͓S0021-8979͑98͒01704-6͔The exchange bias effect, which results from exchange coupling between adjacent ferromagnetic ͑F͒ and antiferromagnetic ͑AF͒ layers, which are either deposited in a magnetic field or cooled down in a magnetic field after heating above the Néel temperature, is a well-established phenomenon. 1 It manifests itself in a shift of the hysteresis loop along the axis of the applied field, the so-called exchange bias field, H eb , and it is often described as an inplane unidirectional anisotropy. Although exchange bias systems have been extensively studied, 2-5 the microscopic origin of the exchange bias effect still remains unclear. The most advanced models propose the formation of magnetic domains in the AF layer causing a macroscopic exchange coupling strength of the experimentally observed size, 6,7 which is two orders of magnitude smaller than the atomic F-AF exchange coupling strength. Key issues to be tested experimentally are, ͑i͒ to prove the assumption of large local spatial variations of the F-AF coupling, and ͑ii͒ apart from the detection of the unidirectional anisotropy, to search for possible additional anisotropy contributions of higher order induced by the exchange coupling mechanism.We have chosen to study the ͑110͒-oriented system of Ni 80 Fe 20 /Fe 50 Mn 50 bilayers, since the lowest order in-plane anisotropy contributions, which are of unidirectional, uniaxial, and fourfold symmetry, can be separately determined for simple symmetry reasons. An unexpected large uniaxial anisotropy contribution caused by the FeMn coverage has already been reported for this orientation. 5,6 We have used Brillouin light scattering ͑BLS͒ from dipolar spin waves ͑Damon-Eshbach modes͒ propagating in the F film parallel to the surface to extract the anisotropy constants. 8 Two molecular beam epitaxy ͑MBE͒ grown staircaseshaped samples have been prepared on Cu͑110͒ single crystal substrates. The samples consist of Ni 80 Fe 20 films with thicknesses of 27, 34, 50, and 70 Å for the first, and 18, 24, 37, and 90 Å for the second staircase-shaped sample. Half of each sample was covered by an 80 Å thick Fe 50 Mn 50 layer. At this thickness the exchange bias effect is saturated. 4 Finally, both samples were covered with a 20 and 30 Å thick protective Au layer, respectively. The samples were characterized using low energy electron diffraction, and their chemical cleanliness was checked by Auger electron spectroscopy. An analysis of the process parameters revealed that the residual oxygen content in the gas used for sputter cleaning was significantly larger during preparation of the second sample compared to the first. Evaporation on a slightly contaminated surface can therefore not be excluded for the second sample. During growth a magnetic field of Ϸ250 Oe was applied in the film plane along the ͓110͔ direction to induce exchange biasing. We note that the ͑110͒-oriented Fe 50 Mn 50 surface is an uncompensated plane with a resultant in-plane magnetization along the Ϯ͓001͔ directions. 4 The Brillouin light scatt...