1: % This is revtex of Safonov and Bertram
2:
3:
4: \documentstyle[amssymb,twocolumn,aps]{revtex}
5: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
6: %TCIDATA{OutputFilter=LATEX.DLL}
7: %TCIDATA{LastRevised=Tue Aug 22 09:57:08 2000}
8: %TCIDATA{<META NAME="GraphicsSave" CONTENT="32">}
9: %TCIDATA{Language=American English}
10: %TCIDATA{CSTFile=revtex.cst}
11:
12: \begin{document}
13: \draft
14: \title{Magnetization reversal in a ``quasi'' single domain magnetic grain: a new
15: numerical micromagnetic technique}
16: \author{Vladimir L. Safonov and H. Neal Bertram}
17: \address{Center for Magnetic Recording Research, University of California \\
18: -- San Diego, \\
19: 9500 Gilman Dr., La Jolla, CA 92093-0401, U.S.A.}
20: \date{\today}
21: \maketitle
22:
23: \begin{abstract}
24: Magnetization reversal in a fine ferromagnetic grain is simulated for the
25: case of an instantaneously applied reversal magnetic field. The Hamiltonian
26: of the system contains the exchange interaction, the uniaxial anisotropy,
27: the Zeeman energy and the dipole-dipole interactions. A cubic grain is
28: discretized into 64 cubic subgrains and the coupled gyromagnetic equations
29: of motion are solved without phenomenological damping. A new scheme to solve
30: these equations is developed that utilizes only two variables per sub-cube
31: magnetization and strictly conserves the absolute magnitude. The initial
32: stage of reversal is uniform rotation followed by a nonlinear excitation of
33: nonuniform magnetic oscillations driven by this uniform mode. An excess of
34: the initial Zeeman energy is transformed into nonlinear spin waves, allowing
35: the average magnetization to substantially reverse. The process of
36: magnetization reversal in fine quasi-single-domain grain exhibits general
37: features of Hamiltonian wave systems with nonlinear diffusion. This
38: nonlinear diffusion is forbidden for either a strong reversal field and/or a
39: small grain size.
40: \end{abstract}
41:
42: \pacs{75.50.Tt, 75.40.Mg, 76.60.Es}
43:
44: \section{Introduction}
45:
46: The problem of magnetization reversal in a fine single-domain grain is of
47: great importance in magnetic recording physics. Characteristic reversal
48: times under strong reversal magnetic field give a physical limitation to
49: both data rates and system signal to noise ratio. Dynamic magnetization
50: reversal depends on the system or magnetic energy relaxation mechanisms. In
51: addition to energy absorption to the ``thermal bath'' of the host lattice,
52: magnetization reversal can occur by excitation of nonlinear `spin-wave' like
53: modes. In the present paper we explore in detail the excitation of these
54: modes in a quasi-single-domain particle. This magnetization reversal process
55: exhibits features of several problems of nonlinear and solid state physics,
56: such as kinetics of an orientational phase transition and energy transfer in
57: nonlinear systems of oscillators. Therefore the problem is of interest to
58: fundamental science.
59:
60: Dynamic magnetization reversal begins when the external magnetic field
61: reaches and exceeds the so-called `nucleation' field and the original
62: remanent state becomes unstable. Typically these dynamical processes in both
63: single and multidomain magnetic systems are solved by integration of the
64: Landau-Lifshitz equations \cite{landau} with phenomenological damping \cite
65: {bertramzhu}. This form of magnetization dynamics describes a coherent
66: rotation of each discretized subgrain, in which the magnetization magnitude
67: is conserved. Such approximation is fair for small excitations, as in linear
68: ferromagnetic resonance {\cite{callen,sparks,gurevich}}, or for some
69: particular model cases when an excess of energy directly goes to the thermal
70: bath without excitation of spin waves.
71:
72: Recently we have explored magnetization reversal in quasi-single-domain
73: grain by neglecting phenomenological damping and specifically exploring
74: nonlinear spin-wave like magnetic excitations \cite{safbert},\cite{safbert1}%
75: . By quasi-single-domain we mean grains whose dimension is on the order of
76: the domain wall width or in some cases the exchange length. Similar analysis
77: has also been done on reversal in thin films \cite{boerner},\cite{zhufilm}.
78: In all these cases, according to Suhl's speculation \cite{suhl2}, the
79: switching process should be rapid, governed by nonlinear spin-wave
80: excitations. Recently spin-waves excitated by large magnetization rotations
81: have been observed experimentally \cite{kabos}.
82:
83: By neglecting phenomenological damping, the total system magnetic energy is
84: conserved. These simulations have demonstrated that the excess of Zeeman
85: energy is transformed to the exchange, anisotropy and magnetostatic energies
86: by nonlinear spin-wave excitations. The purpose of this paper is to present
87: a detailed study of magnetization reversal simulations focusing on magnetic
88: interactions in a single grain.
89:
90: In addition, we describe here a new numerical scheme of solving the
91: Landau-Lifshitz equations. Typically numerical micromagnetic simulations
92: invoke the three Cartesian coordinates of each discretization cube. At each
93: integration time step the magnetization is modified to satisfy the condition
94: of constant magnitude. This scheme is preferable to utilization of the
95: natural two-parameter polar coordinate scheme because of singularities. In
96: polar coordinates there are always two singularities at the poles. In this
97: new scheme we utilize two coordinates per magnetization sub-cube, however
98: here these coordinates are similar to linear combinations of the transverse
99: magnetization components. Using a simple scheme, all singularieties are
100: avoided.
101:
102: In section 2 we discuss the micromagnetic model for our quasi-single-domain
103: grain. In section 3 we describe the implementation of our new numerical
104: model. In sections 4, 5 results of simulations are considered in detail.
105: Then we discuss obtained results and draw conclusions.
106:
107: \section{Model}
108:
109: The magnetic energy (Hamiltonian) of the grain can be written in the form:
110: \begin{equation}
111: {\cal H}=\int ({\cal U}_{{\rm exch}}+{\cal U}_{{\rm anis}}+{\cal U}_{{\rm Z}%
112: }+{\cal U}_{{\rm dmag}})d{\bf r}. \label{energy}
113: \end{equation}
114: Here
115: \begin{equation}
116: {\cal U}_{{\rm exch}}=\frac{A}{M_{{\rm s}}^{2}}\left[ \left( \frac{\partial
117: {\bf M}}{\partial x}\right) ^{2}+\left( \frac{\partial {\bf M}}{\partial y}%
118: \right) ^{2}+\left( \frac{\partial {\bf M}}{\partial z}\right) ^{2}\right]
119: \label{en-exch}
120: \end{equation}
121: is the exchange energy density.
122: \begin{equation}
123: {\cal U}_{{\rm anis}}=K_{{\rm u}}\left[ 1-\left( \frac{{M_{z}}}{M_{{\rm s}}}%
124: \right) ^{2}\right] \label{en-anis}
125: \end{equation}
126: is the uniaxial anisotropy energy density,
127: \begin{equation}
128: {\cal U}_{{\rm Z}}=-{\bf H}\cdot {\bf M} \label{en-Z}
129: \end{equation}
130: is the Zeeman energy density and ${\cal U}_{{\rm dmag}}$ is the
131: magnetostatic energy density.
132:
133: For simplicity we shall consider a micromagnetic grain as a system of $%
134: n=4\times 4\times 4$ (see, Fig.1) coupled cubic subgrains of the volume $%
135: V=L^{3}$, where $L$ is the linear size of sub-cube. Each sub-cube can be
136: characterized as a classical spin:
137:
138: \begin{equation}
139: {\bf S}_{j}\equiv -{\bf M}_{j}V/\hbar \gamma , \label{classspin}
140: \end{equation}
141: where ${\bf M}_{j}$ is the vector magnetization of $j-$th sub-cube, $\hbar $
142: is Plank's constant and $\gamma >0$ is the gyromagnetic ratio. $|{\bf M}%
143: _{j}|=M_{{\rm s}}$, where $M_{{\rm s}}$ is the saturation magnetization, and
144: therefore, $|{\bf S}_{j}|=S\equiv M_{{\rm s}}V/\hbar \gamma $. Each spin is
145: localized in the center of its sub-cube. The characterization of each
146: sub-cube magnetization in terms of an effective spin is our choice: the
147: problem we consider is purely classical. However the spin notation is
148: convenient for the numerical scheme presented in Sec.3. Spin
149: characterization also provides a bridge to quantum spin models.
150:
151: The Hamiltonian Eq.\ (\ref{energy}) of the system in terms of spin notation
152: can be written as:
153: \begin{equation}
154: {\cal H}={\cal H}_{{\rm exch}}+{\cal H}_{{\rm anis}}+{\cal H}_{{\rm Z}}+%
155: {\cal H}_{{\rm dmag}}. \label{hamiltonian}
156: \end{equation}
157: Here
158: \begin{equation}
159: {\cal H}_{{\rm exch}}=-\,{\frac{J}{2}}\,{\sum_{j\neq j\prime }}^{\prime }(%
160: {\bf S}_{j}\cdot {\bf S}_{j\prime }-S^{2}) \label{ham_exch}
161: \end{equation}
162: describes the exchange interaction between the nearest neighbors, $%
163: \,J=2AL/S^{2}$. Note that this form is identical to the usual micromagnetic
164: approximation that assumes a linear variation of the magnetization between
165: grain centers \cite{bertramzhu}. The term $-S^{2}$ in (\ref{ham_exch})
166: yields ${\cal H}_{{\rm exch}}=0$ in the case when all spins are oriented in
167: the same direction.
168:
169: \begin{equation}
170: {\cal H}_{{\rm anis}}=\frac{\hbar \gamma H_{{\rm K}}S}{2}\,\sum_{j=1}^{N}\,%
171: \left[ 1-\left( \frac{S_{j}^{z}}{S}\right) ^{2}\right] \label{ham_anis}
172: \end{equation}
173: is the uniaxial anisotropy energy, where $H_{{\rm K}}=2K_{{\rm u}}/M_{{\rm s}%
174: }$ is the anisotropy field. ${\cal H}_{{\rm anis}}=0$ corresponds to all
175: spins parallel to the anisotropy axis.
176: \begin{equation}
177: {\cal H}_{{\rm Z}}=\hbar \gamma {\bf H}\cdot \sum_{j=1}^{N}{\bf S}_{j}-\hbar
178: \gamma HNS \label{ham_Z}
179: \end{equation}
180: is the Zeeman energy. ${\cal H}_{{\rm Z}}=0$ if either $H=0$ or all spins
181: are oriented along the external magnetic field direction.
182:
183: The magnetostatic interaction of the sub-cubes will be approximated by the
184: dipole-dipole interaction energy
185:
186: \begin{eqnarray}
187: {\cal H}_{{\rm dmag}} &=&{\frac{(\hbar \gamma )^{2}}{2}}\,\sum_{j\neq
188: j\prime }\left[ \frac{{\bf S}_{j}\cdot {\bf S}_{j\prime }}{r_{jj\prime }^{3}}%
189: -\frac{3({\bf S}_{j}\cdot {\bf r}_{jj\prime })({\bf S}_{j\prime }\cdot {\bf r%
190: }_{jj\prime })}{r_{jj\prime }^{5}}\right] \nonumber \\
191: &&-{\cal H}_{{\rm dmag}}^{(0)} \label{ham_dd}
192: \end{eqnarray}
193: where ${\bf r}_{jj\prime }={\bf r}_{j}-{\bf r}_{j\prime }$ and ${\bf r}_{j}$%
194: , ${\bf r}_{j\prime }$ are the radius-vectors of the $j$ and $j^{\prime }$
195: spins, respectively. The constant ${\cal H}_{{\rm dmag}}^{(0)}$ is chosen so
196: that ${\cal H}_{{\rm dmag}}=0$ if all spins are oriented upward along the $z$
197: direction.
198:
199: We shall study the problem of time evolution of the normalized total
200: magnetization
201: \begin{equation}
202: {\bf m}=\frac{1}{NM_{{\rm s}}}\,\sum_{j}{\bf M}_{j}=-\frac{1}{NS}\,\sum_{j}%
203: {\bf S}_{j} \label{magnetization}
204: \end{equation}
205: in the system (\ref{hamiltonian}) from the initial state when $H=0$ and the
206: averaged magnetization at $t<0$ is oriented ``upward'' in the $z$ direction.
207: At the moment $t=0$, a strong reversal magnetic field ${\bf H}%
208: =(H^{x},0,H^{z})$ is applied. The component $H^{z}$ is negative and the
209: transverse component $H^{x}$ is taken to be relatively small ($|H^{x}|\ll
210: |H^{z}|\sim H_{{\rm K}}$). The net vector field is sufficient to give only
211: one energy minimum corresponding to an almost ``downward'' orientation of
212: the averaged magnetization. Thus the system acquires an excess of Zeeman
213: energy which can be later transformed to nonlinear spin excitations
214: containing magnetic anisotropy, exchange and dipole-dipole interaction
215: energies. This transformation will cause the net magnetization magnitude $|%
216: {\bf m}|$ to reduce. Mathematically the question of energy transfer is
217: described in Appendix B.
218:
219: \section{Numerical scheme}
220:
221: The Landau-Lifshitz equation without damping in terms of classical spin (\ref
222: {classspin})\ can be written as
223: \begin{eqnarray}
224: \frac{d{\bf S}_{j}}{dt} &=&-{\bf S}_{j}\times \gamma {\bf H}_{{\rm eff,}j%
225: {\rm \ }}, \label{merm-ll} \\
226: \gamma {\bf H}_{{\rm eff,}j} &\equiv &\frac{\partial {\cal H}/\hbar }{%
227: \partial {\bf S}_{j}}. \label{efffield}
228: \end{eqnarray}
229: This equation describe the evolution of coupled spin components $%
230: S_{j}^{x},~S_{j}^{y}$ and $S_{j}^{z}$. As time evolves, Eqs.(\ref{merm-ll}),(%
231: \ref{efffield}) give two system conservations: the net spin of each subgrain
232: is fixed
233:
234: \begin{equation}
235: (S_{j}^{x})^{2}+(S_{j}^{y})^{2}+(S_{j}^{z})^{2}=S^{2}, \label{spin2}
236: \end{equation}
237: and the total magnetic energy remains constant. Thus only two independent
238: variables exist for each $j$-th site.
239:
240: It is natural to write down only two differential equations for each spin in
241: terms of spin components or other generalized variables. The most popular
242: form of ``two-variable'' Landau-Lifshitz equation is known to be in
243: spherical coordinates. However the use of spherical coordinates for an
244: arbitrary motion of spin gives a stiff system of differential equations with
245: singularities at the poles. Moreover, numerical calculations of
246: trigonometric functions typically take much longer time than simple
247: arithmetic operations. This is why the spherical coordinate approach is not
248: usually used in micromagnetic calculations. In practice the direct method of
249: numerical solution of the Landau-Lifshitz equation is used (see, for
250: example, \cite{Schabes,nakatani}). Dynamic equations for all three variables
251: are integrated with continuum numerical modification so that the Eq. (\ref
252: {spin2}) is maintained.
253:
254: In this paper we present a new scheme for numerical solution of the
255: Landau-Lifshitz equation (\ref{merm-ll}), in which only two variables are
256: utilized and condition (\ref{spin2}) is automatically maintained. We
257: introduce the real and dimensionless variables $q_{j}$ (as a generalized
258: coordinate)\ and $p_{j}$ (as a generalized momentum) as follows
259: \begin{eqnarray}
260: S_{j}^{x} &=&q_{j}\,\sqrt{(S+S_{j}^{z})/2}, \nonumber \\
261: S_{j}^{y} &=&p_{j}\,\sqrt{(S+S_{j}^{z})/2}, \nonumber \\
262: S_{j}^{z} &=&S-(q{_{j}^{2}+}p{_{j}^{2}})/2. \label{hopri-sch}
263: \end{eqnarray}
264: These transformations are exact and can be considered to be a form of
265: classical Holstein-Primakoff \cite{hopri} transformation (see Appendix).
266: Schl\"{o}mann \cite{schlo} also presented a classical spin in a similar form.
267:
268: The equations of motion for $q_{j}$ and $p_{j}$ are the exact analog of the
269: Landau-Lifshitz equations (\ref{merm-ll}) and can be written in the form
270:
271: \begin{mathletters}
272: \begin{eqnarray}
273: \frac{dq_{j}}{d\tau } &=&\frac{{\bf H}_{{\rm eff,}j}}{H_{{\rm K}}}\cdot
274: \frac{\partial {\bf S}_{j}}{\partial p_{j}}, \label{q1-LL} \\
275: \frac{dp_{j}}{d\tau } &=&-\frac{{\bf H}_{{\rm eff,}j}}{H_{{\rm K}}}\cdot
276: \frac{\partial {\bf S}_{j}}{\partial q_{j}}, \label{p1-LL}
277: \end{eqnarray}
278: where a dimensionless time $\tau =\gamma H_{{\rm K}}t$ is introduced.
279: Corresponding spin derivatives are
280:
281: \end{mathletters}
282: \begin{eqnarray}
283: \frac{\partial S_{j}^{x}}{\partial q_{j}} &=&\sqrt{\frac{S+S_{j}^{z}}{2}}-%
284: \frac{q_{j}^{2}}{4}\sqrt{\frac{2}{S+S_{j}^{z}}}, \nonumber \\
285: \frac{\partial S_{j}^{y}}{\partial p_{j}} &=&\sqrt{\frac{S+S_{j}^{z}}{2}}-%
286: \frac{p_{j}^{2}}{4}\sqrt{\frac{2}{S+S_{j}^{z}}}, \nonumber \\
287: \frac{\partial S_{j}^{x}}{\partial p_{j}} &=&\frac{\partial S_{j}^{y}}{%
288: \partial q_{j}}=-\frac{q_{j}p_{j}}{4}\sqrt{\frac{2}{S+S_{j}^{z}}}, \nonumber
289: \\
290: \frac{\partial S_{j}^{z}}{\partial q_{j}} &=&-q_{j},\quad \frac{\partial
291: S_{j}^{z}}{\partial p_{j}}=-p_{j}. \label{sderiv1}
292: \end{eqnarray}
293:
294: It is necessary to mention that Eqs.(\ref{q1-LL}), (\ref{p1-LL}) contain a
295: singularity at $S+S_{j}^{z}=0$. To avoid this singularity we shall use (\ref
296: {q1-LL}), (\ref{p1-LL}) only in the hemisphere $|{\bf S}_{j}|=S$\ with $%
297: S_{j}^{z}\geq 0$. This is illustrated in Fig.2. Following Eq.(5), negative
298: magnetization corresponds to a positive spin and the ($p,q$) coordinates are
299: in the upper half hemisphere. Initially before reversal, when $M_{z}>0$ it
300: is convenient to introduce analogous variables $Q_{j}$ and $P_{j}$ (see
301: Appendix A) for the hemisphere $|{\bf S}_{j}|=S$\ with $S_{j}^{z}<0$ :
302:
303: \begin{eqnarray}
304: S_{j}^{x} &=&Q_{j}\,\sqrt{(S-S_{j}^{z})/2}, \nonumber \\
305: S_{j}^{y} &=&-P_{j}\,\sqrt{(S-S_{j}^{z})/2}, \nonumber \\
306: S_{j}^{z} &=&-S+(Q{_{j}^{2}+}P{_{j}^{2}})/2. \label{hp2}
307: \end{eqnarray}
308: The equations of motion for $Q_{j}$ and $P_{j}$ have the form:
309:
310: \begin{mathletters}
311: \begin{eqnarray}
312: \frac{dQ_{j}}{d\tau } &=&\frac{{\bf H}_{{\rm eff,}j}}{H_{{\rm K}}}\cdot
313: \frac{\partial {\bf S}_{j}}{\partial P_{j}}, \label{q2-LL} \\
314: \frac{dP_{j}}{d\tau } &=&=-\frac{{\bf H}_{{\rm eff,}j}}{H_{{\rm K}}}\cdot
315: \frac{\partial {\bf S}_{j}}{\partial Q_{j}}, \label{p2-LL}
316: \end{eqnarray}
317: and spin derivatives are
318:
319: \end{mathletters}
320: \begin{eqnarray}
321: \frac{\partial S_{j}^{x}}{\partial Q_{j}} &=&\sqrt{\frac{S-S_{j}^{z}}{2}}-%
322: \frac{Q_{j}^{2}}{4}\sqrt{\frac{2}{S-S_{j}^{z}}}, \nonumber \\
323: \frac{\partial S_{j}^{y}}{\partial P_{j}} &=&-\sqrt{\frac{S-S_{j}^{z}}{2}}+%
324: \frac{P_{j}^{2}}{4}\sqrt{\frac{2}{S-S_{j}^{z}}}, \nonumber \\
325: \frac{\partial S_{j}^{x}}{\partial P_{j}} &=&-\frac{\partial S_{j}^{y}}{%
326: \partial Q_{j}}=-\frac{Q_{j}P_{j}}{4}\sqrt{\frac{2}{S-S_{j}^{z}}}, \nonumber
327: \\
328: \frac{\partial S_{j}^{z}}{\partial Q_{j}} &=&Q_{j},\quad \frac{\partial
329: S_{j}^{z}}{\partial P_{j}}=P_{j}. \label{sderiv2}
330: \end{eqnarray}
331:
332: A continuous transition from one set of variable to another one is defined by
333:
334: \begin{eqnarray}
335: q_{j}\, &=&Q_{j}\,\sqrt{\frac{S-S_{j}^{z}}{S+S_{j}^{z}}}{,} \nonumber \\
336: p_{j}\, &=&-P_{j}\,\sqrt{\frac{S-S_{j}^{z}}{S+S_{j}^{z}}}{,} \label{qp-QP}
337: \end{eqnarray}
338:
339: and
340:
341: \begin{eqnarray}
342: Q_{j} &=&q_{j}\,\sqrt{\frac{S+S_{j}^{z}}{S-S_{j}^{z}}}{,} \nonumber \\
343: P_{j} &=&-p_{j}\,\sqrt{\frac{S+S_{j}^{z}}{S-S_{j}^{z}}}{.} \label{QPqp}
344: \end{eqnarray}
345:
346: Thus, we begin the calculations with $P_{j}(0)$ and$\ Q_{j}(0)$. After a
347: short integration time $\Delta t$\ we find $P_{j}(\Delta t)$ and$\
348: Q_{j}(\Delta t)$ and calculate corresponding $S_{j}^{x}$, $S_{j}^{y}$ and $%
349: S_{j}^{z}$. If $S_{j}^{z}<0$, we continue integration of the Landau-Lifshitz
350: equations with $P_{j}$ and$\ Q_{j}$. In the case when $S_{j}^{z}\geq 0$, the
351: $j-$th spin trajectory crosses the equator and enters the upper hemisphere.
352: To avoid the singularity in the $P_{j}$ and$\ Q_{j}$ coordinates in the
353: upper pole (see Fig.2), the new variables $p_{j}$ and $q_{j}$ are introduced
354: by Eq.(\ref{qp-QP}). Integration of the Landau-Lifshitz equations using $%
355: p_{j}$ and $q_{j}$ until the $z$-component of the spin changes sign ($%
356: S_{j}^{z}<0$). At that point we change back to the $P_{j}$ and$\ Q_{j}$
357: coordinates.
358:
359: It should be noted that the Eqs.(\ref{q1-LL}), (\ref{p1-LL}), (\ref{q2-LL})
360: and (\ref{p2-LL}) have simple algebraic operations only and therefore
361: numerical integration of these equations do not contain time consuming
362: operations.
363:
364: A characteristic time scale of the system is $1/\gamma H_{{\rm K}}$. The
365: role of exchange interaction is defined by the parameter $H_{{\rm e}}/H_{%
366: {\rm K}}$, where $\,H_{{\rm e}}=JS/\hbar \gamma =2A/M_{{\rm s}}L^{2}\,$ is
367: the exchange field from the nearest neighbor. One can write
368:
369: \begin{equation}
370: H_{{\rm e}}/H_{{\rm K}}=(n^{2/3}/16)(\delta _{{\rm w}}/\delta _{{\rm s}%
371: })^{2}, \label{wall_length}
372: \end{equation}
373: where $\delta _{{\rm w}}=4(A/K_{{\rm u}})^{1/2}$ is the domain wall width, $%
374: \delta _{{\rm s}}=n^{1/3}L=4L$ is the linear sample size. The exchange
375: interaction energy can also be characterized by, so-called, ``exchange
376: length'' $\,\delta _{{\rm ex}}=A^{1/2}/M_{s}\,$. Therefore one can write
377: \begin{equation}
378: H_{{\rm e}}/H_{{\rm K}}=2n^{2/3}(M_{{\rm s}}/H_{{\rm K}})(\delta _{{\rm ex}%
379: }/\delta _{{\rm s}})^{2}. \label{exch_length}
380: \end{equation}
381:
382: We shall focus on the case of relatively strong uniaxial anisotropy when $M_{%
383: {\rm s}}/H_{{\rm K}}\ll 1$ (which is valid for real magnetic recording
384: grains). In the zeroth order of this small parameter we can neglect the
385: dipole-dipole interaction in the system. However as it will be demonstrated
386: in the next section that even relatively small dipole-dipole interaction,
387: which absorbs very small amount of the Zeeman energy plays an important role
388: in the magnetization reversal.
389:
390: \subsection{Numerical implementation and initial conditions}
391:
392: The system of Landau-Lifshitz equations has been numerically solved for $%
393: n=64 $. High-accuracy solutions to these ordinary differential equations
394: were obtained with minimal computational effort by utilizing the
395: Bulirsch-Stoer method with adaptive stepsize control \cite{numrec}.
396: Integration steps\ were about four orders less than the characteristic time $%
397: 1/\gamma H_{{\rm K}}$.
398:
399: The equations of motion (\ref{q1-LL}), (\ref{p1-LL}), (\ref{q2-LL}), (\ref
400: {p2-LL}) contain relative parameters only. For simulations these parameters
401: were chosen to correspond to real magnetic recording media. For example,
402: with $A=10^{-6}$ erg/cm, $H_{{\rm K}}=7$ kOe and $M_{{\rm s}}=250$ Oe, the
403: scaled parameters are $\delta _{{\rm w}}\simeq 43$ nm, $\delta _{{\rm ex}%
404: }=40 $ nm and $M_{{\rm s}}/H_{{\rm K}}=0.036$ .
405:
406: Nonlinearities can only be excited were there are deviations in the
407: magnetization state from perfect alignment. We include this by allowing the
408: system to initially be in thermal equilibrium. We assign randomly a total
409: energy deviation of $64k_{{\rm B}}T$. Practically we randomly deviate the
410: magnetization of each cell within a fixed limit. This limit is adjusted to
411: achieve the above total energy deviation. To obtain a true thermal
412: equilibrium the Landau-Lifshitz equations are solved, given the above
413: initialization, for a time $\gamma H_{{\rm K}}t_{0}\sim 200$. At this point
414: the system is prepared for the application of the reversal field.
415:
416: \section{Magnetization reversal without dipole interactions}
417:
418: First we shall consider the case when the dipole-dipole interactions in $%
419: \gamma {\bf H}_{{\rm eff,}j}$ (\ref{efffield}) are neglected. Fig.3a,b show
420: the time dependencies of $m_{z}(t)$ and $|{\bf m}(t)|$ in the case when the
421: reverse magnetic field has a very small deviation from the anisotropy axis ($%
422: H_{x}/H_{z}=0.02$). Fig.3b is simply an extension of the time axis of
423: Fig.3a. The field magnitude is slightly less than the nucleation field for
424: uniform rotation ($H_{nuc}\simeq -0.9H_{{\rm K}}$). This field corresponds
425: (by numerical test) to the lowest reverse field required to produce
426: nonlinear excitations. The sample dimension is 16\% and 25\% greater than
427: the domain wall width and exchange length, respectively. Initially the
428: longitudinal component of magnetization decreases monotonically from $%
429: m_{z}\simeq 1$ to $\simeq 0.7$ and then at a `critical' time $\gamma H_{{\rm %
430: K}}t_{c}\simeq 70$, oscillations of $m_{z}$ occur. In contrast, the net
431: magnitude of magnetization remains practically constant $|{\bf m}|\simeq 1$\
432: until $t_{c}$ and then abruptly decreases exhibiting irregular oscillations.
433: For times less than $t_{c}$ the average magnetization simply precesses about
434: the reverse field. For times greater $t_{c}$ the excitation of nonlinear
435: spin waves becomes pronounced. These excitations both absorb Zeeman energy
436: and reduce the magnetization.
437:
438: This behavior is typical of nonlinear spin-wave processes whose amplitudes
439: grow exponentially from an initial thermal level. The decrease in net
440: magnetization becomes observable at virtually a critical time $t_{c}$\ when
441: the magnetic moment of the (transverse) excited waves becomes comparable
442: with the net magnetic moment. At the same time the system nonlinearities
443: restrict the exponential growth. The oscillations in $m_{z}$ that occur for $%
444: t>t_{c}$ primarily correspond to the uniform rotation: the frequency
445: increases with decreasing $m_{z}$. If spin-wave excitations did not occur,
446: oscillations in $m_{x}$, $m_{y}$, $m_{z}$ would reflect uniform rotation
447: only and at every instant sum to give constant $|{\bf m}|\simeq 1$.
448:
449: With increasing time the average $m_{z}$ continually decreases reaching, for
450: this example an asymptotic level in the vicinity of $m_{z}=0$ (at least for
451: times less than $\gamma H_{{\rm K}}t\sim 3000$). There is no reversal of $%
452: m_{z}(t)$. The average magnetization $|{\bf m}(t)|$ also decreases by about $%
453: \approx 60\%$. As time proceeds, the spin-wave oscillations exhibit
454: increased chaotic behavior.
455:
456: Fig.4 shows similar time dependences as in Fig.3 but for the case of
457: increased angle of the reverse magnetic field from the anisotropy axis ($%
458: H_{x}/H_{z}=0.10$). The field magnitude ($H_{z}/H_{{\rm K}}=-0.70$) also
459: corresponds to the onset of spin-wave excitation. As expected, the
460: magnetization dynamics is much faster in this case. We can see the beginning
461: of nonlinear spin-wave excitation at $\gamma H_{{\rm K}}t_{c}\simeq 30$,
462: where the $m_{z}$ has reduced due to pure rotation to $\simeq 0.3$. The
463: decrease of longitudinal magnetization is more substantial, passing through
464: zero at $\gamma H_{{\rm K}}t_{0}\simeq 75$ and reaching about $\approx 55\%$
465: of its nominal value in the opposite direction after $\gamma H_{{\rm K}%
466: }t\gtrsim 1000$. Similar to the case shown in Fig.3, the average
467: magnetization $|{\bf m}(t)|$ decreases to $\approx 60\%$. Note that the
468: amplitudes of chaotic oscillations of $|{\bf m}(t)|$\ are relatively smaller
469: than in the case of Fig.3 (compare the long time behavior of Fig.3b and
470: Fig.4b). Here most of the Zeeman energy is transformed into high frequency
471: exchange-dominant spin-wave modes, which have relatively small magnetic
472: moments.
473:
474: As discussed earlier, the total magnetic energy remains constant in these
475: dynamic processes. In Fig.5a,b we plot the evolution of the Zeeman, exchange
476: and anisotropy energies (Fig.5b is an extended scale of Fig.5a). The reverse
477: angle is intermediate between Fig.3 and 4. Initially, for times less than $%
478: t_{c}$ (in this case $\gamma H_{{\rm K}}t_{c}\simeq 45$), uniform rotation
479: yields a decrease of the Zeeman energy and an increase of the anisotropy
480: energy. At the onset of nonlinear spin-wave excitation we see, in addition,
481: an increase of the exchange energy. The wave-like nature of these
482: excitations can be seen in Fig.5a in the beating of the exchange and
483: anisotropy energies. The nonlinear excitation of chaotic oscillations of the
484: magnetization can be thought of as an ``overheating of the magnetic
485: system''. The more exchange magnetic modes that are involved in the process,
486: the smaller the averaged magnetization $|{\bf m}(t)|$.
487:
488: Fig.6a-d demonstrate the time dependencies of $m_{z}(t)$ and $|{\bf m}(t)|$
489: for various magnetic fields: $-0.78H_{{\rm K}}$, $-0.79H_{{\rm K}}$, $%
490: -0.90H_{{\rm K}}$ and $-1.0H_{{\rm K}}$, respectively. As in Fig.5 the field
491: angle is fixed at $H_{x}/H_{z}=0.05$ and relative sizes are unchanged. For
492: this angle the Stoner-Wohlfarth nucleation field is $-0.83H_{{\rm K}}$. In
493: this case the onset field for nonlinear spin-wave excitation is $H_{z}$ =$%
494: -0.79H_{{\rm K}}$. There is no reversal below this field as shown in Fig.6a.
495: The magnetization evolution at the onset field is shown in Fig.6b (similar
496: to Figs.3,4). Magnetization dynamics for larger fields are shown in
497: Figs.6c,d. A notable features can be seen: the critical time for nonlinear
498: spin-wave excitation increases with increasing field. The increase of $t_{c}$
499: allows for several cycles of uniform rotation to occur prior to the
500: nonlinear spin-wave excitation. No reversal dynamics was observed\ beyond $%
501: H_{z}<-1.1H_{{\rm K}}$ (in this particular case). Essentially as the reverse
502: field is increased beyond the spin-wave onset field, $t_{c}$ rapidly
503: increases and becomes infinite. These phenomena also occur for other field
504: angles ($H_{x}/H_{z}$).
505:
506: As indicated in Figs.6a-d the asymptotic, long time value of $m_{z}$
507: increases with increasing applied field (i.e. less magnetization reversal
508: occurs). In Fig.7 we explicitly plot the asymptotic values of averaged $%
509: \langle m_{z}(t)\rangle $ and $\langle |{\bf m}(t)|\rangle $ versus applied
510: field. It should be noted that the excitation of nonlinear spin waves occurs
511: only for a narrow band of applied fields.
512:
513: Fig.8 shows the evolution of magnetization at various sample sizes. We see
514: that a decrease of the sample size relative to the domain wall width leads
515: to a suppression of magnetization reversal. This fact is also can be
516: interpreted in terms of nonlinear spin-wave processes. By decreasing the
517: sample size, we increase the frequencies of the non-uniform spin-wave modes
518: and therefore restrict the number of resonance nonlinear processes that can
519: contribute to a decay of the uniform mode. Analysis for larger sample would
520: require finer discretization.
521:
522: \section{Magnetization reversal with dipole interactions}
523:
524: In this section we show numerical simulations that include the dipole-dipole
525: interactions in $\gamma {\bf H}_{{\rm eff,}j}$ (\ref{efffield}). The
526: relative role of dipole-dipole interaction is small because we assume a
527: small magnetostatic parameter $M_{{\rm s}}/H_{{\rm K}}\ll 1$. In this case
528: we can not expect a noticeable absorption of the Zeeman energy solely by the
529: dipole-dipole reservoir. Nevertheless, the dipole-dipole interaction plays
530: an important role in the process of energy transfer: it reduces the symmetry
531: of the system and therefore opens new channels of nonlinear spin-wave
532: interactions.
533:
534: Let us compare the results of Fig.3 obtained without dipole interaction and
535: corresponding results in Fig.9 with the dipole interaction. The onset time $%
536: t_{c}$ of nonlinear excitation is about the same. However $m_{z}(t)$
537: decreases more rapidly, reverses and reaches an asymptotic value of $\simeq
538: -0.5$. This is in contrast to Fig.2 where the asymptotic $m_{z}(t)$ does not
539: reverse. In this case the average magnetization $|{\bf m}(t)|$ decreases to $%
540: \approx 50\%$, slightly more than the $\approx 60\%$ in Fig.3. The
541: amplitudes of chaotic oscillations of $|{\bf m}(t)|$ in this case are much
542: smaller than occur in Fig.3. The excess Zeeman energy is transformed into
543: exchange modes, which have smaller magnetic moments.
544:
545: Plotted in Fig.10 is the magnetization dynamics for a larger field angle ($%
546: H_{x}/H_{z}=0.10$). As compared with Fig.4, including magnetostatic
547: interactions slightly increases the reversal rate. The longitudinal
548: magnetization passes through zero at $\gamma H_{{\rm K}}t_{0}\simeq 55$ and,
549: similar to Fig.4, reaches about $\approx 55\%$ of its nominal value in the
550: opposite direction after $\gamma H_{{\rm K}}t\gtrsim 500$. The average
551: magnetization $|{\bf m}(t)|$ is decreased to $\approx 60\%$.
552:
553: Fig.5c,d demonstrate typical transformations of Zeeman, anisotropy and
554: exchange energies for short (c) and long (d) time intervals for
555: magnetization dynamics with dipole interactions. We see faster dynamics of
556: energy transformations than shown in Fig.5a,b. It should be noted that the
557: energy absorbed by dipole interactions is relatively small (within $\pm 0.01$
558: in relative units) as expected and therefore not shown. However the
559: magnetostatic interactions change the balance the exchange and anisotropy
560: energies.
561:
562: In Fig.11 are shown the time dependencies $m_{z}(t)$ and $|{\bf m}(t)|$ for
563: various reversal fields. In agreement with Fig.6, the reversal dynamics is
564: suppressed with increasing the strength of reversal field but the range of
565: reversal dynamics is much broader. We did not observe any reversal dynamics
566: at $H_{z}\leq -3H_{{\rm K}}$ in this particular case. We believe that the $%
567: 4\times 4\times 4$ discretization is not a major limiting factor. For the
568: case of $n=1000$ subgrains no reversal dynamics was observed at $H_{z}\leq
569: -4H_{{\rm K}}$ \cite{safbert1}.
570:
571: In Fig.12 we plot the asymptotic relative change in Zeeman energy $\Delta E_{%
572: {\rm Z}}=(1-\langle m_{z}\rangle )(-H_{z}/H_{{\rm K}})$ versus the applied
573: magnetic field for the parameters of Fig.11 (solid dots). Even though there
574: is less asymptotic magnetization decrease as the field is increased, the
575: relative change in Zeeman energy does increase. After the onset of nonlinear
576: spin-wave excitation, in this case at $H_{z}=-0.79H_{{\rm K}}$, the amount
577: of spin-wave energy that can be excited increases to a maximum value. In
578: addition (open dots), the change in Zeeman energy is plotted corresponding
579: to Fig.5. for the case of no magnetostatic interaction. The same trend
580: occurs but only over a small interval of field values.
581:
582: Fig.13 shows the evolution of magnetization at various sample sizes. As in
583: Fig.8, the decrease of the sample size relative to the domain wall width
584: leads to a suppression of magnetization reversal. The minimum size for
585: spin-wave excitation remains the same, however differences are seen in the
586: amplitude of the chaotic excitations.
587:
588: We have also examined the effect of grain size relative to the exchange
589: length at fixed relative domain wall size ($\delta _{{\rm s}}/\delta _{{\rm w%
590: }}=1.16$). The system evolution is more sensitive to the change of exchange
591: length if the dipole-dipole interaction is taken into account. The primary
592: effect is that with a smaller relative exchange length the rate of decrease
593: of the average magnetization is faster. As discussed above, the presence of
594: magnetostatic interactions allows the excitation of additional spin-wave
595: modes that were otherwise degenerate. These modes are more easily excited if
596: the sample size is large compared to the exchange length.
597:
598: \section{Discussion}
599:
600: Magnetization reversal occurs by reducing the Zeeman energy as the
601: magnetization goes from an initial high energy state to one of low energy
602: (approximately in the field direction). Typically this dynamic process is
603: analyzed using the Landau-Lifshitz equation with an appropriate
604: phenomenological damping term. Here we have excluded phenomenological
605: damping and shown that substantial, rapid reversal can occur. In this case
606: the total magnetic energy is conserved, however subject to constraints of
607: sample size and field range, the Zeeman energy is reduced by the excitation
608: of nonlinear spin-waves.
609:
610: In the present study we have specifically focused on a cubic grain
611: sufficiently small to be almost single domain. In this case after uniform
612: rotation over a short time period, nonlinear spin waves are excited and the
613: average grain magnetization decreases. For the sample sizes examined here
614: the reduction was from positive saturation ($+M_{{\rm s}}$) to at most, $%
615: -0.6M_{{\rm s}}$. Due to numerical limitations we did not examine large
616: grains and it is possible that, with sufficient discretization, almost
617: complete reversal might occur for samples much larger than the domain wall
618: width. In this study we considered only samples with small relative
619: magnetostatic energy contribution. Studies have been performed on high
620: moment low anisotropy thin films, where a large out of plane magnetization
621: reversal torque occurs \cite{boerner},\cite{zhufilm}. In these cases
622: virtually complete reversal almost always occurs with no phenomenological
623: damping.
624:
625: The decrease and suppression of magnetization reversal with increasing the
626: reversal field strength initially seem unusual. In the discussion of Fig.11
627: a limit is suggested as due to the saturation of the spin-wave energy.
628: However, the problem we consider is a general system of nonlinear
629: oscillators (nonlinear spin waves). In the oscillator problem one mode (here
630: corresponding to coherent rotation of the magnetization) is\ initially
631: strongly excited. One might expect that the system would move towards its
632: most probable state (thermal equilibrium) by a chaotic mode mixing. With any
633: large reverse field we would expect reversal to a new equilibrium state.
634: However, the observed reversal decreases and suppression can be
635: qualitatively understood as a result of nonlinear spin-wave interactions.
636:
637: The processes of nonlinear spin-wave decay are defined by several factors.
638: First, it is necessary that the frequency of the initial mode and some
639: portion of the nonlinear spin-wave spectrum overlap. Second, we require
640: sufficiently strong nonlinear interwave coupling. These factors determine a
641: threshold amplitude of the decaying wave (as an effective pumping field).
642: Third, the amplitude of initial (decaying) wave must exceed a threshold
643: value. When the reversal field strength is increased, all these three
644: requirements eventually are not met. As suggested in Ref.\cite{suhl2},
645: increasing the reverse field shifts the spin-wave spectrum away from the
646: uniform precession mode. Also as the reversal field is increased nonlinear
647: coupling decreases. At a certain reversal field the threshold of nonlinear
648: decay will exceed the amplitude of coherent magnetization rotation and the
649: reversal process will be suppressed. Generally speaking, such a suppression
650: of the energy transfer from the excited mode to other modes is a common
651: feature of the theory of Hamiltonian systems of nonlinear oscillators (see,
652: e.g.,\cite{chirikov}). The most simple and popular example of nonlinear
653: coupled oscillators is the Fermi-Pasta-Ulam chain (see, e.g.,\cite{fpu}).
654:
655: The numerical scheme we have presented is valuable here primarily because
656: the number of coordinates is reduced by 2/3. Thus, the computational storage
657: requirement is correspondingly reduced. However, there is an additional
658: advantage to this formulation. These coordinates correspond naturally to
659: nonlinear oscillators that describe spin-wave motions. Recently, the problem
660: of thermally agitated reversal has been analyzed\ in a micromagnetic grain.
661: Using this nonlinear oscillator representation a simple exact solution has
662: been obtained for the first passage time \cite{safonov},\cite{safonbertram},
663: \cite{bertramsaf}.
664:
665: \section{Conclusion}
666:
667: Magnetization reversal in a quasi-single-domain grain can occur by nonlinear
668: spin-wave excitation. In this process an excess of Zeeman energy is
669: transfered to uniaxial anisotropy and exchange energies. The dipole-dipole
670: interaction plays an important role even if small ($M_{{\rm s}}/H_{{\rm K}%
671: }\ll 1$). The process of magnetization reversal in a fine grain exhibits
672: general features of system of interacting oscillators: energy transfer by
673: nonlinear resonance processes. This nonlinear process (diffusion) is
674: forbidden for either a strong reversal field and/or a small grain size.
675:
676: We have presented a new scheme to perform numerical integration of the
677: Landau-Lifshitz equations. This new scheme utilizes only two variables per
678: discretization cell, but strictly conserves the magnetization magnitude.
679: Comparison with conventional techniques show that both schemes require
680: approximately the same computation time at the same level of accuracy (the
681: new scheme takes a slightly shorter time). The main advantage of the new
682: scheme is that it utilizes a smaller number of variables (2/3 of the total
683: magnetization components) and therefore requires less memory.
684:
685: \section*{ Acknowledgments}
686:
687: The authors are grateful to Prof. Harry Suhl for valuable comments and
688: discussions. We also thank E. Boerner, M. E. Schabes and A. N. Slavin for
689: interesting and helpful discussions of obtained results. This work was
690: partly supported by matching funds from the Center for Magnetic Recording
691: Research at the University of California - San Diego and CMRR incorporated
692: sponsor accounts.
693:
694: \begin{references}
695: \bibitem{landau} L. Landau and E. Lifshitz, Phys. Z. Sowjet. {\bf 8}, 153
696: (1935); in Landau L. D. {\it Collected Papers}, ed. ter Haar, D., p.101
697: (Gordon and Breach, New York, 1967).
698:
699: \bibitem{bertramzhu} H. N. Bertram and J. G. Zhu, Solid State Phys. {\bf 46}%
700: , 271 (1992).
701:
702: \bibitem{callen} C. W. Haas and H. B. Callen, in: {\it Magnetism}, eds. G.
703: T. Rado and H. Suhl (Academic Press, New York, 1963), Vol. 1, p.449.
704:
705: \bibitem{sparks} M. Sparks, {\it Ferromagnetic-Relaxation Theory}
706: (McGraw-Hill, NY, 1964).
707:
708: \bibitem{gurevich} A. G. Gurevich and G. A. Melkov, {\it Magnetization
709: Oscillations and Waves} (CRC Press, Boca Raton, Fla., 1996).
710:
711: \bibitem{safbert} V. L. Safonov and H. N. Bertram, J. Appl. Phys. {\bf 85},
712: 5072 (1999).
713:
714: \bibitem{safbert1} V. L. Safonov and H. N. Bertram, J. Appl. Phys. {\bf 87,}
715: 5508 (2000).
716:
717: \bibitem{boerner} E. D. Boerner, H. N. Bertram, and H. Suhl, J. Appl. Phys.
718: {\bf 87},5389 (2000).
719:
720: \bibitem{zhufilm} C. Y. Mao, J. G. Zhu, and R. M. White, J. Appl. Phys.
721: {\bf 85}, 5870 (1999).
722:
723: \bibitem{suhl2} H. Suhl, IEEE Trans. Mag. {\bf 34}, 1834 (1998).
724:
725: \bibitem{kabos} P. Kabos, S. Kaka, S. E. Russek, and T. J. Silva,\ IEEE
726: Trans. Magn. (2000) in press.
727:
728: \bibitem{Schabes} M. E. Schabes and H. N. Bertram, J. Appl. Phys. {\bf 64},
729: 1347 (1988).
730:
731: \bibitem{nakatani} Y. Nakatani, Y. Uesaka, and N. Hayashi, Japanese J.
732: Appl. Phys. {\bf 28}, 2485 (1989).
733:
734: \bibitem{hopri} T. Holstein and H. Primakoff, Phys. Rev. {\bf 58}, 1098
735: (1940).
736:
737: \bibitem{schlo} E. Schl\"{o}mann, Phys. Rev. {\bf 116}, 828 (1959).
738:
739: \bibitem{numrec} W. H. Press, B. P. Flannery, S. A. Teukolsky, and W. T.
740: Vetterling, {\it Numerical Recipes} (Cambridge University Press, Cambridge,
741: 1986).
742:
743: \bibitem{chirikov} B. V. Chirikov, Phys. Reports {\bf 52}, 265 (1979).
744:
745: \bibitem{fpu} T. P. Weissert, {\it The Genesis of Simulation in Dynamics}
746: (Springer, New York, 1997).
747:
748: \bibitem{safonov} V. L. Safonov, J. Magn. Magn. Mater. {\bf 195}, 526
749: (1999).
750:
751: \bibitem{safonbertram} V. L. Safonov and H. N. Bertram, J. Appl. Phys. {\bf %
752: 87}, 5681 (2000).
753:
754: \bibitem{bertramsaf} H. N. Bertram and V. L. Safonov, IEEE Trans. Magn.
755: (2000) in press.
756:
757: \bibitem{mermin} N. D. Mermin, Phys. Rev. {\bf 134}, A112 (1964); J. Math.
758: Phys. {\bf 8}, 1061 (1967).
759:
760: \bibitem{suhl1} H. Suhl, J. Phys. Chem. Solids {\bf 1}, 209 (1957).
761: \end{references}
762:
763: \appendix
764:
765: \section{Equations of motions and canonical variables}
766:
767: \subsection{Classical commutations}
768:
769: Hamilton's equations can be written in a general form
770:
771: \begin{equation}
772: i\hbar \,\frac{d{\cal O}}{dt}=[\![\,{\cal O},{\cal H}\,]\!] \label{hameq}
773: \end{equation}
774: Here ${\cal O}$ is a function of interest and $[\![\,\ldots ,\ldots
775: \,]\!]=i\{\ldots ,\ldots \}$ denotes a classical analog of commutator and $%
776: \{\ldots ,\ldots \}$ are the Poisson brackets. Plank's constant $\hbar $ is
777: used as a dimensional constant to simplify a dimensional structure of the
778: equation and for simple correspondence with quantum mechanics.
779:
780: Taking into account the Mermin's \cite{mermin} formula for the Poisson
781: brackets for a system of classical spins ${\bf S}_{j}$, one can write:
782: \begin{equation}
783: \lbrack \![\,{\cal A},{\cal B}\,]\!]=i\sum_{j}{\bf S}_{j}\cdot \frac{%
784: \partial {\cal A}}{\partial {\bf S}_{j}}\times \frac{\partial {\cal B}}{%
785: \partial {\bf S}_{j}}. \label{merm}
786: \end{equation}
787: It is simple to show that the commutation rules for the classical spins are
788: the same as in quantum mechanics:
789: \begin{equation}
790: \lbrack \![\,S_{j}^{z},S_{j\prime }^{\pm }\,]\!]=\pm S_{j}^{\pm }\delta
791: _{jj\prime },\quad \lbrack \![\,S_{j}^{+},S_{j\prime
792: }^{-}\,]\!]=2S_{j}^{z}\delta _{jj\prime }. \label{spincom}
793: \end{equation}
794: where $S_{j}^{\pm }=S_{j}^{x}\pm iS_{j}^{y}$ are the circular spin
795: components.
796:
797: Substituting ${\cal O}={\bf S}_{j}$ to Eq.\ (\ref{hameq}) and taking into
798: account that
799:
800: \begin{eqnarray*}
801: \lbrack \![\,{\bf S},{\cal H}\,]\!] &=&i{\bf S}\cdot \frac{\partial {\bf S}}{%
802: \partial {\bf S}}\times \frac{\partial {\cal H}}{\partial {\bf S}} \\
803: &=&-i\frac{\partial {\bf S}}{\partial {\bf S}}\cdot {\bf S}\times \frac{%
804: \partial {\cal H}}{\partial {\bf S}} \\
805: &=&-i{\bf S}\times \frac{\partial {\cal H}}{\partial {\bf S}},
806: \end{eqnarray*}
807: one can obtain the Landau-Lifshitz equation (\ref{merm-ll}).
808:
809: \subsection{Holstein-Primakoff transformations}
810:
811: It may be easily shown that the commutations (\ref{spincom}) are valid if
812: \begin{eqnarray}
813: S^{z} &=&S-a^{\ast }a, \nonumber \\
814: S^{+} &=&a\,\sqrt{S+S^{z}}, \nonumber \\
815: S^{-} &=&a^{\ast }\,\sqrt{S+S^{z}}. \label{holstein-primakoff1}
816: \end{eqnarray}
817: These relations describe the classical spin realization in terms of complex
818: variables $a^{\ast }$ and $a$ corresponding to quantum creation and
819: annihilation operators, respectively. The transformation (\ref
820: {holstein-primakoff1}) in the quantum case was introduced by Holstein and
821: Primakoff \cite{hopri}. Suhl \cite{suhl1}\ first utilized approximate
822: Holstein-Primakoff transformation with complex variables to describe
823: magnetic resonance.
824:
825: The classical analog of the commutator for $\,a^{\ast }\,$ and $\,a\,$ is
826:
827: \begin{equation}
828: \lbrack \![{\cal A},{\cal B}]\!]\equiv {\frac{{\partial }{\cal A}}{{\partial
829: }a}}{\frac{{\partial }{\cal B}}{{\partial }a^{\ast }}}-{\frac{{\partial }%
830: {\cal B}}{{\partial }a}}{\frac{{\partial }{\cal A}}{{\partial }a^{\ast }}}.
831: \label{osccomm}
832: \end{equation}
833: The Eq.(\ref{hameq}) with (\ref{osccomm}) can be rewritten as
834:
835: \begin{equation}
836: \frac{da}{dt}=-i\frac{\partial {\cal H}/\hbar }{\partial {\bf S}}\cdot \frac{%
837: \partial {\bf S}}{\partial a^{\ast }}=-i\gamma {\bf H}_{{\rm eff}}\cdot
838: \frac{\partial {\bf S}}{\partial a^{\ast }}. \label{aLL}
839: \end{equation}
840:
841: Another form of Holstein-Primakoff transformation can be written as
842:
843: \begin{eqnarray}
844: S^{z} &=&-S-b^{\ast }b, \nonumber \\
845: S^{+} &=&b^{\ast }\,\sqrt{S-S^{z}}, \nonumber \\
846: S^{-} &=&b\,\sqrt{S-S^{z}}. \label{holstein-primakoff2}
847: \end{eqnarray}
848: The classical analog of commutator for $\,b^{\ast }\,$ and $\,b\,$ is
849:
850: \[
851: \lbrack \![{\cal A},{\cal B}]\!]\equiv {\frac{{\partial }{\cal A}}{{\partial
852: b}}}{\frac{{\partial }{\cal B}}{{\partial b}^{\ast }}}-{\frac{{\partial }%
853: {\cal B}}{\partial b}}{\frac{{\partial }{\cal A}}{{\partial b}^{\ast }}}
854: \]
855: and the corresponding equation of motion has the form
856:
857: \begin{equation}
858: \frac{db}{dt}=-i\frac{\partial {\cal H}/\hbar }{\partial {\bf S}}\cdot \frac{%
859: \partial {\bf S}}{\partial b^{\ast }}=-i\gamma {\bf H}_{{\rm eff}}\cdot
860: \frac{\partial {\bf S}}{\partial b^{\ast }}. \label{bLL}
861: \end{equation}
862: It should be noted that Eqs.(\ref{aLL}) and (\ref{bLL}) represent other
863: forms of the Landau-Lifshitz equation (\ref{merm-ll}).
864:
865: \subsection{Real canonical variables}
866:
867: Let us introduce real variables $q$ and $p$ as
868:
869: \[
870: a=\frac{q+ip}{\sqrt{2}},\qquad a^{\ast }=\frac{q-ip}{\sqrt{2}}
871: \]
872: Thus the Eqs.(\ref{holstein-primakoff1}) become
873: \begin{eqnarray}
874: S^{x} &=&q\,\sqrt{(S+S^{z})/2}, \nonumber \\
875: S^{y} &=&p\,\sqrt{(S+S^{z})/2}, \nonumber \\
876: S^{z} &=&S-(q{^{2}+}p{^{2}})/2. \label{6}
877: \end{eqnarray}
878: These formulas are exact and can be considered as a form of classical
879: Holstein-Primakoff transformation.
880:
881: The classical commutator for $q$ and $p$ is:
882: \begin{equation}
883: \lbrack \![{\cal A},{\cal B}]\!]\equiv {i}\left( \frac{\partial {\cal A}}{%
884: \partial q}\frac{\partial {\cal B}}{\partial p}-\frac{\partial {\cal B}}{%
885: \partial q}\frac{\partial {\cal A}}{\partial p}\right) . \label{pq-comm}
886: \end{equation}
887: The equations of motion for real and dimensionless $q$ and $p$ are
888: \begin{eqnarray}
889: \frac{dq}{dt} &=&\frac{\partial {\cal H}/\hbar }{\partial p}=\frac{\partial
890: {\cal H}/\hbar }{\partial {\bf S}}\cdot \frac{\partial {\bf S}}{\partial p},
891: \nonumber \\
892: \frac{dp}{dt} &=&-\,\frac{\partial {\cal H}/{\hbar }}{\partial q}=-\frac{%
893: \partial {\cal H}/\hbar }{\partial {\bf S}}\cdot \frac{\partial {\bf S}}{%
894: \partial q}. \label{eq-qp}
895: \end{eqnarray}
896: We shall use the transformation (\ref{6}) in the case, when $S_{j}^{z}\geq 0$%
897: .
898:
899: In order to exclude stiff equations, for $S_{j}^{z}<0$ it is convenient to
900: introduce other variables $Q$ and $P$ as follows
901:
902: \[
903: b=\frac{Q+iP}{\sqrt{2}},\qquad b^{\ast }=\frac{Q-iP}{\sqrt{2}}
904: \]
905: The equations of motion for $Q$ and $P$ have the form:
906:
907: \begin{eqnarray}
908: \frac{dQ}{d\tau } &=&\frac{\partial {\cal H}/\hbar }{\partial {\bf S}}\cdot
909: \frac{\partial {\bf S}}{\partial P}=\gamma {\bf H}_{{\rm eff}}\cdot \frac{%
910: \partial {\bf S}}{\partial P}, \nonumber \\
911: \frac{dP}{d\tau } &=&-\frac{\partial {\cal H}/\hbar }{\partial {\bf S}}\cdot
912: \frac{\partial {\bf S}}{\partial Q}=-\gamma {\bf H}_{{\rm eff}}\cdot \frac{%
913: \partial {\bf S}}{\partial Q}. \label{eq-QP}
914: \end{eqnarray}
915:
916: \section{Energy transfer}
917:
918: Mathematically the question of energy transfer between ${\cal H}_{{\rm Z}}$,
919: ${\cal H}_{{\rm exch}}$, ${\cal H}_{{\rm anis}}$ and ${\cal H}_{{\rm dd}}$.
920: is associated with the Eq.\ (\ref{hameq}) with ${\cal O}={\cal H}_{{\rm Z}}$%
921: :
922: \begin{equation}
923: i\hbar \,\frac{d{\cal H}_{{\rm Z}}}{dt}=[\![\,{\cal H}_{{\rm Z}},{\cal H}_{%
924: {\rm Z}}+{\cal H}_{{\rm exch}}+{\cal H}_{{\rm anis}}+{\cal H}_{{\rm dd}%
925: }\,]\!] \label{B1}
926: \end{equation}
927: It is obvious that $[\![\,{\cal H}_{{\rm Z}},{\cal H}_{{\rm Z}}\,]\!]=0$.
928: Direct energy transfer from the Zeeman energy to the exchange energy is
929: always forbidden, because
930: \begin{equation}
931: \lbrack \![\,{\cal H}_{{\rm Z}},{\cal H}_{{\rm exch}}\,]\!]=0. \label{B2}
932: \end{equation}
933:
934: The commutator $\,[\![\,{\cal H}_{{\rm Z}},{\cal H}_{{\rm anis}}\,]\!]\,$
935: depends on the orientation of the external magnetic field. If the external
936: magnetic field is parallel to $z$ axis (${\bf H}=(0,0,H)$) then this
937: commutator is equal to zero. However, even a slight deviation of ${\bf H}$
938: from the $z$ axis permits a possibility to transform in sequence the Zeeman
939: energy to the anisotropy energy and the latter to the exchange energy as
940: long as
941: \begin{equation}
942: \lbrack \![\,{\cal H}_{{\rm exch}},{\cal H}_{{\rm anis}}\,]\!]\neq 0.
943: \label{B3}
944: \end{equation}
945:
946: \newpage
947:
948: Captions
949:
950: Fig.1.
951:
952: Model of micromagnetic grain ($4\times 4\times 4$ subgrains).
953:
954: Fig.2.
955:
956: New coordinates.
957:
958: Fig.3.
959:
960: Short (a) and long (b) time evolution of magnetization in the case of very
961: small deviation of reversal magnetic field from the anisotropy axis.
962:
963: Fig.4.
964:
965: Short (a) and long (b) time evolution of magnetization in the case of
966: moderate deviation of reversal magnetic field from the anisotropy axis.
967:
968: Fig.5.
969:
970: Transformations of Zeeman, anisotropy and exchange energies for short (a,c)
971: and long (b,d) time intervals. The dipole interactions are neglected in
972: (a,b) and are taken into account in (c,d). $H_{z}=-0.79H_{{\rm K}}$ is the
973: onset of reversal, $H_{x}/H_{z}=0.05$, $\delta _{{\rm w}}/\delta _{{\rm ex}%
974: }=1.08$.
975:
976: Fig.6.
977:
978: Time evolution of magnetization for various reversal fields at $%
979: H_{x}/H_{z}=0.05$, $\delta _{{\rm w}}/\delta _{{\rm ex}}=1.08$ (dipole
980: interactions are neglected).
981:
982: Fig.7.
983:
984: Average asymptotic longitudinal component and absolute value of
985: magnetization versus applied reverse field. Conditions correspond to Fig.6.
986:
987: Fig.8
988:
989: Time evolution of magnetization for various grain sizes $\delta _{{\rm s}%
990: }/\delta _{{\rm w}}$. $H_{z}/H_{{\rm K}}=-0.79$, $H_{x}/H_{z}=0.05$, $\delta
991: _{{\rm w}}/\delta _{{\rm ex}}=1.08$ (dipole interactions are neglected).
992:
993: Fig.9.
994:
995: Short (a) and long (b) time evolution of magnetization in the case of very
996: small deviation of reversal magnetic field from the anisotropy axis. The
997: strength of reversal field corresponds to the onset of reversal (dipole
998: interactions are taken into account).
999:
1000: Fig.10.
1001:
1002: Short (a) and long (b) time evolution of magnetization in the case of
1003: moderate deviation of reversal magnetic field from the anisotropy axis. The
1004: strength of reversal field corresponds to the onset of reversal (dipole
1005: interactions are taken into account).
1006:
1007: Fig.11.
1008:
1009: Time evolution of magnetization for various reversal fields at $%
1010: H_{x}/H_{z}=0.05$, $\delta _{{\rm w}}/\delta _{{\rm ex}}=1.08$ (dipole
1011: interactions are taken into account).
1012:
1013: Fig.12.
1014:
1015: Relative change in Zeeman energy versus applied reversal magnetic field.
1016:
1017: Fig.13. Time evolution of magnetization for various grain sizes $\delta _{%
1018: {\rm s}}/\delta _{{\rm w}}$. $H_{z}/H_{{\rm K}}=-0.79$, $H_{x}/H_{z}=0.05$, $%
1019: \delta _{{\rm w}}/\delta _{{\rm ex}}=1.08$ (dipole interactions are taken
1020: into account).
1021:
1022: \end{document}
1023: