Pressure dependent OH yields in the reactions of CH 3 CO and HOCH 2 CO with O 2

OH-formation in the reactions of CH 3 CO (R1) and HOCH 2 CO (R4) with O 2 was studied in He, N 2 and air (27 to 400 mbar) using OH-detection by laser induced fluorescence (LIF). 248 nm laser photolysis of COCl 2 in the presence of CH 3 CHO or HOCH 2 CHO was used as source of the acyl radicals CH 3 CO and HOCH 2 CO. The LIF-system was calibrated in back-to-back experiments by the 248 nm laser photolysis of H 2 O 2 as OH radical precursor. A straight-forward analytical expression was used to derive OH yields ( a ) for both reactions. A Stern–Volmer-analysis results in a 1b (cid:2) 1 (N 2 ) = 1 + (9.4 (cid:3) 1.7) (cid:4) 10 (cid:2) 18 cm 3 molecule (cid:2) 1 (cid:4) [M], a 1b (cid:2) 1 (He) = 1 + (3.6 (cid:3) 0.6) (cid:4) 10 (cid:2) 18 cm 3 molecule (cid:2) 1 (cid:4) [M] and a 4b (cid:2) 1 (N 2 ) = 1 + (1.85 (cid:3) 0.38) (cid:4) 10 (cid:2) 18 cm 3 molecule (cid:2) 1 (cid:4) [M]. Our results for CH 3 CO are compared to the previous (divergent) literature values whilst that for HOCH 2 CO, for which no previous data were available, provide some insight into the factors controlling the yield of OH in these reactions.


Introduction
Acetyl radicals (CH 3 CO) play an important role in atmospheric chemistry. Important sources of acetyl radicals are the photolysis of acetone in the upper troposphere and the reaction of acetaldehyde with OH in the troposphere. The hydroxylsubstituted hydroxy acetyl radicals (HOCH 2 CO) are formed in the reaction of OH with glycol aldehyde (HOCH 2 CHO). The only significant reaction of acetyl and hydroxy acetyl radicals in the atmosphere is with O 2 , forming (mainly) peroxy radicals. Accompanying peroxy radical formation, (R1) displays a second reaction pathway forming OH and an organic by-product. The branching ratio (a) for formation of OH increases from small values (o2%) at standard pressure to unity at pressures close to zero. 1 Reaction (R1) is considered to proceed via an excited peroxy radical CH 3 C(O)O 2 # that is either stabilised by collisions with the bath gas molecules M or decomposes to form OH. [2][3][4][5][6][7][8][9] This is illustrated in reaction Scheme 1 (R = CH 3 ). The pressure dependence of a thus originates from the competition between the pressure-and bath gas-dependent quenching rate [M] Â k M and the pressure-independent decomposition rate k D . A kinetic (Stern-Volmer) analysis of the reaction scheme leads to eqn (1) which can be used to parameterise a: Although OH yields are low at pressures typical for the troposphere, (R1b) has an indirect impact on atmospheric chemistry because of its occurrence in laboratory experiments. The OH product of (R1b) has, for example, been used as spectroscopic marker for CH 3 CO formation in the determination of photo-dissociation quantum yields for acetone, 10 an important source of HOx radicals and PAN (CH 3 C(O)O 2 NO 2 ) in the upper troposphere. 11,12 Recent studies on the yield of OH in the reaction between CH 3 C(O)O 2 and HO 2 observed OH from (R1). 13 The title reaction will also have occurred in and potentially impacted on the results of studies of PAN formation in (R2) at low pressures where the yield of OH is large. For example, in their study of PAN formation, Bridier et al. 14 generated CH 3 CO radical in the presence of O 2 to examine the kinetics of the reaction of CH 3 C(O)O 2 + NO 2 (R2) at pressures down to 20 mbar. As the results of the present publication show, at such pressures 18% of CH 3 CO reacting with O 2 forms OH instead of CH 3 C(O)O 2 . Data recorded at low pressure by Bridier et al. might thus be subject to systematic error since reaction channel (R1b) was not known to take place in 1991.
The reaction of CH 3 C(O)O 2 with HO 2 (R3), which competes with (R2) at low NOx levels, 15 has drawn considerable interest in recent years. 13,[16][17][18] Its main reaction channel (R3a) preserves a HOx species (HOx is OH + HO 2 ) and an organic radical and is hence radical-propagating, which helps sustain atmospheric oxidation capacity.
In experiments on (R3), CH 3 C(O)O 2 and HO 2 are usually generated by reaction of Cl atoms with CH 3 CHO and CH 3 OH in air involving intermediate generation of CH 3 CO and CH 2 OH radicals. Therefore, OH-generation influences the initial [CH 3 C(O)O 2 ]/ [HO 2 ]-ratio in these experiments. In product studies that do not allow for an experimental separation between different OHformation routes, (R1b) must be well known so that discrimination between OH formed in (R1b) and (R3a), respectively, is possible.
In the present work we employ a new experimental approach to quantify the pressure-dependence of the OH forming channels (R1b) and (R4b) of the reactions of O 2 with CH 3 CO and its OH-substituted analogue HOCH 2 CO.
We assume that, for reaction (R4), the same pathways are available as in (R1), i.e. competition between peroxy-radical formation and OH (see Scheme 1, R = HOCH 2 ). The formation of the peroxy radical, its UV-absorption spectrum and its reaction with HO 2 will be subject of a future publication from this group.
Throughout this work the branching ratios of the OHforming reaction channels are defined as follows: k 1b /k 1 = a 1b and k 4b /k 4 = a 4b .

Experimental set-up
The experiments detailed in this publication were performed using the pulsed laser photolysis-laser induced fluorescence (PLP-LIF) apparatus that has been described previously 19,20 and only a short description is given here. Experiments were conducted in a 500 cm 3 reactor at room temperature. The pressure was monitored with a capacitance manometer, and gas flow rates were selected such that a fresh gas sample was available for photolysis at each laser pulse. Reactions were performed at pressures between 27 and 400 mbar in nitrogen and helium bath gases with added O 2 or in air.
Fluorescence from OH was detected by a photomultiplier tube shielded by a 309 nm interference filter and a BG 26 glass cut-off filter. The frequency doubled emission from a Nd-YAGpumped dye laser (Quantel, Lambda Physik) was used to excite the A 2 S(n = 1) ' X 2 P(n = 0), Q 11 (1) transition of OH at 281.997 nm.

Chemicals
Liquid samples of CH 3 CHO (Roth, Z99.5%) were degassed by repeated evacuation, and stored in a blackened glass bulb as B1% mixture in N 2 . HOCH 2 CHO was prepared during the experiments from its dimer (Sigma-Aldrich) by heating the solid sample to 50-75 1C and eluting gaseous HOCH 2 CHO by a continuous flow of N 2 . COCl 2 (Fluka, 499%) was stored in a stainless steel canister as B4% mixture in N 2 or He. H 2 O 2 (AppliChem, 50%) was concentrated in vacuum to 480% and used as liquid sample. He (Westfalen, 99.999%), N 2 (Westfalen, 99.999%) and O 2 (Westfalen, 99.999%) were used as supplied.

Experimental approach
We performed back-to-back experiments in reaction mixtures containing either H 2 O 2 or COCl 2 as photolytic sources of OH radicals or Cl atoms. Addition of CH 3 CHO or HOCH 2 CHO to the COCl 2 experiments converted Cl atoms into CH 3 CO or HOCH 2 CO, which reacted with O 2 to form OH. This allowed us to compare OH formation via title reactions (R1b) and (R4b) directly with OH production from H 2 O 2 -photolysis, a well-characterized source of OH radicals. Formation of acyl radicals by the reaction of Cl atoms with CH 3 CHO (DH = À58 kJ mol À1 ) 18 or HOCH 2 CHO (DH = À49 kJ mol À1 ) 18,21 are exothermic processes and the nascent fragments are expected to be vibrationally and rotationally hot. Assuming an energy transfer efficiency of 300 cm À1 per collision with N 2 , hot CH 3 CO would be deactivated within 16 collisions, ensuring that, at the high pressures of bath gases used in this study, acetyl should, to a good approximation, be thermalized before reaction with O 2 takes place. Experiments in which N 2 was mixed with 1% O 2 yielded the same results as those with 21% O 2 , so that no evidence was obtained for reaction of non-thermalised CH 3 CO with O 2 . Even in the experiments in He (presumably a less efficient energy transfer medium that N 2 ) no dependence of the OH-yield on O 2 partial pressure was obtained. We note also that the existence of a direct channel for OH-formation from excited CH 3 CO and O 2 is considered unlikely. 9 3.1.1 Determination of a. The OH-LIF system was calibrated by photolysing H 2 O 2 that, at 248 nm, generates OH radicals with a quantum yield of 2. 22 Quasi-instantaneous photolytic OH-formation and subsequent OH loss via (R6) result in a mono-exponential decay of [OH] that was recorded by OH-LIF.
The LIF-signal is proportional to [OH] and was fitted by eqn (2) where f cal is a calibration factor that quantifies the sensitivity of the LIF-system.
where a OH and s OH represent the fitted parameters.
Combining eqn (2) and (3) we get: In back-to-back experiments, H 2 O 2 was replaced by COCl 2 and an acyl radical source (CH 3 CHO or HOCH 2 CHO). Photolysis of COCl 2 generates Cl atoms with a quantum yield of 2. 23,24 COCl 2 + hn (248 nm) -2Cl + CO (R7) Reaction of Cl with CH 3 CHO (R8) forms CH 3 CO with a yield very close to unity (k 8 = 8.0 Â 10 À11 cm 3 molecule À1 s À1 ). 17 Cl + CH 3 CHO -CH 3 CO + HCl (R8) OH formation in reaction (R1b) and its main loss via reaction (R9) are both resolved on the time-scale of our experiments and a bi-exponential time-dependence of the LIF-signal is observed (k 9 = 1.5 Â 10 À11 cm 3 molecule À1 s À1 ). 17,18 Reaction with OH generates mainly CH 3 CO (a 9a = 0.95) which is accounted for in the analytical expression of the [OH] time evolution presented below. Only 5% of the OH formed in (R1) is thus converted via (R9b) into CH 2 CHO. Even if CH 2 CHO were converted with unity yield into OH radicals, this would result in a maximum overestimation of no more than 5% in the value of a 1b .
where a Cl1 , a Cl2 and s Cl1 represent the fitted parameters. Under the assumptions that (R1) is fast compared to (R8) and (R9) and that [CH 3 CHO] remains constant on the experimental timescale, an analytical expression for the temporal evolution of [OH] can be derived.
In this expression, and [Cl] 0 is the Cl-concentration initially formed by photolysis.
Conditions were chosen such that reaction (R1) was 5.2-250 times faster than (R8), and 28-1300 times faster than (R9) and thus fast on the experimental time-scale of B1 ms. Combining eqn (5) and (6) we get: f cal can be eliminated from eqn (7) by insertion of eqn (4) because experiments were conducted back-to-back.
This assumes that fluorescence quenching is dominated by the bath gas and that the contribution of reactants is negligible so that switching between H 2 O 2 and COCl 2 /aldehyde does not change the detection sensitivity to OH. The experiments performed in He, which is a weak quencher of OHfluorescence, are the most likely to be influenced, should this not be the case. In Section 3.3 we show however that such quenching effects did not have a measurable effect on the results obtained. The HOCH 2 CHO was used as acyl radical precursor in experiments for the determination of a 4b . Reaction of Cl atoms with HOCH 2 CHO (R10) forms HOCH 2 CO with a yield of a 10a = 0.65 (k 10 = 7.5 Â 10 À11 cm 3 molecule À1 s À1 ). 25,26 Reaction of HOCHCHO with O 2 (R11), is not known to form OH. 25,26 Cl + HOCH 2 CHO -HOCH 2 CO + HCl (R10a) -HOCHCHO + HCl (R10b) Reaction with HOCH 2 CHO (R12) is the main OH loss channel in these experiments.
Reaction with OH generates HOCH 2 CO with a higher yield (a 12a = 0.80) than reaction with Cl. Based on this kinetic scheme, one again expects a bi-exponential time profile of [OH] that can be analysed by eqn (5). The temporal evolution of [OH] is described by the integrated rate law (11) which was derived analytically assuming reaction (R4) to be fast compared to reactions (R10) and (R12) and that [HOCH 2 CHO] was not significantly depleted during the experiments.
Under the experimental conditions applied in this work and assuming the rate coefficients of (R1) and (R4) to be equal, (R4) was 12 to 280 times faster than (R10) and 110 to 2700 times faster than (R12) and, thus, fast on the experimental time-scale of B1 ms. As for the CH 3 CO + O 2 system we can derive an analytical expression for a 4b from eqn (4) and (11).  (9) and (12) can be expanded in a Taylor series that is stopped after the second term. By insertion of eqn (9) and (10) This allows us to separate statistical errors, i.e. reading errors or uncertainties in the determinations of s OH and s Cl which are small, from the systematic errors originating from uncertainties in literature values of the absorption cross sections and, in the case of a 4b , the branching ratio a 10a = 0.65 AE 0.05. 25

CH 3 CO + O 2 (N 2 /O 2 )
Back-to-back PLP-LIF-experiments on (R1) were performed at pressures between 133 and 270 mbar of N 2 or at 27 and 270 mbar of air. Fig. 1 shows a pair of OH-LIF time profiles recorded at 133 mbar in N 2 . Both OH-profiles display the expected kinetics and were analysed using eqn (2) or (5), respectively. Although this work was not performed to re-measure the rate-coefficients of reactions (R8) and (R9), we did derive them from the fit-parameters and the respective [CH 3 CHO] as a check of our experimental approach. We obtained k 8 = (7.4 AE 1.1) Â 10 À11 cm 3 molecule À1 s À1 , and k 9 = (1.8 AE 0.2) Â 10 À11 cm 3 molecule À1 s À1 where the uncertainties represent statistical errors (2s) in the fit-parameters only.
In Fig. 2 these error margins are represented by thin solid lines. We note that the data show no dependence on the O 2 concentration and that values of a 1b determined in air would be higher if CH 3 CO were not thermalized and if there were an additional OH-formation route via CH 3 CO # + O 2 . This observation rules out a significant contribution of hot acetyl radicals.

CH 3 CO + O 2 (He)
Back-to-back PLP-LIF-experiments on (R1) were performed at pressures between 33 and 400 mbar of He. From the fitparameters and the respective [CH 3 CHO] we derived k 8 = (7.6 AE 0.4) Â 10 À11 cm 3 molecule À1 s À1 , and k 9 = (1.9 AE 0.1) Â 10 À11 cm 3 molecule À1 s À1 where the uncertainties represent statistical errors (2s) in the fit-parameters, only. As described above, [CH 3 CHO] was determined from barometric and mass flow readings and carries an additional uncertainty of B20%. These values are, within combined uncertainties, in accordance with the currently recommended literature values. The two data points obtained using 1. We therefore re-fitted the data using eqn (1), i.e. we performed another linear regression with the intercept being fixed to 1.6 (thick solid line in Fig. 3). From this we derived k M k D ¼ ð3:62 AE 0:05Þ Â 10 À18 cm 3 molecule À1 (error statistical only, 2s). We applied eqn (13) to incorporate the systematic uncertainties (2s) and derived a final value of k M k D ¼ ð3:6 AE 0:6Þ Â 10 À18 cm 3 molecule À1 .
In Fig. 3 these error margins are represented by thin solid lines.

HOCH 2 CO + O 2 (N 2 /O 2 )
Back-to-back PLP-LIF-experiments using HOCH 2 CHO as acyl radical precursor were performed at pressures between 33 and 269 mbar in N 2 or air. Fig. 4 shows a pair of OH-LIF time profiles recorded at 133 mbar in N 2 which were fitted using eqn (2) or (5), respectively. As a check for possible error sources  Assuming that (R4b) forms OH with unity yield at pressures approaching 0 mbar, we expect the intercept to be unity in air which, within statistical uncertainty, is the case. In the N 2 -experiments a constant amount of 2.7 mbar of O 2 was added. Thus, at [M] = 0 molecule cm À3 we expect a 4b À1 to approach B1.1, i.e. the value we derived using [M] = 6.5 Â 10 16 molecule cm À3 and k M k D ¼ 1:6 Â 10 À18 cm 3 molecule À1 . This is also confirmed by the data. In both cases we performed linear regressions according to eqn (1), with the intercept being fixed to 1.1 for the data recorded in N 2 , and to unity for the air-data. We derived k M k D ¼ ð1:85 AE 0:16Þ Â 10 À18 cm 3 molecule À1 (solid black line in The slightly larger a 4b -values observed at higher pressures in air are potentially due to experimental scatter. Our data do not however allow us to completely rule out the existence of an additional, O 2 -dependent OH-source as the cause. Therefore, we decided to rely exclusively on the data recorded in N 2 (with 1-10% of O 2 added) which would be less impacted by such an additional OH-source. Doing so we commit a maximum error of 7% in a 4b compared to values derived from all data. The fact that the data obtained at a fixed O 2 -to-N 2 ratio of 21%, but at various pressures (and thus at different O 2 concentrations), display no significant deviation from the expected behaviour, suggests that an additional OH forming channel that is dependent on the O 2 partial pressure is not significant. Incorporation of systematic uncertainties (2s), results in a final value of k M k D ¼ ð1:85 AE 0:38Þ Â 10 À18 cm 3 molecule À1 . The error margins that also enclose the data recorded in air are presented in Fig. 5 by thin solid lines. Our studies on OH formation in the reactions of CH 3 CO (R1) and HOCH 2 CO (R4) with O 2 reveal a strong dependence of the yield on substituents, with k M k D for (R4) a factor of 5 smaller than for (R1). Under the assumption that the collisional quenching of both activated peroxy radicals proceeds at a similar rate this large difference can be attributed to a more efficient decomposition of HOCH the k M k D -values from these studies as well as those from our work. Note that in Fig. 3 the literature data are plotted with an intercept of 1.6 to take into account the presence of O 2 in our experiments at extrapolated zero mbar of He. Tyndall et al. 5 studied the reaction of Cl atoms with CH 3 CHO by irradiation of Cl 2 -CH 3 CHO-mixtures in N 2 or O 2 in environmental chambers and analysed the reaction mixtures by infra-red absorption spectroscopy. They found a pressuredependence of the apparent rate coefficient of (R8) when the experiments were performed in O 2 but none for the measurements in N 2 . The value measured in O 2 increased if the experimental pressure was decreased; at 1.6 mbar the apparent rate coefficient was 2.7 times higher than that derived in N 2 . The authors attributed these findings to OH formation in (R1b). Thus, they did not directly detect OH, but their kinetic and product studies provided strong evidence for OH formation. The k M k D -value shown in Table 1 was derived by Carr et al. 3 based on a personal communication with Tyndall et al. 5 Blitz et al. 4 used the 248 nm pulsed laser photolysis of CH 3 C(O)CH 3 in He to generate CH 3 CO and used OH-LIF for the detection of hydroxyl radicals formed in (R1b) at pressures between 13-533 mbar. Calibration of the LIF-system was achieved by fixing a 1b at zero pressure to unity, which neglects to take into account the fact that the acetyl radical yield is pressure dependent as a significant (but variable) fraction thermally decomposes to CH 3 and CO, at least in nitrogen bath gas. [31][32][33] Blitz et al. could thus have underestimated the value of KM/KD by about 16%. 3 Talukdar et al. 34 used different photolytic schemes (photolysis of acetone, Cl + CH 3 CHO and OH + CH 3 CHO) for CH 3 CO generation coupled to OH-LIF to investigate OH-formation or modification of OH kinetics due to (R1) at experimental pressures between 27-800 mbar in He, N 2 and O 2 . The resulting k M k D values are in good agreement with our results.
Kovács et al. 6 used two low-pressure fast discharge flow tubes (operated at pressures between 1.3 and 11 mbar in helium) that were equipped with LIF or resonance fluorescence detection of OH radicals. CH 3 CO was formed by reacting CH 3  Relative values of a 1b were measured in the pressure range of 7-400 mbar and based on an absolute scale by setting a 1b at 0 mbar to unity. Data were corrected by 25-35% for a pressuredependence 23,31 in the CH 3 CO yield of acetone photolysis. In their N 2 -experiments the authors needed to make an additional correction since they observed a decrease of LIF-sensitivity at elevated pressures. The correction factors were derived in separate experiments by measuring OH-formation from 248 nm photolysis of t-butylhydroperoxide at the same pressure.
Although not a detailed study of the OH yield in the title reaction, we recently published data on OH formation in the reaction of HO 2 with CH 3 C(O)O 2 (ref. 13) and also observed (in this case ''unwanted'') OH-formation via (R1b). This work was conducted in a different apparatus and used a different CH 3 CO-formation scheme (355 nm-pulsed photolysis of CH 3 CHO-CH 3 OH-Cl 2 -O 2 -N 2 -mixtures). OH was detected by an OH-LIF-unit that was calibrated by measuring OH from the reaction of HO 2 with NO. In spite of the different experimental approach we could accurately simulate the OH signals due to reaction (R1b) with the OH yield presented in the current work (see Fig. 8 in Groß et al. 13 ). Use of a from the more recent publication of Carr et al. 2 would have resulted in an overestimation of initial OH-formation by a factor of 3. Tyndall (1997), 5 Blitz (2002), 4 Talukdar (2006), 34 Kovács (2007), 6 Carr (2007), 3 Carr (2011) Our results are in good agreement with those of Talukdar et al. 34 and Kovács. 6 We cannot explain the differences between our work and that of Blitz et al. 4 and Carr et al. 2,3 but we highlight the fact that no correction needs to be applied to our data.
In Table 1  Butkovskaya et al. 35 investigated the OH-initiated oxidation of HOCH 2 CHO in a turbulent flow reactor at 267 mbar of N 2 . A chemical ionisation mass spectrometer was used to detect OH and derive a yield of a 4b = 22%. This high yield may reflect the fact that Butkovskaya et al. 35 were unaware that the reaction of HOCH 2 C(O)O 2 with HO 2 (formed at a yield of 20% from OH + HOCH 2 CHO in the presence of O 2 ) forms OH with a yield of B70%. 16,36 OH formation has also been observed 37 in the reaction of O 2 with CH 3 OCO, which is isomeric with HOCH 2 CO. Similar to (R4), OH-formation is accompanied by CH 2 O and CO 2 by-products. The value of k M k D reported, (7.4 AE 1.9) Â 10 À18 cm 3 molecule À1 , is four times larger than our value for HOCH 2 CO. Given that the products of decomposition are identical the difference must be related to energetic differences in the transition state leading to dissociation.

Conclusion
We determined the pressure-dependence of the OH-forming branching ratios a 1b of reaction (R1a) and a 4b reaction (R4b) using a novel experimental approach. The values for a 1b are in accordance with some earlier studies 6,34 but clearly differ from those from the Leeds group 2-4 that derive much higher OH yields. Our data for a 4b show that hydroxylation of CH 3 CO enhances OHformation in the reaction with O 2 by approximately a factor of five.