VOLUME 81, NUMBER 10 P H Y S I C A L R E V I E W L E T T E R S 7 SEPTEMBER 1998 90± Magnetization Switching in Thin Fe Films Grown on Stepped Cr(001) Ernesto J. Escorcia-Aparicio, Hyuk J. Choi, W. L. Ling, R. K. Kawakami, and Z. Q. Qiu Department of Physics, University of California at Berkeley, Berkeley, California 94720 (Received 18 March 1998) The ferromagnetic/antiferromagnetic interfacial interaction was investigated in thin Fe films grown on stepped Cr(001) with the steps parallel to the [100] direction. Above the Néel temperature of the Cr, the atomic steps induce a uniaxial magnetic anisotropy with the easy axis parallel to the step edges. Below the Néel temperature, the Fe-Cr interfacial interaction favors the Fe magnetization perpendicular to the step edges. The competition between the Fe-Cr interaction and the step-induced magnetic anisotropy re- sults in an in-plane 90± magnetization switching from perpendicular to the step edges at low step-density to parallel to the step edges at high step density. [S0031-9007(98)07012-4] PACS numbers: 75.70.Ak, 75.30.Kz, 75.30.Pd, 75.50.Bb Investigations on the magnetic exchange bias have gen- vide information on the origin of the 90± coupling in the erated great interest in the study of the magnetic interaction compensated F AF systems. In this Letter, we report on between ferromagnetic (F) and antiferromagnetic (AF) thin the study of Fe films grown on a stepped Cr(001) surface. films. There are two types of F AF interfaces: uncom- We found that the Fe-Cr interfacial frustration favors the pensated and compensated, corresponding to the nonzero Fe magnetization perpendicular to the step edges. In com- and zero net magnetic moments at the AF surface, respec- peting with the step-induced magnetic anisotropy, which tively. After the original work of Meiklejohn and Bean favors the magnetization parallel to the step edges, the Fe [1], research was focused for many years on the uncom- magnetization undergoes an in-plane 90± switching from pensated type which was believed to be responsible for perpendicular to the step edges at low step density to par- the exchange bias. However, the recent discovery of the allel to the step edges at high step density. exchange bias in the Fe FeF2 110 system [2], where the A Cr(001) single crystal disk of 2 mm thickness FeF2 110 surface is a compensated AF surface, prompted and 10 mm diameter was mechanically polished with a reevaluation of the physical picture of the F AF inter- 0.25 mm diamond paste finish. Half of the crystal was action in compensated systems. The difficulty in under- kept in the (001) orientation while the other half was standing the F AF compensated systems comes from the polished into a curved shape with step edges parallel to fact that the competition between the intralayer magnetic the [100] crystallographic direction. The curved shape interaction and the F AF interfacial interaction leads to provided a continuous range of the vicinal angle a from magnetic frustration, where not all the nearest-neighbor 0± to 10±. The substrate was cleaned with cycles of Ar ion spins can be in their local minimum energy configurations. sputtering and annealing in an ultrahigh vacuum chamber Therefore, the first important issue in compensated sys- with a base pressure of 5 3 10211 torr. Details on the tems was the following: What is the ground state spin Cr substrate preparation and characterization are presented configuration as a result of the magnetic frustration? elsewhere [8]. A substrate temperature of 480 K was In a recent theoretical simulation, Koon [3] showed that used during the Fe film growth to achieve a smooth the magnetic frustration in a compensated system could film surface without the substrate-overlayer intermixing result in 90± coupling between the F and AF magnetic [6]. Hysteresis loops were obtained by in situ surface moments at the interface although the F AF interfacial magneto-optic Kerr effect (SMOKE) measurements in interaction itself has a collinear form. This prediction is the longitudinal configuration with a He-Ne laser as the supported by several recent experiments [4]. However, light source. For all films studied, no polar loops were the physical origin of this 90± coupling, such as its rela- observed so that the Fe magnetization always remains in tion to the frustration, remains obscure. One experimental the film plane. For measurements on stepped surfaces, the approach to address this issue is to control the magnetic reflection angle of the SMOKE laser beam was used to frustration using atomic steps on an uncompensated AF determine the local vicinal angle. A slit was used on the surface. Evidence of this step-induced magnetic frustra- path of the reflection beam to improve the vicinal angular tion was found in the Fe Cr 001 system where the uncom- resolution to better than 0.25±. pensated Cr(001) surface [5] can be partially compensated We first investigated the Fe films grown on the flat by the presence of random steps [6] in the transverse spin half a 0 of the Cr substrate. Wedged samples were density wave (SDW) regime 120 , T , 311 K . The used to provide a thickness range of 5 50 Å. At T magnetic behavior resulting from this kind of frustration 480 K, which is well above the Néel temperature (TN) of is believed to depend critically on the terrace length of the Cr (311 K), the hysteresis loops of the Fe films exhibit atomic steps [7]. Therefore, a study of the magnetic phases a square shape with full remanence for all thicknesses. with different degrees of step-induced frustration will pro- Below the TN, the remanence of the loops remains 100% 2144 0031-9007 98 81(10) 2144(4)$15.00 © 1998 The American Physical Society VOLUME 81, NUMBER 10 P H Y S I C A L R E V I E W L E T T E R S 7 SEPTEMBER 1998 for thicker films .35 Å , but is reduced for thinner films tion perpendicular to the step edges. Then the interesting (Fig. 1). This remanence reduction has been attributed to question is as follows: Why does this 90± switching not oc- the competition between the Fe-Fe and Fe-Cr interactions cur at high vicinal angles? A simple explanation is that the in the presence of random atomic steps on the Cr surface Fe-Cr interfacial interaction and the step-induced magnetic [6]. For thicker Fe films, the Fe-Fe interaction dominates anisotropy scale differently with the step density: The the Fe-Cr interaction so that magnetic domains will be Fe-Cr interaction scales as the terrace length L , and formed at the Fe-Cr interface to give a full remanence. For the step-induced anisotropy scales as the step density thinner Fe films, the Fe-Cr interaction dominates the Fe-Fe 1 L . Therefore, the Fe-Cr interaction should domi- interaction so that magnetic domains will be formed inside nate the step-induced anisotropy at low step density to the Fe film to result in a low remanence [6,7]. Since we are align the Fe magnetization perpendicular to the step edges, interested in the effect of the Fe Cr interfacial frustration, and the step-induced anisotropy should dominate the Fe-Cr we will focus our attention on the thicker Fe film regime interaction at high step density to align the Fe magnetiza- d . 35 Å . For this purpose, we studied a 40 Å Fe film tion parallel to the step edges. The role of this compe- on the curved side of the Cr(001). tition is further evidenced in the temperature dependence It has been shown that atomic steps can induce an in- of hysteresis loops for fixed a. Figure 5 displays several plane uniaxial magnetic anisotropy [9­12]. To separate loops for a 2± and 5± at different temperatures for mag- the effect of the step-induced anisotropy from the Fe-Cr netic fields parallel to the step edges. At a 5±, where interfacial frustration, we first measured the magnetic the step-induced anisotropy is strong enough to domi- hysteresis loops on the curved substrate at T 480 K, nate the Fe-Cr interaction in the whole temperature range, well above the TN of Cr. The square hysteresis loops the full-remanence easy axis character remains down to for field parallel to the step edges and the split loops for 140 K. At a 2±, however, the easy axis character seen field perpendicular to the step edges (Fig. 2) clearly show at high temperature evolves into a hard axis character as the the existence of the step-induced uniaxial anisotropy with Fe-Cr interaction strength increases by lowering the tem- the easy magnetization axis parallel to the step edges. The perature. The above results clearly demonstrate that this splitting field HS , which is proportional to the strength 90± switching is due to the Fe-Cr interfacial interaction. of the uniaxial anisotropy K , increases with a (Fig. 3), To better understand how the atomic steps result in in agreement with previous work [9,12]. A detailed study the observed 90± switching, we consider a model which P of the a dependence of the step-induced anisotropy will consists of the intralayer Fe-Fe interaction, 2J 0 ij SFe,i ? be published in a subsequent work. P S S The sample temperature was then cooled to 140 K to Fe,j, and the interfacial Fe-Cr interaction, J1 ij Cr,i ? turn on the Fe-Cr interaction. Figure 4 shows hysteresis SFe,j in a stepped Fe Cr 001 system (Fig. 6). Here J0 loops at different vicinal angles for magnetic field paral- and J1 are the strengths of the Fe-Fe and Fe-Cr interactions, lel and perpendicular to the step edges. At high vicinal respectively. The alternating Cr moments in adjacent steps angles a $ 4± , the loops exhibit the same character as produce a periodic Fe-Cr interaction. It has been shown the high temperature loops, indicating that the Fe magne- that the Cr moments are in the film plane in the Fe Cr 001 tization is still parallel to the step edges. At low vicinal angles a # 3± , however, the loops clearly show a 90± switching of the magnetization from a parallel to a perpen- dicular direction to the step edges. This result implies that the Fe Cr interfacial frustration favors the Fe magnetiza- FIG. 2. SMOKE loops measured at 480 K for a 40 Å Fe film grown on a curved Cr(001) at different vicinal angles a . The FIG. 1. SMOKE measurements of remanence as a function of splitting of the loops for field perpendicular to the step edges is Fe film thickness for Fe wedges grown on flat Cr(001). The due to the step-induced uniaxial magnetic anisotropy with the remanence has been normalized to the film thickness. easy axis parallel to the step edges. 2145 VOLUME 81, NUMBER 10 P H Y S I C A L R E V I E W L E T T E R S 7 SEPTEMBER 1998 FIG. 3. Splitting field HS in Fig. 2 as a function of vicinal angle a. The solid line is a guide to the eyes. system [6,13]. For in-plane spin, the spin direction can FIG. 4. SMOKE loops measured at 140 K for the 40 Å film in Fig. 2. The magnetization easy axis for a # 3± switches to only be parallel or perpendicular to the step edges due the direction perpendicular to the step edges. to symmetry considerations. Without knowing the sign of the step-induced uniaxial magnetic anisotropy for the pete with the step-induced magnetic anisotropy energy, Cr, we have assumed that the Cr moments are parallel K L cos2 ¯u, which favors the Fe magnetization parallel to the step edges, as the Fe was at high temperature. to the step edges. Thus, a 90± switching of the Fe magne- Verification of this assumption requires future experiments tization will occur at using other techniques. Taking the Fourier component of K 2J2 this interaction, J 1 L coth pd L 1k sin kx , with k np L, the Fourier . (4) L J component of the Fe film energy per unit length is 0p3a2 K can be estimated from the splitting field in 1 Z 2L dx E Fig. 3, K L d a2 mH k 2 J S d a2 mH0Sa, where 2L 1k sin kx sin u x, 0 0 a m 2.2mB 1.26 3 1025 meV G is the magnetic Z d æ moment of the Fe and H0 1 J S 768 Oe rad is the slope of 0a dz xu 2 1 zu 2 . Fig. 3. Then Eq. (4) gives the following condition for the 0 (1) 90± switching: µ Here d is the Fe film thickness, a is the lattice constant, 2J2 pda a2 1 a coth C . (5) and u x, z is the angular variation of the Fe moments C J0p3mH0Sd a (we assume that the Fe moments vary only in the film plane and are independent of the y coordinate). One can immediately realize that Eq. (1) is similar to the form used by Slonczewski in explaining the biquadratic coupling [14] as long as the periodic interlayer coupling in Slonczewski's model is replaced by our periodic Fe-Cr interfacial interaction. Then, Eq. (1) is expected to result in an Fe magnetization perpendicular to the step edges plus a small fluctuation in u. Indeed, using a step function for the interfacial coupling, J1 x J1 sgn sin px L , the total energy per unit length can be minimized to give J u 1k sin kx cos ¯ u cosh k d 2 z k ¯ u 2 , (2) 2J0ak sinh kd 2J2 L X coth 2m 2 1 pd L E 2 1 cos2 ¯u J0 a2 m 1 p3 2m 2 1 3 2J2 FIG. 5. Temperature dependence of SMOKE loops for a 2 1L coth pd L cos2 ¯u . (3) 2± and a 5± with magnetic field parallel to the step edges J0p3a2 for a 40 Å film. While the magnetization easy axis remains parallel to the step edges for all temperatures for a 5±, the Equation (3) favors the Fe magnetization perpendicu- magnetization easy axis for a 2± switches by 90± as the lar to the step edges ¯u 0 . This energy has to com- temperature is lowered. 2146 VOLUME 81, NUMBER 10 P H Y S I C A L R E V I E W L E T T E R S 7 SEPTEMBER 1998 when the F AF interaction is much stronger than the AF intralayer interaction. In such a case, the spins of the F and AF at the interface will be locked together, and a parallel domain wall inside the AF will be formed in response to the rotation of the F magnetization [17]. In fact, it is thought that it is the unwinding of this domain wall that gives ex- change biasing in the compensated F AF system [3]. In the Fe Cr 001 case, this situation does not occur because the Fe-Cr interaction is much weaker than the Cr-Cr in- teraction. That is probably why exchange biasing was not clearly observed in this system [6]. Another possibility for this 90± switching is that the step-induced anisotropy for the lower step density changes its sign upon cooling. How- ever, this is unlikely because the step-induced anisotropy in other systems does not exhibit such a temperature FIG. 6. Schematic drawing of Fe films on stepped Cr(001). dependence. Arrows indicate the spin directions of Fe and Cr. In summary, Fe on stepped Cr(001) was investigated in the vicinal angle range of 0±­10±. Below the Néel tem- Taking the value of J0 11.9 meV [15], a 1.435 Å, perature of the Cr, we found that the Fe-Cr interfacial in- d 40 Å, and J1 0.31 meV [16], Eq. (5) yields the teraction favors the Fe magnetization perpendicular to the critical angle aC 2.5±, which roughly agrees with our step edges. Its competition with the step-induced mag- experiment. netic anisotropy causes the Fe magnetization to undergo a It is constructive to consider the ultrathin limit d ø L , 90± switching from perpendicular to the step edges at low where there is no twisting of the Fe spins along the normal step density to parallel to the step edges at high step den- direction of the film, to understand the physical meaning sity. This phenomenon can be explained based on a model of the 90± coupling. The twisting of Fe moments across similar to the one used by Slonczewski for biquadratic an atomic step is equivalent to producing a u domain coupling. wall over a length L (Fig. 6). This will cost an energy This work was funded in part by the Department E1 J0u2d L. On the other hand, the twisting will lower of Energy under Contract No. DE-AC03-76SF00098, the Fe-Cr interaction energy by E2 J1uL a. Thus, by the National Science Foundation under Contract the total energy change per unit length by twisting the Fe No. DMR9805222, and by the Hellman Family Faculty moments is E E1 1 E2 L J0u2d L2 2 J1u a. Fund at Berkeley. Minimizing this energy yields u J1L2 2J0ad and E 2J21L2 4J0a2d , corresponding exactly to the ultra- thin limit of Eqs. (2) and (3). Therefore, the 90± coupling [1] W. P. Meiklejohn and C. P. Bean, Phys. Rev. 102, 1413 between the Fe and Cr originates from the angular twisting (1956). of the Fe moments. This phenomenon is also intrinsically [2] J. Nogues et al., Appl. Phys. Lett. 68, 3186 (1996). related to the spin-flop transition in an antiferromagnet: [3] N. C. Koon, Phys. Rev. Lett. 78, 4865 (1997). By applying a strong enough magnetic field along the easy [4] Timothy J. Moran and Ivan K. Schuller, J. Appl. Phys. axis, the antiferromagnet moments switch by 90± to be 79, 5109 (1996); Y. Ijiri et al., Phys. Rev. Lett. 80, perpendicular to the magnetic field plus a small twisting 608 (1998); T. J. Moran et al., Appl. Phys. Lett. 72, 617 to form a canted configuration. From the point of view of (1998). P mean field theory, the Fe-Cr interaction J [5] E. Fawcett, Rev. Mod. Phys. 60, 209 (1988). 1 ij SCr,i ? SFe,j in the Fe Cr system is equivalent to applying a "magnetic [6] A. Berger and H. Hopster, Phys. Rev. Lett. 73, 193 (1994). [7] A. Berger and Eric E. Fullerton, J. Magn. Magn. Mater. field" J1 SFe to the Cr moments at the interface, except 165, 471 (1997). that in this case the Cr moments are fixed and the magnetic [8] Ernesto J. Escorcia-Aparicio et al., IEEE Trans. Mag. 34, field J1 SFe can rotate in space. Then, the spin-flop 1219 (1998). equivalent transition in the Fe Cr system is to rotate the [9] R. K. Kawakami, Ernesto J. Escorcia-Aparicio, and Z. Q. magnetic field J Qiu, Phys. Rev. Lett. 77, 2570 (1996). 1 SFe to the perpendicular direction of the Cr moments. Therefore, it is not surprising to obtain [10] W. Weber et al., Phys. Rev. Lett. 76, 1940 (1996). the 90± coupling between Fe and Cr. [11] J. Chen and J. Erskine, Phys. Rev. Lett. 68, 1212 (1992). In the above discussions, we have ignored any spin di- [12] Hyuk J. Choi et al., Phys. Rev. B 57, R12 713 (1998). rection change in the Cr. The Cr spins should also twist pe- [13] A. Schreyer et al., Phys. Rev. Lett. 79, 4914 (1997). [14] J. C. Slonczewski, Phys. Rev. Lett. 67, 3172 (1991). riodically by the same mechanism as the Fe spins. Koon's [15] C. Kittel, Introduction to Solid State Physics (Wiley, New calculation clearly shows this configuration [3]. Another York, 1996), p. 446. point we have not addressed is the domain wall formation [16] S. T. Purcell et al., Phys. Rev. Lett. 67, 903 (1991). inside the Cr. This situation becomes important especially [17] C. Mauri et al., J. Appl. Phys. 62, 3047 (1987). 2147