CEMS Results

All of the $^{57}$Fe Mössbauer results for the U/Fe multilayers were recorded using CEMS at room temperature. Transmission mode Mössbauer spectroscopy is prohibited by the attenuation of the Mössbauer gamma ray by the glass substrates. The spectra were recorded with the sample plane normal to the incident gamma ray direction. The hyperfine parameters from the least squares fits are given in Table 7.2.

Table 7.2: Hyperfine values for all U/Fe multilayers studied.

$$\Gamma$$

(mms) (mms) (mms) (kG) (%)

$$\left[\mathrm{U}(28\,\mathrm{\AA{}})/\mathrm{Fe}(30\,\mathrm{\AA{}})\right]_{31} +0.02 -0.05 0.31 320.7$$ 14

$$+0.05 -0.02 0.94$$ 288.0 38

$$-0.15 -0.71 0.78$$ 0 48

$$\left[\mathrm{U}(28\,\mathrm{\AA{}})/\mathrm{Fe}(43\,\mathrm{\AA{}})\right]_{31} +0.02 +0.02 0.26 325.3$$ 30

$$-0.10 -0.13 0.90$$ 288.0 34

$$-0.20 -0.76 0.73$$ 0 36

$$\left[\mathrm{U}(40\,\mathrm{\AA{}})/\mathrm{Fe}(60\,\mathrm{\AA{}})\right]_{10} +0.01 +0.00 0.28 328.9$$ 38

$$+0.00 -0.07 0.72 304.6$$ 34

$$-0.18 -0.73 0.73$$ 0 28

$$\left[\mathrm{U}(80\,\mathrm{\AA{}})/\mathrm{Fe}(60\,\mathrm{\AA{}})\right]_{20} +0.01 +0.00 0.27 327.8$$ 35

$$+0.00 -0.07 0.73 304.0$$ 34

$$-0.19 -0.69 0.70$$ 0 31

$$\left[\mathrm{U}(101\,\mathrm{\AA{}})/\mathrm{Fe}(60\,\mathrm{\AA{}})\right]_{20} +0.02 -0.02 0.27 328.3$$ 34

$$+0.00 -0.02 0.76 300.0$$ 34

$$-0.20 -0.69 0.64$$ 0 32

$$\left[\mathrm{U}(42\,\mathrm{\AA{}})/\mathrm{Fe}(113\,\mathrm{\AA{}})\right]_{21} +0.01 -0.01 0.28 329.0$$ 69

$$-0.01 -0.09 0.67 302.0$$ 15

$$-0.19 -0.70 0.69$$ 0 16

$$\left[\mathrm{U}(22\,\mathrm{\AA{}})/\mathrm{Fe}(180\,\mathrm{\AA{}})\right]_{5} +0.02 +0.00 0.29 328.8$$ 53

$$-0.06 -0.12 1.10 299.0$$ 31

$$-0.24 -0.89 0.72$$ 0 16

The parameters for the sextets are consistent with magnetically ordered bcc iron. In all iron layers there are two magnetic components: one having the values of crystalline bcc iron with a linewidth close to that of the calibration (indicating good crystalline growth) and hyperfine fields close to those for bulk bcc iron and a second, broader component with a reduced hyperfine field.

The hyperfine parameters are unaffected by the thickness of the uranium layer. Figure 7.10 shows four spectra, three with the same iron layer thickness but varying thicknesses of uranium. These three spectra are essentially identical in both hyperfine parameters and the relative areas of each component. The fourth spectrum shows a sample with a much thicker iron layer and the relative amount of each component can be seen to have changed considerably.

Figure: CEMS spectra for U/Fe multilayers: $\left[ \mathrm{U}(40)/\mathrm{Fe}(60) \right]_{10}$, $\left[ \mathrm{U}(80)/\mathrm{Fe}(60) \right]_{20}$, $\left[ \mathrm{U}(101)/\mathrm{Fe}(60) \right]_{20}$ and $\left[ \mathrm{U}(42)/\mathrm{Fe}(113) \right]_{21}$. The spectra are unaffected by the uranium layer thicknesses.

The first component is attributed to a layer of crystalline bcc iron in the center of the iron layer. The linewidth is the same as a pure iron foil used for calibrating the spectrometer. The slightly reduced hyperfine field in the samples with iron layers $<60\ensuremath{\unskip\,\mathrm{\AA{}}}$ arises from a reduction in $T_{c}$ as the strength of interactions between the iron layers is diminished by the interspacing of uranium layers and reduced coordination with neighbouring iron atoms. This effect is at its greatest in the sample with the thinnest iron layer, $\left[ \mathrm{U}(28)/\mathrm{Fe}(30) \right]_{31}$, where there is a $3\%$ reduction in hyperfine field. The effect decreases as the iron layer thickness is increased and the hyperfine field reaches a maximum for samples with an iron layer thickness $\geqslant60\ensuremath{\unskip\,\mathrm{\AA{}}}$ of $\approx329\ensuremath{\unskip\,\mathrm{kG}}$, identical to that of pure bulk iron.

The second magnetic component has a much reduced hyperfine field of up to $13\%$ and a very broad linewidth. This is a region of iron at the interfaces with much reduced crystallinity and subject to interface effects. Interdiffusion of iron and uranium atoms, and assuming only the iron atoms carry a moment, gives fewer magnetic neighbours for each iron atom and hence a reduction in hyperfine field. The large linewidth is due to a distribution of iron sites with increasing hyperfine field as the concentration of iron atoms increases further into the iron layer. Thicker iron layers will have more pure iron sites and so the hyperfine field of the second component increases. The proportion of this region to the crystalline iron is very large in the sample with the thinnest layer (the iron layers are only just thick enough to begin fully crystalline growth) but this proportion decreases with increasing iron layer thickness.

This change in the proportion of crystalline iron to the distribution of magnetic iron sites is illustrated in Figure 7.11. The first spectrum from the $\left[ \mathrm{U}(28)/\mathrm{Fe}(30) \right]_{31}$ sample shows the majority of the iron is the poorly-crystalline state or as part of the UFe$_{2}$ intermetallic. As the iron thickness is increased to $43\ensuremath{\unskip\,\mathrm{\AA{}}}$ the crystalline component can be seen to increase in intensity. At $180\ensuremath{\unskip\,\mathrm{\AA{}}}$ the iron layer is predominantly crystalline.

Figure: CEMS spectra for U/Fe multilayers: $\left[ \mathrm{U}(28)/\mathrm{Fe}(30) \right]_{31}$, $\left[ \mathrm{U}(28)/\mathrm{Fe}(43) \right]_{31}$ and $\left[ \mathrm{U}(22)/\mathrm{Fe}(180) \right]_{5}$. Increasing crystallinity is observed within the iron layers as the thickness increases.

The paramagnetic doublet matches well the bulk hyperfine parameters of the Laves Phase intermetallic UFe$_{2}$, with the same crystal structure as the Re/Fe$_{2}$ samples studied in this thesis (Figure 6.1). The values reported in References blow_70 and tsutsui_01 for room temperature measurements show that UFe$_{2}$ is in a paramagnetic phase ( $T_{c}=167\ensuremath{\unskip\,\mathrm{K}}$) with an isomer shift of $\unitfrac[-0.21]{mm}{s}$ and a quadrupole splitting of $\unitfrac[-0.45]{mm}{s}$ (the sign of the quadrupole splitting is obtained from low temperature magnetically ordered spectra). The doublet has a much broader linewidth of $\unitfrac[0.50]{mm}{s}$ than that of magnetically split spectra. The broader linewidth and slightly larger quadrupole splitting observed in the UFe$_{2}$ component within the U/Fe layers studied in this thesis can be explained by the decreased crystallinity compared to the bulk samples studied in Reference blow_70.

The magnetron sputtering process may be forming UFe$_2$ by co-evaporation at room temperature producing the formation of the intermetallic as the atoms of either element are driven into the surface of the newly formed thin film. Other possible materials which could form such as UFe, U$_{6}$Fe and amorphous iron do not match all of the hyperfine parameters consistently recorded for the doublet in all of the multilayer spectra. The current samples can only be recorded using CEMS and this restricts the readings to room temperature. Future work with samples on kapton backings will allow confirmation of the composition of the doublet as these will allow low temperature (below $T_{c}$) measurements.

Analysis of the intensity of each component can give an indication of the amount (roughly corresponding to layer thickness) of the total iron deposited which is in a particular state: crystalline iron, iron with reduced crystallinity or UFe$_{2}$ intermetallic. These values are shown in Table 7.3.

Table 7.3: Calculated thickness of iron layers in a fully crystalline (FE), paramagnetic (PM) or reduced crystallinity (RC) state for all U/Fe samples studied.

Area Layer Area Layer Area Layer

$$(\%) (\%) (\%)$$ (Å) (Å) (Å)

$$\left[ \mathrm{U}(28)/\mathrm{Fe}(30) \right]_{31}$$ 14 9 48 14 38 11

$$\left[ \mathrm{U}(28)/\mathrm{Fe}(43) \right]_{31}$$ 30 13 36 15 34 15

$$\left[ \mathrm{U}(40)/\mathrm{Fe}(60) \right]_{10}$$ 38 23 28 17 34 20

$$\left[ \mathrm{U}(80)/\mathrm{Fe}(60) \right]_{20}$$ 35 21 31 18 34 20

$$\left[ \mathrm{U}(101)/\mathrm{Fe}(60) \right]_{20}$$ 34 21 32 19 34 20

$$\left[ \mathrm{U}(42)/\mathrm{Fe}(113) \right]_{21}$$ 69 78 16 18 15 17

$$\left[ \mathrm{U}(22)/\mathrm{Fe}(180) \right]_{5}$$ 53 95 16 29 31 56

In the sample with the thinnest iron layer, $\left[ \mathrm{U}(28)/\mathrm{Fe}(30) \right]_{31}$, the crystalline iron layer is very thin, only amounting to 3-4 monolayers (using an iron atom spacing of $2.5\ensuremath{\unskip\,\mathrm{\AA{}}}$. The amount of crystalline iron does not increase linearly with the increase of iron layer thickness at first. An increase of $13\ensuremath{\unskip\,\mathrm{\AA{}}}$ in the iron layer only gives an increase of $4\ensuremath{\unskip\,\mathrm{\AA{}}}$ of crystalline iron. Similarly the next increase of $17\ensuremath{\unskip\,\mathrm{\AA{}}}$ in the total iron layer only gives a further $10\ensuremath{\unskip\,\mathrm{\AA{}}}$ at most. Once the total iron layer is above $60\ensuremath{\unskip\,\mathrm{\AA{}}}$, however, the increase is more linear. At low thicknesses the different iron layers appear to not be forming in a set ratio but in an equilibrium between each type of iron site.

Apart from the sample with the very thickest iron layers ( $\left[ \mathrm{U}(22)/\mathrm{Fe}(180) \right]_{5}$, which will be discussed later) the amount of UFe$_{2}$ and non-crystalline iron is relatively constant once there is sufficient iron. Although smaller in the thinner iron layers, as there is less room for their growth, there appears to be a maximum of $\approx18\ensuremath{\unskip\,\mathrm{\AA{}}}$ of UFe$_{2}$ and $\approx20\ensuremath{\unskip\,\mathrm{\AA{}}}$ for the diffuse iron sites. It is not possible to determine from these spectra whether the UFe$_{2}$ and diffuse iron form one on top of the other or whether there is interdiffusion of iron with both uranium and UFe$_{2}$.

Normal CEMS spectra cannot give any information about the relative depths of particular iron sites and so the spectra presented so far cannot determine whether the paramagnetic doublet is from a thin layer on the surface of the sample or whether it is distributed throughout the sample thickness as in the proposed model of UFe$_2$ forming at U/Fe interfaces. To test the model a DCEMS experiment was run on one sample, $\left[ \mathrm{U}(28)/\mathrm{Fe}(43) \right]_{31}$, with one spectrum recorded from approximately 5 bilayers at the top of the sample, and one spectrum recorded from the remaining layers underneath. These spectra are shown in Figure 7.12.

Figure: DCEMS spectra recorded from the $\left[ \mathrm{U}(28)/\mathrm{Fe}(43) \right]_{31}$ sample. Spectrum (a) is obtained from the top $\sim 5$ bilayers in the sample, whilst spectrum (b) was recorded from the remaining layers underneath.

The first conclusion to be drawn is that the paramagnetic doublet is present in both spectra with a substantial intensity. This iron site is thus distributed throughout the sample thickness and supports the model of UFe$_2$ forming at the U/Fe interfaces. The relative intensity of the ordered iron sites and the paramagnetic sites are slightly different in the two spectra. The ratio of UFe$_2$/Fe is 3:2 in the top layers and 2:3 in the lower layers. This may be due to slight differences in growth conditions during sample growth or the central doublet in the top layers may have a small contribution from a surface iron oxide contaminant, such as FeO which is paramagnetic at room temperature.

The relative areas of the components in the spectrum from the sample with the thickest iron layer, $\left[ \mathrm{U}(22)/\mathrm{Fe}(180) \right]_{5}$, give disproportionately large amounts of UFe$_{2}$ and non-crystalline iron compared to the trends seen in the rest of the series. As the iron layer thickness is increased the relative areas of these two components should decrease compared to the crystalline iron. Although the iron layer is very thick the uranium layer is the thinnest of all the samples. At $22\ensuremath{\unskip\,\mathrm{\AA{}}}$ the uranium layer is at the lower limit necessary to deposit homogenous uranium films and hence produce well defined layers. If the layer is not well defined this will be propagated into the iron layers deposited upon it, increasing the thickness of iron necessary to reach a state where fully crystalline growth can start. This explains the larger than expected thickness of the non-crystalline component. This badly defined uranium layer could also make it easier for iron atoms to penetrate during the sputtering process, increasing the probability of forming UFe$_{2}$. The multilayer thus consists of thick layers of crystalline iron between disproportionately large amounts of non-crystalline iron and UFe$_{2}$ and inhomogenous layers of uranium.

The lack of well defined interfaces is supported by specular x-ray reflectivity measurements taken by A.Herring. In Figure 7.13 a reflectivity scan of the $\left[ \mathrm{U}(22)/\mathrm{Fe}(180) \right]_{5}$ sample is compared with one from $\left[ \mathrm{U}(42)/\mathrm{Fe}(113) \right]_{21}$, a sample which Mössbauer shows to have less interdiffusion at the interfaces.

Figure: X-ray reflectivity scans for a) $\left[ \mathrm{U}(42)/\mathrm{Fe}(113) \right]_{21}$ and b) $\left[ \mathrm{U}(22)/\mathrm{Fe}(180) \right]_{5}$ samples. Sharp peaks indicate well defined interfaces.

It can be seen that the reflectivity scan from $\left[ \mathrm{U}(42)/\mathrm{Fe}(113) \right]_{21}$ has many sharp peaks. This is indicative of well defined interfaces[47]. The scan from the $\left[ \mathrm{U}(22)/\mathrm{Fe}(180) \right]_{5}$ sample shows no sharp peaks, indicating either very diffuse or very rough interfaces^7.1.

All recorded spectra show that the iron spins are entirely within the plane of the sample, as expected from the shape anisotropy of a thin film. There appears to be no appreciable perpendicular magnetic anisotropy of any form in the system as might be expected from the coupling of the uranium orbital moment to the lattice if it possessed a magnetic moment. It should be noted however that in the analogous Ce/Fe multilayers perpendicular magnetic anisotropy was only observed for iron layer thicknesses less than $15\ensuremath{\unskip\,\mathrm{\AA{}}}$.[3] This result gives no indication of there being any significant magnetic moment on the uranium atoms at the U/Fe interfaces in the studied samples.^7.2