The effect of spin-orbit coupling on O(P-3)+O-2((3)Sigma(g)(-)) collisions is investigated for J=0 using time-dependent wave packets. The probability of forming O-3 complexes, which is important for understanding the atom exchange reaction mechanism, is calculated in two different ways. The first approach follows the standard treatment in that only the reactive ground electronic state is included. In the second approach all 27 states correlating with O(P-3)+O-2((3)Sigma(g)(-)) and the nonadiabatic transitions induced by spin-orbit coupling are taken into account; all the excited electronic states are repulsive and thus do not lead to complex formation if nonadiabatic transitions are neglected. The required nine diabatic potential energy surfaces (not including spin-orbit coupling) for the electronic states 1 (s)A', 2 (s)A', and (s)A" (s=1, 3, and 5) are constructed by high-level electronic structure calculations in the asymptotic O+O-2 channel with the O-2 bond length being fixed. The two sets of calculations show that spin-orbit coupling is not an important effect. The probability for forming ozone complexes when the oxygen atom is initially in the excited fine structure state O(P-3(j=1)) state is only 10% of that for the lowest state O(P-3(j=2)), and it is below 1% for O(P-3(j=0)). The single-surface calculation, with the excited states phenomenologically taken into account by a statistical factor, gives a rather accurate value for the thermally averaged complex formation rate coefficient. (C) 2003 American Institute of Physics.