While the principal features of the exchange bias between a ferromagnet and an antiferromagnet are believed to be understood, a quantitative description is still lacking. We show that interface spin disorder is the main reason for the discrepancy of model calculations versus experimental results. Taking into account spin disorder at the interface between the ferromagnet and the antiferromagnet by modifying the well known Meiklejohn and Bean model, an almost perfect agreement can be reached. As an example this is demonstrated for the CoFe/IrMn exchange biased bilayer by analyzing the azimuthal dependence of magnetic hysteresis loops from MOKE measurements. Both, exchange bias and coercive fields for the complete 360• angular range are reproduced by our model. The exchange bias system refers to the shift of the ferromagnetic (F) hysteresis loop to positive or negative values when the F system is in contact with an antiferromagnetic (AF) system and cooled in an applied magnetic field through the Néel temperature of the AF system. The exchange bias (EB) phenomenon is associated with the interfacial exchange coupling between ferromagnetic and antiferromagnetic spin structures, resulting in a unidirectional magnetic anisotropy [1]. While the unidirectional anisotropy was successfully introduced by Meiklejohn and Bean (M&B), the origin of the enhanced coercive field is yet not well understood. The details of the EB effect depend crucially on the AF/F combination chosen and on the structure and thickness of the films [2,3]. However, some characteristic features apply to most systems: (1) H EB and H c increase as the system is cooled in an applied magnetic field below the blocking temperature T B ≤ T N of the AF layer, where T N is the Néel temperature of the AF layer; (2) the magnetization reversal can be different for the ascending and descending part of the hysteresis loop [4,5,6,7,8]; (3) thermal relaxation effects of H EB and H c indicate that a stable magnetic state is reached only at very low temperatures [9,10,11].Several theoretical models have been developed for describing possible mechanisms of the EB effect, including domain formation in the AF layer with domain walls perpendicular to the AF/F interface [12], creation of uncompensated excess AF spins at the interface [13], or the formation of domain walls in the AF layer parallel to the interface [14,15]. Another approach is the consideration of diluted antiferromagnets in an exchange field. In the work described in Ref. [16,17,18] the discussion about compensated versus uncompensated interfacial spins is replaced by a discussion of net magnetic moments within the antiferromagnetic layer. Depending on the complexity of the system, the models can explain some but not * florin.radu@bessy.de † Present address: BESSY GmbH, Albert-Einstein-Str. 15, D-12489 Berlin, Germany.all features of experimental hysteresis loops. Here we provide a new model which can describe all features, including the azimuthal dependence of H EB , H c , and the AF thickness dependence. In ...