Ti multilayers

Ti multilayers

ARTICLE IN PRESS Physica B 350 (2004) e241–e244 Polarized neutron reflectivity of FeCoV/Ti multilayers . b, M. Senthil Kumara,*, V.R. Shahb, C. Schan...

252KB Sizes 3 Downloads 23 Views


Physica B 350 (2004) e241–e244

Polarized neutron reflectivity of FeCoV/Ti multilayers . b, M. Senthil Kumara,*, V.R. Shahb, C. Schanzerb, P. Boni c d T. Krist , M. Horisberger a

Department of Physics, Indian Institute of Technology Bombay, Mumbai 400 076, India Physik Department E21, Technische Universitat D-85748 Garching, Germany . Munchen, . c BENSC, Hahn-Meitner-Institut, Glienicker Strasse 100, 14109 Berlin, Germany d Laboratory for Neutron Scattering, ETHZ & PSI, PSI CH-5232 Villigen, Switzerland


Abstract Spin valve multilayers consisting of magnetically hard and soft FeCoV layers separated by Ti spacer layers were investigated for their interface and magnetic properties. The chemical structure and interface roughness were characterized by measuring X-ray reflectivity. Polarized neutron reflectivity (PNR) was performed on selected samples at magnetic saturation, and the spin dependent reflectivities were analysed by modelling the magnetic states of the soft and hard FeCoV layers. PNR suggests the existence of two magnetic phases of FeCoV, i.e, in the bulk and at the interface. The extent of these two regions and their magnetic moments for various Ti thicknesses were estimated from PNR and are utilized to understand the magnetization reversal of these multilayers. r 2004 Elsevier B.V. All rights reserved. PACS: 61.12.Ha; 75.25.+Z; 75.30.Gw; 75.60.Jk; 75.70.Cn Keywords: Magnetization reversal; Magnetic roughness; PNR; Spin-valve multilayers

1. Introduction Magnetic multilayers (MLs) consisting of FeCoV (here FeCoV stands for Fe50Co48V2) and Ti have been previously investigated due to their interesting properties for neutron optics applications [1]. Furthermore, their structural, magnetic and magnetoelastic properties have been investigated in detail [2]. During the course of these investigations, it has been noticed that the coercive field (HC) depends on the FeCoV thickness in these multilayers [2]. We made use of this dependence to design a spin valve consisting of *Corresponding author. Fax: +91-22-2576 7552. E-mail address: [email protected] (M.S. Kumar).

magnetically hard and soft FeCoV layers (by choosing appropriate FeCoV thicknesses) separated by Ti. In this article, we report the investigations on FeCoV (10 nm)/Ti (t nm)/FeCoV (3 nm) MLs to understand how the Ti spacer layer influences their magnetic properties and the magnetization reversal. X-ray reflectivity (XRR) analysis has shown that the present MLs have rather rough interfaces (B2–4 nm r.m.s. roughness) compared to similar samples prepared at the sputter unit of FRM II (B1–2 nm r.m.s. roughness) [3]. The resulting ‘magnetic roughness’ [4] cannot be distinguished in magnetization measurements. We performed polarized neutron reflectivity (PNR) to quantify that and to understand its role in the magnetization reversal.

0921-4526/$ - see front matter r 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.physb.2004.03.060


M.S. Kumar et al. / Physica B 350 (2004) e241–e244

2. Experimental The multilayers [Ti(t nm)/FeCoV(10 nm)/Ti (t nm)/FeCoV(3 nm)]20 (t =3, 5, 8, 10, 12 and 15 nm) + 5 nm Ti capping for the prevention of the oxidation of the top FeCoV layer were sputter deposited on to glass substrates (50 mm  500 mm) by DC magnetron sputtering of Fe50Co48V2 and Ti (base pressure B2  106 mbar) at Paul Scherrer Institute. The interface roughness was determined from specular and off-specular X-ray reflectometry and the bulk magnetic properties were measured by a Quantum Design physical property measurement system (PPMS). PNR measurements were made at the V14 reflectometer, at the Hahn– Meitner Institute, Berlin in the angle dispersive mode. The polarized (96%) beam was monochromatized by pyrolytic graphite (l ¼ 0:472 nm). Flipper coils are used for inverting the beam polarization and the spin analysis was done using a FeCo/Si super mirror (kept in a permanent magnetic field) where the reflected and transmitted intensities represent the non spin-flip (NSF) and spin flip (SF) intensities. Both NSF and SF reflectivities were measured, however we report only the analysis of NSF case due to space restrictions. The SF intensities were much weaker compared the NSF case, nevertheless they reveal some non-collinear spin structures. The MLs were subjected to an external magnetic field of 700 Oe in order to saturate the samples. Finally, the data were corrected for the polarizer and flipper efficiencies and a constant background level was subtracted prior to the normalization to obtain the reflectivities.

3. Results and discussion M–H loops along the easy axis for all the cases of Ti spacer layer thickness (t) are shown in Fig. 1. One would observe three distinct behaviours. At low t (B3–5 nm), the magnetization reversal takes place by a coherent rotation of all the layer magnetizations and the soft magnetic layer (3 nm FeCoV) seems to have identical switching field as the hard magnetic layer (10 nm FeCoV). When t is B8–12 nm, we observe two clear magnetization

Fig. 1. The M–H loops of Ti/FeCoV multilayers for various Ti spacer layer thicknesses.

reversals corresponding to the switching of soft and hard FeCoV layers. When t becomes B15 nm, the magnetization reversal becomes less sharp and the overall coercivity of both hard and soft layers decreased. The magnetization reversal, either by the rotation of the moments or the domain wall motion, depends on the micromagnetic details of the MLs. In the present case, we have estimated, from the XRD experiments, the size of the grains to be B10 nm. However there exists an anisotropic stress distribution in the film plane, very prominent at low t and diminishes as t increases. This anisotropic stress is explained as the reason for in-plane magnetic anisotropy through magnetostrictive effects [2]. We believe that this magnetostrictive energy can orient the moments like in a single domain state at low t. When t as well as the total thickness increases, this magnetostrictive contribution decreases leading to the formation of presumably randomly oriented domains. We notice that the M–H loops become less anisotropic and the values of Mr/MsB0.93 for t ¼ 3 nm to B0.78 for t ¼ 15 nm, is probably an indication of the magnetic disorder with increasing thickness. In Fig. 2 is shown the NSF reflectivities (R++ and R ) for the samples t ¼ 3; 10 and 15 nm, very close their magnetic saturation. A magnetic field of 700 Oe was applied along the easy axis and the two spin configurations of the neutrons (+ + and  ) probe the components of the

ARTICLE IN PRESS M.S. Kumar et al. / Physica B 350 (2004) e241–e244

Fig. 2. PNR of (a) t ¼ 3 nm, (b) t ¼ 10 nm and (c) t ¼ 15 nm samples. The open circles and triangles represent the R+ + and R  and the solid lines represent computed curves.

magnetization parallel to the polarization direction of the neutrons. A model based on a recursive matrix calculation [5] enables one to quantify the layer magnetization in the MLs. We have used the results of the XRR for the chemical structure. The interface roughness was adjusted for the neutron case, since neutrons penetrate deeper compared to X-rays and experience different roughness. With appropriate coherent, incoherent and magnetic scattering lengths, we could achieve a good agreement of the computation with the experimental data, as shown in Fig. 2. The ML periodicities determined from XRR and PNR match very well. The magnetic FeCoV layer was decomposed in to two phases, one phase at the middle part of the FeCoV layer with large magnetic moment and another phase at the two interface regions with a reduced magnetic moment. Introducing a magnetic moment of 2.2mB for the middle FeCoV and 0.8mB for the interface FeCoV, we are able to get very close agreement for the shift


in the critical angle, positions of the Bragg peaks and the contrast difference between the two spin dependent reflectivities for Ti ¼ 3 nm. The interface FeCoV phase extends over 0.5 nm (this value is 1 nm if the thickness of both the interface phases of an FeCoV layer are added) at each interface, in this case. Extending this method for t ¼ 10 and 15 nm, we find the bulk and interface regions assume 2 and 1mB and 1.7 and 1.3mB, respectively. The average widths of the interface regions are 1 and 0.9 nm, respectively. This trend has been verified for the intermediate t as well and they agree well with these observations. Here, we would like to mention that both PNR results and an estimate from the DC magnetization show reduced magnetic moment/f.u for FeCoV compared to its bulk value (2.35mB), however the discrepancy is less when the thicknesses are smaller. The reduced magnetic moments at the interface regions originate from magnetic inhomogeneities like non-collinear magnetic structures, magnetic dead layers or interface alloying. With increasing ‘magnetic roughness’, dipolar fields created at the interface influence the magnetization reversal [6]. Moreover, the reduction of stress anisotropy with increasing thickness results in the formation of domains with random anisotropy axes and can reduce the component of magnetic moments, at the interface as well as at the centre of the layers, along the field axis at moderate fields. The magnetic dead layers in similar films are much smaller (B0.21 and 0.33 nm) than the observed FeCoV/Ti interface layers [2,7], therefore, in the present study we believe that this interface region is most likely due to some interface alloying. However, the analysis of SF and polarization analysis the off-specular reflectivities are essential to separate the chemical and magnetic roughness as demonstrated by van de Kruijs et al., [8] on similar multilayers. Such investigations are underway and will be published elsewhere.

References . D. Clemens, M. Senthil Kumar, C. Pappas, Physica [1] P. Boni, B 267–268 (1999) 320. . [2] M. Senthil Kumar, P. Boni, J. Appl. Phys. 91 (2002) 3750.


M.S. Kumar et al. / Physica B 350 (2004) e241–e244

[3] C. Schanzer, et al., Physica B, in these proceedings. . [4] H. Zabel, K. Theis-Brohl, J. Phys.: Condens. Matter 15 (2003) S505. [5] SimulReflec., Program for reflectivity curve simulations and fitting, Fr!ed!eric Ott, Laboratoire L!eon Brillouin, CEA/ CNRS, Saclay, France, 2002.

[6] C. Tiusan, M. Hehn, K. Ounadjela, Eur. Phys. J. B 26 (2002) 431. . [7] M. Senthil Kumar, P. Boni, M. Horisberger, Physica B 325 (2003) 401. [8] R.W.E. van de Kruijs, et al., Appl. Phys. A 74 (Suppl.) (2002) S1550.