The photodissociation of acetone has been studied over the wavelength (λ) range 279–327.5 nm as a function of temperature (T) and pressure (p) using a spectroscopic method to monitor the acetyl (CH3CO) radical fragment. Above 310 nm the quantum yield (QY) is substantially smaller than previous measurements, and decreases with T. The QYs for production of CH3CO + CH3 and CH3 + CH3 + CO have been parameterised as a function of λ, p and T and used to calculate the altitude dependence of the photolysis frequency. In the upper troposphere (UT) the acetone photolysis lifetime is a factor of 2.5–10 longer, dependent upon latitude and season, than if the previously recommended QYs are used.
 Airborne field campaigns have shown acetone to be an abundant and ubiquitous species in the troposphere [Singh et al., 2001]. Photolysis of acetone in the UT is believed to be an important source of HOx radicals and peroxyacetylnitrate (PAN), and proceeds via two channels with different wavelength thresholds [Atkinson et al., 2002] (see http://www.iupac-kinetic.ch.cam.ac.uk):
The first channel dominates under tropospheric conditions. Inclusion of acetone photolysis has improved the agreement between measured and modelled UT concentrations of OH and HO2 [Jaeglé et al., 2001]. Previous measurements of the QY for acetone photolysis under atmospheric conditions have used gas chromatographic methods to monitor the loss of acetone or the build up of stable products, formed as a result of several reactions of the primary photofragments. Gierczak et al.  monitored the loss of acetone and the production of CO2 to determine the QY in air between 248–337 nm. A p dependence was observed above 270 nm but with almost no T dependence. Emrich and Warneck  measured PAN QYs between 280–330 nm at 298 K formed through the scavenging of the CH3CO(O2) radical by the addition of NO2, and found similar results. Warneck  used the results of both studies to derive a parameterisation that is recommended by IUPAC [Atkinson et al., 2002]. In this work we use the OH radical, formed as a minor product of the reaction between CH3CO and O2 [Blitz et al., 2002; Tyndall et al., 1997]:
as a sensitive spectroscopic marker of the CH3CO fragment. We have measured the acetone photodissociation QY over a wide range of λ, p and T.
 The apparatus uses laser-flash photolysis combined with laser-induced fluorescence (LIF), and has been described previously [Blitz et al., 2002]. Acetone was photolysed using an excimer laser operating at 248 nm or a frequency-doubled pulsed tunable dye laser covering the λ range 279–327.5 nm, generating [CH3CO] ≤ 1011 molecules cm−3. The OH produced by reaction (3b) was probed by LIF at 282 nm using a second dye laser, with a detection limit ≤108 molecule cm−3. The two photolysis lasers counter-propagated through the photolysis cell (one or the other was blocked in a given experiment) at right-angles to the probe laser. Care was taken to ensure that the spatial overlap between the two photolysis lasers and the probe laser was identical for all experiments. The pulse repetition frequency of the lasers was 5 Hz, and the total flow of gas ensured there was a fresh sample of gas for each laser shot. A small concentration of acetone (6 Pa) was photolysed in the presence of varying pressures of He (0.6–540 hPa), N2 (0.6–133 hPa) or air (0.6–80 hPa) in a T-controlled cell (218–295 K, ±1K). For the experiments involving He and N2, a small amount of O2 (13 Pa) was added to ensure rapid conversion of CH3CO into OH via reaction (3b).
3. Determination of Photolysis Quantum Yields
 The kinetics of OH formation from reaction (3b) were measured as a function of p and T to demonstrate unequivocally that this reaction is the sole source of OH in our system and can be used as a marker for CH3CO [Blitz et al., 2002]. The rate coefficients were in excellent agreement with the literature [Tyndall et al., 1997], and the OH yield from CH3CO+O2, denoted by α, was determined between 1–400 Torr for 213 and 295 K. For 295 K, α = 1 and 0.09 at 1 and 400 Torr, respectively, dropping at higher p due to increased competition from channel (3a) to form the adduct CH3CO(O2).
 The acetone photolysis QY at a chosen λ was obtained as follows. For a given p and T, the rise of OH was recorded by changing the delay time between the pump and probe lasers. OH formed in reaction (3b) is removed slowly either by reaction with acetone or by diffusion. A bi-exponential function was fitted to the OH temporal behaviour, from which a relative value of α[CH3CO]0,λ was obtained, where [CH3CO]0,λ is the initial concentration of CH3CO. The experiment was then repeated (always within 5 minutes) at a photolysis λ of 248 nm, and the OH profile again measured and fitted to obtain α[CH3CO]0,248, ensuring that the spatial overlap between the photolysis and probe lasers at the laser-excitation region was kept identical. The [CH3CO]0 ratio at λ and 248 nm is linked to the acetone photolysis yields via:
where Nλ is the number of photons per laser pulse, (determined from an average of ≥50 shots), and σλ is the T-dependent acetone absorption cross-section, measured between 235 and 298 K by Gierczak et al. . Nλ was chosen to give approximately the same [CH3CO]0 at low p, and hence OH signal, as observed at 248 nm. The OH yield, α, and the degree of quenching of the OH fluorescence for a given p are the same at the two wavelengths, and so cancel in equation (4) above and are not required. The degree of excitation within the initially formed acetyl radical will change with λ, but at the total pressures used here, the acetyl fragment will be relaxed rapidly and α is not expected to change with λ. A measurement of a pair of OH profiles enables a value of to be calculated and, as is independent of p [Gierczak et al., 1998], the p dependence of can conveniently be analysed using a Stern-Volmer relationship:
A plot of versus [M] has an intercept I(λ) = and a gradient proportional to A1, a measure of the ratio kM/kD, where kD and kM are the rate coefficients for dissociation and collisional quenching of the excited singlet state of acetone, respectively. Equation (5) is a good representation of the data for λ < 302 nm at all p. However, for λ ≥ 302 nm an extended form of the Stern-Volmer expression is necessary:
reflecting dissociation and quenching from both the singlet and triplet excited states of acetone. For p > 2000 Pa (15 Torr), the last term is ∼1 and equation (6) is linear in [M], but for lower p the last term becomes dependent upon [M] and equation (6) is non-linear. A more detailed discussion of the photophysics of acetone and the extended Stern-Volmer treatment will be the subject of a future publication (M. A. Blitz et al., manuscript in preparation, 2004). Figure 1 shows Stern-Volmer plots of versus [M] and best-fits of either equation (5) or (6) for a variety of photolysis λ. The ratio increases at higher p due to the reduction in caused by collisional quenching of photoexcited acetone, and the gradient increases sharply above 300 nm as the value of kD decreases rapidly [Emrich and Warneck, 2000]. For λ < 300 nm A1 is small and varies little with p, consistent with previous studies [Gierczak et al., 1998; Warneck, 2001]. The curvature at longer λ is clear, but the excellent fit of either equation (5) or (6) to the data allows accurate extrapolation up to p = 1 atm. At the limit p = 0 the total QY for acetone photolysis is given by:
(the rate of fluorescence is negligible compared to that for dissociation) and hence I(λ) = . Previous studies have shown that CO,λ, the yield of CO via channel (2), is independent of p [Gandini and Hackett, 1977; Horowitz, 1991], and a further discussion of this is given in M. A. Blitz et al. (manuscript in preparation, 2004). At 248 nm, ()[M]=0 = and hence I(λ) = 1, with I(λ) decreasing with increasing λ to a limiting value of at longer λ when CO,λ → 0. A plot of I(λ) versus λ gave = 0.35, and hence CO,248 = 0.65 at 295 K. The value of CO,λ at any other λ was obtained from the variation of I(λ) with λ, and ()[M]=0 from 1 − CO,λ. Using the fits to the Stern-Volmer plots the value of at any p could then be calculated.
4. Quantum Yields as a Function of λ and T
 In total ∼50 Stern-Volmer plots of the type shown in Figure 1, each containing data taken at ∼20 pressures, were assembled for λ = 279–327.5 nm, p = 0.66–540 hPa (He, N2 or air) and T = 295 K, 273 K, 248 K and 218 K. For a given λ and T, the values of I(λ) and A1 for λ < 302 nm, or I(λ) and A2, A3, A4 for λ ≥ 302 nm, obtained from the fits for the 3 different buffer gases, did not show any systematic differences. Figure 2 shows total = + CO,λ for λ = 279–327.5 nm at T = 295K and p = 1000 hPa for the buffer gases He, N2 and air. Although there is some scatter, no systematic difference was found between the collision partners. The precise mechanism for collisional quenching of excited acetone is uncertain (M. A. Blitz et al., manuscript in preparation, 2004), but our data suggest all gases quench with similar efficiency. While the precision of individual experiments is very high, the reproducibility is limited by the difficulty in maintaining constant overlap between three laser beams or the calibration of the laser power over long periods.
Figure 3 shows total at 295 K together with the QYs determined by Gierczak et al. , who either measured the loss of acetone (total, λ < 308 nm only) or the production of CO2 (giving ∼20% less than total, all λ), and by Emrich and Warneck , who measured yields of PAN and hence , for p = 1 atm and 295 K. The agreement is very good for λ < 310 nm, but at longer λ our measurements become significantly smaller. At 320 nm and 1 atm the yields are 0.024 (this work), 0.035 [Gierczak et al., 1998] and 0.06 [Emrich and Warneck, 2000]. Our time-resolved experiments are more direct and sensitive than previous studies using end-product analysis, and as they do not rely on a detailed knowledge of the reaction mechanism linking the photolysis event to observed products, may be less subject to interferences.
 The T dependence of total for λ = 280–327.5 nm is shown in Figure 4 for a constant density of 5 × 1018 molecule cm−3 (150 hPa at 218 K, typical of the UT). The T dependence of total below 295 nm is small, and similar to the scatter in the data, but at longer λ is quite striking, with total (295 K)/total (218 K) ∼4 and ∼20 at 310 nm and 322.5 nm, respectively. The data shown are for He buffer gas, but points recorded for M = N2 and air at selected λ and T showed no systematic difference in QYs. In contrast, at 308 nm Gierczak et al.  found a T-independent yield between 298 and 195 K for a given gas density, as recommended by IUPAC [Atkinson et al., 2002].
5. Parameterisation of Quantum Yields for Atmospheric Modelling
 It was not possible to find a single parameterisation to adequately represent the QY for all λ so two regions were used. For λ = 279–302 nm, equation (5) was fitted globally to all the Stern-Volmer plots at the four temperatures studied, using a least-squares routine to minimise χ2, the goodness of fit. The error for each individual point on the Stern-Volmer plots was used to weight the fit. For λ = 302–327.5 nm, a similar routine was used, except equation (6) for the extended Stern-Volmer analysis was fitted to all the data. The optimised parameterisation is as follows:
For λ = 279–327.5 nm, the CO yield is assumed independent of [M], and is given by:
For λ = 279–302 nm
For λ = 302–327.5 nm
where in all cases the units are [M] (molecule cm−3), λ (nm) and T (K). In the limit of [M] = 0, (λ, [M], T) = 1 − CO(λ, T). The parameterisation is valid for 218–295 K, and for p up to 1000 hPa. The error bars for each parameter represent 95% confidence limits, and are propagated errors from the global fit. The overall error in the parameterised curve is 10–15%.
Figure 4 also shows the variation with λ of the parameterised forms of total and CO for T = 295, 273, 248 and 218 K, and for constant [M] = 5 × 1018 molecule cm−3. CO is small in this region, and decreases rapidly with T. The slight discontinuity in the parameterised curve at 302 nm is caused by the switch-over from equations (10)–(11) to describe
6. Variation of Acetone Photolysis With Altitude
 The atmospheric photolysis rate of acetone is given by:
where F is the actinic photon flux that depends upon altitude, z, and solar zenith angle, χ. The CiTTyCAT box model [Evans et al., 2000] was used to calculate the diurnally averaged acetone loss rates due to reaction with OH and photolysis over a range of 13 levels in a column from 1000 hPa to 150 hPa. Figure 5 shows J(New) calculated for 21 June at latitude 50°N using the parameterised form of total(λ, [M], T) given above, together with the rate of loss of acetone by reaction with OH, using rate constants recommended by IUPAC. Temperature was taken from the US Standard Atmosphere (1976), O3 and CH4 were from a 2-D model [Law and Pyle, 1993] and H2O was the saturated vapour pressure at each model level. The acetone absorption cross sections (σ(λ, T)) were taken from Gierczak et al. . Figure 5 also shows J(Recommended) calculated using acetone photolysis yields recommended by IUPAC and the ratio J(Recommended)/J(New). The photolysis rate is significantly slower at all z if the new QY are used, being a factor of 3.5 lower in the UT/lower stratosphere region. The effect is more pronounced at higher z because the new QY drop off sharply at the lower T (see Figure 4). The altitude at which photolysis begins to dominate the acetone loss increases significantly when the new QY are used. The reduction in J using the new QY is slightly less pronounced at 0°N for overhead sun (21 March, factor of 2.65 at 200 hPa), but for 50°N in mid-winter (21 December), J is reduced by a factor of 10 at 200 hPa, and a factor of 3 at the surface (see auxiliary material). Further atmospheric implications of the new T-dependent QY for acetone will be discussed in Arnold .
 A new method to monitor CH3CO has been used to measure the QY for both photodissociation channels of acetone between 279–327.5 nm as a function of T and p. In contrast to previous studies, the QY was observed to decrease significantly at lower T, being significantly smaller (e.g., for 320 nm, 218 K, 150 hPa, by a factor of ∼20) than currently recommended values. From data recorded at ∼1000 combinations of T, p and λ, two parameterised forms of the photodissociation QY were determined for λ below and above 302 nm, and used to calculate acetone photolysis rates as a function of altitude. The new QY show photolysis to be a less important sink for acetone, compared to recommended values, especially in the colder UT region, with important implications for its budget and associated chemistry.
 MAB, DEH and MJP thank the NERC UTLS-O3 programme (award GST/02/2428) for funding this work.