CAPÍTULO III: DIAGNOSTICO DE LA PRODUCCIÓN DE MATERIALES DE LA CONSTRUCCIÓN EN EL TALLER DE
3.2. Potencialidades y limitantes del “Taller de Pedrito” y del Trabajo por cuenta propia en el sector
The logical starting point for investigating the trilayer structures, is with a single layer. This should provide information on the resonant scattering factors, oxidation of the capping layer and whether there is high roughness between the interfaces at the substrate and buffer layer. This information would then in principle be useful
for restricting some of the parameters when fitting the trilayer to help reduce the number of variables included in the fit. As such, specular XRMS data was recorded for single films of SmCo capped with AlZr. The data was fitted using GenX and used to check the quality of the SmCo single layer samples including interface roughness and uniformity. The single layers of SmxCo1−xwere grown as described in Chapter 6.
Reflectivity from a Sm8Co92 film (which is the single film with the closest composition to that in the trilayer) was recorded on beamline X13A at the NSLS, Brookhaven National Laboratory. Figure 7.1(a) shows the sum reflectivity signal recorded at the Co L3 and Sm M4 edges and also using a Cu Kα laboratory source.
Figures 7.1(b) and (c) show the A.R. obtained at the Co L3 and Sm M4 edges respectively. All specular reflectivity data for the single film samples were recorded at 300 K. These were fitted simultaneously which was intended to constrain the parameter set to provide confidence in the fit. As the value of the SLD is given by the product of the density and scattering factors, it is possible when fitting single data sets to find a solution where the parameters can vary by large amounts as long as the total product is the same. By using multiple data sets, the individual values of the scattering factors and densities should be restricted to give more accurate values. It is also important to highlight that many solutions can exist which give very similar looking fits, but they can often give unphysical results in the SLD profile. This must be monitored whilst fitting the data and parameters should be restricted such that they can only take values which give meaningful SLD profiles.
The chemical and magnetic SLD profiles obtained from the GenX fits are presented Figure 7.1(d) and (e) respectively. The chemical SLD shows that the SmCo is uniform across the layer in thezdirection. The resonant scattering factors of the Sm and Co were large and negative resulting in notable changes in the chemical SLD values at the resonant edges. The capping and buffer layers of AlZr are clearly visible, with the dashed lines serving as a visual guide indicating the upper and lower boundaries. To achieve a good fit, it was necessary to add in extra layers between the SmCo and AlZr layers both at the top and bottom where the scattering factors were allowed to vary slightly from those of either the SmCo or AlZr. These layers are approximately 20 ˚A thick and it would be logical to assume that these are regions of chemical intermixing. As a Gaussian error function is unlikely to account for variations in the chemical profiles, this goes some way towards allowing flexibility in the fit. Fitting without these extra layers was very unsuccessful and gave poor fits. The very uppermost layer was assumed to be ice, and as such was allowed to vary slightly (≤5 ˚A) in thickness and roughness between data sets to account for small amounts of ice build up during experiments.
Figure 7.1: Reflectivity data for a single film of Sm8Co92 with a buffer and capping layer of Al70Zr30 on a Si substrate. (a) Specular reflectivity data recorded using (blue) a Cu Kα1 laboratory x-ray source. Additionally the sum signal is shown which was recorded using synchrotron radiation tuned to (red) the Co L3 edge and (green) the Sm M4 edge. A.R. recorded at (b) the Co L3 edge and (c) the Sm M4 edge. The points are the data and lines are the fit obtained using GenX. The SLD profiles obtained from the fits are shown for (d) the chemical and (e) the magnetic parts.
The magnetic SLD profiles in Figure 7.1(e) give unexpected results, with a magnetic moment clearly visible in the capping and buffer layers when AlZr should be non-magnetic. Quite often it is possible that there are not unique solutions for fits like these and so many attempts were made to fit the data with the magnetisation restricted to the SmCo layer only, trying different initial conditions. None of these attempts to fit the data could achieve the fringe periodicity in the A.R. necessary for a good fit which suggested that the magnetic layer needed to be thicker whilst the chemical SLD needed to remain unchanged. The lack of success at fitting the data when the model included a distinct layer of SmCo with a non-magnetic capping and buffer layer showed that the layers in the sample were possibly not exactly of the composition expected from the growth. To account for this in the model, the scattering factors were allowed to vary and the capping and buffer layer were given a magnetic moment of 1, whilst the magnetic scattering factors were fitted.
Figure 7.2: SLD profiles obtained from fits to reflectivity data for a single film of Sm8Co92. (a) Real and imaginary parts of the SLD from data recorded using a Cu Kα1 laboratory x-ray source. (b)-(c) Real and imaginary parts of SLD profiles from data recorded at the Co L3 edge and Sm M4 edge respectively and (d)-(e) are the corresponding magnetic SLD profiles.
to the single layers are shown in Figure 7.2. There are differences between the values of the real and imaginary parts but the structural information is much the same in both. As all the necessary information is contained in the real part and not much more is gained from the imaginary part, Figure 7.1 only shows the real SLD for simplicity. Henceforth, all SLD profiles presented from here on in this chapter will only show the real part. This helps to make a comparison between different SLD profiles easier.
To illustrate why it was that the magnetisation was allowed to extend into the capping/buffer layers, the best fit achieved with the magnetisation in the these layers set to zero is shown in Figure 7.3. In Figure 7.3(a) the data and fit is given and (b) shows the corresponding real parts of the chemical and magnetic SLD. This model allowed the magnetic SLD at the upper and lower interfaces of the SmCo layer to change by increasing or reducing the magnetic SLD value but did not allow it to extend into the capping/buffer layers. The details of how this is incorporated into the model in GenX were given in Chapter 4. No matter how the magnetic SLD was changed at the upper and lower interfaces in the SmCo, however, a suitable fit could not be achieved across the entireqz range.
Figure 7.3: Asymmetry ratio for a single film of Sm8Co92 recorded at the Co L3 edge with data (points) and fit using a model which did not allow magnetisation in the buffer/capping layers.
Particular difficulties were had with fitting the fringes in the region marked by the arrow in Figure 7.3 at around 0.27 ˚A−1, which indicated that a feature of approximately 25 ˚A thickness would be needed to change the phase of the fringes to be able to achieve a good fit in this region. In the SLD profile from the fit in Figure 7.1, the magnetic feature observed in the AlZr layer at the bottom of the sample is 25 ˚A thick which corresponds directly to the feature predicted to achieve the necessary fringe thickness. Attempts to include a feature like this at the bottom of the SmCo layer without extending into the buffer layer were unsuccessful at fitting the fringes in the region marked by the arrow in Figure 7.3. Although there was not the same evidence for a similar magnetic layer at the top of the sample, it was logical that the same could be assumed for the capping layer.
A similar example is given for the data recorded at the Sm M4 edge with the data and fit shown in Figure 7.4(a) and the chemical and magnetic SLD profiles from the fit in Figure 7.4 (b). Again, the best fit achieved whilst keeping the magnetisation in the capping/buffer layers set at zero is shown and shows that the overall shape of the simulated A.R. does not match the data well. The magnetic SLD was allowed to vary at the upper and lower interfaces of the SmCo to try and achieve a suitable fit but was unsuccessful. The particular feature in the data that was difficult to fit was in the region marked by the arrow at around 0.27 ˚A−1. To fit this feature, the magnetic SLD required the surface regions to be altered significantly and in particular, the magnetisation needed to be extended into the capping/buffer layer to move the feature to higher qz whilst retaining the shape of the rest of the
Figure 7.4: Asymmetry ratio for a single film of Sm8Co92 recorded at the Sm M4 edge with data (points) and fit using a model which did not allow magnetisation in the buffer/capping layers.
The magnetic SLD from the fit to the data recorded at the Co L3 edge as presented in Figure 7.1 shows that the magnetisation extends uniformly across the SmCo layer and drops to zero as would be expected. There is then, however, a distinct magnetic layer at the top and bottom suggesting that there is Co in the capping and buffer layers with a magnetic moment. The data recorded at the Sm M4edge does not show the drop to zero magnetisation at the upper and lower edges of the SmCo layer but instead increases to a magnetic SLD value that is much larger than that for the SmCo layer itself. This suggests there is either an increased amount of Sm in the buffer layers or that the composition of the layer has Sm with an enhanced magnetic scattering factor. It appears that there may have been some issues with the growth of the capping and buffer layers for this SmCo single layer sample which has resulted in mixed layers incorporating Sm, Co and AlZr. Although there are issues with the capping/buffer layers, it should be possible to use the resonant scattering factors for the Co and Sm obtained from the fits to the single layers in the fitting of the trilayer data. It will likely be necessary to account for chemical mixing in the trilayer model as well.
Modelling the oxide has also presented many challenges whilst fitting this data and the simulated reflectivity is very sensitive to the SLD profile of the sur- face region, particularly for the resonant energies due to absorption effects. Initial attempts to fit these with a single layer of AlZr with a thin oxide layer on top using published scattering factors obtained from [100] were unsuccessful. The fit suggests that there is a complex oxide formed at the surface of the samples which is not well described by a Gaussian error function. This was modelled by setting the AlZr
layer to actually be three thin layers and allowing the composition to vary for each of these layers. The scattering factors at each energy were then allowed to deviate from the tabulated values slightly to account for the chemical variations in such a layer. As the scattering factors of the mixing layers were allowed to vary to account for different local environments of chemical species as the composition changed, it was difficult to determine the exact chemical profiles. Fits were attempted where the scattering factors were kept constant for each element at each energy and the ratio of elements changed, but produced very poor results.
Work looking at similar amorphous AlZr films showed that the alloy formed regions of polycrystalline Al resulting in other regions of the material becoming Zr rich [204]. This would fit with the AlZr capping layer on the samples in this work exhibiting a complex chemical composition profile. It was also found that the AlZr films had an oxide of up to∼6 nm formed on the surface. This would be the entirety of the 2-3 nm capping layer on the samples presented here. This oxide was shown to have gradients of Zr-enrichment and O-deficiency throughout, plus a mixture of Al and Al2O3 near the surface [205]. This adds extra complications for fitting as it is unlikely that the oxidation of the AlZr will be well represented by a Gaussian error function, especially when fitting data recorded at multiple energies. This also suggests that the SmCo layer is not necessarily protected from oxidation by the capping layer. This information about mixing layers and complex oxides can be carried forward to help with fitting the data from the trilayer.