Reactions of Oxygen Atoms in the Photolysis of CarbonDioxide *BY PETER WARNECKGeophysics Corporation of America, Bedford, Massachusetts, U.S.A.Received 10th January, 1964The vacuum ultra-violet photolysis of CO2 has been investigated in the pressure region 150-760 mm Hg. At 1236 A and around 1600 8, the quantum yields for carbon monoxide, oxygen andozone are consistent with the photodissociation mechanism, while an increase in the quantumyields at 1470A indicates that at this wavelength excited C02 molecules are also involved. Fromthe low rates of ozone formation observed in all cases it is concluded that the photodissociationof C02 produces oxygen atoms in the metastable 1D state. This is confirmed by measurements ofthe ratio of the rate constants associated with the reactions,OfO2-l- M +03+M k3O+ O3 +202 k4An average value of k4/k3[M] = 41 has been found.Although the photodissociation of oxygen and ozone in the Earth’s upperatmosphere is known to generate oxygen atoms in the metastable ID state inaddition to ground-state oxygen atoms, little attention could be given to the fateof the excited species owing to a lack of knowledge concerning the appropriatereaction rates.However, ID oxygen atoms can be expected to show a markedcontrast in reactivity on account of the difference in spin and/or internal energy.It is the principal aim of this paper to show that the photodissociation of carbondioxide in the 1200-1 700 A wavelength regionhvCO,~CO+Oproduces oxygen atoms in the excited 1D state, so that the photolysis of C02 mayprovide a convenient framework for a study of the role of 1D oxygen atoms in thereactions2 0 + M-02 + M0 4- 02( + M)+03( + M)(2)(3)0 + 03-+202 (4)which in combination with (1) constitute the main C02 photodecomposition mechan-ism.There is little doubt that oxygen atoms are indeed produced in the primaryprocess of carbon dioxide photodecomposition, since the COa absorption spectrumin the 1200-1700 A wavelength region exhibits two dissociation continua which areshown in fig. 1. Both continua have been associated 1 with a process of type (1).The participation of 1D oxygen atoms in the CO2 photodecomposition mechanismwas first suggested by Mahan 2 who photolyzed C02 in the presence of small amountsof NO but failed to observe the chemiluminescence characteristic for 3P oxygenatoms.In the present study of carbon dioxide photolysis, further evidence wasderived from the rate of ozone formation, which was measured in addition to CO* This work was supported in part by the National Aeronautics and Space Administration.558 co2 PHOTOLYSISand 0 2 quantum yields. Also, the ratio of the rate constants associated withreaction (3) and (4) was determined and compared with the known ratio for 3Poxygen atoms. These investigations were carried out at pressures between 150 and760 mm Hg, utilizing krypton, xenon and hydrogen light sources.1I I I I I I1300 1400 1500 1600 170 0wavelength (A)FIG. 1 .-Vacuum ultra-violet absorption spectrum and continua (dashed) of carbon dioxide,superimposed by the emission lines : Kr 1236 A, Xe 1470 A, and the many lined hydrogen spectrumaround 1600 A.EXPERIMENTALLIGHT SOURCE AND ACTINOMETRYExperiments were performed with a cylindrical flow reactor consisting of two concentricPyrex tubings, of which the outer one served as the gas inlet.The reactor was attachedto a microwave-powered discharge lamp of similar design as a previously described hydrogenlight source.3 Separate discharge tubes, made of Pyrex or quartz, were employed forkrypton, xenon and hydrogen in order to prevent contamination of the desired spectrumwith foreign lines. The reactor and light source were separated by a window of suitablematerial. Lithium fluoride windows were employed in conjunction with krypton dis-charges, while for the other gases barium fluoride plates were inserted. Since BaF2 doesnot appreciably transmit at wavelengths below 1400 A, it provides a suitable cutoff filtereliminating, among other lines, Lyman-alpha at 1216A and the xenon line at 1295A.Accordingly, the vacuum ultra-violet emission spectra associated with the three lightsources consisted mainly of the lines (a) Kr 1236 A, (b) Xe 1470 A, and (c) of the hydrogenmany-lined spectrum centred near 1600 A.The position of these emission lines relativeto the CO2 absorption continua is shown in fig. 1.Integrated light source intensities ranged from 2x 1015 to 2 x 1016 quanta/sec forkrypton and xenon discharges, and up to 1 x 1016 quanta/sec for the hydrogen lamp.They were determined actinometrically from the rate of ozone formation in a flow oP.WARNECK 59oxygen.3.4 A quantum yield of two 49 5 was assumed in accordance with the view thatat atmospheric pressure alrnost all the oxygen atoms resulting from 0 2 dissociation areconverted into ozone. Only the 11 65 A krypton line is probably ineffective in dissociating0 2 , since it falls into a region between bands where the 0 2 absorption strength is low(k ~ 0 . 5 cm-l).' However, since the CO quantum yields obtained in CO2 photolyseswith the krypton lamp were close to unity, it is concluded that the 1165 A line cannotcontribute more than 10 % of the intensity contained in the 1236A krypton line.QUANTUM YIELDSOzone concentrations were determined from the absorption of the 2537 A mercuryline which lies near the maximum of ozone absorption in this spectral region. A measuredflow of C02 was passed through the reactor, where ozone was formed as a photolyticproduct, and subsequently through a 40.5 cm long absorption tube fitted with quartzwindows at both ends.A Pen-Ray mercury lamp and a rubidium telluride (solar blind)photomultiplier in conjunction with a high voltage power supply and a microammeterserved to measure the extent of light absorption. Owing to the particular spectral charac-teristics of the mercury lamp and the photomultiplier combined, no additional filters wererequired for an adequate isolation of the 2537A mercury line. The ozone quantum yieldis calculated from Q(O3) = [O3]u/l, where [ 0 3 ] is the steady-state ozone concentration inthe absorption tube, II is the bulk flow rate in cm3/sec, and I in quantalsec is the effectiveintegrated intensity of the light source.The amounts of carbon monoxide and oxygen produced during the photodecomposi-tion of C02 were investigated in a closed system in which the gas was circulated by meansof a magnetically-driven rotating stirrer.Prior to the introduction of C 0 2 in the desiredquantity, the system was prefilled with approximately 3 mm Hg of argon which served asa reference in the subsequent mass spectrometric analysis. Each sample drawn aftercompletion of a photolysis run was subjected to liquid nitrogen temperature while beingadmitted to the mass spectrometer in order to minimize the contribution to the mass 28peak originating from carbon dioxide.With the exception of carbon dioxide, research-grade gases were employed withoutfurther purification.C 0 2 was found to contain about 0.05 % oxygen as well as 0.25 %nitrogen and/or carbon monoxide. These gases were removed by subjecting a lecturebottle of C02 to liquid nitrogen temperature and pumping off the volatile components.Repeated application of this procedure was required before the desired degree of purifica-tion was achieved (less than 20 p.p.m. contamination as determined mass spectrometrically).RESULTSco QUANTUM YIELDSThe data obtained by mass spectrometric analysis of samples drawn from theclosed system are summarized in fig. 2. CO quantum yields of unity were foundthroughout the investigated pressure region when C02 was irradiated with lightfrom krypton or hydrogen discharges, but the xenon light source resulted in an increaseof the CO quantum yield from unity at lower pressures to an average of Q(C0) = 1-17at 740 mm Hg.Similarly, a comparatively higher oxygen quantum yield was ob-served in the latter case. This indicates a difference in the photolytic mechanismwhich must be active at the wavelengths of the lines emitted from a xenon dischargeon one hand, and a krypton or hydrogen discharge on the other. The results ob-tained with the krypton lamp can be combined with the data reported by Mahan2to demonstrate that at 1236A the CO quantum yield is insensitive to pressure vari-ations in the region 10-760 mm Hg, thus giving additional support to the conclusionthat carbon monoxide is formed directly in the primary process.The observedoxygen quantum yields (Q(O2) = 0.3 for krypton and hydrogen, ( 3 0 2 ) = 0.4 forxenon) were higher than those reported by Mahan, but still below the limit o60 c02 PHOTOLYSISQ(O2) = 0.5 to be expected if all oxygen atoms produced in the primary processrecombined to yield molecular oxygen. However, the missing oxygen has beenshown to exist in the form of ozone.I I I I I 1 I In n nFIG. 2.-Quantum yields of carbon monoxide and oxygen ; 10-min irradiation with intensities ofabout 5x 1015 quanta/sec. Filled circles: at 1236A; open circles: at 1470A; squares: atabout 1600 A.I10 20time (min)intensity 5 x 101s quanta/sec.FIG.3.-Time dependence of carbon monoxide and oxygen formation at 1470 A ; irradiationThe time dependence of carbon monoxide and oxygen formation is plotted in fig.3 for experiments in which a xenon discharge and pressures close to an atmospherewere employed. The observed linearity substantiates the results presented in fig. 2P. WARNECK 61indicating that for the irradiation intervals and intensities applied the quantum yieldsare time independent. Further, since the straight lines obtained meet at the originof the co-ordinates, the absence of perceptible amounts of impurities in the em-ployed C02 is evidenced.OZONE QUANTUM YIELDSMost of the experiments designed to measure the amount of ozone formationwere carried out at atmospheric pressure.As one could expect on account ofthe stoichometry of the reactions involved, the observed ozone quantum yieldsdepended strongly on the concentration of oxygen contained in the carbon dioxide,and also upon the applied flow rates and irradiation intensities. No ozone couldbe detected when a flow of extensively purified CO:! was irradiated with the kryptonor the hydrogen discharge, even if high light intensities and low flow rates wereemployed. As a consequence, in these cases, only an upper limit quantum yield ofQ(03) <0.01 can be given. The ozone quantum yields observed with the xenon lightsource irradiating pure C02 were in the range of Q(O3) = 0.015, but this was still in-sufficient for an accurate determination of the intensity dependence.photon flux x 10-16 (quanta/cm* sec)I I0.5 1.0 1.5i I I~~0 05 1.0 1.5 Llight source intensity X 10-16 (quantalsec)3FIG. 4.-Ozone quantum yields for COz containing 0 057 % oxygen ; average flow rate 1 -9 cm3/sec.Open circles : at 1236 8, ; filled circles : around 1600 A.Larger concentrations of ozone were found when the employed carbon dioxidecontained a trace of oxygen (0.057 %) making it worthwhile to study the corres-ponding ozone quantum yields as a function of radiation intensity under otherwisesimilar conditions.Fig. 4 shows that the data obtained separately with kryptonand hydrogen discharges overlap quite closely, implying that the mechanism ofozone formation is the same in both cases. The increase in the ozone quantumyield with decreasing intensity is in accord with the kinetic mechanism. As theconcentration of oxygen atoms is reduced, their attachment to molecular oxygenby reaction (3) is favoured in comparison to the second-order direct recombinatio62 c02 PHOTOLYSIS(2).Similarly, reaction (4) is more effective at higher intensities owing to the largerabsolute amounts of ozone involved.No systematic study was performed with the xenon discharge under these con-ditions. However, as was the case with purified COz, the rate of ozone formed asa result of the 1470A irradiation was again greater than that observed at the otherwavelengths. For example, with a xenon light intensity of 7.5 x 1015 quantalsecan ozone quantum yield of Q(O3) = 0.225 was determined, while according tofig.4, the other discharges produce an average Q(O3) = 0.125.The pressure dependence of the ozone quantum yield was briefly explored em-ploying carbon dioxide containing 0.057 % oxygen and irradiating it with a con-stant intensity of light produced in the xenon discharge. According to the linearvariation of oxygen concentration with pressure the production of ozone is expectedto decrease as the pressure is reduced, provided the assumed photolytic mechanismis correct. Indeed, fig. 5 shows that the amount of ozone formation when expressedI '7 I/- I I I2 0 0 400 6 0 0 1pressure (mm Hg)FIG. 5.-Dr~"sure dependence of % ozone formation at 1470 A.in % of that observed at 760 mm Hg exhibits an almost linear pressure dependence.Since with decreasing pressure reaction (4) becomes less effective due to the smallerconcentrations of ozone involved, the pressure dependence of the ozone quantumyield seems to indicate that reactions (2) and (3) have similar third-body requirements.Although all of the above reported results concerning krypton and hydrogenradiation are qualitatively consistent with the invoked mechanism, the observedozone quantum yields are considerably smaller than could be expected if the involvedoxygen atoms were in the 3P ground state.With the use of currently accepted valuesfor the rate constants of ground-state oxygen atoms 6 one estimates, for example,ozone quantum yields in the range Q(O3) = 0-1 for a 1 cm3/sec flow of pure carbondioxide, whereas the observed quantum yields are smaller by at least an order ofmagnitude.However, this result would not be unreasonable if the oxygen atomsgenerated in the C02 photolysis were in the metastable 1D state rather than in thP. WARNECK 633P ground state, because in this case the formation of ozone by reaction (3) is spinforbidden so that the corresponding rate constant would be comparatively smaller.Although the formation of 1s oxygen atoms is energetically feasible at 1236& itcan be precluded since the energy available at the onset of the involved continuumis insufficient (see fig. 1). Thus, it appears that the photodissociation of carbondioxide produces 1D oxygen atoms. To provide a more quantitative test, the ratioof the rate constants of reactions (3) and (4) was measured and compared with theknown value associated with 3P oxygen atoms. A carbon dioxide + oxygen mixturecontaining 2.66 % oxygen was prepared and irradiated with light from the kryptondischarge.The ratio of the absorption cross-sections for carbon dioxide and oxygenat 1236A is quite favourable so that only one-tenth of the available radiation in-tensity is absorbed by the oxygen contained in the mixture. However, the highoxygen concentration promotes reactions (3) and (4) relative to reaction (2) renderingthe latter unimportant. Table 1 demonstrates that the resulting ozone quantumyields are in the vicinity of unity, but a considerable variation with the appliedintensities and flow rates is also noticeable.run no.TABLE 1 .-03 QUANTUM YIELDS AND VALUES FOR y = k4/k3[M] IN THE1236A PHOTOLYSIS OF co2 CONTAINING 2.66 % 0 279/8081/8283/8485/86(A)8 5/86(B)87/8 8(A)87/88(B)89/9091/9293/9495/9697/9899/100101/102V I X 10-16cm3lsec quanta Q<o3)StX1.842.042-064.64.64.54.54.74.73.23-153.1 53.13.21.1350.7260.7 11 ~ 6 81.761 -591 -451-140.5920-3730.3020.2041 a071-540.7220.8 10.7820.8 10-7740.7630-8380.8650.9350.93 80.9881.0320.7850.75524.424.230.523.628.533.221.823.019.019.542.329.525.5-4 corrected431.034.941.534139.344.533-238.145.348.237.831.046.234.034.839.146.538.243.850.141.842.250.853.641.235.051.838.2a corrected assuming OZ+h -+20(1~).b corrected assuming O z + h +0(1D)+o(3P).average y1 = 383i-4.8, y2 = 43.435.3.DISCUSSIONThe low ozone quantum yields could not have been obtained if 1D oxygen atomswere rapidly deactivated by collisions with carbon dioxide.The conclusion thatcollisional deactivation is relatively unimportant is corroborated by a study ofKatakis and Taube 7 who found that 1D oxygen atoms produced in the 2537 Aphotolysis of ozone undergo an exchange reaction with C02 rather than deactivation.Fig. 1 shows that at the wavelengths of the xenon and hydrogen radiation theabsorption is predominantly due to the first continuum appearing at longer wave-lengths, while there is negligible contribution of the second continuum towardsshorter wavelengths. Conversely, the line at 1236A is almost solely absorbed bythe second continuum with negligible contribution from the first.Since the resultsobtained separately ,with the krypton and the hydrogen light source are essentiall64 c02 PHOTOLYSISidentical, it can be inferred that the transitions represented by the two absorptioncontinua lead to identical dissociation products, namely, carbon monoxide and1D oxygen atoms. The effect of the xenon 14701$ line can now be considered.In view of the preceding conclusion it is apparent that the increase in the quantumyields observed at this wavelength must be caused by the partial absorption in theoverlying band system, which at this wavelength accounts for approximately 30 %of the total absorption strength. Accordingly, the results are explained in terms ofreactions involving excited C02 moleculesh v co2+co; (5)(6)which must occur in addition to the photodissociation mechanism (1) through (4).These reactions can account for the increase of the quantum yields observed forall the products.At pressures below 150 mm Hg, where the CO quantum yield isunity, the COJ., molecule presumably predissociates or re-emits the photon whichsubsequently is re-absorbed, preferentially in the continuum. The only finding notentirely consistent with this supposition is the lack of pressure dependence con-cerning the 0 2 quantum yields obtained with the xenon discharge.For a quantitative interpretation of the measurements the physical conditionsof the experiments must be described by a suitable theoretical model.Since thereactions induced by COZ photolysis at atmospheric pressures are confined to anarrow region in the vicinity of the window, which is characterized by a certaindegree of turbulence owing to the reversal flow in this region, it appears appropriateto perform calculations on the basis of stirred reactor theory. In the photolysisof C02 containing 2.66 % oxygen the 0 2 concentration is so large that it remainsapproximately constant even if all the oxygen atoms are consumed in the formationof ozone. Also, reaction (2) can be considered negligible. The stirred reactorequations then read :co: + CO2-+2C0 + 0 2 ,(V/R>[OJ = [OI(k3[02ICMI -k4[03I>,(VIR)COI = (w- COI(~,CO,I[MI + k4c031),where the concentrations refer to the outgoing (steady-state) concentrations, v isthe flow rate, R the reaction volume, and Z the integrated irradiation intensity.Ifthe oxygen atoms are largely consumed within the boundary of the reaction volume,(u/R)[O]<I/R, and the equations can be combined to express the ratio of the rateconstants in terms of the previously defined 0 2 quantum yield :Note that the reaction volume R has cancelled.When this equation is employed in conjunction with the data shown in table 1,one obtains the y values entered in column 5. These are about five times greaterthan the highest value available for 3P oxygen atoms, y(3P) = 5, which can be de-rived using the rate constants k4(3P) = 2.5 x 1014 cm3/molecules sec8 and k3(3P) =2 x 10-34 cm6/molecules2 sec6.However, eqn. (A) does not yet take into accountthe simultaneous photodissociation of oxygen in the mixture so that a correctionis required making the discrepancy even greater. Because of the existing un-certainty regarding the dissociation products of oxygen at 1236& two cases wereconsidered: one in which it was assumed that the dissociation products are two1D oxygen atoms, and a second case involving a 1D and a 3P oxygen atom of whicP. WARNECK 65the latter was supposed to undergo immediate ozone formation. The corres-pondingly corrected y values are shown, respectively, in columns 6 and 7 of table I .The averaged values are seen to be approximately the same in both cases, suggestingy x41. Despite a considerable variation in the intensities and flow rates the averagedeviation from the mean is about 12 %, supporting the employed reaction mechan-ism. The high values obtained for y are considered sufficient evidence that theCO2 photodissociation products are carbon monoxide and 1D oxygen atoms.1 Watanabe, Zelikoff and Inn, Absorption Coefficients of Several Atmospheric GaJes, AFCRC2 Mahan, J. Chem. Physics, 1960, 33, 959.3 Warneck, Appl. Optics, 1962, 1, 721.4 Groth, 2. physik. Chem. B, 1937,37, 307.5 Vaughan and Noyes, J. Amer. Chem. SOC., 1930,52, 559.6 Kaufman, Progress of Reaction Kinetics, vol. 1 , ed. Porter (Pergamon Press, 1961), p. 19.7 Katakis and Taube, J. Chem. PhyAics, 1962, 36, 416.8 Phillips and Schiff, J. Chem. Physics, 1962, 36, 1509.Tech. Rpt. no. 52-23, Geophys Res., paper no. 21, 1953