PHYSICAL REVIEW B VOLUME 54, NUMBER 18 1 NOVEMBER 1996-II Effects of annealing on the magnetoresistance and structure of Fe/Cr 110... superlattices Jose M. Colino* and Ivan K. Schuller Physics Department 0319, University of California at San Diego, La Jolla, California 92093-0319 V. Korenivski and K. V. Rao Department of Condensed Matter Physics, Royal Institute of Technology, 10044 Stockholm, Sweden Received 28 August 1995; revised manuscript received 29 February 1996 We have performed magnetotransport, magnetization, and structural experiments on sputtered Fe 30 Å /Cr 12 Å 110 superlattices that were annealed at temperatures up to 400 °C. Interestingly, their giant magnetoresistance ( ) is enhanced at intermediate temperatures, and strongly decreased at higher temperatures. If normalized to the antiferromagnetic coupling fraction of sample, the magnetoresistance in- creases over the entire annealing range. From low-angle x-ray-diffraction measurements, the enhancement of arises from an interface redistribution because of either a slight interdiffusion, less correlated interfaces, or both. Further annealing causes extreme interdiffusion that is detrimental for the magnetoresistance because of a loss of antiferromagnetic coupling. S0163-1829 96 01642-6 Giant magnetoresistance GMR in artificial magnetic/ show a significant enhancement of GMR at intermediate normal multilayers has attracted much recent attention.1 Ex- temperatures 150 °C, 12 h and 300 °, 30 min , and at all perimental and theoretical work have revealed many of basic temperatures if normalized to the antiferromagnetic coupling ingredients of the GMR, including the interplay of interlayer fraction of the samples. The x-ray diffraction indicates antiferromagnetic AF coupling2 with spin-dependent elec- changes in the interface structure at the annealing tempera- tron scattering,3 the role of the type of elements in the layers, tures for greatest GMR. the crystalline structure,2,4 etc. However, a satisfactory over- Fe 30 Å /Cr 12 Å 10 110 multilayers were grown at all model connecting physical and electronic structure with room temperature on Si 100 wafers by dc magnetron sput- GMR has not yet emerged. For instance, the role of the spac- tering in a 4 mTorr argon atmosphere. Further deposition ers' electronic structure is strongly debated.4­6 Another criti- details have been reported elsewhere.8 The samples's struc- cal issue is the coupling between magnetic layers. The im- ture was characterized by high- and low-angle x-ray diffrac- portance of the Ruderman-Kittel-Kasuya-Yosida interaction tion using a Rigaku rotating anode diffractometer with Cu compared to others such as the classical dipolar interaction7 K radiation, in specular and off-specular scans rocking caused by interface roughness, is still unclear. In this regard, curves . Four lead magnetotransport measurements were per- the roughness and interface structure has been shown to be important factors for increasing the GMR.8­10 To date sev- formed at 4.2 and 77 K and magnetic fields up to 5 T. The eral approaches have been successfully used to enhance the magnetic field was in the film plane and perpendicular to the GMR ( / ) or to some extent. Interface roughness dc electrical current. The absolute value of the resistivity was can be induced during growth by changing the deposition determined either by the Van der Pauw method or directly parameters, for instance by increasing Ar sputtering using photolithographically patterned samples. The magneti- pressures.8,11 This way, the GMR of Fe 30 Å /Cr 18 Å zation was measured by superconducting quantum interfer- 110 at 10 K was increased from 6 to 13 %.8 The in- ence device magnetometry at 10 K. terface structure can also be altered during growth and the Two different thermal treatments were performed. In the GMR dramatically increased by addition of thin Co interface first one, the sample is annealed at 100 °C for 30 min, then layers in Ni cooled to low temperatures and the magnetic and transport 81Fe 19/Cu multilayers.12 Post-deposition Xe ion irradiation of Fe/Cr 110 changes the GMR in a con- properties measured. Then the sample is annealed at 150 °C trolled fashion.13 A reproducible method, which may be of for 30 min, after which magnetic and transport properties are technological importance, is simple annealing in vacuum or measured again. This procedure is repeated several times, inert gases. Petroff et al. reported GMR increases from 14 to increasing the annealing temperature by 50 °C at every step, 27.3 % at helium temperature by annealing an Fe/Cr 100 up to an annealing temperature of 450 °C. Heating and cool- sample at 300 °C for 1 h, and speculated that this was caused ing rates were greater than 40 °C/min. In the second method, by changes in the interface roughness and scattering.10 Ther- the as-prepared film was cut into several samples. Each mal treatments produce structural changes both at interfaces sample was annealed once at a single temperature in the and in the bulk of multilayers, including chemical range 100­400 °C for 12 h heating/cooling rates 10 °C/ interdiffusion,14 changes in crystallinity, etc., which can be min . Structural and magnetotransport measurements fol- studied with x-ray diffraction.15 lowed the annealings. The first method is well suited for In this paper, we present a systematic study of structure following the process in one sample, although the effect is and GMR in Fe/Cr 110 annealed at an inert gas atmosphere cumulative, and the second is useful to study temperature Ar, He for 30 min or 12 h. Magnetotransport experiments dependences. 0163-1829/96/54 18 /13030 4 /$10.00 54 13 030 © 1996 The American Physical Society 54 EFFECTS OF ANNEALING ON THE . . . 13 031 FIG. 2. Full width at half maximum FWHM of the ­2 110 peak of the 12h annealed samples (2 ) . The lines are guides to the eye. (1 MR/MS), where the giant spin-dependent scattering is localized, and plotted the magnetoresistance normalized to such a fraction. Figure 1 c shows a steady increase of nor- malized GMR over all the temperature range tested for samples annealed for 30 min. This shows that the enhance- ment is even more important at higher temperatures and that the decrease of Figs. 1 a and 1 b is caused by a pro- gressive loss of antiferromagnetic order due to magnetic shorts. Since the chemical interdiffusion in Fe/Cr 110 mul- tilayers is significant at 350 °C for 3 h,14 interdiffusion is a likely explanation for our results above 300 °C 1/2 h . High- and low-angle specular and nonspecular rocking curves x-ray-diffraction measurements were carried out for the 12 h annealed samples. As-prepared and annealed samples showed one first-order superlattice peak in the high- FIG. 1. and versus annealing temperature for a 12 h angle spectra about the 110 Bragg reflection. Full widths at annealed pieces of a single sample and b one sample annealed in half maximum FWHM of the 110 reflection are plotted in 30 min steps. The two data at room temperature a are measure- Fig. 2 versus annealing temperature. Above 200 °C, there is ments on different pieces to check the reproducibility. c Magne- a remarkable decrease of the specular peak width Fig. 3 b toresistance normalized to the AF coupling fraction of sample for that, according to Scherrer's formula, corresponds to an en- the multilayer annealed in 30 min steps, and the AF coupling frac- largement of the average crystallite size ( 240 Å by a fac- tion of sample (1 MR /MS). Saturation magnetization (MS) is tor of 1.6 over that of the as-grown sample ( 150 Å , typically 1200 emu/cc, about 70% of the bulk MS value. The lines perhaps resulting from bulk defect annihilation, decrease of are guides to the eye. atomic strain, etc. Nearly constant rocking curve widths, about the 110 reflection 15°­17° , imply minimal changes in crystalline orientation during annealing. Figures 1 a and 1 b show the magnetoresistance at 4.2 K Figure 3 shows low-angle x-ray diffraction LAXRD ( ) and the saturation resistivity ( ) as a function of an- specular spectra for the 12 h annealed samples. Up to third- nealing temperature for the two types of heat treatments de- order superlattice Bragg peaks confirm the multilayer peri- scribed above. Whereas the resistivity exhibits a slight mono- odicity of 42 Å repeated 10 times. An important result is tonic increase in the 20­420 °C range, shows an that, even up to 350 °C, these low-angle spectra exhibit finite interesting nonmonotonic behavior; it rises to a maximum at size and well developed Bragg peaks similar to those of the intermediate temperatures around 300 °C, 30 min and as-prepared samples. 150 °C, 12 h , increasing over initial values of as-prepared Figure 4 a shows the ratio of intensity of second-order samples by 50% a and 28% b , respectively, and thereafter (n 2) superlattice peaks over the background intensity, drops to very small values. Actual values of / are from ­2 scans Fig. 3 , as a function of annealing tem- 8% a and 10% b for as-prepared samples, and upon perature. We chose the second-order peaks since the finite- annealing they reach 11.2% a and 12.2% b at the size peaks around it are small and thus allow a more precise maximum. This increase of is the main focus of this determination of the background level. The decrease in paper. Bragg peak intensities without broadening, for annealing To distinguish the effects of changing coupling behavior temperatures above 100 °C, implies an increased from more intrinsic effects, we have estimated the fraction interdiffusion.15,16 of sample which is antiferromagnetically coupled Further insight from LAXRD can be gained by studying 13 032 COLINO, SCHULLER, KORENIVSKI, AND RAO 54 FIG. 3. Low-angle specular spectra of the as-prepared and an- nealed samples for 12 h . The spectra are offset for clarity. Note the well developed finite-size peaks up to 350 °C. off-specular diffraction rocking curves at low-angle super- FIG. 4. a Intensity of the second-order superlattice Bragg peak lattice reflections. This diffusion scattering results from in- over the background intensity for the 12 h annealed samples. b plane structural inhomogeneities associated with the indi- Ratio of the specular to diffuse intensity (Q) in the second-order vidual interfaces or layers.17,18 In principle, diffuse scattering superlattice peak for the 12 h annealed samples. Lines are guides to arises from vertically correlated and uncorrelated roughness. the eye. The inset shows a rocking curve about a second-order Correlated roughness, i.e., interfacial roughness replicated low-angle Bragg peak as an example from which the factors Q were from layer to layer, originates from defects in the substrate or calculated. is due to growth front effect which give correlations in film morphology. Uncorrelated roughness is random from layer between interdiffusion and correlated or uncorrelated rough- to layer. The inset of Fig. 4 b shows the specular and diffuse ness. However, since interdiffusion may occur at atomic intensities in a low-angle rocking curve, their ratio denoted length scales, interdiffusion is expected to be uncorrelated as Q. Studies of Fe/Cr multilayers grown at different sub- from layer to layer. strate temperatures claimed a correlation of Q with the At moderate temperatures, for which reaches a maxi- GMR.19 Several models were developed for the x-ray scat- mum, the correlated roughness decreases Fig. 4 b , while tering from rough multilayers which exhibit the two kinds of the interdiffusion increases Fig. 4 a . The optimum is roughness mentioned above.18,20 In all these models, despite the result of a competition between two mechanisms; an in- the different functional expressions for Q, a larger Q implies crease in due to a less correlated interface roughness a smaller rms correlated roughness, C . Here, the Q's for and/or a moderate interdiffusion, and a decrease in due to every Bragg peak increase with annealing temperatures up to magnetic shorts which drastically lower the AF coupling 300­350°C and particularly that of the second-order Bragg fraction of the sample. The fact hat the spin-dependent scat- peak Fig. 4 b . According to the models mentioned above, tering at the interface is especially important in the Fe/Cr this implies a decrease of the vertically correlated roughness system23 further strengthens the arguments presented above. with annealing offset ­2 scans proved the existence of A similar, although smaller, overall dependence of on correlated roughness to some extent since there is still Bragg ion dosage was observed in Xe -irradiated Fe/Cr 110 peaks with the same periodicity as the specular reflections . multilayers13 from 7% to a maximum 8% at 77 K . Cr Therefore, due to annealing up to 300­350°C the interfaces alloying of the Fe layer in Fe/Cr 110 can also increase the become less correlated vertically, and at the same time more GMR, from 6.9 to 12.2%.21,22 Further comparison is un- interdiffused. Interdiffusion may increase spin-dependent warranted since the number of bilayers cumulative disor- scattering since chromium alloying in the Fe layers is known der , crystallinity, etc., are different. to enhance GMR.21,22 However interdiffusion is an atomic In summary, annealing can enhance the GMR in scale ``roughness,'' therefore on a scale much shorter than Fe/Cr 110 multilayers, although the annealing temperature the x-ray coherence length. In the absence of detailed struc- range is narrow. Large interdiffusion and bulk crystallinity tural refinement15 which was not possible here it is hard to improvement are generally observed after high-temperature make a definite statement regarding the quantitative relations anneals and lead to magnetic shorts, which is detrimental for 54 EFFECTS OF ANNEALING ON THE . . . 13 033 GMR. A less correlated interface roughness and/or slight in- and Dr. P. Belien for a critical reading of the manuscript and terdiffusion, at moderate temperatures, increases and useful discussions, and acknowledge financial support from therefore the GMR. the U.S. Department of Energy. J.M.C. acknowledges the We thank T. Ryabtseva for help in the initial stages of this Spanish Secretari´a de Estado de Universidades e Investiga- work. We are indebted to Dr. D. Lederman, Dr. R. Schad, cio´n. *Present address: Instituto de Ciencia de Materiales de Madrid- 13 D. Kelly, I. K. Schuller, V. Korenivski, K. V. Rao, K. K. Larsen, C.S.I.C., Cantoblanco 28049, Madrid, Spain. J. Bottiger, E. M. Gyorgy, and R. B. van Dover, Phys. Rev. B 1 M. N. Baibich, J. M. Broto, A. Fert, F. Nguyen Van Dau, and F. 50, 3481 1994 . Petroff, Phys. Rev. Lett. 61, 2472 1988 . 14 H. Nakajima, K. Nonaka, Y. Obi, and H. Fujimori, J. Magn. 2 S. S. P. Parkin, N. More, and K. P. Roche, Phys. Rev. Lett. 64, Magn. Mater. 126, 176 1993 . 2304 1990 . 15 I. K. Schuller, Phys. Rev. Lett. 44, 1597 1980 ; W. Sevenhans, 3 B. A. Gurney, V. S. Speriosu, J.-P. Nozieres, H. Lefakis, D. R. M. Gijs, Y. Bruynseraede, H. Homma, and I. K. Schuller, Phys. Wilhoit, and O. U. Need, Phys. Rev. Lett. 71, 4023 1993 , and Rev. B 34, 5955 1986 ; E. E. Fullerton, I. K. Schuller, H. references therein. Vanderstraeten, and Y. Bruynseraede, ibid. 45, 9292 1992 . 4 E. E. Fullerton, M. J. Conover, J. E. Mattson, C. H. Sowers, and 16 H. Vanderstraeten, D. Neerinck, K. Temst, Y. Bruynseraede, E. S. D. Bader, Phys. Rev. B 48, 15 755 1993 . E. Fullerton, and I. K. Schuller, J. Appl. Crystallogr. 24, 571 5 A. Berger, and H. Hopster, Phys. Rev. Lett. 73, 193 1994 . 1991 . 6 S. S. P. Parkin, T. A. Rabedeau, and A. Modak unpublished . 17 J. B. Kortright, J. Appl. Phys. 70, 3620 1991 . 7 D. Altbir, M. Kiwi, R. Ramirez, and I. K. Schuller, J. Magn. 18 D. E. Savage, J. Kleiner, N. Schimke, Y.-H. Phang, T. Jankowski, Magn. Mater. 149, L246 1995 . J. Jacobs, R. Kariotis, and M. G. Lagally, J. Appl. Phys. 69, 8 E. E. Fullerton, D. M. Kelly, J. Guimpel, I. K. Schuller, and Y. 1411 1991 . Bruynseraede, Phys. Rev. Lett. 68, 859 1992 . 19 P. Belien, R. Schad, C. D. Potter, G. Verbanck, V. V. Mosh- 9 J. Colino, I. K. Schuller, R. Schad, C. D. Potter, P. Belien, G. chalkov, and Y. Bruynseraede, Phys. Rev. B 50, 9957 1994 . Verbanck, V. V. Moshchalkov, and Y. Bruynseraede, Phys. Rev. 20 R. L. Headrick and J.-M. Baribeau, Phys. Rev. B 48, 9174 1993 . B 53, 766 1996 . 21 N. M. Rensing, and B. M. Clemens, J. Appl. Phys. 76, 6617 10 F. Petroff, A. Barthe´le´my, A. Hamzic, A. Fert, P. Etienne, S. 1994 . Lequien, and G. Creuzet, J. Magn. Magn. Mater. 93, 95 1991 . 22 L. H. Chen, T. H. Tiefel, S. Jin, R. B. Van Dover, E. M. Gyorgy, 11 N. M. Rensing, A. P. Payne, and B. M. Clemens, J. Magn. Magn. and R. H. Fleming, Appl. Phys. Lett. 63, 1279 1993 . Mater. 121, 436 1993 . 23 M. J. Hall, B. J. Hickey, M. A. Howson, M. J. Walker, J. Xu, D. 12 S. S. P. Parkin, Appl. Phys. Lett. 61, 1358 1992 . Greig, and P. Wiser, Phys. Rev. B 47, 12 785 1993 .