R. F. C. Farrow, R. F. Marks, M. J. Carey, and B. A. Gurney. IBM Research Division, Almaden Research Center, 650 Harry Road, San Jose, California 95120- ...
RAPID COMMUNICATIONS
PHYSICAL REVIEW B
VOLUME 58, NUMBER 18
1 NOVEMBER 1998-II
Ultrafast optical modulation of an exchange biased ferromagnetic/antiferromagnetic bilayer Ganping Ju and A. V. Nurmikko Department of Physics and Division of Engineering, Brown University, Providence, Rhode Island 02912
R. F. C. Farrow, R. F. Marks, M. J. Carey, and B. A. Gurney IBM Research Division, Almaden Research Center, 650 Harry Road, San Jose, California 95120-6099 ~Received 17 June 1998; revised manuscript received 25 August 1998! Time-resolved magneto-optical Kerr techniques are employed to study optically induced ultrafast modulation in an antiferromagnetic ~AF! exchange biased ferromagnetic ~FM! thin epitaxial bilayer ~NiFe/NiO!. The interface region between the antiferromagnetic NiO and coupled ferromagnetic NiFe layers is photoexcited with femtosecond laser pulses, following which the transient magneto-optical response of the NiFe layer is measured. When compared with optically induced effects on ‘‘bare’’ epitaxial NiFe thin films, we observe a large enhancement to the transient Kerr effect in the exchange biased bilayers, interpreted as a modulation of the AF/FM exchange coupling on a picosecond timescale. @S0163-1829~98!51942-X#
Active research continues to be directed toward improving our understanding of the exchange processes that lead to a unidirectional magnetic bias in coupled antiferromagnetic/ ferromagnetic bilayers, both for fundamental reasons and due to the importance of this system as a sensor component in magnetoresistive ~MR! devices such as spin valves.1–3 Considerable insight has been gained from experimental studies and theoretical micromagnetic modeling, indicating that magnetocrystalline anisotropy, exchange stiffness, and the crystalline texture of the AF medium are important in setting up the unidirectional effective exchange field, with the atomic scale microstructure at the AF/FM interface playing a particularly important role. In this report, we raise the question about the dynamics of the AF/FM exchange bias system at a microscopic level, specifically in terms of the possibility of ‘‘unpinning’’ the coupling under conditions of ultrafast selective electron/spin heating by a subpicosecond laser pulse. Recently it has been shown that femtosecond laser pulses can be used to study the spin dynamics and magnetization kinetics associated with hot, nonequilibrium electrons and spins in ferromagnetic thin films on an ultrashort timescale.4,5 Briefly, energetic electrons (h n .1 eV! are promoted via interband transitions to break the ground-state electron-electron exchange. Subsequent relaxation processes in the nonequilibrium spin system are probed with subpicosecond resolution by using the magneto-optical Kerr effect. In Ref. 5, for example, specific timescales were assigned to dephasing of photoexcited spins in a coherent regime, and the evolution of a thermalized hot spin gas, respectively. In broad terms, these experiments show that a ‘‘window of opportunity’’ exists, roughly on the order of one picosecond, for optically manipulating the spin degrees of freedom in a FM thin film, prior to energy transfer to the lattice from the photoexcited electron/spin system. Here we exploit this opportunity while applying ultrashort laser-pulse excitation to an antiferromagnetic/ferromagnetic exchange-coupled bilayer system, NiO/NiFe. We argue that the experimental results suggest how ultrafast modulation, i.e., partial ‘‘unpinning,’’ of the exchange bias process can occur in a time as 0163-1829/98/58~18!/11857~4!/$15.00
PRB 58
short as 10212 sec, following which the system recovers to its initial state. The recovery involves a combination of spinelectron-lattice interactions and is eventually governed by lattice cooling on a timescale as long as 100 psec. The exchange coupling across the interface of a ferromagnetic and antiferromagnetic film can lead to the presence of an effective unidirectional anisotropy bias field (H EX), observed by a shift of the hysteresis loop, while increasing the easy-axis coercivity (H C ) for the FM layer. Exchangecoupled systems incorporating FeMn, NiMn, PdMn, NiO, CoO, MnPt, and NiCoO antiferromagnetic layers are among the candidates for the AF exchange bias layer for spin-valve MR sensors applications. For the photoexcitation experiments described below, the transparency of NiO makes it possible to photoexcite the interface between the ferromagnet and antiferromagnet efficiently; furthermore, the NiO/ NiFe system possesses a fairly low exchange-bias blocking temperature (T b '200 °C), a feature which is convenient for the types of nonequilibrium experiments in question where we seek to break the exchange bias up by nonthermal means. In terms of an effective temperature concept, the electron temperature T e (t) or spin temperature T s (t) is promoted above the blocking temperature T b , while the lattice temperature T l is kept lower than T b . The samples used in this study were polycrystalline NiFe/ NiO bilayers grown by dc magnetron sputtering on glass substrates. The NiO was grown by reactive sputtering of Ni in Ar1O2 background. The exchange bias was set by heating the samples to ;220 °C in a N2 ambient and cooling in a magnetic field of ;5 kOe. The thickness of the permalloy layers varied in the range of 100–300 Å, at a constant nominal composition of 81% Ni/19% Fe, while the NiO layers were approximately 400 Å thick. For comparison purposes, we also used permalloy films of the same composition and thickness range, deposited directly on ZrO2/glass substrates by the UHV evaporation in a magnetic field of 1 kOe. The experiments were conducted in a geometry shown in Fig. 1. The excitation pulses from a mode-locked Ti:sapphire laser ~pulse duration of 120 fsec, photon energy h n 51.4 eV, repetition rate of 76 MHz! were directed normal to the sample R11 857
©1998 The American Physical Society
RAPID COMMUNICATIONS
R11 858
JU, NURMIKKO, FARROW, MARKS, CAREY, AND GURNEY
FIG. 1. Experiment schematic for time-resolved magnetooptical Kerr measurement of ultrafast magnetization modulation in an exchanged biased AF/FM bilayer system.
through the transparent glass substrate and the NiO AF layer (E g .5 eV!, and absorbed within the NiFe FM layer. The pump beam was acousto-optically modulated at 3.9 MHz for phase-sensitive detection. The optically induced changes in the spin/magnetization of the NiFe layer were recorded by measuring the transient changes in the ~longitudinal! Kerr rotation, DQ K (t), using time delayed weak probe pulses ~with less than 0.2 mW average power and approximately 45° incidence angle! in the blue ~l5430 nm! that were focused to 10 mm spot and reflected from the sample front surface ~NiFe layer!. The Kerr instrumentation employed a polarization-sensitive balanced optical bridge, consisting of Wallaston prism and a low-noise differential detector arrangement, followed by a frequency down converter which transformed the signals at 3.9 MHz to the kHz regime ~where lock-in amplifiers have higher sensitivity! for phase sensitive detection.5,8 We focus here on a specific exchange biased sample ~10 nm NiFe/40 nm NiO/glass! whose static magnetization characteristics indicated the presence of the exchange coupling with H EX5105 Oe and H C 540 Oe. The companion reference sample, without an exchange bias AF layer had the structure consisting of 11.8 nm NiFe/ZrO2 / glass with H EX 50 Oe, and H C 54 Oe. @The measured Kerr loops for the two samples are shown in Fig. 3~a! below.# Apart from the magnetic properties, the choice of an approximately 10-nm NiFe film thickness is useful from the optical point of view. From the measured complex index of refraction for Ni0.81Fe0.19, n(l)1ik(l), we estimate the absorption length at l5860 nm to be approximately 14 nm, and the penetration depth of the reflecting Kerr probe at l5430 nm to be about 7 nm. Thus the photoexcitation can be assumed to be reasonably uniform across the FM layer, while the magneto-Kerr probe mainly measures the spin polarization/magnetization of the NiFe layer away from the AF interface ~i.e., within the ‘‘bulk’’ of the film!. The uniformity of excitation simplifies the hot electron problem in that diffusive processes that result from concentration gradients of hot electrons in thicker films are absent ~electrons are, of course, unable to diffuse into the NiO insulator!. Figure 2~a! shows the transient Kerr data, DQ K (t), for the exchange biased NiFe/NiO bilayer and the reference NiFe film, as well as their difference, taken under nominally
PRB 58
FIG. 2. ~a! Comparison of transient Kerr effect DQ k (t) for permalloy thin films with and without the antiferromagnetic exchange bias layer, and their differential. The curves are offset vertically for clarity. ~b! Comparison of the risetime for transient reflectivity (DR(t)/R) with that for the fractional transient Kerr effect @ DQ k (t)/Q k # in the NiFe/NiO bilayer. ~c! Comparison of the transient reflectivity data for NiFe/NiO bilayer ~solid! and NiFe reference film on a longer time scale, under the same excitation level ~curves offset for clarity!.
the same laser excitation levels and corresponding to an average excitation power density of approximately 6 3103 W/cm2 in the 10 mm spot size at the sample surfaces. The graphs are displaced vertically for clarity. Note that the magnitude of the transient Kerr rotation for the exchange bias bilayer is approximately five times larger than that for the reference NiFe film. The data were taken in an external ~in-plane! magnetic field of H A 5200 Oe, sufficient to bias the samples in positive saturation of magnetization. The magnitude of DQ K /Q K ;1022 is equivalent to an induced Kerr rotation of about 300 microdegrees. If we consider the 120 fsec laser pulses as creating an impulse of excitation, the estimated initial nonequilibrium electron density is less than 1020 cm23. Figure 2~b! compares the rise time and early temporal features for the transient reflectivity DR(t)/R ~solid line, recorded in tandem with the transient Kerr effect! and the Kerr rotation DQ K (t) ~dashed line! for the exchange bias sample. The comparable excitation level for the exchange biased and reference sample is further confirmed by the similarity of their transient reflectivity data, DR(t)/R in Fig. 2~c!. The transient reflectivity, which is insensitive to spin polarization and magnetization effects, is dominated by hot electron contributions to the diagonal elements of the dielectric tensor within the first few psec ~note the initial high amplitude peak of about 5 psec in duration!. On a longer timescale, one observes lattice cooling as well as a contribution by photothermal strain ~oscillations on 20 psec time scale that signify the propagation of very high-frequency ultrasound waves in the films!. The .100 GHz ultrasound waves, which are generated by rapid thermal expansion of the heated layers, show also very high-frequency components, observed as ‘‘ripples’’ on the time scale of a few psec in the transient reflectivity data for the exchanged biased film which are most likely due to the acoustic echoes propagating within the bilayer. That these components in DR(t)/R can usually be identified and separated from each other due to their distinct
RAPID COMMUNICATIONS
PRB 58
ULTRAFAST OPTICAL MODULATION OF AN EXCHANGE . . .
time domain signatures is a practice well established in earlier work on nonmagnetic metals6 as well as magnetic thin films.5 For interpretation of the transient Kerr effect for the reference permalloy film we consider the origin of the magnetoKerr effect, while being also guided by our recent time resolved experiments in CoPt3 thin films.5 Briefly, the Kerr rotation Q K ~or Kerr ellipticity e K ) is related to the offdiagonal component of the conductivity tensor s xy 5 s 1xy 1i s 2xy , so that Q K }C2 s 1xy 2C1 s 2xy , where s xy is proportional to magnetization and the net spin polarization of the medium.7 Since the absorption probability at the pump laser wavelength is unequal for the majority and minority spins ~electrons!, we create by the short pulse photoexcitation at t50 an initial distribution of hot ~i.e., nonthermalized! electrons with finite net nonequilibrium spin polarization. The traces in Fig. 2~a! are proportional to this polarization and provide a means to study the relaxation processes that involve the nonequilibrium spins. The energetic spins initially thermalize by spin-dependent electron-electron scattering processes, following which the interaction with the lattice must be taken into account. In general terms, a rate equation approach can be used to separate the different components of the relaxation process that involve the three degrees of freedom ~spin, electron, lattice!, once the assumption of an effective temperature concept is valid.4,5 Information about spin dependent electron scattering, electron-phonon, and spin-lattice interaction can thus be obtained from this type of data. The time window for the traces in Fig. 2~a! is dominated by the cooling of the lattice by heat conduction into the substrate, once T lattice(t)5T electron 5T spin . The key result of this paper is the much enhanced amplitude in the optically induced modulation of the magnetoKerr signal obtained from the exchange biased NiFe/NiO sample. The time profile of this enhancement is particularly evident on a short timescale ~,5 psec!. We interpret this large enhancement as originating from ultrafast modulation of the unidirectional exchange bias field, DH EX(t), that is, partial ‘‘unpinning’’ of the effective field due to the interaction of the nonthermal spins at the NiFe/NiO interface. In this sense, the coupled system offers an extra degree of freedom for modulation the overall magnetic response, when compared with the NiFe film reference sample. Qualitatively, we begin our argument by noting the generally accepted temperature dependence of the exchange field, as approximately verified by theory and experiment under thermal equilibrium condition: H EX(T);(12T/T B ), 1 where T B is the blocking temperature. Under the transient photoexcitation conditions, we may consider replacing the lattice temperature by T eff , the effective spin temperature which can be reasonably defined ~at least for discussion purposes!, once the hot electron gas has thermalized within the NiFe film ~with T eff .T lattice). The physical origin of the transient modulation most likely involves the random micromagnetic, interfacial origin of the exchange bias field. For explanation of the static magnetic response, such micromagnetic considerations were first pointed out by Malozemoff, who, on the basis of energy arguments expressed the exchange field as H EX 5 d S/2M F t F , where M F and t F are the magnetization of the permalloy film and its thickness, respectively.1 The magnetic
R11 859
FIG. 3. ~a! Static Kerr loops for the NiFe/NiO bilayer (H EX 5105 Oe, and H c 540 Oe! and NiFe single epilayer (H EX50 Oe, and H c 54 Oe!; ~b! transient Kerr loop for the NiFe/NiO bilayer and NiFe epilayer at a pump-probe time delay of 1 ps in the small modulation regime discussed in this paper.
energy difference term dS, in turn, is proportional to the exchange across the interface, of the form d S; ( i j J i j Si•Sj where i and j denote the spins in the monolayers across the interface on the FM and AF sides, respectively. Thus, a mechanism exists whereby hot spins modulate the magnitude of H EX in the nonadiabatic case considered here, by a combination of direct modulation of the interfacial exchange energy d S and the modulation of the magnetization in the NiFe layer by spin/electron heating. The modulation of both H EX and M F by the hot electron system must be viewed in the context of the coupled spin system, however, in the spirit of the random field model with details dependent on the microdomain structure of the NiO near the heterointerface. This coupling, we suggest, provides the crucial feedback mechanism which is responsible for the observed ‘‘amplification’’ effect in our experiments for the bilayer samples, but absent in the NiFe single film sample. In the case of small modulation in the magnitude of the unidirectional exchange bias field, we thus have a simple argument as to why the coupled nature of H EX and M F can lead to the unexpectedly large contribution in the measured DQ K (t) for the bilayer system in saturation, since the Kerr rotation is related to the net spin polarization/magnetization of the permalloy film. In Fig. 3~a! we show the static Kerr hysteresis data for the ~unperturbed! NiFe/NiO bilayer sample, to be compared with the corresponding transient Kerr ‘‘loop’’ in Fig. 3~b! for a time delay of t 51 psec. Within experimental resolution, the two loops are identical, with the same offset along the horizontal axis, showing that we cannot measure the impact of induced changes on DH EX directly under the low excitation conditions that apply to these experiments. @We note that under considerably higher optical excitation, very large and direct changes in DH EX have been seen by us in recent experiments; however, these changes also contain a significant average lattice heating contribution, the elimination of which is an experimental challenge presently under way. No such heating is in evidence in Fig. 3~b!.# Subtraction of the transient Kerr data for the exchange bias and simple permalloy films explicitly isolates the time signature for the contribution from the modulation ~unpinning! of the exchange coupling and gives information about the dynamics of the relaxation processes at issue. In the dif-
RAPID COMMUNICATIONS
R11 860
JU, NURMIKKO, FARROW, MARKS, CAREY, AND GURNEY
ferential ~bottom! trace of Fig. 2~a!, one can identify a very fast rise time of approximately 1.0 psec which is emphatically longer than the risetime ~400 fsec! in the modulated reflectance data @ DR(t)/R in Fig. 2~b!#, the latter corresponding to the hot electron thermalization time. This finite rise time is hence related to the turn-on time for the ‘‘unpinning’’ process triggered by the photoexcitation of nonequilibrium hot electrons and spins, and is presumably controlled by the thermalization of nonequilibrium spins by electronspin interaction within the coupled system. The same relaxation processes also contributes to the similar delay of rise time in the transient Kerr effect of the NiFe single layer @Fig. 2~a!# and the CoPt3 system5 and sets an ultimate limit on the turn-on speed of the unpinning process. This result further underscores the fact that the FM layer provides the key ‘‘trigger’’ mechanism for the transient hot spin effects highlighted in this paper; that is, that the primary input energy channel for the fsec photoexcitation of the spin-coupled system occurs via the NiFe of the exchange biased bilayer system. During the relaxation process, the initial decay time of
W. H. Meiklejohn and C. P. Bean, Phys. Rev. 102, 1413 ~1956!; A. P. Malozemoff, Phys. Rev. B 37, 7673 ~1988!; 35, 3679 ~1987!; N. C. Koon, Phys. Rev. Lett. 78, 4865 ~1997!. 2 C. Tsang, R. E. Fontana, T. Lin, D. E. Heim, V. S. Speriosu, B. A. Gurney, and M. L. Williams, IEEE Trans. Magn. 30, 3801 ~1994!; T. Lin, C. Tsang, R. E. Fontana, and J. K. Howard, ibid. 31, 2585 ~1995!; R. P. Michel, A. Chaiken, C. T. Wang, and L. E. Johmson, ibid. 32, 4651 ~1996!. 3 D. H. Han, J. G. Zhu, J. H. Judy, and J. M. Sivertsen, J. Appl. Phys. 81, 340 ~1997!; R. F. C. Farrow, R. F. Marks, S. Gider, A. C. Marley, and S. S. Parkin, ibid. 81, 4986 ~1997!; K. Takano, R. H. Konama, A. E. Berkowitz, W. Cao, and G. Thomas, Phys. 1
PRB 58
the differential signal indicates the rate of recovery of the nonthermal exchange-coupled system by energy relaxation via electron-phonon, spin-lattice interaction, which can be assumed to dominate the fast components ~;10 psec!. Finally, the much slower process ~;100 ps! is simply due to the lattice cooling by heat flow from the bilayer to the substrate, and sets the ultimate limit of the recovery of the unpinning prior to the next excitation laser pulse. In summary, we have applied ultrashort pulse optical excitation to an AF/FM exchange-coupled system to demonstrate that hot photoexcited spins are effective in modulating the magnetic response of the entire system on a picosecond time scale. We have presented simple physical arguments concerning the mechanism of the effect whose details are presently being studied within a wider range of exchange bias multilayers. The research at Brown was supported by a National Science Foundation Grant No. DMR-9701579 and under the auspices of a Joint Study Agreement with IBM Research.
Rev. Lett. 79, 1130 ~1997!. E. Beaurepaire, J.-C. Merle, A. Daunois, and J.-Y. Bigot, Phys. Rev. Lett. 76, 4250 ~1996!. 5 G. Ju, A. Vertikov, A. V. Nurmikko, C. Canady, G. Xiao, R. F. C. Farrow, and A. Cebollada, Phys. Rev. B 57, 700 ~1998!. 6 C.-K. Sun, F. Vallee, L. H. Acioli, E. P. Ippen, and J. G. Fujimoto, Phys. Rev. B 50, 15 337 ~1994!; G. Tas and H. Maris, ibid. 49, 15 046 ~1994!. 7 W. Rein and J. Schoenes, in Ferromagnetic Materials, edited by K. H. J. Buschlow and E. P. Wohlfart ~Elsevier, New York, 1990!. 8 S. Savikhin, Rev. Sci. Instrum. 66, 4470 ~1995!. 4
