Nuclear Physics B360 (1991) 425-462 North-Holland

DYNAMICS OF MAGNETIC VORTICES*
N. PAPANICOLAOU and T.N. TOMARAS

Department of Physics, Unicersity of Crete, and Research Center of Crete, 71409 Iraklion, Greece
Received 13 December 1990

The canonical conservation laws of linear and angular momentum in the ferromagnetic continuum have been known to be plagued by certain ambiguities which are resolved in this paper by constructing conservation laws as suitable moments of a topological density. The resulting canonical structure is then shown to be analogous to that encountered in the familiar Hall effect and explains the unusual features of the dynamics of magnetic vortices without resorting to a detailed solution of the underlying nonlinear equations. Thus, in the absence of external magnetic fields, a magnetic vortex is shown to be spontaneously pinned around a fixed guiding center. The guiding center would drift in a direction perpendicular to an applied magnetic field gradient, provided that dissipation can be neglected, with a Hall velocity that is calculated explicitly in terms of the initial configuration of the vortex. In the presence of dissipation, the vortex undergoes skew deflection at an angle ~ ~ 90° with respect to the applied field gradient. The angle 6 is related to the winding number of the vortex according to the well-known golden rule of bubble dynamics.

I. Introduction

Topological solitons arise in a variety of physical problems and have been the subject of much study over the last thirty years or so. Among the best known examples are domain walls and magnetic bubbles in a ferromagnetic continuum, vortices in HeII or in a superconductor, topological defects in liquid crystals, as well as skyrmions and monopoles which are particle-like solutions in generic models of high-energy physics. Perhaps the most unexpected feature of topological solitons is their unusual dynamical response to external probes. Thus magnetic bubbles are notorious for their skew deflection under the influence of a magnetic field gradient [1, 2], while vortices in a superconductor behave in a similar fashion [3]. Furthermore recent theoretical studies in relativistic field theories [4-6] have demonstrated that skyrmions and monopoles would exhibit scattering patterns that are highly unusual from the point of view of familiar scattering processes of ordinary particles, but strongly reminiscent of the skew deflection of magnetic bubbles.
* Dedicated to the memory of our colleagu~:s Basilis Xanthopoulos and Stcphanos Pnevmatikos. I)550-3213/91/$(13.5(I ,c> I ~)91 - Elscviur Science Publishers B.V. (Norlh-Holhmd)

426

N. Papanicolaou, T.N. Tomaras / Magnetic cortices

It is clear that a fundamental simplicity underlies the dynamics of topological solitons, the full extent of which has not yet been determined. The purpose of this article is to present some new results in that direction using magnetic vortices and bubbles as examples. To be sure, the earlier work on magnetic bubbles is already on reasonably solid theoretical grounds. Yet we believe that the earlier theoretical arguments have not been sufficiently appreciated by the wider physics community, probably because they had not succeeded in revealing the full simplicity of the underlying dynamics. Our starting point is the semi-phenomenological observation that the motion of bubbles and vortices in external field gradients closely resembles the familiar Hall motion of electrons in external magnetic and electric fields [7]. In our effort to make this analogy precise, we were led to introduce some radical changes in the canonical definition of the conservation laws of momentum and angular momentum which have been known to be plagued by certain ambiguities. Unambiguous conservation laws are constructed here as suitable moments of a topological current density. The resulting canonical structure is then shown to be completely analogous to that encountered in the Hall effect and explains the gross features of vortex dynamics without great effort. The mathematical basis for the study of the terromagnetic continuum is provided by the Landau-Lifshitz equation which is briefly described in sect. 2 as an introduction to our main task, the construction of unambiguous conservation laws given in sect. 3. Subsequent sections are then devoted to a detailed discussion of the implications for dynamics. Hence the possible types of steady motion in the absence of external fields are analyzed in sect. 4. The effect of an external field gradient is studied in sect. 5, and the effect of dissipation in sect. 6. In the concluding sect. 7 we complete some of the theoretical arguments and contemplate generalizations to other field theories. Finally, for comparison purposes, we summarize in appendix A the canonical structure associated with the usual Hall motion.

2. The Landau-Lifshitz equation
The continuum approximation is appropriate for the study of macroscopic domains in which substantial spatial variations occur only over a large number of lattice spacings. Hence a ferromagnetic medium is characterized by the magnetization M = M(x, t) which is some function of position x and time t and obeys the Landau-Lifshitz equation [8] c~M

=MxF,

M 2= 1,

(2.1)

N. Papanicolaou, T.N. Tomaras / Magnetic cortices

427

which describes precession of the magnetization around the effective magnetic field
F= AM-AM + H.

(2.2)

Here the first term is the exchange field, A being the usual Laplace operator. The second term is an abbreviation for the vector
A M = a ( M l, M z, 0),

a > 0,

(2.3) of the anisotropy field an easy axis, but other term in eq. (2.2) is the satisfies the Maxwell (2.4)

and accounts for uniaxial magnetic anisotropy. This choice corresponds to the most interesting case of a magnet with choices can be treated along similar lines. Finally the third magnetostatic field produced by the magnetization and equations V×H=0,
V.(H+4rrM)

=0,

where time derivatives are neglected because typical velocities occurring in magnetic structures of practical interest are smaller than the velocity of light by several orders of magnitude. Note that the constant magnitude of the magnetization vector M can be set equal to unity, as was already done in eq. (2.1), by a suitable rescaling of the time variable and a corresponding choice of units for the magnetic field. Similarly an exchange constant was scaled out of the first term in eq. (2.2) by a rescaling of the spatial coordinates. One may formally solve the magnetostatic equations (2.4) to write

H= vf (V'M)(x',t) ix=x'l

d3x ' ,

(2.5)

which may then be inserted in eqs. (2.2) and (2.1) to obtain an equation that involves only the magnetization. The long-range nature of the magnetostatic field (2.5) is a technical hindrance but clearly not the only difficulty with this highly nonlinear evolution equation. Nevertheless, analytical solutions have been found in important special cases. For instance, the study of one-dimensional solutions (domain walls) is essentially complete [9] and provides much intuition for the analysis of higher-dimensional structures such as magnetic bubbles [1, 2]. The first and most important step is to find static solutions, which satisfy eq. (2.1) with the time derivative absent. The simplest static solution is the one where the magnetization is uniform, pointing in the positive or negative third direction. One is also interested in static magnetic structures in which the magnetization appreciably deviates from the uniform configuration over a more or less finite

2): w= where We is the exchange energy + w..6b) = fWa d3x. (2.6d) where it is understood that the magnetostatic field is expressed in terms of the magnetization through eq.3(W + Wm). On the other hand. (2.e. + Win. (2.6c) and Wm is the magnetostatic energy 1 W m -- f w m d3x. except where ambiguities may arise. + Win) = 0 . 2 or 3.428 N. in eq. where A is some constant. (2. the energy of the configuration M(x/A). + Wm.8) are . w~ = ½a(M? + M ~ ) . so that their energy be finite.6a) ~ = f we d3x . W = 8rr H2 m .7) Furthermore static solutions are stationary points of W.6b).6 W / 6 M . A = 1 is a stationary point of W(A). O denotes i derivative with respect to the spatial coordinate x i. It is not difficult to show that the effective field of eq. (2. The energy functional will be denoted by W and consists of three terms corresponding to the three terms in the effective field F of eq.N. Thus. provided that the constraint M 2 = 1 is taken into account. (2. is given by W(~. W'(A = 1) = 0. Papanicolaou. Tomaras / Magnetic cortices region. and the usual summation convention for repeated indices is invoked. Thus if we assume that M(x) is a static solution with energy W = We + W.2) is related to the energy by F= . (2. Note that we have used vector notation as much as possible.) = AWe + . with i = 1. T. (2. It is thus important to examine at this point the energy functional associated with the Landau-Lifshitz equation.½(aiM'OiM ) .8) which may be thought of as a virial relation that must be satisfied by any static solution. or W~ + 3(W~. (2. (2. all terms on the left-hand side of eq.5).~. a i. Therefore the existence of nontrivial static solutions is subject to limitations deriving from the well-known scaling argument of Derrick [10]. Wa is the anisotropy energy We. (2. By our hypothesis.

However. (2. M 2 = 1. in the sense that significant variations of the magnetization occur only along the fdm directions. These configurations are predominantly two-dimensional. OM dt = M × AM.5).8) can be satisfied only if W e . Therefore. One should also note that the magnetostatic field is no longer given by eq. leading to the familiar barrelshaped magnetic bubbles [1. H E. (2. x E) and ignore in the definition of energy the trivial integrations over the remaining 3 . Such a disappointing conclusion is averted in practice by the t'mite thickness of the film together with the effect of the magnetostatic field and of an applied bias field. Papanicolaou.8) is found to be (d-2)We+d(W~.N. which implies that any such solution coincides with the uniform configuration except possibly at isolated points. 2].4 r i M 1 . the direction perpendicular to the film. With these provisions in mind. T. H = ( .0 which contradicts the existence of nontrivial static solutions at d .3).1). (2.6) and the choice of the anisotropy field (2. if only the exchange field is retained in eq.Wa .9)yields the virial relation We = Wa + Wm which does not exclude nontrivial static solutions and is indeed verified by the familiar domain walls. (2.I0) is too simple to be useful for realistic bubble calculations. Although the model in eq. However.W m = 0.d dimensions.1. (2. known as magnetic bubbles. the virial relation (2. 0 ) . As a result.5) with the kernel replaced by a logarithm. 0) and is given by an equation analogous to (2. One must conclude that nontrivial static solutions with finite energy do not exist in the three-dimensional ferromagnetic continuum without boundaries. By this we mean to search for static solutions of the form M = M(x~) or M = M ( x i.9) reduces to ( d . for d = 2.1 and d . (2.9) Applied for d = 1. the d-dimensional analog of eq. strictly two-dimensional static vortices may exist if the Landau-Lifshitz equation is modified appropriately. The quasi-two-dimensional nature of the magnetic bubbles leads us to reconsider Derrick's scaling argument from the point of view of lower dimension (d = 1 or 2). it is not entirely without practical interest because it describes the . (2.N. 0 . Hence.2 ) W e .3 but not at d = 2. a rich variety of interesting static solutions emerge.+Wm)=O. (2. for any static solution with finite energy. For example. the virial relation reads Wa + Wm = 0 and again excludes nontrivial static solutions. for d .10) then the virial relation (2. However the magnetization varies to some extent also along the third direction. Therefore. strictly two-dimensional static magnetic bubbles are not possible for the type of interactions considered so far.. eq. whereas for d = 2 the magnetostatic field is of the form H = (H~. Tomaras / Magnetic cortices 429 non-negative. as is evident from eq.

(2. eq. ( 2.14). Tomaras / Magnetic vortices long-wavelength limit of the isotropic Heisenberg ferromagnet. Using the complex coordinates z = x t + / x 2 and 2 = x t .90 i ot +an= 2 0 ( I70. In terms of the complex variable O.10) reads . a large class of static solutions can be constructed by choosing O to be any function of z or 2 alone. 1 M3 = " -00 _ =cosO (2.N.~z _ .o)" (2.15) .13) We further restrict our attention to two-dimensional solutions. (2.12) 1+00 where O is the complex conjugate of 0 . Expressing the exchange energy in current notation we find that w. x 2. O . = f 4(10zl2 + 10:12) d2x (l + .430 N. T.11 ) The cartesian components of the magnetization are then given by O+0 _ =sinOcos 1+OO B M t= 1 O-g2 M2 = i 1 + ~ O = sin O sin ~ .14) . (2. or the stereographic variable O = Mt + iM 2 1 +M 3 i = tan ~O e x p ( i ~ ) . VO) 1 + 00 _ . t).10) could in principle be of some value in a semiclassical study of quantum ferromagnets [13. (2./ x 2. Thus we may use either spherical variables O and • for the parametrization of the unit vector M.o. At any rate. Thanks to the conformal invariance of the static sector of eq.13) becomes i &O 2~0~.e. However most of these solutions have infinite energy.~ + O ~ z = 40t 1 + 00 where subscripts denote differentiation with respect to z and 2. A more transparent form of eq. Papanicolaou. this simple model will often be used in the present work for purposes of illustration.~). 0 = O(x~. (2. i. In other words. solutions of eq. eq.O ( z ) or O = O(. (2. 17]. (2.10) is obtained by explicitly resolving the constraint M 2 = 1.

Then the two-dimensional plane (x ~. ~=a/z. q= (2. T.~. as well as to relativistic nonlinear tr-models in 2 + 1 dimensions [6].. + 1. For static configurations of the form O(z) or . We search for solutions approaching at large distances the uniform configuration which points in. the positive third direction (g~--->0 for Izl--* ~). x2) is topologically equivalent to a sphere. For illustration.14).Q(~.. . say. _+ 2 . respectively. we must recall at this point the important notion of a winding number.MXOuM)'M.16) where the Greek indices /. it is sufficient to secure a finite winding number. the magnetization is shown only around the circle Izi = lal. where its third component vanishes.17) Of course.14) but applies to all variations of the Landau-Lifshitz equation examined in the paper. we quote three elementary examples.1. and (~/~)2. while q will be called the topological density. 1. . ~/Y. (2. 1 and 2. and at z = (i.18) are schematically depicted in fig. viewed here as static magnetic vortices in a 2 + 1-dimensional ferromagnetic continuum. For future reference. Papanicolaou.(O. Q is integer valued (Q = 0. (2.N.. Hence. x 2) covers the sphere M z = 1 as x~ and x 2 vary through the plane is given by 1 Q= 4~r f qd2x' i q= ~e#.u..N. the preceding definition of the winding number is not special to the model (2. The configurations (2. which is achieved with the choice ~Q----"/'/'I(Z)/TFE(Z) o r "/'/'l(Z)/'/'/'2(z) where 1r I and "/T2 are arbitrary polynomials. we note the inequality We >_-4~rlQI which follows from a simple comparison of the energy density in eq.17). These are the well-known Belavin-Polyakov instantons [11]. ) for continuous field configurations and will be referred to as the winding number or topological charge.~g])2 -. take the two values 1 or 2 and eu~ is the two-dimensional antisymmetric tensor. (2. where it points toward the south pole. and M(x ~. The number of times the unit vector M(x~. to obtain solutions with finite energy.. x2) defines a map between the plane and the sphere Mt2= 1.15)with the topological density (2. Returning to the problem at hand. Tomaras / Magnetic cortices 431 In order to sort out finite-energy solutions and understand their significance. Here a is an arbitrary complex constant reflecting in part the conformal invariance of the static sector of eq. The magnetization .e~.~sin O0~O0u~. (2.18) with winding numbers Q = . .) the inequality is saturated. We = 4~'IQI. we also express q as 4(la~l 2 .Ig2:l2) (1 +.

1 and 2. Conservation laws We now begin to address some dynamical questions.d ) AW (2. in accordance with our standard convention.6) terms that contain higher-order spatial derivatives. one should keep in mind that the Belavin-Polyakov instantons are marginal examples of (metastable) magnetic vortices because of the underlying conformal invariance.19) 2)We + d ( W a + Wm) = ( 4 .9) becomes (d- AM + OM )4I • (2. Schematic representation of magnetic vortices with winding numbers Q = . Tomaras / Magnetic cortices O O O O 0 0=-1 Q=I 0=2 Fig. not even at d = 3. 3.(x).432 N. The phase of a was chosen equal to 7r/2. M.20) and does not contradict the existence of nontrivial static solutions at d = 2. points toward the north pole at spatial infinity. such as Jw= f The scaling relation (2. T. Although the resulting pictures are essentially identical to those usually drawn for realistic magnetic bubbles. As a first step we interpret the Landau-Lifshitz equation as a hamiltonian system. (3.1) . Such models have attracted some attention in recent literature [12]. More "realistic" strictly two-dimensional magnetic vortices are obtained by adding to the energy functional (2. with hamiltonian W = IV(M) endowed with the Poisson brackets {M. l.(x)8(x- x').N. Thus we conclude our brief survey of basic facts about the Landau-Lifshitz equation and turn to the main task of this paper.(x')} = - M. Papanicolaou.1.

= f ( 1 .N. 19 --. (3. the constraint playing the role of.c o s t9 ai~ only by a total derivative which is added to ensure proper behavior at infinity where. the momentum integral would be taken to be p. 0. the construction of the associated conservation laws would appear straightforward. the simplest one being to use the standard spherical parametrization. and a gauge condition may be chosen that is also algebraic.7). There are many ways of doing so. Thus we note that the field M 2 commutes with the hamiltonian and may be set equal to any prescribed function of the spatial coordinates. say. a convention that will be adopted from now on.2) and coincides with the general form of the Landau-Lifshitz equation (2. An equivalent form of eq.t. (2. by convention. Indeed the Landau-Lifshitz equation may be viewed as a U(1) gauge theory [13].1) with the effective field F given by eq. For instance. T. To complete the canonical description we must also examine the algebraic constraint present in eq. (2.(x')} (3. the Gauss law in a hamiltonian formulation of electrodynamics. Because the constraint is algebraic. Tomaras / Magnetic vortices 433 The hamilton equation is then given by OM OM = ot 6W +M×--=O ' Ot +{M'W} (3.5) Having brought the Landau-Lifshitz equation to a standard canonical form. where cos 19 is the canonical momentum for the azimuthal angle 4): n=cosO. the specific choice M 2= 1 being the one of physical interest. Papanicolaou. The corresponding expression for the angular .4) is given by (sin O)(b = 6W (50' 6W (sin 0 ) 0 = 6-~" (3. {n(x).H Oil= . This situation is reminiscent of a gauge theory. (3.1)..4) where the dot denotes time derivative.3) The Hamilton equations then assume the usual form 6W qb= ~ H ' 6W ~' H= (3. it can be resolved explicitly to express the three components of the magnetization in terms of two real variables that are canonically conjugate.6) where the momentum density differs from the standard definition .cos O) Oi(~d3x.N.

7) Actually the angular momentum would not be conserved unless we add to it the total magnetization . our argument will be carried out in two steps. .) a3x (3. As a consequence.7). However. A superficial objection is that the momentum density in eq. = f(M .M.8) would be separately conserved. . x i 0 . l = f ( l . (3.. the problem persists at the south pole. these conservation laws are ambiguous as they stand. The ambiguities inherent in the definition of the linear momentum (3. The two-dimensional restrictions of the canonical conservation laws discussed above are given by pu = f ( 1 -.c o s O would suppress the ambiguity at the north pole. While the factor 1 . then the angular momentum (3. only the third components would be conserved. Tomaras / Magnetic cortices momentum would then be given by !. Since the magnetization in a magnetic vortex reaches the south pole at least once. If the magnetostatic field is neglected in the Landau-Lifshitz equation. in the case of a nonvanishing uniaxial anisotropy.. and their use for the analysis of the dynamics of magnetic vortices is dubious [12]. the preceding objections to the conservation laws are crucial for our purposes. m = f (cos 6) . The discussion of the three-dimensional case is relegated to sect. ~ d3x.I. T. 7..7) and the total magnetization (3. = f t 1 - cos O)eii . More serious is the non-differentiability of the azimuthal angle qb at points where the magnetization is directed toward the north or the south pole.6) is not invariant under internal rotations.6) and the angular momentum (3.. Because the topological structure is sensitive to space dimension.1) d ' x (3. but it transforms by an additive total divergence. x . 1).N.cos O ) e . #=lor2. (3.. such an instance is commonplace in gauge theories and would not by itself constitute an ambiguity. The bulk of the paper will be devoted to the study of strictly two-dimensional magnetic vortices where the effects of topology are strongest. 0. Papanicolaou.°. ~..8) Here we subtract the trivial magnetization of the uniform configuration M {°} -(0. d:'x .434 N.9) . A complete resolution of these difficulties is achieved here through a deeper appreciation of the topological structure associated with the Landau-Lifshitz equation. Of course.6) were previously analyzed by Haldane [14] and by Volovik [15] but the offered resolutions do not seem to shed light on the current problem. anomalous ~-function contributions arise which prevent the conservation of the linear momentum (3.COS O ) cguc/)d2x.

while the dots stand for possible additional terms originating in interactions of the form given in eq. (3. We thus initially proceed in a somewhat ad hoe manner. Papanicolaou.( F ...14) wh~re the first three terms correspond to the exchange.6) in eq. T. (3. It would be difficult to convey in a few lines our detailed motivation for the construction of the unambiguous conservation laws given below. A relatively simple calculation shows that O . A stronger version of this conservation law is obtained through a closer examination of the vector L =. reserving a detailed discussion for later stages.10) is already in the form of a local conservation law.M i. Hence we consider the time evolution of the topological density q defined in eq.F" a~M) = ej. (3.N.1) restricted to two dimensions. The various terms in eq.a .e~v(a~.14) are calculated by inserting the explicit expression of the energy density from eq. (3..16).12). (3.c^ + tr..a~M = aatr.~aj. anisotropy and magnetostatic contributions..9) are at least as ambiguous as their three-dimensional analogs.. (3. (3. a ~ M ) = --~.11) where the tensor trva may be expressed in terms of the energy density w according to ova = w6va Eq. a~M). which we denote by l. Although only some general properties of the tensor trva will be important for our argument.15) .12) (3.M'O^M).. (2. so that eq.13) which will prove fundamental for all subsequent calculations. + . We may write ". Tomaras / Magnetic vortices 435 Note that there is only one angular momentum component.( F . it is not difficult to illustrate here its explicit form.1o) for any choice of the effective magnetic field F. a .N.~ = ( w~ + w. The quantities (3.. (2.19). (2. whereas m denotes the third component of the total magnetization.. assuming that the magnetization satisfies the Landau-Lifshitz equation in its general form (2.)6. (3.. (3.(O. Thus we find that cr. The right-hand side of this equation is a total divergence.10) is then written in the form 4 = Ow O(cgaMi) O~.

The lowest nontrivial moments are given by t.16) It should be noted that the portion of the tensor trva displayed in eq.13) is that the integrated topological density. for it may easily lead to overinterpretation. (3. the winding number Q of eq. an elementary application of the divergence theorem shows that /'~.M~. An immediate consequence of eq. (3.13) to which we return promptly. (3. M ' O ~ .19) and their conservation is demonstrated by a simple application of eq. Accordingly the tensor o..N... (3. Tomaras / Magnetic rortices where we and wa are the exchange and anisotropy energy densities of eq. /z = 1 or 2. (3. Papanicolaou. appeared earlier in the work of Thiele [16] where it was interpreted as a force density. However we shall not use this terminology further in the present paper. Instead a number of important conclusions will be derived directly from eq. Further terms in eq. M ) + ~ o t ( M ? + M2) • (3.15) is symmetric. (3. as expected.13): L= fx. Furthermore the appearance of a double derivative on the right-hand side of eq.436 N. (2. = 0. is conserved. (3. For simplicity.tM~)cgvMA.A could be called the stress tensor.14) can be computed in a similar fashion. we demonstrate this point by modeling the nonlocal magnetostatic energy density with the local Hubbard-like interaction Wm= ~(O~.17) (oqt. The vector f.18) which is indeed nonsymmetric. ) .13) suggests that some of the low moments of the topological density are also conserved.Od=x = - = ) + d x. for field configurations with reasonable behavior at infinity. By reasonable . but this property is not shared by the magnetostatic contribution try.20) Since both terms in the integrand of eq. (3. so that (Tv~ = W m ~ v a -- ! 2 (3. (3. (3.6) restricted to two dimensions: ! W e + W a = ~ '( O ~ .16).. (2. T. = fx.q dZx.20) are in the form of a total divergence.

(2.21) together with the canonical Poisson bracket (3.. the two definitions of momentum would be identical. (3. provided that the constraint M 2 = 1 is enforced wherever possible. In both cases we find that {Iv.. (3. M} which we calculate in two ways.Oa~. The third ..22) Therefore the moments I v are essentially the generators of space translations. .a.. It is reasonable to ask whether or not there exists a formal relationship between the two definitions of momentum given in eqs. the non-differentiability of the azimuthal angle • is harmless in eq..M } = -eu.N.9).24) + x. (3. Our next concern is to identify the physical content of the conserved quantities (3.19) the topological density q in the form (2. M } . even if we use the spherical parametrization.O u [ x .1).cos 19)0aq)] .23) is invariant under internal rotations. This form of the momentum integral is free of the ambiguities discussed earlier in connection with eq. (3.23). Using eq. as is apparent from the definition of the topological density in eq. (3. q = eva [(1 . (3.16)... as is assumed throughout this paper. one may derive the awkward identity x .9) and (3.-O~. (3. More precisely.cl. (3.(b ] (3. or = -e v f xvq d2x. T. satisfy the Poisson bracket relation {p~.6) or (3. thus using the topological density in the form q = %.( 1 - cos O)(euaOu~A~).M and must thus be identified with the components of linear momentum. sin 0 a.cos O)euaO..21) because the attending factor sin O would suppress ambiguities at both the north and the south pole.. the quantities p# = -evvI ~. Second by resolving the constraint explicitly. Now the second term can be neglected because it is a total divergence and would not contribute to the momentum integral.. First by using in eq.21).23) with ~t = 1 or 2. (3..N. Papanicolaou. If the second and third term were absent in this equation. = e.16) together with the original Poisson brackets (3.M.t.~. The momentum density in eq. ( 1 .lla.19). Tomaras / Magnetic vortices 437 we mean those field configurations for which the energy is finite.. Hence we examine the Poisson brackets {I~. Furthermore.3). This we achieve by examining the symmetry transformations to which the above conservation laws may correspond..

+ 4~rQc. had it not occurred previously in the problem of electron motion in a uniform magnetic field. for Q . t ) and moments I.26) is used together with the definition of momentum (3. Having accepted eq.23) except for field configurations with vanishing winding number. However. (3.c. for Q ~: 0. At first sight. (3. P2} = 4rrQ.A is contracted with the symmetric one O. Let M = M(x.5). It is not difficult to see that e. Tomaras / Magnetic cortices term also appears to vanish because the antisymmetric tensor e. a property that is clearly violated when eq.18) for which q ~ . calculated with eq. aO.23) as the proper definition of the linear momentum. (3.25) with those of eqs. the momentum integral (3.N.Oa.24) does not vanish.23) plays the role of ordinary momentum. Although the latter problem is much simpler than the Landau-Lifshitz equation. so that the new topological density is given by q ' = q ( x .26) One would normally expect the momentum to remain unchanged under rigid translations. the moments (3. For field configurations with vanishing winding number the Poisson bracket (3. the magnetization reaches the south pole at x = x 0 ~ 0 . (3. Therefore the two definitions of momentum are actually not identical. Papanicolaou. However this term need not vanish thanks to the non-differentiability of q~. (3. (3.23) possesses some rather strange properties. where c is some (integer) constant and ~ = arctan(x2/x l) is the polar angle. (2. so that e.438 N. t) be a field configuration with topological density q = q ( x . T. as is evident from a comparison of the Poisson bracket (3. Therefore. The role of the applied magnetic field B in that problem is played here by the winding number Q. 0. (A. (3.Oa~ = 2rrc~(x) ~ 0. Now imagine a rigid translation of this configuration by a constant vector c = (c~. In general..25) This result would appear to be surprising. it offers a useful guide for a physical interpretation of the current findings.cq~.25) vanishes and the conserved quantity (3. To illustrate this point we use the explicit vortex configurations of eq.24)would still be equal to zero because x~6(x) = 0. t) and the new moments by I'~ = I.19) yield a measure of position more so than momentum. A simple calculation of the Poisson bracket between the two components of momentum yields a nonvanishing result: { Pl. it is more sensible to use as conserved .19).^O. however. Nevertheless the third term in eq.3) and (A. Thus a brief survey of the canonical structure associated with the electron problem is given in appendix A. a number of interesting consequences emerge almost immediately. Oa~=27rc6(x-x o) and the third term in eq. c2).

. (3. (3. x2)-plane.28) which is analogous to the second equation of (A. Papanicolaou.5) in the electron problem and suggests that the point defined by the vector R be called the guiding center of the vortex. the vortex is spontaneously pinned around its fixed guiding center.. . T.^) + ~.~] d2x.. 2¢ ¢ . (3. in complete analogy with the familiar cyclotron motion. which may be shown to satisfy the Poisson bracket relation { I .x.9).N.13): = ..o'.29) (3.. so that translation invariance is indirectly restored. (3. spontaneous pinning of the vortex can occur anywhere in the (x l. (3.27) The vector R transforms under translations as R . Ru= Q ~0" (3. This section is completed with a corresponding discussion of the angular momentum. Of course. Yet the Poisson bracket between the two components of R is nonvanishing: {R. In other words.O. R + e and may thus be interpreted as the mean position of the vortex. A more detailed discussion of the implications of this conservation law for the dynamics of magnetic vortices will be given in later sections.. M } = -e~. .29) is the generator of space rotations and may be identified with the angular momentum. The important fact is that in the absence of external field gradients the vector R is time independent. replacing the naive definition given earlier in eq. Thus we consider the second moment of the topological density: l = ½fx2q d2x.N. R2} = 1/4"n'Q.M. ]qd2 x = 4 r r Q .19) normalized by the winding number: fx~qd2x Iv. ) + a~( x.30) Therefore the quantity in eq.31) . Tomaras / Magnetic vortices 439 quantities the moments (3. It should be interesting to examine the time dependence of 1 defined above using again the basic equation (3. even if its detailed structure would exhibit a nontrivial time evolution.

)2 . we consider the explicit examples 0=(~/2)" or (a/z)".27) vanishes and the guiding center coincides with the point where the magnetization is directed "toward the south pole. (3.17) we find that 4. (3. Otherwisd' the conserved quantity is ! + m. r 2 is time independent irrespectively of the time evolution of the detailed structure of the vortex. For Q ~ 0. 2.12 la i2. In particular. unless the tensor tr. the quantity 1 of eq.33) where n is a positive integer..~ is symmetric so that e. as expected.. As in the case of linear momentum. in which case 1 would be by itself conserved. which are simple generalizations of the elementary examples given earlier in eq.~ is symmetric if we neglect the magnetostatic interaction. (3. A more precise definition of the mean squared radius is given by / ( X-.l) q = + (92" + lal2. where m is the total magnetization in the third direction. (3.32) where R is the conserved guiding center vector of eq. T. Calculating the topological density from eq. in fact. . Papanicolaou. (3. the magnetization at the guiding center need not point toward the south pole. For illustration. Tomaras / Magnetic t'ortices The first two terms in the integrand are in the form of a total divergence and do not contribute to the integral.440 N. 1.R ) 2 q d 2 x r 2 __-/'q 1 -. In fact. for vortices with IQI > l. Eq. (2.~2~.29) may be interpreted as ordinary angular momentum only for field configurations with vanishing winding number. (2.32) provides an objective definition of the vortex radius.18). recalling our discussion following eq. dEx 2rrQ R 2 ' (3.34) the vector R of eq. the magnetization may reach the south pole at more than one point. and correspond to a winding number Q = _+n.34) where Q2 = ~.~=O.27). we return to the Belavin-Polyakov instantons discussed in sect.33).~tr. l is rather a measure of the vortex size. (3.N. A byproduct of the preceding discussion is the curious fact that the "angular momentum" l is expected to be nonvanishing even for a static vortex. (3.z = x~ + x 2 = x 2 is the squared polar radius. the tensor tr. However this coincidence is merely an accident associated with the simple examples (3. and in fig. However the last term may be different from zero. In general. especially when the magnetostatic interaction is neglected. Because of tile axial symmetry of the topological density (3. in that case.14).

the main function of the conservation laws obtained in this section is not the description of static magnetic vortices.1 for radial distances smaller than the vortex radius r. Tomaras / Magnetic cortices 441 Now insert the topological density (3. Just as static solutions may be viewed as stationary points of the energy functional. more general steady-state solutions will be viewed as stationary points of the energy under the further condition that one or more of the additional conserved quantities be assigned a definite value. (3.36)we may also write m = -2rrr 2.38). which is indeed different from zero. that is.N. 4. example of a vortex in steady motion. we calculate the total magnetization in the third direction: m = f ( M 3 . a property not shared by more realistic magnetic vortices. reflecting the slow decay of the fundamental Belavin-Polyakov instanton. (3. but the derivation of useful predictions concerning their dynamical behavior. (3.36) coincides with the naive definition r = [al. Needless to say.2 and 0.2 7r[( ~ r / Q ) c o s e c ( r t / Q ) l l a 12 . In the opposite limit. for future reference.1 takes the values .0 : r 2 = [( r r / Q ) c o s e c ( z r / Q ) ] l a 12 . then M 3 .36) This radius is divergent for IQI = 1.34) in eq. and the value + 1 otherwise. respectively. the radius at which the third component of the magnetization vanishes. (3.35) with Q = + n . and the total magnetization is indeed given by eq. the vortex radius (3. .29) and perform the integrations to find 1 = 27rQ[('n-/Q)cosec(~r/Q)]lal 2 . (3. a task taken up in the remainder of this paper. If we imagine a crude model for the vortex in which M 3 takes the value .. 3. Finally. (3.38) (3. In view of eq. Such a procedure was already used in the early discussions of one-dimensional magnetic solitons [17].37) This formula has a simple geometrical meaning.1) d2x = . perhaps contrived. (3.32) using 1 from eq. IQI ~ 0o. Our next aim is to ascertain the possible types of time-dependent generalizations of the notion of a static vortex that are compatible with the conservation laws derived in sect. The corresponding mean squared radius is calculated from eq. Papanicolaou. Steady motion A static vortex is the simplest. T. (3.35) and R .N.

introduced earlier in eq. The simplest question along these lines is whether or not there exist vortices in linear motion with constant velocity. (3.O~O O q~ ) = ~. V ) and ~ = ~ 0 ( x . (4.1) where the constants v~. v ) of the time-independent equations (4. u (4. we find that 6W ~0' 6W ~. which is in general difficult to achieve analytically. sinO(v. . such as solutions with finite energy and nonvanishing winding number? The general discussion of sect.v t .. Leaving aside for the moment the possibility of solutions with Q = 0. t) = ~0(x .3) The left-hand side may be expressed in terms of the topological density q.2).442 N.N.2) Given a solution O = O o ( X .21).cl) . Papanicolaou. Note that the assumption Q ~ 0 is crucial in this argument. s in O ( O O O. T. However a galilean symmetry is not known to exist for the Landau-Lifshitz equation. Therefore. we would like to confirm the above general conclusion directly from eqs. using the explicit expression for the momentum from eqs. with /~ = 1 and 2. v ) and would describe a configuration in rigid linear motion with velocity v. (4.. whereas the right-hand side is but another form for the vector -f.v t .2). v ) and ~ ( x . except for very special onedimensional models [13]. play the role of Lagrange multipliers enforcing a definite value for the momentum. 3 already implies that such solutions do not exist.23) and (3. R ( t ) = R(O) + vt. Thus we seek after stationary points of the extended energy functional U=W-vr.v t . which would contradict the conservation of R unless the velocity vanish.2). Using both of these equations we may write (SW ~SW ) . p~. t ) = Oo(X . v ) and a time-dependent guiding center. But are there any interesting solutions of eqs. (4. for they would yield a topological density of the form q = q ( x .. Tomaras / Magnetic vortices It should be stressed from the outset that the actual construction of steady state solutions still rests with the explicit solution of highly nonlinear equations.~ 0~0 + --6-~Ofl) . the results of this section may be thought of as a qualitative theory of steady motion which may serve as a guide for numerical solutions. One would normally answer such a question by applying a galilean boost to a static vortex. Taking the variations of U with respect to the independent variables (9 and ~. for the most part.O~)= -sinO(v~O.5) would be obtained from O ( x . Nevertheless the question can be asked anyway by exploiting the momentum integral. a solution of the original time-dependent equations (3. v ~.O)= (4.

for configurations with finite energy and thus reasonable behavior at infinity. Using the relation f~ = 0 ~ from eq.6) is equivalent to the virial relation obtained by applying Derrick's scaling argument to the functional U of eq.t' ~ = 0. finite energy. (4. the right-hand side yields zero. (4.16) and its trace by tr tr = a(M~ 2 + M 2) = a sin 20. .3) may be written as (4..5) or the general discussion of sect.4) qev~.14). in agreement with our earlier conclusion. 3. (4.0 are not contradicted by either eq.~tr. (4.N.. Papanicolaou. If we include only the exchange and anisotropy contributions. (4. Then the virial relation (4.11). The left-hand side yields e~c~[x~qd2x = v ' p . the tensor is given by eqs. (3.N.af sin 2 19 d2x. (4.4) does imply certain conditions for the existence of such solutions. while the left-hand side is expressed in terms of the winding number Q: 4~rQev.v# = -a.2) with Q = 0. The absence of a contribution to the trace from the exchange interaction is a reflection of the conformal invariance of the static sector of the pure exchange model. 0. Tomaras / Magnetic vortices 443 (3. eq. and Q = 0. (4. for Q ~ 0.6) takes the explicit form v p.11). Actually eq. and finite momentum. (3. (4.23).5) Hence. and the right-hand side gives Employing the usual notation tr. .. Integrating both sides over all space.7) Interestingly this relation contradicts the existence of nontriviai static solutions (v = 0) but leaves open the possibility of solutions with t .~. (3.15) and (3.1).V-. the resulting relation is v-p=ftrtrd2x. eq. where we used the definition of the momentum from eq.tr tr for the trace of the tensor. Nevertheless finite-energy solutions with nonvanishing velocity and Q . (4. To obtain a more explicit relation one must specify the types of interaction included in the tensor o-~a of eq. On the other hand. We contract both sides of this equation with x~ and then integrate over all space. finite-energy solutions with nonvanishing velocity do not exist.6) which is a virial relation satisfied by all solutions of eqs. T. (3.

we search for stationary points of U.0. the above frequencies must be chosen equal (to = to'). the angular momentum 1 of eq. x2)-plane. (4.10) for which eqs. [18] was apparently rediscovered in ref. p~ = 0.9) where the constants to and to' are the frequencies of rotation and precession. p = 0.to'm. we have already established that in the absence of external field gradients the guiding center of a nontrivial (Q :~ 0) vortex is spontaneously pinned at some arbitrary point in the (x l.M=M×AM. which can be shown to satisfy the . while the magnetization at each point precesses. (4. Nevertheless some progress can be made in the case of the pure exchange model of eq. [19] where an explicit one-soliton solution was also given. As before. at some specified rate. (2. (4. The nonlinear system (4. M 2= 1.444 N. Although the details of that solution are essentially two-dimensional. except possibly for Q -.8) where we have returned to vector notation and A is the two-dimensional laplacian.29) and the total magnetization m of eq.O. Returning to the general problem.8) with finite energy and Q = 0.N. which now takes the degenerate form v . We may thus envisage a steady state in which the vortex rotates as a whole around the fixed guiding center. This is surely a stringent condition from the physical point of view. (4. while angular-momentum conservation yields e.8) do not exist.t'. in the sense that it possesses a Lax pair and an associated dual symmetry. (4. its overall structure is that of a one-dimensional domain wail. However such solutions were never accomplished and the program was eventually abandoned. Such a state would be the analog of the familiar cyclotron motion of an electron in a uniform magnetic field. but it still does not seem to mathematically contradict the existence of solutions of eq. Even in the latter case.~. (3.8) was shown by one of the present authors to be completely integrable [18]. Therefore the linear momentum must vanish for any value of the velocity. The Lax pair of ref. and its energy is infinite. If the magnetostatic interaction is included. Tomaras / Magnetic cortices The construction of explicit solutions is still too difficult a problem to tackle analytically.tol .7). T. respectively. It was then hoped that one could analytically obtain examples of Belavin-Polyakov instantons propagating with nonvanishing velocity. Therefore the only allowed time-dependent generalizations of a static vortex are those that preserve the location of the guiding center and are consistent with the remaining conservation laws.2) are equivalent to -c. The present analysis makes it clear that finite-energy solutions of eq. a severe restriction results from the virial relation (4. One may again consider an extended energy functional of the form U = W . so that only the sum l + m is conserved. namely. (3.9). Papanicolaou.

(4. to. For Q 4= 0.q~ .14) where q = q ( x .Zo). the first moments of the topological density must vanish. t) = ' I ' 0 ( ~ ( x .5)would be obtained from o ( x . (4.10): tox. . q. for solutions with nonvanishing angular frequency.1 ) = -Oao'~a. a. " 3 ..x 0 in eq.14) confirms that the vortex rotates around its fixed guiding center.14). Papanicolaou.o.s i n tot sintot) cos tot ' (4. so the origin of coordinates in these equations has an objective meaning. or the point x . Therefore.11) where ~ denotes symbolically the time-dependent rotation matrix ~q~=( costot . a solution of the original time-dependent eq. (4.OvO = $ W / 6 q ~ .10) are not translation invariant.4) starting from eq. (4. to') and q~ = ~0(x.11) through the appearance of the arbitrary position vector x0.14) is equivalent to stating that the linear momentum vanishes. In all cases. . t) = O o ( ~ ' ( x . (4. Here x o is an arbitrary constant vector reflecting the translation invariance of the original equations.( x . as expected on consistency grounds.03 + o. First we derive the analog of eq. the magnetization precesses with constant frequency to' which does not explicitly enter eq. - to sin 19e~.. coincides with the guiding center of the vortex.N. (3. .0. as is evident in eq.N.10). (4.). For Q = 0. eq. the point x = 0 in eq.to') = 8 W / ~ 1 9 .12) acting on the column vector x . However the reduced equations (4. x2)'plane.'t. As we shall see shortly.13) An immediate consequence of this relation is that tof xvq d2x = O.10) Given a solution 19 = 190(x. Tomaras / Magnetic cortices 445 differential equations sin 19(toe~. (4. to. i. ..q- to'O~(cos 1 9 . to. The physical interpretation of this result depends on whether or not the winding number is different from zero.11).x o. (4. The latter may be pinned anywhere in the (Xl. it is the point around which the vortex rotates with angular frequency to.. (4. (4. to') of these equations. eq. T. (4.~x~... (4.x~.Xo).e..

for essentially all variations of the Landau-Lifshitz equation considered in this paper. the fixed guiding center need not coincide with the point where the magnetization reaches the south pole. and total magnetization. as well as Q ~ 0. (4.O. having gone as far as possible without actually solving the nonlinear differential equations (4. (4.9). and by integrating over all space.10).6). Papanicolaou. further progress may rely on numerical solutions based on the qualitative picture presented in this section.9). we find that tol + to'm = ½f tr o. obtained by contracting both sides with x~. such as 1(0) ~ 0. This virial relation does not contradict the existence of solutions with finite energy. not even for the simple exchange model. The question is then to predict the behavior of the vortex after the field is imposed. Applying partial integration where appropriate and recalling the definitions of the angular momentum 1 from eq. To complete the qualitative description of rotating solutions one must recall the observations made at the end of sect.13) also implies an analog of the virial relation (4. the angular momentum is a measure of the vortex radius more so than a measure of how fast the vortex rotates.N. the applied field points in the third direction and its magnitude h may depend on position as well as on time. As a consequence.1) In words.15) which is again equivalent to the relation obtained by applying Derrick's scaling argument to the functional U of eq. (3. T. 3. These observations conclude our analysis of steady motion. the motion of a vortex resembles the familiar Hall motion.29) and the total magnetization m from eq. Motion in a field gradient The most characteristic features of vortex dynamics become apparent in the presence of an external magnetic field gradient. First. Although analytical solutions do not seem feasible.h(x. 5. Such a field plays the role of an electric field in the analogous electron problem.d2x. while the winding number Q continues to play the role of an applied uniform magnetic field. (4. Specifically we consider the following physical situation. for Q ~: 0. Second. angular momentum. As a result.446 N. We shall find that in the absence of . A vortex initially pinned at some point is subject to an external magnetic field of the form Hext=[O. (3. the dispersion ! = l(to) is expected to exhibit unusual behavior. Tomaras / Magnetic t'oru'ces Eq. just as in the ordinary cyclotron motion the guiding center does not coincide with the actual position of the electron.t)]. (5.

Yet studying the degree to which these conservation laws are violated by the applied field yields direct information on the behavior of the vortex. However the moments of the topological density are no longer conserved. the winding number Q. so that the integrated topological density.1) v in order to prepare a second partial integration.2) Here tr~ is the tensor constructed in sect.O f x.3) may be written as [.2) is still in the form of a total divergence.1). Thus we consider the time evolution of the moments I~. T. is still conserved. a situation made more precise below.5) is not sufficiently explicit to provide a detailed description of the time evolution of the moments.N.19): z.10) to obtain il = e~ag( F. in the absence of the applied field.5) which is now legitimate because M 3 ~ 1. Our starting point is again the fundamental relation (3. eq. To calculate the effect of this field on eq. (5. For a general applied field h .13). = f x. (3. in accordance with our standard convention. (3.( ha M ) d x. 1)d x. (5. (5. (5. (3. (5. fha M3 d2x = -e. (5.13) extended to include the applied field (5. t ) the right-hand side of eq. we just substitute F ~ F + Hext in eq. The right-hand side of eq.3). (5. By partial integration.~a~( Hex t "OvM) = + (5. This situation becomes corn- .4) In the second step of this equation we inserted the trivial identity O M 3 = 0v(M3 . but it already indicates that the moments would tend to vary mostly in the direction perpendicular to the field gradient.N.3) where we use the fact that the first term in the right-hand side of eq. fhO (M3.h ( x .a. Tomaras / Magnetic vortices 447 dissipation the vortex would drift in a direction perpendicular to the field gradient Vh. at large distances. (5. 3. = -e.2) does not contribute in eq. of eq..O~M) + et. and M 3 is the third component of the magnetization. reflecting the conservation of the moments in the absence of the applied field. Papanicolaou.1)d2x.

(5. or Hext = (0. Then eq. with "Hall velocity" VH =/~2.g m / 4 ~ r Q . constant g being an admissible special case. a completely explicit calculation of the time dependence of the right-hand side of eq. so that m would pick up some time dependence during the evolution of the vortex under the influence of the field gradient. i2= -gf(M3.8) These relations stipulate that the vortex will drift as a whole in the direction perpendicular to the field gradient.6) because the latter points in the third direction.448 N.7) is not possible without a detailed solution of the initial value problem.7) is that only the component of the moment in the direction perpendicular to the field gradient varies with time.g m / 4 ~ r Q . only the sum l + m is conserved. discussed in sect. m would be conserved even in the presence of an external field of the form (5. In that case. /~2 = .9) It should be emphasized that the above general result does not depend on the finer details of the time evolution of the vortex configuration.7) provides a completely explicit description of the time evolution of the moments. We shall distinguish two cases depending on whether or not the winding number Q vanishes. Papanicolaou.7) where m is the total magnetization in the third direction. We discuss first the case Q 4:0 where the moments provide a measure of the mean position of the vortex through the guiding center coordinates defined in eq. T. which should still be determined through an explicit solution of the initial value problem. (5. eqs. (5.5)yields i. . For a physical interpretation of the preceding results. or VH = . we must recall the physical content of the moments I~. 3. Tomaras / Magnetic vortices pletely transparent in the case of the simple field h =gx~. (5. (5.27). (5.=o. If we neglect the magnetostatic interaction.1)d2x = -gm. g x l ) . (5. Furthermore the right-hand side of the second equation in eq. (5. (5.7) may be immediately translated in terms of the guiding center coordinates: /~l = 0 . In particular. (3. In the presence of the magnetostatic interaction.N. In such a case. The first important consequence of eq. g=g(t). (5.7) is expressed entirely in terms of the field gradient and the total magnetization m of the vortex in question. Since the winding number remains conserved in the presence of the applied field.0. m can be computed from the initial configuration of the vortex and eq.6) where the gradient g is space independent but may still depend on time.

Note that we currently use the symbol r 0. T.29).. (3. instead of r. The above result also does not depend on whether the initial configuration is a static vortex or a vortex in a steady state of the type studied earlier in sect. This is just one example of the finer details of the vortex that might undergo a nontrivial time evolution. Thus. Because the magnetostatic interaction is absent in this model. If we again restrict our attention to the simple gradient of eq. the guiding center of the vortex would drift in a direction perpendicular to the field gradient. as was already done in eqs.5).h)( M 3 - 1) d2x. the total magnetization m is conserved and may be computed from the initial configuration. (5. 4. (5. the vortex radius r 0.11a)with h -gx~. we find that [ = f (e~. Unfortunately the right-hand side of eq.11b) where R 2 = R2(t) is the second component of the guiding center whose time evolution is governed by eq.38). In fact. If a magnetic field of the form (5.N. even if the details of the vortex undergo a complicated time evolution.37) and (3. and the winding number Q. (5. with velocity given by eq. (5. we may recall its definition (3.6). but such a . We thus find that citd (rE) = 21rQgf(Rz-x2)(M3-.9). The time derivative of l in the presence of the field gradient may again be computed from eq. (5.32) is inferred by using eq.32) in terms of the angular momentum l of eq. in the absence of dissipation.2).10) which is a completely explicit result for the Hall velocity expressed in terms of the field gradient g. 1) d2x. Papanicolaou. (5. (5. Tomaras / Magnetic cortices 449 this result is valid even if the vortex does not reach a steady state.9).8). To get a glimpse of the time evolution of the vortex radius r.14). especially when dissipation is absent. (3.. (3. Hence we find that VH= 2--Q' r2= [(rr/Q)cosec(rt/Q)llal 2. (5. their guiding center will move at a right angle to the field gradient with velocity given by eq. (5.x~. Neglecting the magnetostatic interaction.1 la) which is the analog of eq. for the vortex radius in order to emphasize that r 0 is the initial value of the radius which may subsequently change under the influence of the field gradient.6) is applied to these vortices.33) which are static solutions of the simple exchange model of eq. a steady state is not expected to occur without dissipation which is assumed to be absent in the calculations of this section. we return again to the Belavin-Polyakov instantons of eq.O. (2.N. (5. (3. For an explicit illustration.11b) is not sufficiently explicit to allow substantial analytical progress. (5. the time derivative of the mean squared radius r 2 defined in eq.

In this respect. (3. (3. the fact that the guiding center is a sensible definition of the mean position of the vortex was already established in sect. is atypical because it lacks an intrinsic scale. the moments I~. also in analogy with the electron motion in a uniform electric field and an arbitrarily weak uniform magnetic field. or p~. 3 following eq. (5.2].1) d2x. by analogy with Hall electrons which. namely. One may push this analogy further to conjecture that vortices with large Q. Nevertheless this result suggests that the vortex radius decreases with increasing Q.1) may again be inferred from eq.23). the positive (negative) values of the topological density become dominant.e u ~ l V. in the adiabatic limit. often used in this paper for explicit illustrations. Papanicolaou. There remains to examine the extreme nonadiabatic limit Q = 0. without working out explicit solutions. (3.450 N. increasingly so for increasing Q.10) by the arbitrary scale lal. On the other hand. 2. that the trajectory of the guiding center is a faithful description of the motion of the vortex as a whole. (3.N. Instead the moments Iu at Q = 0 may be used to calculate the linear momentum according to eq. compared to some fundamental scale in each model. (5. Tomaras / Magnetic cortices relation may prove useful for approximate estimates of the time evolution of the vortex radius. a fact made apparent in the expression for the vortex radius in eq. especially for Q = 0 when it integrates to zero.32). The choice of the magnetic field made in eq. and q becomes a bonafide measure of the vortex position. (5. However it is difficult to decide a priori whether or not the vortex radius is small. T. Furthermore a measure of the spread of the vortex around its guiding center is provided by the mean squared radius of eq.the so-called adiabatic limit.5): = f (o. even for small nonvanishing Q. for Q 4: 0. which may be called hard vortices [1.26) applied for Q = 0. In a generic model. (5. fail to provide an estimate of the mean position of the vortex. just as the guiding center trajectory in the Hall motion becomes an accurate description of the actual electron trajectory in the limit of a large applied magnetic field . Such a property is completely lost for Q = 0. Therefore the time evolution of the linear momentum of a Q = 0 vortex subject to an external field of the form (5. one would expect that a particle like description of vortices in terms of the guiding center would become increasingly faithful with increasing winding number. namely the xE-axis. Of course. the simple exchange model.26). To conclude our discussion of vortices with Q 4= 0. drift along lines of constant electrostatic potential. as is evident from eq. we must comment on a tacit assumption made throughout this section.12) . subject to a general field h = h(x) would drift along lines of constant h.h )( g3 . A vortex with vanishing winding number is shown in fig. A related fact is that the topological density q is not positive (negative) definite. In this case.6) is special in that the guiding center follows a line of constant h. = .

Hence the deflection angle for realistic mag- . Nevertheless the above picture does not contradict the experimental fact that Q = 0 magnetic bubbles move in the direction of the field gradient. in this case. just as the Hall conductivity acquires a longitudinal component in the presence of dissipation. the velocity of the vortex would develop a nonvanishin~ component in the direction of the field gradient. strictly speaking. unlike the case of ordinary particle dynamics. 2. Tomaras / Magnetic cortices 451 ® O--O Fig. However. dissipation will be shown to be responsible for the experimentally observed Q-dependent deviations of the deflection angle from the theoretical value of 90 ° derived in this section. /~2=0. The effect of dissipation The deflection of magnetic vortices with Q ~: 0 in a direction perpendicular to the field gradient predicted in the previous section is analogous to the absence of longitudinal Hall conductivity in the absence of dissipation. the relation between momentum and velocity is not a priori known in the present problem. this result does not necessarily imply that a Q = 0 vortex would actually move in the direction of the field gradient. T. for. 6. Furthermore.N.N. Papanicolaou. a situation that will be clarified after including dissipation. 6. in sect. (5.6) we find that /~! = g m . Schematic representation of a magnetic vortex with vanishing winding number. On the other hand. for magnetic bubbles with Q ~: 0. in analogy with the momentum of an electron which would vary only in the direction of the electric field if the magnetic field were absent. the momentum of a Q = 0 vortex would vary only in the direction of the field gradient. where the right-hand side may be interpreted as the force acting on the vortex due to the field gradient. If we further restrict our attention to the simple gradient (5.13) Hence.

452 N. In a phenomenological description of dissipation.=--f[(MxF). T T2 = ~ I+T 2' (6.1) may also be written as OM Ot = M x G.N. are not conserved.f(F. (6. M" 1 (6.10) with the simple substitution F --->G: q = ~u~au(G • J~M). (6. Thus we find that t =fx qd x- d2x.a. (3.2) and is similar in form to the original equation (2.M]d2x. (6.4) where we may insert the explicit expression of G from eq.a. G=T~F+T2(MXF). Tomaras / Magnetic cortices netic vortices is expected to be different from 90 °. Eq. Because of the formal analogy of eq. (6.5) is assumed to contain an . would vanish in the absence of an external field gradient. D. To avoid introducing a complicated notation.M × F . 1 Yl= I + T - . as is discussed more precisely in the remainder of this section.1) is modified according to OM Ot TM× OM Ot . (6.. the time derivative of the topological density q may be inferred from eq. (2.1).3) The winding number Q continues to be conserved in the presence of dissipation.1) with the field G playing the role of the effective magnetic field F. (6. the Landau-Lifshitz equation (2.5) This equation has been arranged so that the left-hand side coincides with the time derivative of the linear momentum.-. but the field F in eq.1) where y is called the dissipation constant. Papanicolaou.1). (6. M 2 = 1. the field F in eq. but the moments I.2)will be assumed to contain an external magnetic field of the form (5.2) to obtain c. The vector C~.2) with eq.M)d2x. (6. T.

. In fact. we return to eq..M ( x . (6.r. (5.6)we find that C ! =gm.6) which is the force due to the external field gradient discussed earlier in eq.10) is called the dissipation tensor [16]. a prediction that is now modified by the appearance of the vector Du in eq.h)(M3 .O + sin 20oJ)o~. .o0.N. is not sufficiently explicit to allow for further analytical progress.5). (6. C2=0.N. = 4~-Qc.r..7) The force C~. constructed above is responsible for the 90 ° deflection of magnetic vortices discussed in sect..f ( H e x t ' a ~ M ) d 2 x = f (O. + yd..9) =-f (O. v).1) and assume that the steady-state magnetization is of the form M . it is not known whether or not a simple gradient of the form (5.t. (6. which is used in eq. T. (6. Thiele [16] seems to suggest that the field gradient may have to exhibit a carefully selected time dependence. Tomaras / Magnetic cortices 453 external field of the form (5. = 0 . (6. (6.vt...(O. We shall not examine here conditions on the form of the external field that would guarantee the development of a steady state.M.. (6.. Hence we find that C z = .. With these unsettling comments in mind.O . so that -v~. If we further restrict the external field to the form (5. M and integrating over all space we find that D u = du.M-TMXO~.t:~.8) Contracting both sides of this equation with .11) .5) together with eq. where (6.6) would actually lead to a steady state. Papanicolaou.1). (6. 5. in particular. without further assumptions. * C. 5. For a steady-state configuration of the above form we may further set [.O. matching that of sect.12).M)d2x = f (o. Unfortunately the expression for the vector D.9) to obtain the algebraic relation 4zrQe.1) d2x. But the essential point of this argument can be made plausible by assuming that in the presence of a field gradient as well as dissipation a steady state is eventually reached in which the vortex moves rigidly with constant velocity v.M) = M X F .y(47rQ)e~.~)d2x (6.'.

4rrQ -. it is described in its rest frame by spherical variables of the form O = O(~}) and tp = const.g m ) . (6. (6. Then the elements of the dissipation tensor (6.2 ~ r .15) and the deflection angle by V2 sin ~ = . Relation (6.g i n ) . dl I = d22 = ~(dll + d22 ) 1 - We.10) reduce to di2 = O = d 2 ! . which may be viewed here as reasonable approximations. as is where we may further substitute the total magnetization tn = .16) may also be written as gm 4rrV sin 8. where ~ and cp are the usual polar variables defined from x~ =~} cos ~p and x 2 = ~ sin ~p.N.-V 4~'QV gm = .7) and the linear system (6. (6. For the special choice of the external magnetic field (5.12) A simpler version of these relations is obtained by invoking some further assumptions about the steady state profile of the vortex.12) becomes y We Cl = y2We2 + (4~'Q) 2 ( .g i n ) . x cp. T.Ydl2 v2 = y2 det d + (4~rQ) 2 ( .454 N Papanicolaou. (6.e.g i n ) . The speed of the vortex is then given by V = (u21 -l-v2)1/2= 4rrQ v 2 = y2We2 + (47rQ) 2 ( . is given by eq.13) where We is the exchange energy of the vortex. and eq.17) 2. the vector C~. Q= (6. i. Tomaras / Magnetic cortices Applying this relation for/z = 1 and 2 yields a system of linear equations for the components of the velocity t. (6. (6. Thus we assume that the vortex retains approximately its axial symmetry.14) Igml [y2We2 + (4rrQ)2] 1 / 2 ' (6. t and v 2.6).16) Here 8 is the angle between the trajectory of the vortex and the direction of the field gradient.11) is solved to yield yd22 t'! = y2 det d + (47rQ) 2 ( .

Indeed a natural extension of the conservation laws of sect. . Papanicolaou. Thus we see that the vortex moves in the direction of the field gradient for Q . an equivalent form of eq. (6. Thus we consider the current density ' J.18) might prove possible only in the adiabatic (large-Q) limit. (3.N. the vortex radius r. (6. Actually it is more or less evident that these assumptions are expected to be valid only in some approximate sense. (5. (7. the magnetic field gradient g. (6. T. Tomaras / Magnetic vortices 455 suggested by eq. and the deflection angle 6. In the latter case. the speed of the vortex is always smaller than the absolute value of its characteristic Hall velocity. but it undergoes skew deflection for Q ~ 0. in the presence of dissipation. In particular. 3 exists in three dimensions and is briefly described in the following. Concluding remarks We have thus completed our discussion of the two-dimensional ferromagnetic continuum and now turn our attention to some more general issues.9). (6.1) . Such solitons may occur under certain conditions. 7. Some steps in this direction were taken recently by Maslov and Chetverikov [20]. Therefore.~jM) "M.17) is given by V = Vn sin 6. to obtain the golden rule of bubble dynamics [1.19) where VH is the Hall velocity defined earlier in eq. In conclusion. we must emphasize that some of the arguments of this section are based on plausible assumptions whose validity remains to be justified.18) might also prove feasible.jk(OkM × . at least in principle [21]. the answer of this question could prove to be of some practical significance in the study of threedimensional topological solitons which are characterized by a topological charge of a different nature. namely the Hopf invariant. we address the question of constructing conservation laws for the three-dimensional theory which are natural generalizations of their two-dimensional analogs derived in sect.N. 2] gr 2 Q = 2---Vsin 6. namely. Hence a rigorous justification of the golden rule (6.~e. One would then envisage an adiabatic perturbation theory where the calculation of nonadiabatic corrections to eq.38) and the ensuing discussion.18) which expresses the integer valued winding number Q in terms of directly measurable quantities. but a complete adiabatic theory is still lacking.0. Aside from its obvious academic interest. 3. the vortex speed V.

namely 1 Qi= 4~r fJi d3x (7.4) {I~j. This is already an indication that nothing special happens in the dynamics of a generic three-dimensional situation.16).M} =%. (2.. The same relation implies that the three independent elements of Iij may be identified with the three components of the linear momentum according to P i = --123. It is then not difficult to show that the analog of eq. with i = l.13) reads ~. Nevertheless we may proceed examining further consequences of eq.. (7.16). However. Tomaras / Magnetic vortices where Latin indices i.J d3x are conserved.2) where trkt is a straightforward three-dimensional extension of the tensor constructed earlier in eq.5) which makes it apparent that the tensor li~ is antisymmetric. (7. Thus we find that tile moments f .456 N.M. (7. while they satisfy the Poisson bracket relation (7. which may be thought of as the analogs of the winding number Q of eq. An immediate consequence of eq.7) . unlike the winding number in two dimensions. the Qi are trivial in the three-dimensional ferromagnetic continuum without boundaries.O.6) Similarly the three components of the angular momentum can be shown to be given by li = ~ fx2Ji d3x. The current density J~ is a simple generalization of the topological density q of eq. (2. (7. i = 1. (3. in the sense that they are either divergent or vanishing.12). Papanicolaou. P3 = --I12" (7.2) is that the integrated current density. lij = -lyi.3) leads to three conserved charges. take over three distinct values and eij k is the usual antisymmetric tensor.N. P2 = --131. T. In particular. (3. 2 or 3. = --e~skaSal~rkl. j. the Qi vanish for all field configurations with finite exchange energy. (7.2 or 3..2).

23] 1 f ( X . x 3 playing the role of a parameter characterizing the specific field configuration. The picture changes drastically in the case of the quasi-two-dimensional ferromagnetic continuum. consider the charge Q3 which we may write as Q3 f. Although the density J3 depends in general on x 3. One should also note that Derrick's scaling argument yields severe restrictions on the existence of finite-energy static solutions in three dimensions.lr--------~ jeijkJi(x) ix_x. the moments (7. one should not conclude from the above general argument that the third direction does not play any role in the dynamics of magnetic bubbles. while the charges Qi vanish for a Hopf configuration with finite exchange energy.9) yields Q 3 . Of course.x')j Jk(X') d3x d3x ' QH = 16. a case of special practical interest because of the magnetic bubbles known to occur in thin magnetic films. 2.N..4) are invariant under rigid translations and the Poisson brackets among the components of the linear momentum (7. The latter may also be expressed in terms of the current density Ji according to [22. One can further show that Q3 is bounded by the exchange energy according to We >/47rlQal. where the x3-axis is taken to be perpendicular to the film. Therefore. Hence there is no evidence that such topological solitons would exhibit unusual dynamical behavior.. For instance. 3 and now comment on their possible implications for dynamics.. as long as the film thickness remains finite. This observation applies. (7. it can still be interpreted as the two-dimensional topological density q of eq.8) However the topological charge QH does not enter directly the construction and immediate properties of the conservation laws. the gross features of the dynamics of quasi-two-dimensional magnetic bubbles are similar to those of strictly two-dimensional magnetic vortices. +_ 1. (2. finite-energy solutions may exist in which Q3 "-QL :/: O.[3 (7.16). which does not depend on x3. As a result.N. but a detailed discussion of these finer points will not be given in this paper.QL where L is the thickness of the fiim. (7. As mentioned already.9) is still integer valued.. Therefore the three-dimensional dynamics does not appear special in any way. taking a value Q = 0.3) implies that the above conservation laws do not display unusual behavior. 79.6)vanish. Thus the double integral in eq. dx3 ~ [1 ! 1. (7. to topological solitons characterized by a nontrivial value of the Hopf invariant. . Tomaras / Magneticcortices 457 We have thus obtained natural three-dimensional extensions of the conservation laws of sect. Papanicolaou. which is a simple generalization of a similar inequality discussed in sect. the vanishing of charges Qi of eq. Then the charges Qi need not vanish for field configurations with finite energy. As an example. but such solutions could be stabilized by precession of the magnetization [21]. T. in particular. and eq.

10) is the familiar expression for the Berry phase.16). we briefly describe here the canonical structure associated with two-dimensional electron motion in a uniform magnetic field B perper~dicular to the plane and an in-plane . We shall be content to notc here that the dynamics of domain walls is completely understood through an explicit solution of the equation of motion [9]. However. T. This case appears singular because the current density Ji of eq. (2. We thus return to the canonical definition of the linear momentum p= _ (l-cosO)~xdX (7. Appendix A TWO-DIMENSIONAL ELECTRON MOTION IN A UNIFORM MAGNETIC FIELD In order to illustrate the analogy with the Hall effect more explicitly. 870009). rather than as a result of the detailed form of the equation of motion.10) and discuss first the case of pulse solitons where the magnetization reaches the north pole at both ends of the line. We are grateful to Victor Bar'yakhtar and Boris lvanov for conversations emphasizing the phenomenological analogy of this problem with the Hall effect. This work was supported in part by two grants from the EEC (Science SC1"-0250-C and Esprit 3041) and by a NATO travel grant (No. 1. (7. Then the integral (7. namely. we believe to have presented a more or less complete framework for the study of the dynamics in the ferromagnetic continuum.458 N. noting that the integral (7. We thus expect that the methodology of this paper will prove useful also in the discussion of the related topics mentioned in sect. A detailed discussion of this case may be found in the article of Haldane [ 14]. vanishes when restricted to one dimension.1). The situation is less clear in the case of domain walls where the magnetization reaches the north (south) pole at x = + oo and the south (north) pole at x = -0o. the solid angle traced by the magnetization as x varies from to + ~. or the density q of eq. Tomaras / Magnetic cortices The discussion of conservation laws would not have been complete without commenting on the situation in the one-dimensional ferromagnetic continuum. it seems natural to define the linear momentum as the generic value of this integral. and to Michael Marder for a number of related discussions. Papanicolaou. the closest example being that of vortices in a superfluid or in a superconductor [3]. In conclusion.10) would be ambiguous if the magnetization happens to reach the south pole at some intermediate point.N. The unusual dynamical features of magnetic vortices are understood as a result of a radical change in the canonical structure of the theory due to the underlying topology.

U(x). x 2) is the electrostatic potential. We use the symbol B for the magnetic field to distinguish it from the magnetic fields occurring in the main text. The corresponding Poisson bracket relations read as 1 {'g'l. (A. + "rrt/B). 2 2--772.4) while retaining rr~ and 772 as independent variables. Tomaras / Magnetic vortices 459 electric field E = . (A. A more transparent set of variables is obtained by introducing the guiding center coordinates 77 2 'IT1 R1--Xl + B ' R2=x2 B ' (A."rr:/B.U(Rt . {-r/'1. 71"2= . x2} . The electron motion is then governed by the equations 21 --Tr 1.B . (A. x2) ".5) and the Hamilton equations as OU (rl = B T r 2 .OR---~l' ! OU OU +2 = .R2} = ~-.6) where U --. ~'1 = B ' t r 2 . The mass and the charge of the electron as well as the speed of light are set equal to unity. In fact.0 2 U .2) endowed with the Poisson bracket relations {'rrl. while the electrostatic potential U corresponds to the magnetic field h of eq. T..N. I~2 - B OR l . the magnetic field B in the electron problem corresponds to the winding number Q of a vortex. . rather than to denote magnetic induction.0R--~2 I ~U l~l - B OR 2 . (5.1). Rv} = 0 .U(x~. {.N. {R.1) where ~-~ and 772 are the components of the mechanical momentum.B ' r r I . Papanicolaou.B .'Tr2} = .B ' n ' ! . (A.VU.t J l U .772} = . These equations may be understood as a hamiltonian system with hamiitonian ' ' (A.a-u.3) where we display only the nonvanishing Poisson brackets. R. where U . Xl} = 1 = {'r/'2.

the last two equations in eq. the last two equations in (A. /~2 = (A. which is the conservation law associated with invariance under space translations.7) which is also seen to be different from its standard mechanical value x F r 2 -x2"/'/" 1.n ' ~ 2 . which requires that U be axially symmetric. Finally we consider the effect of dissipation modeled here by adding a term -3. with a Hall velocity whose magnitude is given by I VUI lB.6) cannot be decoupled from the rest.yrr~.x2) neither the guiding center remains fixed nor the angular momentum is conserved. n "l.N.-~-B-(zr2 +'rr2). R2). Nevertheless some general properties of the electron motion can be derived without a detailed solution of the equations of motion. T.6) becomes 7:r I = B r r 2 + E . E 3. Then the system of equations (A. Tomaras / Magnetic vortices In the absence of an electric field (U = const.e. except in the adiabatic (large-B) limit where 10U /~1-- 10U J~2 B 0R 2 B oar I U = U( R I .9) This system can be solved explicitly. the guiding center again drifts along equipotential lines. where 3' is the dissipation constant. but the essential result may be obtained simply by assuming that a steady state is eventually reached where "/r~ = 0 = "h-2. R 2 ) . (A.6) become /~ = 0 =/~2. g l = -. For a more general potential U.460 N. in the case of a uniform electric field pointing in the xrdirection.7T2.zr to the Lorentz force. For simplicity we assume that the electric field is uniform. As a result. Now in the presence of a general electrostatic potential U = U(x~. 3. i. Papanicolaou. U = -Ex~.8) which may be viewed as a hamiltonian system for the pair of canonically conjugate variables R~ and R2 with hamiltonian U(R~.6) read /~ = 0 and /~2 = -E/B.~-Tr I . ~2 = -Brr~ . in a direction perpendicular to the electric field.3. and imply that the guiding center drifts in a direction perpendicular to the electric field with constant Hall velocity VH = -E/B. is given by the angular momentum B B 1 l=(XiTF2--X27"I'I) -~.) the last two equations in eq."~(X I + X 2 ) = -~-(R~ + R 2 ) . (A. The conserved quantity associated with invariance under space rotations. For instance. Then the first two equations become algebraic for the unknowns "n'~ and "rr2 which are inserted in the last two equations to yield the terminal velocity of the guiding . (A. implying that the guiding center remains fixed. (A.

M. G. 45 (1974) 5406 [10] G. Springer Series in Solid-State Sciences (Springer.P. V. Usp.A. (A. Gann. References [1] A. Derrick. Ferromagnetodynamics. Magn.A.A. T. Walker. Press. Zakrzewski. 5 (1964) 1252 ill] A.H. 1979) [4] N.1 1) may be written as V= VH Sin S. O'Dell. the dynamics of magnetic bubbles. Magnetic domain walls in bubble materials (Academic Press.G. Phys. Mat. Papanicolaou. (A. 88 (1990) 116: B.A. Lett.V. Phys. Atiyah and N.H. Math. 1988) [6] R.R. 1981) [3] R. Bll0 (1982)54 [5] M. Princeton.B 2 ' V2 = 1~ 2 = BE 72+ E 2 " v = (v. 22 (1975) 245 [12] B. Smolenskii and B. Appl. Bar'yakhtar. Magn. Slonczewski. J. (6. J.N. Lifshitz. Yu. The same results are obtained by explicitly solving the initial value problem for eq. Peyrard and W. Phys.~ 2 -.10) Thus we see that the guiding center undergoes skew deflection at an angle 8 with respect to the direction of the electric field. Nonlinearity 3 (1990) 773 [7] V. Stephanovich and A. (6. M. Hitchin. Recalling also that the Hall velocity in the absence of dissipation would now be given by Vn = .A. The geometry and dynamics of mabaetic monopoles (Princeton Univ. Lett. lvanov.A. Huebener. Schryer and L. I. J. Landau and E. Gorobets.A. Phys. Leese. 20 (1977) 298 [81 L.11) This relation is analogous to eq. AI41 (1989) 89 . Malozemoff and J. Magnetic flux structures in superconductors.A. Zhmudskii.J. New York.S. eq. Berlin. Sov. Belavin and A. V. JETP Left.N. New York.19). lv:mov and V. domains and domain walls (Wiley.P. Polyakov. 1979) [2] T. Physik A (Soviet Union) 8 (1935) 153 [9] N. Filippov. Stephanovich.I.12) which is the analog of eq. Tomaras / Magnetic cortices 461 center: yE V! = g l = . (A.16) given in the main text.C. sin~= I/2 V B E V" (A. Phys.L. Manton.E / B .9) and then taking the limit t ~ ~.N." + E (y2 + B2)l/2 " (A.

Let[.) 120 (1979) 107 [14] F. Chetverikov. Wen and A. Let[. 30 (1973) 230: J. Differential forms in algebraic topology (Springer. Papanicolaou. Tjon and J. Phys. Appl. Maslov and V.M.462 N. Papanicolaou. Phys. JETP Let[. Thiele. B15 (1977) 3470 [18] N. Theor. Phys. 20 (1987) L83 [16] A. C: Solid State Phys.V. Wright.D. Volovik. Phys. 61 (1988) 1025 . A84 (1981) 151 [19] A. Phys. 77 (1988) 1180 [21] J. Rev. 57 (1986) 1488 [15] G. Kiseliev. Borisov and V.A. Rev.G.M. Rev. Jevicki and N. Rev.P.Y.A. Phys. Tomaras / Magnetic cortices [13] A. New York. J.E. Bott and L W Tu.N. 1982) [~] X. 45 (1974) 377 [171 J. Dzyloshinskii and B. 29 (1979) 540 [22] R.B. Phys. Zee. Ann. Phys. T. Papanicolaou. Ivanov. Math. Haldane. (N. Phys. A107 (1985) 161 [20] V. Phys. Let[. Let[. Let[.E.