Absence of mean-free-path effects in the current-perpendicular-to-plane magnetoresistance of magnetic multilayers

The series resistor and Valet-Fert models widely used to describe the current-perpendicular-to-plane (CPP) magnetoresistances of ferromagnetic/nonmagnetic (F/N) metal multilayers were recently claimed to be valid only for mean-free paths shorter than layer thicknesses; otherwise the mean-free path was claimed to be an important length scale in the CPP magnetoresistance (MR). This claim was based on observations of differences in the CPP MR's, after the samples were taken to above their saturation magnetic fields, of two different kinds of multilayers involving Co and Cu: interleaved [Co(6)/Cu(20)/Co(1)/Cu(20)](N) and separated [Co(6)/Cu(20)](N)[Co(1)/Cu(20)](N), with N repeats and thicknesses in nm. The maximum CPP MR's of separated samples were only about half as large as those for interleaved ones. In two short papers, we provided experimental evidence that mean-free paths are not important length scales in the CPP MR by showing that the differences in CPP MR's upon which the above claim was made did not change when the mean-free paths in the N and F layers were reduced from well above to well below their layer thicknesses. We ascribed part of the behaviors of interest to finite spin-memory loss (spin flipping) in the F and N metals, and proposed that the rest might be due to spin nips at F/N interfaces. In the present paper we (a) present further experimental evidence against mean-free-path effects, (b) provide details of the calculations we use to analyze the data, and (c) use measurements of magnetization and polarized neutron reflectivity to show that the differences in CPP MR are not due to spurious differences in magnetic structure between interleaved and separated multilayers, but only to the differences in the relative magnetic alignment of adjacent layers. Additional evidence for this last point is our observation that the CPP MR's of separated samples in their as-prepared states are as large as those of the equivalent interleaved samples after they are taken to above their saturation fields. We show that similar differences between interleaved and separated data appear also in the current-in-plane (CIP) MR's and when the Cu is replaced by Ag.