Discussion and conclusions

In this section, we summarize the results from the two different groups of samples that were presented in the last two sections. Figure 3. 10 and Figure 3. 7 are plotted in order to compare these clearly. We don’t discuss He of the sample with bottom layer with integer number of atomic layers here, the details of this discussion will be presented in sec. 3.3.

First we should clarify some different phenomena in the results. a), there should be two antiferromagnetic maxima of the RKKY-type coupling in trilayer samples in the studied Mn thickness range, i.e., at around 2.5 ML and 8.2 ML Mn thickness. The first maximum was missing in the 10.0 ML bottom Co layer trilayer samples, and the coupling strength was weaker than we

oxidization of the films could influence the roughness. b), the coercivity of the 10.5 ML bottom Co layer bilayer and trilayer samples decreases quickly after the maximum, whereas the 8.5 ML case shows a slow decrease after the peak maximum. We explain the latter by a slithtly inhomogeneous thickness of the bottom layer. Even a difference of 0.1 ML Mn thickness will influence the coercivity of bilayer and trilayer a lot. This will be discussed in a latter section.

Figure 3. 10 Coercivity Hc (top), exchange bias field He (middle), and remanent Kerr signal of 10.0 ML (black squares) and 10.5 ML (red stars) bottom Co bilayers and trilayers as a function of Mn thickness.

Now we can list some results from the above samples.

1. tAFM: Bilayers with Co film with half-filled topmost atomic layer show smaller values of tAFM than in the filled case, i.e., tAFM = 3.4 ML and 4.1 ML for 10.5 ML and 8.5 ML Co thickness, respectively, while tAFM = 4.3 ML and 4.5 ML for 10.0 ML and 8.0 ML Co thickness, respectively. All trilayers have a larger tAFM than the corresponding bilayer, and the difference in tAFM between the bilayers and trilayers is larger for filled than for half-filled Co bottom layers (Table 3.1).

2. Hc: The coercivity of the bilayers with completely filled topmost atomic layer of the Co film shows two times an increase with increasing Mn thickness, i.e., at 4 ML - 7 ML and at 7 ML -10 ML, and then keeps a constant value (~50 mT for 10.0 ML and ~80 mT for 8.0 ML Co layer). Hc of the trilayers displays a similar behavior. Due to the larger tAFM

value, these two increasing slopes phenomena appear at thicker Mn layers. Furthermore, the loops of the trilayers show a different behavior due to antiferromagnetic RKKY-type coupling, and the decoupled magnetization reversal of the two Co layers appears at higher Mn thicknesses. The second difference between the two samples that can be related to the different interface roughness is that the Hc of the 10.5 ML bottom Co layer sample shows a sharp maximum at around 5-6 ML Mn thickness and then strikingly decreases to nearly half the value of this maximum (~10 mT for 10.5 ML and ~20 mT for 8.5 ML Co layer).

This behavior is consistent with data of a wedged Mn/20 ML Co/Cu(001) bilayer,²⁰ and may thus be explained by assuming a similar interface roughness of our 10.5 ML and the 20 ML Co layers in Ref. 20. A dependence of Hc on the AFM/FM interface roughness has been observed before in Mn/Co bilayers, and has been attributed to a biquadratic exchange interaction between FM and AFM spins due to roughness.^59,60 For the trilayer case, a major difference to the case of the atomically filled bottom Co layer is that magnetization loops with only one step are observed for all Mn thicknesses under study. This could be due to the lower coercivity of the bottom Co layer, as seen from the Mn/10.5 ML Co or Mn/8.5 ML Co bilayers. It is more similar to the coercivity of the top Co layer, which could lead to a merging of the magnetization reversals, possibly also mediated by stray fields from propagating domain walls. We will talk about these later in wedged Co/Mn/10.0 ML and 10.5 ML Co trilayers in sec. 3.6.2. Furthermore, above 8 ML Mn thickness the Hc of 10 ML Co/Mn/10.5 ML Co shows an oscillation with a period of around 1 ML Mn thickness, which can be attributed to the layer-by-layer growth of Mn on Co. We will discuss this further down in connection with the behavior of the 10 ML Co/Mn/10.0 ML Co trilayers at higher Mn thicknesses in sec. 3.5.

3. He: The onset Mn thickness for the appearance of the exchange bias field He should be higher than that of the maximum of the coercivity. Unfortunately, the onset thickness of

He could not be determined from the above results. One reason is that the coercivity dramatically decreases after the maximum with increasing Mn thickness, and the MOKE signal averages different thicknesses of Mn due to the large Mn slope. This will lead to an big error, such that the data for He become unreliable. Another error arises from the RKKY coupling. Errors also come from the not fully separated two magnetization reversals. One reliable conclusion about He is that its magnitude for the trilayers is always higher than in the bilayers, and it seems that the intensity decreasing with top Co thickness increasing.

Bottom layer

8.0 ML Co 8.5 ML Co 10.0 ML Co 10.5 ML Co Bilayer

tAFM (ML) 4.5 4.1 4.3 3.4

Trilayer

tAFM (ML) 5.4 4.8 5.2 3.9

ΔtAFM (ML) 0.9 0.7 0.9 0.5

Table 3.1 The onset thickness (tAFM) for AFM order in Mn/Co bilayer, and Co/Mn/Co trilayer films.