Depositing the multilayers at different substrate temperatures is an obvious way of influencing the composition and crystallinity of the spacer layer phase in the Fe/Si multilayers. Fullerton has suggested that the interlayer of Fe/Si multilayers is improved by high-temperature growth.[36] We have grown films on glass substrates at various temperatures between -150 and +200°C. The effect of substrate temperature on the interlayer coupling of the films is illustrated in Figure 9, where magnetization curves for three (Fe40Å/Si14Å)x40 multilayers grown at -150°C, +60°C (nominal RT) and +200°C are shown. The data show that as the substrate temperature increases the saturation field increases indicating larger AF coupling. The saturation magnetization also decreases, suggesting a larger degree of interdiffusion in the films grown at higher temperatures.
The suspicion that more interdiffusion occurs at higher substrate temperatures is confirmed by examination of the SAXS spectra for the three films, shown in Figure 10. The film grown at reduced temperature has 7 peaks while the film grown at nominal RT has 5 and the film grown at +200°C has only 4. Quantitative modelling of low-angle x-ray data has shown that the suppression of higher-order peaks may be due to either interdiffusion or cumulative roughness.[28,37] Certainly larger cumulative roughness could also occur at higher growth temperatures, but one would expect very rough growth to suppress AF coupling due to an increased number of pinholes and larger magnetostatic interlayer coupling.[38] Since higher growth temperatures seem to enhance rather than suppress the coupling, it seems more likely that high substrate temperatures are promoting interdiffusion rather than roughness. Studies of Mo/Si multilayers showed that a growth temperature of 150°C gives maximum SAXS reflectivity, which the authors attribute to greater interface smoothness than for RT deposition.[39] Smaller bilayer periods in multilayers grown at higher temperatures support the claim of increased interdiffusion. Fitting Eqn. 1 to peak positions from Figure 10 gives lambda = 52.7, 49.3, and 43.8 Å\ respectively for the -150°, +60°, and +200° multilayers versus the nominal value of 54Å.
Higher substrate temperatures may also promote ordering of the Fe and Si atoms in the crystalline spacer layer. In the fully ordered B2 phase, the Fe and Si atoms sit on different simple cubic sublattices. The sublattice order can occur irrespective of whether or not the Fe to Si ratio is 1:1. It is interesting to speculate whether the AF coupling is dependent on the degree of ordering in the spacer layer. An ordering-dependent coupling seems plausible in light of the Fermi-surface theories of coupling in metal/metal multilayers.[40,41] A well-ordered B2 or DO 3 phase would have more well-defined Fermi surface features than a random solid solution. Unfortunately the (001) silicide peak has only been observed by TEM, making experimental attempts to address this issue difficult. Further studies with x-ray diffraction and soft x-ray fluorescence are underway.
The crystallinity of the films also varies with growth temperature. Surprisingly, films grown at both low and high temperatures on glass substrates always have only the (011) texture, while films grown at nominal RT often have mixed (001) and (011) textures. The multilayers deposited on heated and cooled substrates do differ greatly in that those grown at low temperature have amorphous spacer layers, while those grown at high temperatures have long crystalline coherence lengths. The reasons for the strange temperature dependence of growth texture are not understood, although one presumes that they have to do with the kinetics of growth. It is not clear why the (001) texture should appear at all, although it has also been seen in Mo/Ge multilayers.[3] An oscillatory dependence of film texture on spacer layer thickness and deposition conditions has been reported for NiFe/Cu multilayers grown by IBS.[42] The (001) texture has not been reported in polycrystalline magnetron-sputtered Fe/Si multilayers, and may be due to some peculiarity of IBS growth.
A logical extension to the growth temperature studies is to try annealing the Fe/Si multilayers grown at lower substrate temperatures to see if their properties evolve towards those of the multilayers grown at higher temperatures. As far as the magnetic properties are concerned, the answer is ``no.'' Annealing the uncoupled RT-grown (Fe30Å/Si20Å)x40 and low-temperature-grown (Fe40Å/Si14Å)x40 multilayers at +200°C for two hours had almost no effect on their magnetic properties beyond a slight magnetic moment reduction. A subsequent 300°C anneal for two hours once more produced a moment reduction and a decrease in coercive field in the uncoupled multilayers. A very low coercive field for annealed Fe/Si films is not surprising given the well-known softness of Fe-Si alloys. A 300°C anneal even eliminated the interlayer exchange coupling of a RT-grown (Fe40Å/Si14Å)x50 film used as a control. For this (40/14) multilayer, the 300°C anneal caused the SAXS peaks to narrow and reduced their number from 5 to 4. At the same time the bilayer period decreased from 49.4Å to 46.0Å. High-angle x-rays spectra (not shown) indicated that the Fe lattice constant slightly decreased, which is consistent with increased diffusion of Si in the Fe layer.[31] These x-ray and magnetization results imply that annealing primarily promotes interdiffusion of the Fe and silicide layers. With sufficient interdiffusion the spacer layer may become ferromagnetic, which would explain the suppression of antiferromagnetic interlayer coupling. These Fe/Si multilayers show less thermal stability than Mo/Si multilayers with comparable layer thicknesses, which do not show changes in SAXS spectra until 400°C.[39] There was no sign of the solid-state amorphization previously observed in Fe/Si multilayers with thicker layers.[43]
Whatever process occurs during annealing, it does not enhance the interlayer coupling the way that +200°C growth does. This is hardly surprising given that annealing will tend to drive the multilayer towards its equilibrium state, presumably a mixture of different iron silicide phases. There is no reason to think that the crystalline Fe/FexSi1-x multilayer should be an intermediate phase during the annealing. In the future the kinetics of Fe/Si multilayer growth at different substrate temperatures will be investigated further by employing an ion-assist gun to improve atomic surface mobility.