Relative quantum yields for O1D production in the photolysis of ozone between 301 and 336 nm: evidence for the participation of a spin-forbidden channel

Abstract Relative quantum yields for the formation of O1D from the photolysis of ozone have been measured between 301 and 336 nm. O1D was monitored indirectly, using laser induced fluorescence detection of vibrationally excited OH. The OH was produced by the reaction of O1D with H2. The observation of ‘blue-shifted’ laser induced fluorescence provided high detection sensitivity and eliminated probe laser interference effects. The results confirm that there is a significant quantum yield for O1D production at wavelengths longer than 320 nm. A measurable yield is observed out to 336 nm providing evidence for spin-forbidden production of O1D.


Introduction
While photolysis in the peak of the Hartley (200-300 nm) bands is important in the stratospheric chemistry of ozone, absorption in the weak tail (300-350 nm) dominates the photochemical activity of ozone in the troposphere. Throughout most of the Hartley bands, ozone photolysis occurs through two spin-allowed processes. The major channel reaction 1 a is thought to have a quantum yield of 0.9-0.95, producing O]D [1][2][3]. 03 + hv-"~ OlD + O2(IA), (la) 0 3 + hu ~ O3p + O2 (3]~-). (lb) The region, > 300 nm, in which the Hartley continuum passes into the more structured Huggins bands, is responsible for the production of tropospheric O I D, and hence the production of OH via the process OlD + Hat ~ 2OH.
The O I D production rate is a product of the actinic flux, ozone absorption cross-section and the quantum yield for O ID production. The actinic flux is rising rapidly and both the ozone absorption cross-section and O]D quantum yield, q)o~o, are falling rapidly in this region. Ozone absorption cross-sections, O-o~, are thought to be known with an uncertainty of 1% [1][2][3], however the uncertainty in the O ID quantum yield is much greater, particularly in the long wavelength tail of the spectrum. The NASA evaluation [2] recommends an uncertainty of 1.3 in the region 305 nm<A<320 nm and suggests that q~O~D=0 at 0009-2614/97/$17.00 Copyright © 1997 Elsevier Science B.V. All rights reserved. PII S0009-261 4(96)01342-5 wavelengths longer than 320 nm. This recommendation is based on a series of experiments performed by Moortgat and coworkers [4][5][6]. Subsequently there have been several studies [1] using laser photolysis which have measured a significant ~o'a at )l.photolysi s > 320 nm. Nevertheless the two widely used evaluations of atmospheric rate constants, NASA [2] and CODATA [3] consider the long wavelength 'tail' in q~o,o to be an artifact, while also acknowledging the need for further experiments. More recent field and laboratory measurements have raised new questions about the NASA/IUPAC recommendation [7][8][9]. In addition a recent photofragment-LIF study of jetcooled ozone in the 317-327 nm region provides strong evidence for the spin-forbidden formation of O ~O [10].
One of the major difficulties in determination of the ~o'I) is the difficulty in monitoring O ID, coupled with the dynamic range required, i.e. the product {O-o3 × ~o,i)} varies by a factor of > 103 between 300 and 330 rim. Laser based techniques offer the only reasonable approach to this determination but it is critical to ensure that non-linear phenomena or other artifacts associated with high photon densities do not perturb the measurements. In this work we present results based on the detection of vibrationally excited OH as a spectroscopic marker for O J D. We find a long wavelength tail in the O I D quantum yield which is suggestive of the participation of spin-forbidden production of O 1D.

Experimental
Our experimental studies used detection of X 2H OH (v = 1) as a 'spectroscopic marker' for O I D.
The experiments utilized pulsed, tunable laser photolysis of an O3/O2/H 2 gas mixture between 301 and 336 nm. OH (v---1) was detected by a second pulsed tunable laser, using laser induced fluorescence (LIF) on the OH A-X (0-1) Qll transition at 347 nm. Fluorescence was detected in the (0-0) band at 308 nm. All experiments were performed under slow flow conditions. Ozone was generated by passing 02 through a commercial ozonizer. The 02/0 3 mixture was combined with a flow of H 2 and passed through a jacketed Pyrex cell. The total pressure was 55 Torr, the ozone concentration was determined photometrically and was 2 × 10 ~5 molecules cm -3. The partial pressure of H 2 was 10 Torr and the balance of the mixture was 0 2. The photolysis laser was a frequency doubled, Nd-YAG pumped dye laser with an estimated UV bandwidth of 0.5 cm-i. By using several dyes or dye mixtures it was possible to continuously tune the UV output between 300-336 nm. The probe laser output was obtained by mixing the fundamental of an injection seeded Nd-YAG laser with the output of a dye laser, the estimated bandwidth of the probe laser was 0.06 cm -~ The energies of the photolysis and probe pulses were measured using photodiodes. The response of the diode which monitored the photolysis laser was corrected for the change in sensitivity as a function of wavelength by comparison with a thermoelectric power meter. In a typical experiment the photolysis laser fired and after a fixed delay of two microseconds the probe laser fired. The LIF and photodiode signals were monitored using a 500 MHz digital oscilloscope, typically averaging 100 shots. The 'noise', i.e. the signal due to scatter from the photolysis and probe beams, was also measured but was negligible until close to the longest wavelength, 336 nm, at which an OH LIF signal was observable. The photolysis laser wavelength was then incremented by 1 nm and the experiment repeated. The LIF signal was found to be linear with photolysis laser power over the range of laser powers available with our system. The LIF signal was linear with probe power at lower powers and showed characteristics of partial saturation at higher powers. The normalized LIF signal, i.e. the relative yield of OH (v = 1), is shown as a function of wavelength in Fig.  1. This is the product of the 03 absorption coefficient, tro3, and q~o~o. Relative values of ~O'D were obtained by dividing this product by the absorption cross-sections of Molina and Molina [11]. They were then placed on an absolute basis by normalizing to a value of q0o, o = 0.95 at 302 nm 2. The total pressure, partial pressures of H E and 0 3 and the probe laser power were essentially constant over the course of an experiment. As shown in Fig. 1, the photolysis energy varied by a significant factor. The only other experimental variable which was adjusted was the pmt detection sensitivity which was altered by adjusting the pmt high voltage. In these cases common data points were taken at each voltage and the data normalized appropriately.
.,-4 .L4.oo m 0 t2.00 ~ 0 o.oo ~ 34o Photolysis Wavelength (nm) Fig. 1. The relative yield of OH (v = 1), shown as a function of wavelength between 301 and 336 nm. This yield has been normalized for variations in the photolysis laser power. The relative photolysis laser power is also shown. There is a considerable degree of scatter in the values around 301 nm and 310 nm, including two points with values (including ltr errors) of qbo, D which exceed 1. This is suggestive of some systematic error in the normalization procedure at the very low photolysis laser powers that were available around these wavelengths. The 'tail' in ~O'D 1S clearly evident and the threshold for O ~D production appears to be to the red of 336 nm.

Results
The elimination of experimental artifacts associated with the use of the high photon fluxes is a significant concern in this type of experiment. Some of the potential problems associated with this type of experiment are as follows:

Is OH (v = 1) a true spectroscopic marker for 0 ZD?
The reaction of O I D with H 2 to produce vibrationally excited OH is well characterized [12] and is extremely fast with a rate coefficient [2] of 1.0 × 10-1° cm 3 molecule-I s-I. The product OH is formed with a high degree of vibrational and rotational excitation with greater than 50% being formed in v = 1 or higher levels [12]. After photolysis we expect rapid rotational thermalization together with relatively fast vibrational deactivation via collisions with 0 2. Rensberger et al. [ 13] reported a rate coefficient for 0 2 vibrational deactivation of OH (v = 2) of 2.6 x 10-13 cm 3 moleculel s-i with the deactivation rate with H 2 being immeasurably slow. The formation of OH, population of v = 1 via collisions from v> 1 and the deactivation of v= 1 are pseudo-first order in H 2 or 0 2. Hence, for a fixed delay between the pump and probe lasers, the relative OH (v = 1) LIF signal should accurately reflect relative O ~D formation.
If, as we suggest below, the mechanism of photodissociation involves both spin-allowed and spinforbidden channels the translational energy disposed into the O t D fragments in likely to be very different in the two cases. If O ~ D is produced with significant translational energy and this affects the energy disposal in the OH product of its reaction with H 2 then a significant artifact could be introduced into this work.
The only other potential sources of OH (v = 1) would require a hot atom reaction of O 3p with H 2. We consider 'hot atom' artifacts to be unlikely, "l+ l Cr " particularly at the pressures used here, but we plan to examine the pressure dependence of (/)OLD in a bath gas which has a low efficiency for O Io deactivation, i.e. He or Ar, but which will efficiently thermalize atoms with excess translational energy.

Multiphoton photolysis of 03
Even if the observed OH LIF signal is indeed due to O I D, multiphoton dissociation of 0 3 could produce an artifact. Such a process would produce a non-linear dependence of the LIF signal with photolysis power, experimentally, however, a linear dependence was observed.

Probe laser photochemical effects
If the probe laser photolyzed electronically excited 0 2 to produce O I D, this could react to produce OH (v = 1) within the 6 nsec duration of the probe pulse. Such a photolytically generated signal would initially show a quadratic dependence on the probe laser power. The experimentally observed probe power dependence of the LIF signal was linear at lower powers, showing the 'roll-off' characteristic of partial saturation at higher powers.
An example of the signal-to-noise which characterizes these experiments is shown in Fig. 3. In this case the signal averager was used as a digital boxcar integrator set to a ten shot exponential average of the LIF signal which was recorded in 'strip chart' form. The points were sampled at 5 Hz, hence each point corresponds to an instantaneous average of two laser shots, smoothed by the ten shot exponential average. The full scale of 1200 x 2 shots corresponds to a 4 min measurement period. It can be seen that at 329 nm the observed LIF signal greatly exceeds the background noise and disappears when the photolysis laser is blocked, the H 2 flow is tumed off or the ozonizer is switched off. The total signal is a factor of six larger than the background and the ratio of the signal to the l tr standard deviation in the background is 35. At 336 nm a small OH signal is observable above the background. We estimate that our signal-to-noise ratio can be improved significantly by increasing [O a ], switching to He buffer gas which does not quench the A state of OH and increasing averaging times. This experimental ap-proach is well suited to determining the form of the long wavelength tail for O I D production as well as the temperature dependence and threshold.

Discussion
Our data are shown in Fig. 4, together with the recent results of Amerding et al. [8] Also shown are the currently recommended NASA/IUPAC evaluation, the recommendation of Steinfeld et al. [1] and the model calculations of Michelsen et al. [14]. The Michelsen et al. model assumes that the long wavelength tail is solely due to dissociation of vibrationally excited molecules. Early experiments on the yield of O ID were reviewed by Steinfeld et al. [1]. Those workers recommended a 'tail' in @O'D based on the laser photolysis experiments of Brock and Watson [15] and Trolier and Weisenfield [16]. However, both the NASA and IUPAC panels have continued to regard this tail as an artifact. In addition to our work and that of Amerding et al.  late their results but their reported yields are in very close agreement with the yields of Trolier and Weisenfeld [16]. Recent measurements of Ball and coworkers [17][18][19] show that the yield of O2(~A) is significant out to 325 nm. In recent, unpublished data [20] they report O I D yields which broadly agree with the O2(IA) yields. In this work a (2 + 1) REMPI technique is used to directly monitor O 1D. In recent field experiments, Muller et al. [7] simultaneously measured spectrally resolved actinic fluxes between 280 and 330 nm and J(O 1D) using a chemical actinometer. They compared calculated and measured O 1D formation rates using both the NASA recommended ~O,D, and one that included a long wavelength tail. Their results are consistent with a significant long wavelength tail.
It is clear that there is a significant body of laboratory and field evidence which supports the existence of a long wavelength tail in ¢PO'D although its precise form and magnitude is not well defined and there is little data on its temperature dependence. The thermodynamic threshold for the spin-allowed channel la is 310 nm, and two mechanisms provide likely explanations for a significant ¢PO'D at longer wavelengths. Each quantum of excitation in the (001) or (100) vibrational modes decreases the photolysis threshold by = 1000 cm -l. This shifts the photoly-sis threshold by ~-10 nm to the red. Photolysis of vibrationally excited 0 3 could be responsible if the absorption spectrum of these levels is significantly red shifted.

o3 + O'D + 02(3: -). (Ic)
The participation of the spin-forbidden channel I c, which has a thermodynamic threshold of 411 nm, would also be consistent with long wavelength O J D production.
The measurements of Ball and coworkers [17][18][19] on the wavelength and temperature dependence of the yield of O2(IA) seem consistent with an O2(IA) yield which has a temperature dependent spin-allowed component, together with a temperature independent spin-forbidden component, channel ld. 03 +hv--* O3p + O2('A ). (ld) The temperature independent channel provides strong evidence for the participation of the spin-forbidden channel and time of flight measurements on the kinetic energy of O2(IA) produced in long wavelength (> 320 nm) photolysis are consistent with this. However, if O2(1A) is produced from both spin-allowed and forbidden channels then it is not possible to infer O ~D yields from these measurements. Data on (PO'D beyond 325 nm is limited to the recent data of Amerding et al. [8] and the results reported here, both of which determined relative yields using OH as a spectroscopic marker for O J D. The agreement between these results is reasonable up to 328 nm, however we find a measurable OH yield out to 336 nm. Amerding et al. [8] use an extrapolation to place the threshold for O ~D formation between 331 and 333 nm. They claim that this makes the participation of a spin-forbidden channel unlikely. They cite an unpublished communication from Banichevich [22] reporting calculations which suggest that reaction lc is the least likely dissociation channel for electronically excited 03 . The most significant differences between the experimental approaches are our use of H 2, rather than H20, as the co-reactant and the use of the A-X (0-1) transition to pump vibrationally excited OH, followed by detection of blue shifted fluorescence. Frost and Vaida [21] have recently suggested that H20-O 3 complexes may photolyze at long wavelengths to produce OH directly. If their calculations are correct, such a complex would have been present at significant concentrations in the experiments of Amerding et al. [8] and would have contributed to the observed OH (v = 0) LIF signal. In order to avoid such complications we used H E. If photolysis of such a complex was significant, the I~)OI D measured by Amerding et al. [8] would have been larger than ours and should have been measurable at longer wavelengths. The agreement between the results suggests that photolysis of H20-O 3 complexes is not a significant source of OH. Our detection scheme has two major advantages over excitation of the A-X (0-0) transition at 308 rim. Observation of blue shifted fluorescence allows for very effective discrimination against scattered probe laser light. In addition, since the O s absorption cross section and q~O'D are small at 347 nm, the probe beam produces a negligible OH LIF background signal. This is in strong contrast to the use of 308 nm excitation which produces a strong background OH LIF signal with the same laser pulse producing O~D and detecting the OH reaction product. This noise signal is a two-photon process and scales quadratically with laser power.
During the revision of this paper, a preliminary report on the photolysis of jet-cooled ozone appeared which provides strong evidence for the formation of O ID via the spin-forbidden process lc. Takahashi et al. [10] photolyzed jet-cooled 03 between 317-327 nm and obtained the O I D photofragment excitation spectrum using laser induced fluorescence to monitor O ID. At 320.05 and 323.80 nm, wavelengths which correspond to peaks in the photofragment excitation spectrum, they obtained broad O I D Doppler profiles of 1.5 cm -1 . This necessarily implies spin-forbidden O ID production if the excitation is from the vibrationless level of 03 . It also implies that an insignificant fraction of the available energy is partitioned into vibrational excitation of O 2 (3E-). The Doppler profiles of O I D produced by photolysis at two wavelengths in the underlying continuum are narrower and the authors identify these as being formed from vibrationally excited O 3. However narrow profiles do not necessarily imply spin-allowed production of O ID. Such profiles would be consistent with spin-forbidden production in which a significant fraction of the available energy is partitioned into excitation of 0 2 (3•-). The structure which the authors identify with the spin-forbidden process is apparent in the photofragment spectrum of thermalized 03 at 227 K [10] but is not apparent at 298 K in the 317-325 nm region [9].
The model of Michelsen et al. [14] predicts spinallowed production of O ~D out to 325 nm, although the predicted quantum yield at 325 nm is 0.01, lower than any of the experimental studies. At 324 nm approximately 0.2% of the 0 3 population has sufficient internal energy to produce O 1D via a spin-allowed process. At 334 nm this fraction has dropped by greater than two orders of magnitude to 0.0014%. Our observations imply that ~OtD is essentially constant in this region. For our observations to be consistent with production of O ~D via channel la the cross-section of vibrationally excited ozone molecules would have to increase dramatically with wavelength between 324 and 334 rim. At 334 nm it would require a cross section of --2 × 10 -17 cm 2. It seems more likely, in our view, that the spin-forbidden channel lc is operative. Temperature dependent studies and the measurement of the quantum yield at longer wavelengths should shed more light on this question. As we have noted above our experimental approach should be able to address both questions.
With some experimental modifications we should be able to increase our signal-to-noise ratio significantly and extend the measurements to temperatures as low as 215 K.
These results have significant implications for both tropospheric and stratospheric models. We have used a time-dependent photochemical model [23,24] to examine the effects of the increase in ~o'D on tropospheric OH levels. The full details of these calculations will be published elsewhere [25], however we find that at 298 K and 0 km the use of our ~O'O increases the average OH concentration by 10-20% over that computed with currently recommended NASA values. The exact value depends on latitude and is greater at higher latitudes. At 298 K the inclusion of a spin-forbidden channel with ~O'O = 0.05 for 325 nm< A < 350 nm has essentially no effect on this result. At lower temperatures the effects of a spin-forbidden tail could be more significant. A preliminary assessment suggests that at a temperature of 220 K and a height of 15 km an addition of a temperature independent 'tail' with • O,D = 0.05 for 320 nm < A < 350 nm would increase the calculated JO~D by 10% over the model of Michelsen et al. [14] at high zenith angles. By contrast both models predict rates which are a factor of 3 larger than those calculated using the current NASA values.