Eur. Phys. J. B 18, 413­419 (2000) THE EUROPEAN PHYSICAL JOURNAL B EDP Sciences c Societ a Italiana di Fisica Springer-Verlag 2000 Effect of the structural quality of the buffer on the magnetoresistance and the exchange coupling in sputtered Co/Cu sandwiches A. Dinia1,a, N. Persat1, S. Colis1, C. Ulhaq-Bouillet1, and H.A.M. van den Berg2 1 IPCMS-GEMMEb, ULP-ECPM, 23 rue du Loess, 67037 Strasbourg Cedex, France 2 Siemens AG, ZT MF 1, Paul Gossenstrasse 100, 91052 Erlangen, Germany Received 11 July 2000 Abstract. The effect of the structural quality of the buffer stack on the structural properties, giant magnetoresistance (GMR) and the quality of the antiferromagnetic coupling has been investigated for Co/Cu/Co sandwiches prepared by DC-magnetron sputtering. Three kinds of buffers were employed: type A: Cr(6 nm)/Co(0.8 nm)/Cu(10 nm), type B: Fe(6 nm)/Co(0.8 nm)/Cu(10 nm) and type C: Cr(4 nm)/Fe(3 nm)/Co(0.8 nm)/Cu(10 nm). For B and C type buffers, the antiferromagnetic alignment is very interesting at zero field with a coupling strength larger than 0.4 erg/cm2 and a GMR signal reaching 5% at room temperature. However, for the A type buffer the antiferromagnetic coupling completely disap- pears, while the GMR drops to about 0.8%. X-ray diffraction, atomic force microscopy and transmission electron microscopy have been performed in order to understand the origin of the observed difference in the magnetic properties. The results show a strong difference in the average surface roughness, 1.15 nm and 0.35 nm, respectively for the A and C types buffers, and demonstrate that the quality of the surface of the buffer is the key to optimize both the GMR and the indirect exchange coupling. PACS. 75.70.-i Magnetic films and multilayers ­ 75.70.Cn Interfacial magnetic properties (multilayers, magnetic quantum wells, superlattices, magnetic heterostructures) ­ 75.70.Pa Giant magnetoresistance 1 Introduction by the same technique, AF coupling strength (JAF) which vary from 0.3 erg/cm2 [12] to 0.45 erg/cm2 [13] oscillation In the past decade, improved sample preparation tech- periods between 1 [3] to 1.3 nm [14] and GMR values be- niques have made it possible to study magnetic multilayer tween 51% [15] and 80% [3,16]. Such a dispersion in the systems of unprecedented quality, revealing new physi- results for the same system, prepared by the same tech- cal properties. In particular, the antiferromagnetic (AF) nique, demonstrates clearly how sensitive to the prepara- exchange coupling [1] between ferromagnetic layers sep- tion conditions the magnetic and the transport properties arated by a nonmagnetic metallic spacer layer with cer- are. tain specific thicknesses as well as giant magnetoresistance In order to obtain a perfect antiferromagnetic coupling (GMR) [2]. (which means 100% antiparallel alignment of the adjacent The Co/Cu multilayers have been studied widely. It Co layers through the Cu nonmagnetic layer), some effort has been reported that polycristalline Co/Cu multilayers has to be put into the samples preparation and the choice with (111) texture prepared by sputtering exhibit antifer- of the buffer layer. This can lead to a general improvement romagnetic exchange coupling and a GMR that can reach in structured quality by reducing structural defects such 65% at room temperature [3,4]. In contrast, unless special as surface roughness, direct contact between the magnetic techniques are used, single-crystal superlattices grown by layers etc. which may lead to direct or ferromagnetic cou- molecular beam epitaxy (MBE) very often fail to yield a pling instead of the desired antiferromagnetic coupling. GMR at all [5]. This apparent anomaly was clarified when This is exactly what has been done in this study, which the first observation of GMR in (111)Co/Cu superlattices aims to show that the choice of an adequate buffer layer grown by MBE was reported [6]. A large variety of re- makes it possible to obtain nearly perfect AF coupling sults on antiferromagnetic coupling and GMR in Co/Cu with complete AF alignment with a negligible amount of system [7­11] has been reported for thin films prepared remanent magnetization for a Co/Cu sandwich prepared by sputtering. Firstly, magnetoresistance curves are pre- a e-mail: aziz@ipcms.u-strasbg.fr sented in order to demonstrate the quality of the AF cou- b UMR 7504 du CNRS pling. As the GMR is very sensitive to structural defects, 414 The European Physical Journal B the shape of this signal is a good indicator of growth qual- ity. Then, X-ray diffraction, atomic force microscopy, and transmission electron microscopy analysis are reported to explain the origin of the optimized antiferromagnetic cou- pling and GMR. 2 Experimental details The sandwiches have been prepared by DC-magnetron sputtering with a base pressure of 5 × 10-8 mbar and deposited on glass or silicon substrates at room temperature. The sandwiches consist of Co(1.2 nm)/ Cu(tCu)/Co(1.2 nm) with tCu = 0.83 nm near the first AF peak in the oscillatory interlayer coupling. Three types of buffers were employed [14]: type A: Cr(6 nm)/Co(0.8 nm)/Cu(10 nm), type B: Fe(6 nm)/ Fig. 1. Magnetoresistance curves measured at room tem- Co(0.8 nm)/Cu(10 nm) and type C: Cr(4 nm)/Fe(3 nm)/ perature with the magnetic field in the film plane and par- Co(0.8 nm)/Cu(10 nm). The purpose of the 10 nm thick allel to the current direction for the following sandwich: Cu layer is to exchange decouple the Co/Cu/Co sandwich Fe(6 nm)/Co(0.8 nm)/Cu(10 nm)/Co(1.2 nm)/Cu(0.83 nm)/ from the magnetic part of the buffer stack, and, in addi- Co(1.2 nm)/Cu(2 nm)/Cr(2 nm). tion, to provide a smooth surface for the growth of the Co/Cu/Co sandwich. The samples were protected by a Cu(2 nm)/Cr(2 nm) capping layer. periods [3]. This is, however, hard to achieve for a single The GMR curves were measured at room tempera- period stack. The Co(1.2 nm)/Cu(0.83 nm)/Co(1.2 nm) ture by the standard four-point method with orthogonal sandwich has been grown on the following buffer: sensing current and applied magnetic field in the plane of Fe(6 nm)/Co(0.8 nm)/Cu(10 nm). The GMR curves are the layers. Magnetization curves have been measured us- reported in Figure 1. It is clearly seen that both GMR ing an alternating gradient force magnetometer (AGFM) and AF coupling are very interesting. The GMR value is with the magnetic field applied in the film plane. around 6%, which is relatively high for a sandwich stack. The X-ray measurements were performed at room There are only a few previous studies on GMR in Co/Cu temperature using a Siemens powder diffractometer with sandwiches [11,14] and these report lower GMR values monochromatic Cu or Co K radiation. The geometry than the 6% observed in our case. This is a first indica- 1 of the diffractometer allows only experiments in reflection tion of the high quality of our samples. Using the satura- mode. tion field value HS, the saturation magnetization of bulk The surface roughness was studied by analyzing many Co, MS (since in this case the presence of a 6 nm thick different line scans with various lengths using an Atomic Fe buffer layer makes the determination of the magnetiza- Force Microscope. The roughness distribution on the sur- tion saturation of the Co very difficult) and the following face of the samples is well fitted by a Gaussian function expression: JAF = HSMStCo/2, where tCo is the ferro- described by its root mean square (RMS). magnetic thickness of the Co layer and the factor 2 ex- TEM observations were performed at Strasbourg with plains that only two surfaces are involved in the exchange a high resolution electron microscope TOPCON EM002B coupling, we found JAF = 0.43 erg/cm2. This value is operating at 200 kV with point to point resolution of among the highest values observed in sputtered or in MBE 0.18 nm at Scherzer defocus. While at Munich a Philips grown Co/Cu systems. A large GMR and exchange cou- TM 200 FEG electron microscope has been used which is pling strength do not guarantee that the antiferromagnetic equipped with field emission gun and operating at 200 kV coupling is complete. This is made, however, difficult since providing a resolution of 0.2 nm. The spatial resolution the analysis of the magnetization curve is hampered by the for the diffraction patterns was better than 10 to 20 nm. thick Fe layer and particularly in the case of a sandwich Only samples deposited on silicon substrates have been containing two very thin Co layers [14]. As a consequence, prepared for TEM plan-view and cross-section observa- the estimation of the amount of the remanent magnetiza- tions, either by an ionless tripod technique [17] or by ion tion in the Co/Cu/Co system itself is made difficult. Al- milling at liquid nitrogen temperature. though the magnetic signal of the Fe layer can in principle be easily evaluated separately, difficulties are encountered when subtracting from the total magnetization, the signal 3 Magneto-transport properties of the separately grown Fe layer. This arises from the fact that the signal of this layer, in particular its coercivity, is The use of a Fe buffer layer is known to lead to a strong affected by the stack grown on top of it. and complete antiferromagnetic coupling in sputtered Therefore, we attempted to replace the magnetic layer Co/Cu multilayers, which consequently exhibit high mag- by a nonmagnetic one. A Cu buffer layer is known to in- netoresistance ratio for multilayers with a large number of duce rough interfaces in Co/Cu systems [3]. Cr exhibits A. Dinia et al.: Buffer structure and magneto-transport properties of Co/Cu sandwiches 415 Fig. 2. Magnetoresistance curves measured at room tem- perature with the magnetic field in the film plane and par- Fig. 3. Magnetoresistance curves measured at room tem- allel to the current direction for the following sandwich: perature with the magnetic field in the film plane and par- Cr(6 nm)/Co(0.8 nm)/Cu(10 nm)/Co(1.2 nm)/Cu(0.83 nm)/ allel to the current direction for the following sandwich: Co(1.2 nm/Cu(2 nm)/Cr(2 nm). Cr(4 nm)/Fe(3 nm)/Co(0.8 nm)/Cu(10 nm)/Co(1.2 nm)/ Cu(0.83 nm)/Co(1.2 nm)/Cu(2 nm)/Cr(2 nm). much crystallographic resemblance to Fe and has high switched after reversing the applied magnetic field, have affinity to the oxygen of the glass substrate. The shown that the remanence of the AF sandwich is negligi- Cr(6 nm)/Co(0.8 nm)/Cu(10 nm) buffer has been used for bly small. Since there is no in-plane magnetic anisotropy the artificial antiferromagnetic subsystem (AAF) growth. in our sandwiches, the absence of the remanence is a good The results reported in Figure 2, show unfortunately that indication that the antiferromagnetic alignment is perfect the AF coupling vanishes completely and the GMR drops and as a consequence the coupling is perfect. to 0.8%. This effect is attributed to the roughness as will These results have clearly shown that using appropri- be evidenced later by transmission electron microscopy. ate growth conditions and optimal choice of the different In order to reduce the magnetic contribution of the layers, which constitute the buffer, we can obtain a strong buffer layer and to optimize the GMR and the exchange AF coupling strength and a complete AF alignment at coupling, we have decided to use both Fe and Cr in the H = 0 and a large GMR. In the next part, structural buffer layer. Several combinations have been made and analysis are developed in order to understand the physi- the best results have been obtained with a tiny Fe layer cal mechanism at the origin of the optimized AF coupling (3 nm) between a 4 nm Cr layer and 10 nm Cu layer. This and GMR. buffer makes it possible to re-establish both the antiferro- magnetic coupling and the GMR. As shown in Figure 3, the GMR and the interlayer coupling strength are close 4 Structural investigations to the values obtained with the Fe buffer. Moreover, the shape of the GMR curve presents a parabolic variation To characterize the physical parameters at the ori- around zero field, which is a good indication of better in- gin of the observed difference in the magneto-transport terlayer coupling than for the case with the Fe buffer [14]. properties between the different buffers, two of them This means that the magnetization vectors of the adjacent have been selected for the structural analysis. The Co layers are fully antiparallel and follow a small angular Cr(4 nm)/Co(0.8 nm)/Fe(3 nm)/Cu(10 nm) buffer, variation for small applied magnetic fields. In addition, which gives the best AF coupling and GMR, and the the Cr/Fe/Cu buffer contributes less to the sample to- Cr(6 nm)/Co(0.8 nm)/Cu(10 nm) for which the coupling tal magnetic moment as compared to the usual 6 nm Fe vanishes and the GMR is strongly reduced. These buffers buffer. The magnetic layer of the buffer stack constitutes have been analyzed by X-ray Diffraction, atomic force mi- a disadvantage for the analysis of the magnetic behavior croscopy and transmission electron microscopy (TEM). of the antiferromagnetic sandwich. On the other hand, it For the B type buffer, the structural analysis has shown provides a useful tool for testing the completeness of the similar characteristics to the C type buffer, which explains antiparallel alignment of the Co layers at H = 0. This the similarity between their magnetotransport properties. has been done [12] previously using different buffers. Thin magnetic layers of 0.8 nm Co and 1.8 nm Ni80Fe20 have been inserted between the 3 nm Fe layer and the 10 nm 4.1 X-ray diffraction Cu layer. The analysis of the magnetoresistance curves around H = 0 and particularly the contribution to the Figure 4 shows the high angle X-ray spectrum obtained GMR of the magnetic part of the buffer layer, when it for the Co(1.2 nm)/Cu(0.84 nm)/Co(1.2 nm) sandwich 416 The European Physical Journal B 600 800 Cr /Fe /Co /Cu /Co /Cu /Co 4nm 3nm 0.8nm 10nm 1.2nm 0.84nm 1.2nm [Cu(3nm)/Co(3nm)] multilayer 700 500 300 2 =51.21 ° 600 B 2 =1.48 ° 250 400 500 .) I (a.u.) .u 400 =11.8° (111)Cu 200 Gauss approximation I (a.u.) 300 I(a 300 48 50 52 54 Gauss 200 200 approximation 100 100 0 30 40 50 60 70 80 90 100 110 120 5 10 15 20 25 30 35 40 2 (degree) (degree) Fig. 4. High angle X-ray spectrum recorded at room Fig. 5. Rocking curve performed on the Cu/Co multi- temperature for the sandwich: Cr(4 nm)/Fe(3 nm)/ layer deposited on C type buffer: Cr(4 nm)/Fe(3 nm)/ Co(0.8 nm)/Cu(10 nm)/Co(1.2 nm)/Cu(0.84 nm)/Co(1.2 nm) Co(0.8 nm)/Cu(10 nm). using the Co k 1 ( = 0.1789 nm) wavelength. 5000 deposited on Cr(4 nm)/Fe(3 nm)/Co(0.8 nm)/Cu(10 nm) Cr /Co /Cu /Rh buffer. This experiment has been performed in the re- 6nm 0.8nm 10nm 2nm flection mode. Only one Bragg peak is observed around (111)Cu 2 = 51 . This peak is mainly the result of the 10 nm (110)Cr Cu decoupling layer since the other layers are too thin 4000 to contribute to the signal. Using the Bragg angle and the wavelength values, we find the average parameter I (a.u.) d0 = 0.205 nm, which corresponds to fcc structure with (111) texture along the growth direction. From the full width at the half maximum (FWHM), we obtain the co- 3000 herence length along the growth direction L = 7.2 nm. This is an indication of the size of the crystallites, which contribute to the diffraction. To confirm the observed tex- 40 45 50 55 60 65 ture, we have grown a [Cu(3 nm)/Co(3 nm)]20 multilayer 2 (degree) on the same substrate. The results are similar, with only one well resolved Bragg peak observed at the same angular Fig. 6. High angle X-ray spectrum recorded at room position. These results give an indication that there is a temperature for the sandwich: Cr(6 nm)/Co(0.8 nm)/ preferential growth orientation along the (111) direction. Cu(10 nm)/Rh(2 nm)/Cr(12 nm) using the Co k 1 ( = However it does not constitute an absolute proof of the ex- 0.1789 nm) wavelength. istence of the unique (111) texture because it corresponds to the strongest diffraction peak that we can detect on a fcc polycrystalline powder. Moreover, the rocking curve performed on the multilayer shows FWHM, of about 12 (Fig. 5). This large value is a good indication that the Cr with bcc (110) texture, the epitaxial relationship is in layers are not perfectly textured as will be confirmed by favor of fcc Cu with (111) texture. transmission electron microscopy. On the basis of the X-ray diffraction spectra, it is Figure 6 shows the high angle X-ray spectrum ob- clearly seen that both Fe/Cr and Cr buffers stabilize the tained for the following sandwich: Cr(6 nm)/Co(0.8 nm)/ fcc structure of Cu with a slight (111) preferential texture. Cu(10 nm)/Rh(2 nm)/Cr(12 nm). Rh gives a well defined Thus, the difference in the magneto-transport properties contrast with Cu, hence the 2 nm Rh layer has been de- cannot be attributed to the difference in the texture of posited in order to identify the nature of the Cu surface the different buffers. This conclusion has to be carefully by transmission electron microscopy. The figure shows two considered, since we know that the X-ray diffraction in Bragg peaks at 2 = 50.6 and 2 = 52.3 corresponding, /2 mode does not constitute an absolute proof of the respectively, to the fcc structure of Cu with (111) texture observed texture. For this reason, these samples have to and the bcc structure of Cr with (110) texture. Such a be analyzed by transmission electron microscopy in order result is not surprising, since it is well known that for the to confirm the previous observation. A. Dinia et al.: Buffer structure and magneto-transport properties of Co/Cu sandwiches 417 10 nm nm ists in a Cu-thickness window with a width smaller than 7.5 1 nm for an ideal interlayer with uniform thickness. This 5 nm thickness window with perfect antiferromagnetic coupling, 0 shrinks when the Cu thickness varies laterally. For large 0 nm characteristic lengths of these variations, the maximum -7.5 µm 0.00 0.25 0.50 0.75 1.00 coupling strength and the saturation field remain unaf- (a) Cr 2 fected, while the AF-coupling strength reduces at lengths 6nm/Co0.8nm/Cu10nm/Rh2nm 1x1µm RMS = 1.148 nm that are small as compared to the lateral coherence length 10 nm nm of the magnetization vector. This is exactly what happens 5 in the case of the Cr buffer where the average roughness 5 nm is of the same order as the oscillation period. Therefore, 0 the average interlayer exchange coupling is mainly ferro- 0 nm µm magnetic with a small antiferromagnetic component and -50.00 0.25 0.50 0.75 1.00 1.25 explains well the strong decrease of the exchange cou- (b) Cr 2 pling strength. Moreover, since the GMR is directly re- 4nm/Fe3nm/Co0.8nm/Cu10nm/Rh2nm 1x1µm RMS = 0.395 nm lated to the amount of the antiparallel alignment between Fig. 7. Atomic force microscopy scan images per- the magnetization vectors of the adjacent Co layers, the formed on Cr(6 nm)/Co(0.8 nm)/Cu(10 nm)/Rh(2 nm) GMR is strongly decreased in this case, as shown in the and Fe(4 nm)/Cr(3 nm)/Co(0.8 nm)/Cu(10 nm)/Rh(2 nm) GMR curve. However, for the Fe/Cr buffer the average buffers. The topography and the section analysis were per- roughness is around 0.35 nm, which leads to spacer Cu formed on 1 × 1 µm2 surfaces. thicknesses with mainly antiferromagnetic exchange cou- pling, and therefore, explains the high exchange coupling strength and the large GMR value observed for the sand- 4.2 Atomic force microscopy wich deposited on the Fe/Cr buffer type. In order to understand the physical origin of the difference in the exchange coupling and the GMR, 4.3 Transmission electron microscopy atomic force microscopy study have been performed on the Cr and Fe/Cr buffers with the following struc- ture: Cr(6 nm)/Co(0.8 nm)/Cu(10 nm)/Rh(2 nm) and High Resolution Transmission Electron Microscopy Fe(4 nm)/Cr(3 nm)/Co(0.8 nm)/Cu(10 nm)/Rh(2 nm). (HRTEM) allows the direct observation of atomic-scale This technique will bring some information on the rough- details of crystalline interfaces. This technique will ness of the buffer layers. The 2 nm Rh capping layer was bring some information on the effect of the buffer used to protect the Cu layer against oxidation. Scans were layer and its roughness on the coupling and GMR. made on 1 × 1 µm2, and at different positions on the sam- First of all, a cross-section has been performed on a ples in order to have an average roughness value and to AAF sandwich deposited on the buffer similar to the test the homogeneity of the surface. The distribution of C type buffer in order to confirm the (111) texture the tip position on the vertical axis has been analyzed as suggested by X-ray diffraction (see Fig. 4). The to obtain the average roughness. The tip-surface distance thickness of the different layers of this sample i.e., distributions are very well fitted by a Gaussian function, Cr(4 nm)/Fe(6 nm)/Cu(5 nm)/Co(5 nm)/Cu(5 nm)/ and the full width at the half maximum (RMS) values Co(5 nm)/Cr(4 nm) has been chosen in order to perform obtained for different scans are very similar, with a dis- easily optical diffractograms (Fig. 8). There is almost no crepancy between the values never exceeding 15%. This Z-contrast between Cr, Fe and Cu, due to their close scat- is a first indication that our surfaces are homogeneous. tering potential values. In this case we have used the thick- Figure 7 presents the topography of the two samples us- ness of the different layers as determined by Electron Spec- ing surface 1 × 1 µm2 area. The average RMS values troscopy Imaging to determine the interfaces. The optical are about 1.16 nm and 0.35 nm, respectively for Cr and diffractogram of the Cu and Fe/Cr layers of the buffer Fe/Cr buffers, which means that the average roughness reported in Figure 8 clearly shows the fcc (111) and bcc is very different between the two buffers. Thus, the sur- (110) respectively as growth direction. Nevertheless, the face of Fe/Cr/Co/Cu buffer is relatively flat with very (100) growth direction for both layers has also been ob- localized islands, which is in favor of a nice antiferromag- served as shown in the Figure 8. This is a proof that the netic coupling. This is a direct experimental evidence of fcc (111) direction is not the unique growth direction. the difference in the buffer surface roughness between the Two other sandwiches used for the HRTEM cross- two buffers, which is at the origin of the difference in the section have been especially chosen in order to make magneto-transport properties. Indeed, as already reported the understanding of the HRTEM images easier. They for Co/Cu system, the exchange coupling oscillates be- correspond to C and A types buffers with the fol- tween ferromagnetic and antiferromagnetic coupling over lowing structures: Cr(4 nm)/Fe(3 nm)/Co(0.8 nm)/ a thickness range of the Cu interlayer of typically 0.5 nm. Cu(10 nm)/Rh(2 nm)/Cr(12 nm) and Cr(6 nm)/ In other words, the antiferromagnetic coupling only ex- Co(0.8 nm)/Cu(10 nm)/Rh(2 nm)/Cr(12 nm). The 2 nm 418 The European Physical Journal B 020 -111 11-1 -111 000 020 000 [101] 11-1 Cu [101] Cu Fe -220 -202 02-2 000 200 Fe [111] 1-10 110 000 Cr [001] Cr SiO SiO 3 nm 3 nm Si Si Fig. 8. Cross-sectional HRTEM pictures and their optical diffractograms for the C type buffer: Cr(4 nm)/ Fe(3 nm)/Co(0.8 nm)/Cu(10 nm) sandwiches for: (a) with (111) Cu growth direction and (b) (100) Cu growth direction. thin Rh layer has been deposited onto the Cu to high- light the roughness at the Cu surface, thanks to strong Z- contrast between Rh and Cu. On the other hand the 12 nm Cr layer has been deposited onto the Rh layer to protect Cr the Cu/Rh interface during the mechanical milling. It also Rh allows a better Cu/Rh interface resolution as compared to Co, Cu an uncovered layer, close to the glue used for cross-section processing. Fe Cr Figure 9 shows the HRTEM cross-sectional images ob- SiO tained for both sandwiches. These images clearly show the difference in the Cu surface roughness between the Si two buffers. The roughness of the Cu surface is more pro- nounced in the case of the Cr buffer (type A) than in the case of Cr/Fe buffer (type C). From these images we ex- tract an average roughness values, which correspond to 0.3 Cr and 1.2 nm, respectively for the Fe/Cr and the Cr buffers. Rh These values are in good agreement with those determined by atomic force microscopy. Co, Cu In order to have precise information on the texture between these two buffers and also on the size of the Cu Cr SiO grains, plan-view diffraction has also been performed on both samples and their Selected Area Diffraction (SAD) Si patterns are reported in Figure 10. The ring diffraction patterns suggest that both samples are polycrystalline giv- Fig. 9. Cross-sectional HRTEM images of (a) C type buffer: ing rise to the same diffraction planes. To support the Cr(4 nm)/Fe(3 nm)/Co(0.8 nm)/Cu(10 nm)/Rh(2 nm)/ polycrystalline character of our samples, we have slightly Cr(12 nm) and (b) A type buffer: Cr(6 nm)/Co(0.8 nm)/ tilted the electron beam by about 10 with respect to Cu(10 nm)/Rh(2 nm)/Cr(12 nm) deposited on silicon sub- the growth direction. The results are exactly the same strates. A. Dinia et al.: Buffer structure and magneto-transport properties of Co/Cu sandwiches 419 5 Conclusion To conclude, this work allows us to show a direct correlation between surface roughness and magneto- transport properties. The best giant magnetoresistance and the largest antiferromagnetic coupling strength for the Co/Cu/Co AAF sandwich have been obtained us- ing the Cr/Co/Fe/Cu buffer which gives a smaller Cu surface roughness around 0.35 nm. X-ray diffraction and Transmission Electron Microscopy results show no sensi- tive difference in the crystalline quality and the prefer- ential growth texture between the buffers, which exclude this hypothesis as a possible origin for the difference in the magneto-transport properties. References 1. P. Gr¨unberg, R. Schreiber, Y. Pang, M.B. Brodsky, H. Sowers, Phys. Rev. Lett. 57, 2442 (1986). 2. M.N. Baibich, J.M. Broto, A. Fert, F. Nguyen van Dau, F. Petroff, P. Etienne, G. Creuzet, A. Friederich, J. Chazelas, Phys. Rev. Lett. 61, 2472 (1988). 3. S.S.P. Parkin, R. Bhadra, K.P. Roche, Phys. Rev. Lett. 66, 2152 (1991). 4. D.H. Mosca, F. Petroff, A. Fert, P.A. Schroeder, W.P. Pratt Jr, R. Laloe, J. Magn. Magn. Mater. 94, L1 (1991). 5. W.F. Egelhoff Jr, M.T. Kief, Phys. Rev. B. 45, 7795 (1992). 6. D. Greig, M.J. Hall, C. Hammond, B.J. Hickey, H.P. Ho, M.A. Howson, M.J. Walker, N. Wiser, D.G. Wright, J. Magn. Magn. Mater. 110, L239 (1992). 7. M.T. Johnson, S.T. Purcell, N.W.E. McGee, R. Coehoorn, Fig. 10. TEM planar selected area diffraction patterns J. aan de Stegge, W. Hoving, Phys. Rev. Lett. 68, 2688 for: (a) C type buffer Cr(4 nm)/Fe(3 nm)/Co(0.8 nm)/ (1992). Cu(10 nm)/Rh(2 nm)/Cr(12 nm) and (b) A type buffer: 8. M.T. Johnson, R. Coehoorn, J.J. de Vries, N.W.E. McGee, Cr(6 nm)/Co(0.8 nm)/Cu(10 nm)/Rh(2 nm)/Cr(12 nm) de- J. aan de Stegge, P.J.H. Bloemen, Phys. Rev. Lett. 69, 969 posited on silicon substrates. The rings are compared to the (1992). expected radii for the diffraction Bragg peaks of Cu and Fe. 9. A. Schreyer, K. Br¨ohl, J.F. Anker, C.F. Majkrzak, Th. Zeidler, P. B¨odeker, N. Metoki, H. Zabel, Phys. Rev. B. 47, 15334 (1993). with the strongest ring corresponding to the fcc (111) Cu 10. C. Dupas, E. Kolb, K. Le Dang, J.P. Renard, P. Veillet, E. and the bcc (110)Cr/Fe, and confirm our hypothesis that V´elu, D. Renard, J. Magn. Magn. Mater. 128, 361 (1993). there is no well defined preferential texture for both sam- 11. N. Persat, A. Dinia, J.P. Jay, C. M´eny, P. Panissod, J. ples that can be observed in this geometry. However, the Magn. Magn. Mater. 164, 37 (1996). presence of the (111) diffraction peak is not surprising 12. G. Rupp, K. Schuster, J. Magn. Magn. Mater. 121, 416 since the FWHM of the rocking curve around the (111) Cu (1993); H.A.M. van den Berg, Fundamentals, Analysis and Industrial Application, edited by U. Hartmann (Springer diffraction peak is of about 12 . This corresponds mainly Verlag, Berlin, 1999), pp. 179­262. to the large angular distribution of the (111) Cu grains 13. N. Persat, H.A.M. van den Berg, A. Dinia, J. Magn. Magn. around the normal to the film and not to the grain sizes, Mater. 165, 446 (1997). which are of the order of 30 nm as determined by dark 14. N. Persat, H.A.M. van den Berg, K. Cherifi-Khodjaoui, A. field images. Dinia, J. Appl. Phys. 81, 4748 (1997); N. Persat, H.A.M. The structural analysis has shown that there is no van den Berg, A. Dinia, Phys. Rev. B. 62, 3917 (2000). clear difference in the crystalline character between the 15. S. Schmeuber, A. Hubert, G. Rupp, J. Magn. Magn. Cr and Fe/Cr buffers. Both give rise to a polycrystalline Mater. 166, 267 (1997). Cu layer with a slight texture along the (111) direction. 16. G. Rupp, H.A.M. van den Berg, IEEE Trans. Magn. 29, However atomic force microscopy and transmission elec- 3102 (1993). tron microscopy have clearly evidenced a strong difference 17. J. Benedict, R. Anderson, S.J. Klepeis, Mater. Res. Soc. in the average roughness between these two buffers. Proced. 254, 121 (1992).