We have investigated the processes which lead to the deactivation of active chlorine (ClO+2Cl2O2) and the formation of ClONO2 in and around the Arctic polar vortex. By using a range of three-dimensional and trajectory models, forced by meteorological analyses for the 1991/1992 winter, we have been able to separate processes associated with the model formulation and resolution, from true chemical processes. The rate of ClO deactivation can be overestimated in a three-dimensional model due to numerical diffusion from a poor transport scheme or to low resolution. In our experiments we found that vertical resolution is at least as important as horizontal resolution. The use of isentropic levels reduced the spurious vertical mixing by providing a true separation between horizontal and vertical motion. Increasing the horizontal resolution of the model reduced mixing and produced a sharper, narrower ClONO2 collar region with slightly higher maximum values. The domain-filling trajectories produce a ClONO2 collar structure with no mixing between parcels. These results show that it is in situ chemical deactivation of ClO and not mixing which produces the ClONO2 collar. Results from the high-resolution isentropic simulations and domain-filling trajectories have been compared with Cryogenic Limb Array Etalon Spectrometer data and diagnosed for the chemical cause of ClO deactivation. The models generally reproduce the observed features of the collar region but underestimate the highest ClONO2 mixing ratios. As the polar lower stratosphere is heavily denoxified the cause of the ClONO2 recovery is release of NO2 from HNO3. For this process, reaction of HNO3 with OH can be as important as photolysis at high solar zenith angles.¿ 1997 American Geophysical Union