Magnetic properties are strongly dependent on crystal structure of the material. Thus, the crystal properties of these trilayers films were analyzed in order to identify the causes of the variation in magnetic properties.

Figure 1 shows the GIXRD patterns of as-deposited [Fe (30 Å)/Si (15 Å - 30 Å)/Fe (30 Å)] trilayer samples. All the patterns exhibited only single peak and it corresponds to the diffraction from α-Fe (110). The absence of any Si peak indicates the amorphous nature of deposited Si film in all the cases. However, the peak corresponding to α-Fe (110) of trilayer samples was significantly different from that of bulk Fe. We interpret that the peak position shift is caused by the elongation of the (110) interplanar distance ‘d’ due to large internal stress in the Fe layers induced by adjacent Si layers, and their intermixing during deposition causing the formation of iron silicide layer at the interface. Further, the intensity of (110) peak reduces substantially and FWHM increases with increasing Si layer thickness. Particularly for Si ≥ 30 Å, the peak shows a broad hump around 2θ = 43.56˚, due to further growth of iron silicide layer along with appearance of an amorphous Si (α-Si) layer at the interface.

Figure 2 shows the t_Si dependence of particle size and

elongation in the interplanar distance of α-Fe (110), Δd/d₀, in the trilayer samples evaluated from the GIXRD patterns as shown in Figure 1. Here d₀ denotes the (110) interplanar distance of bulk α-Fe crystallites with a perfectly cubic lattice and Δd denotes the difference between the (110) interplanar distance, d, of the trilayers and d₀. The particle size obtained from Scherrer formulism decreases substantially with increasing t_Si, indicating that trilayers with smaller t_Si have larger and more oriented α-Fe (110) crystallites and t_Si beyond 25 Å appeared to be composed of micro-crystallites or amorphous Fe grains, which we assume, are due to the appearance of pure amorphous Si (α-Si) layer in the spacer. On the other hand, Δd/d₀ increase monotonically with increase of t_Si and saturated at about 2.42% for t_Si > 25 Å. This suggests that the elongation may be attributed to the large compressive stress in the Fe layers imposed by adjacent Si layer.

The quality of surface and interface structure plays decisive role in achieving optimum performance for which the multilayer structure is designed. Therefore, further information about the structure and surface morphology can be obtained from AFM studies conducted on Fe/Si/ Fe trilayer samples. Figure 3 shows the three-dimensional AFM images of [Fe (30 Å)/Si (15 Å)/Fe (30 Å)] and [Fe (30 Å)/Si (30 Å)/Fe (30 Å)] trilayer samples obtained from 2 × 2 μm² sample area using contact mode. It is clearly seen from Figure 3(a) that the deposited layers are nearly continuous showing a valley like features with wide distribution of the sizes and separation, giving rise to a very large rms surface roughness value. These valleys like features are clearly indicating the crystalline and oriented growth of Fe. However, as the thickness of Si layer increases to 30 Å (see Figure 3(b)), the separation of these features decreases and show the formation of more continuous and denser layers compared

to the above-mentioned case, which is due to the appearance of pure amorphous Si (α-Si) layer in the spacer and as a result the value of surface roughness decreases. This is in correlation with the XRD data’s. Hence, the obtained AFM pictures provide us more clear information about different growth stages, as the Si layer thickness is increased from 15 Å to 40 Å.

In order to see the variation in AF coupling with increasing spacer layer thickness t_Si, MOKE measurements have been performed. For all the samples, magnetic field was applied parallel to the surface of the film and hysteresis loops were recorded upto the saturation magnetization. Figure 4 shows the evolution of MOKE loops for Fe/Si/Fe trilayer samples as a function of Si layer thickness. As can be seen, the magnetic parameters like the remanent (M_R) to saturation (M_S) magnetization ratio (the so-called F_AF = 1 – M_R/M_S parameter) and saturation field (H_S) are strongly influenced by the spacer thickness. In metallic MLS the F_AF parameter is usually treated as proportional to the AF coupled fraction of the sample, and F_AF = 1 when ML is completely AF coupled and F_AF = 0 for a ferromagnetically coupled MLS.

Figure 5 displays the values of F_AF and the saturation field (H_S) for the Fe/Si/Fe trilayer samples as a function of spacer layer thickness t_Si. F_AF dependence shows that the strong AF coupling occurs at t_Si = 25 Å where F_AF = 0.38. Similar dependence and occurrence of only single maximum of F_AF are reproduced by H_S (t_Si). For t_Si = 25 Å, H_S reaches its maximum value of about 158 Oe, which may also indicate the occurrence of a very strong AF interlayer coupling. For t_Si > 25 Å, both F_AF and H_S values are strongly reduced, F_AF < 0.1 and H_S < 90 Oe. The reduction of F_AF and H_S values for t_Si < 25 Å points out that the neighbouring Fe layers become gradually connected through pinholes. However, it is noted that the rate of change of AF coupling with spacer layer thickness on both sides of the AF peak is not the same. The difference is mainly due to changes in interlayer structure according to different sub-layer thicknesses.

From Figure 4 one can see the existence of different interlayer coupling that modifies the shapes of hysteresis loops from square-like to step-like. For t_Si = 15 Å, the hysteresis loop is square in shape indicating that the distribution of anisotropy is rather sharp which makes the domain magnetization switching beyond certain applied magnetic field. The large vertical jump with retentivity almost equal to saturation magnetization and lower coercivity value suggest the soft magnetic behaviour of the sample with strong anisotropy leading to in plane easy direction of magnetization. However, shape of the hysteresis loop changes from square to smoother one as the spacer layer thickness increases to t_Si = 25 Å, indicating strong AF coupling at this thickness. Further on increasing the thickness of spacer layer from 25 Å to 30 Å and 40 Å, shape of the loop again changes to step-like, characterized by two “coercive” fields: H_C1 and H_C2, indicating the presence of different phases at these thicknesses. This peculiar behaviour is mainly due to the presence of silicide layer along with pure α-Si layer at the interface. Due to weak coupling the correct identification of the saturation fields is very difficult. Therefore, in the inset of Figure 5, we also present the values of H_C1 and H_C2 fields as a function of spacer thickness. From the figure, one can see that H_C2 appears only for thicker Si layer, i.e., for t_Si ≥ 30 Å. Hence, above magnetic measurements conclude that Fe layers in the trilayers generally consist of Fe layers doped

with Si, ferromagnetic Fe-Si silicide layers and nonmagnetic Fe-Si silicide interface layers, while the nominal Si spacers turn out to be Fe-Si compound layers with additional α-Si sublayers only under the condition where t_Si ≥ 30 Å. A strong AF coupling is observed with iron silicide spacers and disappears when α-Si layers appear in the spacer.

In order to further investigate the chemical nature of layers at surface and interface in the as-deposited Fe/Si/Fe trilayer samples, XPS measurements have been performed after repeated sputtering of sample surface by Ar⁺ ions. This gave the chemical composition of the film at different depths, averaged over the escape depth of photoelectrons (~40 Å). The interfaces are analyzed when the top layer was Fe or Si rich and also when top layer was Si (Si/Fe interface). It was performed on purpose to note the possibility of inequivalence of the interfaces, as an AES study by Strijkers et al. indicated some differences of interfaces at iron silicide formation [24].

Figure 6 shows the survey scans of as-deposited sample surface of Fe(30 Å)/Si(15 Å)/Fe30 Å) trilayer before and after 10 min sputtering. From the spectrum (before

sputtering) one can see that peaks corresponding to carbon (C), oxygen (O) and iron (Fe) are clearly visible, whereas very small contribution due to silicon (Si) is observed. This small presence of Si peaks is expected, as Si is the second layer after Fe (30 Å) top layer. Since the sample was exposed to air, there are large quantities of absorbed carbon and oxygen on the surface as evident from the spectrum. However after sputtering the sample for 10 min, the intensity of carbon and oxygen drops to minimum with appearance of buried Si peaks, indicating that signals are coming from the first interface region. We also note that the reduction in the carbon content is more rapid then oxygen content and may be attributed to preferential sputtering of carbon with respect to reacted oxygen.

Figure 7 shows the fitted (a) Fe-2p and (b) Si-2p core level spectra of Fe(30 Å)/Si(15 Å)/Fe30 Å) trilayer sample as a function of sputtering time. For comparison, core levels from clean Fe and Si single layer films are also given in the same figure. We note that these spectra are in excellent agreement with previous reports [25]. Before sputtering the observed spectrum of Fe presents significant differences, which are characteristic of the oxidized state. Fe-2p_3/2 peak is shifted to binding energy (B.E) of 710.5 eV by the change in electrostatic potential at the Fe sites. This chemical shift is caused by the partial transfer of electrons from the Fe³⁺ to the O^2– ions. The 2p_3/2 peak is broadened by multiplet splitting involving the core hole and the valence electrons. Finally, the 2p_3/2 peak shows a distinct satellite feature around 714.6 eV B.E, caused by charge-transfer processes. The main peak corresponds to so-called well-screened 2p⁵3d⁶L final state configuration, while the satellite corresponds to so-called poorly screened 2p⁵3d⁵ final state configuration (L denotes a ligand hole) [26]. However after sputtering for 5 min, the presence of sharp peak around 707.0 eV and tail towards higher B.E indicates that signals are coming mostly from elemental Fe region. The spectra are split by the 2p spin-orbit effect into the 2p_3/2 and 2p_1/2 regions and continuous tail is caused by the electron-hole pair excitations and is a significance of metallic states. However, this peak is shifted by 0.2 eV towards higher B.E as compared to clean Fe-2p core level. This shift indicates the formation of iron silicide (FeSi) layer at the interface in as deposited trilayer sample. Similarly, after 10 min sputtering, it is observed that this peak is further shifted by 0.1 eV towards higher B.E of 707.1 eV, suggesting that signals are now coming mostly from the first interface region i.e., more from silicide layer. However after further sputtering the sample for 15 min, no noticeable changes are observed in the B.E and FWHM of this peak.

Similarly, figure. 7b shows the Si-2p core level spectra of Fe(30 Å)/Si(15 Å)/Fe30 Å) trilayer sample with respect to sputtering time. In this case, before sputtering peaks due to Si are very small in intensity as it is the second layer after 30 Å top Fe layer as discussed earlier. However, after 5 min sputtering, spectrum clearly shows the presence of buried Si-2p peaks. A closer inspection of this peak shows small presence of SiO₂ peak along with elemental Si-2p (99.7 eV) peak at a B.E position of 102.1 eV. The set of typical Si-2p spectra consist of spin orbit doublet peaks corresponding to Si-2P_3/2 and Si-2p_1/2 level states. However, spectrometer resolution does not allow distinction of their structure. Similar to the case of Fe-2p after 5 min sputtering, Si-2p peak is also found to be shifted by 0.2 eV towards higher B.E as compared to elemental Si-2p further suggesting the formation of silicide layer at the interface. Upon further sputtering the sample for 10 and 15 min, there are no changes (B.E and

and FWHM values) in the Si-2p spectra. The above results show that there seems to be no changes in the core level spectra of both Fe-2p and Si-2p with respect to sputtering, suggesting that silicide layer is formed at the interfaces during deposition and also complete Si layer is interdiffused into Fe layers thereby converting as deposited Fe/Si/Fe trilayer structure to Fe/FeSi/SiFe/Fe structure.

Figure 8 shows the fitted (a) Fe-2p and (b) Si-2p core level spectra of Fe(30 Å)/Si(40 Å)/Fe30 Å) trilayer sample with respect to sputtering time. In this case also, before sputtering, the surface is completely oxidised and the spectrum mostly shows oxidised states of Fe-2p core levels. After 5 min sputtering, similar to earlier case, Fe-2p and Si-2p core levels are shifted by 0.3 eV and 0.2 eV respectively towards higher B.E as compared to core levels of pure Fe and Si layers, suggesting the formation of silicide layer at the interface. After further sputtering for 5 min (total 10 min) a drop in the Fe-2p and increment in the Si-2p core level intensities is observed. Apart from this, no shift is observed in the B.E position of Fe-2p core level, whereas slight shifts of 0.1 eV towards lower B.E (99.6 eV) is observed in case of Si-2p core level. Upon further sputtering for 5 min (total 15 min) one can clearly see more reduction in Fe-2p with observable enhancement in Si-2p core level intensity. However, an important point to be noticed here is that Si-2p core level is further shifted by 0.1 eV towards

lower B.E (99.5 eV), and still no shift is found in the B.E of Fe-2p core level The shift only in Si-2p core level towards clean Si-2p core level indicates the appearance of elemental Si layer at this situation and still small presence of silicide layer from the first interface causes Fe-2p core level to remain shifted.

Based on the above-mentioned XRD, MOKE and XPS results we have proposed a model for Fe/Si/Fe trilayer sample having the Si layer thickness 15 Å and 40 Å, respectively as shown in Figure 9. The model shows the formation of silicide layer at the interfaces in both the cases. However, apart from silicide layer, pure α-Si layer is also observed only in case of [Fe(30 Å)/Si(40 Å)/Fe(30 Å)] trilayer sample.

Figure 10 show the variation of the resistivity (ρ) versus nominal Si layer thickness t_Si at room temperature. It is shown that the resistivity increases slowly with increasing t_Si for the films with t_Si < 25 Å. Near t_Si ³ 25 Å, the resistivity increases rapidly, and when t_Si > 30 Å it is nearly constant with a slightly increasing tendency. The t_Si dependence of the resistivity in this system is quite peculiar and different from many metallic multilayered systems [27]. It is known that FeSi silicide is easily formed at the interface due to atomic intermixing and interdiffusion and the resistivity of FeSi silicide is much higher than the resistivity of Fe [28]. Thus peculiar t_si dependence of the

resistivity for t_Si < 25 Å can be explained as mainly due to the increase of interdiffusion and the silicide formation with increasing t_si. The rapid increase of the resistivity and changes in GIXRD and MOKE data at t_Si ³ 25 Å can be understood by that in addition to the FeSi layer, a α-Si layer appears in the spacer as also revealed form our XPS results. The appearance of α-Si layer in the spacer stops further interdiffusion and results in no appreciable change in the resistivity and also prevents AF coupling after t_Si ³ 30 Å.