Since Iwamoto's discovery of an unusually high activity of Cu/zeolites in NO decomposition, [1] there has been an ongoing experimental and theoretical quest to understand the details of the mechanism of this process. Despite this tremendous effort, there are still a number of unsettled questions, including the uncertainty about the nature of the reaction intermediates and about the structure of the active site; [2] either the involvement of only a single extra-framework Cu cation [3][4][5][6][7][8] or the concerted effect of two nearby Cu cations, the Cu pair, [9][10][11][12][13][14] was proposed. From the vast amount of relevant experimental data available today, it is clear that the activity of the Cu/zeolite catalysts depends on 1) zeolite topology, 2) zeolite composition (Si/Al ratio and Cu loading), and 3) Cu-exchange procedure, including the sample pretreatment. [4,6,8,12,13,15] However, the understanding of the relationship between active site structure (the metal coordination and localization) and catalytic activity is far from complete; it is our goal to increase our knowledge in this respect.The lack of direct experimental evidence about the transition-state structures and about the details of the reaction mechanism justifies the use of quantum chemistry to provide the missing details. However, a reliable description of the reactions catalyzed by transition metals in a complex environment (such as zeolites) represents a major challenge for contemporary computational chemistry since it requires the use of a model that realistically represents the active site (including its environment) and the use of a method that can consistently describe the electronic structure of the system along the reaction path (see the Methods section). Many reaction steps possibly involved in the direct NO decomposition on Cu/zeolites have been investigated previously by employing various models and methods. [16][17][18][19][20] Several reaction paths proposed in the literature have been compiled (Scheme 1) along with the reaction energies and activation barriers (where available) reported for the elementary reaction steps. Since these energies were obtained at various levels of theory using different types of models for the active site, a direct comparison should be taken with some caution. For example, the reaction energies of À107 and À173 kJ mol À1 were reported for reaction step D based on the BP86 functional (using a 1-T cluster model) and based on the B3LYP functional (using a 3-T cluster model), respectively. [16,17]