You are on page 1of 12

Coupled Dzyaloshinskii walls and their current-induced dynamics by the spin Hall

effect
Eduardo Martnez and scar. Alejos

Citation: Journal of Applied Physics 116, 023909 (2014); doi: 10.1063/1.4889848
View online: http://dx.doi.org/10.1063/1.4889848
View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/116/2?ver=pdfcov
Published by the AIP Publishing

Articles you may be interested in
Current-driven dynamics of Dzyaloshinskii domain walls in the presence of in-plane fields: Full micromagnetic
and one-dimensional analysis
J. Appl. Phys. 115, 213909 (2014); 10.1063/1.4881778

Current-driven domain wall motion along high perpendicular anisotropy multilayers: The role of the Rashba field,
the spin Hall effect, and the Dzyaloshinskii-Moriya interaction
Appl. Phys. Lett. 103, 072406 (2013); 10.1063/1.4818723

Switching of a single ferromagnetic layer driven by spin Hall effect
Appl. Phys. Lett. 102, 212410 (2013); 10.1063/1.4808092

Current-induced motion of a transverse magnetic domain wall in the presence of spin Hall effect
Appl. Phys. Lett. 101, 022405 (2012); 10.1063/1.4733674

The influence of the Rashba field on the current-induced domain wall dynamics: A full micromagnetic analysis,
including surface roughness and thermal effects
J. Appl. Phys. 111, 07D302 (2012); 10.1063/1.3671416


[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
130.92.9.55 On: Thu, 28 Aug 2014 04:43:08
Coupled Dzyaloshinskii walls and their current-induced dynamics
by the spin Hall effect
Eduardo Martnez
1,a)
and

Oscar Alejos
2
1
Dpto. de Fisica Aplicada, Universidad de Salamanca, Plaza de los Cados s/n, E-37008 Salamanca, Spain
2
Dpto. de Electricidad y Electr onica, Universidad de Valladolid, Paseo de Bel en, 7, E-47011 Valladolid, Spain
(Received 29 May 2014; accepted 28 June 2014; published online 11 July 2014)
The nucleation of domain walls in ultrathin ferromagnetic/heavy-metal bilayers is studied by means
of micromagnetic simulations. In the presence of interfacial Dzyaloshinskii-Moriya interaction, the
nucleated walls naturally adopt a homochiral conguration with internal magnetization pointing
antiparallely. The interaction between these walls was analyzed and described in terms of a classical
dipolar force between the magnetic moments of the walls, which couples their dynamics.
Additionally, the current-induced motion of two homochiral walls in the presence of longitudinal
elds was also studied by means of a simple one-dimensional model and micromagnetic modeling,
considering both one free-defect strip and another one with random edge roughness. It is evidenced
that in the presence of pinning due to edge roughness, the in-plane longitudinal eld introduces an
asymmetry in the current-induced depinning, in agreement with recent experimental results.
VC
2014 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4889848]
I. INTRODUCTION
Nowadays spin-orbit driven phenomena at heavy-metal/
ferromagnet interfaces are the focus of intense research
efforts.
17
The inuence of spin-orbit coupling (SOC) on
spin transport and magnetization textures leads to rich funda-
mental phenomena that apart from their intrinsic interest can
be also exploited to enable high-performance and low-power
spintronic devices.
811
In particular, recent experiments
1216
have shown that in ultrathin metallic ferromagnets sand-
wiched between a heavy metal and an oxide, SOC and the
broken inversion symmetry lead to chiral domain walls
(DWs) through the Dzyaloshinskii-Moriya interaction
(DMI).
10,1724
Also in these systems, SOC can generate
much stronger current-induced torques (spin orbit torques,
SOTs) than conventional spin-transfer torques (STTs),
25,26
and recent works
14,15
suggest that the spin Hall effect
(SHE)
2731
is dominant on driving DMI-stabilized homochi-
ral Neel DWs with high efciency.
Experimental setups to study the current-induced DW
motion (CIDWM) usually start from an uniform state in the
ferromagnetic strip, which is magnetized along the high per-
pendicular magnetocrystalline anisotropy (PMA) direction
(out-of-plane, z-axis). By means of injection of current
pulses along a conducting and transversal wire,
9,14,32,33
the
direction of the magnetization can be locally reversed and
therefore two DWs are nucleated. Although there are few
theoretical works analysing these DW nucleation processes
for in-plane
33
and for out-of-plane
34
ferromagnetic mono-
layers, this numerical analysis has not been studied in high
PMA multilayers with broken inversion symmetry.
Moreover, most of the theoretical studies are focused on the
description of the CIDWM of a single DW, but the dynamics
of two coupled walls is currently interesting from both theo-
retical and technological points of view.
In the present work, the nucleation processes of two
DWs are micromagnetically studied in high PMA strips with
zero and nite DMI. The force between homochiral walls is
micromagnetically analysed by applying perpendicular elds
antiparallel to the magnetization in the central domain
between the two walls. After quantifying this repulsion
force, the CIDWM by the SHE is studied by means of a one-
dimensional model (1DM) which takes into account the
coupled motion of these homochiral walls along perfect sam-
ples. The CIDWM is also micromagnetically studied for
samples with weak and strong DMI in the presence of longi-
tudinal elds. Finally, and in order to describe realistic con-
ditions, the inuence of edge roughness is evaluated. Our
analysis of the coupled CIDWM of homochiral DWs is rele-
vant for describing experimental measurements, and it could
be also useful to design DW based spin-orbitronics devices.
II. MICROMAGNETIC MODEL: DIMENSIONS AND
MATERIAL PARAMETERS
In the framework of the micromagnetic model (lM), the
magnetization
~
M(~r; t) is a continuous vectorial function and
its dynamics is governed by the augmented Landau-Lifshitz
Gilbert (LLG) equation
d~ m
dt
= c
0
~ m
~
H
ef f
a ~ m
d~ m
dt

~s
SO
; (1)
where ~ m(~r; t) =
~
M(~r;t)
M
s
is the normalized local magnetization
with M
s
the saturation magnetization, c
0
is the gyromagnetic
ratio, and a the Gilbert damping parameter.
~
H
ef f
is the effec-
tive eld, derived from system energy density
(
~
H
ef f
=
1
l
0
M
s
d
d~ m
), which apart from the standard exchange,
magnetostatic, uniaxial anisotropy, and Zeeman
a)
Author to whom correspondence should be addressed. Electronic mail:
edumartinez@usal.es
0021-8979/2014/116(2)/023909/11/$30.00 VC
2014 AIP Publishing LLC 116, 023909-1
JOURNAL OF APPLIED PHYSICS 116, 023909 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
130.92.9.55 On: Thu, 28 Aug 2014 04:43:08
contributions also includes the anisotropy exchange
DMI.
1719
In the thin-lm approach (L
z
L
y
, L
x
), the
interfacial DMI energy density
DM
is given by
19,23

DM
= D[m
z
\ ~ m (~ m \)m
z
[; (2)
where D is the DMI parameter describing its intensity. The
DMI effective eld
~
H
DM
is therefore
~
H
DM
=
2D
l
0
M
s
@m
z
@x
~u
x

@m
z
@y
~u
y

@m
x
@x

@m
y
@y

~u
z

: (3)
In the absence of DMI (D=0), the exchange interaction
imposes boundary conditions at the surfaces of the sample
35
such that the magnetization vector does not change along the
surface (
@~ m
@n
=
~
0, where
@
@n
indicates the derivative in the out-
side direction normal to the surface of the sample).
However, in the presence of the interfacial DMI, this bound-
ary condition has to be replaced by
36
@~ m
@n
=
D
2A
~ m ~n ~u
z
( ); (4)
where ~n represents the local unit vector normal to each sam-
ple surface.
The last term in (1) is the SOT, ~s
SO
. Based on experi-
mental observations,
1316
here, we assume a Slonczewski-
like SOT which originates from the SHE,
27,28
where spin-
dependent scattering of an in-plane charge current
(
~
j
a
= j
a
~u
x
) in a heavy metal (e.g., Ta, Pt) generates an out-
of-plane spin current in the ferromagnetic layer. This spin
current is polarized along the transversal direction ~u
y
, per-
pendicular to both the electrical current
~
j
a
= j
a
~u
x
and the
direction of broken inversion symmetry (~u
z
). The corre-
sponding SHE-SOT~s
SO
is
3743
~s
SO
= c
0
~ m (~ m H
SH
~u
y
); (5)
where the amplitude of the SHE effective eld H
SH
is given
by
2,30,31,37
H
SH
=
hh
SH
j
a
2l
0
eM
s
L
z
; (6)
where e <0 is the electron charge, L
z
is the thickness of the
ferromagnetic layer, and h
SH
is the spin Hall angle, which is
dened as the ratio between the spin and charge current
densities.
2,30
We consider an ultrathin ferromagnetic CoFe strip with
high PMA and a cross section of L
y
L
z
=160 nm0.6 nm
on top of a heavy metal layer, either Ta or Pt. Common and
typical high PMA parameters were considered for both
Ta/CoFe and Pt/CoFe samples:
16
saturation magnetization
M
s
= 7 10
5
A=m, exchange constant A = 10
11
J=m, and
uniaxial anisotropy constant K
u
= 4:8 10
5
A=m
3
. The spe-
cic parameters for the Ta-sample (Ta/CoFe/MgO) are a
weak DMI parameter D = 0:05 mJ=m
2
, a negative spin Hall
angle h
SH
=0.11 (see Refs. 14 and 16) and a Gilbert damp-
ing a =0.03.
44,45
On the other hand, the values to mimic the
Pt-sample (Pt/CoFe/MgO) are a strong DMI parameter
D = 1:2 mJ=m
2
, a positive spin Hall angle h
SH
=0.07 (see
Refs. 14 and 16), and a high Gilbert damping a =0.3 (see
Ref. 46). For both Ta and Pt samples, the amplitude of the
DMI ([D[) is high enough to overcome shape anisotropy, and
therefore, the DWs adopt a Neel equilibrium conguration.
14
The micromagnetic (lM) results described hereafter
were obtained by numerically solving the LLG equation (1)
by using GPMagnet, a commercial parallelized nite-
difference micromagnetic solver.
47,48
Samples with
L
x
=4.8 lm in length were discretized in 2D with cells of
Dx =Dy =4 nm in side, and thickness equal to the ferromag-
netic strip thickness (L
z
=0.6 nm). A 6th-order Runge-Kutta
scheme with a time step of 0.5ps was used to solve (1).
Several tests were performed with a cell size of 2 nm con-
rming that similar micromagnetic results are obtained.
III. NUCLEATION OFACHIRAL AND HOMOCHIRAL
WALLS
Before analyzing the CIDWM, we rstly performed a
micromagnetic study of the DW nucleation process by
means of a current pulse owing through an adjacent con-
ductive wire, which is perpendicular to the ferromagnetic
strip and placed above its center (see Fig. 1(a)). The ferro-
magnetic strip is initially uniformly magnetized along the
z >0 direction, and the conductive wire has a cross section
of w
c
t
c
=200 nm640 nm. A current density
~
j
c
= j
c
~u
y
owing through it induces an Oersted eld
~
H
Oe
which was
numerically computed from Biot-Savarts law
35
and included
as an additional contribution to the total effective eld
~
H
ef f
in
the LLG equation (1) (see Ref. 33 for further details). A
square current pulse
~
j
c
(t) = j
c
(t)~u
y
is injected through the
current wire with an amplitude of j
c
= 1:7 10
12
A=m
2
and
duration of 2.5 ns. Except this current pulse, the nucleation
processes were studied in the absence of any other driving
eld nor current.
In the absence of DMI (D=0), two walls are nucleated
at the end of the current pulse with a central domain between
them magnetized along the negative z-axis (see rst set of
snapshots in Fig. 1(b), which shows the three components of
the local magnetization at 2.5 ns, when the current pulse is
switched off). At this instant, the magnetization in the left
DW is not uniform across the strip width, and the right DW
shows a Neel conguration with internal moment magne-
tized along the positive x-axis. The second set of snapshots
in Fig. 1(b) corresponds to t =25 ns, where the internal mag-
netization of both walls points along the transversal direction
(y-axis). These two Bloch walls depict an antiparallel
conguration with ~ m
L
= ~u
y
for the left wall (i =L) and
~ m
R
= ~u
y
for the right wall, and the third set of snapshots in
Fig. 1(b) (at t =500 ns) indicates that the distance between
the two Bloch walls has increased, so the DWs repel each
other. This repulsion between these Bloch walls in the ab-
sence of DMI has been already discussed by Kim et al. in
Ref. 34. However, in their analysis, the repulsion force was
described in term of the dipolar eld that lateral domains
exert on the central one, without considering the relative ori-
entation of the internal DW moments.
More interesting is the analysis of the nucleation proc-
esses in samples with nite DMI, where the internal structure
023909-2 E. Martnez and

O. Alejos J. Appl. Phys. 116, 023909 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
130.92.9.55 On: Thu, 28 Aug 2014 04:43:08
of the nucleated DWs is completely different. Both for
weak-DMI (D = 0:05 mJ=m
2
, Fig. 1(c) for a =0.03, and Fig.
1(d) for a =0.3) and for strong DMI (D = 1:2 mJ=m
2
and
a =0.3, Fig. 1(e)), the left and the right walls present Neel
states with internal magnetization along the x-axis. Now,
these walls are antiparallel each other as due to the chirality
imposed by the DMI. The left wall, which is an up-down
wall, has internal moment ~ m
L
= ~u
x
, whereas the right wall
(down-up wall) is magnetized along the antiparallel direction
(~ m
R
= ~u
x
). The same homochiral Neel conguration was
also obtained for other nucleation current pulses. Our simula-
tions indicate (see third snapshots in Figs. 1(c)1(e) 500 ns
after pulse was switched off) that even in the absence of any
applied current nor eld, the central domain between the
walls, magnetized along the negative z-axis, expands even
when the nucleating pulse is switched off. Therefore, the
homochiral Neel walls move away, indicating that they ex-
perience a repulsion force. A similar repulsion force was
very recently described by Vernier et al.
49
for extended thin-
lms of Ta/CoFeB/MgO samples. Our modeling for conned
multilayers in the presence of the DMI also indicates that the
internal structure of the homochiral DWs does not change
during this repulsion. It was also checked that the central do-
main monotonously expands even when the simulations start
from two homochiral walls without being nucleated by the
pulse. Therefore, the repulsion is not related to the inertia of
the walls gained during the nucleation process, and it must
be related to the antiparallel internal moments of these
homochiral walls, as it will be shown in Sec. IV.
Besides, we noted that the repulsion force depends on the
damping parameter a: the smaller the damping the larger the
distance between walls at a given instant of time (see both
Figs. 1(c) and 1(d) at t =500 ns for a =0.03 and a =0.3,
respectively). As it be will shown in the Sec. IV, the repulsion
force can be described by an effective out-of-plane eld
which pushes the wall along the longitudinal axis (x-axis).
Differently from the model by Kim et al.,
34
where the repul-
sion dipolar eld was assumed to be only dependent on the
relative size of the domains, our simulations indicate that this
longitudinal effective eld depends on the relative orientation
of the internal DW moments and on the distance between
their centers. For a given instant of time after the switching
off of the nucleation pulse, the distance d(t) between walls is
larger for smaller damping parameter a because as in the
standard eld-driven DW motion in the rigid regime,
50
the
repulsion velocity is proportional to the longitudinal effective
eld and inversely proportional to the damping.
IV. REPULSION FORCE BETWEEN HOMOCHIRAL
DWs: MICROMAGNETIC STUDYAND ANALYTICAL
DESCRIPTION
In order to describe and quantify the repulsion force
between the two homochiral walls (D,=0), we considered an
initial state with two homochiral Neel walls separated by
d =256 nm from each other in the Pt-sample (Fig. 2(a)). Left
and right walls depict up-down and down-up congurations,
respectively, so the lateral domains are magnetized along
~u
z
and the central one along ~u
z
. As described above, if
the system evolves freely in the absence of driving forces
(B
z
=0), the two walls would separate from each other (not
shown). However, if a perpendicular eld is applied parallel
to the magnetization in the lateral domains (
~
B
z
= B
z
~u
z
with
B
z
>0), the repulsion between these chiral DWs can be bal-
anced with the force due to the B
z
, which tries to shrink the
central domain. This is clearly seen in Fig. 2(b), which
shows the temporal evolution of the distance between walls
(d(t) =X
R
(t) X
L
(t), with X
R
and X
L
being the position of the
Left and Right DWs, respectively) under different elds. The
distance between walls nally reaches a terminal value, and
this terminal distance decreases as B
z
increases.
Figs. 2(c)2(f) show the terminal magnetization congu-
ration under several applied elds B
z
. These snapshots and a
detailed evaluation the magnetization proles indicate that
the internal magnetization ~ m
i
and the width D of each wall
do not signicantly change with B
z
, and only the distance is
FIG. 1. (a) Schematic illustration of
the ferromagnetic strip, including the
position and dimensions of the current
line, required to nucleate the walls as
described in the text. A square current
pulse
~
j
c
(t) = j
c
(t)~u
y
is injected
through the current wire with an ampli-
tude of j
c
= 1:7 10
12
A=m
2
and du-
ration of 2.5 ns. Micromagnetic
snapshots of the nucleated DWs when
the pulse is switched off (at t =2.5 ns)
at t =25 ns and at t =500 ns for: (b) in
the absence of DMI (D=0) with
a =0.3, (c) for weak DMI D =
0:05 mJ=m
2
with a =0.03, (d) for
strong DMI D = 1:2 mJ=m
2
with
a =0.03, and (e) again for weak DMI
D = 0:05 mJ=m
2
but now with a =0.3.
023909-3 E. Martnez and

O. Alejos J. Appl. Phys. 116, 023909 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
130.92.9.55 On: Thu, 28 Aug 2014 04:43:08
modied. The maximum eld which could be applied with-
out collapsing the two walls was B
z
=106 mT.
The terminal distance between walls (d
eq
) as a function
of B
z
is shown in Fig. 3(a), which indicates that B
z
scales
with d
4
eq
. This is reminiscent of a classical repulsion/attrac-
tion dipolar force between two antiparallel/parallel magnetic
moments aligned along the direction which contains both
moments.
51
These magnetic moments ~l are related to the in-
ternal magnetization of each wall as ~ l
L
= M
s
V~ m
L
and
~ l
R
= M
s
V~ m
R
, where V=DL
y
L
z
is the volume of each wall
and ~ m
i
is the unit vector along the internal DW magnetiza-
tion. It has to be noted that our micromagnetic results of Fig.
3(a) cannot be reproduced by the approach considered by
Kim et al.
34
describing the repulsion force between achiral
walls by just considering the magnetostatic eld generated
by the lateral domains on the central domain (see Eq. (1) in
Ref. 34 and green solid curve in Fig. 3(a)), and therefore,
neglecting the internal magnetization of the walls, which is
essential to describe the full micromagnetic results of Fig.
3(a) as explained below. The results of Fig. 2 indicate that
~ m
i
= 7~u
x
, but in general, the internal DW magnetization
could point along any in-plane direction (for instance, under
in-plane elds B
x
,=0 or B
y
,=0). Therefore, it is worthy to
describe the dipolar force
~
F
dip;ij
which the wall i exerts on
the wall j in terms of their internal magnetization angles U
L
and U
R
, as schematically represented in Fig. 3(b). For the in-
plane internal DW magnetization of these walls, the dipolar
force
~
F
dip;ij
would have both longitudinal x and transverse
y components, but for our purpose of studying the DW
motion along the x-axis, the only relevant one is the x com-
ponent, which can be written as
~
F
dip;ij
= F
dip;ij
~u
ij
=
3l
0
M
s
V ( )
2
4pd
4
sin U
i
sin U
j
2 cos U
i
cos U
j [ [~u
x
; (7)
where i, j: L, R and the unit vector pointing from wall i to
wall j is ~u
ij
= ~u
x
= ~u
ji
. For the case of Fig. 1,
U
L
=180

and U
R
=0

, and therefore (7) provides the man-


ner in which the repulsion force depends on d:
~
F
dip;LR
=
3l
0
M
s
V ( )
2
2pd
4
~u
x
=
~
F
d;RL
. The applied eld
~
B
z
=B
z
~u
z
with B
z
=l
0
H
z
>0, produces an attractive force between the
two walls which is given as
~
F
B
=2l
0
L
y
L
x
M
s
H
z
~u
x
,
5254
and
at the equilibrium, the repulsion dipolar force is balanced
with the compressing external force ([F
dip
[ =[F
B
[).
Therefore, the repulsion between homochiral walls can be
quantied in terms of an effective out-of-plane eld along
the z-axis (
~
H
dip
=H
dip
~u
z
) given by
~
H
dip;ij
= H
dip;ij;z
~u
z
=
F
dip;ij
2l
0
L
y
L
x
M
s
~u
z
; (8)
where F
dip,ij
was dened in (7). The full micromagnetic
results for d
eq
vs B
z
=l
0
H
z
(open symbols in Fig. 3(a)) are
remarkable tted to (7) and (8) (solid line in Fig. 3(a)) with
the DW width D as the only tting parameter, which results
to be D
t
~12 nm. This value is a factor (2p)
1
4
~ 1:6 larger
than the DW width micromagnetically computed according
to the Thiele denition
55
(D ~ 7.6 nm).
The analysis for the strong-DMI Pt-sample
(D = 1:2mJ=m
2
) depicted in Figs. 2 and 3(a) was also per-
formed for a weak-DMI Ta-sample (D = 0:05mJ=m
2
),
obtaining similar results (d
eq
vs B
z
and D
t
~ 12 nm). The
only difference was the maximum eld B
z
which can be
applied without collapsing the two walls, which for the Ta-
sample was B
z
~ 2.2 mT due to the smaller DMI in this sam-
ple. Although here we are studying conned strips, it is also
FIG. 3. (a) Equilibrium distance between DWs d
eq
under static perpendicular
eld B
z
. Symbols correspond to full lM results, whereas the lines (solid black
for the Pt-sample and the dashed red for the Ta-sample) are the 1DM predic-
tions. The solid green line depicts the prediction by the analytical equation (1)
in Ref. 34. Black squares and solid black line correspond to the strong-DMI
Pt-sample with D = 1:2 mJ=m
2
, whereas open circles and dashed red line cor-
respond to the weak-DMI Ta-sample with D = 0:05 mJ=m
2
. (b) Schematic
representation of the DWs and their variables in the 1DM.
FIG. 2. (a) Micromagnetic snapshot of the evaluated system depicting two
homochiral walls separating three perpendicular magnetized domains (up,
down and up, from left to right). The initial distance between the walls is
d =256 nm. The internal DW moments (yellow arrows) point along the
x-axis antiparallel each other: ~ m
L
~u
x
for the left wall and ~ m
R
~u
x
for
the right wall. (b) Temporal evolution of the distance d(t) =X
R
(t) X
L
(t)
between the left and right walls under different perpendicular elds B
z
.
(c)(f) Micromagnetic snapshot of the equilibrium conguration under
static perpendicular elds: B
z
=0.2 mT, B
z
=2.5 mT, B
z
=10 mT, and
B
z
=106 mT. These results correspond to the strong-DMI Pt-sample: a =0.3
and D = 1:2 mJ=m
2
.
023909-4 E. Martnez and

O. Alejos J. Appl. Phys. 116, 023909 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
130.92.9.55 On: Thu, 28 Aug 2014 04:43:08
remarkable that similar values of the destruction eld
(~2 mT) were recently measured by Vernier et al.
49
in
extended Ta/CoFeB/MgO thin-lms.
V. COUPLED CURRENT-INDUCED DW DYNAMICS
ALONG PERFECT SAMPLES
We focus now on the current-driven dynamics of these
coupled chiral walls by the SHE in the presence of longitudi-
nal elds (B
x
). When a static current along the x-axis
(
~
j
a
= j
a
~u
x
) is injected in the multilayer consisting on a heavy
metal and a ferromagnetic strip, the SHE in the heavy metal
layer deects electrons with opposite spin in opposite direc-
tions, giving rise to a spin current that is injected into the
ferromagnet, and exerts a Slonczewski-like torque on the fer-
romagnet magnetization.
2
This SHE torque is the last term at
the rhs in the LLG equation (1) and it is given by (5) and (6).
Therefore, it generates an out-of-plane effective eld on
the DMI-stabilized Neel walls which promotes their dynam-
ics along the strip. This out-of-plane SHE effective eld
acts on each wall (i: L, R) and can be written as
~
H
SHE;i
= ~ m
i
H
SH
~u
y
= H
SH
m
x;i
~u
z
, where m
x,i
represents the
x component of the internal DW moment ~ m
i
. In the absence
of in-plane elds (B
x
=0), both up-down and down-up Neel
walls move along the current in the Pt-sample due to the pos-
itive SHE angle (h
SH
>0). On the contrary, the walls move
along the electron ow along the Ta-sample (h
SH
<0). We
rst considered the weak-DMI Ta-sample with two homochi-
ral walls: up-down at left side and down-up at right side. In
this weak-DMI sample, the CIDWM can be described by a
rigid 1DM. In this 1DM framework, the DW dynamics is
described by four coupled differential equations in terms of
two time-dependent variables: the DW position X
i
=X
i
(t) of
each wall (i: L, R for left and right walls) along the strip axis
(x-axis), and the corresponding DW angles U
i
=U
i
(t), which
are dened as the in-plane (x-y) angle of the internal DW
magnetization (~ m
i
) with respect to the positive x-axis (see
Fig. 3(b)). These rigid-1DM coupled equations, a pair for
each wall (i: L, R), are
14,42,43
(1 a
2
)
_
X
i
= D(QX
A;i
aX
B;i
); (9)
(1 a
2
)
_
U
i
= aX
A;i
QX
B;i
; (10)
the functions X
A,i
and X
B,i
depend on the DW positions
(X
L
, X
R
) and DW angles (U
L
, U
R
), and are given by
X
A;i
U
i
; U
j
; d ( ) =
1
2
c
0
H
K
sin 2 U
i
( ) [ [

p
2
c
0
H
x;i
Q
i
H
D
( ) sin U
i

p
2
c
0
H
y;i
cos U
i
; (11)
X
B;i
U
i
; U
j
; d ( ) = c
0
H
z;i

p
2
c
0
QH
SH
cos U
i
; (12)
where Q
i=L
=1 (up-down wall) and Q
i=R
=1 (down-up
wall). The applied current ows along strip axis,
~
j
a
= j
a
~u
x
with j
a
>0 along the x >0 axis. The amplitude of DMI
effective eld is H
D
=
D
l
0
M
s
D
with D the magnitude of the
DMI.
23
D is the DW width, which is estimated to be
D =

A
K
ef f
q
~ 7:62nm, where K
ef f
= K
u

l
0
M
2
s
2
. The shape
anisotropy eld is H
K
=N
x
M
s
, where N
x
is the magnetostatic
factor given by Ref. 56, N
x
=
L
z
log 2
pD
= 0:0174. (H
x,i
, H
y,i
,
H
z,i
) are the total Cartesian components of the magnetic eld
that each wall i: L, R experiences, including the applied mag-
netic eld, which has Cartesian components (H
x
, H
y
, H
z
),
and the eld due to the coupling between walls. In particular,
the dipolar force (
~
F
dip;ij
= F
dip;ij
~u
ij
) that every DW
exerts on each other is taken into account in the 1DM model
by an additional contribution to the perpendicular applied
eld H
z
as
H
z;i
U
j
; d ( ) = QH
z
H
dip;ji;z
= QH
z

3M
s
V
8pd
4
sin U
j
sin U
i
2 cos U
j
cos U
i ( );
(13)
where H
dip,ji,z
was dened in (8).
Additionally, every wall exerts not only dipolar forces on
the internal moment of each other but it could also exert a
dipolar torque.
51
This dipolar torque (~s
dip;ij
= ~ m
j

~
B
dip;ij
)
modies the externally applied longitudinal H
x
and transverse
H
y
in-plane elds so that
H
x;i
U
j
; d ( ) = H
x

B
dip;ji;x
l
0
= H
x

M
s
V
4pd
3
2 cos U
j
; (14)
H
y;i
U
j
; d ( ) = H
y

B
dip;ji;y
l
0
= H
y

M
s
V
4pd
3
sin U
j
: (15)
While the dipolar force (
~
F
dip;ij
) and the corresponding
out-of-plane eld (H
dip,ji,z
) were found to be essential to
reproduce the micromagnetically computed repulsion
between homochiral walls, the dipolar torques (~s
dip;ij
) and
the corresponding in-plane elds (H
dip,ji,x
and H
dip,ji,y
)
were found negligible, and the results described hereafter
were also obtained by imposing H
dip,ji,x
=H
dip,ji,y
=0.
These 1DM equations describe coupled DW dynamics
in perfect samples (no pinning) and at zero temperature.
However, these pinning and thermal effects can be taken into
account in this 1DM.
52,53,57
In such a case, the out-of-plane
eld H
z,i
has to be replaced by H
z,i
H
p,i
(X
i
) H
th,i
(t), which
includes in addition to H
z,i
the spatial dependent pinning
eld (H
p,i
(X, i)) accounting for local imperfections, and the
thermal eld (H
th,i
(t)) describing the effect of thermal uctu-
ations.
52,53,57
The pinning eld can be derived from an effec-
tive spatial-dependent pinning potential
52,53,57
V
pin,i
(X
i
) as
H
p;i
(X
i
) =
1
2l
0
M
s
L
y
L
z
@V
pin;i
(X
i
)
@X
i
. The thermal eld is assumed
to be a random Gaussian-distributed stochastic process with
zero mean value (H
th;i
(t)) = 0) and uncorrelated in time
(H
th;i
(t)H
th;i
(t
/
)) =
2aK
B
T
l
0
c
0
M
s
DL
y
L
z
d(t t
/
), where K
B
is the
Boltzmann constant and T the temperature). Except where
otherwise indicated, perfect samples (H
p,i
(X
i
) =0) at zero
temperature (H
th,i
(t) =0) are considered. The system of four
coupled equations (9) and (10) for i =L, R with (11), (12),
(14), (15) and (13), were numerically solved by means of a
4th Runge-Kutta algorithm with a time step of 1 ps. Before
describing the current-driven motion, it is also interesting to
mention here that the similar curves to the ones shown in
023909-5 E. Martnez and

O. Alejos J. Appl. Phys. 116, 023909 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
130.92.9.55 On: Thu, 28 Aug 2014 04:43:08
Fig. 3(a) were also obtained by numerically solving these
coupled 1DM equations, which further supports the validity
of the model.
A. Temporal evolution of the DW positions, velocities,
and angles: Weak-DMI
We consider an up-down wall at the left side
(U
L
(0) =180

) and a down-up wall at right side (U


R
(0) =0

).
The initial distance between them is d(0) =X
R
(0) X
L
(0)
=1200nm. The 1DM results for the temporal evolution of the
DWpositions (X
L
(t), X
R
(t)), DWvelocities (v
L
(t), v
R
(t)) and DW
angles (U
L
(t), U
R
(t)) are shown in Fig. 4 under a xed current
density [j
a
[ = 10
9
A=m
2
with either one or the inverse polarity.
In the absence of longitudinal eld (B
x
=0) (see Fig.
4(a)), both left and right DWs reach a steady regime, where
they move with identical velocities along the electron ow
(against the current), and maintaining their initial distance
(d =1200 nm). Note that the initial distance between walls is
very large so the repulsion at rest is negligible. When the
current is applied, the internal DW magnetization rotates
oppositely within the two walls. Under positive current
(along ~u
x
), the internal DW magnetization of the Left wall
rotates clockwise, whereas the right wall rotates the same
angle counter-clockwise. Once the steady regime is reached,
both walls have a no-null and constant component along the
positive transverse axis (y-axis). The sense of rotation
changes when the current is along ~u
x
. This rotations are
due to the SHE, which drives Neel walls but it also exerts a
torque on their internal magnetization. When a longitudinal
eld is applied (B
x
,=0), the dynamics of the two DWs are
different because this eld supports the internal DW magnet-
ization along the x-axis within one wall but opposes to it
within the other wall. If a positive longitudinal eld B
x
>0 is
applied, it favourably acts to the DMI eld in the right wall,
and opposingly to the DMI eld in the left one. In other
words, a positive longitudinal eld is parallel (antiparallel)
to the internal magnetization of right (Left) DW. If the posi-
tive longitudinal eld is smaller than the DMI effective eld
(l
0
H
D
~10 mT) (see Fig. 4(b)), both DWs move along the
same direction, but with different velocities. The DW at the
right side propagates faster than the one at the left, because
the terminal DW angle within the right wall is closer to the
x-axis, while the terminal angle within the left wall has a
larger transverse y component. Therefore, under positive cur-
rent, the central domain contracts, and the two DWs collapse
at a time around 280 ns, whereas under negative current, the
central domain expands. If a longitudinal eld larger than
the DMI effective eld is applied (B
x
>10 mT), the internal
magnetization in the Left wall reverses, and therefore, it prop-
agates along the current sense, and against the right wall (see
Fig. 4(c) for B
x
=15 mT). Note that for B
x
=15 mT>l
0
H
D
,
U
R
~ 0

for the terminal DW angle within the right wall,


whereas U
L
<90

. Under positive current, the walls collapse


at a time around 90 ns. Oppositely, the central domain
expands under negative current.
All these 1DM predictions were also veried by means
of full micromagnetic simulations of a strip with L
x
=4.8 lm
FIG. 4. 1DM results for the temporal
evolution of the DW positions X
L
and
X
R
(top panel), DW velocities v
L
and v
R
(central panels) and DW angles U
L
and
Phi
R
(bottom panels) under a xed cur-
rent density [j
a
[ = 10
9
A=m
2
with both
polarities and several longitudinal elds:
(a) B
x
=0, (b) B
x
=5mT<l
0
H
D
, and
(c) B
x
=15mT>l
0
H
D
for a weak DMI
sample with D = 0:05 mJ=m
2
and
a =0.03.
023909-6 E. Martnez and

O. Alejos J. Appl. Phys. 116, 023909 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
130.92.9.55 On: Thu, 28 Aug 2014 04:43:08
and equal cross section. Representative magnetization snap-
shots in the steady-state regime are shown in Fig. 5 for iden-
tical values of current (j
a
= 610
9
A=m
2
) and elds as those
in preceding Fig. 4.
B. Temporal evolution of the DW positions, velocities,
and angles: Strong-DMI
A similar micromagnetic study was also performed for
the Pt-sample, which is representative of a strong-DMI case
(D = 1:2 mJ=m
2
). As in the former study, the initial distance
between walls was also chosen to be d(0) =1.2 lm, in order
to minimize their repulsion at rest. In the absence of longitu-
dinal elds (B
x
=0), and due to the positive SHE angle
(h
SH
=0.07), the walls propagate now along the current
sense (j
a
= 610
11
A=m
2
), but maintaining the initial distance
(see Figs. 6(b) and 6(c)). The internal DW moments inside
each wall rotate during the transient period before reaching
the terminal steady-regime. However, the DW moments
rotate oppositely, if compared with the case of weak-DMI,
due to the opposite sign of the SHE angle. The most remark-
able difference with the Ta-sample case is the rotation of the
DW normal plane (~n
i
) with respect to the x-axis due to the
strong-DMI. This DW tilting derives from the fact that the
DMI tends to maintain ~ m
i
parallel to ~n
i
,
16,42
but when a cur-
rent is applied, the SHE also cants ~ m
i
away from 6~u
x
: strong
DMI causes the DW to tilt in order to minimize the angle
between ~ m
i
and ~n
i
. For positive (negative) current, the left
wall tilts counter-clockwise (clockwise), whereas the right
wall tilts clockwise (counter-clockwise). Note also that these
trends for the DW tilting mimic the transient rotations of the
internal DW angles (see m
x
(~r) and m
y
(~r) in 6(b) and 6(c)).
Panels (d) and (e) in Fig. 6 correspond to propagating
steady-states for equal positive and negative currents
([j
a
[ = 10
11
A=m
2
) as (b) and (c), but in the presence of a posi-
tive longitudinal eld of B
x
=100 mT, which is around a 50%
smaller than the DMI effective eld (l
0
H
D
~ 200mT).
16
This
longitudinal eld is also of the same order of magnitude than
the uniaxial PMA effective eld (l
0
H
eff ;z
=
2K
ef f
M
s
~ 492 mT),
and therefore, the domains at both sides of the walls depict
some canting from the easy z-axis, with signicant component
along the longitudinal x direction (note the red color of the
domains of the m
x
density plot in Figs. 6(d) and 6(e)). As
expected and described for the Ta-sample, a eld B
x
of
100 mT supports the internal DW magnetization of the right
wall (~ m
R
), and opposes to the one within the left wall (~ m
L
).
Therefore, the right wall moves faster than the left one
([v
R
[ > [v
L
[). For a positive current (j
a
= 10
11
A=m
2
), the
two walls move along the current sense (~u
x
), but the central
domain expands since [v
R
[ > [v
L
[ (see Fig. 6(d)). For a nega-
tive current (j
a
= 10
11
A=m
2
), the two DWs move along
~u
x
, but the central domain contracts because [v
R
[ > [v
L
[, as
for positive currents (see Fig. 6(e)).
Due to the strong-DMI of this Pt-sample, much higher lon-
gitudinal elds are needed to reverse the internal DW magnet-
ization within the left wall and, consequently, its direction of
motion. This can be done by applying a eld of B
x
=300mT,
which is around a 50% larger than the DMI effective eld
(l
0
H
D
~200 mT).
16
The steady-state snapshots are shown in
panels of Figs. 6(f) and 6(g) for the positive and negative cur-
rents, respectively. Now, the central domain contracts under the
positive current, and expands under the negative one. Note
again that in both cases [v
R
[ > [v
L
[, since the internal magnet-
ization of the right wall is supported by the positive longitudi-
nal eld. It is also interesting to see that DW tilting is larger for
the left wall than for the right wall by the same reason.
VI. COUPLED CURRENT-INDUCED DW DYNAMICS
ALONG REALISTIC SAMPLES
Previous results have been performed under ideal condi-
tions, without disorder. However, it is well known that in out-
FIG. 5. Steady-state micromagnetic congurations corresponding to a weak-
DMI Ta-sample with D = 0:05 mJ=m
2
; h
SH
= 0:11, and a =0.03. The initial
distance between walls is d(0) =1.2lm. The applied current is j
a
= 610
9
A=m
2
.
FIG. 6. Steady-state micromagnetic congurations corresponding to a
strong-DMI Pt-sample with D = 1:2 mJ=m
2
; h
SH
= 0:07, and a =0.3. The
initial distance is d(0) =1.2 lm. The applied current is j
a
= 610
11
A=m
2
.
023909-7 E. Martnez and

O. Alejos J. Appl. Phys. 116, 023909 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
130.92.9.55 On: Thu, 28 Aug 2014 04:43:08
of-plane magnetized samples, the DWs can interact with local
defects, and therefore, a description of the coupled CIDWM
under realistic conditions is of high interest. Here, the disorder
was taken into account by assuming a random edge roughness
with a typical grain size of D
g
=4 nm at both sides of the
strips (see Ref. 52 for details). The effect of such random dis-
order (edge roughness) is also qualitatively consistent with
nano-scale defects distributed throughout experimental sam-
ples,
14,16
and it produces a pinning of the DWs motion, thus
avoiding their free propagation. For the Ta-sample, a mini-
mum current density of j
dep
= 1:1 10
11
A=m
2
is required to
promote DWs propagation at zero temperature in the absence
of longitudinal elds. From this point on, our attention will be
restricted to this weak-DMI sample.
The micromagnetically computed temporal evolution of
the left X
L
and right X
R
DW positions are shown in Fig. 7
under the effect of a xed current of [j
a
[ = 5 10
10
A=m
2
and
several longitudinal elds B
x
. In the absence of longitudinal
eld (see Fig. 7(a) for B
x
=0), the two DWs do not propagate
along the strip, and both of them remain pinned, close to their
corresponding initial positions, due to the local pinning
derived from local edge roughness. Under a small positive
eld close to the effective DMI eld (B
x
=10 mT, see Fig.
7(b)), only the right wall (with internal DW magnetization ~ m
R
parallel to B
x
) propagates, whereas the left wall (with internal
DW magnetization ~ m
L
antiparallel to B
x
) remains pinned (see
Fig. 7(b)). Under positive current (j
a
= 5 10
10
A=m
2
) and
positive eld (B
x
=10 mT), the right wall propagates with
negative velocity toward the position of the pinned left wall,
and they nally annihilate because the driving SHE force is
larger than the repulsion force. Under negative current
(j
a
= 5 10
10
A=m
2
), the right wall propagates along the
electron ow (with v
R
>0), and therefore, it monotonously
separates from the still pinned left wall. If the eld B
x
is
reversed, so that B
x
=10 mT (see Fig. 7(c)), the left wall
now propagates, while the right wall remains pinned. These
results indicate that in real samples, a longitudinal eld paral-
lel to the internal DW moment assists the depinning and
subsequent propagation. Similar results are also obtained
when the amplitude of the longitudinal eld doubles
([B
x
[ = 20 mT), for an identical value of the current. Only the
wall whose internal magnetization is aligned parallel to the
longitudinal eld propagates along the strip with a direction
imposed by the sign of the current (see Figs. 7(d) and 7(e)).
However, if the longitudinal eld is further increased
([B
x
[ = 30 mT), both the left and the right walls become
depinned and they propagate along the strip (see Figs. 7(f) and
7(g)). Several micromagnetic snapshots corresponding to
these micromagnetic simulations at 20 ns are shown in Fig. 8.
Note again that for (j
a
>0, B
x
<0) and (j
a
<0, B
x
>0), the
walls move away from each other (the central domain
expands), whereas for (j
a
>0, B
x
>0) and (j
a
<0, B
x
<0), the
walls come closer (the central domain shrinks). Finally
for (j
a
= 65 10
10
A=m
2
; B
x
= 730 mT), the walls collapse
after 20ns.
The micromagnetically computed averaged velocities of
the left and right walls along a rough Ta strip are shown in
FIG. 7. Micromagnetic results for the temporal evolution of the DW positions
X
L
and X
R
along a weak-DMI strip (Ta-sample: D=0:05mJ=m
2
; h
SH
=0:11,
and a=0.03) with edge roughness (D
g
=4nm). The initial distance is
d(0)=1.2lm. The applied current is j
a
=6510
10
A=m
2
and reversal longitu-
dinal elds are also applied: (a) B
x
=0, (b) B
x
=10mT, (c) B
x
=10mT, (d)
B
x
=20mT, (e) B
x
=20mT, (f) B
x
=30mT, and (g) B
x
=30mT.
FIG. 8. Micromagnetic congurations after 20 ns corresponding to a weak-
DMI Ta-sample (D = 0:05 mJ=m
2
; h
SH
= 0:11, and a =0.03) along a strip
with edge roughness (D
g
=4 nm). The initial distance between the up-down
(left) and the down-up (right) walls is d(0) =1.2 lm. The applied current is
(a) j
a
= 5 10
10
A=m
2
and (b) j
a
= 5 10
10
A=m
2
. Several positive
and negative longitudinal elds B
x
are also applied.
023909-8 E. Martnez and

O. Alejos J. Appl. Phys. 116, 023909 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
130.92.9.55 On: Thu, 28 Aug 2014 04:43:08
Fig. 9 as a function of B
x
, for a xed amplitude of the current
([j
a
[ = 5 10
10
A=m
2
) with the two possible polarities. It
must be noted that the averaged velocities of each DW were
calculated from their total displacement after 50 ns when the
two DWs move away from each other. On the contrary,
when the two DWs come closer to each other, the averaged
velocities depicted in Fig. 9 were estimated from the dis-
placement before collapsing. A null velocity is plotted when
the DW remains pinned. These results indicate that the longitu-
dinal eld introduces an asymmetry for the propagation of the
DWs. For instance, under positive current
(j
a
= 5 10
10
A=m
2
), the up-down left wall is depinned and
propagates with positive velocity only if a critical positive eld
of B
x,c
~ 25mT is applied. Conversely, a much more reduced
eld (B
x
=10mT) is needed to overcome the local depinning
and promote its propagation with negative velocity for the
same positive current j
a
= 5 10
10
A=m
2
(see lled black
squares in Fig. 9). The same behavior is observed for the
down-up right wall, but in this case, the critical depinning eld
is smaller when it is applied along the negative direction (see
lled red circles in Fig. 9). These results reverse with the polar-
ity of the current (see open symbols in Fig. 9 for
j
a
= 5 10
10
A=m
2
): the up-down left wall propagates with
positive velocity (along the current) under a negative eld with
smaller amplitude (B
x
=10mT) than in the case of propaga-
tion with negative velocity (against the current) forced by a
positive eld (B
x
=20 mT).
The asymmetry in the DW depinning observed for the
left and the right walls in the presence of longitudinal elds
is not related to the details of the random pinning potential.
This was conrmed by evaluating other randomly generated
edge roughness patterns and obtaining results similar to
those of Fig. 9. Indeed, the same asymmetry was also
observed by analyzing the CIDWM in the framework of the
1DM, considering a xed periodic pinning potential
V
pin
(X
i
) = V
0
sin
2
(p
X
i
p
) as described in Sec. V, and studying
the dynamics for up-down and down-up walls independently
(single-wall case). In the 1DM, the pinning potential V
pin
(X
i
)
is periodic, and in order to describe the micromagnetic
results obtained for D
g
=4 nm, an amplitude of this pinning
potential of V
0
=20 10
21
J and a spacial periodicity of
p =21 nm have been considered (see details in Sec. V). The
results are depicted in Figs. 10(a) and 10(b) for both positive
(j
a
= 5 10
10
A=m
2
) and negative (j
a
= 5 10
10
A=m
2
)
currents, respectively.
In the absence of pinning (perfect samples with
V
pin
(X) =0, solid lines in Fig. 10), both the up-down and the
down up-walls move with equal velocity for B
x
=0. The ve-
locity of the up-down and down-up walls is modied by
B
x
,=0 as formerly described and as it is also clearly shown
in Figs. 10(a) and 10(b). However, in the presence of pinning
(rough samples with V
pin
(X) ,=0, lines with small symbols in
Fig. 10), there is a minimum threshold for the longitudinal
eld B
x
in order to achieve sustained DW motion, and this
threshold value changes for positive and negative longitudi-
nal elds. For instance, under positive current (Fig. 10(a)),
the up-down wall remains pinned for 2 mT_B
x
_
20 mT, whereas the down-up wall does for 20 mT_B
x
_
2 mT. The same pinning ranges (2 mT_B
x
_20 mT
FIG. 9. Micromagnetic results for the average DW velocities as a function
of B
x
along a weak-DMI strip (Ta-sample: D = 0:05 mJ=m
2
; h
SH
= 0:11,
and a =0.03) with edge roughness (D
g
=4 nm). The initial distance is
d(0) =1.2 lm. The applied current is j
a
= 65 10
10
A=m
2
.
FIG. 10. 1DM description of CIDWM
under longitudinal elds B
x
for both
positive (j
a
= 5 10
10
A=m
2
) and
negative (j
a
= 5 10
10
A=m
2
) cur-
rents. Solid lines correspond to the
1DM results in the absence of pinning
(V
pin
(X) =0). Small symbols corre-
spond to the 1DM results, including
pinning, as described along the text
with V
0
=20 10
21
J and p =21 nm.
Big symbols correspond to the full
micromagnetic (lM) results of former
Fig. 9, when the two walls do not
collapse.
023909-9 E. Martnez and

O. Alejos J. Appl. Phys. 116, 023909 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
130.92.9.55 On: Thu, 28 Aug 2014 04:43:08
for the up-down wall and 20 mT _ B
x
_ 2 mT for the
down-up wall) also appear for negative current (Fig. 10(b)).
These results are in a good qualitative agreement with full
micromagnetic simulations of the two coupled walls (see
Fig. 9 and big symbols in Figs. 10(a) and 10(b)), and there-
fore, we can conclude that in general, a longitudinal eld
parallel to the internal DW magnetization assists the DW
depinning, whereas an antiparallel eld opposes to it. The
driving force here is always the SHE, which, as discussed in
Secs. II and V, produces an effective out-of-plane eld
~
H
SHE;i
= H
SH
m
x;i
~u
z
which is proportional to the current (H
SH
j
a
) and to the x component of the internal DW moment
(m
x,i
), and a longitudinal eld B
x
supports or acts against m
x,i
being parallel or antiparallel to each other, respectively.
Therefore, B
x
can increase or reduce the depinning current.
These numerical results can be relevant to describe very
recent experimental observations,
58
which also depict similar
asymmetries in the longitudinal eld required to promote the
depinning for up-down and down-up walls under positive
and negative currents.
VII. CONCLUSIONS
The nucleation of DWs in ultrathin strips with high
PMA has been micromagnetically studied both in the ab-
sence and in the presence of DMI. In this latter case, the
DWs naturally adopt an homochiral conguration, with in-
ternal DW moment antiparallel each other. It was found that
they repel when they are close to each other, and this effect
has been accurately described in terms of the dipolar force
between the magnetic moments of the walls. This force,
which couples the dynamics of these chiral walls, has been
described by means of micromagnetic modeling and 1DM
points of views.
The coupled current-induced dynamics by the spin Hall
effect of these two homochiral walls has been studied. In
perfect samples and in the absence of longitudinal eld, both
DWs move with the same velocity, along the current or
against it depending on the sign of the spin Hall angle. When
a longitudinal eld smaller than the DMI effective eld
([B
x
[ < [B
D
[) is applied, the DWs still move along the same
direction as for B
x
=0 but now with different velocity, being
always faster the wall with initial internal magnetization par-
allel to the longitudinal eld B
x
. In this case, and although
both walls move in the same direction, the different DW
velocities can result in a monotonously separation or in col-
lapsing the walls. If B
x
is larger than the effective DMI
([B
x
[ > [B
D
[), the chirality of the wall initially magnetized
antiparallely to B
x
reverses, and therefore, its sense of
motion.
These dynamical processes were also studied in the
presence of edge roughness, which results in local pinning
and opposes to the free DW propagation of the walls. Our
micromagnetic and 1DM results indicate that also the depin-
ning current is inuenced by the longitudinal eld, taking
place in a different manner for both walls, which introduces
an asymmetry in the current-induced depinning. It was found
that, in general, the DW with internal moment initially paral-
lel to B
x
depins under smaller eld. Very recent
measurements by Lo Conte et al.
58
depict similar trends, and
therefore our results are of relevance to provide a plausible
interpretation of these experimental observations. Besides,
the analysis carried out here also suggests that the coupled
current-induced DW motion by the spin Hall could be used
to develop novel DW-based devices, where either the rela-
tive motion of the two walls or even their collapsing can be
simply controlled by means of longitudinal elds.
ACKNOWLEDGMENTS
This work was supported by Project No. MAT2011-
28532-C03-01 from Spanish Government and Project No.
SA163A12 from Junta de Castilla y Leon.
1
P. Gambardella and I. M. Miron, Philos. Trans. R. Soc., A 369, 3175
(2011).
2
L. Liu, C.-F. Pai, Y. Li, H. W. Tseng, D. C. Ralph, and R. A. Buhrman,
Science 336, 555 (2012).
3
X. Fan, J. Wu, Y. Chen, M. J. Jerry, H. Zhang, and J. Q. Xiao, Nat.
Commun. 4, 1799 (2013).
4
P. M. Haney, H.-W. Lee, K.-J. Lee, A. Manchon, and M. D. Stiles,
Current induced torques and interfacial spin-orbit coupling: Semiclassical
modeling, Phys. Rev. B 87, 174411 (2013).
5
J. Kim et al., Layer thickness dependence of the current-induced effective
eld vector in Ta vertical bar CoFeB vertical bar MgO, Nature Mater. 12,
240245 (2013).
6
K. Garello et al., Symmetry and magnitude of spin-orbit torques in ferro-
magnetic heterostructures, Nat. Nanotechnol. 8, 587593 (2013).
7
X. Fan, H. Celik, J. Wu, C. Ni, K.-J. Lee, V. O. Lorenz, and J. Q. Xiao,
Nat. Commun. 5, 3042 (2014).
8
S. S. P. Parkin, M. Hayashi, and L. Thomas, Science 320, 190 (2008).
9
K.-J. Kim, J.-C. Lee, S.-J. Yun, G.-H. Gim, K.-S. Lee, S.-B. Choe, and K.-
H. Shin, Appl. Phys. Express 3, 083001 (2010).
10
A. Fert, V. Cros, and J. Sampaio, Nat. Nanotechnol. 8, 152156 (2013).
11
J. Sampaio, V. Cros, S. Rohart, A. Thiaville, and A. Fert, Nat.
Nanotechnol. 8, 839844 (2013).
12
I. M. Miron, T. Moore, H. Szambolics, L. D. Buda-Prejbeanu, S. Auffret,
B. Rodmacq, S. Pizzini, J. Vogel, M. Bonm, A. Schuhl, and G. Gaudin,
Nature Mater. 10, 419 (2011).
13
P. P. J. Haazen, E. Mure, J. H. Franken, R. Lavrijsen, H. J. M. Swagten,
and B. Koopmans, Nature Mater. 12, 299 (2013).
14
S. Emori, U. Bauer, S.-M. Ahn, E. Martinez, and G. S. D. Beach, Nature
Mater. 12, 611 (2013).
15
K.-S. Ryu, L. Thomas, S.-H. Yang, and S. Parkin, Nat. Nanotechnol. 8,
527 (2013).
16
S. Emori, E. Martinez, U. Bauer, S.-M. Ahn, and G. S. D. Beach, Spin
Hall torque magnetometry of Dzyaloshinskii domain walls, e-print
arXiv:1308.1432 (2013).
17
T. Moriya, Phys. Rev. Lett. 4, 228 (1960).
18
M. Bode, M. Heide, K. von Bergmann, P. Ferriani, S. Heinze, G.
Bihlmayer, A. Kubetzka, O. Pietzsch, S. Blugel, and R. Wiesendanger,
Nature 447, 190 (2007).
19
M. Heide, G. Bihlmayer, and S. Blugel, Phys. Rev. B 78,140403 (2008).
20
X. Z. Yu, Y. Onose, N. Kanazawa, J. H. Park, J. H. Han, Y. Matsui, N.
Nagaosa, and Y. Tokura, Nature 465, 901 (2010).
21
S. Heinze, K. von Bergmann, M. Menzel, J. Brede, A. Kubetzka, R.
Wiesendanger, G. Bihlmayer, and S. Blugel, Nature Phys. 7, 713, (2011).
22
S. X. Huang and C. L. Chien, Phys. Rev. Lett. 108, 267201 (2012).
23
A. Thiaville, S. Rohart, E. Jue, V. Cros, and A. Fert, Europhys. Lett. 100,
57002 (2012).
24
G. Chen, J. Zhu, A. Quesada, J. Li, A. T. NDiaye, Y. Huo, T. P. Ma, Y.
Chen, H. Y. Kwon, C. Won, Z. Q. Qiu, A. K. Schmid, and Y. Z. Wu,
Phys. Rev. Lett. 110, 177204 (2013).
25
S. Zhang and Z. Li, Phys. Rev. Lett. 93, 127204 (2004).
26
A. Thiaville, Y. Nakatani, J. Miltat, and Y. Suzuki, Europhys. Lett. 69,
990 (2005).
27
M. Dyakonov and V. Perel, JETP Lett. 13, 467 (1971).
28
J. E. Hirsch, Phys. Rev. Lett. 83, 1834 (1999).
29
B. Gu, I. Sugai, T. Ziman, G. Y. Guo, N. Nagaosa, T. Seki, K. Takanashi,
and S. Maekawa, Phys. Rev. Lett. 105, 216401 (2010).
023909-10 E. Martnez and

O. Alejos J. Appl. Phys. 116, 023909 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
130.92.9.55 On: Thu, 28 Aug 2014 04:43:08
30
L. Liu, T. Moriyama, D. C. Ralph, and R. A. Buhrman, Phys Rev. Lett.
106, 036601 (2011).
31
K. Kondou, H. Sukegawa, S. Mitani, K. Tsukagoshi, and S. Kasai, Appl.
Phys. Express 5, 073002 (2012).
32
T. Koyama, D. Chiba, K. Ueda, K. Kondou, H. Tanigawa, S. Fukami, T.
Suzuki, N. Ohshima, N. Ishiwata, Y. Nakatani, K. Kobayashi, and T. Ono,
Nature Mater. 10, 194 (2011).
33
J. L. Prieto, M. Munoz, and E. Martinez, Phys. Rev. B 83, 104425 (2011).
34
K.-J. Kim, K.-W. Moon, K.-S. Lee, and S.-B. Choe, Nanotechnology 22,
025702 (2011).
35
G. Bertotti, Hysteresis in Magnetism: For Physicists, Materials Scientists
and Engineers (Academic Press, San Diego, CA, 1998).
36
S. Rohart and A. Thiaville, Phys. Rev. B 88, 184422 (2013).
37
S.-M. Seo, K.-W. Kim, J. Ryu, H.-W. Lee, and K.-J. Lee, Appl. Phys. Lett.
101, 022405 (2012).
38
E. Martinez and G. Finocchio, IEEE Trans. Magn. 49(7), 3105 (2013).
39
E. Martinez, G. Finocchio, L. Torres, and L. Lopez-Diaz, AIP Adv. 3,
072109 (2013).
40
E. Martinez, S. Emori, and G. S. D. Beach, Appl. Phys. Lett. 103, 072406
(2013).
41
A. V. Khvalkovskiy, V. Cros, D. Apalkov, V. Nikitin, M. Krounbi, K. A.
Zvezdin, A. Anane, J. Grollier, and A. Fert, Phys. Rev. B 87, 020402(R) (2013).
42
O. Boulle, S. Rohart, L. D. Buda-Prejbeanu, E. Jue, I. M. Miron, S. Pizzini,
J. Vogel, G. Gaudin, and A. Thiaville, Phys. Rev. Lett. 111, 217203 (2013).
43
E. Martinez, S. Emori, N. Perez, L. Torres, and G. S. D. Beach, J. Appl.
Phys. 115, 213909 (2014).
44
S. Ikeda, K. Miura, H. Yamamoto, K. Mizunuma, H. D. Gan, M. Endo, S.
Kanai, J. Hayakawa, F. Matsukura, and H. Ohno, Nature Mater. 9,
721724 (2010).
45
S. Iihama, Q. Ma, T. Kubota, S. Mizukami, Y. Ando, and T. Miyazaki,
Appl. Phys. Express 5, 083001 (2012).
46
A. J. Schellekens, L. Deen, D. Wang, J. T. Kohlhepp, H. J. M. Swagten,
and B. Koopmans, Appl. Phys. Lett. 102, 082405 (2013).
47
GPMagnet a commercial parallelized micromagnetic solver. See http://
www.goparallel.net/index.php/en/gp-software.html for details about the
GPMagnet.
48
L. Lopez-Diaz, D. Aurelio, L. Torres, E. Martinez, M. A. Hernandez-
Lopez, J. Gomez, O. Alejos, M. Carpentieri, G. Finocchio, and G.
Consolo, J. Phys. Appl. Phys. 45, 323001 (2012).
49
N. Vernier, J.-P. Adam, S. Eimer, G. Agnus, T. Devolder, T. Hauet, B.
Ocker, F. Garcia, and D. Ravelosona, Appl. Phys. Lett. 104, 122404
(2014).
50
A. Thiaville and Y. Nakatani, in Spin Dynamics in Conned Magnetic
Structures III, edited by B. Hillebrands and A. Thiaville (Springer, 2005),
Vol. 101, pp. 161205.
51
D. J. Grifths, Introduction to Electrodynamics (Pearson Education,
2008).
52
E. Martinez, L. Lopez-Diaz, L. Torres, C. Tristan, and O. Alejos, Phys.
Rev. B 75, 174409 (2007).
53
E. Martinez, L. Lopez-Diaz, O. Alejos, L. Torres, and C. Tristan, Phys.
Rev. Lett. 98, 267202 (2007).
54
E. Martinez, L. Lopez-Diaz, O. Alejos, and L. Torres, J. Appl. Phys. 106,
043914 (2009).
55
A. A. Thiele, Phys. Rev. Lett. 30, 230 (1973).
56
S. V. Tarasenko, A. Stankiewicz, V. V. Tarasenko, and J. Ferre, J. Magn.
Magn. Mater. 189, 19 (1998).
57
E. Martinez, J. Phys.: Condens. Matter 24, 024206 (2012).
58
See Fig. 1 of abstract titled Efcient current-induced domain wall motion
and magnetization switching due to spin orbit torques by R. Lo Conte, T.
Schulz, Su-Jung Noh, M. Klaeui, A. Mihai, A. Hrabec, C. Marrows, and
T. Moore, http://intermag2014.ifw-dresden.de/index.php?id=33 (Session
HC-Spin Orbitronics: DW motion and switching), Intermag 2014.
023909-11 E. Martnez and

O. Alejos J. Appl. Phys. 116, 023909 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
130.92.9.55 On: Thu, 28 Aug 2014 04:43:08

You might also like