The Origin for Training Effects in Exchange Bias Systems: Frustration and Multiple Anisotropy Axes at the Interface

摘要

The interaction of an antiferromagnet with a ferromagnet can establish a unidirectional anisotropy, which is referred to as an exchange bias. Many exchange bias systems show an instability that, after initial field cooling, results in irreversible changes between subsequent hysteresis loops. These field-training effects are suspected to be due to irreversible changes in the magnetic microstructure of the antiferromagnet, but a comprehensive theoretical understanding is still missing. In general one can distinguish two types of training effects. Generally there is an initial dramatic change of the hysteresis loop after the first reversal, and a more gradual change with diminishing exchange bias for subsequent loops after the second hysteresis loop. I will present numerical simulations based on a simple coherent rotation model, which imply that the initial irreversibilities observed between the first and second hysteresis loop are driven by the symmetry of the antiferromagnet [1]. The availability of more than one easy axis in a high symmetry antiferromagnet can give rise to a spin-flop transition during the first field reversal, after which the antiferromagnetic spins are trapped in a metastable configuration. As can be seen in Fig. 1, the calculations show a very pronounced training effect for the case of antiferromagnets with two anisotropy axes, while no training is observed for antiferromagnets with only one anisotropy axis. This explains quite naturally why training effects are only observed for exchange bias systems with high symmetry antiferromagnets, while they are absent for antiferromagnets with uniaxial anisotropy. This simple and universal model reproduces many of the experimentally observed initial training effects in exchange bias systems, such as the change in loop shape and the different domain structures at the two coercive fields during the first hysteresis loop.