In a recent paper [M.-C. Ciornei et al., Phys. Rev. B 83, 020410(R) (2011)], it has been shown by mesoscopic nonequilibrium thermodynamics that, for relatively fast magnetization dynamics, Gilbert's equation of motion for the magnetization has to be supplemented by an ``inertial'' damping term that contains the second derivative of the magnetization with respect to time. In the present Brief Report, it is shown that such an additional damping term can be derived and evaluatedvery naturally also within a slightly extended breathing Fermi surface model for magnetization damping.
17 November 2011 in Journal of Physics: Condensed Matter
For metallic magnets we review the experimental and electron-theoretical investigations of fast magnetization dynamics (on a timescale of ns to 100 ps) and of laser-pulse-induced ultrafast dynamics (few hundred fs). It is argued that for both situations the dominant contributions to the dissipative part of the dynamics arise from the excitation of electron-hole pairs and from the subsequent relaxation of these pairs by spin-dependent scattering processes, which transfer angular momentum to the lattice. By effective field theories (generalized breathing and bubbling Fermi-surface models) it is shown that the Gilbert equation of motion, which is often used to describe the fast dissipative magnetization dynamics, must be extended in several aspects. The basic assumptions of the Elliott-Yafet theory, which is often used to describe the ultrafast spin relaxation after laser-pulse irradiation, are discussed very critically. However, it is shown that for Ni this theory probably yields a value for the spin-relaxation time T(1) in good agreement with the experimental value. A relation between the quantity α characterizing the damping of the fast dynamics in simple situations and the time T(1) is derived.
Despite intensive research, the underlying mechanisms for ultrafast demagnetization after laser irradiation in transition metals are still not understood. We discuss the possible processes which have been suggested in order to explain the ultrafast demagnetization within several hundreds of femtoseconds and argue that the spin angular momentum has to go to the lattice in the end. Based on this argument, we consider spin-flip electron-phonon scatterings. The demagnetization time ${$\tau${}}_{M}$ and the demagnetization rate $dM/dt$ due to spin-flip electron-phonon scattering is calculated for fcc Ni and bcc Fe. Thereby, the electronic states and phononic states are calculated ab initio. We find that the demagnetization rates for fcc Ni and bcc Fe are too small to explain experimental demagnetization rates, which is in agreement with earlier publications. In addition, the demagnetization rates for band structures with reduced exchange splitting are calculated, however, also these demagnetization rates are too small. Finally, the phase space for scattering processes which is related to the maximum possible demagnetization is estimated for band structures with ground-state exchange splitting and with reduced exchange splitting. The maximum possible demagnetization is too small for bcc Fe and fcc Co but not necessarily for fcc Ni. We suggest to include magnons and to consider independent combinations of spin-flip electron-phonon and spin-flip electron-magnon scattering processes as a possible explanation for the ultrafast demagnetization.
Possible contributions of electron-magnon scatterings for the ultrafast demagnetization after femtosecond laser irradiation of films of Ni and Fe are investigated by the ab initio spin-density-functional electron theory. The calculations are based on Fermi's golden rule for transition rates, and the nonequilibrium after the action of the laser pulse is modeled by different chemical potentials for electrons which are in a ``dominant spin-up'' or in a ``dominant spin-down'' state. First, it is shown that the experimentally observed demagnetization cannot be described by electron-magnon spin-flip scatterings with concomitant changes of the electronic orbital moments which are immediately quenched by the crystal field, a mechanism which has been suggested by Carpene et al. Second, it is argued that the experimentally observed demagnetization possibly can be explained by a combination of individual spin-flip electron-phonon and spin-flip electron-magnon processes. A precondition for this is (among others) that the magnon emission rate is much larger than the magnon absorption rate. It is shown that this precondition is indeed fulfilled for Ni and especially for Fe.
Usually, the adiabatic magnetic exchange-energy hypersurface is parametrized in terms of the bilinear Heisenberg interactions in pairs of atoms. For general magnetic configurations, this model is not complete even if it includes pairs with up to infinite interatomic distances. In contrast, the modeling by an in principle infinite spin-cluster expansion is complete for all conceivable magnetic configurations. In the present Letter, it is shown for bcc and fcc iron that a very accurate representation can be reached with a finite expansion with 20 terms which include biquadratic or multispin non-Heisenberg interactions.
Three-dimensional micromagnetic simulations of the switching of a magnetic vortex core in a cylindrical nanodisk are performed, for excitations with out-of-plane fields (symmetric switching) or with various types of time-dependent in-plane fields (asymmetric switching). Although the switching mechanisms are different in detail, all switching events must involve the movement of a Bloch point through the disk, because the switching leads to a change of the Skyrmion number which is a topological invariant as long as there is no action of a Bloch point. The momentary magnetization configurations are different in different layers of the disk. Because of the three-dimensionality it is often difficult to decide whether the asymmetric switching is caused by the splitting of the dip close to the vortex core into a vortex-antivortex pair, and the annihilation of the original vortex with the antivortex (whereby a Bloch point moves). It is suggested that there are situations for which such a switching occurs by the formation of a Bloch point in a configuration which is already similar to a vortex-antivortex configuration, but by a movement of this Bloch point before the formation of a complete pair and without the annihilation of the original vortex with an antivortex.