Introduction

Since the giant magnetoresistance (GMR) of Fe/Cr multilayers was discovered by Baibich et al. in 1989,[1] scientists around the world have made great strides towards understanding the fundamental physics of this phenomenon. There have been dozens of experimental studies. Theorists have studied the magnitude of the magnetoresistance as a function of layer thickness,[2,3,4] interface morphology,[2] temperature,[5,6] configuration of the magnetic moments in the layers,[7] and impurity scattering,[8] all with some degree of success. In these models, the microscopic parameters which describe the detailed electronic potential of the metallic layers and their interfaces have often been lumped into phenomenological spin-dependent mean-free paths or transmission coefficients. This approach has been successful in describing the layer-thickness dependence and temperature dependence of the magnetoresistance, but it does not explain why some ferromagnet/paramagnet multilayers intrinsically have larger magnetoresistance than others.

Recently a pair of theoretical attempts have been made to relate the size of the magnetoresistance to band-structure parameters.[4,9] These efforts took very different approaches, one emphasing the density of states in the ferromagnetic layers,[4] and another emphasizing band offsets at the ferromagnet/paramagnet interface.[9] The question of which microscopic properties of the ferromagnet and paramagnet affect the size of the GMR in a given multilayer system remains an open one although it is of great importance both for fundamental reasons and for applications. The intellectual problem can be succintly put forth by noting that the magnetoresistance is large in the Fe/Cr[1] and Co/Cu[10] systems, but small in the Co/Cr[11] and Fe/Cu[12] systems. Another suggestive series of experiments has been performed by B. Dieny et al. on NiFe/Cu/ferromagnet sandwiches[13,14] where the type of ferromagnet in the second magnetic layer was varied. In these studies, a sandwich with a second NiFe layer has a smaller GMR than one with a Co layer, while the GMR of a sandwich with a Ni layer is smaller yet. Most recently, Saito and coworkers have found that the GMR is larger in Co₉Fe/Cu multilayers than in Co/Cu multilayers or Co₃Fe /Cu multilayers.[15]

Practical issues related to crystal structure and growth morphology tend to complicate the interpretation of the multilayer and sandwich data. After all, Cu and Co are typically fcc in the sputtered superlattice structures, whereas Fe and Cr are typically bcc. Therefore one might suppose that the GMR is larger in the all-bcc and all-fcc superlattices, and smaller in those with mixed crystal structures. However, this hypothesis does nothing to explain the extensive data of Dieny and coworkers[13,14] since their Cu, NiFe, Co and Ni are all presumably fcc. In the Co_xFe_1-x/Cu multilayer work of Saito et al.,[15] both fcc and hcp Co were observed in all 3 types of samples.

The question of the maximum size of the magnetoresistance in a given ferromagnet/paramagnet system deserves more attention from both theorists and experimentalists. ("Maximum size" here means the room temperature magnitude of the magnetoresistance at the paramagnet and ferromagnet thicknesses which give the largest effect.) Ideally one would like to eliminate any issues about crystal structure and growth morphology in order to define clearly the relationship between band-structure effects and the GMR. One way of minimizing such concerns is to study the Co_xFe_1-x alloy system, which can be grown epitaxially in the bcc crystal structure throughout the range of composition.[16] In order to explore a possible link between electronic structure and GMR, a series of Fe/Ag/Co_xFe_1-x/Ag sandwiches have been grown with different Co_xFe_1-x stoichiometries, and their transport properties have been studied.