purity using FTIR spectroscopy, similar to the method of Danielache et al. (2008a).The concentrations of infrared active impurities were determined from the IR spectra using the global nonlinear least squares spectral fitting procedure developed by Griffith (1996).Line parameters were taken from the HITRAN database (Rothman et al., 2005).The samples used in this study contain CO 2 at mole fractions of 0.04 %, 0.04 %, 0.03 %, and 0.04 % for OC 32 S, OC 33 S, OC 34 S and O 13 CS, respectively; other impurities could not be detected.

Comparison with previous studies
The absorption cross section of OC 32 S is shown in Fig. 1.Vibrational structure can be seen superimposed on a broad Gaussian-like absorption band typical of direct dissociation.The measured absorption cross section of OC 32 S is consistent with previous studies using natural abundance OCS samples such as Molina et al. (1981) and Wu et al. (1999) (Fig. 1).In particular, the difference between the present study and the data of Wu et al. (1999) was less than 1 % in the high absorption region (210-230 nm).Despite the lower spectral resolution of our study in comparison with Wu et al. (1999), the vibrational structure was resolved and reproduced (Fig. 1).At wavelengths shorter than 210 nm, the cross sections of Wu et al. (1999) are larger than those of OC 32 S in this study, but this difference is due to their sample's contamination by CS 2 as discussed in Wu et al. (1999).

Comparisons of absorption of cross sections of OCS isotopologues
The isotopologue's absorption cross sections are shown in Fig. 2. While similar structures are observed for OC 32 S, OC 33 S, and OC 34 S, O 13 CS has a slightly different structure.
Peak positions vary systematically from OC 32 S, OC 33 S, OC 34 S to O 13 CS, the shift being approximately linear in transition energy (Fig. 3).A similar pattern of shifting was observed for SO 2 isotopologues (Danielache et al., 2008a) and N 2 O isotopomers and isotopologues (Selwyn and Johnston, 1981;Schmidt et al., 2011).The maximum in the absorption cross sections, σ max , occurs at 223.44 nm, 223.50 nm, 223.54 nm, and 223.54 nm for OC 32 S, OC 33 S, OC 34 S, and O 13 CS, respectively (Table 1).
It is noteworthy that the largest changes in integrated intensity, width of absorption envelope and shift in position of vibrational structure was seen for 13 C which is expected based on the change in the ground state vibrational wavefunction (Danielache et al., 2009) and comparison to N 2 O (Johnson et al., 2001).In both molecules the transition is forbidden when the molecule is linear, and becomes allowed as the system bends.As a result there is an unusually large increase in the transition intensity with bending excitation.Substitution at the carbon atom changes the degree of bending excursion much more than sulfur substitution, resulting in weaker intensity across the band.
We compare our OC 32 S and OC 34 S cross sections with those of Colussi et al. (2004).Although the red-shift of σ max for OC 34 S relative to OC 32 S is consistent with this work, the reported σ absorption intensity of OC 34 S relative to OC 32 S is opposite to the results of this study.Colussi et al. (2004) compared the spectrum of natural abundance OCS (about 90 %) and commercial OC 34 S after distillation using an Hg(II)SO 4 solution, and their sample of OC 34 S contained about 30 % CO 2 .This level of impurity likely affected the determination of the absolute cross sections; we note that the absolute cross section of their natural abundance sample is about 10 % larger than the cross section reported by Wu et al. (1999).As presented above, our samples were S. Hattori et al.: Ultraviolet absorption cross sections of carbonyl sulfide isotopologues carefully purified and the impurities were documented by several methods.In addition to the absence of impurities in the samples in the present study, the carbon and oxygen isotopic purity is the same for OC 32 S, OC 33 S and OC 34 S. Colussi et al. (2004) predict a large photolytic isotopic fractionation of sulfur and this is not consistent with our result.

Shifted spectra of OCS isotopologues
The positions of the vibrational peaks of the heavier isotopologues (OC 33 S, OC 34 S and O 13 CS) were generally redshifted relative to the most abundant species (OC 32 S) cf.Fig. 3.This red-shift in vibrational structure for heavier isotopologues is consistent with a Franck-Condon model in which isotopic substitution changes the positions of the vibrational levels of the ground and excited electronic states.
At the lowest transition energies the zero point energy (ZPE) shift of the ground state is larger than the shift in the excited state.With increasing energy transitions are made to upper state levels with increasing vibrational excitation.For these states the isotope shift in frequency is larger than the change in the ground state, leading to red-shifting of the vibrational structure extending over most of the band.The exact position of the absorption maximum (Fig. 2) is due to a vibrational peak, and so the position of the maximum in Table 1 is red shifted for isotopic substitution.
For direct photodissociation, the absorption cross section is often fairly well approximated as an energy weighted Gaussian function.The reflection principle states that the absorption spectrum can be approximated by reflecting the initial vibrational wavefunction onto the energy axis across the repulsive excited state surface (Schinke, 1993).Energy weighted Gaussian functions (Eq.12) were fitted to the isotopologue cross sections in order to discuss changes in the overall shape of the band.
where E max indicates the wavenumber of σ max , E ph is the photon energy, and w is a width parameter.Plots of ln (σ (E ph )/E ph ) vs. E ph in the high absorbance region (210-235 nm) show the center (Center) and the full width at half maximum (FWHM) of the fitted spectra as summarized in Table 2.This shows that the heavy isotopologue band centers are blue-shifted, in accordance with the prediction of the reflection principle and the ZPE theory (Miller and Yung, 2000).In addition, the width (FWHM) is more narrow for the heavy isotopes as predicted by Liang et al. (2004).This is also expected since, for direct dissociation, the width of the cross section is proportional to width of the ground state vibrational wavefunction (Schinke, 1993).Consequently, the isotope effect on the UV absorption cross sections of OCS isotopologues is seen to arise from three effects: blueshifting and a more narrow overall band for heavier isotopo- logues, and red-shifting of the weak vibrational structure for heavier isotopologues.

Integrated absorption cross sections
Integrated absorption cross sections for the 195-260 nm band vary with isotopic substitution (Table 1).In order to determine the fractionation constants from the measured cross section as described in next section, it is important to accurately determine the overall magnitude of the individual cross sections.This in turn, requires pure samples and an accurate determination of the pressure in the cell.To estimate the reliability of the magnitudes of the measured cross sections we compare the experimental integrated σ (E ph )/E ph with theoretical results.
The theoretical results are obtained using a recent ground state potential energy surface and transition dipole moments (Danielache et al., 2009).
The integrated intensity, s, is calculated using The quantity s can also be obtained from the following expression (Schinke, 1993): where R are the nuclear coordinates, i is the i-th vibrational wavefunction in the electronic ground state, µ FI is the transition dipole moment, C is a universal constant and W i is the Boltzmann weighting factor, where d i is the degeneracy of state i.Only the vibrational ground state and first excited bending and stretching states were considered.Table 1 shows integrated energy weighted cross sections relative to the most abundant isotopologue.The relative integrated absorption cross section measurements are in excellent agreement with theory, showing deviations of less than 0.3 %.The differences are due to several factors.One is the difficulty of accurately predicting transition intensities within quantum chemical methods.In addition there are systematic errors in the measurement as described above.

Isotopic fractionation constants for OCS photolysis
Using Eqs.(2-4), wavelength dependent isotopic fractionation constants were calculated and are presented in Fig. 4 for sulfur isotopes and Fig. 5 for carbon isotopes.Since OCS does not have a highly structured absorption, the isotopic fractionation constant does not show the highly structured fractionation spectrum seen for SO 2 isotopologues which varies from −1000 ‰ to +1000 ‰ (Danielache et al., 2008a).
It is worth noting that the large variation in the isotopic fractionation at low and high energies may be due to measurement errors occurring at regions of low absorptivity ( Isotopic fractionation constants for sulfur species ( 33/34 ε (λ)) and for 33 S excess ( 33 E (λ)) plotted as a function of wavelength; this study (a), theoretical calculation (Danielache et al., 2009) (b), ZPE estimation (Miller and Yung, 2000;Danielache et al., 2009) (c).and 5), but the trend of the wavelength dependence of isotopic fractionation can be seen.Generally it is close to zero on average on the higher energy side (195-230 nm) in both 33 ε and 34 ε.In contrast, on the lower energy side, negative fractionation constants are observed that are in good agreement with the ZPE estimation and Danielache et al. (2009) (Fig. 4).In contrast, 13 ε on the higher energy side was lower than that of the ZPE estimation, suggesting the effect of narrowing is dominant in this region (Fig. 5).6) as a function of wavelength.Black line: this study; black dotted line: theoretical calculation (Danielache et al., 2009); gray line: ZPE estimation (Miller and Yung, 2000;Danielache et al., 2009).

Prediction of sulfur isotopic fractionation of OCS photolysis in the modern atmosphere
The calculated value of I (λ) 32 σ (λ)ϕ (λ) shows a maximum at 210-211 nm (Fig. 6).This finding is largely consistent with previous work by Colussi et al. (2004) which found the maximum of I (λ) 32 σ (λ) at approximately 209 nm.Calculated 32/33/34/13 J values and estimated fractionation constants ( 33 ε, 34 ε and 13 ε) at 20 km altitude are summarized in Table 3.The estimated 32 J is approximately 1.3 × 10 −8 molecule s −1 , being of the same order as a previous estimate (Chin and Davis, 1995).As described above, our measurements have not only random errors but also systematic errors.Therefore, as a total estimated error, we propagate a standard deviation from repeated measurements and a systematic error of 3 ‰ which is estimated from the differences between integrated cross sections and theory.
The predicted isotopic fractionation constants in OCS photolysis at 20 km altitude are (−3.7 ± 4.5) ‰, (1.1 ± 4.2) ‰, (−26.8 ± 4.3) ‰, and (−4.2 ± 6.6) ‰ for 33 ε, 34 ε, 13 ε and 33 E, respectively (Table 3).This 34 ε is inconsistent with the observed isotopic fractionation in 34 S of (73.8 ± 8.6) ‰ in OCS decomposition for the lower stratosphere derived using stratospheric OC 34 S and OC 32 S concentration profiles from the infrared limb transmittance spectra acquired by the JPL MkIV instrument (Leung et al., 2002).First, consider that the sulfur isotopic fractionation in 34 S of 73.8 ‰ may not be due to OCS photolysis as a significant amount of OCS is removed by radical reactions.However the 34 ε of OCS + OH is estimated to be small (−2.56 ‰; Danielache et al., 2008).Further, the apparent fractionation derived from concentration profiles is usually about half of the process fractionation constant due to mixing  (McLinden et al., 2003), so the isotopic fractionation in 34 S of (73.8 ± 8.6) ‰ from the MkIV instrument analysis (Leung et al., 2002) does not match the previously estimated isotopic fractionation in OCS photolysis in 34 S of (67 ± 7) ‰ (Colussi et al., 2004).One concern is that these retrievals are difficult and have yet to be confirmed in other studies (either via remote sensing or sampling).We conclude that further observations are required to confirm or disprove the observed OCS isotopologues profiles reported by Leung et al. (2002).

OCS as a candidate of background SSA sulfur origin
δ 34 S values of 11 ‰ in tropospheric OCS were estimated by mass balance of oceanic sulfur and terrestrial sulfur (Krouse and Grinenko, 1991).The OCS oxidation product, background SSA, is reported to be 2.6 ‰ (Castleman et al., 1974), indicating that background SSA is less enriched in 34 S than tropospheric OCS.Our results show that OCS photolysis gives only a small isotopic fractionation of 34 ε ((1.1 ± 4.2) ‰) and Danielache et al. (2008b) reported that the OCS reaction with OH also has a small and negative 34 ε (−2.56 ‰).Consequently, sulfate from OCS sink reactions will not be fractionated more than 5 ‰ compared to the initial material, showing that OCS is an acceptable source of background SSA.
Recently Lin et al. (2011) reported the results of laboratory experiments on OCS.They found that small values of 34 ε (−10.5 to +5.3 ‰) in λ > 280 nm experiments is consistent with the ZPE fractionation estimate.Their negative values of 34 ε on the low energy side (λ > 230 nm) due to the ZPE shift of heavier sulfur isotopologues are largely consistent with this study (Fig. 4), although it is noteworthy that UV light from 205 to 215 nm is predicted to be most important for OCS photolysis in the modern atmosphere (Fig. 6).In addition, recently, the sulfur isotopic Table 3. Photolysis rate of each OCS isotopologues at 20 km altitude ( 32/33/34/13 J ) and isotope fractionation constants ( 33/34 ε and 13 ε) and non-mass-dependent fractionation constant ( 33 E).  and Grinenko, 1991) also should be reconsidered.Therefore, in the future, high precision determinations of the sulfur isotopic composition atmospheric OCS are needed to confirm both whether δ 34 S value in OCS of 11 ‰ is correct and whether OCS can be the major source of background SSA.

Non-mass-dependent signatures in sulfate in the modern and Archean atmospheres
Although there are no reports of 33 S values in background SSA, polar ice core records of sulfate after stratospheric sulfate aerosols show initially positive 33 S values (Baroni et al., 2008), suggesting there is non-mass-dependent fractionation in stratospheric sulfur chemistry.SO 2 photolysis and photo excitation, and SO 3 photolysis have been suggested to cause the non-mass-dependent signature observed in sulfate aerosol (Savarino et al., 2003;Pavlov et al., 2005;Baroni et al., 2008).In this study, 33 E at 20 km altitude is (−4.2 ± 6.6) ‰, mass-dependent to within the error (Table 3), and this is also consistent with the OCS photolysis experiment in the laboratory (Lin et al., 2011).Thus, OCS photolysis at 20 km altitude cannot explain the often observed positive 33 S values in sulfate.Additionally, both δ 34 S and 33 S have positive values in glacial sulfate resulting after a stratospheric volcanic eruption (Baroni et al., 2007), and this also suggests that OCS photolysis does not contribute sulfate after stratospheric volcanic eruption.
Non-mass-dependent fractionation is found in Archean sedimentary sulfide and sulfate, suggesting atmospheric origin of the anomalous fractionation (Farquhar et al., 2000).The reducing Archean atmosphere may have contained elevated levels of OCS (Ueno et al., 2009), and it is therefore interesting to test whether or not photolysis of OCS may cause non-mass-dependent fractionation.Our results demonstrate that OCS photolysis does not cause large non-massdependent fractionation in 33 S, and thus this can be ruled out as the source of the non-mass-dependent signatures seen in the geological record, in agreement with the prediction of Lyons (2009) and laboratory experiment of OCS photolysis (Lin et al., 2011).

Conclusions
The main result of this study is that OCS photolysis at 20 km altitude is not expected to produce isotopic fractionation larger than ca. 5 ‰ in 34 ε.The results do not support the large isotopic fractionation predicted by Leung et al. (2002) and apparently observed by Collusi et al. (2004).Additionally, the small isotopic fractionation for OCS photolysis as well as the OCS sink reaction with OH indicates that OCS can be the major source of background SSA.Investigation of isotopic fractionations in OCS sink reactions with O( 3 P) radicals, and high precision measurement of isotopic composition of OCS in atmosphere are needed in future studies.Finally, based on our measurements OCS photolysis is not responsible for the positive 33 S values observed in sulfate aerosols.

Fig. 1 .
Fig. 1.Measured absorption cross section of the OC 32 S isotopologue compared to the spectra of natural isotopic abundance samples measured by Molina et al. (1981) (long dashed line) and Wu et al. (1999) (short dashed line) in the region of 190-260 nm.

Fig. 3 .
Fig. 3. Isotopic shift for the 210-230 nm region.The shift in the position of vibrational structure for the OC 33 S, OC 34 S, and O 13 CS isotopologues relative to OC 32 S are shown.Dotted line shows shifts using the ZPE model calculated in Danielache et al. (2009).

Fig. 6 .
Fig.6.Photolysis rates as a function of wavelength for OC 32 S at 20 km in arbitrary units.

Table 1 .
Wavelength at σ max , shifting of σ max , σ max ratio for each isotopologue, and the ratio of the areas under spectral curve for the measured spectra(195-260 nm).All comparisons are relative to OC 32 S.

Table 2 .
Reconstructions of absorption cross sections of OCS isotopologues using Gaussian fitting.
* Full width at half maximum.