Domain phases in antiferromagnetically coupled sandwiches N. Persat IPCMS-Gemme, 23 rue du Loess, F-67037 Strasbourg, France and Siemens AG, ZT MF 1, Paul Gossenstrasse 100, D-91052 Erlangen, Germany H. A. M. van den Berg Siemens AG, ZT MF 1, Paul Gossenstrasse 100, D-91052 Erlangen, Germany K. Cherifi-Khodjaoui and A. Dinia IPCMS-Gemme, 23 rue du Loess, F-67037 Strasbourg, France The impact of buffer stacks containing Cr/Fe sandwiches on the quality of the antiferromagnetic coupling in Co 1.2 nm /Cu 0.83 nm /Co 1.2 nm sputtered trilayers has been investigated. Coupling strengths J larger than 0.4 erg/cm2 have been realized. The completeness of the antiferromagnetic alignment of both Co layers at zero field has been probed with the soft magnetic layer in the buffer stack. No remanence could be detected, demonstrating the completeness of the AF coupling across the sandwiches. The lack in remanence is partly due to the 111 texture imposed by the buffer and polycrystallinity of the Co layers, causing low rotational friction. The freedom in the sense of rotation of the Co layers causes a dense domain configuration with domain walls having their magnetization in the centers parallel to the original saturation field Hs . This causes an increment in the magnetization component in the direction of Hs and a reduction in the resistance. The state of high domain density converts into a low density one by annihilation of domains at positive field, so that no remanence due to these domains is detected. © 1997 American Institute of Physics. S0021-8979 97 47708-3 INTRODUCTION the growth of the AAF. The samples were protected by a Cu 2 nm /Cr 2 nm capping. The polycrystalline Co­Cu The Co/Cu multilayer system has been intensively inves- layers are 111 textured. tigated because of the high magnetoresistive level.1 This is The GMR signals were measured at room temperature partly due to the high scattering asymmetry of the Co/Cu by the standard four-point method with orthogonal sensing interfaces and partly due to the completeness of the antifer- current and applied field H in the plane of the layers. The romagnetic AF coupling in the first maximum. In many magnetization curves were measured by AGFM. cases, the complete AF coupling is only achieved after sev- eral periods, while the first few ones are not perfect. BUFFER LAYERS AND COUPLING QUALITY The recently introduced GMR sensors contain a Co/ Cu/Co sandwich, the so-called AAF artificial antiferromag- As criteria for the quality of the AF coupling, we con- netic subsystem , for which the ideal AF alignment is sider its strength J, the degree of completeness, i.e., the ab- prerequisite.2 It consists of at least one detection layer which sence of defects, and the coupling distribution.3 The use of a is decoupled from the AAF, i.e., this type of sensor requires Fe buffer layer is known to lead to a strong and complete AF the perfect AF alignment right from the first Cu spacer layer coupling in sputtered Co/Cu multilayers, which consequently in the AAF. In this paper, we present a number of buffers, exhibit high MR ratio.1 However, the analysis of the magne- which allow the growth of AAFs that show perfect alignment tization curve is hampered by the thick Fe layer, particularly at large AF coupling strength. Furthermore, a method for in the case of a sandwich containing two very thin Co layers. probing the completeness of the AF alignment is presented. Therefore we attempted to replace the Fe layer by a non- Various causes of remanence are discussed and experimen- magnetic one. A Cu buffer layer is known to induce rough tally demonstrated. interfaces in Co/Cu systems.1 Cr exhibits much crystallo- graphic resemblance to Fe and also has high affinity to the oxygen of the glass substrate. Unfortunately, the AF cou- EXPERIMENTAL TECHNIQUES pling vanishes completely for deposition on a Cr 4 nm / Cu 10 nm buffer, probably due to roughness. However, a The AAFs presented here consisted of two 1.2 nm Co tiny Fe layer 1.5 nm between Cr and Cu reestablishes the layers AF coupled through a 0.83 nm Cu layer. The samples occurrence of AF coupling type A buffer . The buffer stacks were prepared by sputtering with a base pressure of 5 of types B and C are also well suited and are characterized 10 8 mbar and deposited on glass substrates. Several kinds by excellent reproducibility. Types A, B, and C buffers con- of buffers were employed-type A: Cr 4 nm /Fe 1.5 nm / tribute much less to the sample total magnetic moment as Cu 10 nm , type B: Cr 4 nm /Fe 1.5 nm /Co 0.8 nm /Cu 10 compared to the usual 6 nm Fe buffer. Let us briefly describe nm , and type C: Cr 4 nm /Fe 1.5 nm /Ni80Fe20 1.8 nm / the effect of the magnetic part of the buffer on the magne- Cu 10 nm . The purpose of the 10 nm thick Cu layer is to toresistive response of the sample. exchange decouple the AAF from the magnetic part of the In the case of an ideal isolated AF coupled system see buffer stack, and, in addition, to provide a smooth surface for Fig. 1 a , the angle between H and the magnetization M 4748 J. Appl. Phys. 81 (8), 15 April 1997 0021-8979/97/81(8)/4748/3/$10.00 © 1997 American Institute of Physics Downloaded¬29¬May¬2001¬to¬148.6.178.13.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FM is Fe, Co, and Ni80Fe20 for types A, B, C, respectively. The level of RINT (H), and consequently the slope around H 0 of the total GMR curve see Fig. 1 b is proportional to ( Co 1)( FM 1), where FM / is the effective spin-asymmetry ratio of the buffer. As shown in Fig. 2, the slope at H 0 is the largest for the type B buffer, which suggests that Co NiFe Fe . It is in accordance with the fact that the MR ratio of Co/Cu mul- FIG. 1. Stylized M(H) and R/R(H) curves. In a , the full line corre- tilayers is known to be five times larger than that of sponds to the AAF and the dotted line to the soft magnetic layer. In b , the Ni80Fe20/Cu and Fe/Cu.4 However, the shape of the magne- full line is RAAF(H), the dotted line is RINT (H) and the dashed line is toresistive signal of the isolated AAF is very sensitive to their superposition. In each figure, the full and dotted arrows correspond to small variations of the spacer and magnetic layer thick- the M vectors within the AAF and soft magnetic layer, respectively. nesses. If the monoatomic steps have lateral extensions larger than the lateral coherence length, the magnetoresis- of each magnetic layer of the AAF satisfies cos H /H tance curve is the superposition of several parabola, each s for H H characterized by a different saturation field.3 The deforma- s . The normalized GMR signal see Fig. 1 b of the AAF is R tion of the ideal parabola can modify the slope at H 0 of AAF(H) 1-(H/Hs)2. Let us now consider the situation that the AAF is no longer isolated and interacts the total signal, whose value can thus not be used to estimate with the magnetic layer of the buffer, which is supposed to quantitatively the effective spin-asymmetry ratio of the have an ideal soft-magnetic stepwise response see Fig. 1 a . buffer. The normalized GMR signal resulting from this interaction is The magnetic layer of the buffer stack constitutes a dis- roughly given by R advantage for the analysis of the magnetic behavior of the INT(H) 1 H/Hs for H Hs and adds to R AAF. On the other hand, it provides a useful tool for testing AAF(H) see Fig. 1 b . The parabolic response R the completeness of the antiparallel alignment of the Co lay- AAF(H) is modified, and the slope of the curve around H 0 is related to the actual level R ers at H 0. INT (H). Figure 2 presents the GMR curves obtained for the AAF Let us now consider the case of a small lag , for deposited on the buffer stacks of types A, B and C, respec- example due to friction, in the magnetic response of the tively. A strong hysteretic behavior is observed in the signal, AAF, so that cos 1(H/Hs) . RAAF(H) deviates by but we now focus on the upper R(H) branch, for reasons that 2 sin(2 ) from the perfect signal. The response is mostly will be detailed further. This R(H) branch presents in all sensitive to this modification for values around 45° 90°. cases the expected form superposition of a parabolic curve On the contrary, the signal is not sensitive to the deviation with a triangular one . The slopes of the three curves at H from the complete antiparallel alignment of the AAF at H 0 differ significantly. It is attributed to the nature of the 0 ( 90°), i.e., to remanence. In the case of a small lag magnetic layer of the buffer. The aim of the three different , RINT(H) deviates by sin from the perfect signal. buffer stacks is to modify the interaction between the mag- The effect of on the GMR is the largest for values netic soft layer and the AAF. The level of the signal around 90°, i.e., near H 0. A sudden jump in the GMR R signal will occur upon switching the soft magnetic layer. INT(H) is determined by the electron scattering events of both spin-current channels at both the Co/Cu and the FM/Cu in this example originates in homogeneous friction in the interfaces and also in the magnetic bulk of the buffer. Here AAF. It is obvious that any other source of remanence in the AAF will produce a similar effect. Let us focus on the experimental magnetoresistive re- sponse of the samples, and particularly near H 0, for ex- ample in the case of an AAF deposited on the buffer stack of type B Fig. 3 . The AAF presents a large AF coupling (J 0.4 erg/cm2, and even larger has been achieved . The buffer has also been grown without AAF on top, to sepa- rately investigate its M(H) response. The Fe 1.5 nm /Co 0.8 nm bilayer switches at about 50 Oe inset in Fig. 3 b . As shown in the inset in Fig. 3 a , the R(H) reduces upon switching the detection layer, indicating that the mean M of the AAF has already changed sense, i.e., the remanence of the trilayer is negligibly small. This is confirmed by the M(H) measurement in Fig. 3 b . The Co/Cu/Co sandwich deposited on type A or C buff- ers also presents a strong AF coupling strength J 0.4 erg/cm2 and a complete AF alignment at H 0. Noting, that most multilayers only exhibit perfect AF coupling after FIG. 2. The magnetoresistance signal as a function of the normalized field H/H growing several periods, the present results demonstrate the s for identical AAFs deposited on three types of buffer stacks. The vertical scale is the same in all cases, shifted by 1% for clarity. excellent quality of the buffers. J. Appl. Phys., Vol. 81, No. 8, 15 April 1997 Persat et al. 4749 Downloaded¬29¬May¬2001¬to¬148.6.178.13.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp than the longest mean free path of the electrons so that the walls constitute channels with reduced . In other words, at some given field, is lower at the domain walls because of the parallelism of the moments, compared to the hypothetical configuration with uniform M in each of the layers. The higher R(H) branch is characterized by a much lower wall density due to irreversible domain annihilations as will be explained . The lower R(H) curve in Fig. 3 a and the branch with the higher mean moment along H in Fig. 3 b correspond. Remembering that the moments in the middle of the domain walls always lie in the direction of the original H, the M(H) branch with the high domain-wall density should lie above its pendant. The domains constitute states of high energy that be- come unstable when the domain-wall energy increases. This energy increases strongly upon reducing H since the domain- wall angles grow. The smaller domains in regions with strong AF coupling, i.e., with high wall angles and wall- energy density, collapse first. The angle between the M's of both layers of the AAF grows at the former wall sides, and, as a consequence, the local resistivity is increased. After re- ducing H to zero, the domain density is diminished to a low level. Increasing H to the negative saturation leads to the occurrence of the high R(H) branch Fig. 3 a and to the low M(H) branch Fig. 3 b . In the previous section, we FIG. 3. Magnetoresistance a and magnetization b curves of the sample have focused on the upper R(H) branch. The corresponding Co/Cu/Co on type B Cr4 nm/Fe1.5 nm/Co0.8 nm/Cu10 nm buffer stack. In magnetic configuration low density of domain walls exhib- b , the dotted lines show the magnetization of the soft magnetic Fe1.5 its much resemblance with the ideal configuration of Fig. 1 nm/Co0.8 nm bilayer, with no AAF on top. The insets detail the signals between 150 and 150 Oe. single domain layers , provided that the mean size of the domains is large enough. DOMAINS IN AF COUPLED SYSTEMS CONCLUSION Various branches are recognized in both the GMR and The use of different buffers has allowed us to obtain a M(H) curves. The branch with the lowest R(H) occurs very high coupling quality in sputtered 111 Co/Cu/Co when reducing H from positive saturation towards zero in sandwiches with complete antiferromagnetic alignment of Fig. 3 a . The relatively low resistivity is attributed to the the Co layers at zero field. The buffer provides a useful tool development of a dense domain structure, which originates to probe by means of GMR signal the amount and direction in the freedom in the sense of the rotation of the M of the of remanence of the Co/Cu/Co part. AAF. Upon reducing H, the M's of both AAF-magnetic lay- The various branches in both the GMR and M(H) sig- ers rotate in opposite directions until they reach an AF align- nals correspond to a general phenomenon originating in the ment at H 0. Because of polycrystallinity, there is no glo- freedom of the sense of rotation of the Co magnetization bal anisotropy and at zero field, the AF alignment should be upon reducing H from saturation. The studied samples are perpendicular to the original H. The M in a specific layer has magnetically isotropic and the occurrence of the state with the freedom to rotate either clockwise or anticlockwise. high domain density might be avoided when a global anisot- Probably, asymmetries in the local anisotropy energies be- ropy is present which differs for both AAF layers. tween both AAF-magnetic layers determine the local sense of rotation. Consequently, domains distinguishing them- 1 S. S. P. Parkin, R. Bhadra, and K. P. Roche, Phys. Rev. Lett. 66, 2152 selves by the sense of rotation of M develop. Domain walls 1991 . 2 H. A. M. van den Berg, W. Clemens, G. Gieres, G. Rupp, W. Schelter, separate the regions with opposite rotation sense. The mo- and M. Vieth, IEEE Trans. Magn. 32, 4624 1996 . ments in the middle of the walls always lie in the direction of 3 H. A. M. van den Berg, S. Zoll, K. Ounadjela, D. Stoeffler, and A. Dinia the original saturation field. The walls in both magnetic lay- unpublished ; H. A. M. van den Berg, in Magnetic Thin Films and ers are just above each other because of their magnetostatic Multilayer Systems, edited by U. Hartmann Springer, Berlin, 1997 . 4 J. Inoue, H. Itoh, and S. Maekawa, J. Magn. Magn. Mater. 121, 344 coupling. The width of these walls is of the order or larger 1993 . 4750 J. Appl. Phys., Vol. 81, No. 8, 15 April 1997 Persat et al. Downloaded¬29¬May¬2001¬to¬148.6.178.13.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp