CHAPTER 1

Introduction

1.1 Preamble

Over the recent years, the study of ferromagnetic materials has gained a great at-
tention due to its enormous application in modern technological devices such as
hard-disks, magnetic sensors, cellular phones, magnetic random access memory
(MRAM), ferroelectric random access memory (FeRAM) etc. In view of the en-
hancement of these applications, one of the most promising innovation is to provide
a cheap storage of digital information with less space consumption and it opens the
door for nanostructures. Due to its unprecedented significance in both the applica-
tions and mathematical interest, it is extremely essential and yet more challenging
to study the magnetic behavior of these nanostructures mainly in respect of its sta-
bility and controllability features. Apart from its uses in storage sector, ferromag-
netic materials also play a inevitable role in other sectors of engineering as well
viz. biomedical sector, automotive engines, nuclear systems and environmental
sciences etc ([15, 149]).

In particular, ferromagnetic nanowires (nanostrips) are extremely advantageous

in nanotechnological devices where the information is encoded as a pattern of mag-
netic domains along the wire (strip). Domains are the uniformly magnetized re-
gions which are separated by a thin continuum zones referred as domain walls
(DWs) or simply walls. More precisely, in case of MRAMs (FeRAMs) the data
bit is achieved by passing the magnetic field (electric current pulses) through the
material (see [109, 117]). Features like reading and writing of the data are closely
related to the formation and evolution of domain walls. Therefore, for the devel-
opment of these devices, it becomes indispensable to understand the DW motions
under the action of magnetic fields and spin-polarized electric currents. For the
complete description of the physical state-of-the-art of ferromagnetic material and
DWs, we navigate the reader to the excellent monographs [3, 73, 129].
1.2 Magnetic Domain Walls

Before moving on to the formal treatment of the DWs and its characteristics, we
describe the ferromagnetic materials characterized by a time varying spontaneous
magnetization vector field. Ferromagnetic material is highly attracted to magnets
and has a tendency to become permanently magnetized. We consider a finite homo-
geneous ferromagnetic body denoted by Ω ⊂ R3 . The body is made of infinitesimal
physical volumes δV . These volumes are supposed to be large enough to contain
huge number N of elementary magnetic moments µi , i = 1, 2, . . . , N . The mag-
netization vector M is defined as the vector sum of the magnetic moments present
in the elementary unit volume δV , i.e.

N
X
µi
i=1
M (r) = . (1.2.1)
δV

where r ∈ Ω represents the position vector. We assume the elementary physical

volumes to be sufficiently small so that the magnetization vector can be treated as
continuous in space. Moreover, we suppose that the material is get saturated and
attains its saturation value Ms being constant in both the time and space variables.
After renormalization, we define the time varying unitary magnetization vector
field as:

M (t, r)
m(t, r) = . (1.2.2)
Ms

More precisely, magnetization vector field m is given as:

m : R+ × Ω → R3 ,
(t, x, y, z) 7→ m(t, x, y, z).

The magnetization vector m links the magnetic induction B and the magnetic field
H by the relation B = m̄ + H, where m̄ is the extension of m by zero outside Ω.
We denote by · the Euclidean scalar product on R3 and | · | the associated norm and
× is the vectorial product. Also, m satisfies the physical constraint:

|m| = 1 at any point. (1.2.3)

2
 Experimentally, it has been observed that below a critical temperature referred
as Curie temperature, ferromagnetic material has a propensity to split up into a
small uniformly magnetized regions. In fact each material is made of atoms and
atoms themselves have magnetization direction due to the spins and orbital mo-
ments of the atom’s electron. Especially, in ferromagnetic material, due to the
exchange interaction (detail appears in sequel), group of atoms with like magne-
tization direction accumulate with each other. These regions of magnetic atoms
are termed as domains and the transition zone between them is known as domain
wall.

DWs are further classified in two main categories depending on the modes of
rotation of the magnetization vector.

1.2.1 Bloch Walls

In case of Bloch walls, the rotation of the magnetization from one domain through
a wall to another domain takes place in a plane parallel to the wall plane. This
type of walls mainly occur in bulk materials or thick films. Fig. 1.1, exhibits a
180◦ Bloch-type DW separating two faraway domains in which the orientation of
magnetization differs by an angle 180◦ .

Figure 1.1: Change of orientation of magnetization vector in 180◦ Bloch-type

DWs.

3
1.2.2 Néel Walls

In these types of walls, the rotation of the magnetization from one domain through
a wall to the other domain happens perpendicular to the wall plane. Néel walls
usually appear in thin magnetic films. A 180◦ Néel-type DW has been depicted in
Fig. 1.2.

Figure 1.2: Change of orientation of magnetization vector in 180◦ Néel-type DWs.

In addition to Bloch and Néel walls, there exists a rich pattern of DWs such as
transverse DWs, chiral DWs, vortex, spiral and skyrmions. We shall discuss about
these various patterns in the course of this Chapter. Also, experimental evidence
shows that DW motion experiences two different dynamical regimes in presence
of external sources (magnetic fields or electric currents). For small value of fields
(currents), DW profile moves rigidly with constant velocity which leads to a steady-
state regime of high DW mobility. The minimum value of the field (current) for
which the wall motions are observed in steady-state regime is known as thresh-
old value. The maximum value of external sources for which DW motion remains
steady is referred as Walker breakdown. As the field (current) exceeds the break-
down value, DW structure gets altered due to the internal deformation which leads
to the oscillatory motion of magnetization configuration between Bloch and Néel
structures. Such a dynamic regime is called precessional [100, 121].

The formation and evolution of domain walls is one of the most celebrated
topic among the researchers. The formation of domain walls is determined by
the interplay between the various competing energy contribution. Discussion of
these various energy contributions of total micromagnetic free energy is of our
next interest which we detail in the subsequent section.

4
1.3 Micromagnetic Free Energy

In this section, we embark on the various energy contributions of free energy which
delineates the inherent magnetic properties of the material. Formation of domains
allows a metallic ferromagnet to minimize its total micromagnetic free energy
which comprises of several energy components. The main energy contributions
are the exchange energy, anisotropy energy, demagnetizing (stray field) energy,
Zeeman (applied field) energy which determine the shape and size of the domains.
Also, another component of total free energy is magnetostrictive energy (elastic
energy) which we have left out and is beyond the scope of this thesis. Technical
details of these energies are as follows:

1.3.1 Exchange Energy

The ferromagnetic behavior is essentially due to a quantistic force which tends

to align the magnetic dipole moment parallel to each other. Exchange energy is
minimized if magnetic moments inside the material are parallel or close to parallel
to each other. In what follows, this energy favors the wide walls. We write the
exchange energy as (cf. [2, 7, 23, 44]):
Z
A
Eexch = |∇m|2 dΩ,
2
Ω

with exchange field

Hex (m) = A∆m.

where the Eexch and Hex (m) corresponds to exchange energy and exchange field
respectively. Also, the constant A is expressed in terms of the exchange constant
Aex as:
2Aex
A= .
µ0 Ms2
µ0 is the magnetic permeability of the vacuum.

5
1.3.2 Anisotropy Energy

Magnetic behavior of material is highly influenced by the direction in which we

apply the external source. In the absence of external sources, the orientation of
magnetization tends to align itself along one or more specific directions in ferro-
magnetic crystal. These energetic preferred directions of magnetization are called
"easy-axes". The amount of energy require to switch the magnetization away from
the easy-axis is defined as anisotropy energy. We denote it by Ean .
Z
Ean = fan dΩ,
Ω

where energy density fan : R3 → R+ is a convex function which depends on the

material property (see [142]). The value of energy density fan is optimized if the
moments are parallel or close to parallel to one of the easy axis directions of crystal
lattice. In the course of this thesis, we have used the metallic ferromagnet with
high perpendicular magnetocrystalline anisotropy and in this case the anisotropy
field Han is defined as (cf. [44, 48, 110, 138]):

Han (m) = β(m · e)e,

where the unit vector e denotes the direction of the easy axis and the constant
2K
β = , K being the anisotropy constant, depends on the material property.
µ0 Ms2
However, for more details related to the uniaxial anisotropy and cubic anisotropy,
we refer the reader to Sec. 3.2.3 in [73] and [40, 49].

1.3.3 Demagnetizing (Stray field) Energy

A magnetized medium of ferromagnetic material comprise of single domain pos-

sess a macroscopic magnetization. In fact, due to this magnetization the medium
tends to act like a magnet with a magnetic field lines around it. The direction of
these field lines is in the opposite direction from the direction of the magnetiza-
tion inside the medium. In what follows, field outside the medium magnetize the
material in the opposite direction from its own magnetization. In view of this, it
is referred as demagnetizing field and we denote it as Hd . The associated energy

6
determines the shape of the sample and can be optimized by reducing the exter-
nal demagnetizing field. This energy is also known as magnetostatic energy or
stray field energy. Therefore, in simple words, it is the energy which is connected
with the magnetic field generated by the medium itself and denoted by Edem (cf.
[2, 7, 22]: Z Z
1 2 1
Edem = |Hd (m)| dΩ = − Hd (m) · m dΩ.
2 2
R3 Ω

The demagnetizing field (stray field) Hd (m) is characterized by the Maxwell equa-
tions:

 curl Hd (m) = 0 in R3 ,


 div (Hd (m) + m̄) = 0 in R3 .



It is worth to mention that for uniformly magnetized samples, the demagnetizing

field is given as (cf. [14, 37, 40, 45, 106]):

Hd (m) = −D · m.

where D is a diagonal matrix depending on the material property. For example:

• From the classical results (see [2, 7, 106] and Sec. 3.2.5 in [73]), we know
that the demagnetizing (stray) field generated by a uniformly magnetized
ellipsoid is given by:
 
α 0 0
Hd (m) = −D · m = −  0 β 0  m,
0 0 γ

where D is a symmetric positive matrix and the constants α, β and γ depend

on the ellipsoid geometry.
• For a straight ferromagnetic nanowire with circular cross section and with
the assumption that the radius of the wire is very small in comparison to the
length of the wire, an expression for demagnetizing field is derived using the
Γ-convergence arguments (for details see [22, 32, 120]) and given as:

Hd (m) = −m2~e2 − m3~e3 ,

where (m1 , m2 , m3 ) are the coordinates of m in R3 .

Remark 1.3.1. In ferromagnets, except for the aforementioned favorable cases,

the stray field calculations are highly complex and we need to draw our attention
towards its numerical evaluation. For details, we refer the interested reader to [1,

7
9, 12, 73, 141] and the references listed therein. Moreover, the distribution of non-
uniform demagnetization fields in paramagnetic bulk solids for the various sample
geometries (cones, ellipsoids, paraboloids and hyperboloids) are investigated in
[52].

1.3.4 Zeeman (Applied field) Energy

The applied field energy Ea reflects the effect of an applied magnetic field Ha on
the magnetic medium and is defined as (cf. [2, 22, 23, 49]):
Z
Ea = − Ha · m dΩ.
Ω

This energy is optimized if every magnetic moment of the magnetic system tends
to align parallel or close to parallel to the applied magnetic field. In nature, the
applied magnetic field can be a static magnetic field, a time varying field pulse or a
harmonic high frequency field. This energy is also termed in literature as Zeeman
energy.

Thus, we define the total micromagnetic energy E as the sum of exchange en-
ergy, anisotropy energy, stray field energy and applied field energy.

E = Eexch + Ean + Edem + Ea .

It is evident that different energy contribution favours different domain arrange-

ment, however the total energy is optimized by a compromise domain arrangement
which has been portrayed in Fig 1.3 (cf. Sec. 3.2 in [73] and chap. 7 in [129]).

Next, we describe the static aspects of micromagnetism which was proposed

by W.-F. Brown in the 60’s (cf. [19]). The theory of micromagnetism is based
on the fact that the equilibrium states (metastable states or relevant configurations)
of the magnetization minimize the total micromagnetic energy. We obtain these

8
equilibrium states using the variational principle for the total energy:

 δE(m) = 0,

(1.3.1)
 δ 2 E(m) > 0.


where the first order variation δE(m) vanishes for any variation δm of the magneti-
zation vector m which satisfies the physical constraint |m+δm| = 1 (cf. [49, 130]).
Eqn. (1.3.1) renders Brown’s Equation:

 m × Hef f = 0,

(1.3.2)
 m × ∂m = 0 on the boundary ∂Ω.

∂n

where n is a unit normal vector to the boundary ∂Ω. Also, Hef f represents the total
effective field which is the sum of individual field come from the various energy
components, i.e.

Hef f = Hex (m) + Han (m) + Hd (m) + Ha . (1.3.3)

However, in presence of electric current field, there are some other contributions
which participate in total effective field such as Rashba field and Dzyaloshinskii
Moriya field. We shall discuss these field components in the later part of this Chap-
ter. Moreover, effective field is connected with the total micromagnetic energy by
the relation Hef f = −δE, the first variation of the energy functional.

Figure 1.3: Schematics of a domain arrangement that optimizes the total micor-
magnetic energy.

9
 The physical significance can be interpreted from the Brown’s equation in a
way that the torque exerted on the magnetization by the effective field must vanish
at the equilibrium state. Also, it is worth to bring into the notice that equilibrium
states can be obtained using the variational method which does not require infor-
mation about the dynamic behavior of magnetization (that is, how does it reach the
equilibrium as time evolves). Technical insights of the dynamic features of magne-
tization under the influence of external sources (magnetic field or electric current
pulses) are discussed in the following section.

1.4 Theory of Governing Dynamics

The dynamic behavior of magnetization namely its precessional motion and dis-
sipative processes is the prime concern of this section. The dynamic aspects of
magnetization were first introduced by L. Landau and E. Lifshitz in 1935 (cf.
[84]). They derived an equation which governs the evolution of magnetization
inside the ferromagnetic medium. Later, W. F. Brown [19] laid the foundation of
micromagnetic theory which rely on the fact that the equilibrium states of mag-
netization minimize a given energy functional. T. L. Gilbert [61], modified the
Landau-Lifshitz (LL) equation by replacing the dissipation term with an equivalent
torque term arising from a viscous force. This modified equation known as Landau-
Lifshitz-Gilbert (LLG) equation describes the motion in metallic ferromagnets with
no crystallographic defects. Such materials are termed as ideal ferromagnets.

In order to understand the dynamics in the materials with structural disorder

(non-ideal ferromagnets), the Gilbert linear dissipation term is generalized through
the two additional dissipation torque terms which account for nonlinear viscous and
dry-friction dissipation [44, 47, 48, 113, 143]. This equation is known as extended
Landau-Lifshitz-Gilbert (ELLG) equation. Moreover, it has been observed that
below the typical time scale (of the order of picosecond) LLG equation violates.
In this time domain LLG equation is modified through an additional inertial term
which describes the motion in a more realistic way (cf. [42, 59, 105, 146]). The
dynamic equation with an inclusion of inertial effects is referred as inertial Landau-
Lifshitz-Gilbert (ILLG) equation.

10
 In the above discussed dynamic models, the magnetization is driven by the
applied magnetic field. It was J. C. Slonczewski [127] who first introduced and
further modified the LLG equation and included the effect of spin-polarized electric
current into the dynamics. The modified equation bears his name and is referred as
Landau-Lifshitz-Gilbert-Slonczewski (LLGS) Equation. In the sequel, we discuss
these various dynamic models and their theory in detail.

1.4.1 Landau-Lifshitz (LL) Equation

In 1935, L. Landau and E. Lifshitz proposed the governing dynamics which de-
scribes the precession of magnetization around the total effective field and is given
as (cf. [84]):

∂m
= −γm × Hef f . (1.4.1)
∂t

However, Eqn. (1.4.1) fails to provide insights about the dissipative processes.
In order to make the magnetization dynamics lucid, they meliorate the equation
through an additional torque term that describes the dissipation processes in mag-
netic system. The modified equation takes the following form:

∂m
= −γm × Hef f − λm × (m × Hef f ) . (1.4.2)
∂t

where the constant γ = Ms µ0 γe is defined in terms of the magnetic permeability

ge
of the vacuum µ0 and the gyromagnetic ratio γe = . The constants g, e and
me
me represent the Landè factor, electron charge and the electron mass respectively.
Also, λ > 0 is a phenomenological damping parameter. Eqn. (1.4.2) is referred in
literature as Landau-Lifshitz (LL) equation.

We remark that, given a solution m of (1.4.2) with a static applied field follow-
ing holds (cf. [6, 83, 112]):
Z
d
(E(m(t, .)) = −λ |Hef f (m(t, x)) − (Hef f (m(t, x)) · m(t, x)) m(t, x)|2 .
dt
Ω

This energy functional is nonincreasing along a solution of (1.4.2) and reflects that
the second term appearing in (1.4.2) describes the dissipative processes. However,

11
the first term is responsible for the deviation of magnetization towards total effec-
tive field and referred as Larmor precession. Fig. 1.4, depicts the undamped and
damped precessional motion of magnetization vector m around an effective field
Hef f . Also, τp and τd denote the precessional and dissipation torques appear in
R.H.S. of (1.4.2) in the respective order.

Figure 1.4: Variation of magnetization vector m around an effective field Hef f (a)
undamped precession (b) damped precession.

Remark 1.4.1. Every steady-state of (1.4.2) satisfies m×Hef f (m) = 0 (since both
terms appearing in the right hand side of (1.4.2) are orthogonal to each other)
which is precisely the Brown’s equation. Therefore, this leads to the conclusion
that set of steady-states of magnetization coincides with the set of extremal points
(relevant configurations) of the micromagnetic energy E (see [6]).

1.4.2 Landau-Lifshitz-Gilbert (LLG) Equation

To define the dissipative processes, T. L. Gilbert used a different approach to mod-

ify the conservative equation (1.4.1). More precisely, he included the torque term
which depends on the rate of change of magnetization.
 
∂m ∂m
= −γm × Hef f + ξG m × . (1.4.3)
∂t ∂t

where ξG represents the standard phenomenological Gilbert damping constant.

Eqn. (1.4.3) is generally termed as Landau-Lifshitz-Gilbert (LLG) equation.

12
It is indeed worth to mention that from the physics point of view, Landau-Lifshitz
and Landau-Lifshitz-Gilbert equations share a substantial differences (cf. [77, 90]),
however mathematically they are very similar. In fact both the equations are equiv-
alent. To observe this, we take a cross-product of both sides of (1.4.3) by m which
yields:
 
∂m ∂m
m× = −γm × (m × Hef f ) + ξG m × m × . (1.4.4)
∂t ∂t

∂m
Under the impression of the physical constraint (1.2.3), we have m · = 0.
∂t
Using this, Eqn. (1.4.4) becomes:

∂m ∂m
m× = −γm × (m × Hef f ) − ξG . (1.4.5)
∂t ∂t

By substituting back the equation (1.4.5) in the R.H.S. of (1.4.3), we obtain:

∂m γ γξG
=− 2
m × Hef f − 2
m × (m × Hef f ) . (1.4.6)
∂t 1 + ξG 1 + ξG

Eqn. (1.4.6) is referred as Landau-Lifshitz equation in Gilbert form. It is clear

from (1.4.2) and (1.4.6) that LL and LLG equations are equivalent up to the con-
stants (cf. [6, 40, 49]). Since the inception of LL and LLG equations, there have
been a phenomenal contribution in the theoretical results concerning to existence
of solutions, stability, controllability and DW motions related to one-dimensional
to higher dimensional models. In the sequel, we discuss some of the main contri-
butions related to these results.

Several results on existence of solutions concerning to LL and LLG equations

have been established in literature. First theoretical result related to LL equation
was investigated by Visintin [142]. In this, he studied the weak solutions of LL
equation equipped with Maxwell’s equation under the influence of exchange and
anisotropy energy. Thereafter, the theory of existence of weak solutions has been
evolved rapidly and investigated [8, 16, 20, 25, 26, 29, 54, 66]. However, the
strong solutions of LL equation for the various models have been considered in
[27, 30, 39, 41, 53, 151]. Also the asymptotic behavior of LL equation coupled
with Maxwell’s equation can be found in [28, 75]. In addition, existence of global
weak solutions of Landau-Lifshitz-Gilbert equation has been studied under differ-

13
ent framework in [67, 101, 111, 140]. Theoretical results concerning to numerical
aspects of micromagnetics can be sought in [1, 9, 10, 11, 12, 82, 107] and the
references listed therein.

Over the last decade, an enormous investigation related to the theory of sta-
bility and controllability analysis have been carried out by many researchers. To
be specific, stability and controllability results relevant to one-dimensional models
(nanowires) of LL equation have been established in [22, 23, 32, 33, 34, 35, 36,
83, 112, 120]. In [22], author has inspected several papers concerning the DW
motions in ferromagnetic nanowires and listed many other important literature. In
addition, various results relevant to higher dimensional models have been derived
in [21, 51, 97, 131]. Next, we highlight the studies related to other nanostruc-
tures mainly the magnetization dynamics in nanoparticles (magnetic dots). In most
of the literature, authors have considered the particles geometry as an ellipsoidal
shape, perhaps due to the well known expression of stray-field for a uniformly
magnetized ellipsoid. Some significant works related to magnetization reversal for
various models of an ellipsoidal shape particles can be found in [2, 7, 95, 106].

1.4.3 Extended Landau-Lifshitz-Gilbert (ELLG) Equation

In this section, we describe the dynamics which reflects the relaxation processes in
a more realistic way. It is worth mentioning that in LLG equation dissipative phe-
nomena are described via a linear viscous type Gilbert dissipation. However, LLG
equation does not provide sufficient description related to a large angle magneti-
zation precession occurring in the nonlinear regime. To describe such phenomena,
we need to introduce an additional phenomenological nonlinear viscous dissipation
torque into the dynamics. In addition, to take into account the effect of crystallo-
graphic defects of the material, we need to further include the dry-friction dissipa-
tive torque into the governing equation. We denote the general phenomenological
dissipative torque which include these effects by Td and write it as:

Td = Td1 + Td2 = m × F1 (∂t m) + γξd (m × F2 (∂t m)) ,

14
with  2k
X ∂t m
F1 (∂t m) = ξG ξk ∂t m, (1.4.7)
k≥0
γ

∂
where ξk ’s are dimensionless coefficients with ξ0 = 1. Also, the notations ∂x
= ∂x ,
∂2
∂x2
= ∂xx are used for the partial derivatives. Taking the first order approximation
of Taylor series expansion of R.H.S. term appearing in (1.4.7), we obtain
  
ξv 2
Td1 = ξG 1 + 2 (∂t m) (m × ∂t m) . (1.4.8)
γ

In (1.4.8), the first term represents the linear viscous-type Gilbert dissipative torque
which describes the dissipation phenomenon in the case of magnetic materials
without any intrinsic defects i.e. ideal materials (cf. [44, 138]). Second term de-
scribes a rate dependent nonlinear viscous dissipation (cf. [45, 48, 135]). Here the
positive dimensionless constants ξG and ξv represent the classical Gilbert damping
and nonlinear viscous dissipation coefficients respectively (for the sake of notation,
we use ξv in place of ξ1 ).

Dissipation torque Td2 accounts for the presence of magnetic inclusions such as
dislocations, impurities and other metallic defects in the ferromagnetic materials.
Due to such inclusions an internal friction, also known as dry-friction, takes place
in the material which in turn opposes the DW propagation. This mechanism is
rate-independent and responsible for the hysteresis and Barkhausen effect (see [24,
38, 44, 55, 116, 143, 144, 145]). This additional term is described by the maximal
monotone operator F2 defined as (cf. [24]):

∂m
 t if ∂t m 6= 0,


F2 (∂t m) = |∂t m|

 B(0, 1) if ∂t m = 0.


where B(0, 1) denotes the unit ball centered at origin in R3 . ξd is a positive dry-
friction parameter which accounts for the average distributions of defects in the
material.
Thus, Td takes the following form:
   
ξv 2 γξd
Td = ξG 1 + 2 (∂t m) + (m × ∂t m) . (1.4.9)
γ |∂t m|

15
Equation (1.4.9) is valid for ∂t m 6= 0. For ∂t m = 0, we choose the arbitrary con-
stant vector which belongs to B(0, 1) in such a way that the dry-friction dissipation
torque vanishes. For simplicity, we choose such an arbitrary vector to be as m. We
replace the second term in (1.4.3) by the more general form of dissipative torque
Td given in (1.4.9). Thus, we get (cf. [44, 48]):
   
∂m ξv 2 γξd
= −γm × Hef f + ξG 1 + 2 (∂t m) + (m × ∂t m) . (1.4.10)
∂t γ |∂t m|

Eqn. (1.4.10) is generally known as Extended Landau-Lifshitz-Gilbert equation

which describes the magnetic behavior under the action of applied magnetic field.

1.4.4 Inertial Landau-Lifshitz-Gilbert (ILLG) Equation

In general, the governing dynamics which have been discussed so far viz. LL,
LLG, ELLG equations are found inadequate to describe the connection between
the magnetic properties and the inertial properties of ferromagnetic systems. It
has been observed that, in case of ultrafast magnetization dynamics, the dissipa-
tive processes need to be further modified through an additional phenomenological
torque (inertial damping torque). To capture the inertial effects, the generalized
LLG equation is derived within the framework of mesoscopic nonequilibrium ther-
modynamics theory (see [42, 59, 105, 134, 146]). To be precise, a relaxation time τ
is introduced as a phenomenological parameter which bifurcates the magnetization
dynamics in two different regimes: diffusion regime (t  τ ) and the precessional
regime (t ≈ τ ). However, the usual LLG is recovered in the limiting case as τ → 0.

The inertial Landau-Lifshitz-Gilbert equation which accounts the nonlinear dis-

sipations is described as follows (cf. [69, 104, 146]):
   
ξv 2 γξd
∂t m + ξG 1 + 2 (∂t m) + (m × ∂t m) + ξG τ (m × ∂tt m)
γ |∂t m|
= γ (m × Hef f ) . (1.4.11)

The third term appears in L.H.S. of (1.4.11) depicts the inertial effects in the system
in which τ accounts for the relaxation time of the angular momentum characteriz-
ing the relaxation of magnetization acceleration from the inertial regime toward the

16
precessional regime.

1.4.5 Landau-Lifshitz-Gilbert-Slonczewski (LLGS) Equation

In the current induced DW motion, spin-polarized electric current that enters into a
ferromagnet exerts a torque on the magnetization due to the transfer of spin angular
momentum. This torque term is called spin transfer torque (STT). In 1996, Berger
[13] and Slonczewski [127] proposed the inclusion of STT term into the dynamics
that describes the interplay between the spin-polarized electrical current and local
magnetic moments.

The equation which describes the current-driven domain wall motions in fer-
romagnets with an inclusion of nonlinear dissipations is given as (cf. [44, 45, 48,
113]):

  
ξv 2 γξd
∂t m − ξG 1 + 2 (∂t m) + (m × ∂t m)
γ |∂t m|
= −γm × Hef f + (−1 + ζm×)(vs · ∇)m.
(1.4.12)

The last two terms on the R.H.S. of (1.4.12) represent the STTs, in which, the
first term depicts the adiabatic contribution and describes a reactive STT whereas
the second term is non-adiabatic and represents the dissipative torque (cf. [44, 99,
127, 132]). The constant ζ accounts for the phenomenological non-adiabatic spin-
torque parameter. Also, vs stands for the spin-torque velocity which is directly pro-
portional to the current density J. Eqn. (1.4.12) is known as generalized Landau-
Lifshitz-Gilbert-Slonczewski equation. However, in literature, Eqn. (1.4.12) with-
out an inclusion of nonlinear dissipations is usually referred as Landau-Lifshitz-
Gilbert-Slonczewski equation.

Next, we move our attention towards the rich theory related to current induced
DW motion in ferromagnetic materials. Due to its huge applications in spintronics
(spin transport electronic devices), numerous theoretical descriptions of current
induced DW motion in magnetic nanowire (nanostrip) have been reported recently.
Mainly, the global weak solutions, classical solutions and stability results related
to one-dimensional model of LLGS equation have been investigated in (cf. [31,

17
98, 136, 148]). In addition, current driven magnetization dynamics, namely, the
description of DW velocity, threshold and Walker breakdown value of the external
sources in different dynamic regimes in magnetic nanowires can be found in [44,
46, 48, 93, 100, 139] and the references listed therein. In all these literature, authors
have studied the current induced DW motions with the inclusion of STT only and
have not included the spin-orbit torque (SOT) into the dynamics. Some of the
recent works with the inclusion of SOT into the governing dynamics are listed in
the sequel.

Recent experiment observations show that in case of metallic ferromagnet with

a strong structural inversion asymmetry, a strong spin-orbit coupling (SOC) leads to
the additional spin-orbit torque which contributes in total effective field and termed
as Rashba field (cf. [99]). The spin-orbit Rashba field can induce the DW mo-
tion along the current direction and delays the Walker breakdown (for instance see
[71, 91, 92, 99, 103, 113, 114, 119]). In addition, in case of ferromagnetic multi-
layers in which the metallic ferromagnet is sandwiched between a heavy metal and
oxide, this strong SOC renders an additional SOT which is due to a spin current
generated by the spin Hall effect (SHE) in a heavy metal layer and injected into
the ferromagnet (cf. [68]). Also, these ferromagnetic multilayer with strong SOC
is expected to experience an interfacial anisotropic exchange interaction which is
due to the Dzyaloshinskii-Moriya interaction (DMI). The inclusion of DMI into
the dynamics gives different magnetization patterns namely chiral DWs, spiral and
skyrmions (cf. [60, 94, 150]). The current induced DW motions with effect of SHE
and DMI are investigated in [64, 72, 79, 81, 85, 87, 122, 118, 133]. It is worth to
point out that magnetization dynamics in ferromagnetic multilayers is beyond the
scope of this thesis and left out for further studies.

In the field of spintronics, plenty of innovative proposals have been presented

in which DWs in magnetic nanowire are used to encode the data in the form of bit
in memory (cf. [108, 109, 117]) or logic devices (cf. [4, 5, 74]). Among these,
the most promising innovation is the magnetic racetrack memory [109] due to its
advantage over current silicon based memories and magnetic hard disk drives. In
the racetrack memory, we obtain a three-dimensional storage device by using U-
shaped magnetic nanowires normal to the plane of silicon wafer. In these nanowires
also referred as racetracks, data is encoded as a pattern of magnetic domains along

18
a portion of the nanowire by passing the electric current pulses through it (for
details see [108, 109]).

1.5 Preliminaries

In the present section, we review some standard results from the classical theory
of ordinary differential equation (ODE) which have been used in the forthcoming
Chapters. To begin with, first we state the Gronwall’s inequality (integral form) in
the following Theorem.

Theorem 1.5.1. (Gronwall’s inequality) Let u, α : [a, b] → R be bounded non-

negative measurable functions and β : [a, b] → R be a non-negative integrable
function with the property that
Z t
u(t) ≤ α(t) + β(s)u(s)ds, ∀ t ∈ [a, b],
a

then,
Z t Z t 
u(t) ≤ α(t) + α(s)β(s)exp β(τ )dτ ds, ∀ t ∈ [a, b].
a s

Moreover, in addition, if the function α is non-decreasing, then

Z t 
u(t) ≤ α(t)exp β(s)ds , ∀ t ∈ [a, b].
a

In the sequel, we briefly describe the local and global existence of the solution
to the Cauchy problem for a nonlinear nonautonomous system.

We consider an initial value problem for a nonlinear nonautonomous system as:


 u0 (t) = f (t, u(t)),

(1.5.1)
 u(t0 ) = u0 ,


where (t0 , u0 ) ∈ D ⊆ R × Rn is fixed, with D being an open set and f : D → Rn

is a continuous function.

Definition 1.5.1. By a solution to the problem (1.5.1), we mean a function u : I →

19
Rn defined on an open interval I ⊆ R, t0 ∈ I, which is differentiable and satisfies
(1.5.1) i.e. u ∈ C 1 (I, Rn ), u0 (t) = f (t, u(t)) for all t ∈ I and u(t0 ) = u0 .

Definition 1.5.2. By a sub-solution to the problem (1.5.1), we mean a function

u : I → Rn defined on an open interval I ⊆ R, t0 ∈ I such that u ∈ C 1 (I, Rn ),
u0 (t) ≤ f (t, u(t)) for all t ∈ I and u(t0 ) ≤ u0 . Moreover, solution is said to be
strict sub-solution if u ∈ C 1 (I, Rn ), u0 (t) < f (t, u(t)) for all t ∈ I and u(t0 ) < u0 .

Definition 1.5.3. By a super-solution to the problem (1.5.1), we mean a function

u : I → Rn defined on an open interval I ⊆ R, t0 ∈ I such that u ∈ C 1 (I, Rn ),
u0 (t) ≥ f (t, u(t)) for all t ∈ I and u(t0 ) ≥ u0 . Moreover, solution is said to be
strict sup-solution if u ∈ C 1 (I, Rn ), u0 (t) > f (t, u(t)) for all t ∈ I and u(t0 ) > u0 .

Definition 1.5.4. We say that the problem (1.5.1) has a local solution if there exist
an open interval I ⊆ R and a function u : I → Rn such that u is a solution of
(1.5.1) on I.

Definition 1.5.5. We say that the problem (1.5.1) has a global solution if for every
open interval I ⊆ R such that

t0 ∈ I, {x ∈ Rn |(t, x) ∈ D} =
6 φ, ∀ t ∈ I.

there exists a function u : I → Rn which is a solution of (1.5.1) on I.

Where φ represents the null set. Next, we present the statement of the local and
global existence and uniqueness Theorems.

Theorem 1.5.2. (Existence theorem) If f : D → Rn is continuous, then for every

(t0 , u0 ) ∈ D, there exist δ > 0 (depending on (t0 , u0 )) and a function u : (t0 −
δ, t0 + δ) → Rn which is a solution of (1.5.1) on (t0 − δ, t0 + δ).

Definition 1.5.6. By a unique local solution to the Cauchy problem (1.5.1), we

mean if there exist an interval J and a solution v : J → Rn on J such that, for
every other solution u : I → Rn , we have u = v on I ∩ J, where I and J are open
intervals containing t0 .

Definition 1.5.7. The continuous function f : D → Rn of the problem (1.5.1) is

said to be locally Lipschitz continuous in x ∈ Rn uniformly with respect to t ∈ R

20
if for every (t0 , u0 ) ∈ D there exist a neighborhood of (t0 , u0 ), D̃ ⊆ D, and a
constant L > 0 such that

||f (t, x) − f (t, y)||Rn ≤ L||x − y||Rn , whenever (t, x), (t, y) ∈ D̃. (1.5.2)

Theorem 1.5.3. (Uniqueness theorem) If f : D → Rn is continuous and satis-

fies the local uniform Lipschitz condition (1.5.2), then for every (t0 , u0 ) ∈ D, the
Cauchy problem (1.5.1) has a unique local solution in the sense of Definition 1.5.6.

Definition 1.5.8. We say that the Cauchy problem (1.5.1) has a unique global
solution if it has exactly one solution, in the sense that there exists a solution
v : J → Rn , such that for every other solution u : I → Rn we have: I ⊆ J
and u = v on I.

Theorem 1.5.4. (Global existence and uniqueness on a strip) Let us consider

the Cauchy problem (1.5.1), with D given by the strip D = (a, b) × Rn , where
−∞ ≤ a < b ≤ +∞. If f satisfies the global uniform Lipschitz condition i.e.
there exists L > 0 such that

||f (t, x) − f (t, y)||Rn ≤ L||x − y||Rn , ∀ t ∈ (a, b), ∀ x, y ∈ Rn .

then, for every (t0 , u0 ) ∈ D, the problem has a unique global solution, in the sense
of Definition 1.5.8. Moreover, the solution is defined on the whole interval (a, b),
that is, the problem is globally solvable.

Now, we describe the qualitative properties of solutions of differential equations

viz. stability, asymptotic stability and exponential stability.

Let us assume a physical phenomena whose governing dynamics is given by

the autonomous system:

u0 = f (u), (1.5.3)

where f : D → Rn with f ∈ C 1 (D) and D is some region in Rn . Let u = u0 be

an equilibrium or fixed point i.e. (f (u0 ) = 0).

Definition 1.5.9. The equilibrium solution u0 of (1.5.3) is said to be stable if for

each number ε > 0 we can find a number δ > 0 (depending on ε) such that if ψ(t)

21
is any solution of (1.5.3) having ||ψ(t0 ) − u0 || < δ, then the solution ψ(t) exists
for all t ≥ t0 and ||ψ(t) − u0 || < ε for t ≥ t0 .

The equilibrium solution u0 is said to be unstable if it is not stable.

Definition 1.5.10. The equilibrium solution u0 is said to be asymptotically stable

if it is stable and if there exists a number δ0 > 0 such that if ψ(t) is any solution of
(1.5.3) having ||ψ(t0 ) − u0 || < δ0 , then ||ψ(t) − u0 || → 0 as t → ∞. Moreover,
it is said to be globally asymptotically stable if lim ψ(t) = u0 is satisfied for any
t→∞
ψ(t0 ).

Definition 1.5.11. The equilibrium solution u0 of (1.5.3) is said to be exponen-

tially stable if there exist positive constants c, k and λ such that if ψ(t) is any
solution of (1.5.3), then

||ψ(t) − u0 || ≤ k||ψ(t0 ) − u0 ||e−λt , ∀ ||ψ(t0 ) − u0 || < c. (1.5.4)

and globally exponentially stable if (1.5.4) is satisfied for any ψ(t0 ).

For the exhaustive theory of ordinary and partial differential equations, we refer
the reader to the excellent books [18, 43, 58, 76, 128].

1.6 A Brief Description of the Problem

In this section, we give a brief introduction and motivation behind the magnetic
phenomena discussed in the subsequent Chapters. As pointed out earlier, the major
contribution in the theoretical results concerning to stability, controllability and
DW motions have been reported mainly for the one-dimensional models (cf. [2,
22, 44, 48, 62, 91, 113]). However, these issues in higher dimensional models
have not been studied in great detail. Apart from their mathematical interest, these
higher dimensional models are highly applicable in variety of technological areas
including photonics, electronics and optoelectronic devices. This motivated us to
study these problems in higher dimensional models. In particular, the problems
concerning to stability and controllability (magnetization reversal) in monodomain
and one-dimensional network models of ferromagnetic nanoparticle of ellipsoidal

22