Journal of Magnetism and Magnetic Materials 190 (1998) 187-192 High sensitivity GMR with small hysteresis in Ni-Fe/Cu multilayers M. Urbaniak*, T. Lucin´ski, F. Stobiecki Institute of Molecular Physics, Polish Academy of Sciences, PL 60-179 Poznan&, Smoluchowskiego 17, Poland Received 5 May 1998; received in revised form 6 August 1998 Abstract Giant magnetoresistance (GMR) effect and magnetisation reversal processes have been investigated in Py/Cu(Py"NiFe, permalloy) multilayers (Mls) obtained by face-to-face sputtering method. The investigated films had constant sublayer thicknesses both for Py and Cu (d!"2 nm, d."2 nm) and various numbers of ferro- magnetic sublayers. It has been shown that for such Mls a high field sensitivity of GMR effect (S+0.4%/Oe) and negligible hysteresis can be obtained for a low number of Py layers. 1998 Elsevier Science B.V. All rights reserved. Keywords: Exchange coupling; GMR sensitivity; Magnetoresistance; Multilayer; Permalloy 1. Introduction as magnetic layer display moderately high GMR amplitudes and low saturation fields [3-7]. The Multilayered structures consisting of ferro- above leads to very promising, from the application magnetic layers separated by a nonmagnetic, con- point of view, values of GMR field sensitivity which ducting spacer are the subject of a very intensive in our samples attain 0.6%/Oe [7]. In this paper we study. One of the reasons is a phenomenon of a present an analysis of the influence of the number of giant magnetoresistance (GMR), i.e., a considerable magnetic layers, N, in a stack on the GMR effect change of resistance upon the application of mag- field sensitivity of Ni netic field [1]. The occurrence of GMR effect in Fe/Cu multilayers ob- tained by double face-to-face sputtering [8]. It is multilayer systems is often the result of the exist- shown that the GMR saturation field decreases ence of an antiferromagnetic interlayer exchange with decreasing N much more than predicted the- coupling [2]. It turns out that multilayers with Py oretically for ideal multilayer. We show that due to a low uniaxial anisotropy of Py sublayers and a relatively weak interlayer exchange coupling in Py/Cu multilayers a negligible hysteresis in R(H) behaviour can be accompanied by GMR field sen- * Corresponding author. E-mail: urbaniak@ifmpan.poznan.pl. sitivity close to 0.4%/Oe. 0304-8853/98/$ - see front matter 1998 Elsevier Science B.V. All rights reserved. PII: S 0 3 0 4 - 8 8 5 3 ( 9 8 ) 0 0 3 3 3 - 3 188 M. Urbaniak et al. / Journal of Magnetism and Magnetic Materials 190 (1998) 187-192 2. Experimental a longitudinal magneto-optical Kerr effect (MOKE). The DC magnetoresistance measurements were The glass/Py!d./[Cu!2 nm/Py!d.];(N!1) performed at RT with the conventional four-point (where N"2, 6, 11, 21, 51, 101 and d. denotes method while in situ resistance characteristics were Py"NiFe thickness, d."2 and 2.3 nm. Cu measured with the two-point method. We define sublayer thickness is denoted by d!) multilayers the field dependence of the GMR as: have been obtained at room temperature (RT) by double face-to-face sputtering [8]. In this sputter- R(H)!R(H GMR(H)"100; ), ing geometry the substrate is placed outside the R(H ) plasma discharge. It has two advantages: the sub- strate temperature is lower than for other sputter- where H +700 Oe denotes the maximum mag- ing methods and the in situ resistance measurements netic field applied in our experiment. In this paper are possible. The Cu and Py sublayer thicknesses the maximum resistance, not its zero-field value, were determined by X-ray fluorescence method determines the GMR amplitude. The external mag- (XRF) [9]. For samples with a greater number of netic field in all measurements was applied in- repetitions a well-defined periodic structure was plane, while the sensing current was always perpen- confirmed by low and high angle X-ray diffraction dicular to the magnetic easy axis (EA) direction. which allowed us to determine the concentration modulation wavelength ( "d.#d!). The X-ray measurements revealed that our polycrystalline 3. Multilayers growth samples show a dominant (1 0 0) texture. The mag- netisation reversal processes were examined with Fig. 1 gives a representative example of the a vibrating sample magnetometer (VSM) and by in situ conductance (G) of the Py(2.3 nm)/Cu(2 nm) Fig. 1. Conductance of Py(2.3 nm)/Cu(2 nm) multilayer as a function of the total thickness. M. Urbaniak et al. / Journal of Magnetism and Magnetic Materials 190 (1998) 187-192 189 multilayer as a function of the thickness. The onset for Py(2 nm)/Cu(2 nm) multilayer with 101 mag- of conductivity was observed to occur at the nom- netic sublayers. The fitting procedure was per- inal Py thickness of about 1.2 nm. It suggests an formed within the two-layer model proposed by island growth mode and implies that the initial Dieny et al. [14]. It was assumed that a total energy mesoscopic roughness exceeds 1.2 nm [10]. The of a system consists of exchange coupling energy, minimum of G(d.) during Py growth on Cu surface Zeeman energy and anisotropy energy (it was which indicates the completion of the first homo- shown previously that in our multilayers a distinc- geneous Py layer [11,12], appears at a thickness of tive uniaxial in-plane anisotropy is present [7]). about 0.2 nm (about 1 ML) suggesting a layer-by- Thus, the energy of bilayer per surface unit has layer growth (see insert in Fig. 1). The decrease of a form (with magnetic field applied parallel to EA conductance is due to an increase of diffuse scatter- direction) ing of Cu conduction electrons at the sample sur- face [13]. Conductance varies almost linearly with E"!BM1t(cos #cos ) the number of deposited Py/Cu bilayers which indi- !J cos( cates that irrespective of the position in the stack ! )!K3t(cos #cos ), the transport properties are similar (perfectly linear where M variation corresponding to parallel resistors would 1 and t are saturation magnetisation and thickness of magnetic layers, respectively; K mean no giant magnetoresistance). Our in situ res- 3 is a uniaxial anisotropy constant, istance measurements do not allow us to estimate  and  are angles between magnetisations and magnetic field the quality of Py/Cu interface, i.e. for Cu deposited direction and J is a bilinear coupling constant. In on Py, since in this case no minimum in G(d!) the calculation a steepest descent method with the dependence is seen. It could be argued that there is basic step equal to 0.0005 rad was used to find the no evidence for an island growth since in a 3D local energy minimum. Dieny et al. [14] have growth mode a plateau in G(d!) should be ob- shown that a calculation performed for a bilayer served [12]. (only two magnetic sublayers) could be used to describe the behaviour of a multilayer with large, odd number of magnetic layers provided that the 4. Interlayer exchange coupling coupling constant is multiplied by 2. Magnetisation and thickness values used in our Fig. 2a presents a typical Kerr rotation )(H) calculations were determined experimentally from dependence together with a model curve obtained VSM and XRF measurements, respectively. The Fig. 2. The field dependence of (a) the Kerr rotation )(H) (line shows a fit according to the model of Dieny (see text)) and (b) magnetisation M(H) of Py(2 nm)/Cu(2 nm) multilayer with 101 magnetic sublayers. 190 M. Urbaniak et al. / Journal of Magnetism and Magnetic Materials 190 (1998) 187-192 obtained antiferromagnetic exchange coupling con- stant values are small, about 0.5-0.8;10\ J m\, which as shown later allowed us to obtain R(H) characteristics with small saturation fields. The relatively high hysteresis present in )(H) depend- ence is not observed in M(H) curves obtained with VSM (see Fig. 2b) and in R(H) measurements. It reflects the fact that in contrast to the resistance and VSM measurements a MOKE signal is col- lected from a thin (&20 nm) surface layer and thus it is much less affected by thickness inhomogenei- ties inevitably present in our Mls. The loss of the modulation periodicity (different Cu spacer thick- nesses) causes the coupling energy between dif- ferent pairs of neighbouring Py layers to vary. As a result, different layers rotate at different field values and the GMR effect field sensitivity is diminished. 5. The influence of number of repetitions on R(H) behaviour Exemplary GMR(H) curves obtained for multi- layers with different number, N, of magnetic sub- layers are displayed in Fig. 3. Decreasing of GMR amplitude with N (Fig. 4a) is partly caused by an Fig. 3. Exemplary GMR(H) curves for Py(2 nm)/Cu(2 nm) increased contribution of outer boundary scatter- multilayers with different number of magnetic sublayers. ing to conducting processes and a lower number of magnetic-nonmagnetic interfaces within electron mean free path [1]. The N"2 stack shows no coupling is stronger for layers more distant from GMR effect. In this case we observe only a small the substrate. As can be seen from Fig. 4a, AF- magnetoresistive signal coming from the scattering coupled fraction of the sample F of the conduction electrons by paramagnetic and/ $ (where F$" 1!M or superparamagnetic fluctuations localised near 0/M1, M0 denotes remanence magnetisa- tion) increases with increasing N, i.e., as the relative Py/Cu interfaces, as discussed by Lucinski et al. contribution of first layers to the total magnetic [6]. We conclude that coupling between first two moment of the sample decreases. sublayers is absent or favours parallel alignment of We have observed a decrease of the saturation their magnetic moments [15]. The absence of anti- field, H ferromagnetic interlayer coupling can be explained 1, with lowering N (Fig. 4b). This effect was theoretically explained by Dieny [14]. Note that by initial roughness (island growth) (see Fig. 1) H which can change the effective spacer thickness. In 1 changes much more than by a factor of 2, as predicted by Dieny et al. for an ideal stack, on our samples even very small, less than 0.3 nm, de- decreasing N. It suggests, similarly to the GMR parture from a nominal Cu thickness can reduce amplitude dependence on N, that the coupling is GMR amplitude to zero [6]. stronger for layers at larger distances from the Complementary magnetisation reversal measure- substrate [15]. Comparing Fig. 4a and Fig. 4b one ments performed with VSM (Fig. 2b) seem to con- can see that there is a range of N in which GMR firm that antiferromagnetic interlayer exchange amplitude is almost constant while H1 decreases M. Urbaniak et al. / Journal of Magnetism and Magnetic Materials 190 (1998) 187-192 191 Fig. 5. GMR(H) dependence for Py(2 nm)/Cu(2 nm) multilayer with six magnetic layers. The magnetic field was applied parallel to the EA direction. GMR(H) characteristic with small hysteresis. Nevertheless, it must be noted that the GMR(H) curve is no longer smooth when N is small as opposed to Mls with high number of magnetic sublayers. It may result from the N dependence of the shape of M(H) curves, which is the case even for a structure with identical layers [1], and on the other hand, from the fact that in our samples anti- ferromagnetic exchange coupling is weaker in the Fig. 4. (a) The GMR effect amplitude and F$ dependence on first layers of the stack. From the application point the number N, of magnetic layers in Py(2 nm)/Cu(2 nm) multi- layers for field applied parallel to EA direction. (b) The GMR of view it is also advantageous that our samples are effect 50% saturation field, H thin and have thus high resistance values (for N"6 1, dependence on the number N, of magnetic layers in Py(2 nm)/Cu(2 nm) multilayers. Dots show sheet resistivity is about 15 / ). H1 for field applied parallel,while squares for field perpendicu- lar to EA direction. (c) The GMR effect field sensitivity depend- ence on the number N, of magnetic layers in Py(2 nm)/Cu(2 nm) multilayers. Dots show S 6. Conclusions  for field applied parallel, while squares for field perpendicular to EA direction. The results of magnetic, in situ resistance and magnetoresistance measurements performed at RT considerably. We define the GMR field sensitivity, on Py(2 nm)/Cu(2 nm) multilayers presented above S, in a usual way: S"GMR/(2H1) (where allow us to determine the influence of the number H1 is the field change necessary to reduce the of magnetic layers on their magnetoresistive prop- GMR value from maximum to 50% of its ampli- erties. We conclude that: tude). Fig. 4c shows that when N is decreased from 101 to 6, S increases about two times and nearly 1. small values of interlayer exchange coupling reaches the value of 0.4%/Oe. Sensitivity can be constant, equal to about 0.8;10\ J m\ in further increased by increasing magnetic layer topmost sublayers, were observed for Cu spacer thickness, but unfortunately the GMR amplitude thickness of 2 nm; decreases simultaneously and hysteretic effects be- 2. strong dependence of GMR saturation field on come more pronounced (due to the domination of the number of magnetic layers in the stack al- anisotropy over exchange coupling) [7]. As can be lowed us to obtain Mls with high sensitivity and seen in Fig. 5, we have obtained high sensitivity relatively small hysteresis; 192 M. Urbaniak et al. / Journal of Magnetism and Magnetic Materials 190 (1998) 187-192 3. due to the island growth mode we do not ob- [7] M. Urbaniak, T. Lucinski, F. Stobiecki, Mol. Phys. serve GMR effect in a bilayer consisting of two Rept. 21 (1998) 167. magnetic Py sublayers. [8] F. Stobiecki, J. Dubowik, T. Lucin´ski, B. Szyman´ski, H. Rohrmann, K. Ro¨ll, M. Schmidt, Acta Phys. Polon. A 91 (1997) 277. The sensitivity in our samples is much lower than in [9] J. Baszynski, F. Stobiecki, B. Szyman´ski, K. Chrzumnicka, exchange-biased spin valves [16] but the host of Phys. Stat. Sol. (a) 141 (1994) K23. unresolved technological aspects (read noise, elec- [10] G. Reiss, H. Bru¨ckl, Surf. Sci. 269/270 (1992) 772. tromigration, thermal stability) [17] still makes [11] Th. Eckl, G. Reiss, H. Bru¨ckl, H. Hoffmann, J. Appl. Phys. them potentially attractive from the application 75 (1994) 362. [12] T. Lucinski, G. Reiss, N. Mattern, L. Van Loyen, J. Magn. point of view [18]. Magn. Mater. 189 (1998) 39. [13] C. Bellouard, C. Senet, B. George, G. Marchal, J. Phys.: Condens. Matter. 7 (1995) 2081. [14] B. Dieny, J.P. Gavigan, J.P. Rebouillat, J. Phys.: Condens. References Matter. 2 (1990) 159. [15] H.A.M. van den Berg, G. Rupp, IEEE Trans. Magn. 30 [1] B. Dieny, J. Magn. Magn. Mater. 136 (1994) 335. (1994) 809. [2] S.S.P. Parkin, Ann. Rev. Mater. Sci. 25 (1995) 357. [16] C.-M. Park, K. Ho Shin, Appl. Phys. Lett. 70 (1997) [3] S.S.P. Parkin, Appl. Phys. Lett. 60 (1992) 512. 776. [4] A.M. Zeltser, N. Smith, J. Appl. Phys. 79 (1996) 9224. [17] M.A. Parker, K.R. Coffey, J.K. Howard, C.H. Tsang, [5] T. Dei, R. Nakatani, Y. Sugita, Jpn. J. Appl. Phys. 32 (1993) R.E. Fontana, L. Hylton, IEEE Trans. Magn. 32 (1996) 1097. 142. [6] T. Lucinski, F. Stobiecki, D. Elefant, D. Eckert, G. Reiss, [18] P.P. Freitas, M.C. Caldeira, M. Reissner, B.G. Almeida, B. Szymanski, J. Dubowik, M. Schmidt, H. Rohrmann, K. J.B. Sousa, H. Kung, IEEE Trans. Magn. 33 (1997) Ro¨ll, J. Magn. Magn. Mater. 174 (1997) 192. 2905.