Pressure dependent CO and CO₂ quantum yields in the laser pulse photolysis of acetone at 248 nm and T = 298 K have been measured directly using quantitative infrared diode laser absorption. The experiments cover the pressure range from 50 to 900 mbar. It is found that the quantum yields show a significant dependence on total pressure, with Φ_CO decreasing from around 0.5 at 20 mbar to approximately 0.3 at 900 mbar. The corresponding CO₂ yields as observed when O₂ exists in the reaction mixture, exhibit exactly the opposite behaviour. For the sum of both a value of 1.05^+0.02_−0.05 independent of pressure is obtained, showing that the sum of (Φ_CO + Φ_CO2) is a measure for the primary quantum yield in the photolysis of acetone. In addition, CO quantum yields and corresponding pressure dependences were measured in experiments using different bath gases including He, Ar, Kr, SF₆, and O₂ as third body colliders.

The theoretical framework in which we discuss these data is based on our previous findings that the pressure dependence of the CO yield is a consequence of a stepwise fragmentation mechanism during which acetone decomposes initially into methyl and a vibrationally ‘hot’ acetyl radical, with the latter being able to decompose promptly into methyl plus CO. The pressure dependence of the CO yield then originates from the second step and is modelled quantitatively via statistical dynamical calculations using a combination of RRKM theory with a time-dependent master equation (ME) approach.

From a comparison of experiment with theory the amount of excess energy in the vibrationally hot acetyl radicals (E* ≈ 65 kJ mol⁻¹) as well as the characteristic collision parameters for interaction of acetyl with the different bath gases were derived. Values of 90, 280, 310, 545, 550 and 1800 cm⁻¹ for the average energy transferred per downward collision for the bath gases He, Ar, Kr, O₂, N₂, and SF₆, respectively, are obtained. The calculations also considered different models for the energy transfer kernel P(E,E′) and best fits were obtained with a ρ-weighted exponential down model.