Journal of Magnetism and Magnetic Materials 240 (2002) 497­500 Effect of interface roughness on magnetic multilayers of Fe/Tb and Fe/Cr Amitesh Paul* Institut f.ur Festk.orperforschung, Forschungszentrum J.ulich GmbH, D-52425 J.ulich, Germany Abstract The effect of systematic variation in the correlated interface roughness on perpendicular magnetic anisotropy (PMA) and giant magnetoresistance (GMR) has been studied in Fe/Tb and Fe/Cr multilayer systems, respectively. Multilayers for each system were deposited simultaneously on a set of float glass substrates pretreated with varying rms surface roughness. In both the systems the amount of intermixing at the interfaces and other morphological parameters are found similar, thus allowing one to separate out the effect of interface roughness only. X-ray reflectivity, diffuse scattering, conversion electron M.obbauer spectroscopy and superconducting quantum interference device magneto- metry are used to characterise the systems. With the increase in s; the PMA in Fe/Tb as well as the GMR in Fe/Cr shows a small decrease. The observed effects are mainly due to the changes in the correlated part of the roughness of the multilayers, while the uncorrelated part of the s of different multilayers are expected to remain similar. r 2002 Elsevier Science B.V. All rights reserved. 1. Introduction [2]. Therefore, in the present study, MLs for both the systems of Fe/Tb and Fe/Cr are deposited on substrates Magnetic multilayers (MLs) showing properties like pretreated with varying surface roughnesses. It has been perpendicular magnetic anisotropy (PMA) in systems seen that except for the interface roughnesses, other like Fe/Tb MLs or giant magnetoresistance (GMR) in microstructural features of the ML like grain size, Fe/Cr MLs are significantly affected by their interfacial coherence length, grain texture, intermixing at the structures [1,2]. interface, internal stresses etc. are similar, thus allowing Earlier studies on Fe/Tb MLs, to see the effect of one to selectively study the effect of interface structure interfacial modifications on PMA, were mainly done by (varied systematically) on PMA in Fe/Tb and on GMR post-deposition treatments like thermal annealing [3] or in Fe/Cr MLs. ion irradiation [1]. However, the induced effects include changes in geometrical roughness as well as intermixing/ demixing at the interface. Therefore, it has not been possible to separate the effects of interface roughness (s) 2. Experimental details from that of intermixing/demixing [1]. Experimental results on the effect of s on GMR are also conflicting. It Substrates with varying surface roughness were has been seen that depending upon the ratio of the spin prepared in two sets by etching the float glass (FG) asymmetry for the interface and bulk scattering and the substrates in dilute HF for different periods of time. various techniques used to modify the interfaces there is Set1: Eight substrates with increasing etching times of 0, either an increase or decrease of GMR with roughness 15, 30, 60, 90, 120, 150 and 180 s, designated as S1­S8, respectively, were taken. The multilayer consisted of 20 bilayers of composition 3.0 nm Fe/2.0 nm Tb, were *Fax: +49-2461-61-4443. deposited on FG substrates. Set2: A set of substrates E-mail address: a.paul@fz-juelich.de (A. Paul). were prepared for 14 different etching times which are 0304-8853/02/$ - see front matter r 2002 Elsevier Science B.V. All rights reserved. PII: S 0 3 0 4 - 8 8 5 3 ( 0 1 ) 0 0 9 1 2 - X 498 A. Paul / Journal of Magnetism and Magnetic Materials 240 (2002) 497­500 numbered as S1­ S14 and which show similar results as roughness with increasing etching time. The fitting of that of Set1. In this set MLs consisted of the following the reflectivity and rocking curve patterns for the deposition sequence: substrate/Cr (10.0 nm)/[Fe substrates, done by simulations following theories (3.0 nm)/Cr (1.2 nm)] 20/Fe (5.0 nm). Deposition con- [4,7], gives the value of s; x (lateral correlation length) ditions are similar as reported in Refs. [2,3] B450750 nm and h (Hurst parameter measuring A powder X-ray diffractometer model D5000 of jaggedness)=0.270.1. Siemens with Cu Ka radiation was used to measure CEMS spectra for the MLs are fitted with two the specular (XRR) and diffuse scattering geometry subspectra: one sharp (a-Fe) and other broad one (XDS) [4]. 57Fe conversion electron M.obbauer Spectro- corresponding to the Fe atoms at the interface. scopy (CEMS) was used to get information about the Intermixed layer thickness inferred from the area under intermixing at the interface and the PMA at room the sharp sextet shows no significant increase temperature using a gas flowing (95% He, 5% CH4) (1.070.3 nm) with etching time and is essentially being proportional counter. The spectral profiles were ana- used as an input parameter in XRR curve fitting. The lysed by means of the NORMOS code developed by probability of hyperfine field distribution PðBhfÞ and the Brand [5]. The magnetic texture of the sample is revealed average /BhfSðTÞ was also similar. Fig. 1 shows the by the intensity of the 2nd and 5th peaks relative to the specular (subtracted off the off-specular) X-ray patterns inner ones of the M.obbauer spectrum. RF SQUID for the specimens. The patterns clearly show the first- measurements were done at 4.2 K (QUANTUM DE- order Bragg peak due to ML periodicity and a distinct SIGN model MPMSR2) with the field being in the film oscillatory variation of the s with increasing etching plane and the ratio of magnetic remanence Mr and the time. The values of the substrate (ss) and the interface magnetic saturation Ms was used to infer the extent of roughness (si) for the MLs with sample nos. S1, S2, S4, antiferromagnetic coupling fraction (AFF) given by S5 and S6 are also given with the figure. The similar (1 Mr=Ms) [6]. increment in s2 (s2i2s2) from substrate to the ML interfaces signifies that the change in roughness by substrate roughness variation is only affecting the correlated part of the roughness of the MLs while the 3. Results and discussion uncorrelated part of the roughness remains unaffected. The parameters could not be extracted for the sample 3.1. Fe/Tb MLs nos. S3, S7 and S8 as the intensity of the off-specular scans is comparable to that of the specular scans, which The XRR pattern of the float glass substrates of Set1 also signifies that the peak at the first Bragg position for and Set2 subjected to different etching time (see Ref. [2]) the specimens with higher substrate roughness is arising show distinct oscillatory variation of the substrate due to the correlated part of the roughness [8] only. The experimental S 1 theoretical s= 0.6 5 ±0.0 5 n m i= 1.7 0 ± 0.0 5 n m S1 = 0 .9 4 5 0 S 2 s= 1 .2 0 n m i = 1 .9 0 n m ~ 3 0 0 n m c S 4 (± 5 0 n m ) s= 1 .9 5 n m i = 2 .4 5 n m = 0 .9 7 5 0 S 5 s= 1 .1 5 n m Log intensity (arb.units) i = 1 .7 0 n m ~ 5 0 n m u S 6 s= 1 .2 5 n m (± 1 0 n m ) i = 1 .9 5 n m -0 .8 0 .0 0 .8 0 .2 0 .4 0 .6 0 .8 1 .0 1 .2 - (deg.) Incident angle (deg) Fig. 1. XRR scans of [Fe(3.0 nm)/Tb(2.0 nm)] 20 multilayers along with their fit deposited on FG substrates with different etching times. The substrate roughness (ss) and interface roughness (si) are shown. The inset shows the transverse (o) scan for S1 along with the fit at two different angles of y corresponding to the position at the Bragg peak and at an off-set to it. At o2y the specular peak is seen over a diffuse background. For clarity, various curves are shifted relative to each other along the y-axis. A. Paul / Journal of Magnetism and Magnetic Materials 240 (2002) 497­500 499 fit to the rocking curve (inset for S1) shows the correlation length for the correlated and uncorrelated 4.0 part of s as xcB300 nm and xuB50 nm with h ¼ 0:570:02 and do not change with sU The polycrystalline 3.5 nature of the Fe layer in the Fe/Tb MLs is confirmed for 3.0 all the specimens from the XRD measurements. GMR (%) The angle f (73.01) (between the film normal and the 2.5 average direction of the magnetic moments) as obtained 3.0 from the fit of the room temperature CEMS spectra of 2.0 2.5 four representative samples S1, S4, S5, and S8 is B421 [9]. A small decrease in PMA though may be observed 2.0 for sample no. S8 (fB541) whose roughness is compar- able with the thickness of the layer. It may be noted Substrate 1.5 roughness (nm) from an earlier study [1] that a small change in s (B0.45 nm) by 80 MeV Si ion-irradiation causes the 1.0 angle f to decrease by 6.31, whereas in the present case 0 200 400 600 800 the roughness has been increased by 2.5 nm resulting in a Time of etching (s) similar change in f: The intermixed layer thickness has Fig. 2. The plot of surface roughness of the FG substrates and also been found to increase (upon 150 MeV Ag ion- the corresponding GMR ratio as obtained from the fit to the irradiation [1]) or decrease (by thermal annealing [3]) XRR data and the magnetoresistance measured. The arrows causing the PMA to decrease largely. Therefore, in both indicate the points for variation maximum/minimum in rough- post-deposition treatments where the s variation is ness/GMR. expected to be uncorrelated is seen to be associated with a possible stress relaxation within the bulk of the layers causing the PMA to decrease but a correlated change in roughness does not affect the PMA significantly. 0.70 3.2. Fe/Cr MLs 6 ) 0.65 s /M In Set2 the substrate roughness variation, the s and x r ) -M (from XRR and XDS) and the intermixed layer s/M (1 thickness (from CEMS) of the MLs are found to behave r 0.60 5 ) / -M similarly as in Set1. XRD measurements have shown (1 (%R that the structural coherence length (z), grain size, M 0.55 G internal stresses and the texture (1 1 0) do not vary from 4 sample to sample. Furthermore, since all the films were deposited simultaneously, the deposition conditions like 0.50 deposition rate and substrate temperature are identical 1.0 1.5 2.0 2.5 3.0 Substrate roughness (nm) for all the specimens; therefore, the individual layer thicknesses as well as the density of defects in the bulk of Fig. 3. The change in AFF (') with increase in substrate the layers is expected to be similar. roughness as obtained from SQUID measurements. Also Thus, the only difference between various MLs shown is the GMR normalised to AFF (") with increasing deposited on different substrates is in their s; and the roughness. observed variation in GMR can solely be attributed to the variation in the s: The GMR ratio is defined as ðR roughness in all the MLs is expected to be similar in 0 RsÞ=Rs 100ð%Þ; with R0 and Rs being, respec- tively, the resistance values at zero and saturating fields. magnitude because of the identical conditions of It is interesting to note that with increase in etching time deposition. The AFF showing a saturating behaviour as shown in Fig. 2, the variation in GMR is highly with increase in roughness is plotted in Fig. 3. Normal- correlated with that in the roughness. The difference in ising the GMR (%) ratio with AFF gives the contribu- the interfacial roughness in different MLs is essentially tion due to the interfacial scattering with the increase in due to the difference in the roughness of their substrate roughness, which is also plotted. One may see from the which is transmitted to the successive layers. Thus, the figure that while the decrease in the AFF is B20%, the difference among various MLs is expected to be in their interfacial scattering alone can bring B40% decrease in correlated part of the interfacial roughness (similarly as GMR ratio for a change of B70% in correlated in case of PMA in Fe/Tb MLs). The uncorrelated roughness in a range of few nm. This change in GMR 500 A. Paul / Journal of Magnetism and Magnetic Materials 240 (2002) 497­500 is smaller as compared to B65% decrease due to References 200 MeV Ag ion irradiation effects as observed in an earlier study [10]. The interface structure modification [1] A. Gupta, R. Amitesh Paul, D.K. Gupta, G. Avasthi, due to ion irradiation effects are expected to be Principi, J. Phys. Condes. Mater. 10 (1998) 9669 and uncorrelated and thus a small increase in roughness references therein. can cause a large decrease in GMR. [2] A. Gupta, Amitesh Paul, S.M. Chaudhari, D.M. Phase, J. In conclusion it has been seen that keeping all other Phys. Soc. Jpn. 69 (2000) 2182 and references therein. parameters unchanged a large change only in the [3] Amitesh Paul, A. Gupta, J. Alloys Compounds 326 (2001) 246. correlated part of the interface roughness can be caused [4] D.E. Savage, J. Kleiner, N. Schimke, Y.H. Phang, T. by etching the substrates for different periods of time. Jankowski, J. Jacobs, R. Kariotis, M.G. Lagalley, J. Appl. This correlated variation is expected to have a smaller Phys. 69 (1991) 1411; effect on the RE­TM bonds at the interface of Fe/Tb D.K.G. de Boer, Phys. Rev. B 49 (1994) 5817. systems or on the interfacial scattering in Fe/Cr MLs [5] R.A. Brand, Nucl. Instrum. and Methods B 28 (1987) 398. compared to the uncorrelated changes at the interfaces [6] R. Schad, P. Beli.en, G. Verbanck, V.V. Moshchalkov, Y. caused by other post-deposition treatments as ion Bruynseraede, H. Fisher, S. Lefebvre, M. Bessiere, Phys. irradiation, thermal annealing and in situ modification Rev. B 59 (1999) 1242. of roughness. Thus a small decrease in PMA and in [7] L.G. Parratt, Phys. Rev. 95 (1954) 359. GMR is observed with increase in roughness. [8] A. Gupta, Amitesh Paul, S. Mukhopadhyay, Ko Mibu, J. Appl. Phys. 90 (2001) 1237. [9] Amitesh Paul, Ajay Gupta, Prasanna Shah, K. Kawaguchi, Hyperfine Interaction 2001, in press. Acknowledgements [10] A. Amitesh Paul, S.M. Gupta, D.M. Chaudhari, Phase, Vacuum 60 (2001) 401. The work was done at Inter-University Consortium for DAEF (Indore), India.