Another way of obtaining unequal coercivities in two layers of a sandwich is simply to make them out of different ferromagnets. Previous workers employed this strategy with NiFe-Cu-Co sandwiches.[7] In the present case, Fe-Cu-Co structures have been grown on glass or Si substrates using electron-beam evaporation. Each sandwich was capped with a Cu protection layer about 30Å thick. The specifics of the deposition and basic characterization have been discussed previously.[8] Under appropriate growth conditions, as detailed below, the Co films have a coercive field of about 200 Oe, while the Fe films have a weakly growth-dependent coercive field of about 30 Oe. The B-H loop of a properly prepared Fe-Cu-Co sandwich is shown as the tFe=40Å curve in Figure 1a). The identification of the higher coercivity part of the B-H loop with the Co film is supported by a Co-layer-thickness study, which showed that the moment associated with the higher-coercivity part of the loop increases linearly with Co thickness.[8]
As a result of the sequential switching of the two ferromagnetic layers, the magnetization curves of these structures display a plateau of nearly constant moment about 150 Oe wide. In this plateau region, the moments of the Fe and Co films are macroscopically antialigned. The resistance of the structure is enhanced in the plateau region, as demonstrated by the tFe=40Å curve in Figure 1b). The enhanced resistance associated with the antialignment of the moments has been called the spin-valve effect.[5]
The solid curves in Figures 1a) and 1b) show the corresponding data for a sample with a thinner Fe layer, 16Å thick. Figure 1a) shows that reducing the thickness of the Fe layer has little effect on the magnetic properties of the Fe, but has the unexpected consequence of greatly reducing the coercivity of the Co layer. Previous studies of the effect of varying the Co layer thickness and growth temperature have shown that the coercivity of the Co layer is very sensitive to both parameters.[3,4,8] Nonetheless, it is surprising that the Fe layer thickness should impact the Co coercivity. Since one expects that the length scale of the grains in a polycrystalline film should be approximately the same as the layer thickness, this result may be an indication that the coercivity of the Co film may be strongly dependent on the grain size. Another possibility is that the Co growth may variously be fcc or hcp depending on the deposition conditions. X-ray diffraction and electron microscopy studies are underway to investigate these hypotheses.
Previous groups of investigators have observed antiferromagnetic interlayer coupling in Co-Cu and Fe-Cu superlattices.[9] The coupling in these superlattices has the form of an exponentially decaying sinusoid versus Cu thickness. Ferromagnetic resonance (FMR) measurements of the Fe-Cu-Co sandwiches have been performed to look for signs of a similar interlayer coupling. No antiferromagnetic coupling is seen over a wide range of Cu thicknesses, although the possibility of ferromagnetic interlayer coupling cannot be ruled out. The lack of antiferromagnetic coupling in these samples is not understood, since some of the Cu layers fall in the thickness range where it has been found earlier.[9] For low Cu thicknesses only one resonance is observed, rather than the two distinct resonances that would be expected for completely decoupled Fe and Cu films. Even samples with plateaus in their magnetization curves sometimes have only a single FMR line. The observation of only one peak is suggestive of a small degree of coupling between the Fe and Co. Ferromagnetic interlayer coupling may be responsible for the square shape of the magnetic hysteresis loops observed for sandwiches with thick Fe and Co layers, but with tCu < 30Å. On the other hand, since the Co coercivity is obviously very sensitive to the growth mode, interlayer coupling cannot be definitely identified as the cause.
The magnetoresistance in the Fe-Cu-Co sandwiches has two components, the spin-valve part, which is unique to multilayers, and the anisotropic magnetoresistance (AMR), which is observed in both multilayers and single ferromagnetic films. The AMR is entirely responsible for the magnetoresistance of the tFe = 16Å sample since, as Figure 1a) shows, there is no applied field region where the Fe and Co moments are antialigned, and thus no spin-valve effect. Because the AMR is typically about a tenth the size of the spin-valve MR in these samples, the magnetoresistance of the tFe=16Å sandwich is much smaller than that of the tFe=40Å sandwich. By measuring the angular dependence of the resistance at a large constant applied field, the two components of the MR can be separated out even for samples where both contribute.[5,6] The spin-valve MR and AMR thus obtained are shown in Figure 2 for a series of Fe-Cu-Co sandwiches with varying Fe thickness. The AMR increases linearly with increasing Fe thickness, which is the straightforward consequence of the increasing fraction of ferromagnetic material in the sample.
The broad peak in the spin-valve MR versus Fe thickness has more physical content. According to a semiclassical model of spin-dependent transport in these systems by Barnas et al.,[3] a maximum in the spin-valve magnetoresistance versus ferromagnetic layer thickness is indicative of a "bulk" origin for the spin-dependent scattering. In this context, "bulk" means that the microscopic scattering events which give rise to the spin-valve effect occur primarily within the ferromagnetic layers, not at the ferromagnet-paramagnet interface. The Fe-Cu-Co samples have a maximum near 60Å in the spin- valve MR versus Co thickness as well as versus Fe thickness.[8] In addition, Dieny et al. have observed similar peaks in the value of the spin-valve MR versus ferromagnetic layer thickness in NiFe-noble metal-transition metal sandwiches.[5] The occurrence of a peak in the spin-valve MR versus ferromagnetic layer thickness thus appears to be a general result in uncoupled sandwiches, consistent with the model of Barnas et al.