PHYSICAL REVIEW B VOLUME 60, NUMBER 18 1 NOVEMBER 1999-II Aperiodical oscillation of interlayer coupling in epitaxial Co/Ir 001... superlattices H. Yanagihara* and Eiji Kita Institute of Applied Physics and Center for Tsukuba Advanced Research Alliance (TARA), University of Tsukuba, Tsukuba, 305-8573 Japan M. B. Salamon Department of Physics, University of Illinois, Urbana, Illinois 61801 Received 12 May 1999 High quality epitaxial cobalt-iridium superlattices were successfully grown on MgO 001 substrates via molecular-beam epitaxy and were found to exhibit anomalous oscillatory interlayer exchange coupling with a strong coupling constant. The aperiodical antiferromagnetic AF peaks appear at Ir layer thicknesses of 5 Å, 15 Å, and 33 Å. Maximum magnetoresistance MR ratios of 1.1% are obtained and are almost the same for the first and the second AF peaks. The interlayer coupling corresponding to the first AF peak is so strong that the magnetization did not saturate even in an applied field up to 90 kOe. The estimated value of the interlayer exchange constant by MR curves at room temperature is more than 2.2 erg/cm2 close to the first AF peak. A distinction of the in-plane magnetization process between 100 and 110 directions is observed and the easy axis of magnetization is found to be along the 110 axis in plane. The aperiodic oscillations are thought to indicate multiperiodicity of the interlayer exchange constants. S0163-1829 99 01742-7 I. INTRODUCTION caused by large lattice misfits between spacer metals and ferromagnetic metals. There are numerous spanning vectors Since the discovery1 of interlayer coupling between adja- on the Fermi surface of the transition metals, each of which cent ferromagnetic layers that oscillates with the thickness of may be associated with an oscillation period.6 In the case of nonmagnetic spacers, numerous experimental and theoretical multilayers composed of transition metal spacers, we expect, studies of the effect have been reported. Many combinations therefore, to observe multiperiodicity consistent with two or of materials, in sandwiches and multilayers, have been ex- more long-range periodicities. amined in terms of their coupling and the oscillation Superlattices composed of a 3d ferromagnetic metal Co periods.2 In multilayers, the oscillation periods are between 8 and 5d transition metal spacer Ir have been found to ex- Å and 12 Å for almost any metal spacer3 and no multiperi- hibit magnetic properties commonly observed in other odicity has been observed. From a theoretical perspective, ferromagnetic/transition or noble metal systems, i.e., perpen- the oscillation periods are found to be equal to 2 / q dicular magnetic anisotropy,9 and magnetic exchange cou- i , where q pling that depends on spacer layer thickness.3,10­12 Although i is one of the extremal spanning vectors that con- nects stable points of the Fermi surface of the spacer material the interlayer coupling in Ir systems is much stronger than along the growth direction.4­6 Multiperiodicities that imply that in other 3d or noble-metal systems Co/Ir is reported to both short- about 2 monolayers and long-range about 5­8 be the second strongest antiferromagnetic AF coupling sys- monolayers periodicities of interlayer coupling have been tem in Parkin's pioneering work3 , there have been few re- reported in several well-controlled sandwich systems. In ports of magnetic properties or superlattice structure. Re- multilayer systems, however, short-range coupling oscilla- cently, additional work on the Co/Ir multilayer systems has tions have not been observed, despite theoretical predictions, appeared, focusing on the interlayer coupling and magne- presumably a consequence of interfacial roughness.4 In gen- toresistance MR .10,11 However, the samples studied were eral, in order to study interlayer coupling, most reports fo- made by sputtering method, so that the multilayers were cused on noble metals or on 3d transition metals, which composed of 111 preferred crystalline texture. In order to show antiferromagnetism as spacer materials. Noble metals study interlayer coupling, it is much better to treat single have rather simple Fermi surfaces where there are a few crystal samples; that is, epitaxial superlattices. This poses an extremal spanning vectors, and less lattice misfit with a experimental challenge, as there is a large lattice misfit be- magnetic-layer element, so that one can easily compare the- tween fcc-Co and fcc-Ir, expected to be about 8%. In order to oretical predictions and experimental results; there are re- grow Co/Ir epitaxially, then, well-controlled and optimized markable correspondences between theoretical and experi- growth conditions are necessary. In this paper, we report the mental oscillation periods.7 Recently, Unguris et al., using growth of Co/Ir 001 superlattice structures prepared with carefully grown Fe/Au/Fe trilayers, showed experimental re- molecular-beam epitaxy MBE , and on the dependence of sults consistent in not only the periodicity, but also the mag- magnetic properties on Ir layer thickness. nitude of exchange coupling.8 Unfortunately, there are fewer studies of superlattices consisting of 4d or 5d transition met- II. EXPERIMENT als as a spacer, possibly because of the more complicated All samples were grown at the University of Tsukuba Fermi surface and prospective changes in electronic structure using a conventional MBE system SEO-5 made by Japan 0163-1829/99/60 18 /12957 6 /$15.00 PRB 60 12 957 ©1999 The American Physical Society 12 958 H. YANAGIHARA, EIJI KITA, AND M. B. SALAMON PRB 60 FIG. 1. RHEED patterns of a Co 7 Å/Ir 11 Å 001 superlattice during the growth process. The incident electron beam is parallel to MgO 100 and accelerated to 15 kV. a After driving off the impurities by substrate heating. b Reflection pattern of Ir buffer layer. c The tenth Co layer. d The tenth Ir layer. e The fiftieth Co layer. f The fiftieth Ir layer. The final Ir layer thickness is 50 Å for prevention of oxidiza- tion. SEED LAB. Ltd. , which includes reflected high-energy graphic films. In order to measure MR, samples were cleaved electron diffraction RHEED equipment. Single crystal along MgO 100 and 010 directions, and were shaped in a MgO 001 substrates, polished for epitaxial growth, were rectangle 2 8 mm2 in size. MR measurements were car- used. Layer thickness and deposition rates for each source ried out at room temperature and magnetic fields were ap- were monitored by crystal oscillator and controlled by a plied in plane up to 15 kOe in usual cases; some samples, commercial deposition controller. Films for nominal thick- which show high saturation fields, were measured up to 90 ness calibration were measured via multiple beam interfer- kOe. Both the measurement current of 0.1 mA and typical ometry and contact step meter. The difference between both applied magnetic fields were parallel to the in-plane methods is negligible. MgO 100 . Magnetization measurements were performed at Prior to growth, the substrates were heated to 600 °C for room temperature with a commercial superconducting quan- three hours or more in a growth chamber. Layers of 12 Å Fe tum interference device magnetometer with magnetic fields and 50 Å Pt were subsequently grown on the MgO 001 up to 50 kOe applied mainly along the in-plane MgO 100 . substrate as seed layers at 600 °C in order to relax the misfit between MgO 001 substrates and Ir 001 buffer layers and to facilitate the growth of a flat surface.13 After deposition of seed layers, an Ir layer of 500 Å was deposited at the same III. RESULTS AND DISCUSSIONS temperature. Growth rates were 0.1 Å/sec, 0.1 Å/sec, and 0.2 Figure 1 shows RHEED patterns of each step during the Å/sec, for Fe, Pt, and Ir, respectively. Co/Ir superlattices growth of Co 7 Å/Ir 11 Å 50(001). The RHEED pattern were grown at 60 °C to suppress alloying and interdiffu- shown in Fig. 1 a is the MgO 001 substrate after heating sion. The deposition rate was controlled to be 0.15 0.02 for 3 h. At first, the pattern of seed layer Fe growth appears Å/sec. A series of superlattices were designed with fixed Co as smeary spots. After deposition of 50 Å Pt, the patterns layer thickness of 7 Å, while Ir layer thicknesses were varied become broad streaks, meaning that a flat film surface has nominally from 5 Å to 40 Å. The repetition number of bi- developed. Figure 1 b shows the pattern of the 500 Å Ir layers for each sample was 50 and a final 50-Å-thick Ir layer buffer layer. The sharp streak pattern of Ir buffer layers was deposited in order to prevent oxidation. The base pres- grown over 300 Å indicates a sufficiently flat surface. We sure of the deposition chamber was approximately 1 10 9 could not observe a 5 1 surface reconstruction of Ir 001 torr and the typical deposition pressure was better than 4 from the pattern of 500-Å-thick Ir layer. The crystal relation- 10 8 torr during superlattice growth. We determined each ship MgO 100 001 Ir 100 001 is deduced from these layer thickness using intervals of superlattice peaks obtained RHEED patterns. by x-ray diffraction patterns by fitting the slope and intercept The RHEED patterns from the tenth Co layer, the tenth Ir of the superlattice intervals versus nominal Ir layer thickness. layer, the final 50th Co layer, and Ir capping layer are The intercept and the slope correspond respectively to the Co shown in Figs. 1 c , 1 d , 1 e , 1 f , respectively. No remark- layer thickness and the Ir layer correction factor between able change of RHEED pattern, except the brightness, was nominal and actual thickness. The obtained intercept Co observed in the whole process of the superlattice growth, that layer thickness is around 6.7 Å and the slope is 0.9. While is, from the first Co layer deposition to the final Ir layer the difference of Co layer thickness between nominal and deposition. The streak patterns indicate that the growth mode that estimated by the intervals is less than 5%, we denote the of superlattices is layer-by-layer-like growth. During Co Co thickness as 7 Å. On the other hand, the difference for the layer growth, the streak patterns blur a little and their bright- Ir layers is not negligible, and we express the rounded off ness dims. On the other hand, during Ir layer growth, the thickness after correction by the factor 0.9 in the following. streak patterns tend to brighten and slightly shorten. This The structure of superlattices was examined in situ with tendency was observed in all samples. RHEED and ex situ with x-ray diffractometry by Cu-K X-ray diffraction patterns of some samples comprised of radiation. The RHEED patterns were recorded on photo- different Ir layer thicknesses are shown in Fig. 2. A strong, PRB 60 APERIODICAL OSCILLATION OF INTERLAYER . . . 12 959 FIG. 2. X-ray diffraction patterns of Co 7 Å/Ir (t FIG. 3. Typical MH curves of Co 7 Å/Ir (tIr) Å 001 superlat- Ir) Å 001 su- perlattice. The t tices at room temperature. The tIr are given in the figure. In samples Ir's value is denoted in each figure. designed with thin Ir thickness, the magnetization does not saturate sharp peak placed at 2 47.4° originates from the buffer Ir even at 90 kOe. layer. Weak, broad peaks around 2 20° and 2 74° are contributions from the seed layers considered as FePt or- the Co layer has bcc structure, the envelope center could not dered alloy. Superlattice peaks are observed in the whole be observed in this region. The XRD profiles around the region from small to high angle; however, there is no evi- 004 reflection support the picture of fcc-Co diminished dence of 011 or 111 stacking in any samples. In the case along the growth direction. Our preliminary results of off- of Co 7 Å/Ir 14 Å axis x-ray diffraction measurements using a four circle dif- 50(001), the full width at half maximum of a Co/Ir 002 peak is less than 0.35°, which yields a crystal fractometer also support fcc structure of Co layers. coherence length larger than 250 Å. The existence of sharp Figure 3 shows some typical MH curves of and higher-order satellite peaks indicates that the superlat- Co 7 Å/Ir(tIr) 50(001) measured at room temperature. From tices have clear and sharp interfaces with little interface mix- the results of in-plane and perpendicular magnetization pro- ing and high superlattice periodicity. cesses, it is found that the easy axis of magnetization lies in Co thin films have the possibility of two different meta- the film plane irrespective of the Ir layer thickness. All stable crystalline structures at room temperature, i.e., fcc samples designed with tIr 9 Å show little remanence, and phase and bcc phase. It is believed that film growth condi- have such remarkably high saturation fields (HS) that we tions and growth orientation decide the Co structure. If a Co could not saturate their magnetization even in H 50 kOe. layer has bcc structure, the lattice misfit between fcc-Ir and This behavior means that the adjacent Co layers couple bcc-Co is expected to be 4%,14 while in the case of strongly antiferromagnetically and the layer structures of the fcc-Co, the misfit is 8%. In the step model of x-ray superlattices must be sufficiently perfect that direct coupling diffraction XRD pattern analysis of superlattices,15 the in- between adjacent Co layers is negligible even for Ir layer tensity of superlattice peaks is mainly determined by the thickness of only 2­5 monolayers. The shape of the MH form-factor envelopes of the layers that comprise each bi- curves is strongly dependent on Ir layer thickness. The layer. Although the superlattice peaks around Ir 002 appear squareness a ratio of the remanence (MR) to the saturation to be composed of only one envelope because the two en- magnetization (MS) and the saturation field (HS) are velopes for Co and Ir layers are close to each other , the strongly related to each other. The samples around tIr 15 peaks around 004 are composed of two distinguishable en- Å, corresponding to the second AF peak, show typical, ideal velopes. The center of one envelope coincides with the peak antiferromagnetically coupled MH curves, with little rema- of buffer Ir 004 and the other one is close to fcc-Co 004 nence, high saturation field, and constant susceptibility. On but at an obviously larger angle position. If we suppose that the other hand, MH curves of the samples around tIr 11 Å 12 960 H. YANAGIHARA, EIJI KITA, AND M. B. SALAMON PRB 60 FIG. 4. MR curves of Co/Ir 001 superlattices with high satura- tion field and little remanence. or tIr 23 Å show large remanence and low saturation field. They are typical of ferromagnetically coupled or noncoupled hysteresis curves. The MR curves for antiferromagnetically coupled super- lattices are shown in Fig. 4, measured at room temperature with the applied field and current parallel to the 100 direc- tion. Samples designed with Ir layer thickness thinner than 9 Å were measured up to 90 kOe and the others were measured FIG. 5. MH and MR curves measured along 100 and 110 in up to 15 kOe. All samples were saturated at fields lower than plane for Co 7 Å /Ir 31 Å 50(001) superlattices. The upper figure 90 kOe except those films with the thinnest Ir layer thick- shows MH curves obtained with applied field parallel to 100 ness; i.e., tIr 5 Å. Rounded shapes at low field and little solid line and 110 broken line . The bottom figure shows MR hysteresis in the MR curves indicate ideal antiparallel align- curves measured with applied field parallel to 100 solid line and ment of the magnetic moments of adjacent Co layers at low 110 broken line . field. The MR ratios at the first and the second AF peaks are about 1.1%. On the other hand, MR ratios of ferromagneti- same field of H 500 Oe. The magnetization curve for cally coupled samples are very small, for instance, typically H 110 saturates at around 1400 Oe, while the magnetiza- less than 0.1%. This MR value is larger than previous tion for H 100 does not saturate and saturate below 3 kOe. reports10­12 on Co/Ir multilayers prepared by sputtering on The difference in MH curves of tIr 31 Å along 100 and thinner buffer layers. The saturation field HS estimated from 110 directions implies that this superlattice shows an ideal MR curves agrees well with that estimated from MH curves magnetization process with competition between cubic an- at the second and the third AF peaks, so it is reasonable to isotropy and antiferromagnetic coupling.16 The inflection believe that we determined HS from the MR curve of the points at H 500 Oe indicate that two different magnetiza- thinner Ir samples with high HS . tion jumps occur for the two MH loops, respectively; that is, The magnetization process is dependent on the direction a spin-flop type jump occurs in the 110 MH loop and a of the applied field, the magnetocrystalline anisotropy, and nonsymmetric-type jump occurs in the 100 MH loop. The coupling strength, including its sign, between adjacent ferro- easy axis in fcc-Co is known to be parallel to the 111 magnetic layers. In the cubic case, samples have fourfold direction and the 110 axis is harder than the 111 , but symmetry in the film plane. The easy axis is expected to be easier than 100 .17 Taking demagnetization fields into ac- parallel to either 100 or 110 directions. The magnetiza- count, it is reasonable that the easy axis is parallel to the tion process along different directions for superlattices at the 110 in plane. third AF peak is shown in Fig. 5. The measurements were Other antiferromagnetically coupled samples, designed carried out at room temperature and MR and MH curves of with thinner Ir layer thickness than above, around tIr 5 Å or tIr 31 Å sample were obtained along 110 and 100 direc- tIr 15 Å, displayed inflection points in the MH curves remi- tions. In both MR and MH curves, a clear difference is niscent of spin-flop type or nonsymmetric-type magnetiza- found. Slopes of the magnetization curve for H 110 and tion jump, but we could not observe a clear distinction be- H 100 look similar at low field. They jump at almost the tween the two directions in the film plane. If a large PRB 60 APERIODICAL OSCILLATION OF INTERLAYER . . . 12 961 tion periods are generally 10­20 Å for long periods and around 4 Å for short periods. From theoretical consider- ations, the dependence of the coupling energy on spacer thickness can be expressed in the following general form: 1 Jn A n2 m sin 2 n/ m m , m where n is the spacer thickness in monoatomic layer units and m denotes spanning vectors that contribute to interlayer coupling. The wavelength m corresponds to one of the os- cillation periods and m is a characteristic phase of each oscillation period. As we mentioned in Sec. I, multiperiodic- ity has previously been found only in sandwich systems with noble metal spaces. Short period oscillations have not usu- ally been found in multilayer systems because of the imper- fections of layer structure, e.g., accumulative interface roughness and/or discrete thickness fluctuations.4 We have observed three AF coupling peaks at around tIr 5, 15, and 33 Å in 5­40 Å region. Compared with previous reported multilayer systems, present Co/Ir 001 system behaves pecu- liarly in terms of its oscillation periods; i.e., the AF peaks do not appear periodically. According to the calculation by Stiles,6 the longest period is about 10 monolayer FIG. 6. The Ir layer thickness dependence of the AF coupling oscillation. a The t ( 19 Å). Although this value corresponds well to the spac- Ir dependence on squareness of Co/Ir 001 su- perlattices. b The t ing between the second and the third AF peaks found in our Ir dependence on saturation field of Co/Ir 001 superlattices. result, it is hard to elucidate the origin of the first AF peaks without considering multiperiodicity because no AF peak, difference exists between the AF coupling and the magneto- associated with the first AF peak, occurs around 20­28 Å. crystalline anisotropy energies, the MH curves are domi- Superposition of several periodicities with periods close to nated by the stronger term. Evidently the magnetocrystalline each other could also explain this peculiarity. In any case, it anisotropy is so much smaller than the strong interlayer cou- is necessary to explain this aperiodic oscillation by introduc- pling that no clear difference is observed between the mag- ing multiperiodicity to multilayer system. The band structure netization processes for different directions. of Ir, which belongs to the 5d transition metal group is not Figure 6 shows the Ir thickness dependence of the satura- so simple, even in the bulk state, so that many spanning tion field and normalized remanence at room temperature of vectors could exist along the 001 direction.6,19 Although Co/Ir 001 superlattices. The saturation fields were deter- this aperiodic coupling oscillation is likely to be associated mined by the MH and MR measurements along the hard with multiperiodicity originating in the complicated Ir band axis, H 100 . As mentioned above, in the thinner Ir samples structure, we cannot consider its band structure to be that of the saturation fields obtained along 100 and 110 look very the bulk because of the large misfit. We believe these pecu- similar to each other because the saturation field is almost liarities are not features of only this system but would be completely dominated by strong AF coupling. An unusual found in other high-quality superlattices with 4d or/and 5d striking inharmoniously is found in both normalized rema- transition metal spacers. We would like to encourage theo- nence squareness plot and saturation field (H retical investigation. S) plot. The intervals between AF peaks in the Ir thickness dependence of HS or between valleys in the squareness plot of the Ir thick- ness dependence reveal an irregular period of antiferromag- IV. CONCLUSION netic coupling oscillations. The samples with Ir layer thick- We succeeded in the growth of high-quality Co/Ir 001 nesses near 5 Å­ 7 Å possess a very strong epitaxial superlattices via MBE. Co and Ir layers both pos- antiferromagnetic coupling constant J 2 erg/cm2 at tIr 6 sess fcc structure. The magnetic easy axis lies in plane for all Å, while the system must become ferromagnetic in the limit samples and the easy axis is parallel to the 110 direction. tIr 0. Here, we suppose that the coupling constant can be MR ratios are small, even for AF coupled samples, while the expressed as J HSMCotCo/4 and that the anisotropy energy coupling constants were particularly strong compared with is negligible. The sample with thinnest Ir layer thickness other typical systems. The most interesting result is the ape- (tIr 5 Å) possesses a saturation field in excess of 90 kOe, riodic coupling oscillation, which may be due to a superpo- corresponding to J 2.2 erg/cm2. This estimated coupling sition of several periods corresponding to spanning vectors strength is exceeded only by the Co/Ru system.18 of the complicated Ir Fermi surface. In order to make clear In many magnetic multilayer and sandwich systems that these peculiarities, i.e., strong AF coupling and aperiodic show interlayer exchange coupling, the coupling oscillation coupling oscillation, further studies on sandwich samples results in periodic AF peaks or MR maxima, and its oscilla- with thicker Ir layers are in progress. 12 962 H. YANAGIHARA, EIJI KITA, AND M. B. SALAMON PRB 60 ACKNOWLEDGMENTS sions. This study was partially supported by Grant-in-Aid for We would like to thank Dr. T. Taniyama and Dr. I. Na- Scientific Research on Priority Areas No. 10130204 , and katani at NRIM for MR measurement by PPMS and to K. by a Distinguished Visiting Professorship M.S. , from the Ono at University of Tsukuba for sample growth. One of the Japan Ministry of Education, Science, Sports, and Culture. authors H.Y. would like to acknowledge Dr. T. Katayana, Partial support was received from the U.S. Department of Dr. Y. Suzuki, and Dr. S. Yuasa at ETL for helpful discus- Energy under Grant No. DEFG02-91ER45439. *Present address: Dept. of Physics, University of Illinois at 10 Y. Luo, M. Moske, and K. Samwer, Europhys. Lett. 42, 565 Urbana-Champaign, Urbana, IL 61801. 1998 . 1 J. Kwo, E. M. Gyorgy, D. B. McWhan, M. Hong, F. J. DiSalvo, 11 A. Dinia, M. Stoeffel, K. Rahmouni, D. Stoeffler, and H. A. M. C. Vettier, and J. E. Bower, Phys. Rev. Lett. 55, 1402 1985 ; C. van den Berg, Europhys. Lett. 42, 331 1998 . F. Majkrzak, J. W. Cable, J. Kwo, M. Hong, D. B. McWhan, Y. 12 H. Yanagihara, K. Pettit, M. B. Salamon, Eiji Kita, and S. S. P. Yafet, J. V. Waszczak, and C. Vettier, ibid. 56, 2700 1986 ; P. Parkin, J. Appl. Phys. 81, 5197 1997 . Gru¨nberg, R. Schreiber, Y. Pang, M. B. Brodsky, and H. Sow- 13 R. F. C. Farrow, G. R. Harp, R. F. Marks, T. A. Rabedeau, M. F. ers, ibid. 57, 2442 1986 . Toney, D. Weller, and S. S. P. Parkin, J. Cryst. Growth 133, 47 2 For a review, see in Ultrathin Magnetic Structures II, edited by J. 1993 . A. C. Bland and B. Heinrich Springer, Berlin, 1994 . 14 3 G. A. Prinz, Phys. Rev. Lett. 54, 1051 1985 . S. S. P. Parkin, Phys. Rev. Lett. 67, 3598 1991 . 15 4 Y. Fujii, in Metallic Superlattices, Artificially Structured Materi- P. Bruno and C. Chappert, Phys. Rev. Lett. 67, 1602 1991 ; 67, als, edited by T. Shinjo and T. Takada Elsevier, New York, 2592 E 1991 . 5 P. Bruno, Phys. Rev. B 52, 411 1995 . 1987 , p. 33ff. 16 6 M. D. Stiles, Phys. Rev. B 48, 7238 1993 . B. Dieny, J. P. Gavigan, and J. P. Rebouillat, J. Phys.: Condens. 7 A. Fert and P. Bruno, in Ultrathin Magnetic Structures II Ref. 2 , Matter 2, 159 1990 . 17 Sec. 2, p. 82. T. Suzuki, D. Weller, C.-A. Chang, R. Savoy, T. Huang, B. A. 8 J. Unguris, R. J. Celotta, and D. T. Pierce, Phys. Rev. Lett. 79, Gurney, and V. Speriosu, Appl. Phys. Lett. 64, 2736 1994 . 18 2734 1997 . S. S. P. Parkin, N. More, and K. P. Roche, Phys. Rev. Lett. 64, 9 F. J. A. den Broeder, W. Hoving, and P. J. H. Bloemen, J. Magn. 2304 1990 . Magn. Mater. 93, 562 1991 . 19 O. Krogh Anderson, Phys. Rev. B 2, 883 1970 .