RAPID COMMUNICATIONS PHYSICAL REVIEW B VOLUME 61, NUMBER 14 1 APRIL 2000-II Dynamics of surface magnetization on a nanosecond time scale Fausto Sirotti Laboratoire pour l'Utilisation du Rayonnement Electromagnetique, CNRS-CEA-MESR, F-91405 Orsay, France Simone Girlando INFM, Dipartimento di Fisica dell'Universita di Modena, I-4100 Modena, Italy Pilar Prieto Departamento Fisica Aplicada, C-XII, Facultad de Ciencias, Universidad Autonoma de Madrid, E-28049 Cantoblanco, Spain Luca Floreano INFM, TASC Laboratory, Area di Ricerca, Padriciano 99, Trieste, Italy Giancarlo Panaccione APE Project, INFM, TASC Laboratory, Area di Ricerca, Padriciano 99, Trieste, Italy Giorgio Rossi INFM, Dipartimento di Fisica dell'Universita di Modena, I-4100 Modena, Italy and Laboratorium fu¨r Festko¨rperphysik, ETH-Zu¨rich, CH-8093 Zu¨rich, Switzerland Received 5 November 1999 The dynamics of surface magnetization is measured with ns time resolution by spin-polarimetry of the total photoemission yield excited by synchrotron radiation pulses. The surface response is compared to the bulk magnetization dynamics as obtained by induction measurements. The surface and the bulk show distinct magnetization dynamics indicating weak coupling during the reversal process in the ns- s time domain. Ultrathin layers of Fe as well as three-layer Fe/Cu/Fe exchange coupled structures were grown on top of an amorphous soft-ferromagnetic substrate Vitrovac and showed different reversal dynamics. The surface of a ferromagnet has a special magnetic be- behavior of the surface, i.e., integrated over the domain havior with respect to the bulk.1,2 The electronic structure of structure. Atomically clean surfaces of a soft-magnetic rib- the surface layer implies that exchange is anisotropic, i.e., bon, as well as iron monolayers or iron/copper/iron inter- the exchange stiffness is different on a path within the layer faces, were prepared by ion sputtering and e-beam evapora- compared to a path perpendicular to it.3 The surface magne- tion techniques in an ultrahigh vacuum environment, and tization of the 3d ferromagnetic metals is characterized by measured at room temperature. The primary and secondary enhanced spin and orbital magnetic moments with respect to photoejected electrons from the sample surface were col- the bulk.4,5 The critical behavior of the magnetization near lected by an electrostatic accelerator lens and directed to the the Curie temperature (TC) is described by surface critical thin Au target of a 100 KV Mott scattering detector. The spin exponents which differ from the bulk ones. The polarization of the ejected electron beam SP (Iup ferromagnetic/paramagnetic transition itself may occur at a Idown )/(Iup Idown), where Iup(down) are the spin-up or different temperature with respect to the bulk TC .6 Based on -down intensities, was measured and independently regis- these considerations alone one expects the dynamics of mag- tered after each pulse of synchrotron radiation, i.e., at 8.333 netization and of magnetization reversal at a ferromagnetic MHz rate.7,8 The measure of the bulk magnetization reversal surface to be different than in the bulk. dynamics was obtained by means of an induction search coil Our experiments are based on the surface sensitivity of in otherwise identical experimental and timing conditions. the measurement of spin polarization SP of the photoemis- The external magnetic field was applied to the whole sion yield as excited by synchrotron radiation SR in the UV sample by means of a current pulse in a low inductance coil. and soft x-ray energy range, and on the pulsed structure of A steady magnetization state saturation was maintained by the radiation from an undulator source on a positron storage a constant bias current in the coil. At the time t 0 of each ring. 500 ps-long pulses of SR at time intervals of 120 ns, experiment the current in the coil was reversed and set to a were obtained exploiting the ``two bunch mode'' of the Su- fixed value of choice between a few mA up to about 40 perAco storage ring at Orsay. The polychromatic undulator amperes . The reversed applied field stabilized typically radiation or monochromatic radiation of energy h 200 eV within 70 ns, after which it remained constant during the from the undulator source DOMINO at SuperAco was fo- data acquisition. In Fig. 1 we present a scheme of the time cused to a spot of about 3 1 mm and impinged on the structure of the experiment with the SR pulses at 120 ns sample surface. This sets the lateral scale of the experiment intervals real time mode . By applying a delay of 1 ns or which is therefore representative of a macroscopic magnetic multiples to the magnetizing pulse with respect to the SR 0163-1829/2000/61 14 /9221 4 /$15.00 PRB 61 R9221 ©2000 The American Physical Society RAPID COMMUNICATIONS R9222 FAUSTO SIROTTI et al. PRB 61 FIG. 1. Time structure of the experiments: The grid represents the SR pulses 500 ps every 120 ns . The upper curve is the current in the magnetic circuit, triggered on a SR pulse at t0. The lower data are SP measurements for zero delay with respect to t0. Delays of 1 ns or longer have been applied shifting the data acquisition with respect to the applied field. FIG. 3. Maximum slope of the magnetization reversal curves as pulses and by repeating the experiment the overall time reso- a function of the applied magnetic field obtained for the surface lution of about 1 ns is obtained pump and probe mode .7,8 open symbols and the bulk solid symbols . In Fig. 2 we present the surface sensitive and bulk sensi- tive magnetization data as a function of time t after the ap- field.14 By comparing the surface and bulk experimental plication of an external field antiparallel to the previous satu- curves for the same applied field it is found that the surface rated magnetization state (t0). Various field magnitudes were magnetization reversal of a 100 m-thick Vitrovac ribbon applied to the sample: they are given in units of the coercive advances the bulk one: the inequality tDSurf tDBulk becomes field Hc0 as measured in a standard i.e., slow hysteresis progressively larger as the applied field increases and the loop for the Vitrovac substrate. All curves have the same reversal time decreases. The magnetization reversal speed general shape: the demagnetization state zero spin polariza- can also be represented by the value of the time derivative of tion in the surface data is reached at a time tD and the the magnetization reversal curves evaluated at tD as shown reversed saturation magnetization is reached in a time ex- in Fig. 3. The maximum magnetization reversal speed as ceeding 2tD . The analysis of these M(t) curves could be well as the dynamical coercivity of bulk and surface appear attempted to some extents by means of models involving quite different in the high magnetic-field regime. An applied energy barriers9 physically related to domain nucleation and field of 130 times Hc0 is sufficient to switch the magnetiza- domain-wall motion.10,11 This kind of analysis leads to a tion of the surface with tD 110 ns and a maximum speed of large set of fitting parameters whose physical meaning is 37 s 1, but the bulk magnetization reversal process takes vague and to the conclusion that the local domain-wall mo- place with tD 180 ns and maximum speed of 8 s 1. In tion is governed by viscous motion in an external magnetic the time interval tDSurf t tDBulk the net orientation of sur- face and bulk are therefore opposite. This important point must be addressed carefully since the bulk reversal data are affected by the induction of eddy currents as the reversal proceeds and the effect of the demagnetizing field for a step- like reversal of the applied field is hard to evaluate quantita- tively. On the other hand, the comparison of the SP data measured for different surfaces grown on the same Vitrovac substrate, and therefore exchange coupled to it, is indepen- dent on the actual applied field and magnetization dynamics of the substrate. The thermal decrease of the surface magnetization is larger compared to the bulk due to the double probability of finding spin waves at the surface than in the bulk.3 Various experiments have been performed to establish the spin-wave stiffness of surfaces, including permalloy and Fe 100 .15­17 It has been found that indeed the exchange interaction along a path perpendicular to the surface is reduced and that it can be FIG. 2. Magnetization reversal curves measured on a further modified by modifying the chemical composition or 100 m-thick Vitrovac sample for the surface upper panel and the structure of the surface. This means that the surface is the bulk lower panel . The curves are obtained applying a field 8, ``weakly'' exchange coupled to the bulk and that this cou- 12, 20, 28, 50, 84, 112, 209, 331, 477, and 654 times larger than pling can be artificially modified. A further hint to the weak Hc0. coupling of surface and bulk was given by the comparison of RAPID COMMUNICATIONS PRB 61 DYNAMICS OF SURFACE MAGNETIZATION ON A . . . R9223 FIG. 5. Maximum slope of the surface magnetization reversal curves measured on the surface of two exchange coupled system: 20 Å Fe / 4 Å Cu / 20 Å Fe and 20 Å Fe / 10 Å Cu / 20 Å Fe vs the applied field. FIG. 4. Surface magnetization reversal curves measured on the narrow hysteresis of Permalloy or Vitrovac which is there- surface of a three-layer Fe/Cu/Fe system. The copper thickness was fore called the magnetic ``driver'' of the surface iron film. 4 Å upper panel and 10 Å lower panel . The curves are obtained The present data show that, by inserting an intralayer of a applying a field 12, 16, 20, 24, 32, 40, 48, 60, 80, and 160 times nonmagnetic material such as copper, one can control the larger than Hc0. reduction of the magnetic coupling between the top iron layer and the substrate and this is reflected in a very different surface and bulk hysteresis loops.12 We have investigated the relative dynamical behavior. The Fe/Cu/Fe system shows a magnetization dynamics of exchange coupled structures by delay of the onset but a faster reversal transition with respect comparing SP(t) data for different surfaces deposited onto to the more strongly coupled Fe layer and Vitrovac surfaces. atomically clean Vitrovac: nanometer-thick layers of iron The results of an independent experiment addressing the and Fe/Cu/Fe three-layer structures with different layer surface to bulk dynamical coupling are shown in Fig. 7. Here thicknesses. The bottom and top iron layers were 20 Å thick the dynamical response of the magnetization of a Vitrovac and the copper spacers were 4 Å and 10 Å thick. The SP of film to an applied field of 12 Hc0 is studied as a function of secondary electrons measured after the deposition of the cop- the time duration tneg of the previous opposite magnetizing per spacer layer was 50% and 5% of the value measured for field whose amplitude is 32 times Hc0. The time scale of the bottom iron layer, respectively . The magnetization re- the magnetization reversal processes is aligned taking t 0 versal curves are shown in Fig. 4 where the results obtained when the 32 Hc0 / 12 Hc0 field inversion takes place. for the 4 Å and 10 Å Cu thicknesses are identified by the SP The SP data show that the reversal dynamics is extremely values across the spacer. Equilibrium experiments show that dependent on tneg , that is on the previous history of the the Fe surface is exchange coupled parallel to the Vitrovac sample. For a very short tneg 0.48 s the surface had not surface and that it displays a square hysteresis loop with the reached the saturation before the new reversal was induced. same coercive field as the substrate. The upper iron layer is exchange coupled to the substrate through the copper intra- layer, but the coupling vanished for Cu thicknesses larger than 20 Å. The comparison between the maximum magnetization re- versal speed of the two Fe surfaces is shown in Fig. 5. The magnetization reversal process of the less coupled Fe layer is faster then the more strongly coupled one. The experimental curves measured for an applied field 80 times larger than Hc0 are shown in Fig. 6. It appears that the ferromagnetic sur- faces follow different time patterns showing also a variable delay of the onset of reversal. In particular, the less coupled Fe/Cu/Fe surface layer open symbols presents a magneti- zation reversal transition with a delayed onset but a faster transition which gives a reduced tDSurf value. The demonstration and understanding of exchange coupled artificial structures is at the basis of current magnetic multilayer technology. An iron film can be made magneti- FIG. 6. Magnetization reversal curves measured for an applied cally ``soft'' by exchange coupling to a soft-ferromagnet field 80 Hc0 for two Fe/Cu/Fe systems for which the SP after Cu substrate such as Permalloy or Vitrovac. One obtains in that deposition was reduced to 50% and 5% of the SP of the clean iron case the high magnetic moment of pure iron and the very substrate. RAPID COMMUNICATIONS R9224 FAUSTO SIROTTI et al. PRB 61 ration, in a out-of-equilibrium state of the ferromagnet, the next surface reversal is the fastest, with a doubled speed with respect to the case of initial surface-bulk saturation equilib- rium state. This behavior mimics a spring coupling between the surface and bulk magnetization. This experiment inde- pendently confirms that the magnetization dynamics at the surface is faster than in the bulk and removes all uncertain- ties connected to the effective value of the applied field in- side the bulk. It does show that, during reversal, the surface and bulk of a ferromagnet are two subsystems out of equi- librium. The ``weak coupling'' between the surface and the bulk in ferromagnets is put in evidence by the present experiments on the dynamics of the surface magnetization reversal in the 100 1000 ns time scale. The details of the magnetization FIG. 7. Surface magnetization reversal curves obtained for reversal mechanism are not directly retrievable from the ex- negative pulse duration ranging between 0.48 s and 3.72 s. periments. The fact that the magnetization reversal starts The magnetic field values are 32 H promptly at the surface is possibly related to the noncollinear c0 before t0 and 12 Hc0 after t0. The continuous line represents bulk magnetization reversal curve alignment of the surface and bulk magnetic moment due to for the transition to 32 Hc0. the surface anisotropy as suggested in Refs. 12 and 13. As the reversed field is applied, the torque exerted on the mag- But for values of t netic moments can be finite only for surface moments, if they neg 0.72 s the surface was fully satu- rated. The reversal dynamics nevertheless is very different are even slightly misaligned with respect to the bulk mo- for values of pulsewidth up to 3.72 s solid symbols in ments. The bulk magnetization reacts then to the surface re- Fig. 7 . For 1.3 s t versal with a spring-coupling behavior. Structural and do- neg 3.7 s the magnetization rever- sal process has a delayed onset but a higher speed with re- main distribution informations are needed in order to attempt spect to the behavior measured in equilibrium conditions. a useful micromagnetics analysis of this phenomenon, but The corresponding bulk dynamics is shown by the thick solid the present results clearly show how surface magnetization line. These results are quite important since they show that reversal is different with respect to bulk magnetization rever- the bulk cannot reach an equilibrium state, with the present sal. value of applied field, before 3.72 s. This phenomenon The main message of the present experiments is that the could be explained by a domain-structure-shape memory ef- surface magnetization is not in equilibrium with the bulk fect as in Ref. 14. The ``weak'' surface to bulk coupling is magnetization when a sudden change of applied field occurs. clearly observed in this experiment: the surface magnetiza- Modified surfaces show different dynamical response to ex- tion reversal is the slowest when at t 0 both bulk and sur- ternal fields but always appear to switch faster than the bulk face are saturated, but it is much faster when at t 0 the bulk of a 100 m-thick amorphous soft ferromagnet. had not yet reached saturation in the previous magnetization We thank Ch. Back for discussion and H. C. Siegmann direction. After the shortest pulse that produces surface satu- for stimulating suggestions and continuous support. 1 H.C. Siegmann, J. Phys.: Condens. Matter 4, 8935 1992 . Mater. 80, 211 1989 . 2 U. Gradmann, in Handbook of Magnetic Materials, edited by 11 B. Raquet, R. Mamy, and J.C. Ousset, Phys. Rev. B 54, 4128 K.H. Buschow Elsevier Science Publishers, 1993 , Vol. 7. 1996 . 3 J. Mathon and S.B. Ahmad, Phys. Rev. B 37, 660 1988 . 12 R. Allenspach, M. Taborelli, M. Landolt, and H.C. Siegmann, 4 C.L. Fu et al., Phys. Rev. Lett. 54, 2700 1985 . Phys. Rev. Lett. 56, 953 1986 . 5 G. Panaccione, F. Sirotti, and G. Rossi, Solid State Commun. to 13 A.P. Popov and D.P. Pappas, Phys. Rev. B 56, 3222 1997 . be published . 14 J. Ferre´, V. Grolier, P. Meyer, S. Lemerle, A. Maziewski, E. 6 F. Gauthier, in Rayonnement Synchrotron Polarise´ Electrons Po- Stefanowicz, S.V. Tarasenko, V.V. Tarasenko, M. Kisielewski, larise´s et Magnetisme, edited by E. Beaurepaire, B. Carrie re, and D. Renard, Phys. Rev. B 55, 15 092 1997 . and J.P. Kappler IPCMS, Strasbourg, 1989 . 15 7 H.C. Siegmann, D. Mauri, D. Scholl, and E. Kay, J. Phys. Paris , F. Sirotti, R. Bosshard, P. Prieto, G. Panaccione, L. Floreano, A. Colloq. 49, C8-9 1988 . Jucha, J.D. Bellier, and G. Rossi, J. Appl. Phys. 83, 1563 1998 . 16 8 D. Mauri, D. Scholl, H.C. Siegmann, and E. Kay, Phys. Rev. Lett. S. Girlando, thesis Tesi di Laurea , Universita di Modena e Reg- gio Emilia, 1998. 61, 758 1988 . 17 9 R.H. Victora, Phys. Rev. Lett. 63, 457 1989 . F. Sirotti, G. Panaccione, and G. Rossi, Phys. Rev. B 52, R17 063 10 M. Labrune, S. Andrieu, F. Rio, and P. Bernstein, J. Magn. Magn. 1995 .