One puzzling aspect of the interlayer exchange coupling in the Fe/Si system has been the dependence of its strength on the number of bilayers in the multilayer. This trend is illustrated in Figure 11, where magnetization curves for (Fe40Å/Si14Å)xN multilayers with 2, 12 and 25 repeats are displayed. (The 2-repeat multilayer is just an Fe/Si/Fe trilayer.) Although the trilayer has magnetic properties like bulk Fe, the 25-repeat multilayer data has a magnetization curve similar to the 40-repeat multilayer data shown above. The magnetization curve for the 12-repeat multilayer falls in between that for the thicker and thinner films. Evidence for AF coupling which is stronger near the top of an Fe/Si multilayer than near the substrate has previously been described by Fullerton et al.[44] Presumably the increase of coupling with bilayer-number is a manifestation of the same phenomenon. The interlayer coupling in Co/Cu multilayers also increases with the number of bilayer periods up to about 25 bilayers.[45]
One would not expect interlayer coupling that is quantum-mechanical in nature to be affected much by total film thickness. The unusual thickness dependence therefore raises the question of whether there is quantum-mechanical coupling at all, or whether some other mechanism might determine the shape of the magnetization curves. Disordered magnetic materials such as small amorphous Fe particles can have low remanence and high saturation fields without any layering at all. The magnetization curves of these Fe particles are in fact quite similar to those of the Fe/Si multilayers.[46] This resemblance might lead to speculation that the topmost Fe layers in Fe/Si multilayers are discontinuous and that the magnetic properties are dominated by particle shape. However, the existence of half-order peaks in polarized neutron reflectometry measurements in the IBS-grown Fe/Si multilayers[26] and the magnetron-sputtered multilayers[44] gives unambiguous evidence that the magnetic properties are due to magnetic order rather than structural disorder. In addition, TEM pictures such as Figure 6 show that the Fe layers are continuous in films with both high and low saturation fields.
How then does the number of bilayer periods influence the AF coupling strength? It has been suggested that the difference between thin and thick multilayers grown at nominal RT is that the substrates of thick multilayers have time to rise to a higher temperature (about +60°C for our system) during the longer growth.[36] This idea seems reasonable in light of the larger coupling in samples grown on heated substrates as described above. In order to investigate this idea, a (Fe100Å/Si14Å/Fe100Å) film was grown on glass at +200°C. The magnetization curve for this film is shown in Figure 12. Also shown in this figure are data for a (Fe100Å/Si14Å/Fe100Å) trilayer deposited at nominal RT and for a (Fe100Å/Si14Å/Fe100Å) trilayer deposited at +200°C, both grown on a 500Å-thick a-Si buffer. The trilayer deposited directly on glass at elevated temperature has only slightly less remanence and higher saturation field than the trilayer grown at RT whose data are shown in Figure 11. This result implies that it is not substrate temperature alone which causes bilayer-number effects. The magnetization curves of the trilayers grown on buffer layers, on the other hand, look much more like typical t Fe = 40Å 40-repeat multilayer results. An epitaxial (Fe100Å/Si14Å/Fe100Å) trilayer grown directly on an MgO(001) at +200°C substrate also has strong AF coupling (data not shown). Undoubtedly the strong AF coupling of the trilayer grown directly on the MgO is due the superior surface quality of the single-crystal substrate.
The take-away lesson from all of these results is that substrate roughness is probably responsible for the reduced interlayer coupling in (Fe40Å/Si14Å) multilayers with a low number of bilayers. Conformal growth may propagate this roughness up from the substrate into the multilayer. Parkin et al. have found that the interlayer coupling in MBE-grown Co/Cu multilayers is very sensitive to the substrate and the buffer layer type, perhaps due to pinholes through the Cu layers.[47] Presumably thin Fe layers grown directly on glass are so wavy that pinhole and magnetostatic coupling dominate the interlayer interactions for the first few bilayer periods. Recent calculations show that magnetostatic effects associated with propagating roughness can give interlayer ferromagnetic coupling of the same order of magnitude as the coupling derived from quantum-well effects.[38] Ongoing polarized neutron reflectivity experiments may give more information on the variation of the coupling with position in the thicker multilayers.[26]