Atomic and Magnetic Structure of the Interface in Multilayers

Hyperfine Interactions 141/142: 13–20, 2002.

© 2002Kluwer Academic Publishers. Printed in the Netherlands. 13

Atomic and Magnetic Structure of the Interface in Multilayers

J. BALOGH, D. KAPTÁS, T. KEMÉNY, L. F. KISS, T. PUSZTAI and I. VINCZE

Research Institute for Solid State Physics and Optics, P.O. Box 49, 1525 Budapest, Hungary

Abstract. Temperature dependence of the magnetic properties of Fe/Ag vacuum evaporated multi- layers was studied in a wide range of layer thickness. For Fe thickness larger than 1 nm continuous magnetic layers can be found, but its hyperfine field is significantly lower than that of pureα-Fe at elevated temperatures. It is attributed to a decrease of the Curie temperature due to Ag impurities in the Fe layer. Below 1 nm Fe thickness magnetic relaxation and the formation of a granular alloy with 35 T average hyperfine field was observed. Magnetoresistance results indicate the presence of Fe clusters in the Ag matrix, as well.

Key words:multilayer, diffusion amorphization, superparamagnetic relaxation.

Mössbauer spectroscopy is widely used to study chemical mixing at the interface of different phases. The width of the interdiffused region can be estimated from the intensity of the appropriate spectral components or by a systematic study of the hyperfine parameters as a function of the layer thickness of the constituents [1]. It is expected that the interface structure does not change any more when the layer thickness exceeds the characteristic length for interdiffusion of the constituents.

This approach is useful when the hyperfine parameters of the interface and the layer material strongly overlap.

Hyperfine parameters of epitaxial Fe/Ag multilayers have been extensively stud- ied [2–4] in the past, but the assignment of the different possible Fe sites of the interface is still contradictory. It was generally supposed that the interface is sharp and the observed hyperfine field distributions are related to Fe atoms sitting at step edges of atomic terraces [3]. On the other hand magnetic relaxation effects were observed in case of thin Fe layers [5], which was explained [6] by an island struc- ture of these films. Besides magnetic relaxation the hyperfine field distribution can originate from atomic scale mixing of Fe and Ag under non-equilibrium conditions, which was recently raised by several experiments [7–9]. To elucidate the respective role of chemical mixing and magnetic relaxation in the observed hyperfine field distributions we performed temperature dependent studies on Fe/Ag multilayers over a wide range of layer thickness, as a function of temperature.

The Fe/Ag multilayer samples were prepared by electron beam evaporation of the elements in a vacuum of 10⁻⁷ Pa. The samples prepared contained Fe layers of nominal thickness, dFe = 0.2, 0.7, 1.4, 2.8, 5.6 and 11.2 nm with various Ag thickness, which was controlled by a vibrating-quartz oscillator. The top and bot-

14 J. BALOGH ET AL.

Figure 1. Room temperature Mössbauer spectra of the indicated multilayer samples. The upper one is fitted with a single six-line pattern, while the bottom fit contains 5 six-line patterns and a doublet belonging to impurities in the Al substrate.

tom layers were Ag and the overall thickness was between 200 and 300 nm. The substrate was silicon single-crystal or Al foil at room temperature. In many case the sample could be removed from the silicon substrate with a scotch tape and transmission Mössbauer measurements could be performed between 4.2 K and 800 K. The atomic structure of the samples were examined by X-ray diffraction and reflectometry. The magnetisation measurements were performed by a Quantum Design MPMS-5S SOUID magnetometer with a maximum field of 5 T.

X-ray reflectivity peaks were observed on samples with Fe layer thicknessdFe 1.4 nm. The periodicity calculated from the reflectivity peaks agrees with the nom- inal layer thickness within 10 to 15 percent. Two Mössbauer spectra characteristic to this thickness range is shown in Figure 1. The 3:4:1:1:4:3 relative line intensities indicate a perfect alignment of the magnetic moments in the sample plane. For dFe > 2.8 nm the spectrum can be fitted with a slightly broadened ( = 0.05 mm/s fordFe = 5.6 nm) six-line pattern and its hyperfine field (hf) at 12 K agrees well with that of bcc-Fe. However, the temperature dependence of the hf significantly deviates from the bulk behaviour, as it is shown in Figure 2 for

INTERFACE IN MULTILAYERS 15

Figure 2. Temperature dependence of the average hyperfine field of the indicated samples. Estimated Curie temperatures are shown in the top corner.

dFe =11.2, 5.6 and 2.8 nm. The gradual heat treatment of the samples during the measurements above 500 K results in the disappearance of the line broadening and an increase of the hf to the bulk value. This behaviour can be explained by a lower Curie temperature of the as received multilayer samples. The Curie temperatures estimated from the data points below 500 K are indicated in Figure 2. Similar results on thinner layers were interpreted as due to the quasi two dimensional nature of the layers [10] and the temperature dependence of the hf was also dependent on the Ag thickness. In our experiment the temperature dependence of the hyperfine field does not seem to depend on the Ag thickness and the lower Curie temperature is rather attributed to Ag impurities in the Fe layers. A few percent dissolved Ag impurity may certainly explain the estimated decrease of the Curie temperature.

Spectral components significantly different from the bulk value at low tem- peratures can only be found for samples withdFe 2.8 nm. Room temperature spectrum of the [1.4 nm Fe/2 nm Ag]60sample with the fitted subspectra is shown in Figure 1. The same parameters could be fitted for a [1.4 nm Fe/4 nm Ag]60

sample and these are in good agreement with those of [3] for a Fe/Ag trilayer containing about seven monolayers of Fe on a thick Ag(001) single crystal.

No layer structure could be observed by X-ray reflectivity for samples with dFe 0.7 nm. The Mössbauer spectra of two representative samples with dif- ferent Ag thickness are shown in Figure 3 for different temperatures. The loss of continuous layers can be correlated to a decrease of the 2–5 line intensities as compared to the previous samples. The average hf at 12 K and at 300 K as a function of the average Ag concentration is shown in Figure 4, together with the

16 J. BALOGH ET AL.

Figure 3. Mössbauer spectra of Fe/Ag granular samples with the indicated nominal multilayer structure at different temperatures.

Figure 4. Average hyperfine fields and isomer shift with respect to α-Fe for the samples with dFe = 0.7 nm at 12 K (full circles) and at 300 K (open circles) and withdFe = 0.2 nm at 4.2 K (stares) and at 300 K (crosses) as a function of the average concentration.

INTERFACE IN MULTILAYERS 17

Figure 5. (a) X-ray diffraction pattern of the [0.2 nm Fe/0.6 nm]₃₂₄, (b) the [0.2 nm Fe/2.6 nm Ag]₁₆₂with the indicated angle of incidence and (c) that of the [0.2 nm Fe/2.6 nm Ag]₁₆₂sample measured in Bregg–Brentano geometry.

average isomer shifts relative toα-Fe. At low temperature the average hf is close to 35 T in the whole concentration range, but an increase of the width of the hf distribution with increasing Ag content can be observed. At 300 K the magnetic splitting sharply depends on both nominal thickness. FordFe =0.7 nm the higher the Ag concentration the lower the average hf and the larger the width of the dis- tribution. FordFe =0.2 nm the sample is paramagnetic at high Ag concentration.

The concentration dependence of the average parameters is very similar to that observed on co-deposited granular alloys [11] where the increased low temperature hf and the apparently lower Curie temperature was attributed to the formation of non-equilibrium fcc Fe1−xAgxalloy.

X-ray diffraction pattern of the two samples with dFe = 0.2 nm is shown in Figure 5. In this thickness range the variation of the silver concentration pro- duces highly different sample morphology. The [0.2 nm Fe/0.6 nm]324 sample

18 J. BALOGH ET AL.

Figure 6. Magnetisation of the [0.2 nm Fe/0.6 nm]₃₂₄ sample measured both in parallel and per- pendicular applied field at 5 K and 300 K. The full line in the bottom panel is the fitted curve, as explained in the text.

(Figure 5(a)) exhibits broad peaks. The estimated average grain size is about 4 nm.

The [0.2 nm Fe/2.6 nm]162sample (Figure 5(b)) shows relatively narrow diffraction lines and very low intensity around 2θ ≈45^◦, where the fcc (200) and the bcc (110) reflections are to be found. Measurement in the Bregg–Brentano (θ–2θ) geometry (Figure 5(c)) shows that it can be attributed to a (111) preferred orientation of the Ag layers and a small grain size of the bcc granules if at all they are present in the sample.

The low temperature Mössbauer spectra are also different, as shown in Figure 3.

Though the average saturation hyperfine fields are similar, the higher Ag concen- tration is reflected in significantly larger isomer shifts, as it is shown in Figure 4, and the average quadrupole splitting also increases. Strong correlation of these parameters results in the significant asymmetry of the spectrum lines. It is worth

INTERFACE IN MULTILAYERS 19 while to note that similar Mössbauer spectrum was observed on ultrathin Fe(100) films on Ag(100) prepared by molecular beam epitaxy [12].

The temperature dependence of the spectra shown in Figure 3 indicates super- paramagnetic behaviour with blocking temperature TB ≈ 50 K for the [0.2 nm Fe/2.6 nm Ag]162 sample. This was also confirmed by magnetisation measure- ments [13]. The [0.2 nm Fe/0.6 nm]324 sample shows magnetic splitting in the whole temperature range, but a small paramagnetic component appears already at 12 K and the width of the hyperfine field distribution increases significantly with increasing temperature. The magnetisation measurements support the presence of a superparamagnetic component, as shown in Figure 6. The applied field dependence of the magnetisation at 300 K could be fitted as the sum of a ferromagnetic compo- nent and a Langevin function with 750µBcluster moment, which can be attributed to an about 2 nm grain size. This agrees well with the value for the [0.2 nm Fe/2.6 nm Ag]162sample [13].

The granular behaviour was also examined by magnetoresistance measurements [14] which is a sensitive tool to indicate the presence of magnetic clusters. All the samples exhibited giant magnetoresistance (GMR) behaviour, i.e., negative magne- toresistance independently from the angle between the measuring current and the applied magnetic field. The non-saturating negative magnetoresistance observed on a 25 nm thick Fe layer covered on both sides by Ag [14] was explained by the presence of small Fe clusters in the Ag layer.

In conclusion we observed that the magnetic behaviour of vacuum evaporated Fe/Ag multilayers drastically changes around 1 nm Fe layer thikness. FordFe 1.4 nm the magnetic layers are continuous and have a decreased Curie temperature.

In this range the magnetic properties are not influenced by the Ag layer thickness.

FordFe <1.4 nm the magnetic behaviour is similar to that found in granular alloys and the temperature dependence of the hyperfine parameters strongly depend on the Ag thickness. Superparamagnetic behaviour was indicated both by magnetisation and Mössbauer measurements. The results show that the grain size of the magnetic granules can be varied on the nanometer scale by appropriate layer deposition.

The role of magnetic relaxation for dFe 1.4 is to be further studied, but the magnetoresistance results indicate that small magnetic clusters are formed at the interface in case of thick continuous Fe layers, as well.

Acknowledgement

This work was supported by the Hungarian Research Fund (OTKA T034602 and T030753).

References

1. Balogh, J., Bujdosó, L., Kemény, T., Pusztai, T., Tóth, L. and Vincze, I.,Appl. Phys. A65 (1997), 23.

20 J. BALOGH ET AL.

2. Freeland, J. W., Keavney, D. J., Storm, D. F., Grigorov, I. L., Walker, J. C., Pini, M. G., Politi, P.

and Rettori, A.,Phys. Rev. B54(1996), 9942.

3. Schurer, P. J., Celinski, Z. and Heinrich, B.,Phys. Rev. B51(1995), 2506.

4. Liu, G. and Gradmann, U.,J. Magn. Magn. Mater.118(1993), 99.

5. Volkening, F. A., Jonker, B. T., Krebs, J. J., Prinz, G. A. and Koon, N. C.,J. Phys.49(1988), C8-1699.

6. Morup, S., Christiansen, G. and Koon, N. C.,J. Magn. Magn. Mater.104–107(1992), 1793.

7. Neubauer, M., Lieb, K. P., Schaaf, P. and Uhrmacher, M.,Phys. Rev. B53(1996), 10237.

8. Bürgler, D. E., Schmidt, C. M., Schaller, D. M., Meisinger, F., Hofer, R. and Güntherodt, H.-J., Phys. Rev. B56(1997), 4149.

9. Scholl, A., Baumgarten, L. and Ebhardt, W.,Phys. Rev. B56(1997), 747.

10. Gutierrez, C. J., Qiu, Z. Q., Tang, H., Wieczorek, M. D., Mayer, S. H. and Walker, J. C.,Phys.

Rev. B44(1991), 2190.

11. Kataoka, N., Sumiyama, K. and Nakamura, Y.,Trans. Jap. Inst. Met.26(1985), 703.

12. Koon, N. C., Jonker, B. T., Volkening, F. A., Krebs, J. J. and Prinz, G. A.,Phys. Rev. Lett.59 (1987), 2463.

13. Kiss, L. F., Balogh, J., Bujdosó, L., Kaptás, D., Kemény, T., Pusztai, T. and Vincze, I., J. Metastable and Nanocrystalline Materials10(2001), 505.

14. Balogh, J., Gábor, A., Kaptás, D., Kiss, L. F., Csontos, M., Halbritter, A., Kézsmarki, I. and Mihály, G., In:Nato Advanced Research Workshop, Size Dependent Magnetic Scattering, Pécs, Hungary, May 28–June 1, 2000.