The flash photolysis/UV absorption technique has been used to study the self-reaction of BrO radicals over the temperature range 222-298 K and the pressure range 100-760 Torr of N 2 or O 2 . Two chemical sources of BrO radicals were used: photolysis of Br 2 in the presence of excess ozone and photolysis of O 2 in the presence of excess Br 2 . The overall rate constant, k 1 , for the BrO self-reaction (defined by -d[BrO]/dt ) 2k 1 [BrO] 2 ) was found to be temperature and pressure independent at T g 250 K, with k 1 ) (2.88 ( 0.20) × 10 -12 cm 3 molecule -1 s -1 . At temperatures below 250 K, k 1 was found to be pressure dependent, due to the emergence of a new termolecular channel of the BrO self-reaction 1c, -1c forming the BrO dimer, Br 2 O 2 (BrO + BrO + M h Br 2 O 2 + M). Channel-specific rate constants were determined for the two bimolecular channels of the BrO self-reaction above 250 K, giving for (1a) (BrO + BrO f 2Br + O 2 ) k 1a ) (5.31 ( 1.17) × 10 -12 exp{(-211 ( 59)/T} cm 3 molecule -1 s -1 and for (1b) (BrO + BrO f Br 2 + O 2 ) k 1b ) (1.13 ( 0.47) × 10 -14 exp{(983 ( 111)/T} cm 3 molecule -1 s -1 . Below 250 K, the overall rate coefficient of the two bimolecular channels is reduced as the dimer forming channel emerges. At 235 and 222 K, rate constants for the formation (k 1c ) and decomposition (k -1c ) of Br 2 O 2 were recorded. Using the values for K 1c , ∆H r for reaction 1c was estimated as -58.6 ( 0.1 kJ mol -1 . A UV absorption spectrum attributed to Br 2 O 2 was also recorded over the wavelength range 300-390 nm. The cross section of the smooth Br 2 O 2 spectrum was found to be 1.2 × 10 -17 cm 2 molecule -1 at 320 nm. These results are rationalized in terms of a mechanism of the BrO self-reaction that shows competition, at low temperatures, between collisional quenching and unimolecular dissociation of an excited BrOOBr* intermediate. The rate constant for the reaction of oxygen atoms with molecular bromine was also determined in the course of these experiments [O + Br 2 f BrO + Br (5)], giving k 5 ) (5.12 ( 1.86) × 10 -13 exp{(989 ( 91)/T} cm 3 molecule -1 s -1 . All errors are 1σ. The atmospheric implications of these results are discussed.