3B2v7:51c ED:Chanakshi=brr GML4:3:1 MAGMA : 8451 Prod:Type:com pp:124ðcol:fig::NILÞ PAGN: chris SCAN: nil ARTICLE IN PRESS 1 3 Journal of Magnetism and Magnetic Materials ] (]]]]) ]]]­]]] 5 7 Epitaxial growth, alloying and magnetic structure of interfaces 9 in Fe/Cr (0 0 1) superlattices 11 V. Uzdina, W. Keuneb,*, M. Walterfangb 13 a St. Petersburg State University, ICAPE, 14 linia V.O. 29, 199178, St. Petersburg, Russia b Laboratorium f.ur Angewandte Physik, Gerhard-Mercator-Universit.at Duisburg, Lotharstr. 65, D-47048 Duisburg, Germany 15 17 Abstract 19 Fe/Cr(0 0 1) superlattices containing two-monolayers thick 57Fe probe layers at the Fe/Cr (Fe-on-Cr) or Cr/Fe (Cr- 21 on-Fe) interfaces were studied using conversion electron M.ossbauer spectroscopy (CEMS). For the interpretation of the CEMS data of superlattices annealed at different temperatures, we performed theoretical modeling of their chemical 23 and magnetic structure. Roughness and interface alloying were introduced to the model by algorithms of epitaxial growth, which included ballistic deposition with subsequent floating of some atoms on the surface. Self-consistent 25 calculations of magnetic moments within the periodic Anderson model confirmed the proportionality between hyperfine fields and magnetic moments. For the explanation of the evolution of CEM spectra versus annealing temperature, 27 the difference in the melting points of bulk Fe and Cr has to be taken into account. r 2001Published by Elsevier Science B.V. 29 Keywords: M.ossbauer spectroscopy; Multilayers; Annealing; Magnetic structure; Fe­Cr 31 33 In Fe/Cr multilayers antiferromagnetic interlayer recent calculations of magnetic moments in Fe/Cr 57 35 exchange coupling and giant magnetoresistance multilayers with interface alloying [4] did not confirm (GMR) were discovered for the first time. Despite the the assumption that a large spectral intensity of the 20-T 59 37 large efforts which were undertaken for studying the satellite means that the interface is atomically smooth. correlation between their interface structure and macro- On the contrary, our results showed that this peak 61 39 scopic magnetic and transport properties, many aspects corresponds to short-range interdiffused Fe atoms inside are still contradictory. In particular, the value of the the Cr spacer layers, but not far away from the interface. 63 41 GMR effect was found to correlate with the spectral Therefore, its relative spectral intensity has to increase contribution (spectral area) of the satellite Zeeman upon interface alloying. According to our recent 65 43 sextet in the conversion electron M.ossbauer (CEM) findings, earlier conclusions about the role of interface spectra [1] that corresponds to a hyperfine field, Bhf; of and bulk scattering in the GMR effect should be revised. 67 45 20 T. According to the traditional interpretation for Fe/ We emphasize in this context that the interpretation of Cr multilayers, the peak at 20 T in the hyperfine field experimental data is an ambiguous problem, and for the 69 47 (hff) distribution, PðBhfÞ; originates from Fe atoms at understanding of real mechanisms of epitaxial growth the ideally flat (0 0 1) interface with four nearest and interface alloying in Fe/Cr multilayers, experimental 71 49 neighbours 73 51 75 53 *Correspo UNCORRECTED PROOF and one second neighbour Cr atoms [1­3]. and theoretical studies have to be used together to This interpretation seems to allow the correlation of achieve self-consistency in details. transport properties, interface roughness and the ratio Here, we present results of a CEMS investigation after between interface and bulk scattering [1]. However, our isochronal annealing of Fe/Cr(0 0 1) superlattices with two-monolayers (ML) thick 57Fe probe layers (95% nding author. Tel.: +49-203-379-2387; fax: +49- enriched) placed either at Fe/Cr interfaces (Fe-on-Cr or 77 55 203-379-3601. ``lower'' interfaces) or at Cr/Fe interfaces (Cr-on-Fe or E-mail address: keune@uni-duisburg.de (W. Keune). ``upper'' interfaces), and of theoretical modeling of 79 0304-8853/01/$ - see front matter r 2001Published by Elsevier Science B.V. PII: S 0 3 0 4 - 8 8 5 3 ( 0 1 ) 0 0 8 8 8 - 5 MAGMA : 8451 ARTICLE IN PRESS 2 V. Uzdin et al. / Journal of Magnetism and Magnetic Materials ] (]]]]) ]]]­]]] 1 Table 1 57 Position of the fitted Gaussian peaks (satellites) in the hff distribution (first column) and their relative spectral area (in %) for the 3 sample with the 57Fe probe layers at the upper (Cr-on-Fe) and lower (Fe-on-Cr) interfaces after isochronal annealing at different 59 temperatures 5 B 61 hf (T) RT 2001C 3001C 4001C 4501C 5001C 5501C 7 Upper interface: Cr on Fe 63 33.7(%) 17.8 21.6 21.9 22.2 23.3 21.4 22.5 9 31.3(%) 16.1 13.3 12.8 12.1 12.5 20.2 28.3 65 28.5(%) 13.9 13.9 14.7 16.3 14.0 17.1 30.4 25.6(%) 11.3 12.9 12.0 12.6 13.0 10.6 11 23.1(%) 11.3 12.0 13.1 10.1 11.2 8.5 67 20.3(%) 15.1 11.1 10.0 9.9 11.2 11.9 15.5 13 16.5(%) 12.1 12.4 13.9 12.2 8.2 69 15 Lower interface: Fe on Cr 71 33.8(%) 29.6 30.7 30.4 32.3 31.3 30.0 25.6 17 31.4(%) 15.8 13.9 15.3 14.5 14.9 18.1 21.9 73 28.7(%) 11.3 12.9 13.1 13.9 13.2 15.1 17.2 19 25.5(%) 10.9 11.6 10.5 9.8 9.7 10.1 13.6 75 23.0(%) 10.4 9.7 9.6 8.9 11.1 7.2 21 20.0(%) 10.7 11.1 11.9 13.7 10.4 11.1 17.4 77 23 79 25 interface alloying in these systems during the sample to the sum of the areas of the fitted Gaussian peaks) 81 preparation and annealing process. obtained for each annealing temperature is only slightly 27 Superlattices of composition MgO(0 0 1)/Cr(50 (A)/ less than 100% of the total experimental spectral area 83 [57Fe(2 ML)/natFe(12 ML)/Cr(8 ML)]15 (lower interfa- according to Table 1. 29 ces) or MgO(0 0 1)/Cr(50 (A)/[natFe(12 ML)/57Fe(2 ML)/ The hff distributions for the two types of interfaces 85 Cr(8 ML)]15 (upper interfaces) were epitaxially grown prove to be remarkably different (Table 1). For the Fe- 31 at Ts ¼ 433 K by ultrahigh-vacuum (UHV) depo- on-Cr interface, the relative area of the ``bulk'' peak 87 sition of the metals on epipolished MgO(0 0 1) sub- (near 33.8 T) is found to be about 30%, whereas for the 33 strates, as described in detail in Ref. [4]. (natFe=Fe Cr-on-Fe interface it is only about 20%. Other satellite 89 metal of natural 57Fe abundance, 2.1%). Characteriza- peaks are narrower and generally yield less contribution 35 tion by low and high angle X-ray diffraction demon- (except for the peak near 31.4 T) to the total spectrum 91 strates the high-quality superlattice structure of our for the lower interface as compared with the upper 37 samples [4]. interface. The largest difference was detected for the 93 CEM spectra were measured at room temperature as peak near 20 T: before annealing it contributes less than 39 described in Ref. [4]. Typical CEM spectra and hff 11% of the total spectral area for the lower interface and 95 distributions before annealing of samples with either more than 15% for the upper one. An additional peak 41 ``upper'' or ``lower'' interfaces were shown in Fig. 2(a) of corresponding to a hff of 16.5 T appears in PðBhfÞ of the 97 Ref. [5]. One-hour isochronal annealing of the samples upper interface after annealing at 2001C and above. In 43 was performed in UHV in steps of increasing tempera- general, for the upper interface the low-field distribution 99 tures between 2001C up to 5501C, when strong bulk was found to be essentially broader and the amplitude of 45 diffusion starts. The CEM spectra and hff distributions the 20 T peak itself was smaller than that for the lower 101 (not shown here) are of similar statistical quality as interface. Annealing of the samples up to a temperature 47 those displayed in Refs. [4] (Figs. 4 and 5) and [5] of 4501C does not crucially modify the CEM spectra. A 103 (Fig. 2(a)). The obtained hff distributions have been small increase of the ``bulk'' contribution (33.8 T peak) 49 decomposed 105 51 107 53 UNCORRECTED PROOF by least-squares fitting into several (here: and of the low-field contribution (20 T peak for the six or seven) Gaussian functions (satellites) with lower interface, and the sum of the 20.3 and 16.5 T peaks individual widths and spectral areas. Our results, i.e. for the upper interface) up to a temperature of 4001C the average position of the Gaussian satellite lines and reflects the weak (short-range) diffusion of 57Fe atoms the corresponding relative spectral areas (in %, relative from the interface towards the inside of the Fe and Cr 109 to the total experimental spectral area), obtained after slabs, respectively. A similar result concerning a weak 55 annealing both types of samples at different tempera- increase in area of the 20 T satellite line after annealing 111 tures, are given in Table 1. The total area (corresponding at 3001C was reported by Kopcewicz et al. [3]. Their MAGMA : 8451 ARTICLE IN PRESS V. Uzdin et al. / Journal of Magnetism and Magnetic Materials ] (]]]]) ]]]­]]] 3 1 interpretation is based on the assumption that this hff between atoms of different chemical elements, and gives 57 corresponds to Fe atoms in the ``flat'' interface, and they a natural explanation of the change of the hff distribu- 3 had to conclude that there is in-plane diffusion inside the tion on 119Sn atoms in V/Cr superlattices versus the 59 superlattice during annealing which leads to smoothing position of the 119Sn probe layer inside the Cr spacer [10]. 5 of the interfaces. However, this is an unlikely process, We conclude that the difference between the hff 61 and our finding that the 20 T peak (as well as the 16.5 T distributions from lower and upper interfaces in Fe/Cr 7 peak) originates from Fe atoms embedded in the Cr superlattices can be explained without consideration of 63 spacer near the interface [4] gives a more natural Fe and Cr melting points. However, the different melting 9 explanation of the annealing experiment. points play an important role in the evolution of PðBhfÞ 65 Another problem is the explanation of the differences versus annealing, especially at high temperature, when 11 in PðBhfÞ for lower and upper interfaces. This difference bulk diffusion starts. After annealing at 5001C, we 67 was reported earlier [5,6], and it was assumed that observe a remarkable increase of the spectral area of the 13 interface alloying is governed by the binding energies second satellite line (31.3 T­31.4 T ) in both types of 69 between the substrate and ad-atom material, which, as a interfaces. This means that individual Cr atoms start to 15 first approach, are proportional to the melting points of penetrate into the Fe layers. There is no similarly 71 the solids [6]. The melting point of Cr is higher than for remarkable increase of the low-field contribution at 17 Fe and, therefore, interface mixing for the lower 5001C, which would correspond to isolated Fe atoms in 73 interface might be less significant as compared with the the Cr spacer. The latter process starts only after 19 upper one [6]. However, recent investigations of epitaxy annealing at 5501C for the lower (Fe-on-Cr) interface 75 of Fe on Cr [7] and Cr on Fe [8] using scanning whereas for the upper (Cr-on-Fe) interface, the total 21 tunneling microscopy demonstrate the occurrence of low-field contribution (at 20.3 and 16.5 T) decreases 77 alloying on both interfaces. Modeling of the epitaxial strongly with annealing at 5501C. Consequently 57Fe 23 growth with the algorithm of simple ballistic deposition atoms at the upper interface do not penetrate deeply (far 79 cannot reproduce the differences between interfaces [4]. away from the interface) into the Cr spacer even at such 25 Now, we developed a new algorithm [9] for interface a high annealing temperature. Although the observed 81 alloying, which includes ballistic deposition with con- starting temperature of 500­5501C for bulk diffusion is 27 sequent rising up of some atoms on the surface. It allows remarkably lower then the melting points of bulk Fe and 83 to reproduce the main differences between lower and Cr, it is closer to the Fe melting point. That is why Cr 29 upper interfaces. We assume that site exchange of atoms atoms can diffuse into the Fe slabs, which are nearer to 85 and their diffusion take place only at the surface during the liquid state, but Fe atoms do not diffuse into the 31 the epitaxial growth and there is no internal bulk solid Cr spacer. This conclusion is in agreement with 87 diffusion. We start from the structure obtained by the recent calculations of vacancy formation energies which 33 algorithm of simple ballistic deposition. Then in every are found to be larger for Cr than for Fe [11]. 89 layer, we choose a definite fraction (B) of atoms using a 35 random procedure, and layerwise, starting from the V.U. appreciates financial support from the Alexan- 91 bottom, this fraction of atoms was exchanged in every der von Humboldt foundation and the MML01 37 pair of neighbouring layers. The parameter Bo1 symposium organizers. 93 determines the intensity of interface alloying. Such a This work was partially supported by the Russian 39 scenario automatically leads to the asymmetry of the Ministry of Higher Education (grant E00-3.4-547), the 95 interface: atoms could float up on several layers, but did program ``Universities of Russia: fundamental re- 41 not move down due to suppression of diffusion in the searches'' (project 015.01.01.083) and the Deutsche 97 inner layers below the surface. For the probe layer at the Forschungsgemeinschaft (SFB 491Bochum/Duisburg). 43 lower (Fe-on-Cr) interface, 57Fe atoms will float and 99 move into the natFe slab, thus increasing the intensity of References 45 the bulk-like peak in PðBhfÞ: At the upper (Cr-on-Fe) 101 interface, these 57Fe atoms will float and move into the [1] R. Schad, P. Belien, G. Verbanck, K. Temst, V.V. 47 Cr spacer, thus increasing the low-field contribution in Moshchalkov, Y. Bruynseraede, B. Bahr, J. Falta, 103 PðBhfÞ: Self-consistent calculations of the magnetic J. Dekoster, G. Langouche, Europhys. Lett. 44 (1998) 379. 49 moment [2] F. Klinkhammer, Ch. Sauer, E.Yu. Tsymbal, 105 51 107 53 UNCORRECTED PROOF distribution performed within the periodic Anderson model for these structures show a consider- S. Handschuh, Q. Leng, W. Zinn, J. Magn. Magn. Mater. able number of Fe atoms which have a magnetic 161 (1996) 49. moment corresponding to the hff of about 20 T [9], [3] M. Kopcewicz, T. Lucinski, F. Stobiecki, G. Reiss, J. Appl. Phys. 85 (1999) 5039. contrary to the case of structures obtained by a simple [4] V. Uzdin, W. Keune, H. Schr.or, M. Walterfang, Phys. 109 ballistic deposition algorithm [4]. Note that such a Rev. B 63 (2001) 104407. 55 scenario of epitaxial growth is very general. It does not [5] T. Shinjo, W. Keune, J. Magn. Magn. Mater. 200 (1999) 111 assume any differences in the strength of interactions 598. MAGMA : 8451 ARTICLE IN PRESS 4 V. Uzdin et al. / Journal of Magnetism and Magnetic Materials ] (]]]]) ]]]­]]] 1 [6] B. Heinrich, J.F. Cochran, T. Monchesky, R. Urban, Phys. [9] V. Uzdin, W. Keune, Phys.Metals, Metallurgy (2001) in Rev. B 59 (1999) 14520. press. 11 [7] A. Davies, J.A. Stroscio, D.T. Pierce, R.J. Celotta, Phys. [10] M. Almokhtar, K. Mibu, A. Nakanishi, T. Kobayashi, 3 Rev. Lett. 76 (1996) 4175. T. Shinjo, J. Phys.: Condens. Matter 12 (2000) 9247. 13 [8] Y.J. Choi, I.C. Jeong, J.-Y. Park, S.-J. Kahng, J. Lee, [11] P.A. Korzhavyi, I.A. Abrikosov, B. Johansson, 5 Y. Kuk, Phys. Rev. B 59 (1999) 10918. A.V. Ruban, H.L. Skriver, Phys. Rev. B 59 (1999) 11693. 15 7 9 UNCORRECTED PROOF