J. CHEM. SOC. FARADAY TRANS., 1994, 90(13), 1819-1829 Mechanism of Atmospheric Oxidation of 1,I ,I ,2=Tetrafluoroethane (HFC 134a) Oliver V. Rattigan,* David M. Rowley, Oliver Wild and Roderic L. Jones Centre for Atmospheric Science, Department of Chemistry, University of Cambridge, Lensfieid Road, Cambridge,UK CB2 IEW R. Anthony Cox NERC, Polaris House, North Star Avenue, Swindon, UK SN2 IEU The chlorine-initiated photooxidation of hydrofluorocarbon 134a (CF,CH,F) has been studied in the temperature range 235-318 K and at 1 atm total pressure using UV absorption. Trifluoroacetyl fluoride [CF,C(O)F] and formyl fluoride [HC(O)F] were observed as the major products. IR analysis of the reaction mixture also showed car- bony1 fluoride [C(O)F,] as a product. By measurement of the yields of HC(0)F from the photooxidation as a function of [O,] and temperature, the rate of the unimolecular decomposition of the oxy radical, CF,CHFO, reaction (5),was determined relative to its reaction with O,, reaction (4): CF,CHFO + 0, +CF,C(O)F + HO, (4) CF3CHF0+CF3 + HC(0)F (5) The results were treated using both an arithmetic derivation and numerical integration with a detailed reaction scheme, Inclusion of other recently published kinetic data leads to the following recommended rate expression for reaction (5)at 1 atm k, = 7.4 x 10" exp[( -4720 f220)/77 s-' The errors are la.The observation of enhanced product yields in the present work is attributed to the reaction of the CF,O radical with HFC 134a leading to further peroxy radical formation.The results have been incorporated into a 20 atmospheric model to assess the environmental implications of HFC 134a release in the troposphere. It is now widely accepted that chlorofluorocarbons (CFCs) are directly responsible for the increasing levels of strato- spheric chlorine observed over the past decade.' This has in turn led to large losses of stratospheric ozone, particularly over the polar regions (see, e.g. ref. 2). Hydro-chlorofluorocarbons (HCFCs) and hydrofluorocarbons (HFCs) have been proposed as replacements for CFCs. These compounds contain one or more carbon-hydrogen bonds and are therefore susceptible to attack by the hydroxy radical in the troposphere leading to shorter atmospheric lifetime^.^ These shorter lifetimes, coupled with reduced chlorine substi- tution, lead to a lower release of chlorine in the stratosphere and hence a lower ozone-depletion potential.In order to assess the environmental impact of HCFC and HFC release it is necessary to quantify the nature and yield of the products from atmospheric oxidation. Although the reaction of the hydroxy radical with these compounds has been the subject of several the detailed mechanis- tic and kinetic pathways are somewhat uncertain. HFC 134a (CF,CH,F) is a proposed replacement for dichlorodifluoromethane (CFC-12) which is widely used as a refrigerant and in air-conditioning systems. In the tropo- sphere OH-initiated oxidation of HFC 134a leads to the for- mation of the peroxy radical CF,CHFO, via the following reactions CF,CH,F + OH -,CF,CHF + H,O (1) CF3CHF+ 0, + M +CF,CHFO, + M (2) The reaction of CF3CHF0, with NO [reaction (3)] has been found to be and is expected to be a major loss reac- tion for the peroxy radical for most situations in the tropo- sphere CF,CHFO, + NO -+ CF,CHFO + NO, (3) Subsequently, CF,CHFO radicals formed in reaction (3) may either react with 0, [reaction (4)] or undergo unimolecular decomposition [reaction (5)] CF,CHFO + 0, -+ CF,C(O)F + HO, (4) CF,CHFO +CF, + HC(0)F (5) The primary product distribution from the oxidation of HFC 134a therefore depends on the relative rates of reactions (4) and (5).We have recently measured the UV spectra and absorption cross-sections, 0, for CF,C(O)F and HC(0)F9-" and in the present work we have used this information to determine the yields of these products in the chlorine photosensitised oxida- tion of HFC 134a by UV spectroscopy.Measurement of the yields as a function of [O,] and temperature gave informa- tion on the reaction mechanism and allowed a determination of the rate constant ratio k,/k, in the temperature range 235- 318 K. During the course of this work a number of other studies of the photooxidation of HFC 134a have been carried out and the combination of these results has greatly aided in the understanding of the complex reaction mechanism and determination of the kinetic parameters. Wallington et a1.l' studied the chlorine-initiated photooxi- dation of HFC 134a over the temperature range 261-353 K and at pressures of 15-5650 Torr using product analysis with FTIR.An expression for k,/k, of 1.58 x exp(3600/T) cm3 molecule-' s-' was reported at 2 atm and a strong pres- sure dependence below 1 atm was found due to fall-off in the 1820 rate of the unimolecular reaction (5).Tuazon and Atkinson', have also studied the photooxidation using FTIR and their branching ratio of k,/k, = 3.2 x exp(3510/T) over the range 273-320 K and 1 atm pressure is in excellent agree- ment. Edney and Driscoll'3 have reported a similar study at room temperature with results that are broadly consistent. Maricq and Szente', have recently reported a flash pho- tolysis UV absorption kinetics study of HFC 134a photooxi- dation.Diode array spectroscopy was used to obtain time-resolved spectroscopic information of the radicals CF,CHFO and CF3CHF0,. They determined the rate of the unimolecular decomposition of the CF,CHFO radical, k, = 3.7 x lo7 exp(-2200/T) s-' over the temperature range 210-372 K and at 250 Torr, from the time dependence of growth and decay of an absorption assigned to CF,CHFO. Zellner et ~1.'~studied the pulsed laser photolysis of HFC 1 34a-C1,-NO-02 mixtures with time-resolved measurement of [OH] and [NO,] using optical methods. By simultaneous fitting to the NO, and OH profiles, they were able to deter- mine values for k,, k, and k, of (1.7 f0.6) x lo-', cm3 molecule-' s-', (2.7 f0.6) x lo-', cm3 molecule-' s-' and (1.8 f0.4) x lo4 s-', respectively, at 295 K and 38 Torr total pressure.An absolute uncertainty of a factor of two was reported for the individual values of k, and k, , but the ratio k,/k, was well determined. There is considerable uncertainty about the fate of the CF, radical produced in reaction (5) and its effect on the kinetics and products in the system. Under the conditions employed in this work, CF, is likely to form CF302 radicals by the addition of 0,. The self- and cross-reactions of this peroxy radical lead to CF30.16317Good evidence has also been reported" for the formation of trioxide species such as CF,O,CF,, formed in the reaction of CF30 with CF,O,, although these substances are unlikely to be formed in the atmosphere in view of the low radical concentrations.Sehes- ted and Wallington found evidence that CF,O reacts with CF,CHFO, to form a trioxide which is less stable than CF,O,CF, ." Recent studies have also shown that the rate constant for the reaction of the CF,O radical in H-atom abstraction from hydrocarbons is rapid, with room-temperature rate coeffi- cients similar to those for OH Sehested and Wallington found that CF,O reacts with HFC 134a to form CF,OI-I which decomposes to give COF, and HF." It seems likely that CF,OH formation is the fate of CF,O in the tro- posphere. In the present work the relative importance of the thermal decomposition of CF,CHFO compared to its reaction with 0, has been computed using a 2D atmospheric model2, The fate of the product CF,C(O)F in the atmosphere is also dis- cussed.Experimental The dual-beam diode array spectroscopy system used in this study has been described in detail previo~sly.~~ A 1 m long jacketed quartz cell connected to a standard greaseless vacuum system was used as the reaction vessel. The cell was thermostatted at 235-318 K using flowing ethanol, and evac- uated dual-window assemblies prevented frosting of the optical faces at low temperatures. Four interchangeable pho- tolysis lamps were mounted adjacent and parallel to the cell. In the present study Philips TL/09 lamps of spectral output in the range 300-400 nm were used. The source of UV for absorption measurements was a deu- terium lamp (30W, Hamamatsu L 1636) collimated output from which was passed through a beam splitter (Oriel Scien- tific, model 78150) producing two beams (reference and J. CHEM.SOC. FARADAY TRANS., 1994, VOL. 90 sample) which were collected in optical fibre couplers (Oriel Scientific model 77800) either directly (reference) or after passage longitudinally through the cell (sample). These two beams were then resolved and imaged by a 275 mm Czerny- Turner spectrograph, at a spectral resolution of 1.2 nm, separately onto two 512 channel unintensified silicon diode arrays (Reticon). The spectrograph and detector were con- trolled by a microcomputer (Dell 316SX) and software pack- ages (Spectroscopy Instruments Ltd.) were used for linearisation of the wavelength scale, for background subtrac- tion and averaging of the data and for calculation of the absorbance using the reference spectrum to correct for changes in the source lamp intensity.Despite this correction, however, the main limitation in absorbance measurements was baseline drift, attributed to inhomogeneities in the source output affecting the reference and sample beams differently. This limited the precision of measurements to f0.0005 absorbance units. CF,CH,F (99.8%) was obtained from ICI Chemicals and Polymers Ltd. and CF,C(O)F (97%) was obtained from Flu- orochem. Both samples were purified by trap-to-trap distilla- tion prior to use. Research grade samples of C1, (5% in N,), 0, and N, obtained from BOC were used without further purification. Formyl fluoride was prepared by the reaction of formic acid with benzoyl chloride using dry KHF, at 333 K.,, The formyl fluoride was first purified by passing through a trap at 255 K to remove benzoyl chloride vapours and then trapped at 78 K.Several successive distillations of the formyl fluoride were carried out. Results Reference UV absorption spectra of pure samples of CF,C(O)F and HC(0)F were recorded over a range of condi-tions, prior to the photooxidation study.'*'' Example spectra are shown in Fig. 1. The absorption cross-section for CF,C(O)F was 13.8 x lop2' cm2 molecule-' at the maximum, 214 nm, and was found to have a small tem- perature dependence.' For HC(0)F the cross-section at 230 nm was 6.85 x lo-,' cm2 molecule-', using a resolution of 1.2 nm (FWHM), in good agreement with the results of Gid- dings and Innes.26 The absorption cross-section was found to be independent of temperature in the range 233-318 K.IR spectra of the two main products were also recorded, in order to confirm their purity. For the photooxidation study, mixtures of CF,CH,F (8-12 Torr), Cl,, (1 Torr, 5% in N,) and 0, (20-730 Torr) were made up to a total pressure of 760 Torr using N, (17-731 Torr) and pre-mixed in the dark for several hours. Irradiation in the wavelength range 300-400 nm was then used to drive the C1-atom-initiated photooxidation of the hydro-fluorocarbon and at regular intervals prior to and during this photolysis, UV spectra of the reaction mixture were recorded.Fig. 2 shows three sequential spectra taken following 11 1, 171 and 243 s of irradiation. As can be seen from these spectra, there is a build-up of absorption around 230 nm, attributable to product formation and a reduction of absorption around 300 nm, attributable to chlorine consumption, with increas- ing irradiation time. Spectral stripping routines were applied to the sequential spectra for the identification and quantification of products formed during the photooxidation. Reference spectra re-ported above were used, allowing for the temperature depen- dence of the CF,C(O)F spectrum. However, owing to the very weak UV absorption of HFC 134a in the wavelength range used in this study, consumption of the hydro-fluorocarbon could not be monitored. Yields of the two main J.CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 10.0 I I A 'I 4.0 I Gu-../0.01 . , , . , . . ,--, II 1 ,I1 I I. I I I I I I I I I 225.0 232.5 240.0 247.5 255.0 262.5 wavelength/nm 200 220 240 260 280 wavelength/nm Fig. 1 Reference UV absorption spectra of, A, HC(0)F and, B, CF,C(O)F at (a)238 and (b)293 K products were therefore expressed in terms of the molecular chlorine consumed, given that the experimental conditions were arranged such that chlorine atoms were converted stoi- chiometrically into CF,CHFO, radicals. Chlorine cross-sections were taken from the current NASA evaluation.' A graph of product formation against chlorine consumption gave the 'chlorine based' yield as the gradient, and the small non-zero intercept, where present, was attributed to differen-tial baseline shifting of the sample and reference beams and neglected.Errors on individual yields were taken from these graphs and combined with a +_5% uncertainty in the cross- sections. Typically, this resulted in a total error in the yield of 0.12 1 1 1 A 0.0 ' . ' wavelength/nm Fig. 2 Sequential UV absorption spectra of a photolysed HFC 134a-C12-0,-N, mixture at (a) 111, (b)171 and (c) 243 s f8%. Following spectral stripping of Cl,, HC(0)F and CF,C(O)F, the baseline of UV absorption showed no addi- tional significant absorbance. The photooxidation study was carried out at four experi- mental temperatures in the range 235-318 K, and at a range of oxygen partial pressures. This range of oxygen pressures used at each experimental temperature was chosen such that comparable and therefore relatively easily measurable con- centrations of the two main products were obtained.Chlorine based yields of HC(0)F and CF,C(O)F obtained are shown in Table 1. Not all spectra were analysed for the CF,C(O)F yield, however, since this quantity proved to be both difficult to ascertain and of limited use in extracting branching ratios. The problem is discussed further below. In addition to the UV yield analysis, qualitative experi- ments to determine the composition of the reaction mixture following photolysis were carried out using IR spectroscopy. A spectrum of a typical photolysed mixture of HFC 134a, Cl,, 0,and N, is shown in Fig.3. In addition to the major products, absorptions attributable to COF, , HF and SiF, were observed. Discussion In the photooxidation study, the C1-atom-initiated formation of peroxy radicals, CF,CHFO,, in the absence of NO, leads to the self-reaction CF,CHFO, + CF,CHFO, +2CF,CHFO + 0, (6a) -,CF,CHFOH + CF,C(O)F + 0, (6b) This reaction provides a convenient way of studying the reac- tion of the oxy radical in the absence of chain processes initi-ated by the presence of NO in the system. Oxy radicals formed in the non-terminating step (6a) then either react further with 0, [reaction (4)] or unimolecularly decompose [reaction (5)J The final product distribution reflects this branching.The basic mechanism for the photooxidation, in the absence of NO, is illustrated diagrammatically in Fig. 4. Table 2 shows the full mechanism, which is similar to that Table 1 Yields of products from the photooxidation of HFC 134a yield" T/K [0,]/10'8 molecule cm-, HC(0)F CF,C(O)F 293 5.98 2.48 1.35 9.89 1.96 1.87 19.3 1.35 1.65 24.0 1.15 1.85 235 0.822 0.98 b- 2.06 0.75 1.42 3.25 0.60 1.26 4.11 0.59 1.34 4.77 0.60 b- 6.1 1 0.36 1.57 273 5.71 1.54 2.13 10.59 1.07 1.67 14.27 0.725 1.82 17.73 0.754 1.75 318 9.78 3.78 b- 12.20 4.50 2.26 17.80 3.12 b- 22.20 3.00 1.44 Product yields are expressed in terms of the amount of C1, con-sumed. Not analysed, see text for details.J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 t 1 I I I I 1 IllIlII1 11....1..-1...1---1...1--.1-I...I.. * -''.I.-. I ---1 . -.I. -'I --. 1-4000 3600 3200 2800 2400 2000 1800 1600 1400 1200 1000 800 600 wavenumber/cm-' Fig. 3 IR spectrum of a photolysed HFC 134a-Cl,-O,-N, mixture presented by Wallington and co-workers.' '919 Analysis of the depends on oxygen concentration measured product yields to obtain the CF,CHFO rate con- stant ratio k,/k, was carried out using two separate tech- B = k4Co,lM~4Co,l+ k,) niques. These methods are discussed individually below. Both of these branching ratios affect the observed product distribution. A preliminary analysis of the product data can Arithmetic Derivation of kJk, be undertaken by arithmetically expressing the yields.Thus, for a nominal photodissociation of x chlorine molecules in The branching ratio for the non-terminating channel of the the presence of excess of HFC and oxygen, 2x CF,CHFO,CF,CHFO, radical self-reaction is defined as 01 = k6d(k6a are produced.+ k6&. Similarly, the branching ratio for the CF,CHFO radical reaction with oxygen is defined as B. B therefore xC1, +2xC1+ 2xHCl+ 2xCF,CHFO, Table 2 Full mechanism for the photooxidation of HFC 134a in the absence of NO" reaction A EIR reaction c1, + hv +2c1 --fitted C1+ CF3CH,F +HCl + CF3CHF 1 x lo', 1958 CF3CHF + 0, +CF3CHF0, (2) 2CF3CHF0, -+ 2CF3CHF0 + 0, a6.7 x -700 (64 -+ CF,CHFOH + CF,COF + 0, (1 -a)6.7 x -700 (6b) CF3CHF0+ 0, -+ CF3COF+ HO, 6.0x 10-14 925 (4)CF,CHFO -+ CF, + HC(0)F fitted HO, + CF3CHF0, -+ CF,CHFO,H + 0, 5.7 x 10-13 -700 (7)HO, + HO, +H,O, + 0, (9)CF, + 0, +CF30, (10) CF,O, + CF30, +2CF,O + 0, 1.2 x 10-13 -800 (11) CF30, + HO, +CF30,H + 0, 5.7 x 10-13 -700 (12) CF,O, + CF3CHF0, -+ CF30 + CF3CHF0 + 0, a7.6 x lo-', -700 (13) -+ CF,OH + CF3COF + 0, (1 -a)7.6 x -700 (14) CF,O + HC(0)F --+ CF,OH + C(0)F 1.2 x lo-', 2030 (15)CF,O + CF30, +CF,O,CF, (16)CF,O + CF,CHFO, --+ CF3CHF03CF3 fitted CF,O + CF3CH,F -+ CF30H + CF3CHF 1.2 x lo-', 2030 (8) a Reactions of HO, with peroxy radicals are assumed to lead to stable hydroperoxides, with rate coefficients equal to that for the CH30, + HO, reaction.26 Temperature dependence of CF,CHFO, branching ratio, (1 -a)= 3.0 x loT3exp(1200/T).J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 CF3CH2Fexcess 2CI + 202I CF30 CF,C(O)F + CFSCHFOH L---cF302 1I CF3CH F02H 1 CF3CHF02 f CF30H ___t C(0)FZ + HF + CF3C(O)F Fig. 4 Photooxidation mechanism for HFC 134a in the absence of NO Similarly, in terms of a and p, the reaction of the peroxy radical can be expressed 2xCF3CHF0, +2xaCF3CHF0 (64 -+ 2x(1- a){CF,CHFOH + CF,C(O)F} (6b) For the oxy radical: 2xaCF,CHFO -+ 2xafiCF,C(O)F + 2xafiH0, (4) -+ 2xa(l -p)HC(O)F + 2x41 -/?)CF, (5) Considering the secondary chemistry, HO, radicals are known to react rapidly with RO, radical^,'^ usually some- what faster than the R02 self-reaction, and the most prob- able fate of HO, generated in reaction (4) is therefore the cross-reaction with the 'parent' RO, species.Thus 2xapH0, + 2xaflCF,CHFO, +2xaBCF,CHFOOH( +0,) (7) In the scheme described above, the formation of HO, can therefore be assumed to give a further consumption of xap photodissociated chlorine molecules. For CF, radicals, the chemistry is somewhat more complex. The radicals are likely to combine rapidly with oxygen to form CF302 radicals, which then undergo self- and cross-reactions with all other RO, type radicals present. A number of these reactions are expected to give rise to CF,O and, as has been recently these species react ana- logously to C1 atoms in hydrogen abstraction from hydrocar- bons. Thus, in the excess of HFC 134a used in these experiments, the CF,O radicals regenerate CF,CHFO, rad-icals via reactions (8) and (2), in this case without any con- comitant C1, consumption.The chemistry of the CF, radical is therefore simplified for the purposes of this model. It is assumed that, apart from the regeneration of CF,CHFO,, no other effect on the measured product distribution arises: 2xa(l -fi)CF, O2 2x41 -P)CF,O, CF302IR02 t 2x41 -B)OCF,O 2xa(l -p)6CF3O CF3CHzF excess 2x41 -P)BCF,OH+ 2xa(l -B)BCF,CHF (8) 2x41 -/?)6CF,CHF 02excess * 2xa( 1 -B)6CF3CHF0, (2) A further term, 0, is therefore defined here as the fractional efficiency of production of CF,O radicals from CF, . Sehested and Wallingt~n'~ have shown that CF,O radicals react with HFC 134a in this photooxidation and assign an IR absorption feature to CF,OH.The CF,OH is observed to decay slowly in the dark, giving C(0)F2 and HF. These pro- ducts were also identified in the IR spectrum of the pho- tolysed mixture from this study see Fig. 3. Furthermore, the observation of SiF, in this study presumably arises from the hetergeneous reaction of HF on the quartz cell walls. A chain reaction is therefore taking place following the ini- tiation of reaction by dissociation of chlorine, whereby both products CF,C(O)F and HC(0)F are formed and CF,CHFO, is regenerated. However, since only a fractional regeneration of the peroxy radical takes place, the chain length of this process is limited. The total amount of CF,CHFO, reacting, t(RO,), following initiation by disso- ciation of a nominal number, n, of C1, molecules can there- fore be expressed as the sum of a geometric series of the form a, ar, ar2, ar3, ..., where r is the fractional regeneration of CF,CHFO, and convergence is defined by r c 1.This sum is given by t(R0,) = a/(1 -r). The total yield of a given product, y(prod), can then be expressed as a fraction,f, of this total RO, reacting, divided by the nominal chlorine concen- tration considered to initiate reaction. Thus Y(PW =fCt(ROz)l/n wherefand t(R0,) are defined in terms of a and B. Solving in this way for the yield of the HC(0)F leads to HC(0)F yield = [2a(l -B)]/[1 -a8 -q?(1 + O)] Substituting for p in terms of k,/k, then gives the following linear relationship l/YCHW)FI = C(1 + a)/2QIC021(k4/k5)+ "1 -4/2a1 Plots of l/y[HC(O)F] us.LO,] for data at each experimen- tal temperature are presented in Fig. 5, and show reasonable linearity. Fitted parameters for these plots are shown in Table 3. Using values of a from the work of Wallington et a/.' ' values of 8 and k,/k, have been calculated and are also shown in the table. k,/k, is shown in Arrhenius form in Fig. 6. There is an indication of curvature in the Arrhenius plot, but an unweighted fit to the k,/k5 data gives k,/k, = 1.18 x exp(2860 380)/T cm3 molecule-' The errors are la. Table 3 Parameters obtained from plots of [HC(O)F]-' us. [O,] T/K intercept gradient" ab e k4k" 235 0.697 28.3 0.50 0.606 18.0 273 0.312 6.30 0.76 0.692 5.44 29 3 0.252 2.56 0.82 0.716 2.31 318 0.165 0.778 0.87 0.819 0.724 Units cm3 molecule-'. Cakulated from Wallington et al." J.CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 0.51 0.oLL 0 5 10 15 20 25 [02]/10'8molecule ~m-~ 0.1j 0 5 10 15 20 0l8molecule CM-~[0,]/10'8molecule ~m-~ [02]/1 Fig. 5 HC(0)Fyield-' us. [O,] at (a)293, (b) 235, (c) 273 and (d)318 K 8 increases almost linearly with temperature. The efficiency of regeneration of RO, radicals from CF, is therefore enhanced at higher temperatures. Thus, given that an excess of HFC 134a and oxygen was always present in these experi- ments, this can be attributed to an increased conversion of CF, into CF,O radicals at higher temperatures, presumably resulting from both an increased rate of the CF302 self- reaction relative to CF302 +RO, and a higher branching ratio for CF,O formation in the CF302+RO, cross-reactions.A similar approach was taken to try to quantify the CF,C(O)F yield in terms of a and fi. However, unlike HC(O)F, this compound is formed in small amounts in a number of different channels in the photooxidation of HFC 134a. The sensitivity of the yield to changes in a and fi is consequently reduced. Furthermore, because the apparent experimental yields of CF,C(O)F could be erroneous because of the presence of similar absorbing species, as discussed below, no meaningful estimates of branching ratios could be produced from this yield. lo-'*kj 10-21""1"1"'1'1 I'II"'I 1"""" 3.c 3.5 4.0 4.5 lo3 KIT Fig.6 kJk, deduced from arithmetic analysis (0)and computer simulation (0)of the product yields expressed in Arrhenius form In conclusion, the preliminary analysis of yields to give estimates of the branching ratio for the CF,CHFO radical reaction provides a useful indication of the likely chemistry involved in the HFC 134a photooxidation. The observed yields of HC(0)F confirm the presence of a chain reaction regenerating CF,CHFO, without chlorine consumption. Kinetic Modelling In order to provide more- exact- kinetic parameters for the chlorine-initiated photooxidation of HFC 134a, in particular the branching ratio for the reactions of the CF,CHFO radical, a computer simulation of the experimental system was carried out using numerical integration.The kinetic model used was based on the reaction mechanism shown in Table 2, together with the source references for the rate coeffi- cients. In order to define the efficiency of regeneration of CF,CHFO, radicals, 8, the degradation of the CF, radical produced in reaction (4) is elaborated in some detail. The degradation chemistry is in accordance with the product analysis work of Sehested and Wallingt~n'~ and the dis- covery in the present work of enhanced product yields attrib- uted to secondary attack on HFC 134a by the CF,O radical formed from CF302. The rate coefficients for the self- and cross-reactions of the peroxy radicals CF,O, and CF,CHFO, are taken from the recent direct kinetic studies of Maricq and Szente16 and from Nielsen et aI.17 The tem- perature dependence of the branching ratio for the formation of radical (CF,CHFO) and molecular products [CF,C(O)F +CF,CFHOH] from the CF,CHFO, self-reaction is taken from the work of Wallington et al." The same branching ratio was assumed for the cross-reaction of CF,O, and CF,CHFO, .The model also contained the newly discovered reaction of CF,O with peroxy to form stable trioxides, CF,O,R (R =CF, and CF,CHF). Rate coefficients for the formation and decomposition of the trioxides were taken from Sehested and Wallington." The temperature depen- dence of the decomposition of CF,O,CHFCF, was based on J. CHEM. SOC. FARADAY TRANS., 1994, VOL.90 the 295 K value of k = 2.7 x s-' and an assumed A factor of lo', s-'; this gave an activation energy of 99 kJ mol-'. Preliminary results from Sidebottom2' and Chen et a!.'' using a relative rate technique, and Zellner et using a flash photolysis-LIF method, have shown that the reactivity of CF30 in H-abstraction from hydrocarbons is comparable to that for OH radicals. Sehested and Wallington'' have shown that the primary product of the reaction of CF30 with HFC 134a is CF30H which decomposes heterogeneously to C(O)F, and HF. The rate coefficient for the reaction of CF,O with HFC 134a is based on the 295 K value of k =(1.1 -t 0.7) x lo-', cm3 molecule-' s-' determined by these workers, and an assumed A factor of 1 x lo-', cm3 molecule-'s-'. The rate coefficient for the reaction of C1 atoms with HFC 134a at 298 K, k = 1.4 x lo-'' cm3 molecule-' s-', is well establi~hed.~~~~~The temperature dependence is based on an assumed A factor of 1.2 x lo-', cm3 molecule-' s-'.The secondary removal of HC(0)F by reaction with C1 and CF30 was also included in the reaction scheme. The rate coefficients for the reaction of CF30 with HC(0)F were assumed to be the same as for CF30 with HFC 134a. The rate coefficient for C1 + HC(0)F at 295 K has been determined by Wall- ington et d.," k = 2.0 x lo-'' cm3 molecule-' s-'. The temperature dependence is based on an assumed A factor of 1.0 x lo-', cm3 molecule-' s-'. Using these rate coeffi- cients, the calculated secondary loss of HC(0)F was very small at the extent of reaction of HFC 134a used in the experiments. This was in accordance with the observations.The expression for the temperature dependence of the rate coefficient for the reaction of CF3CHF0 with 0, is based on the room-temperature value, determined by Zellner et a/.,' k, = 2.7 x lo-'' em3 molecule-' s-', and an A factor of 6.0 x lo-', cm3 molecule-' s-', equal to that recommended for the C,H,O + 0, reaction6 -a The unknown rate coefficient, k, (unimolecular decomposi- tion of CF,CHFO) and RCI2(the rate of Cl, photolysis) were obtained by fitting computer-generated concentration-time data for Cl,, HC(O)F, CF,C(O)F and other products, to experimental data obtained in 16 experiments covering a range of temperatures and 0, partial pressures.The kinetic equations were integrated using the FACSIMILE pr~gram,~' which contained an optimization routine for fitting unknown parameters to experimental data, using a non-linear least- squares criterion. In exploratory experiments it was found that the computed yields of CF,C(O)F were always substantially smaller than the experimentally determined yields, when values of k, and RC12were adjusted to fit the HC(0)F and C1, concentrations. This observation was consistent with results from the arith- metic analysis of yields and was therefore attributed to an experimental overestimation of the CF,C(O)F yield rather than a problem with the analyses. Such an erroneous obser- vation could arise as a result of further UV absorptions underlying the smooth CF,C(O)F spectrum.A number of candidates for these products exist, of which hydroperoxides (CF,CHFOOH, CF300H, H,O,) showing broad-band absorption in the UV region and trioxides (CF303CF,, CF,O,CHFCF,), for which no UV spectra have been re-ported, are strong possibilities. Interestingly, adding the mod- elled hydroperoxide products to the modelled CF,C(O)F yield significantly improved the fit to the experimental data, as shown in Fig. 7. However, because of this uncertainty, only HC(0)F and C1, experimental data were used to determine Rc12and k, . The effect of varying the branching ratio for the cross-reaction of CF,O, and CF,CHF02 was also tested since the amount of the molecular channel in this reaction influences the fraction of CF, radicals forming CF30, and hence the regeneration of CF3CHF0, ,i.e.the regeneration efficiency 6. As expected it was found that the value of k, needed to fit the I0 -0 100 200 300 400 500 -,111II. 0 100 200 300 400 500 600 700 800 50 100 150 200 250 300 350 Fig. 7 Computed (-) and experimental (0,HC(0)F; .,CF,C(O)F and +,Cl,) concentration-time profiles from HFC 134a photooxida- tion. Computed CF,C(O)F (---), (a) 235 K, 100 Torr; (b) 293 K, 589 Torr; (c) 273 K, 299 Torr; (d)318 K, 729 Torr. Computed CF,C(O)F + hydroperoxides (see text for details) (. . .). Table 4 Fitted values of RCl2,k, and fraction, 8, of CF, radicals recycled in the photooxidation of HFC 134a" 235 148.6 6.69 3.01 0.37 100.0 5.33 2.46 0.39 79.0 5.41 2.41 0.40 50.0 4.64 2.45 0.42 5.52 f0.85' 0.40 f0.02' 273 761.3 24.0 4.16 0.56 50 1.O 23.7 4.3 1 0.50 403.0 22.8 3.82 0.5 1 299.1 20.5 4.45 0.52 22.8 f1.6' 0.52 f0.02' 293 726.5 96.1 5.34 0.61 584.1 102 5.26 0.62 300.0 125 4.76 0.66 181.3 165 4.52 0.67 122 f31.3' 0.64 k0.03' 318 729.0 340 7.66 0.72 586.0 28 1 7.72 0.72 400.0 679 7.57 0.73 322.0 342 7.40 0.74 411 181' 0.73 f0.01* The errors are lo.Average values. HC(0)F yields increased as the branching ratio for formation of molecular products from this reaction increased. 8 and hence k, were also dependent on the relative values of the rate constants for the CF302 and the CF30 reactions, e.g.hydroperoxide and trioxide formation. Fig. 7 illustrates some of the computed and experimental concentration-time curves. Good fits to the experimental data for HC(0)F and C1, were obtained in all experiments and the values of RCll and k, were well and independently determined, except at 318 K and low 0, where the corre- lation coefficient for the fitted parameters reached 0.88. Table 4 summarizes the fitted values obtained at four temperatures for RC,2and k,, together with the values of 8, obtained from the ratio of the calculated yields of CF,OH from reactions (8), (14) and (15) and HC(0)F from reaction (5). The uncer- tainties are la values obtained from averaging four experi- ments at each temperature.The values of k,/k, obtained from the simulation are plotted in Fig. 6, together with the data from the arithmetic analysis. The agreement between the values of the ratios obtained from the two methods is good, and both data sets indicate possible curvature in the Arrhenius plot. The unweighted fit to the ratio k,/k, from the simulation gives k,/k, = 9.37 x exp(2960 & 620/T) cm3 molecule-' and a fit to results obtained from both analyses gives k,/k, = 1.05 x exp(2910 & 310/T) cm3 molecule-' The errors are la. Discussion of Kinetic Parameters Expressions for the temperature dependence of the relative rate of the decomposition of CF,CHFO compared to the reaction with 0, have been reported by Wallington et d." and by Tuazon and Atkinson', from data in the range 261-353 K and 273-320 K, respectively.In order to compare results for the unimolecular decomposition, k, was calculated in each case using the expression k, = 6.0 x lo-', exp(-925/T) cm3 molecule- s-' as used for the reference reaction in this study. Fig. 8 shows an Arrhenius plot of the J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 lU3.0 3.5 4.0 4.5 103 KIT Fig. 8 Arrhenius plot for the CF,CHFO decomposition reaction. 0, kinetic modelling. Fitted values delimited Arithmetic analysis; 0, by A,Wallington et al.;" Turazon and Atkinson;I2 +, Maricq.,and Szente.14 The dashed line is the recommended fit (see text for details). data from these sources, together with data from the present work using both the arithmetic approach and the computer simulations.Over the temperature range in which the data overlap, the value of k, from the present study appears to be significantly greater than that obtained by Wallington et al.," which is, in turn, larger than the value of Tuazon and Atkinson," both determined at pressures 2 1 atm. The determination of Tuazon and Atkinson', is based on measurements of CF,C(O)F formation, assuming that this product is produced only in reactions (4) and (64. Other potential sources of this product exist, however, e.g. the HO,+ CF3CHF0, reaction, which could lead to an underesti- mation of the value of k, relative to k,. Wallington et al." also used the CF,C(O)F yields to determine k,/k,: these results could be similarly affected.HC(0)F yields were also used and indeed a decrease of ca. 20% in the ratio k,/k, was reported in this case implying a higher value of k, . In both studies, a reasonably complete carbon balance among the measured products was reported, particularly at high [O,]. However, in neither study was the fate of HO, formed in reaction (4) considered, whereas in the present work, substantial yields of CF,CHFOOH and CF300H were predicted by the simulations using the full mechanism. Heterogeneous decomposition of the hydroperoxide CF,CHFOOH could provide another source of CF,C(O)F which is included in the carbon balance. In the present work, only the HC(0)F yields were used in the extraction of kinetic parameters.This product has a unique source in reaction (5) and analysis of the yields using two independent techniques was used to determine k,/k, . Nevertheless, an overestimation of k, cannot be ruled out in this work, in view of the sensitivity of k, to the regeneration factor, 8. In the temperature range over which these studies overlap, the temperature dependence of k, from the three relative rate studies is in good agreement. However, at the lowest temperature (235 K) in the present work, the value of k, obtained is a factor of three higher than that predicted by linear extrapolation of the Arrhenius expressions based on all data from T > 261 K. Furthermore, both Wallington et al." and Tuazon and Atkinson'* have shown that the ratio k,/k, is pressure dependent, presumably owing to unimolecular fall-off in reaction (5) at pressure below 1 atm at room tem- perature.Although the results of Wallington et al." would be consistent with a high-pressure limit value of k, up to a factor of 2 higher than the 1 atm pressure value, it seems unlikely that fall-off effects could account for the apparent J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 curvature in the Arrhenius plot implied by the 235 K results in the present study. An alternative explanation for this cur- vature is non-Arrhenius behaviour of the reference rate coeffi- cient, k,, a possibility which cannot be discounted since no determinations of the temperature dependence of CF,CHFO + 0, kinetics have been reported. Direct measurements of k, have been reported by Maricq and Szentel, and Zellner et Maricq and Szente14 used a flash photolysis time-resolved UV spectroscopic study of the CF3CFH0, self-reaction over the temperature range 211-372 K.The production and loss of the CF,CFHO radical was monitored directly from the time dependence of an absorption feature centred near 270 nm, assigned to CF,CFHO. Rate constants were derived by fitting a mecha- nism similar to that used in the present work, which is based on the work of Wallington et a!. l1 The expression derived for k, for a total pressure of 230 Torr was k, = (3.7 f0.7) x lo7 exp[-(2200 & 150)/T] s-'. These data are shown in Fig. 8. The pressure dependence obtained by Wallington et al.l1 indicates that the value of k, at 230 Torr would be about a factor of two lower than the high-pressure limit value.However, even allowing for this effect it can be seen that the room temperature value, corrected for the pressure depen- dence, is a factor of three lower than that from the relative rate determinations, and that the temperature dependence of k, from the direct determinations is much less pronounced. As pointed out by Maricq and Szente,14 their reaction condi- tions would be expected to lead to values of A, and E, lower than the high-pressure limit values. However, if fall-off effects are used to account for their reported values, they would have to be even larger than those needed to account for the results of the present study at 235 K, which already seem inconsistent with the pressure dependence observed by Wall- ington et al." Maricq and Szente', also noticed that their data for CF,CHFO decomposition were not consistent with the branching ratio for the CF,CHFO, self-reaction, Q, extrapolated to low temperatures, indicating some mechanis- tic complications at low temperature.Zellner et a1." obtained a value of k, at 298 K and 230 Torr of (1.8 & 0.4) x lo4 s-' from the time dependence of OH and NO, formation in the pulsed photolysis of C1,-HFC 1 Ma-NO-0, mixtures. Using the pressure dependence of Wallington et al." this corresponds to a value of ca. 8 x lo4 s-l at 760 Torr, in better agreement with the relative rate studies than that of Maricq and Szente.', However, at present, both the direct determinations and the relative rate studies show considerable uncertainty and the reasons for these discrepancies are not apparent.Table 5 shows the Arrhenius parameters for reaction (5) obtained from the various temperature-dependent studies. All values except those of Maricq and Szente14 are close to the high-pressure limit according to the data of Wallington et al. Nevertheless, the A factors are all significantly lower than the typical value of 1013-10'4 s-expected for a radical decomposition at the high-pressure limit.,' This suggests that Table 5 Arrhenius parameters for the reaction CF,CHFO -+ CF,+ HC(0)FO 16.7 4820 12.0 this work (arithmetic) 180 3.8 1.9 5570 4525 4435 9.98 7.5 5.1 this work (modelling) Wallington et al." Tuazon and AtkinsonI2 0.00037 7.4 2200 4720 2.0 7.5 f 2.4 Maricq and Szente14 recommended a All results, except those of Maricq and Szente, are calculated rela- tive to k, = 6.0 x exp(-925/T) cm3 molecule-' s-'.the reaction may not be as close to the high-pressure limit as the data of Wallington et al.' indicate. Further experimental investigation of the pressure dependence is clearly needed to establish if this is the case. For the purposes of calculating the rate of CF,CFHO decomposition at atmospheric pressure the expression in Table 5 is recommended. This is based on the mean of the experimental values at 293 K and the E/R values from Wall- ington et a/.," Tuazon and Atkinson', and the present work, excluding the data at 235 K.The temperature dependences obtained from the present work if the 235 K data are included, and from the lower pressure data of Maricq and Szente,14 both lead to unrealistically low A factors and are therefore not included in the evaluation. Errors Errors in individual Arrhenius expressions in this work have been quoted, at the la level solely for the temperature depen- dence (E/R)of the returned k,/k, or the k, value from the analysis. This is in accordance with other recent published on this reaction and reflects the fact that although the relative rate k,/k, is well established, the pre-exponential factor and hence the absolute values of k, and k, are not. Furthermore, because of the limited number of data points obtained in this study, a full statistical analysis of errors is inappropriate. The error given in the final recommendation for k, has been obtained using the procedure adopted by the NASA panel for chemical data evaluation.' This method uses the error in the ambient temperature value of k, [f(293)] and that in the temperature dependence (AEIR)to determine the approximate la errors over the entire temperature range.The errors are obtained by multiplying or dividing the value of k, at any given temperature, T, byf(T). Thus AE 1f(T) =f(293) exp 1 -R (-T -L,1293 Adopting this procedure, the percentage errors in k, for this study are f30% at ambient temperature, rising to f56% at 235 K and f38% at 318 K.The average errors in the recommended k, value are &40% across the entire tem- perature range. Atmospheric Modelling There is considerable interest in the relative rate of pro- duction of CF,C(O)F and HC(0)F in the atmospheric oxida- tion of HFC 134a. Neither of these compounds will be significantly photolysed in the tropo~phere~?~ and their major fate is likely to be physical removal in the aqueous phase. Hydrolysis of CF,C(O)F leads to CF,C(O)OH, tri-fluoroacetic acid (TFA), which is the subject of some environ- mental concern., TFA is also likely to be physically removed in rain water3, and the source region of CF,C(O)F will therefore determine the atmospheric distribution of CF,C(O)OH produced from HFC 134a photooxidation.A 2D model2, was used to calculate the latitude-height dis-tribution of the OH-induced oxidation of HFC 134a. Coupled with a knowledge of the temperature and pressure dependence of the relative rate of formation of CF,C(O)F and HC(O)F, the distribution of the source term of these two products in the atmosphere was computed. The model2, is a classical zonally averaged Eulerian model. It extends horizontally from 90" S to 90" N in 19 discrete latitude boxes and from ground level vertically up to 60 km in 17 levels with a resolution of 3.5 km. The model includes a representation of tropospheric photochemistry in order to produce a seasonally varying OH concentration 50 40 E Y,$ 30 c.-e-lu 20 I, I l/+9;:--fI I'"1, ,-75 -50 -25 0 25 50 75 latitude Fig.9 Modelled atmospheric abundance of HFC 134a in ppt after photochemical loss by the hydroxy radical field, In the model HFC 134a was released mainly between latitudes 33 and 62" N at ground level at a fixed rate of 29.5 kt year-' i.e. 10% by volume of the 1986 CFC 11 emission rates. The calculated tropospheric lifetime of HFC 134a in the model due to loss by the hydroxyl radical, using a value for k(OH + HFC 134a) of 8.4 x lo-', exp -153513, was 12 years. This is somewhat shorter than the value of 15 years calculated by Prather et al.,, the differences probably arising from the higher hydroxy group concentrations in the present model. Fig. 9 shows the distribution of HFC 134a in ppt following a 40 year integration with photochemical destruction by the hydroxy radical.The even global distribution reflects the relatively long tropospheric lifetime (12-15 years) of this hydrofluorocarbon. Fig. 10 shows the logarithm of the annual average destruction rate of HFC 134a in molecule cm-, s-l due to loss by the OH radical. The major loss (ca. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 10.0-E5 7.5. Q1-0 c.-c-(Ll 5.0 2.5 -75 -50 -25 0 25 50 latitude Fig. 11 Modelled fraction of CF,C(O)F formed from the CF,CHFO radical in the atmosphere 1 molecule cm-' s-') occurs in the tropical lower tropo- sphere from 40" S to 40" N corresponding to both high OH fields and high concentration of the hydrofluorocarbon source.Following the destruction of HFC 134a the oxy radical, CF,CHFO, formed may either react with molecular oxygen to form CF,C(O)F and HO, ,reaction (4) or undergo unimolecular decomposition to CF, and HC(O)F, reaction (5) CF,CHFO + 0, +CF,C(O)F + HO, (4) CF,CHFO +CF, + HC(0)F (5) Fig. 11 shows the fraction of CF,C(O)F formed in the tropo- sphere from the degradation of the CF,CHFO radical using the recommended parameters for k, and k, from Table 5 and the pressure dependence of the ratio k,/k, from Wallington et 1 2.5I -75 -50 -25 0 25 50 75 -75 -50 -25 0 25 50 75 latitude latitude Fig. 10 Logarithm (base 10) of the modelled annual destruction rate of HFC 134a (molecule cm-,s-') average Fig.12 cm-, s-l) formed from the photooxidation of HFC 134a Annual average production rate of CF,C(O)F (molecule J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 1829 al." assuming the same pressure dependence at low tem- peratures. The fraction of CF,C(O)F produced increases with altitude (lower temperatures and decreasing pressure both favour this channel, the temperature effect being dominant), from a value of 20% at ground level to ca. 80% at the tropo- pause. However, the maximum loss of HFC 134a occurs in 10 11 12 0.V. Rattigan, D. M. Rowley, 0. Wild, R. L. Jones and R. A. Cox, in Kinetics and Mechanisms for the Reactions of Haloge-nated Organic Compounds in the Troposphere, CEC-AFEAS Workshop, University College Publishers, Dublin, 1993, p.88. T. J. Wallington, M. D. Hurley, J. C. Ball and E. W. Kaiser, Environ. Sci. Technol., 1992, 26, 13 18. E. Tuazon and R. Atkinson, J. Atmos. Chem., 1993, 16,301. the tropical lower troposphere (Fig. 10) and the rate of CF,C(O)F formation is a maximum in this region because the destruction rate of HFC 134a falls off more rapidly with altitude. Fig. 12 shows the computed rate of CF,C(O)F pro- duction from HFC 134a degradation. Preliminary results by DeBruyn et suggest that efficient removal of CF,C(O)F 13 14 15 E. 0.Edney and D. J. Driscoll, Int. J. Chem. Kinet., 1992, 24, 1067. M. M. Maricq and J. J. Szente, J. Phys. Chem., 1992, %, 10862. R. Zellner, A. Hoffmann, D. Bingemann, V. Mors and J. P. Kohlmann, Kinetics and Mechanisms for the Reactions of Halo-genated Organic Compounds in the Troposphere, STEP-HALOCSIDE/AFEAS Workshop, Dublin, 1991.by hydrolysis in cloud water to form CF,C(O)OH occurs with a lifetime of ca. 1 month. Most of the hydrolysis will occur in the tropical lower troposphere and hence the hydro- lysis product, CF,C(O)OH, will be formed in this region. 16 17 18 M. M. Maricq and J. J. Szente, J. Phys. Chem., 1992,%, 4925. 0. J. Nielsen, T. Ellermann, J. Sehested, E. Bartkiewicz, T. J. Wallington and M. D. Hurley, lnt. J. Chem. Kinet., 1992, 24, 1009. T. J. Wallington, J. Sehested, M. A. Dearth and M. D. Hurley, J. Photochem. Photobiol. A, 1993, 70, 5. The authors wish to thank the Alternative Fluorocarbon 19 J. Sehested and T. J. Wallington, Environ. Sci. Tech., 1993, 27, Environmental Acceptability Study SPA-AFEAS, Inc. and the Department of the Environment, UK for financial support.Thanks are also due to T. J. Wallington for the com- munication of his work prior to publication. 20 21 22 146. H. Sidebottom, personal communication. J. Chen, T. Zhu, H. Niki and G. J. Mains, Geophys. Res. Lett., 1992,19,2215. H. Saathoff and R. Zellner, Chem. Phys. Lett., 1993,206, 349. 23 K. S. Law and J. A. Pyle, J. Geophys. Res., 1993,98, 18377. References 24 0.Rattigan, E. Lutman, R. L. Jones, R. A. Cox, K. Clemitshaw and J. Williams, J. Photochem. Photobiol. A, 1992,66, 31 3. Scientijic Assessment of Ozone Depletion: 1991, WMO Global Ozone Research and Monitoring Project, Report No. 25, 1992, Geneva, Switzerland. R. S. Stolarski, M. R. Schoeberl, P.A. Newman, R. D. McPeters 25 26 27 G. A. Olah and S. J. Kuhn, J. Am. Chem. Soc., 1960,82,2380. L. E. Giddings Jr. and K. K. Innes, J. Mol. Spectrosc., 1961, 6, 528. P. D. Lightfoot, B. Veyret and R. Lesclaux, J. Phys. Chem., 1990, and A. J. Krueger, Geophys. Res. Lett., 1990, 17, 1267. M. Prather and C. M. Spivakovsky, J. Geophys. Rex, 1990, 95, 18723. 28 94, 708. T. J. Wallington and M. D. Hurley, Chem. Phys. Lett., 1992, 189, 437. Scientific Assessment of Stratosphere Ozone, WMO Global Ozone Research and Monitoring Project, Report No. 20, 1989, vol. 2, AFEAS report. W. B. DeMore, S. P. Sander, D. M. Golden, R. F. Hampson, M. J. Kurylo, C. J. Howard, A. R. Ravishankara, C. E. Kolb and M. J. Molina, Chemical Kinetics and Photochemical Data for Use in Stratospheric Modeling, NASA Evaluation No. 10, 1992, Jet Propulsion Laboratory Publication 92-20, Pasadena, CA, USA. R. Atkinson, D. L. Baulch, R. A. Cox, R. F. Hampson, J. A. Kerr and J. Troe, J. Phys. Chem. R@. Data, 1992,21, 1125. T. J. Wallington and 0.J. Nielsen, Chem. Phys. Lett., 1991, 187, 33. 29 30 31 32 33 J. P. Sawerysyn, A. Talhaoui, B. Meriaux and P. Devolder, Chem. Phys. Lett., 1992, 198, 197. A. R. Curtis and W. P. Sweetenham, FACSIMILEICHEKMAT Users Manual, 1988, AERE-R12805, Harwell, Laboratory, Oxfordshire. S. W. Benson, Thermochemical Kinetics, Wiley-Interscience, New York, 2nd edn., 1976. J. H. Hu, Jeffrey A. Shorter, P. Davidovits, D. R. Worsnop, M. S. Zahniser and C. E. Kolb, J. Phys. Chem., 1993,93, 11037. W. DeBruyn, J. A. Shorter, P. Davidovits, D. R. Worsnop, M. S. Zahniser and C. E. Kolb, Atmospheric Wet and Dry Deposition of Carbonyl and Haloacetyl Halides, AFEAS Workshop, 1992, J. Peeters and V. Pultau, CEC-AFEAS Workshop, September Brussels, AFEAS, New York, 1993, p. 12. 1992, Leuven, CEC Air Pollution Research Report 45. 0. V. Rattigan, 0. Wild, R. L. Jones and R. A. Cox, J. Photo- chem. Photobiol. A, 1993, 73, 1. Paper 3/07325D; Received 13 December, 1993