3050 IEEE TRANSACTIONS ON MAGNETICS, VOL. 36, NO. 5, SEPTEMBER 2000 Metastable States in Large Angle Magnetization Rotations Pavel Kabos, Senior Member, IEEE, Shehzaad Kaka, Stephen E. Russek, and Thomas J. Silva, Member, IEEE Abstract—Large angle rotations were studied in Permalloy thin films and spin valve devices. It is shown that edge effects or inhomo- geneities strongly influence the magnetization behavior under cer- tain initial conditions. Complicated intermediate states can form which significantly increase the magnetization response time. Index Terms—Small particles, spin valves, switching, spin dy- namics and relaxation, magnetocrystalline anisotropy. I. INTRODUCTION T HE operation of write heads, giant magnetoresistance recording heads, and magnetic random access memory in the gigahertz frequency range will require a detailed un- derstanding of the underlying dynamics. However, very little information is available in the literature about the magneti- zation dynamics for large angle rotations and the gigahertz frequency range. Recently, excitation of spin waves in the switching process was examined theoretically [1] and through micromagnetic calculations [2], [3]. These calculations predict that: (a) in the process of switching, nonlinear spin waves are excited, with significant short time scale influence on the average magnetization, and (b) the presence of inhomogeneities should significantly influence the switching process. In this paper, we experimentally address some aspects of these theoretical predictions in thin films and spin valves. It is demonstrated that inhomogeneities strongly influence the magnetization dynamics. As a consequence, complicated intermediate states can form in both small magnetic devices and single-layer films. The formations of these intermediate states significantly increase the device response time. II. EXPERIMENT We experimentally studied two types of samples. The first was a spin valve GMR device with dimensions m m. The spin valve structure Ta(5 nm)/Ni Fe (5 nm)/Co(1 nm)/ Cu(1 nm)/Co(3 nm)/Ru(0.6 nm)/Co(1.5 nm)/FeMn (10 nm)/Ta(5 nm) was sputter-deposited on high resis- tivity Si wafers covered with a thermal SiO layer. In this device structure the exchange interactions between the free and pinned layers are minimized due to the low net moment of the pinned layer. The magnetization dynamics are predominantly Manuscript received February 15, 2000. The authors are with the National Institute of Standards and Technology, Boulder, CO 80303 USA (e-mail: kabos@lamar.colostate.edu). Publisher Item Identifier S 0018-9464(00)08510-1. confined to the free layer, and the device may be viewed, to first approximation, as a single, independent magnetic layer. Measurements on the spin valve device were performed with the pin direction and the drive field along the hard axis, trans- verse to the long, easy axis of the device. A constant 1.0 to 2.0 mA bias current was applied to the device and the high frequency magnetoresistance signal was separated with a dc bias “tee” and measured with a commercial 20 GHz bandwidth sampling oscilloscope [4]. Current pulses, supplied by a pulse generator connected to a 50 characteristic impedance mi- crostrip write line above the device, create the driving magnetic field. The pulse rise-time was 100 ps and the pulse duration was 150 ps. The second sample was a macroscopic cm cm square, 50 nm thick Permalloy film. The film was fabricated by sputter deposition onto a 100 m thick sapphire substrate. The sub- strate was glued to a 50 characteristic impedance coplanar waveguide, with the film side up, to permit stroboscopic mea- surements of the dynamic response with the second harmonic magneto-optic Kerr effect (SHMOKE) [5], [6]. The film was biased along the easy axis with an external mag- netic field of about 80 A/m (1 Oe) to sweep any domain walls out of the sample. Driving pulses of 1040 A/m (13 Oe) and 150 ps rise time and 2 ns pulse duration were applied along the hard axis in a manner similar to that of the spin valve. The field is supposed to be relatively uniform over the 500 m wide center conductor of the coplanar waveguide. Vectorial SHMOKE was used, allowing for the separation of and components. The SHMOKE apparatus is described in detail in [5], and its modification for vectorial SHMOKE measurements is shown in [6]. In the vectorial SHMOKE exper- iment, 60 fs light pulses at a wavelength of 800 nm, produced by a mode-locked Ti:sapphire laser, are focused on the sample in a -incidence geometry in the middle of the center conductor of the coplanar waveguide. The spot size was 5 m, and the peak intensities exceed 100 GM/cm . At such high power densities, the Permalloy surface acts as a second harmonic generator, pro- ducing reflected pulses at half the wavelength and collinear with the reflected pulses at the fundamental frequency. The funda- mental component of the reflected light is then filtered out, and the second harmonic light passes through a photo-elastic mod- ulator (PEM), an analyzer, and is directed into a photo-detector. The signal from the photo-detector is processed by photon counting electronics, and, at the same time, serves as an input to the lock-in amplifier. In the -incidence geometry, the signal from the photo-detector is proportional to the transverse (per- pendicular to the plane of incidence) component of the magneti- zation. The signal from the lock-in amplifier (at the fundamental 0018–9464/00$10.00 © 2000 IEEE