Non-equilibrium transport in topologically non-trivial

systems

Dissertation by

Sumit Ghosh

In Partial Fulfillment of the Requirements

For the Degree of

Doctor of Philosophy

King Abdullah University of Science and Technology

Thuwal, Kingdom of Saudi Arabia

February, 2019
 2

EXAMINATION COMMITTEE PAGE

The dissertation of Sumit Ghosh is approved by the examination committee

Committee Chairperson: Prof. Aurelien Manchon

Committee Co-Chair: Prof. Udo Schwingenschlögl
Committee Members: Prof. Ying Wu, Prof. Branislav K. Nikolic
 3

©February, 2019

Sumit Ghosh

All Rights Reserved

 4

ABSTRACT

Non-equilibrium transport in topologically non-trivial systems

Sumit Ghosh

One of the most remarkable achievements of modern condensed matter physics is

the discovery of topological phases of matter. Materials in a non-trivial topological
phase or the topological insulators can be distinguished by their unique electronic and
transport properties which are indifferent to different types of perturbations and thus
open new routes towards the dissipationless transport. Explaining their properties
requires proper involvement of relativistic approach as well as topological analysis.
Among different classes of topological insulators, the Z2 topological insulators have
drawn special attention due to their strong spin-orbit coupling which makes them
a promising candidate for spintronics application, especially for magnetic memory
devices. Due to their inherent strong spin-orbit coupling, they provide an efficient
way to manipulate electronic spin with an applied electric field via spin orbit torque.
The topological insulators have been found to be far more superior in manipulating
the magnetic order parameter of a ferromagnet compared to the conventional heavy
metals like platinum or tantalum.
Another milestone in magnetic memory devices is marked by the introduction of
antiferromagnetic memory devices which has not drawn any attention for long time
as they can not be controlled by an applied magnetic field. Recently it has been found
that in case of a non-centrosymmetric antiferromagnet, the magnetic order parameter
can be manipulated by with spin-orbit torque which also have been verified experi-
mentally. The advantages of antiferromagnetic devices over ferromagnetic devices are
that they allow faster switching speed and they are immune to an external magnetic
field which are two highly solicited properties for next generation spintronic devices.
 5
This thesis is focused on understnading the transport properties in topologically
nontrivial materials and their interface with different magnetic matrial. We use sim-
plifed contninuum model as well as tight binding models to capture the salient features
of these systems. Using non-equilibrium Green’s function we explore their transport
properties as well as spin-charge conversion mechanism. Our finding would provide a
better understanding of these new class of materials and thus would be instrumental
to discover new mechanisms to manipulate their properties.
 6

ACKNOWLEDGEMENTS

To begin with, I would like to thank my mentor Prof. Aurelien Manchon for his
guidance. Without him, the thesis would never have come to reality. I would like to
thank Prof. Udo Schwingenschlögl for his support and his sporting attitude. I am
thankful to all my friends and colleagues with whom I travelled this long journey.
From that list, I would especially like to mention Hang Li, Sergiy, Joshua, Maxim,
Nirpendra, Maha, Aleksandra, Fengjun, Guilhem and Joana. I would like to thank
Frederic Piechon for the illuminating discussion. I would like to thank Prof. Branislav
Nikolic for his warm hospitality and insightful views. I would like to thank the KAUST
supercomputing team for their assistance. I would like to thank all the housekeeping
staffs for keeping me accompanied and my desk clean. Finally, I would like to thank
the rest of the universe for being as it is.
 7

TABLE OF CONTENTS

Examination Committee Page 2

Copyright 3

Abstract 4

Acknowledgements 6

List of Figures 10

List of Tables 19

1 Introduction 20
1.1 The hall of Hall effects . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.1.1 Quantum Hall effect . . . . . . . . . . . . . . . . . . . . . . . 21
1.1.2 Anomalous Hall Effect . . . . . . . . . . . . . . . . . . . . . . 23
1.1.3 Spin Hall Effect . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.1.4 Skyrmion/topological Hall effect . . . . . . . . . . . . . . . . . 25
1.2 Spin orbit coupling in different materials . . . . . . . . . . . . . . . . 26
1.2.1 Heavy metal interfaces . . . . . . . . . . . . . . . . . . . . . . 27
1.2.2 Topological Insulator surfaces . . . . . . . . . . . . . . . . . . 28
1.3 Spin transfer and spin orbit torque . . . . . . . . . . . . . . . . . . . 32
1.4 Spin-orbit torque in different materials . . . . . . . . . . . . . . . . . 33
1.4.1 Spin-orbit torque with ferromagnets . . . . . . . . . . . . . . . 33
1.4.2 Spin-orbit torque with antiferromagnet . . . . . . . . . . . . . 35
1.5 Motivation and outline of the thesis . . . . . . . . . . . . . . . . . . . 36

2 Theoretical Prerequisites 40
2.1 Berry phase, Berry curvature and Chern number . . . . . . . . . . . . 40
2.2 Linear response theory . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.3 Transport in quasi one dimensional system . . . . . . . . . . . . . . . 48
2.4 Square Lattice in uniform magnetic field: Tight binding model . . . . 52
 8
2.5 Honeycomb Lattice . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.5.1 Haldane model . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.5.2 Kane-Mele model . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.5.3 Buckled Honeycomb Lattice . . . . . . . . . . . . . . . . . . . 60
2.6 Two dimensional topological insulator . . . . . . . . . . . . . . . . . . 63
2.7 Three dimensional topological insulators . . . . . . . . . . . . . . . . 65

3 Bulk and edge modes in honeycomb lattice and their impact on

wavepacket dynamics 68
3.1 Effective Hamiltonian for zigzag graphene nano ribbon . . . . . . . . 70
3.2 Evolution of wave packet and zitterbewegung in graphene zigzag nanorib-
bon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

4 Probing topological phases with wavepacket dynamics 80

4.1 Low energy Hamiltonian, eigenvalues and eigenstates of buckled hon-
eycomb lattice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.2 Evolution of wavepacket and zitterbewegung in buckled honeycomb lattice 82
4.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

5 Ferromagnetic and antiferromagnetic topological insulator 90

5.1 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.2 Robustness of different magnetic configurations . . . . . . . . . . . . 94
5.3 Non-equilibrium spin density and Spin-orbit torque . . . . . . . . . . 96
5.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

6 Topological insulator-ferromagnet heterostructure 103

6.1 Model and Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.1.1 Tight Binding Model . . . . . . . . . . . . . . . . . . . . . . 106
6.1.2 Non-equilibrium transport formalism . . . . . . . . . . . . . . 110
6.2 Spin Hall current, non-equilibrium spin density and SOT in topological
insulator - ferromagnet heterstructure . . . . . . . . . . . . . . . . . . 113
6.2.1 Spin Hall conductance in the bulk . . . . . . . . . . . . . . . . 114
6.2.2 Surface versus Bulk Transport . . . . . . . . . . . . . . . . . . 115
6.2.3 Intrinsic versus Extrinsic Spin-Orbit Torque . . . . . . . . . . 120
6.2.4 Understanding the torque efficiency . . . . . . . . . . . . . . . 121
6.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
 9
7 Topological insulator-antiferromagnetic heterostructure 127
7.1 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
7.2 Result and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
7.2.1 Spatial distribution of spin density and conductivity . . . . . . 133
7.2.2 Fermi surface and sea contributions to Sx . . . . . . . . . . . . 136
7.2.3 Effect of AF-TI coupling . . . . . . . . . . . . . . . . . . . . . 139
7.2.4 Coupling to surface vs bulk TI states . . . . . . . . . . . . . . 141
7.2.5 Effect of impurity . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.2.6 Angular dependence of torques . . . . . . . . . . . . . . . . . 143
7.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

8 Concluding Remarks 148

8.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
8.2 Future Research Work . . . . . . . . . . . . . . . . . . . . . . . . . . 149

References 150

List of publications 172

 10

LIST OF FIGURES

1.1 Schematic representation of Hall effect (HE), quantum Hall effect (QHE),
anomalous Hall effect (AHE) and spin Hall effect (SHE). . . . . . . . 21
1.2 Klitzing’s measurement and notes on Quantum Hall effect . . . . . . 22
1.3 Optical detection of spin hall effect. . . . . . . . . . . . . . . . . . . . 25
1.4 Schematic of topological Hall effect. The non-collinear magnetisation
of skyrmion creates an emergent electromagnetic force for an incoming
electron which creates a Hall effect. The resulting spin transfer torque
causes the skyrmion to undergo a deflection in opposite direction [42]. 26
1.5 Variation of ζ with atomic number. Data taken from [51]. Red, green,
blue and gray region denote s, p, d and rare earth elements. . . . . . 28
1.6 (a) Spin polarisation at Fermi level Au(111) surface state excited by p-
polarized 6.05 eV photons. (b) Spin resolved dispersion along kx axis.
(c) and (d) shows the spin texture and colour code [57]. . . . . . . . . 29
1.7 Example of genus 0, 1, 2, 3 (orientable) surfaces and knots. . . . . . . 30
1.8 Two different topological state in Kane-Mele moddel . . . . . . . . . 30
1.9 Detection of spin-momentum locking of spin-helical Dirac electrons in
Bi2 Se3 and Bi2 T e3 using spin-resolved ARPES. a , b , ARPES inten-
sity map at EF of the (111) surface of tuned stoichiometric Bi2−δ Caδ Se3 (a)
and of Bi2 T e3 ( b ). Red arrows denote the direction of spin projec-
tion around the Fermi surface. c , d , ARPES dispersion of tuned
Bi2−δ Caδ Se3 ( c ) and Bi2 T e3 ( d ) along the kx cut. The dotted red
lines are guides to the eye. The shaded regions in c and d are our pro-
jections of the bulk bands of pure Bi2 Se3 and Bi2 T e3 , respectively,
onto the (111) surface [81]. . . . . . . . . . . . . . . . . . . . . . . . . 31
1.10 Schematic of spin transfer and spin orbit torque . . . . . . . . . . . . 32
 11
1.11 Schematic of generation and set up for detection of spin orbit torque
in topological insulator. (a) Schematic of shift of equilibrium Fermi
surface due to applied bias which produces the nonequilibrium spin
density. (b) Schematic diagram topological insulator (Bi2 Se3 ) and
permalloy (Py) heterostructure. The red and yellow arrows denote the
non-equilibrium spin in corresponding position. c, Schematic of the
circuit used for the ST-FMR measurement [117]. . . . . . . . . . . . . 34
1.12 Schematic of antiferromagnetic memory device . . . . . . . . . . . . . 36

2.1 Origin of geometric phase due to adiabatic evolution on a sphere. . . 40

2.2 Integration contour for summing Matsubara frequency. . . . . . . . . 46
2.3 (Left) A two terminal device configuration. The central red region
corresponds the scattering region and the blue regions corresponds semi
infinite leads. (Right) Equivalent finite system where the effects of the
electrodes are replaced by corresponding self-energy. . . . . . . . . . . 49
2.4 Schematic for incoming and outgoing modes of a two terminal device.
µL and µR are the chemical potential of the left and right leads. S
corresponds the scattering matrix. . . . . . . . . . . . . . . . . . . . . 50
2.5 Tight binding scheme for a square lattice in magnetic field. A is the
vector potential and Φ is the total flux per unit cell of area a2 . . . . . 52
2.6 (a) Bulk band structure over the two dimensional Brillouin zone. The
legend shows the Chern number of each band. (b) Distribution of
Berry curvature over the Brillouin zone for each bulk band. (c) Band
structure of the quasi one-dimensional ribbon. The black lines shows
the edge bands. (d) Wavefunctions for some selected states of the
ribbon. Corresponding energy eigenvalues are marked in (b) with same
colour. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.7 Robustness of edge states in QHE system in a square lattice with 0.02
flux quantum per square. The scattering region is made of 300 × 120
sites and we choose the transport energy such that only the lowest
Landau level contribute to the transport. The black crack shows a
defect and the grey bushes are the scalar impurity with amplitude
0.2t, t being the hopping parameter. Red and blue colours show the
right and left moving modes. . . . . . . . . . . . . . . . . . . . . . . . 54
 12
2.8 (a) Lattice structure of a honeycomb lattice. Red and blue dots corre-
spond A and B sublattices. a1 and a2 are the lattice vectors and the
green region shows the unit cell. (b) Corresponding Brillouin zone and
first positive band. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.9 Structure and energy spectrum of zigzag and armchair nanoribbons.
The blue bands for armchair nanoribbon corresponds a ribbon with
3n+2 hexagon which forms a cone. The red bands corresponds an
armchair ribbon with 3n hexagons which does not form a cone. . . . . 56
2.10 (a) Staggered flux in Haldane’s model. Solid and dashed line corre-
sponds first and second nearest neighbour hopping. Filled and empty
circles denotes two different sublattices. (b) Variation of Chern num-
ber (ν) in the phase space. φ, M and t2 are the phase change due to
second nearest hopping, staggered onsite energy and the second nearest
hopping amplitude [65]. . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.11 Different topological phases in Kane-Mele model. (a) QSH phase for a
staggered potential λv = 0.1t and (b)the insulating phase for λv = 0.4t
where t is the hopping parameter. In both case intrinsic spin orbit
coupling (λSO = 0.06t) and Rashba spin orbit coupling (λR = 0.05t) is
fixed [68]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.12 Side and top view of a buckled lattice. Red and blue spheres correspond
A and B sublattices. . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.13 (Left) Different topological phases for silicene in E (electric field) -
M (magnetisation) space. (Right) Bandstructure of a silicene zigzag
nanoribbon in different topological phases [179]. . . . . . . . . . . . . 62
2.14 (A)Band edges of HgTe and CdTe and (B) band inversion of HgTe with
thickness [70]. (C) Measurement of quantised Hall conductance [71]. . 63
2.15 (b)Bulk (grey region) and edge band structure of a quasi one dimen-
sional BHZ system. ψL and ψR denote two edge states with opposite
velocity. (a) probability density along the width W for ψL and ψR . (c)
Conductance (G) in unit of conductance quantum G0 (blue) and DOS
(red) of the system. The blue dashed line shows the conductance in
presence of impurity. . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
 13
2.16 (a) Lattice structure of Bi2 Se3 multilayer (Bi=purple, Se=green). (b)Schematic
of band inversion in Bi2 Se3 [73]. (I), (II) and (III) represent the effect
of turning on chemical bonding, crystal-field splitting and SOC. (c)
Band structure of Bi2 Se3 multilayer obtained from density functional
theory. Dark and light blue lines corresponds edge and bulk bands.
Red line shows the bulk band structure. . . . . . . . . . . . . . . . . 66
2.17 Energy spectrum and spin texture of a Dirac Hamiltonian without
(Eq.2.59) and with (Eq.2.60) magnetisation. . . . . . . . . . . . . . . 67

3.1 Variation of bandgap and probability density of lowest unoccupied

state of graphene ring with zigzag(a,b) and armchair(c,d) edges. x =
rx /b and y = ry /a where rx,y is the actual distance along x, y directions,
a = 1.42Å and b = 2.46Å [183]. . . . . . . . . . . . . . . . . . . . . . 68
3.2 (Color online) Schematic of initial wavepacket at time t = 0 in a zig-zag
graphene nano ribbon. . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.3 Band structure of a ZGNR near K and K 0 point. The blue lines are
due to the bulk state and the red lines are due to the edge states. The
green region denotes the continuous spectrum for an infinite graphene
sheet. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.4 Variation of ky with kx at K(kx = 0) point. The solid blue lines are real
part of kx and the red dashed line is the imaginary part. The states on
the left side of the vertical grey dashed line are bulk states (kx is real)
and the states on the right side are edge states (the imaginary part of
kx is finite). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.5 (a)Norm of the wavepacket at t = 0 for three different values of kx0 with
d = L/4. For kx0 = −8/L the bulk states are dominant. For kx0 = 1/L
both bulk and edge gives the same contribution and for kx0 = 8/L the
states are localized at the edges. The width of the ribbon is L and
same scale is used for the length. The blue dashed line denotes x = 0.
(b) Evolution of a wavepacket with d=L/4 and kx0 = −4/L in time
where the time is measured in units of L/vF . . . . . . . . . . . . . . . 74
3.6 (Color online) Expectation value for x and y for the complete wavefunc-
tion (Ψ) and for positive and negative component (+/−) respectively.
Here we use kx0 = 1/L and d = L/4 . . . . . . . . . . . . . . . . . . . 74
 14
3.7 Zitterbewegung component of (a) hx(t)i and (b) hy(t)i for a wavepacket
with Gaussian width d=L/4. (c) and (d) shows the variation of ampli-
tude with the central momentum. . . . . . . . . . . . . . . . . . . . . 75
3.8 Variation of (a) xM
Z
ax
and (b) Γ with kx0 for different d’s (given in legend). 77

4.1 (a) Geometrical structure of a buckled honeycomb lattice. (b) Forma-

tion of Dirac cones at K and K 0 valleys with opposite spin represented
by red and blue colour. . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.2 Top: Dirac cone for EZ = ±EC at K and K 0 valley. Bottom: Variation
of band gap for spin up and down at K valley for M = 0. The green
and orange regions correspond to quantum spin hall phase (Cs = 1)
and bulk insulator (Cs = 0) phase where Cs = (C↑ − C↓ )/2 is the spin
Chern number. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.3 Formation of Gaussian wavepacket. The blue and red lines corresponds
up and down spin bands and the green line shows the Gaussian distri-
bution. The shaded region shows the portion of bands that contributes
to the wavepacket. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.4 Zitterbewegung component of x and y coordinates for different values
of EZ and M . The left panels show the corresponding band structure. 85
4.5 Formation of beating in charge (top) and spin (bottom) density waves. 86
4.6 Nature of beating when spin up and down are in the same and opposite
topological phases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.7 Product of the beat (↑ + ↓) amplitude of longitudinal Zitterbewegung
(xAmp
ZB ) and relative phase factor (Θ) over the EZ −M plane for different
width of momentum distribution. The top (bottom) panel corresponds
to a broad (narrow) spatial distribution, i.e. a narrow (broad) momen-
tum distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

5.1 Schematic of FTI and different types of AFTI. The green region shows
one unit cell of the lead. Blue and red dots represent positive and
negative m~ i. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.2 Conductance (G) of TI with different magnetic configurations against
random disorder for (a) out of plane and (b) in plane magnetic mo-
ments. The conductance is normalized to the conductance quantum
G0 = 2e2 /h. The boxed portion of (a) is enlarged in the inset. . . . . 94
 15
5.3 (a) Total density of states (DOS) for TI, FTI and different AFTI. (b)
Conductance against random disorder at EF = 0.05A (filled symbols)
and at EF = 0.25A (open symbols) for different configurations with an
in-plane magnetic moment 0.2Aŷ. . . . . . . . . . . . . . . . . . . . . 95
5.4 Non-equilibrium spin density for (a) FTI and (b) G-AFTI. We use a
magnetization strength m = 0.2A along ŷ and the spin densities are
evaluated at EF = 0.25A with a bias voltage (µL − µR ) = 0.02A. (c,d)
shows the corresponding field like (τF , magenta) and antidamping (τD ,
green) SOT evaluated at the top edge for FTI and G-AFTI respectively. 97
5.5 Variation of uniform (Sux,z ) and staggered (Sstx,z ) spin densities as a
function of disorder strength with an in plane magnetic order |m ~ i| =
0.2A for (a) FTI and (b) G-AFTI. The green dot-dashed line shows
the corresponding conductance. . . . . . . . . . . . . . . . . . . . . . 98
5.6 Evolution of Sx at top layer with disorder strength for (a)FTI and
(b)AFTI. The magnetic order and Fermi level is |m ~ i | = 0.2A and
EF = 0.25A respectively. The spin density is averaged over 1280 con-
figurations and calculated with a bias voltage 0.02A. . . . . . . . . . 100
5.7 Staggered Sstx at V0 =1.5A for different length (L) and width (W , given
in legend). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.8 The uniform and staggered order parameter for (a) AFTI at EF =
0.05A, (b) AFTI at EF = 0.25A and (c) for TAF at EF = 0.25A. For
TAF we use B = 1.0A, M = −2.2A. |m ~ i | = 0.2A for all three cases. . 101

6.1 Schematic of the interlayer coupling procedure and origin of the bulk
spin Hall current and surface texture. . . . . . . . . . . . . . . . . . . 107
6.2 Band structure for 5F M + 20T I with ∆ = 0.25A and for different
coupling parameter : (a) tTM =0, (b) 0.1A, (c) 0.2A, (d) 0.3A, (e) 0.4A,
(f) 0.5A. The red, blue and green colors correspond to contributions
from FM layers, top TI layer and bottom TI layer. The black lines
correspond to the bulk TI bands. . . . . . . . . . . . . . . . . . . . . 109
6.3 (a) Schematic of the band structure of a typical FM-TI heterostructure.
The blue and red colors correspond to contributions from TI and FM
layers, respectively. (b, c, d) Corresponding non-collinear spin texture
computed at different positions in the band structure, as denoted by
the vertical lines in (a). . . . . . . . . . . . . . . . . . . . . . . . . . . 110
 16
6.4 Schematic of the origin of different non-equilibrium spin components
in an FM-TI heterostructure. In the bulk of the TI away from the
interface, SHE spatially separates the flowing electrons with spin ori-
ented along ±y (large blue and red arrows). Electrons polarized along
+y penetrates into the FM layer, generating an effective spin density
oriented along +x (small green arrows). In addition, ISGE at the in-
terface directly generates a spin density oriented along +y (small blue
arrows). Both spin densities exert a torque on the magnetization of
the FM layer (yellow arrows). . . . . . . . . . . . . . . . . . . . . . . 113
6.5 (a) Band structure of bulk TI (thick green lines) and a 20-layer TI slab
(black lines). The corresponding spin texture of the top and bottom
layers of the slab at E = 0.25A is represented on the insert of (b).
(b) Spin Hall conductances as a function of the energy computed in
the bulk of the same TI. The spin conductance is constant as long as
the energy lies in the bulk gap and decreases when bulk states become
conductive. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6.6 Layer-resolved non-equilibrium Sx , Sy and longitudinal charge conduc-
tivity σxx across the FM-TI heterostructure, with tTM = 0.5A and
∆ = 0.25A. The FM-TI interface is marked by a vertical dashed line.
The color scale gives Sx,y in units of 1010 V−1 ·m−1 and σxx in 10−4 Ω−1 .115
6.7 (a) Sx , (b) Sy and (c) σxx for a 5FM+20TI heterostructure for differ-
ent coupling constant (tTM ) with exchange coupling ∆ = 0.25A and
broadening η = 0.01A. The vertical dashed lines show the beginning of
decoupled magnetic bands (E = 0.3A) and bulk TI bands (E = 1.0A).
Sx is given in units of 109 V−1 ·m−1 , Sy in 1010 V−1 ·m−1 and σxx in
10−4 Ω−1 . The bottom panel shows the color code for different layers. 116
6.8 (a1, a2, a3) Band structure and Spin texture of (b1, b2, b3) M1 and
(c1, c2, c3) M2, and (d1, d2, d3) TItop at (1) E = 0.25A, (2) E = 0.75A
and (3) E = 1.25A for tT M = 0.5A, ∆ = 0.25A and η = 0.01A. The
texture colors correspond the Sz component of the spin density. The
band colors have the same meaning as Fig. 6.2. . . . . . . . . . . . . 119
 17
6.9 non-equilibrium (a) SyF M , (c) SxF M in FM and TI conductance (b) σxx TI

with tTM = 0.5A and ∆ = 0.25A for different disorder broadening η.

The blue lines correspond to the surface-dominated/magnetic insula-
tor regime, the red lines correspond to the surface-dominated/magnetic
metal regime, and the green lines correspond to the bulk-dominated/magnetic
metal regime. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
6.10 Non-equilibrium spin density summed over the magnetic layers and the
conductance summed over TI layers of the full system for 5 FM + 20
TI layers with tTM = 0.5 and η = 0.1A. . . . . . . . . . . . . . . . . . 124

7.1 (a) Schematic of an AF-TI heterostructure. Red and blue spheres de-
notes up and down magnetic moment of the AF layer and the green
spheres denotes the TI. (b) Top view of an AF lattice. The gray bound-
ary denotes the unit cell. a is the interatomic distance. (c) Correspond-
ing first Brillouin zone (gray region) with lattice vector L1 and L2 and
high-symmetric points M , Γ and X for the AF-TI heterostructure. . . 129
7.2 Band structure of a heterostructure with 10 layers of TI and 5 layers of
AF with a coupling strength (a) tC = 0, (b) tC = 0.5 and (c) tC = 0.75.
The red, green, blue and black colors correspond to the contributions
from AF layers, bottom TI layer, interfacial TI layer and bulk TI layers,
respectively. The gray region shows the gap due to exchange splitting
for decoupled AF. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
7.3 (a) Band structure, non-equilibrium spin components (b) Sx (c) Sy and
(d) conductivity for sites 1 to 12 with out-of-plane magnetization. The
red, blue, green and black colors in (a) show the contribution from AF,
interfacial TI, bottom TI and bulk TI bands. The horizontal dashed
lines denote the peaks of spin densities. In (b), (c) and (d) the vertical
dashed green lines show the AF layers and the thick green line the
position of the interface. Corresponding configuration is shown in the
bottom panel where the blue and red boxes denote AF sites with up
and down magnetic moments, and the green boxes correspond to TI
sites. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
 18
7.4 (a) Band structure and spin texture at site (b) AFB B
2 , (c) AF1 and (d)
TIB1 (see bottom panel in Fig. 7.3) at different energies [(1) E/A =
0.156, (2) E/A = 0.582, (3)E/A = 0.724]. The arrow indicates the
in-plane component and the color of the arrow represents the out of
plane component. The dotted lines correspond to the degenerate bands.135
7.5 Contribution from Fermi surface (orange) and Fermi sea (green) for
the B sublattice of first three AF layers. The equilibrium Sz is shown
in the boxed regions where red and blue color correspond negative and
positive values. The maximum value in red box is 0.35 and maximum
value in the blue box in 0.1. . . . . . . . . . . . . . . . . . . . . . . . 136
7.6 Band structure of a Rashba-ferromagnet system (7.14) along kx with
projected (a) σy , (b) σz , (c) Berry curvature and (d) integrated Berry
curvature or the anomalous Hall coefficient. The dashed lines show the
exchange gap. Here we use ~ = 1, m0 = 0.5, α = 1 and ∆ = 0.05. . . 138
tot
7.7 (a) Sxstg , (b) Sytot and (c) σxx of the AF layer as a function of coupling
strength whose value is shown in the colorbar. (d) The band structure
of decoupled (red) AF-TI and with coupling tC = 0.5A (blue). The
dashed lines show three different regions I, II and III. . . . . . . . . . 140
7.8 Energy dependence of (a) Sxstg , (b) Sytot and (c) σxx tot
of the AF layer as
a function of the onsite energy of the AF layer with tC = 0.5A. The
inclined dashed line shows the region spanned by the decoupled AF
bands. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
7.9 (a) Sxstg , (b) Sytot and (c) σxxtot
as a function of broadening. Different
line colors correspond to different energies as shown in panel (d) along
with the band structure. . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.10 Variation of τ̄F L (a,b,c) and τ̄DL (d,e,f) with respect to θ (a,c;d,f) and
φ (b;e). The inset value shows the fixed angle. (g), (h) and (i) shows
Sx , Sz and Sy where solid and dashed line denotes the contribution
from AF sites with up and down spin. The colors represent different
energies as described in Fig. 7.9d. . . . . . . . . . . . . . . . . . . . . 145
7.11 Variation of τD in xy plane for (a) tC = 0.25, (b) tC = 0.50 and
(c) tC = 0.75 and corresponding non-equilibrium Sz (d,e,f). Different
colors correspond the energies at which Sz have peaks. The band
structures are shown in Fig. 7.2. . . . . . . . . . . . . . . . . . . . . . 146
 19

LIST OF TABLES

1.1 Conductivity(σ), spin Hall conductivity(σS ), spin-Hall angle (θS ) and

switching current (JSW ) for different systems [141]. . . . . . . . . . . 35

3.1 Comparison between different characteristics of the Zitterbewegung

in various systems: InAs nanowire, Graphene (Gr) mono and bilay-
ers with and without a magnetic field. The last line is our results.
The quantities represented are the width of the wavepacket d, the ini-
tial wavelength λ0 = 2π/k0 , amplitude xMax
Z , oscillation period τ , and
−1
damping rate Γ . In the case of semiconducting nanowire and ZGNR,
we use the values are resonance. . . . . . . . . . . . . . . . . . . . . . 78

4.1 Different parameter values for graphene (Gr), silicene (Si), germanene
(Ge) and stanene (Sn) [105, 181]. . . . . . . . . . . . . . . . . . . . . 81

6.1 Tight-binding parameters used in the main text. . . . . . . . . . . . . 109

6.2 Room temperature bulk conductivity, spin conductivity and SOT ef-
ficiency measured in various Bi2 Se3 (t)/FM(d) systems, where t and d
are the thicknesses in nm. The bulk conductivity is in 104 Ω−1 ·m−1
and the spin conductivity is in (~/2e) 104 Ω−1 ·m−1 . The numbers in
the last column are computed using Eqs. (6.19) and (6.20). . . . . . . 122

7.1 Torque coefficients τFi and τDi for zx, xy and yz plane. . . . . . . . . . 144
 20

Chapter 1

Introduction

Based on their ability to conduct electric current, materials can be roughly classified
into two distinct species - metal and insulator. Contrary to an insulator, a metal can
conduct electricity when subjected to a small bias voltage which can be explained
with classical model. With the technological advancement it is now possible to probe
into the dimension comparable to the atomic scale where the physical phenomena
are driven by quantum mechanical laws. Depending on the associated energy and
length scale, a large class of observed metallic properties can be explained within
classical, semiclassical or quantum framework [1]. In the last couple of decades a new
class of phenomena emerged in condensed matter physics which are governed by their
inherent topology. One of the most fascinating manifestation of the impact of topology
on material is the so called topological insulator - a special class of materials which
is insulator in the bulk and conducting at the edges. Physical laws when revisited
within the topological framework, reveals new features and opens new possibilities
for devices with high efficiency and low power consumption. In the following sections
we present a brief introduction of the journey from classical to topological properties.

1.1 The hall of Hall effects

In presence of a magnetic field, a charged particle deviates from its initial path due to
the Lorentz force, which gives rise to more interesting features. In 1879 Edwin Hall
[2] observed that when subjected to a perpendicular magnetic field, a metallic system
 21
develops a potential difference perpendicular to the direction of current. This effect is
known as “Hall Effect” or more commonly “Ordinary Hall Effect”. The phenomena is
a manifestation of Lorentz force felt by the electrons in an magnetic field and can be
explained within the framework of classical electrodynamics. At low temperature and
smaller dimension (ballistic regime, i.e. when the Fermi wavelength is larger than the
mean free path) the physics is governed by the laws of quantum mechanics. As a result
instead of monotonic dependence between current and voltage one observes quantised
steps [3] which can be explained in terms of quantised modes with Landauer-Büttiker
formalism [4]. This is known as the quantum Hall effect (QHE). Soon it is discovered
that it is possible to simulate Hall-like effects even in absence of an applied magnetic
field by exploiting the other degrees of freedom. Among these field-free Hall effects,
the anomalous Hall effect (AHE) and spin Hall effect (SHE) construct the core of
this thesis (Fig.1.1). For completeness we also give a brief description of topological
Hall effect. It is worth mentioning that although these descriptions are presented for
electronic systems, analogous effects can be observed in magnonic [5], photonic [6] or
phononic [7] systems as well.

HE/QHE AHE SHE

Figure 1.1: Schematic representation of Hall effect (HE), quantum Hall effect (QHE),
anomalous Hall effect (AHE) and spin Hall effect (SHE).

1.1.1 Quantum Hall effect

“The unexpected discovery of the quantum Hall effect was the result of ba-
sic research on silicon field-effect transistors combined with my experience
in metrology, the science of measurements.”
 22
-Klitzing [8]

Figure 1.2: (Left) The first measurement of Quantum Hall plateau by Klitzing, Dorda
and Pepper [9]. (Right) Klitzing’s note on the experiment [8].

In 1980 while trying to conduct a high accuracy measurement of fine structure

constant with silicon metal-oxide-semiconductor field-effect transistor, Klitzing [9]
found that the longitudinal resistance of the system vanishes over a range of energy
1 h
and the transverse or the Hall resistance shows plateaus with height ν e2
(where ν ia
an integer) the effect being insensitive to the geometry of the sample. Similar results
were also obtained by Tsui and Grossard [10] at much lower temperature (50mK) and
with smaller magnetic field (∼ 50kG) [11].
Theoretically the behaviour of quantised conductances had already been studied
by Ando [12, 13, 14] in a series of papers and their physical origin was explained in
terms of Landau levels. In 1981 Laughlin [15] described the quantised Hall conduc-
tance as a manifestation of the gauge invariance of the system. Halperin [16] provided
an alternative description which shows the quantised Hall conductance coming from
robust edge states. However, it was well known that in presence of a uniform magnetic
field or periodically varying potential the Landau levels split into multiple sub-bands
 23
[17, 18] and each sub-band, according to Laughlin’s argument, must carry an integer
multiple of Hall current caused by the entire Landau level. Besides both Laughlin’s
and Halperin’s argument relied on the existence of a boundary and they didn’t clarify
whether the Hall conductance depends on the boundary condition. These limitations
were overcome by Thouless, Kohmoto, Nightingale and Nijs [19] in 1982 when they
explicitly derive the expression for Hall conductance in the bulk using Kubo formula
[20, 21]. Their result shows that the origin of the quantised Hall conductance is
embedded in the momentum space topology of the system [22] and hence is quite
insensitive to any smooth deformation (for example rough boundary or random im-
purity), which facilitates high accuracy measurement. Although they produce the
same result, the quantised Hall conductances calculated in bulk [19] and edge [16]
have different topological origin. The first one is given by the Chern number of the
bulk whereas the second one given by the winding number of the edge [23].

1.1.2 Anomalous Hall Effect

In continuation of his experiments Hall observed that in magnetic materials the trans-
verse conductance remains finite even at zero magnetic field [24, 25]. The behaviour
was quite anomalous and hence eventually acquire the name “Anomalous Hall Ef-
fect”. Although the anomaly is resolved the name survived. It was soon realised
that the heart of the effect is the so called spin-orbit coupling. However, the exact
mechanism was still in darkness at that time. Karplus and Luttinger [26] showed
that in presence of spin-orbit coupling electrons acquire an anomalous velocity per-
pendicular to the applied bias which appear as a perturbation effect. This anomalous
velocity was later identified as the Berry Curvature[27] of the ground state. The
model was based on a clean system and therefore faced a strong criticism. Based on
the asymetric scattering by impurity, two alternative models were proposed by Smit
(skew scattering)[28, 29] and Berger (side jump)[30]. The mechanism proposed by
 24
Karplus and Luttinger is termed ‘intrinsic’ as it depends only on the band structure
and eigenstates of the material, not on the nature of the disorder. The mechanisms
proposed by Smit and Berger on the other hand depend on the interaction between
the electron and the impurity and hence are termed ‘extrinsic’. At this point it is
worth mentioning that the anomalous Hall measurements had been done in two dif-
ferent types of materials. They are either ferromagnet with uniform magnetisation or
nonmagnetic material with magnetic impurity (dilute Kondo system). Both of them
demonstrate anomalous Hall current either due to inherent magnetisation or due to
induced magnetisation. One major difference between these two systems is that at
zero applied magnetic field the former still breaks the time reversal symmetry while
the later preserves the time reversal symmetry. Since the Berry curvature vanishes
in presence of time reversal symmetry, the governing mechanism of anomalous Hall
effect would be different for these two types. Much of the debate about the dominant
mechanism was caused by an attempt to explain such different cases with a common
theory.

1.1.3 Spin Hall Effect

In a nonamgnetic system, the situation is quite different. Since the system has the
same number of up and down spin, the asymmetric scattering due to spin orbit
coupling (SOC) can not produce a finite charge current. Dyakonov and Perel [31,
32] first predicted the existence of a finite spin current in a system with zero net
magnetisation. The experimental verification of this new effect was hindered long due
to lack of any prescription to detect the spin current. In 1999 Hirsch [33] proposed
a experimentally feasible method to detect the spin current by connecting two edges.
In fact, the term spin Hall effect was first introduced by Hirsch. Although Hirsch’s
method remained an experimental challenge, it opened new possibilities for theoretical
and experimental apprehension. A more simplified version of this method was later
 25
introduced by Hankiewicz et.al. [34] in 2004 using a H-bar geometry. The technical

Figure 1.3: (A and B) Two-dimensional images of spin density ns and reflectivity R,

respectively, for the unstrained GaAs sample measured at T = 30 K and E = 10 mV
µm−1 [35].

complication in detecting the electrical signal in spin Hall effect was initially bypassed
by using optical method. The first successful optical detection was done by Kato et.
al. [35] (extrinsic spin Hall effect) and by Wunderlich et. al. [36] (intrinsic spin
Hall effect) in 2004. Zhang [37] generalised the spin Hall effect in presence of spin
diffusion and proposed using a ferromagnetic probe to detect the spin accumulation.
This prescription was soon experimentally realised by Valenzuela [38] and Kimura
[39]. In 2010 Brüne et.al. [40] detected spin Hall effect using Hankiewicz’s H-bar
geometry.

1.1.4 Skyrmion/topological Hall effect

The topological Hall effect or the skyrmion Hall effect [41] is a rather younger in-
troduction to the Hall family. Here the Hall effect is produced with a non-collinear
magnetisation. When an electron moves through the spatially varying magnetisation,
 26
it feels an emergent electromagnetic field [42] and undergoes a deviation due to the
effect of the Lorentz force (Fig.1.4). During the process the electron also exerts a
torque on the skyrmion causing it to deviate in opposite direction. Such a deviation
has been observed experimentally as well [43].

Electron flow
Topological
Hall Effect

Spin transfer
torque
Skyrmion
Hall Effect

Emergent
electromagnetic induction

Figure 1.4: Schematic of topological Hall effect. The non-collinear magnetisation of

skyrmion creates an emergent electromagnetic force for an incoming electron which
creates a Hall effect. The resulting spin transfer torque causes the skyrmion to un-
dergo a deflection in opposite direction [42].

1.2 Spin orbit coupling in different materials

The spin orbit coupling in real materials is a relativistic phenomena. When a charged
~ ∼
particle moves in an electric field, it feels a magnetic field B 1 ~
(E × p~). This
mc2

effective field acts on the spin of the system resulting in a change in energy given by
~ × p~) where σ denotes the spin of the system. Dresselhaus first proposed an
∼ µB ~σ .(E
effective Hamiltonian for spin-orbit coupling for non-centosymmetric bulk material
[44]. A more comprehensive version was presented by Rashba and Sheka [45] where
they showed the an energy splitting linear in momentum ner the Γ point in a bulk
wurzite lattice. They also show the effective form of the SOC Hamiltonian in other
high symmetric points as well. Bychkov and Rashba conducted further studies on the
k-linear SOC Hamiltoninan HSO = α[σ × k] · ez and showed how to obtain the value
of α from the experimental data obtained by Stein et. al. [46] and Stormer et. al.
 27
[47]. In 1979 Vas’ko [48] first deduced the wellknown effective form σ · [ez × pk ] of the
SOC interaction due to short range scalar potetial created by a surface. Depending
on the lattice symmetry some possible forms of SOC interactions are [49] given by

100
HSO = α(ky σx − kx σy ) + γ(kx σx − ky σy ) (1.1)
110
HSO = αky σx + βkx σy + λkx σz (1.2)
111
HSO = (α + γ)(ky σx − kx σy ) (1.3)

where the superscripts denote the surface plane and α, β, γ and λ are the material
parameters. In presence of a broken inversion symmetry, only the α term survives
and we obtain the well known Rashba spin-orbit Hamiltonian

HR = α~z.(~σ × ~k) (1.4)

From Rashba and Sheka’s derivation one can see that the spin momentum locking
is actually originates from the coupling between the spin and orbital angular momen-
tum known as RussellSaunders coupling or L.S coupling and the atomic potential
generated by the nucleus. Therefore the SOC strength of material is often measured
in terms of atomic SOC Hamiltonian [50]

α2
 
1 ∂V ~ ·S
~ = ζL
~ · S,
~
HSO = L (1.5)
2 r ∂r

where HSO is the SOC energy for an one electron atom.The value of the parameter ζ
for various atoms with respect to their atomic number is given in Fig.1.5.

1.2.1 Heavy metal interfaces

The initial studies on the SOC and spin non-degenerate k-linear dispersion (Rashba
effect) were conducted on the the bulk zincblend or wurzite crystals. Vas’ko first
 28

Hf-Hg
Ga-Kr

La-Lu
Al-Ar
10 000

Sc-Zn

In-Xe
B-Ne

Y-Cd
8000

ζl / cm -1
6000
4000
2000
0
20 40 60 80 100
Atomic Number

Figure 1.5: Variation of ζ with atomic number. Data taken from [51]. Red, green,
blue and gray region denote s, p, d and rare earth elements.

not only demonstrated the existence of spin non-degenerate states near the surface of
these lattices but also pointed out the possibility to produce a non-equilibrium spin
density with an applied in plane electric field [48]. This effect was later reported by
Edelstein [52] and also known as Edelstein effect. The quest for the Rashba SOC
soon turns towards the metal surfaces from the semiconductors. Due to their heavy
nuclei and strong ζ values (Fig. 1.5), the 4d (Y-Cd) and 5d (Hf-Hg) elements become
quite popular in this regard. The first detection of Rashba SOC in metallic system
was made with in Au(111) surface using angle resolved photo emission spectroscopy
(ARPES) in 1996 [53]. Similar band splitting was observed in Li covered 110 surfaces
of W and Mo [54]. A direct measurement of the Rashba parameter becomes possible
with the introduction of spin resolved ARPES [55, 56, 57] which also makes it possible
to visualise the resulting spin texture (Fig. 1.6). These experimental findings along
with the density functional theory (DFT) based study [58] showed that the Rashba
SOC arises as a consequence of inversion symmetry breaking of the wavefunction.

1.2.2 Topological Insulator surfaces

A topological insulator is identified by conducting boundary modes with a insulating

bulk [59, 60, 61] which distinguishes it from a normal insulator (one without con-
ducting edges) and can be characterised by a topological invariant associated with its
 29

Figure 1.6: (a) Spin polarisation at Fermi level Au(111) surface state excited by p-
polarized 6.05 eV photons. (b) Spin resolved dispersion along kx axis. (c) and (d)
shows the spin texture and colour code [57].

band structure [62, 63, 64]. Contrary to quantum Hall systems, no applied magnetic
field is required (time reversal symmetry is preserved) in this case and therefore one
has to define a different topological invariant for these systems. The method is similar
to classifying a closed Gaussian surface or a knot by its genus. The term ‘topological
insulator’ was first introduced by Moore and Balents [64].
Topological insulators draw enormous attention because for a long time, it was
believed that notrivial topology can only be observed in a system with broken time
reversal symmetry and therefore magnetic field is indispensable for edge state physics.
In 1987 Haldane [65] proposed a model where one can observe quantum Hall effect in
absence of any net magnetic field. The model was based on a honeycomb lattice with
alternative magnetic flux such that the resultant flux through the unit cell is zero and
the time reversal symmetry is lifted by a second nearest hopping. Although the phys-
 30

Figure 1.7: Example of genus 0, 1, 2, 3 (orientable) surfaces and knots.

ical realisation of Haldane’s model remains an experimental challenge [66], it opened

new theoretical possibilities. In 2004 Kane and Mele [67] generalised Haldane’s idea
for spinful graphene monolayer and showed the existence of robust spin filtered edge
states (Fig.1.8). The nontrivial topological state can be characterised by the Z2 in-
variant of the system [68]. There are two crucial features of Kane-Mele model. First,

Figure 1.8: Energy bands for a one-dimensional zigzag strip in the (a) QSH phase
and (b) the insulating phase in Kane-Mele model [68].

it possesses edge states in presence of time reversal symmetry and second, in addition
to quantised charge current it also predicts the existence of quantised Hall current.
This quantised Hall current has a completely different topological origin compared to
the spin Hall current predicted in semiconductor [69] which does not show any edge
state. In two dimensions, the seminal work by Kane and Mele [67] founded the corner-
stone for the search of the time reversal invariant topological insulators, its detection
 31
remaining an experimental challenge at that time. Bernevig, Hughes and Zhang [70]
predicted the existence of nontrivial topological phase in a HgTe/CdTe/HgTe quan-
tum well which was experimentally observed [71] subsequently. Soon it was realised
that the topological states can also survive in three dimensions [72] and can be char-
acterised by a triplet of Z2 invariants. Unlike two dimensional topological insulators,
in three dimensions one can not define individual topologically protected states for
different spin species and therefore spin-orbit coupling is indispensable for a strong
topological insulator in three dimension. Three dimensional topological insulators
were originally predicted [73] and observed [74, 75, 76, 77, 78] in Bi2 Se3 and Bi2 T e3 .
Unlike d block elements, in these materials only a single spin-texture arises on one
surface which can be directly observed by ARPES measurement. Quest for new
topological materials [79, 80] has been a sensation ever since.

Figure 1.9: Detection of spin-momentum locking of spin-helical Dirac electrons in

Bi2 Se3 and Bi2 T e3 using spin-resolved ARPES. a , b , ARPES intensity map at
EF of the (111) surface of tuned stoichiometric Bi2−δ Caδ Se3 (a) and of Bi2 T e3 ( b
). Red arrows denote the direction of spin projection around the Fermi surface. c
, d , ARPES dispersion of tuned Bi2−δ Caδ Se3 ( c ) and Bi2 T e3 ( d ) along the kx
cut. The dotted red lines are guides to the eye. The shaded regions in c and d are
our projections of the bulk bands of pure Bi2 Se3 and Bi2 T e3 , respectively, onto the
(111) surface [81].
 32
1.3 Spin transfer and spin orbit torque

Spin Transfer Torque Spin Orbit Torque Components of Torque

Figure 1.10: Schematic of spin transfer and spin orbit torque. FM and HM denote
ferromagnet and heavy metal. The spacer is an insulating layer. jc shows the direction
of the charge current. TF and TD denotes field like and damping like torques.

When a spin polarized current is injected into a ferromagnet, the incoming spins
exert a torque on the local magnetization. In other words, the non-equilibrium
spin density carried by the incoming electrons is coupled to the local magnetiza-
tion through exchange interaction, which forces the non-equilibrium spin density, i.e.
the itinerant moement, and the local magnetic moments to align with each other.
This spin density can be obtained by passing a charge current through a primary
ferromagnet [82, 83], or through spin-orbit processes [84, 85] detailed in the next sec-
tion. From the ferromagnet point of view, this electrical torque can be parsed into
two components: a field-like torque, that acts like an effective magnetic field, and a
damping torque that acts like an non-equilibrium magnetic (anti)damping [86, 87].
In case of a two terminal spin transfer torque device, the reading and writing use
same channel which reduces the reliability and lifetime of the device. This limitation
is overcome in a spin orbit torque device where the electron polarisation is generated
by passing a charge current through a sigle layer with strong spin orbit interaction
[84, 85]. This not only offers better better reliability but also provides better effi-
ciency for device. Like spin transfer torque, spin orbit torque also can be classified
into two category namely field like and damping like [88, 49, 89, 90] which might
 33
have different forms depending on the symmetry of the system and spin orbit cou-
pling [91, 92]. Experimentally spin orbit torque was first observed in 2009 [93] in
(Ga,Mn)As. Ever since there is a growing interest in detecting spin orbit torque by
both electrical [94, 95, 96, 97, 98, 99, 100, 101] and optical [102, 103] means. However
there is still a ongoing debate about the physical origin of the torque. While some
studies suggests the origin to be the interfacial Rashba spin-orbit coupling [94], others
indicates the bulk spin-Hall [96] effect to be the governing mechanism. The spin-Hall
effect [104, 105, 106] cause (anti)damping torque whereas the interfacial spin orbit
coupling contributes the field like torque [107, 108] and therefore the relative strength
of these two torques bears the signature of the dominant mechanism. Experimental
results [109, 110, 111, 112, 113] suggests that the debate is yet to be resolved.

1.4 Spin-orbit torque in different materials

The heart of a SOT device is made of two main elements - (i) a material with broken
inversion symmetry as a source of strong spin-orbit coupling (SOC) and (ii) a magnetic
layer (Fig.1.11). When a charge current moves through the material with SOC it
produces a non-equilibrium spin density which penetrates into the magnetic layer
and exerts a torque on the magnetisation [84, 85]. Initially only ferromagnets were
used as the magnetic material. Recently antiferromagnets have been found to exhibit
faster switching [114, 115, 116].

1.4.1 Spin-orbit torque with ferromagnets

The first generation of SOT devices was fabricated with heavy metals like platinum
[95] tantalum [96] or tungsten [118]. Due to their inherent strong spin-orbit cou-
pling topological insulators soon made a prominent entry in the study of spin-orbit
torque [117, 119, 120, 121]. Compared to their heavy metal counterpart, they show
much better spin charge conversion efficiency [122, 123, 124] and therefore are consid-
 34

Figure 1.11: Schematic of generation and set up for detection of spin orbit torque
in topological insulator. (a) Schematic of shift of equilibrium Fermi surface due to
applied bias which produces the nonequilibrium spin density. (b) Schematic diagram
topological insulator (Bi2 Se3 ) and permalloy (Py) heterostructure. The red and
yellow arrows denote the non-equilibrium spin in corresponding position. c, Schematic
of the circuit used for the ST-FMR measurement [117].

ered as promising candidates for spintronics application [125]. Topological insulator-

ferromagnet heterostructures can operate with a switching current ∼ 105 A/cm2 at
room temperature [126, 127, 128] which is two order of magnitude smaller compared
to that of a heavy metal ferromagnet heterostructure [95, 96]. The actual source of
their superior efficiency and the impact of their band topology on their efficiency are
still a matter of active research. Most of the theoretical studies have been focused
only on the surface states assuming a simplified Hamiltonian [129, 130, 131, 132, 133].
Although the simplified Hamiltonian suits well for a clean surface of topological in-
sulator [134, 135, 136], there applicability is questionable when a ferromagnetic layer
is brought to its vicinity. DFT based study shows that [137] along with the spin or-
bit coupling, the band structure is also modified in presence of ferromagnetic layers.
However a fully first principle calculation to study spin orbit torque is computation-
ally quite expensive and has not been done so far. In between the accurate and
computationally heavy density functional methods and the oversimplified analytical
approach, lies tight binding models . A well-parametrised tight binding model can
 35
capture the essential features of a real system [138, 139] and can produce results in
good agreement with DFT [140] and thus has become a favourable tool for theoretical
investigations.

Parameters Bix Se1−x [141] Bi2 Se3 [127] Bi2 Se3 [128] β-Ta [96] Pt [95]
σ(Ω−1 m−1 ) 0.78 × 104 9.43 × 104 2.43 × 104 5.3 × 105 4.2 × 106
~ −1 −1
σS ( 2e Ω m ) 1.5 × 105 0.15 × 105 0.43 × 105 −0.8 × 105 3.4 × 105
θS 18.62 0.16 1.75 -0.15 0.08
JSW (A cm−2 ) 4.3 × 105 3 × 106 6 × 105 5.5 × 106 2.85 × 107−8

Table 1.1: Conductivity(σ), spin Hall conductivity(σS ), spin-Hall angle (θS ) and
switching current (JSW ) for different systems [141].

1.4.2 Spin-orbit torque with antiferromagnet

“A large number of antiferromagnetic materials is now known ... They

are extremely interesting from the theoretical viewpoint, but do not seem
to have any applications.”

-Louis Néel, Nobel Lecture 1970.

The physics of spintronic memory device undergoes a paradigm shift with the
introduction of antiferromagnetic spintronics [142, 143, 144, 145, 146, 147, 148]. Due
to their indifference to an applied magnetic field, the antiferromagnetic memories are
more robust against magnetic perturbation which also constitutes a major obstacle to
manipulate their order parameter. In 2006 Núñez et. al. [149] theoretically showed
the possibility of switching antiferromagnetic order parameter using spin transfer
torque. Z̆elezný et. al. [150] first proposed using spin orbit torque for manipulat-
ing the antiferromagnetic order parameter in a noncentrosymmetric antiferromagnet.
Experimentally it was first demonstrated by Wadley et. al. in 2015 [151] in CuM-
nAs. Similar results also obtained in ferromagnet/antiferromagnet bilayer by Fukami
[100] and Oh [152]. The room-temperature functional antiferromagnetic memory was
demonstrated by Olejnik [153] in 2017. Apart from having extra protection against
 36

Figure 1.12: a. Left: schematic of antiferromagneetic chain and right: 0 and 1

configuration with respect to an STM tip. b, Left: Optical microscopy image of
the solid-state CuMnAs memory device and schematic of the electrical writing and
readout geometry. Right: Change in the transverse resistance with 50 ms writing
pulses [146].

magnetic field antiferromagnets shows superior switching speed [153, 154, 155, 156].
However, the role of topological phases on antiferromagnetic spintronics is still an
uncharted territory and attracting blooming interest [157, 158]. Although presence of
ferromagnetism destroys the topological states by breaking the time reversal symme-
try, it is possible to preserve them in presence of antiferromagnetism [159]. GdPtBi
has been predicted to be a member of this newly predicted antiferromagnetic topo-
logical insulator class [160, 161]. Although some introductory studies have been done
with tight binding model [162, 163], a proper study in this field is yet to be done.

1.5 Motivation and outline of the thesis

Topologically nontrivial materials have opened new horizons in designing the next
generation spintronic devices. The characteristic feature that distinguishes topolog-
ical insulators from normal metals and semiconductors is their immunity against
 37
various types of perturbations which can provide a new route to the long-anticipated
dissipationless transport. The topological insulators already have proven their su-
perior efficiency in switching magnetic moment. Compared to a heavy metal based
magnetic memory device, they require two orders of magnitude less switching current.
Although the experimental findings are well established, the underlying physics is still
subjected to strong debate. One particular question that has raised a lot of ambiguity
in understanding the source of SOT in heavy metal-ferromagnet heterostructure is
its origin. Where some studies suggest that it comes from the interfacial SOC, other
studies indicate towards the bulk spin Hall effect to be the origin. In case of a topolog-
ical insulator, due to the bulk-boundary connection, these issues become even more
complicated. So far the most accurate theoretical description of these materials can
be provided by DFT. However, DFT deals with the equilibrium properties whereas
for practical applications the non-equilibrium properties are more important which
within the DFT framework is computationally very demanding. Besides, due to the
inherent nonlinearity of DFT, it becomes very difficult to identify the source of some
particular effect. Therefore, for a proper theoretical understanding, it is crucial to
building minimal models which can capture all the essential features of these systems
and yet should be simplified enough to be tractable. This is the central theme of
this thesis. These models can be further fine-tuned with the data obtained from the
DFT calculation. One major advantage of using such simplified models is that it can
reduce the computation time significantly and thus provides us with an opportunity
to study different non-equilibrium phenomena and their subtle features. Another ad-
vantage of such a simplified model-based approach is that we can simulate the effect
of impurity scattering which is very useful to understand and interpret experimental
results. Majority of the DFT based calculations are based on clean systems. Im-
purity is often introduced as a single substitutional or doping atom which can not
simulate the effect of decoherence. In the case of topologically nontrivial systems
 38
impurity scattering requires special attention due to their robustness. Therefore a
proper framework to study the effect of impurity scattering is highly solicited.
This thesis is focused on addressing these problems. Here we study different
nonequilibrium phenomena by using simplified model Hamiltonians. Depending on
the nature of the problem, we employ either a continuum or lattice Hamiltonian. The
models are kept as general as possible so that they can be used for a wide variety
of materials rather than a particular material. At this point, it is more prudent to
explore new possibilities with these materials which can lead to new paradigms in
device engineering. In this thesis, we focus more on providing a qualitative under-
standing of different aspects and speculate new features of these materials, rather
than providing quantitatively accurate data. However, we also demonstrated that
with certain parameter values our models can produce similar values of different
observables as predicted by DFT which establishes their credibility. We have particu-
larly emphasised on demonstrating the connection between different equilibrium and
non-equilibrium observables. Our approach allows us to identify the impact of partic-
ular interactions on different observables like the conductivity or the nonequilibrium
spin density. These connections can be exploited for an equilibrium DFT based study
to predict the corresponding non-equilibrium scenario.
The thesis is divided into eight chapters. Chapter 1 gives a brief introduction.
In chapter 2 we discuss relevant theoretical formalisms. Chapter 3-7 are the main
constituents of this thesis. While chapter 3 and 4 are focused on identifying and char-
acterising the non-trivial topology, chapter 5-7 demonstrate the impact of nontrivial
topology on different non-equilibrium observables. Chapter 3 deals with honeycomb
lattices where an edge state appears from the lattice symmetry. Here we study the
wave-packet dynamics and identify the admixture of bulk and edge modes from the
oscillation of wavepacket known as zitterbewegung. In chapter 4 we demonstrate that
zitterbewegung can also bear the signature of different topological phases. In chapter
 39
5 we study the transport properties of magnetic topological insulators and show a
comparison between ferromagnetic and antiferromagnetic configurations. In chapter
6 we introduce a combination of topological insulator and ferromagnetic materials
and study the non-equilibrium spin density and SOT. Chapter 7 deals with a topo-
logical insulator-antiferromagnet heterostructure. The overall conclusion and future
prospects are kept in chapter 8.

2. Theoretical
1. Introduction
prerequisite

3. Bulk and edge modes in honeycomb

lattice and their impact on
wavepacket dynamics

8. Conclusion 4. Probing topological phases

with wavepacket dynamics

7. Topological insulator -
antiferromagnet heterostructure 5. Ferromagnetic and
antiferromagnetic
topological insulator
6. Topological insulator -
ferromagnet heterostructure
 40

Chapter 2

Theoretical Prerequisites

In this chapter we introduce some of the most essential theoretical concepts and
formalisms used in subsequent chapters.

2.1 Berry phase, Berry curvature and Chern number

The concept of the Berry phase can be best explained by a hypothetical experiment.
Imagine a pendulum is transported adiabatically (so that the plane of oscillation does
not change with time) from the north pole to the equator, then along equator and
finally brought back to the north pole (Fig.2.1).

Figure 2.1: Origin of geometric phase due to adiabatic evolution on a sphere.

One can see that although the pendulum is brought back to the same position,
the initial state is no longer the same as the initial state. This is happening because
at the north pole (and also at south pole) the coordinate is not uniquely defined. If
the pendulum is transported along a path which encircles the pole, the initial and
 41
final state of the pendulum would be different by 2π. This total phase gain due to
travelling along a closed path is known as geometric phase. Similar phenomena can
be observed when a quantum mechanical system is transported in a closed path in
parameter space. The final state might differ from the initial state by a constant phase
factor which cannot be avoided by any gauge transformation. This problem was first
studied by Pancharatnam [164, 165] in 1956 and later by Berry [27] in 1984. Hence
the geometric phase is also known as PancharatnamBerry phase or more commonly
Berry phase.

“Slowly, I realized that Pancharatnam’s phase was something I had to

understand. This was not easy because his arguments made heavy use
of the geometry of the Poincaré sphere, which I knew about but had little
facility with. ... I learned that not only had this young fellow of twenty-two
created the simplest example of the geometric phase, but that he had also
pointed out a feature (the definition of phase difference described below)
that had not, by 1987, been perceived in any of the many papers developing
my work of 1984.”

-Berry [166]

The most comprehensive way to understand the Berry phase is to follow Berry’s
original derivation [27]. Let us consider a Hamiltonian H(R(t)) which depends on a
set of time dependent parameter R(t). Let us assume that the parameters R(t) are
changed adiabatically such a way that R(T ) = R(0). The eigenstates (|n(R(t))i) can
be defined as

∂
i |n(R(t))i = H(R(t))|n(R(t))i = En |n(R(t))i, (2.1)
∂t

For simplicity, we set ~ = 1. Here we do not write the coordinate dependence explic-
itly. For a time independent Hamiltonian the evolution of the state can be simply
 42
Rt 0 0
denoted by an exponential ei 0 dt En (t ) . Since the parameters, and hence the Hamil-
tonian, is changing adiabatically we can express the eigenstates as

Rt
En (t0 )dt0 iγn (t)
|Ψn (t)i = ei 0 e |n(R(t))i (2.2)

where eiγn (t) is the geometric phase. Inserting this wavefunction in Schrödinger equa-
tion (Eq.2.1) one can easily show that the geometric phase change from time ti to
time tf is given by

Z R(tf )
γn (tf − ti ) = i ~ R |n(R(t))i.dR,
hn(R(t))|∇ ~ (2.3)
R(ti )

Since the explicit dependence on time is no longer used, we can drop t from the
~ R |n(R(t))i is called Berry connection. Note
expression. The quantity ihn(R(t))|∇
that Berry connection is not gauge invariant and therefore not an observable. If we
change the parameters in a cyclic way (R(tf ) = R(ti )) then the phase change due to
the cyclic evolution can be written as

I
γn = i ~ R |n(R)i.dR,
hn(R)|∇ ~
Z
= i ~ R × hn(R)|∇
∇ ~ R |n(R)i.dS,
~ (2.4)

where we have used the Stokes theorem. This expression can be further simplified in
 43
a gauge invariant form as

Z
γn = i ~ R n(R)| × |∇
h∇ ~ R n(R)i.dS,
~
Z Xh i
= i ~ ~ ~
hn(R)|∇R |m(R)i × hm(R)∇R |n(R)i .dS,
m6=n
Z X" ~ R H|m(R)i × hm(R)|∇~ R H|n(R)i
#
hn(R)|∇ ~
= i .dS, (2.5)
m6=n
(En − E m ) 2

Z
= i Ω ~ n .dS,
~ (2.6)

where

X  hn(R)|∂Ra /∂H|m(R)ihm(R)∂Rb /∂H|n(R)i 

~ a,b = i
Ω − (a ↔ b) , (2.7)
n
m6=n
(En − Em )2
~ R × (ihn(R)|∇
= ∇ ~ R |n(R)i),

is known as Berry curvature for the nth state. If we consider the Berry connection
as a gauge potential in parameter space, then the Berry curvature plays the role of a
magnetic field. Berry curvature is a gauge invariant quantity and a large number of
physical quantities like polarisation, orbital magnetisation, anomalous Hall coefficient
[167, 168, 169] can be calculated from it. In the following sections we are going to
see some more example.
Berry curvature plays a special role in the analysis of two dimensional topologically
nontrivial systems. For an n band system one can define a quantity

Z
1 ~ n (~k),
Cn = dkΩ (2.8)
2π

know as the Chern number of the band. This number is an integer and the total
 44
Chern number for all bands is zero. When summed over the valence bands only

X
σxy = f (E − En )Cn , (2.9)
n

where f (E−En ) is the Fermi-Dirac distribution, it gives the anomalous Hall coefficient
of the system which is a topological invariant of the system. We are going to present
more discussion on this in the following sections.

2.2 Linear response theory

Let us consider a system given by a Hamiltonian H0 and subjected to a perturba-

tion H 0 (t). We assume that the perturbation is turned on at time t = −∞. The
expectation value of an operator Ô at any time t is given by

O(~r, t) = hψ0 |S † (t, −∞)Ô(~r, t)S(t, −∞)|ψ0 i, (2.10)

where |ψ0 i is the eigenstates of H0 and S(t, −∞) is the S matrix which evolve the
states to time t. The S matrix is given by

Rt 0 0 0
S(t, −∞) = T [e−i −∞ H (t )dt ],
Z t
= 1−i H 0 (t0 )dt0 + O(H 02 ). (2.11)
−∞

For a weak perturbation it is sufficient to consider the linear term only (hence the
name linear response) and the expectation value can be given by

Z t
O(~r, t) = hψ0 |Ô(~r, t)|ψ0 i − i dt0 hψ0 |[Ô(~r, t0 ), H 0 (t0 )]|ψ0 i, (2.12)
−∞

This is the general form of Kubo formula [170]. The first term is the unperturbed
 45
expectation value and is not affected by the source. If the source term is a time
~ r, t) = E
dependent electric field E(~ ~ 0 ei(~q.~r−ωt) , then the perturbation Hamiltonian can

be given by

~ r, t) = i
Z Z
0
H (t) = d~r~j(~r, t) · A(~ ~ r, t),
d~r~j(~r, t) · E(~ (2.13)
ω

where ~j(~r, t) is the current operator which is proportional to the velocity only and
~ r, t) is the vector potential for the electric field such that E(~
A(~ ~ r, t) = ∂t A(~
~ r, t).

Instead of the expectation value of an operator, we are often interested in the response
function, i.e. the expectation value per unit source strength. For a homogeneous
system one can express the correlation function in Fourier space as

Z ∞
1
χO q , ω)
ab (~ = dteiωt hψ0 |[Ôa (~q, t), jˆb (~q, 0)]|ψ0 i, (2.14)
ωV 0

where the indices a, b corresponds the vector components and V is the volume of the
system. The DC response can be defined as

1
χO q ) = lim
ab (~ (−=[Πab (~q, ω)]) , (2.15)
ωZ
ω→0
1 ∞ 0
ΠR
ab q
(~ , ω) = −i dteiω(t−t ) Θ(t − t0 )hψ0 |[Ôa (~q, t), jˆb (~q, t0 )]|ψ0 i, (2.16)
V −∞

where ΠR
ab (ω) is the retarded Matsubara correlation function and = denotes the imag-

inary part. The easiest way to calculate it is through the Matsubara correlation
function in imaginary frequency defined as

Z β
1
π(~q, iω) = − dτ eiωτ hψ0 |S(τ, 0)Ôa (~q, τ )jˆb (~q, 0)|ψ0 i, (2.17)
V 0

where β is the inverse equilibrium temperature (1/kB T ). The Matsubara S matrix is

 46
defined as

 Z τ 
0 0 0
S(τ, 0) = Tτ exp − dτ H (τ ) . (2.18)
0

Considering the lowest order of expansion we can write the correlation function as

1 X
π0 (~q, iω) = Ôa (~q)G(~q, iν + iω)jˆb (~q)G(~q, iν), (2.19)
βV iν

where ν is the Matsubara frequency which is given by the poles of the thermal dis-
tribution function. Using this property we can express the summation as a contour
integral,

I
1 1
π0 (~q, iω) = − f (z) Ôa (~q)G(~q, z + iω)jˆb (~q)G(~q, z), (2.20)
2πi CR V

where CR is a circle of radius R. We must consider R → ∞ to make sure all the

Matsubara frequencies have been considered.To avoid the poles of the Green’s function
at z = 0 and z = −iω, we choose the integration contour like Fig.2.2.

Im

z=0 Re
z=-iω

Figure 2.2: Integration contour for summing Matsubara frequency.

As shown in the figure the singularities at z = 0, −iω are avoided by an infinites-

imal shift ±δ. The integration along the circumference will vanish for R → ∞. The
remaining contributions are
 47

∞
i
Z
π0 (~q, iω) = df ()[
2πV −∞

Ô(~q)G(~q,  + iω)jˆb (~q)G(~q,  + iδ) − Ôa (~q)G(~q,  + iω)jˆb (~q)G(~q,  − iδ)

+ Ôa (~q)G(~q,  + iδ)jˆb (~q)G(~q,  − iω) − Ôa (~q)G(~q,  − iδ)jˆb (~q)G(~q,  − iω)].
(2.21)

Using the analytic continuation iω → ω + i0+ and using the identity G(ω ±
i0+ ) = GR,A (ω) we can get the retarded correlation function in terms of retarded and
advanced Green’s functions.

∞
i
Z
ΠR q , iω)
ab (~ = df ()[
2πV −∞

Ôa (~q)GR (~q,  + ω)jˆb (~q)GR (~q, ) − Ôa (~q)GR (~q,  + ω)jˆb (~q)GA (~q, )

+ Ôa (~q)GR (~q, )jˆb (~q)GA (~q,  − ω) − Ôa (~q)GA (~q, )jˆb (~q)GA (~q,  − ω)].

(2.22)

The correlation function Πab (~q, ω) (2.16) is given by the negative imaginary part
of this retarded correlation function ΠR (~q, ω) (2.22). Since we are interested in DC
response we can use the limiting condition

∂ R,A
lim GR,A (~q,  ± ω) = GR,A (~q, ) ± ω G (~q, ). (2.23)
ω→0 ∂

The first term (GR,A (~q, )) gives real value and therefore does not contribute to
 48
the response function. Therefore the response function can be simplified as

=[ΠR (~q, ω)]

χab (~q) = − lim
Z ∞ω
ω→0
∂GR (~q, ) ˆ

1
= − df ()Re Ôa (~q) jb (~q)(GR (~q, ) − GA (~q, ))
2πV −∞ ∂
∂GA (~q, )

R A ˆ
−Ôa (~q)(G (~q, ) − G (~q, ))jb (~q) (2.24)
∂

This is know as Kubo-Bastin formula [171]. It can be further simplified by inte-

gration by parts and can be written as

χab (~q) = χsurf

ab
ace
(~q) + χsea q)
ab (~
1 h i
χsurf
ab
ace
q
(~ ) = Re Ôa q
(~ )G R
q
(~ , )jˆb q
(~ )(GA
q
(~ , ) − G R
q
(~ , ))
πV EF
Z ∞  R
1 ∂G
χsea
ab (~q) = df ()Re Ôa (~q)GR (~q, )jˆb (~q)
πV −∞ ∂
∂GR ˆ

R
−Ôa (~q) jb (~q)G (~q, ) (2.25)
∂

where we use that ∂f /∂ = −δ(E − EF ). This form is known as Kubo-Streda or

Streda-Smrcka formula [172]. χsurf
ab
ace
and χsea
ab is known as Fermi surface and Fermi

sea terms and gives the contribution from states near Fermi level and states from
the valence bands. For topologically trivial system the Fermi surface terms gives
the dominant contribution whereas for the nontrivial systems the Fermi sea terms
dominates.

2.3 Transport in quasi one dimensional system

For practical purpose we often use a two terminal device configuration where a finite
scattering region is connected to semi-infinite electrodes. While dealing with such
configuration one can divide the whole system in three regions - (i) a central region
 49
characterised by a finite dimensional Hamiltonian HS , (ii)a left (HL ) and (iii) a right
(HR ) electrode. The overlap between the scattering region and the electrodes are em-
bedded in the connection matrices VL,R . The electrodes are periodic in one direction
and the unit cells of the electrodes are connected by VR,L (Fig.2.3).

VL VLS VRS VR ΣL ΣR

HL HS HR HS

VL VLS VRS VR
Figure 2.3: (Left) A two terminal device configuration. The central red region cor-
responds the scattering region and the blue regions corresponds semi infinite leads.
(Right) Equivalent finite system where the effects of the electrodes are replaced by
corresponding self-energy.

Such system can be represented by an infinite dimensional matrix as

...
 
VL
 
 † 
VL HL VLS 
 
 
H= † (2.26)
 VLS HS VRS 

 
†
VRS HR VR 
 

 
† ..
VR .

However it is not possible to work with an infinite dimensional matrix. Therefore

the effect of the semi-infinite electrodes are added as self energy [4] to the interfacial
sites of the scattering region. Another way is to start with the unperturbed plane
wave states in the lead far from the scattering region and calculate the scattering
state in subsequent blocks using the transfer matrix. In this way one can calculate
all incoming and outgoing mode as well as the scattering wavefunction within the
central region. The incoming and outgoing modes can be utilised to construct the S
matrix which in turns can be used to calculate the conductance. For example let us
 50
consider a set up (Fig.2.4) where the left and right modes are given by

N
X
ψL (z) = ai φi eiki z + bi φi e−iki z , (2.27)
i
N
X
ψR (z) = ci φi eiki z + di φi e−iki z . (2.28)
i

a1 c1

b1 d1
mL S mR
aN cN

bN dN

Figure 2.4: Schematic for incoming and outgoing modes of a two terminal device.
µL and µR are the chemical potential of the left and right leads. S corresponds the
scattering matrix.

One can define the S matrix for the system as

   
b a
  = [S]   , (2.29)
c d

where S is an unitary 2N × 2N matrix which connects the outgoing modes (b, c) to

the incoming modes (a, c). The transmission probability from ith incoming channel
to j th outgoing channel is Tij = |Sij |2 and corresponding current is given by

e
Iij = (µi − µj )Tij , (2.30)
h

where µi and µj are the chemical potential of the lead attached to the ith and j th
channel. The total current is given by summing over all the possible incoming and
outgoing modes.
An alternative way to calculate the transmission and conductance is by using
Green’s function where the semi-infinite leads are replaced by their self energy (Fig.2.3).
 51
The transmission probability from ith to j th lead is given by

Tij = T r[Γi GR Γj GA ], (2.31)

where GR,A = [EI − HS − ΣR,A

All ]
−1
is the retarded/advanced Green’s function of the
scattering region. ΣR,A
All is the retarded/advanced self energy summed over all the
R,A
leads. Γj = i[ΣR A
j − Σj ] where Σj is the retarded/advanced self energy of the j th
lead.
In most of the cases we are interested in non-equilibrium quantities. In that
case one has to use a new type of Green’s function know as lesser Green’s function
(G< ). The retarded and lesser Green’s function for a two terminal device (Fig.2.4)
are defined by [173]

−1
GR (E) = [EI − HS − ΣR R
Lef t − ΣRight ] , (2.32)

G< (E) = GR (E)[ifLef t (E)ΓLef t (E) + ifRight (E)ΓRight (E)]GR (E)† , (2.33)

where f (E) is the Fermi-Dirac distribution and Γj = i[ΣR − ΣA ]. The equilibrium,

non-equilibrium and current driven density matrix is defined as

1 ∞
Z
ρeq = − dEGR (E)f (E), (2.34)
π −∞
Z ∞
1
ρneq = dEG< (E), (2.35)
2πi −∞
ρCD = ρneq − ρeq . (2.36)

In case of symmetric leads the current driven density matrix can be further sim-
plified to [173]

ρCD = GR (EF )[ΓLef t (EF ) − ΓRight (EF )]GR (EF )† , (2.37)

 52
where EF is the Fermi energy of the system. Once the proper density matrix is
obtained, the expectation value of any physical observable Ô is given by

hOi = T r[Ô · ρ]. (2.38)

2.4 Square Lattice in uniform magnetic field: Tight binding

model

Most of our studies have been conducted on a lattice model with tight binding ap-
proximation. In this section we introduce the formalism and use it to characterise
a free electron gas in an uniform magnetic field. The first step is to discretise the
system over a square grid which resembles a square lattice. Then we assume that an
electron can either sit on the lattice point or jump from one lattice point to another,
each action being characterised by an energy. The Hamiltonian of the system can be
written as

-A.dl A.dl

eikya e-ikya

e-ikxa eikxa
a

Figure 2.5: Tight binding scheme for a square lattice in magnetic field. A is the
vector potential and Φ is the total flux per unit cell of area a2 .

X X X
H= iα c†iα ciα + tijα c†iα cjα + t0ijα c†iα cjα + · · · , (2.39)
i,α hi,jiα hhi,jiiα

where the summation goes over all the sites. c†iα , ciα is the creation,destruction oper-
ator for electron at ith site with other degrees of freedom α. The first term in Eq.2.39
is called the onsite energy and gives the energy of an electron sitting on the atomic
 53
site. The second term describes the hopping energy of an electron from one site to
one of its closest sites. The third term corresponds to hopping to the second nearest
site and so on. Since the probability, i.e. the wavefunction overlap, of the jumping
to another site decreases exponentially with distance, it is often sufficient to consider
hopping to the first nearest and second nearest sites. For a periodic system we can
use Bloch theorem and define the Hamiltonian in momentum space. In this case when
~ to another lattice point R
an electron jumps from a lattice point R ~ + ~r, it picks up
~
a phase eik.~r . In presence of a magnetic field it requires an additional phase term
e ~ r).d~ ~ r) is the corressponding vector potential. This
R
given by ei ~ A(~ r
(Fig.2.5) where A(~
is known as Peierels substitution [174, 175].
This simple toy model is sufficient to demonstrate the bulk-boundary correspon-
dence and the role of topology in the formation of edge states. We start with a
two dimensional square lattice with a magnetic field pointing out of plane with such
1h
a magnitude that each unit square contains 3e
flux. Next, we can create a quasi
one-dimensional chain from this system. Corresponding band structures and wave-
functions for the ribbon is shown in Fig.2.6.
One can see that the edge states are appearing at a energy where there is no states
in the bulk bands. Since the bulk bands are well separated, we can characterise each
bulk bands with a Chern number which is the integrated Berry curvature over the full
Brillouin zone. The edge states appear between band n and n + 1 if ni=1 Ci 6= 0, the
P

value denoting the number of edge states. In Fig.2.6(d) we plot the wave functions
of the right moving edge states which is localised at only one edge and hence they
are called chiral edge states. These chiral edge states are the source of quantised
quantum Hall conductance which can be studied using Landauer-Büttiker formalism.
One interesting feature of these edge channels is that they are immune to defect and
impurities as long as they are not connecting the opposite edges. This can be verified
by studying the scattering wave function within the scattering region (Fig.2.7).
 54

Figure 2.6: (a) Bulk band structure over the two dimensional Brillouin zone. The
legend shows the Chern number of each band. (b) Distribution of Berry curvature
over the Brillouin zone for each bulk band. (c) Band structure of the quasi one-
dimensional ribbon. The black lines shows the edge bands. (d) Wavefunctions for
some selected states of the ribbon. Corresponding energy eigenvalues are marked in
(b) with same colour.

Figure 2.7: Robustness of edge states in QHE system in a square lattice with 0.02
flux quantum per square. The scattering region is made of 300 × 120 sites and we
choose the transport energy such that only the lowest Landau level contribute to the
transport. The black crack shows a defect and the grey bushes are the scalar impurity
with amplitude 0.2t, t being the hopping parameter. Red and blue colours show the
right and left moving modes.
 55
2.5 Honeycomb Lattice

The Honeycomb lattice has played an important role in both condensed matter physics
[176] and high energy physics [177]. Contrary to square lattice, a honeycomb lattice
has two sites per unit cell commonly known as A and B sublattices. The most
important feature that distinguishes honeycomb lattice from square lattice is that
the bands touch each other two points in the Brillouin zone namely K and K 0 points,
where it mimics the dispersion relation of a massless Dirac fermion. The effect of mass
can be simulated by breaking the sublattice symmetry. In last few decades honeycomb
lattices have played a crucial role in the evolution of topology in condensed matter
physics. Here we are going to see three main models based on honeycomb lattice.

Figure 2.8: (a) Lattice structure of a honeycomb lattice. Red and blue dots correspond
A and B sublattices. a1 and a2 are the lattice vectors and the green region shows the
unit cell. (b) Corresponding Brillouin zone and first positive band.

A two dimensional honeycomb lattice can be defined by a simple tight-binding

model involving only onsite and nearest neighbour hopping elements,

X X
H= εi c†i ci − t c†i cj , (2.40)
i hiji

where εi,α is the onsite energy and t is the hopping energy. If the lattice is made of
equivalent sublattices, then one can neglect the onsite term. For a choice of lattice
 56
defined in Fig.2.8(a) the dispersion relation reads

√ √
v ! !
u
u 3kx − 3ky 3kx + 3ky √ 
E(kx , ky ) = ±t3 + 2 cos + 2 cos + 2 cos ,
3kx (2.41)
2 2

√ √
which is zero at 4π/(3 3)(±1/2, 3/2). These two points are known as K and K 0
point (Fig.2.8(b)). At these two points the low energy effective Hamiltonian is given
by

H(kx , ky ) = σx kx + ησy ky (2.42)

where η = ±1 correspond K and K 0 valleys.

Zigzag Armchair
1.0 1.0

0.5 0.5
Armchair

a1 a2
E/t

0.0 E/t 0.0

-0.5 -0.5

Zigzag -1.0 -1.0

Γ X Γ X Γ X

Figure 2.9: Structure and energy spectrum of zigzag and armchair nanoribbons. The
blue bands for armchair nanoribbon corresponds a ribbon with 3n+2 hexagon which
forms a cone. The red bands corresponds an armchair ribbon with 3n hexagons which
does not form a cone.

One can further construct a quasi one dimensional ribbon from an infinite two di-
mensional honeycomb lattice. Depending on the direction of truncation one can have
mainly two types of nanoribbon namely zigzag and armchair nanoribbons (Fig.2.9).
In case of a zigzag nanoribbon the edges are made of one type of sublattice and they
supports localised zero energy states at the edges. In case of an armchair nanoribbon,
depending on the width, the valence and conduction bands can touch each other. It
forms a cone if the ribbon contains 3n+2 numbers of hexagons, where n is an in-
teger, along the transverse direction, otherwise it possesses a gap which is inversely
 57
proportional to the width of the ribbon.

2.5.1 Haldane model

This model was introduced by Haldane in 1988 in his seminal paper [65] and become
one of the founding stone in the study of topological phases. The idea was to create
a model which can demonstrate quantum Hall effect without any applied magnetic
field. To obtain that he introduced a staggered magnetic flux in the unit cell in such a

(a) (b)

Figure 2.10: (a) Staggered flux in Haldane’s model. Solid and dashed line corresponds
first and second nearest neighbour hopping. Filled and empty circles denotes two
different sublattices. (b) Variation of Chern number (ν) in the phase space. φ, M
and t2 are the phase change due to second nearest hopping, staggered onsite energy
and the second nearest hopping amplitude [65].

way (Fig.2.10(a)) that the total flux enclosed by the unit cell remains zero. As a result
the flux encircled by the first nearest neighbour hopping (t1 ) is zero and therefore one
can choose real hopping parameters. However the flux encircled by the second nearest
neighbour hopping (t2 ) is not zero and which makes them accumulate a complex phase
term. By varying these parameters along with a staggered onsite energy (M ) it is
possible to create a non-trivial topological phase (Fig.2.10(b)). One major difficulty
in experimentally realising Haldane model is to construct the staggered flux within a
the unit cell. This hurdle was later overcome using ultracold Fermions [66].
 58
2.5.2 Kane-Mele model

Based on Halden model Kane and Mele introduced a new model to study quantum
spin Hall effect in graphene [67, 68]. Unlike Haldane model, Kane-Mele model take
electronic spin into account and the role of staggered flux is played by the spin-orbit
coupling. The nontrivial topological phases manifest themselves as zero energy modes
in a zigzag nanoribbon (Fig.2.11). In this construction, it is possible to preserve time
reversal symmetry and as a result it is not possible to characterise the topological
phases with Chern number. Kane and Mele showed that in presence of time reversal
symmetry, one can use a Z2 classification.

1
E/t

λ /λ
5 R SO
I
QSH
0 0

−5 λv/ λ SO

(a) −5 0 5
(b)
−1
0 π ka 2π 0 π ka 2π
Figure 2.11: Different topological phases in Kane-Mele model. (a) QSH phase for a
staggered potential λv = 0.1t and (b)the insulating phase for λv = 0.4t where t is
the hopping parameter. In both case intrinsic spin orbit coupling (λSO = 0.06t) and
Rashba spin orbit coupling (λR = 0.05t) is fixed [68].

The Kane-Mele Hamiltonian consists of three main terms and can be written as

H(~r) = H0 (~r) + HR (~r) + HSO (~r), (2.43)

where H0 is the normal hopping term, HR corresponds Rashba SOC and HSO cor-
responds the intrinsic SOC. In addition one can also introduce a magnetic exchange
term (HM ) and a staggered potential term (HU ). With these new terms the complete
 59
Hamiltonian becomes [178]

H(~r) = H0 (~r) + HR (~r) + HSO (~r) + HM (~r) + HU (~r), (2.44)

where each term is given by

X
H0 (~r) = −t c†iα cjα ,
hiji;α
X
HR (~r) = itR êz ·(sαβ ×dij )c†iα cjβ ,
hiji;α,β
2i X †
HSO (~r) = √ tSO ci s·(dkj ×dik )cj ,
3 hhijii
X †
HM (~r) = M ciα szαβ ciβ ,
i;α,β
X
HU (~r) = c†iα Vi ciα . (2.45)
i;α

Here, c†iα and ciα are the creation and annihilation operators for a pz electron with spin
α on site i. t in H0 is the nearest neighbour hopping amplitude which is of the order of
2.6 eV for graphene. HR describes the Rashba SOC where dij is the vector pointing
from site j to site i. The third term HSO is the intrinsic spin-orbit coupling with
k connecting the next-nearest neighbor sites i and j through an intermediate site k.
hi/hhii denotes sum over nearest/next-nearest neighbours. The fourth term and the
last term correspond to the exchange field and staggered AB -sublattice potentials,
respectively. To introduce a staggered potential one can use Vi = +U at A-type
sublattices and Vi = −U at B -type sublattices. α and β denote spin indices, and
s is a vector whose components are Pauli matrices. Since we are mostly interested
in the physics near the K and K 0 point, it is convenient to construct a low energy
Hamiltonian given by [178]
 60

h0 (~k) = v(ητx kx + τy ky )1s ,

λR
hR (~k) = (ητx sy − τy sx ),
2
hSO (~k) = ηλSO τz sz ,

hM (~k) = M 1τ σz ,

hU (~k) = U τz 1s . (2.46)

Here, η = ±1 denotes K and K 0 valleys. σ and τ are Pauli matrices corresponding

to the spin and the pseudo-spin/sublattice degrees of freedom. v = 3t/2 is the Fermi
√
velocity, λR = 3tR is the Rashba SOC and λSO = 3 3tSO is the intrinsic spin-orbit
coupling.

2.5.3 Buckled Honeycomb Lattice

While graphene is a two dimensional structure, its other siblings like silicene, ger-
manene, stanene etc. has a buckled structure [179, 180, 181]. Here the A and B
sublattices are located in two different planes (Fig.2.12) which gives rise to additional
spin-orbit interactions. In addition an electric field perpendicular to the lattice plane
creates different onsite potential energies at A and B sublattices, breaking the sub-
lattice degeneracy. This can be exploited to tune the band gap and the topological
phases of the system as well.
A buckled honeycomb lattice can also be described by a Kane-Mele type Hamilto-
nian (Eq.2.47) [179]. In addition to the nearest neighbour Rashba SOC, the buckling
gives rise to second nearest neighbour Rashba SOC which is an order of magnitude
 61

Figure 2.12: Side and top view of a buckled lattice. Red and blue spheres correspond
A and B sublattices.

stronger compared to the first nearest neighbour Rashba SOC.

X λSO X
H= −t c†iα cjα + i √ νij c†iα σαβ
z
cjβ
hi,jiα
3 3 hhi,jiiαβ
X †  z
+iλR1 (Ez ) ciα σ × d̂ij cjβ
αβ
hi,jiαβ
2 X  z
−i λR2 µi c†iα σ × d̂ij cjβ
3 αβ
hhi,jiiαβ
X X †
+` µi Ez c†iα ciα + M ciα σz ciα , (2.47)
iα iα

where the symbols have same meaning as Eq.2.45 and Eq.2.46. νij = ±1 if the second
nearest neighbour is reached with a anticlockwise/clockwise rotation. µi = ±1 for
A/B sublattice. ` is the buckling distance, i.e. the vertical distance between the planes
containing A and B sublattices. λR2 is the Rashba SOC coefficient corresponding
to second nearest hopping. Like plannar two dimensional homeycomb lattices, the
interesting features appear near the K and K 0 point where one can construct a low
energy effective Hamiltonian. The effective Hamiltonian at K valley in the basis
 62
{ψA↑ , ψB↑ , ψA↓ , ψB↓ }T is given by [179]

 
 E(1, 1) ~vF k− iaλR2 k− 0 
 
 ~vF k+ E(1, −1) −iλR1 −iaλR2 k− 
H(kx , ky ) =  , (2.48)
 
 −iaλ k iλR1 E(−1, 1) ~vF k− 
 R2 + 
 
0 iaλR2 k+ ~vF k+ E(−1, −1)

where k± = kx ± iky , and the diagonal elements are

E(sz , tz ) = λSO sz tz + `Ez tz + M sz , (2.49)

with the spin sz = ±1 and the pseudospin/sublattice tz = ±1. The sublattice sym-
metry can be broken using a perpendicular electric field (EZ ). When simultaneously
applied with magnetisation, they can show new exotic phases like quantum anoma-
lous Hall (QAH) phase and valley polarised metallic (VPM) phase (Fig.2.13) [179].

Figure 2.13: (Left) Different topological phases for silicene in E (electric field) -
M (magnetisation) space. (Right) Bandstructure of a silicene zigzag nanoribbon in
different topological phases [179].
 63
2.6 Two dimensional topological insulator

In 2003 Murakami, Nagaosa and Zhang [182] first predicted the existence of a dis-
sipationless spin current in hole doped semiconductors like Si, Ge and GaAs. In
a subsequent paper [69] they predicted the existence of a quantum spin current in
absence of a net charge current and proposed HgTe and PbTe to be a potential can-
didate. In 2006 Bernevig, Hughes and Zhang [70] showed that in a CdTe-HgTe-CdTe
quantum well one can manipulate the topological phase by tuning the thickness of the
HgTe layer. In 2007 König et al [71] experimentally observe the existence of quantised
Hall current in HgTe quantum wells, confirming the prediction of Bernevig, Hughes
and Zhang.

Figure 2.14: (A)Band edges of HgTe and CdTe and (B) band inversion of HgTe with
thickness [70]. (C) Measurement of quantised Hall conductance [71].

An effective Hamiltonian can be constructed by considering four bands near the

Fermi level. These bands are known as E1↑,↓ and H1↑,↓ . The E1 bands consists of |s, ↑i
and |s, ↓i orbitals and the H1 bands are coming from the combination of |px + ipy , ↑i
and |px − ipy , ↓i. The effective Hamiltonian near the Γ point is given by
 64

 
h(kx , ky ) 0
H(kx , ky ) =  , (2.50)

∗
0 h (−kx , −ky )

where

X
h(kx , ky ) = ε(k) + di σi , (2.51)
i=1,2,3
d1 ± d2 = A(kx ± iky ), (2.52)

d3 = M − B(kx2 + ky2 ), (2.53)

ε(k) = C − D(kx2 + ky2 ). (2.54)

This effective Hamiltonian was first introduced by Bernevig, Hughes and Zhang
[70] and commonly known as BHZ Hamiltonian. One can easily convert this Hamil-
tonian into a lattice Hamiltonian using the following transformations

d1 ± d2 = A[sin(kx ) ± i sin(ky )] (2.55)

d3 = (M − 4B) + 2B[cos(kx ) + cos(ky )] (2.56)

ε(k) = (C − 2D) − 2D[cos(kx ) + cos(ky )] (2.57)

Using this lattice model one can straightforwardly verify the existence of helical
edge states by constructing a quasi one-dimensional system (Fig. 2.15).
From Fig.2.15(a) one can see that at both edges left and right moving states posses
opposite spins. Therefore in order to reverse its direction of motion, an electron
one has to flip its spin as well. Since a scalar impurity can not change the spin,
the conduction due to the edge channels remains unaffected against scalar impurity
(Fig.2.15(c)). Since the bulk states do not have any such protection they easily
fall down against scalar impurity. The presence of strong impurities can create a
 65

Figure 2.15: (b)Bulk (grey region) and edge band structure of a quasi one dimensional
BHZ system. ψL and ψR denote two edge states with opposite velocity. (a) probability
density along the width W for ψL and ψR . (c) Conductance (G) in unit of conductance
quantum G0 (blue) and DOS (red) of the system. The blue dashed line shows the
conductance in presence of impurity.

connection between states moving with opposite velocity and same spin from opposite
surface, which eventually can kill the quantised edge transport.

2.7 Three dimensional topological insulators

In case of a three dimensional topological insulator the topologically protected states

appears at the surface and is characterised by a spin texture. Theoretically this
phase was first predicted in Bi2 Se3 , Bi2 T e3 and Sb2 T e3 [73] where the topologically
protected states manifest themselves as a Dirac cone at the Fermi level (Fig.2.16)
Experimentally such states were first observed in Bi2 T e3 [75]. Unlike heavy met-
als, the surface states in topological insulators are characterised by a non-zero Z2
topological invariant which can be probed experimentally [74]. From Fig.2.16 one
can see that near Fermi level the band structure forms a Dirac cone. The low energy
effective Hamiltonian near Γ point can be expressed in the basis (|P 1+ +
z , ↑i, |P 2z , ↑
 66
(a) (b) (c)
0.5
0.4
0.3

E-EF (eV)
0.2
0.1
0
-0.1
-0.2
K M

Figure 2.16: (a) Lattice structure of Bi2 Se3 multilayer (Bi=purple, Se=green).
(b)Schematic of band inversion in Bi2 Se3 [73]. (I), (II) and (III) represent the effect
of turning on chemical bonding, crystal-field splitting and SOC. (c) Band structure
of Bi2 Se3 multilayer obtained from density functional theory. Dark and light blue
lines corresponds edge and bulk bands. Red line shows the bulk band structure.

i, |P 1+ +
z , ↓i, |P 2z , ↓i) and is given by [73]

 
M(k) A1 kz 0 A2 k− 
 
 A1 kz −M(k) A2 k− 0 
H(k) = 0 (k)I4×4 +   + O(k2 ), (2.58)
 
 0
 A2 k+ M(k) −A1 kz  
 
A2 k+ 0 −A1 kz −M(k)

with k± = kx ± iky , (k) = C + D1 kz2 + D2 k⊥

2 2
and M(k) = M − B1 kz2 − B2 k⊥ .
The values of these parameters can be obtained by fitting this Hamiltonian with the
data obtained from the first principle calculations. For Bi2 Se3 the typical values
are [73] M = 0.28eV, A1 = 2.2eVÅ, A2 = 4.1eVÅ, B1 = 10eVÅ2 , B2 = 56.6eVÅ2 ,
C = −0.0068eV, D1 = 1.3eVÅ, D2 = 19.6eVÅ. In the case of a slab geometry one
can use a proper boundary condition at the surface which gives the effective surface
Hamiltonian
 
0 kx − iky 
HSurf (kx , ky ) = A2  , (2.59)

kx + iky 0
 67
where the surface wavefunctions are superpositions of |P 1+ +
Z , ↑ (↓)i and |P 2Z , ↑ (↓

)i. This represents a Dirac cone where the spin of a state is perpendicular to its
momentum (Fig.2.17). The presence of an out of plane magnetisation opens a gap in
the surface energy spectrum and the Hamiltonian can be written as
 
∆
∆ A2 (kx − iky )
HSurf (kx , ky ) =  (2.60)


A2 (kx + iky ) −∆

where ∆ is the effective exchange coupling.

Δ=0 Δ≠0

Figure 2.17: Energy spectrum and spin texture of a Dirac Hamiltonian without
(Eq.2.59) and with (Eq.2.60) magnetisation.
 68

Chapter 3

Bulk and edge modes in honeycomb lattice and their impact

on wavepacket dynamics

In this Chapter we are going to study a honeycomb lattice and the motion of a
wavepacket in it to identify any signature of their nontrivial topology. A honeycomb
lattice can have mainly two types of terminations namely a zigzag edge and an arm-
chair edge. The electronic properties of the system crucially depends on the nature of
the termination. For example, in case of an armchair termination, depending on the
width, the bad gap shows three distinct branches. The wavefunction of the highest
occupied as well as lowest unoccupied state is distributed all over the system. In case
of zigzag termination the band gap is much smaller and the wavefunction is localised
at the edges. This feature is independent of the system geometry and can be observed
even in quantum dots (Fig.3.1).
(a) (b) (c) (d)

xin=3, xout=6 yin=8, yout=13

Figure 3.1: Variation of bandgap and probability density of lowest unoccupied state
of graphene ring with zigzag(a,b) and armchair(c,d) edges. x = rx /b and y = ry /a
where rx,y is the actual distance along x, y directions, a = 1.42Å and b = 2.46Å [183].

Similar features can be observed in a nanoribbon as well. In this chapter we

 69
choose a zigzag nanoribbon. We are going to study a special feature of the motion
called Zitterbewegung. This is an oscillatory motion of the wavepacket in a relativistic
system [184] caused by the interference of positive and negative energy states. Such
oscillation has been predicted in a large class of multiband systems [185] and has
been observed experimentally as well [186, 187]. Here we are going to see how this
oscillation bears the signature of structural topology.
The main characteristic feature that distinguishes a zigzag graphene nano ribbon
(ZGNR) from a semiconducting nanowire is the presence of edge states which appears
as a result of coupling between transverse and longitudinal momenta caused by the
boundary condition [188, 189]. The edge states are distinguished by their low energy
which is a result of their strong localization. Although the properties of the edge
state has been extensively studied [190, 191], the effect of their interaction with the
bulk state has not been thoroughly addressed. In this work, we demonstrate that
the Zitterbewegung effect in ZGNR arises from the overlap between bulk and edge
states and enters in resonance when properly weighting the contributions of these
two types of states. Even more interestingly the resonance occurs in the direction of
propagation, in sharp contrast with other systems (extended graphene, nano wires
etc.).

Figure 3.2: (Color online) Schematic of initial wavepacket at time t = 0 in a zig-zag

graphene nano ribbon.
 70
3.1 Effective Hamiltonian for zigzag graphene nano ribbon

The effective Hamiltonian for graphene is given by [192]

Hη = −i~vF (σx ∂x + ησy ∂y ) (3.1)

where η = ±1 for (K, K 0 ) point and vF is the Fermi velocity. In the following section
we consider ~ = vF = 1, for simplicity. For the ZGNR with zigzag edge along the x
0
direction (see Fig. 3.2), we use the boundary condition φK K
A (x = L) = φA (x = L) =
0
φK K
B (x = 0) = φB (x = 0) = 0 [192], L being the width of the ribbon and the suffix

A, B denotes the type of sublattice. For ZGNR, the solution around K and K 0 points
are decoupled and we can proceed with any one of them separately (see Fig. 3.3).
In the following sections we will consider only K (η = 1) point. The wavefunction is
given by
 
s sin(kn (y − L)) ikx x
ψs (kn , kx , x, y) = An  e (3.2)
sin(kn y)
 − 1
sin(2kn L) 2
An = L− ,
2kn

where s = ±1 for positive/negative energy states. kn corresponds to the nth mode of

transverse momentum ky which is related to kx by the relation

kn
kx = , (3.3)
tan(kn L)

and the corresponding energy eigenvalue is given by

 
s
 kn 
E = s
 . (3.4)
sin(kn L) 
 71
The band structure is displays in Fig 3.3. The two K and K 0 valleys are clearly
separated and possess both bulk states (blue lines), and edge states (red lines).

K K¢

EHÑ vF LL

kx L

Figure 3.3: Band structure of a ZGNR near K and K 0 point. The blue lines are
due to the bulk state and the red lines are due to the edge states. The green region
denotes the continuous spectrum for an infinite graphene sheet.

We now build the Gaussian wave packet accounting for all available states. For
Zitterbewegung to take place, the initial wavepacket must contain both positive and
negative energy components. The wavepacket with an initial momentum (kx0 , ky0 )
reads

Ψ(x, y, t) (3.5)
1 d
Z
1 X − 1 ((kx −kx0 )+(kn −ky0 ))2 d2
=√ √ dkx e 2
2π π ζ(kx )
kn [kx ]
1
√ ψ+ (kn , kx , x, y)e−iEn t + ψ− (kn , kx , x, y)eiEn t ,


2
q 0 2 d2
√d e−1/2(kn −ky )
P
where ζ(kx ) = π kn [kx ] . For large ZGNR, the transverse momenta
(kn ) will be more closely spaced and ζ(kx ) → 1. For an infinite graphene sheet kn is
continuous and ζ(kx ) = 1. To any given kx corresponds a discrete set of kn which is
taken into account by the summation. The expectation values of the position along
 72
hx(t)iΨ and perpendicular hy(t)iΨ to the direction of motion are defined

Z
hx(t)iΨ = dxdyΨ(x, y, t)† xΨ(x, y, t), (3.6)
Z
hy(t)iΨ = dxdyΨ(x, y, t)† yΨ(x, y, t). (3.7)

Since the wave packet consists of both positive and negative energy states, the ex-
pectation value of an operator as well will contain two parts [193], one non oscillatory
part coming from the individual contribution of positive and negative energy states
and a second oscillatory part coming from the interference between the positive and
negative states which we are interested in. We define the Zitterbewegung component
of an operator as

1
OZ = hOiΨ − (hOi+ + hOi− ), (3.8)
2

where hOi+,− corresponds the expectation value evaluated for a wavepacket made of
only positive/negative energy states with the same Gaussian weight as Ψ(x, y, t).

3.2 Evolution of wave packet and zitterbewegung in graphene

zigzag nanoribbon

The zigzag boundary condition not only quantizes ky but also couples kx and ky [see
Eq. (3.3)]. However in the limit of large kx , i.e. |kx L|  1, kn → nπ/L, which is
close to a harmonic oscillator confinement, as seen in Fig. 3.4.
For our analysis we have chosen ky0 = 0, −10 ≤ kx0 L ≤ 14 as well as different
widths d of the Gaussian distribution. The Gaussian width controls the number of
transverse mode in the wavepacket. The range of kx0 covers both the regions where kn ’s
are almost independent of kx (|kx L − 1|  0) and where kn ’s are strongly dependent
on kx (kx L ∼ 1) (Fig. 3.4). Due to the Gaussian weight, for kx0 L  1, the wavepacket
 73
5Π
4Π
3Π

ky L
2Π
Π
0
-15 -10 -5 0 5 10 15
kx L

Figure 3.4: Variation of ky with kx at K(kx = 0) point. The solid blue lines are real
part of kx and the red dashed line is the imaginary part. The states on the left side
of the vertical grey dashed line are bulk states (kx is real) and the states on the right
side are edge states (the imaginary part of kx is finite).

will be dominated by bulk states and tends to localize at the center of the ribbon. For
kx0 L  1 the wavepacket will be dominated by edge states and tends to localize near
the edges. Near kx0 L = 1 both states are present, their relative weight depending on
d. These three types of wavepackets (bulk dominated, edge dominated and bulk-edge
overlapping) are represented in Fig. 3.5 at t = 0. One can see from Fig.3.5(b) that
each wavepacket consists of two parts with opposite velocity which get separated in
time. The overlap between bulk and edge states is the driving factor behind the
Zitterbewegung in the present system. When both states have comparable share,
the resultant wavepacket is be more delocalised and displays maximum oscillation
amplitude.
Let us now investigate the time-dependence of the position of the wavepacket.
Fig. 3.6 shows the expectation value hx(t)iΨ and hy(t)iΨ for kx0 = 1/L and d = L/4.
As mentioned above, one can see that the total expectation value is not simply the
sum of the individual contributions of positive and negative energy wavepackets, but
also displays an oscillatory contribution arising from the interplay between these
two energy states. The wavepacket consists of both bulk and edge states with both
positive and negative energy which produces five types of interference. (i) B± ↔ B± ,
 74

Figure 3.5: (a)Norm of the wavepacket at t = 0 for three different values of kx0 with
d = L/4. For kx0 = −8/L the bulk states are dominant. For kx0 = 1/L both bulk
and edge gives the same contribution and for kx0 = 8/L the states are localized at the
edges. The width of the ribbon is L and same scale is used for the length. The blue
dashed line denotes x = 0. (b) Evolution of a wavepacket with d=L/4 and kx0 = −4/L
in time where the time is measured in units of L/vF .

(ii)B± ↔ B∓ , (iii) B± ↔ E± , (iv) B± ↔ E∓ , (v) E± ↔ E∓ where B(E)+(−) means

bulk(edge) state with positive(negative) energy. Since the edge states have energy
close to zero (v) does not give any significant contribution to oscillation and (iii) and
(iv) almost similar contribution. The bulk states with same sign of energy have a
smaller gap with respect to that with opposite sign which makes the contribution
from (i) also negligible. Hence the dominant contribution will come from (ii) and
(iii). Notice that here we considered states at the K valley only. Similar result can

0.65
1.0
0.60
0.5
XxHtL\L

XyHtL\L

0.55
0.0 Y
0.50
-0.5 +
-1.0 0.45
-
-1.5 0.40
0 1 2 3 4 0 1 2 3 4
tHLvF L tHLvF L

Figure 3.6: (Color online) Expectation value for x and y for the complete wavefunction
(Ψ) and for positive and negative component (+/−) respectively. Here we use kx0 =
1/L and d = L/4

be obtained for K 0 valley as well, but in that case the values will have opposite sign.
However so long the order of magnitude of d is same as L which is roughly 100nm, i.e.
 75
1000 times more than the interatomic distance, there is no possibility of inter valley
mixing. Notice that for a large width or large value of kx0 there can be contribution
from both valleys which can cancel each other.

HaL HbL

0.04
0.4
0.02 0.3
xZ HtLL

yZ HtLL
0.00 0.2
-0.02 0.1
-0.04 0.0
0 1 2 3 4 0 1 2 3 4
tHLvF L tHLvF L

HcL HdL

0.04 0.4
0.03 0.3
Z L

Z L
xMax

yMax

0.02 0.2
0.01 0.1
0.00 0.0
-10 -5 0 5 10 -10 -5 0 5 10
k0x L k0x L
k0x L

-10 -5 0 5 10

Figure 3.7: Zitterbewegung component of (a) hx(t)i and (b) hy(t)i for a wavepacket
with Gaussian width d=L/4. (c) and (d) shows the variation of amplitude with the
central momentum.

Figures 3.7(a) and (b) show xZ (t) and yZ (t) for a wavepacket with a Gaussian
width d = L/4. The maximum amplitude xMax
Z and yZMax are reported on Figs. 3.7(c)
and (d), respectively. One can see that both xZ (t) and yZ (t) present a damped os-
cillatory behaviour. In the case of xZ (t), the decay length is much longer than the
oscillation period, while in the case of yZ (t), the decay is much stronger which sup-
presses the oscillatory behaviour. Note also that the period of oscillation of xZ (t) is
essentially constant ∼ π/2. In addition, yZ (t) shows a saturation towards L/2, which
 76
is quite expected as the amplitude can’t exceed the ribbon width. One remarkable dif-
ference of the Zitterbewegung effect for ZGNR with respect to other Dirac systems is
the existence of oscillation along the direction of motion, i.e. xZ (t). If the longitudinal
and transverse momenta were not coupled, one can readily check, using Heisenberg
equation of motion, that the acceleration of a Dirac particle is always perpendicu-
lar to its initial momentum. However if the momenta are coupled, a longitudinal
component of acceleration develops causing a longitudinal Zitterbewegung.
The other striking characteristic of the Zitterbewegung effect is the appearance
of a resonance in xZ (t) oscillations, as shown in Fig. 3.7(c). From Fig. 3.4 one can
see that the dependence between kx and ky is strongly nonlinear near kx L = 1, which
corresponds to the rise of edge states. We already showed in Fig. 3.5, that bulk and
edge states tend to localize in opposite ends along the transverse direction. From Fig.
3.3 one can also see that for a particular kx the bulk and edge states have opposite
propagation velocity as suggested by the slope of the energy bands. Hence maximum
oscillation is expected when these two states contribute equally.
This feature can be further clarified when considering different Gaussian widths.
Following earlier studies [193] we assume that the envelope of the oscillating xZ (t) is
of exponential nature, i.e.

xEnvelope
Z = xM
Z
ax −Γt
e , (3.9)

where xEnvelope
Z is the value of xZ (t) at the peaks which describes the envelope for
xZ (t) and xM
Z
ax
is the maximum peak value. Figure 3.8(a) and (b) show xM
Z
ax
and Γ,
respectively, for different values of the width d.
From Fig. 3.8, one observes that as d decreases (i.e. the wavepacket becomes
more localized in real space and broader in momentum space), both the amplitude
(xM
Z
ax
) and damping (Γ) increase. The resonant amplitude however doesn’t increase
 77
HbL
HaL
æ 0.35 æ
æ
æàæ
0.30 æ
à
0.04 æ
à àæ ì
ò
ô æ L 4
à ìì æ L 4 0.25
àæ æ
Z L 0.03 ì à à L 3.5
æì
à à L 3.5 0.20 ì
xMax

G
æ
òò ìà ò
0.02 ìò ì L 3 0.15 à ô ì L 3
ò æ ì
ì æ à
æ
ò ò ò L 2.5 0.10 æ
æ
à ì
ò ò L 2.5
0.01 à
ô
ôô à
æ à ì ò
ô
æ ì ô ô ò ì
æ
à ô L 2
0.05æà æ
à
à
ì ì ò ô ô L 2
à ò ôô ò æ ì
ò ì
ò ô
ò ò
ô ô
0.00 ìôòæà ìôòæà ìôòæà ìôò ô ì
ò à
ì æ
à
ì
ô ô ò
ô ò
ô 0.00 ô
-10 -5 0 5 10 -4 -2 0 2 4 6
k0x L k0x L

Figure 3.8: Variation of (a) xM

Z
ax
and (b) Γ with kx0 for different d’s (given in legend).

monotonically and rather tends towards a saturation which is clear from Fig. 3.8(a).
The change of Γ is more prominent when xM
Z
ax
undergoes the resonance. The reso-
nance takes place roughly within the region 2 . kx0 L . 3. The larger d, the closer the
resonance is to kx L = 1. This behaviour can be explained based on our earlier ar-
gument. As d increases, less bulk states contribute to the wavepacket. Consequently
the bulk and edge share similar amount of contribution and the wavepacket becomes
more delocalised causing an increase in the amplitude of xZ (t). Since the contribution
from the bulk decreases, the condition when the bulk and edge share equal contri-
bution becomes closer to kx L = 1, and hence the resonant position will also shift
accordingly. The damping also depends on the amount of interference among states.
For smaller d more states interfere with each other which results in faster damping
[see Fig. 3.8(b)].
It is worth mentioning that the present analysis also holds with the current density
jx = evF hσx i. But one can readily see that vF hσx i = hdx/dti = hdxi /dt and hence
the qualitative nature of the resonance and damping will be the same.
It is instructive to compare the present results with previous works [194, 193, 195],
which are summarized in Table 3.1. Although the values provided in this Table
are orders of magnitude, they indicate that in extended graphene monolayers and
bilayers (Gr monolayer and Gr bilayer), the frequency of oscillation is extremely fast
(τ ∼ fs) compared to nanoribbons (100nm ZGNR) and semiconducting nanowires
 78
System d (nm) λ0 (nm) xMax
Z (nm) T (fs) Γ−1 (fs)
50nm InAs[194] ? 95 35 2500 -
Gr Monolayer [193] 40 5 5 2.5 15
Gr Bilayer [193] 30 18 5 2.5 50
Gr+2.5T [195] 8 18 5 650 900
Gr+10T [195] 8 10 5 280 350
100nm ZGNR 25 16 5 100 1000
Table 3.1: Comparison between different characteristics of the Zitterbewegung in
various systems: InAs nanowire, Graphene (Gr) mono and bilayers with and without
a magnetic field. The last line is our results. The quantities represented are the width
of the wavepacket d, the initial wavelength λ0 = 2π/k0 , amplitude xMax Z , oscillation
period τ , and damping rate Γ−1 . In the case of semiconducting nanowire and ZGNR,
we use the values are resonance.

(50nm InAs) (τ ∼0.1-1ps). Furthermore, the state quantisation in extended graphene

results in a strong damping of the Zitterbewegung effect. This damping can be
dramatically reduced by opening Landau levels using a magnetic field (Gr+2.5T and
Gr+10T), which is accompanied by a slowdown of the oscillation frequency. The
results discussed in the present work (100nm ZGNR) compare favourably with the
results obtained using semiconducting nanowire or magnetic field in graphene. It
therefore demonstrates that using bulk-edge state overlap to achieve Zitterbewegung
resonance is a valuable route to observe this effect experimentally.

3.3 Conclusion

In this work we investigated the Zitterbewegung effect in ZGNR. In sharp contrast

to other Dirac systems such as extended graphene or asymmetric semiconducting
nanowires, the trembling motion appears along the direction of propagation of the
wavepacket. We argue that this peculiar feature results from the coupling between
the momenta along longitudinal and transverse direction of motion caused by the
zigzag boundary condition. Such a coupling does not exist in armchair nanoribbons.
Furthermore, by tuning the initial longitudinal wavevector, the wavepacket enters in
resonance which results in a significant increase of the magnitude of the oscillation.
 79
This resonance is attributed to the interference between the bulk and edge states and
appears when these states equally contribute to the wavepacket. The main physical
reason behind the resonant longitudinal Zitterbewegung is associated with the differ-
ence of propagation velocity of bulk and edge states. In support of our argument we
further demonstrate that the resonance can be controlled by modifying the Gaussian
width and therefore tuning the relative contributions of bulk and edge states. In fact,
the longitudinal Zitterbewegung is thus a true signature of interference between bulk
and edge states. This study shows that graphene nanoribbons with zigzag edges are
suitable for the observation of Zitterbewegung effect.
 80

Chapter 4

Probing topological phases with wavepacket dynamics

In this chapter we demonstrate that Zitterbewegung can be used to probe the different
topological phases of two-dimensional hexagonal lattices. For our study we choose
silicene, a two dimensional buckled honeycomb lattice with strong spin-orbit coupling,
where one can tune the topological phases with an external electric field and onsite
magnetization [179]. We present a systematic analysis of Zitterbewegung in different
topological phases that enable us to recover the topological phase diagram of the
material. The formalism we adopted here is quite generic and hence applicable to
any Dirac material
In a buckled honeycomb lattice like silicene, germanene or stanene the two sub-
lattices lies in two different planes (Fig.4.1) which gives them stronger spin orbit
coupling. In addition one can further tune the band gap and spin polarisation at
each valley by applying a perpendicular electric field (Fig.4.1).

Figure 4.1: (a) Geometrical structure of a buckled honeycomb lattice. (b) Formation
of Dirac cones at K and K 0 valleys with opposite spin represented by red and blue
colour.
 81
4.1 Low energy Hamiltonian, eigenvalues and eigenstates of
buckled honeycomb lattice

The Hamiltonian for near K and K 0 points is given by [179]

Ĥη = ~vF (ηkx τ̂x + ky τ̂y ) + ητ̂z ĥ11 − `EZ τ̂z + M σ̂z

+λR1 (ητ̂x σ̂y − τ̂y σ̂x )/2,

ĥ11 = λSO σ̂z + aλR2 (ky σ̂x − kx σ̂y ), (4.1)

where η = ±1 corresponds K and K 0 valley, σ̂ and τ̂ are Pauli matrices for spin and
valley, a is the interatomic distance and ` is the buckling height. EZ is an external
field applied perpendicular to the plane and M is the onsite magnetization. λSO
is the spin-orbit coupling, λR is the second nearest Rashba parameter. The system
undergoes a topological phase transition at a critical electric field EC = λSO /` for
M = 0. The values of the parameters for different materials are given in Table 4.1.

Table 4.1: Different parameter values for graphene (Gr), silicene (Si), germanene (Ge)
and stanene (Sn) [105, 181].
Atom a ` λSO λR vF EC
5
(Å) (Å) (meV) (meV) 10 m/s (meV/Å)
Gr 2.46 0.00 10−3 0.0 9.8 ∞
Si 3.86 0.23 3.9 0.7 5.5 17
Ge 4.02 0.33 43.0 10.7 4.6 130.3
Sn 4.70 0.40 100 9.5 4.9 250

In practice λR  λSO and we can drop this term. In that case the spin up and
down Hamiltonians decouple and we can write them as 2 × 2 matrices,
 82

 
−`EZ + s(M + ηλSO ) vF (−iky + ηkx ) 
Hη,s =  .
vF (iky + ηkx ) `EZ + s(M − ηλSO )
(4.2)

This Hamiltonian is analytically solvable, and the eigenvalues and eigenfunctions

read

± 1 A
Eη,s = (m + mB η,s )
2 η,s
1q A
± (mη,s − mB 2 2 2
η,s ) + 4vF (kx + ky ), (4.3)
 2 
±
± ~ cos(Θη,s )
ψη,s (k) =  , (4.4)
sin(Θ±
η,s )

A/B vF (iky +ηkx )

where, mη,s = ∓`Ez + s(M + ηλSO ), Θ±
η,s = tan
−1
±
Eη,s −mB
. Due to the valley
η,s

dependence of spin-orbit coupling, for EZ = ±EC carriers with opposite spin projec-
tion form Dirac cones at opposite valleys (Fig. 4.2). Consequently, at a particular
valley carriers with opposite spin projection undergo topological transition at a dif-
ferent critical electric field. In other words, as illustrated in Fig. 4.2, at K (K’) point,
spin up (down) band exhibits a Dirac cone at EZ = EC , while spin down (up) band
presents an orbital gap. The situation is reversed for EZ = −EC .

4.2 Evolution of wavepacket and zitterbewegung in buckled

honeycomb lattice

Since there is no position dependent term in the Hamiltonian, momentum is a con-

served quantity and we can use the momentum eigenstates to create a wave packet.
To do so, we use a Gaussian envelop for the momentum distribution so that the
wavepacket in the real space is also Gaussian. By choosing a narrow width of the
 83

EZ = EC EZ = -EC

K¢ K K¢ K

¯ ­ ­ ¯

2.5
C­ =+1
2.0
C¯ =-1
DΛSO
1.5
C­ =1 C­ =-1
1.0
C¯ =1 C¯ =-1
0.5 ¯ ­
0.0
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5
EZ EC

Figure 4.2: Top: Dirac cone for EZ = ±EC at K and K 0 valley. Bottom: Variation
of band gap for spin up and down at K valley for M = 0. The green and orange
regions correspond to quantum spin hall phase (Cs = 1) and bulk insulator (Cs = 0)
phase where Cs = (C↑ − C↓ )/2 is the spin Chern number.

momentum distribution, we can avoid valley mixing. From now on we will focus on
K (η = 1) valley only and drop the valley index. The wavepacket can be expressed
as [193]

Z
1 1 +
Ψs (~r, t) = g(~k, k~0 ) √ [ψs+ (~k)e−iEs t
N 2
− ~ −iEs− t i~k.~
+ψs (k)e ]e r d~k, (4.5)
d ~ ~ 2 2
g(~k, k~0 ) = √ e−1/2(k−k0 ) d , (4.6)
π

where d is the width of momentum distribution and N is the normalization factor.

Eq. 4.6 applies to a two band system, but can be generalized to multiband system
by adding up all contributing states. Due to the Gaussian envelop, only a selected
portion of the bands contributes to the wavepacket as shown in Fig. 6.3. If the Fermi
 84
level lies in the middle of two bands, then both bands contribute to the wavepacket
(Fig. 4.3a). On the other hand if the Fermi level cuts one of the bands, only finite
region of the selected portion can make a contribution (Fig. 4.3b).

HaL HbL

Figure 4.3: Formation of Gaussian wavepacket. The blue and red lines corresponds
up and down spin bands and the green line shows the Gaussian distribution. The
shaded region shows the portion of bands that contributes to the wavepacket.

Once we construct the wavepacket, we can easily calculate the Zitterbewegung

component of the position as described in Ref. [196]. Let Ψ+ r, t) and Ψ−
s (~ r, t) be
s (~

two wavepackets with the same momentum distribution and made of only positive
and negative energy states. The Zitterbewegung component of an operator Ô is given
by

1 − −
OZB (s, t) = hΨs |Ô|Ψs i − (hΨ+ +
s |Ô|Ψs i + hΨs |Ô|Ψs i), (4.7)
2

where the subscript ZB denotes the Zitterbewegung component. Note that our defi-
nition is same as [193] where they define Zitterbewegung component as the sum over
expectation values due to overlap integral between different states.
As mentioned above, depending on the values of M and EZ silicene exhibits dif-
ferent topological phases, which have been described, for instance, in Ref. [179]. The
topological phase diagram of silicene when varying EZ and M is reported in Fig.
4.7 by the dashed lines. In the following, we explore the nature of Zitterbewegung
throughout the phase diagram and establish a correlation between the different topo-
 85
logical phases and Zitterbewegung features. To do so, we construct wavepackets in
each of the regions of the topological phase diagram of silicene, and calculate the
Zitterbewegung component of the coordinates. We choose the central momentum of
the wavepacket to be k~0 = 0.00025/ax̂ and the width of the Gaussian distribution to
be d = 5000a.
First let us focus on two points on the M = 0 axis, say EZ = 1.5EC and EZ =
0.5EC . The first case corresponds to a topologically trivial phase for both spin up
and spin down bands. In the second case, spin up is in topologically nontrivial phase
while spin down is in topologically trivial phase. The band structures for these two
cases are reported on the top and bottom left panels of Fig. 4.4, respectively. We also
compute the Zitterbewegung contributions to the position of the wavepacket (xZB ,
yZB ) as a function of time. These results are reported on the top and bottom right
panels of Fig. 4.4 for the EZ = 0.5EC and EZ = 1.5EC , respectively.

EZ = 0.5 EC, M = 0 ΛSO

E 0.2 0.3
0.2
yZB HΜmL

0.1
xZB HΜmL

0.1
0.0 0.0
­ -0.1 -0.1
¯ -0.2 -0.2
-0.3
0 2.5 5 0 2.5 5
t HpsL t HpsL
EZ = 1.5 EC, M = 0 ΛSO
E 0.2 0.3
xZB HΜmL

0.2
yZB HΜmL

0.1 0.1
­ 0.0 0.0
-0.1 -0.1
¯ -0.2
-0.2
-0.3
0 2.5 5 0 2.5 5
t HpsL t HpsL

Figure 4.4: Zitterbewegung component of x and y coordinates for different values of

EZ and M . The left panels show the corresponding band structure.
 86
One can readily see that the qualitative nature of the band structures is the same
in both cases - up spin having a smaller gap and the down spin having a larger gap.
xZB however behaves completely differently in these two cases. We can see that xZB
for spin up states undergoes a π phase shift when moving from topologically nontrivial
to trivial phase while the phase for down spin remains the same. yZ on the other
hand does not show any qualitative change.
Let us focus on xZB for a spin unpolarized wavepacket. The change of phase in xZ
oscillation is simply related to the inversion of the band gap through the topological
transition. Depending on the values of EZB and M , different spin components oscillate
with different amplitude and frequency resulting in beating in charge and spin density
waves, as illustrated on Fig. 4.5.

­ ¯ H­+¯L

­ ¯ H­-¯L

Figure 4.5: Formation of beating in charge (top) and spin (bottom) density waves.

Modeling the oscillations for individual spins as damped sinusoids (Ae−Γt sin(ωt +
δ0 )) [196] we can easily evaluate the amplitude (A), damping factor (Γ), frequency
(ω) and epoch (δ0 ) from which we can calculate the beating frequencies. When the
amplitudes of both oscillations are the same, the beating frequency is simply the mean
of two frequencies. For different amplitudes, however, the beats would not be exactly
periodic and in that case we would consider the average frequency. One can also detect
the topological states of either spins from the beating pattern (Fig. 4.6). If spin up
and down belong to opposite topological states, then the oscillations are out of phase.
 87
Consequently the charge density increases initially while the spin density decreases.
When the states are in the same topological phase, we observe the reverse pattern.
One should note that the current is directly proportional to the time derivative of the
position, and hence one can see the same oscillation in current as well.

Topologically Topologically
different same

­ ¯ H­+¯L

­ ¯ H­-¯L

Figure 4.6: Nature of beating when spin up and down are in the same and opposite
topological phases.

Let us now look at the amplitude of the beating. From Fig. 4.6 one can see
that (↑ − ↓) oscillation does not provide any new information compared to (↑ + ↓)
oscillations and hence we focus on (↑ + ↓) wavepacket only. We choose different
EZ , M combination and calculate the amplitude frequency and initial phase of the
resultant beating. We define a relative phase factor Θ = ±1 which indicates whether
the two spin projections have the same (Θ = +1) or opposite (Θ = −1) initial phase,
and plot the product of Θ and amplitude of the beating over the whole EZ − M
space. Fig. 4.7 displays the modulated amplitude of the longitudinal Zitterbewegung,
ΘxZB , when varying both EZ and M for for (a) d=10000a, (b) d=5000a, and (c)
d=2000a. The corresponding topological phase diagram calculated by Ezawa [179]
is indicated by the dashed lines. We find a good match between our analysis and
the analytical phase diagram. Notice though that since the Zitterbewegung effect
 88
involves interference among states within a range of momentum, the accuracy of
the boundaries between different regions of the phase diagram are sensitive to the
width of the wave packet. For a spatially wide wave packet [d=10000a≈300-500nm,
Fig. 4.7(a)], a small number of states are involved and the boundaries are well
defined. However, upon reducing the wavepacket width, more states are involved
in the Zitterbewegung process and the boundaries deteriorate [d=2000a≈ 60-100nm,
Fig. 4.7(a)]. Therefore, a good definition of the boundaries of the topological phase
diagram requires the use of a spatially wide wavepacket.

4.3 Conclusion

We have demonstrated that Zitterbewegung features correlate with the various topo-
logical phases of two-dimensional hexagonal lattices. By analyzing the longitudinal
jittering motion of the wavepacket, we were able to reconstruct the phase diagram of
silicene up to a good accuracy, providing that the wavepacket considered is spatially
wide. An interesting aspect of this analysis is that is provides access to the bulk
properties of the material directly without the need for searching for quantized edge
states. The unpolarized wavepacket described in the present work can be realized
and detected using optical techniques such as pump-probe method. Such techniques
have been recently exploited to investigate the ultrafast dynamics of Dirac electrons
in graphene [197, 198, 199, 200]. In this context, the search for the Zitterbwegung ef-
fect and potential signatures of topological phase transition in hexagonal honeycomb
lattices constitute an appealing experimental challenge.
 89

Figure 4.7: Product of the beat (↑ + ↓) amplitude of longitudinal Zitterbewegung

(xAmp
ZB ) and relative phase factor (Θ) over the EZ − M plane for different width of
momentum distribution. The top (bottom) panel corresponds to a broad (narrow)
spatial distribution, i.e. a narrow (broad) momentum distribution.
 90

Chapter 5

Ferromagnetic and antiferromagnetic topological insulator

In this chapter, using scattering wave function formalism implemented on a tight-

binding model, we explore the nature of spin transport and torque in two dimensional
ferromagnetic (FTI) and antiferromagnetic (AFTI) topological insulators. We find
that AFTI is more robust against disorder than FTI, such that topological edge states
are preserved even under weak disorder. Most importantly, SOT possesses two com-
ponents: a field-like torque (odd under magnetization reversal) and an antidamping
torque (even under magnetization reversal). While the former is directly generated
by the spin-momentum locking at the edges, the latter arises upon scattering and is
quite sensitive to disorder and size effects.

5.1 Method

We start from the Bernevig-Hughes-Zhang model [201] on a square lattice. We use

the basis (1 ↑, 2 ↑, 1 ↓, 2 ↓)T , where 1, 2 refer to two orbitals and ↑, ↓ refer to spin
projections, and define the TI Hamiltonian by a 4 × 4 matrix,

 
 h(k) 0
H(k) =  , (5.1)

0 h∗ (−k)

where h(k) is given by

 91

h(k) = [M + B(cos(kx ) + cos(ky ))]σz

+A sin(kx )σx + A sin(ky )σy . (5.2)

Here A, B, M are model parameters whose values depend on the real structure [201,
61]. The topologically nontrivial phases appear for B > |M/2|, which is manifested
as gapless edge states in quasi one dimensional systems. In case of a CdTe-HgTe
quantum well this is achieved by tuning the width of the quantum well. For our
calculations we choose A to be the unit of energy and consider B = 1.0A, M = −1.5A
that ensures the existence of topologically protected edge states for nonmagnetic TI.
To map this bulk Hamiltonian (5.1) on a finite scattering region, we first extract the
tight-binding parameters [159, 202] by expressing

H(k) = H0 + Hx̂ eikx + Hŷ eiky + Hx̂† e−ikx + Hŷ† e−iky ,

(5.3)

with
 
 M 0 0 0 
 
 0 −M 0 0 
H0 =  ,
 
 0
 0 M 0 

 
0 0 0 −M
   
 B −iA 0 0   B −A 0 0 
   
1  −iA −B 0
 0   A −B 0
1 0 
Hx̂ = , Hŷ =  .
 
2 0 2 0

 0 B iA 
  0 B −A 

   
0 0 iA −B 0 0 A −B
 92
We can use these hopping elements to construct a real space Hamiltonian for a finite
system as

X X
H = c†i H0 ci + c†i tij cj (5.4)
i i6=j

where tij = Hr̂ij , r̂ij (=±x̂, ±ŷ) being the unit vector between nearest neighbor sites i
and j, and c†i (ci ) is the creation (destruction) operator for the state (1 ↑, 2 ↑, 1 ↓, 2 ↓)T
at i-th site. The coupling between itinerant spins ~s and the local magnetization (m
~ i ),
as well as the disorder potential Vir are introduced in the onsite energy as

X X
imp
Hm = c†i (H0 + m
~ i · ~s + Vir I4 )ci + c†i tij cj , (5.5)
i i6=j

where In is the n-th rank identity matrix, and ~s = (ŝx , ŝy , ŝz ), with

     
 0 I2   0 −iI2   I2 0 
ŝx =   , ŝy =   , ŝz =  . (5.6)
I2 0 iI2 0 0 −I2

F B A G
y −→

x −→

Figure 5.1: Schematic of FTI and different types of AFTI. The green region shows
one unit cell of the lead. Blue and red dots represent positive and negative m
~ i.
In the following, we consider five different configurations, defined by the spatial
modulation of m
~ i : an ordinary, non-magnetic TI as a reference (referred to as O), an
FTI (F ), and A-, B-, and G-type AFTI configurations [(A), (B) and (G) in Fig. 5.1].
The total system can be divided into three parts - (i) left lead, (ii) scattering region
and (iii) right lead. The leads are semi-infinite and can be characterized by one unit
 93
cell (green shaded region in Fig. 5.1) whereas the scattering region is defined by Eq.
(5.5). Note that for A, G type AFTI we need to double the unit cell of the lead to
maintain the translation symmetry. For this work, we consider a scattering region
composed of 40 × 20 sites arranged on a square lattice. To calculate the transport
properties we adopt the wavefunction approach, as implemented in the tight-binding
software KWANT [203]. This approach is equivalent to the non-equilibrium Green’s
function formalism [204]. In this method one starts by defining the incoming modes at
a particular energy in terms of eigenstates of an infinite lead and subsequently obtain
the wavefunction within the scattering region by using the continuity relations. By
applying this method throughout the scattering region one can obtain the outgoing
modes that can be exploited to construct the S-matrix of the system. The scattering
wavefunction and the S-matrix are two basic outputs one can obtain from KWANT
for any given system (see Section 2 of Ref. [203] for details). The conductance of
the system is calculated from the S-matrix using Landauer-Büttiker formalism. To
calculate the non-equilibrium spin density at some given energy EF , we use a small
bias voltage VBias = µL −µR , where eµL(R) = EF ±eVBias /2 are the chemical potential
of the left (right) lead. We use the scattering wavefunction calculated by KWANT
and evaluate the expectation values of different spin components integrated over the
bias window to get the total non-equilibrium spin density as,

Z µL
~ neq =
S hψi (E)|~s|ψi (E)idE, (5.7)
i
µR

where ~s is the onsite spin operator defined in Eq. (5.6), and ψi (E) is the scattering
wavefunction for i-th site at energy E. Once we get the non-equilibrium spin density
we can calculate the onsite SOT as

τ~i = m ~ neq .
~i × S (5.8)
i
 94
Finally, in order to introduce nonmagnetic disorder in the system we add to the
Hamiltonian, Eq. (5.5), a random onsite energy Vir uniformly distributed over the
range [−V0 , V0 ]. This gets rid of any possible shift of energy spectrum that might
appear if one chose only positive amplitudes for the disorder potential. The transport
properties are then averaged over 1280 random disorder configurations.

5.2 Robustness of different magnetic configurations

Let us first compute the impact of disorder on the conductance in the various magnetic
configurations. Fig. 5.2(a,b) displays the behavior of conductance as a function of
the disorder strength when the direction of the magnetic moments m
~ i is (a) out of
plane and (b) in the plane. Here, the transport energy is taken EF = 0.25A.

(a) m~ = 0. 2Aẑ (b) m~ = 0. 2Aŷ

2.0 2.0

1.5 1.5 O
2.0 G
A
G/G0

G/G0

1.0 O 1.0
G 1.8 B
A F
0.5 B 1.61.0 1.5 2.0 2.5
0.5
F
0.00.0 0.5 1.0 1.5 2.0 2.5 0.00.0 0.5 1.0 1.5 2.0 2.5
V0 (A) V0 (A)

Figure 5.2: Conductance (G) of TI with different magnetic configurations against

random disorder for (a) out of plane and (b) in plane magnetic moments. The con-
ductance is normalized to the conductance quantum G0 = 2e2 /h. The boxed portion
of (a) is enlarged in the inset.

From Fig. 5.2(a) we see that when the magnetic moments lie out-of-plane FTI (F)
is comparatively more sensitive to disorder than AFTI (A, B, G) and nonmagnetic TI
(O), although the difference of robustness between the nonmagnetic and magnetic TI
is not very large. The initial quantized conductance of FTI starts decreasing around
V0 ∼ 1.5A due to the progressive quenching of the topological protection of the edge
modes, while in contrast, AFTI and nonmagnetic TI maintain their topological egde
 95
states up to V0 ∼ 2A. The difference becomes quite significant when we set the
magnetic order in the plane, see Fig. 5.2(b). Noticeably, F and B cases are very
sensitive to disorder, while A and G cases are much more robust. This indicates that
AFTIs with an in-plane staggered magnetic character along the transport direction
remain topological insulators even for weak disorder.
For a better understanding of this effect, we calculate the density of states of
the TIs in the absence of disorder, see Fig. 5.3(a). An in-plane magnetic order
opens a gap for F and B, while A and G can preserve their gapless states similarly
to the nonmagnetic TI (O). Since the topological protection is stronger at lower
energy, we calculate the robustness at EF = 0.05A [Fig. 5.3(b)] and find the quantized
conductance due to the edge states of A, G and O cases survives longer compared to
that evaluated at EF = 0.25A. B and F types have a gap at that energy and hence
show zero conductance.

60 (a)m~ = 0. 2Aŷ (b)m~ = 0. 2Aŷ

O 2.0
50 G
A 1.5
40
Total DOS

B
30 1.0
G/G0

F O
20 G
0.5 A
10 B
0.0 F
00.3 0.2 0.1 0.0 0.1 0.2 0.3 0.0 0.5 1.0 1.5 2.0 2.5
E(A) V0 (A)

Figure 5.3: (a) Total density of states (DOS) for TI, FTI and different AFTI. (b)
Conductance against random disorder at EF = 0.05A (filled symbols) and at EF =
0.25A (open symbols) for different configurations with an in-plane magnetic moment
0.2Aŷ.

From now on, we proceed with only FTI and G-type AFTI as they qualitatively
behave similarly as B-type AFTI and A-type AFTI, respectively.
 96
5.3 Non-equilibrium spin density and Spin-orbit torque

As mentioned in the introduction, spin transfer torque [149, 205, 142] as well as SOT
[150, 206] can be used to control the direction of the antiferromagnetic order param-
eter. The order parameter can be controlled in two ways [143, 146, 145]: either using
a time-dependent (ac) field-like torque (i.e., a torque that is odd under magnetization
reversal), or using a time-independent (dc) antidamping torque (i.e., even under mag-
netization reversal). Our intention is to investigate the nature of SOT in the AFTI
case, where both topologically protected edge transport and antiferromagnetic order
parameter coexist.
First we calculate the total non-equilibrium spin density and the associated SOT
in FTI [Fig. 5.4(a,c)] and AFTI [Fig. 5.4(b,d)], with an in-plane magnetic order
~ i ∼ ŷ) and in the absence of disorder. In these calculations, we set |mi | = 0.2A,
(m
EF = 0.25A and (µL − µR ) = 0.02A.
Fig. 5.4(a,b) display the spatial distribution of the different components of the
non-equilibrium spin density in FTI and AFTI, respectively. The middle panels show
the spatial profile of Sy , the spin density component that is aligned along the magnetic
order. This component is uniform in FTI and staggered in AFTI, as expected from
the magnetic texture of these two systems. We observe finite Sz on both edges, which
is a characteristic feature of a TI (bottom panels) and more interestingly finite and
oscillatory Sx on both edges (top panels). The oscillation is caused by the scattering
at the interfaces between the conductor and leads. Since Sx and Sy are not immune
to scalar perturbation, the potential steps at the interfaces mix these components
depending on the chirality of each edges. Note that the amplitude of oscillation of
Sx in G-AFTI is two orders of magnitude smaller compared to that in FTI which
denotes that the scattering in G-AFTI is weaker compared to that in FTI.
From the symmetry we can easily recognize that Sz produces the so-called field-like
torque [~τFi ∼ m
~ i ×~z] and Sx gives rise to the antidamping torque [~τDi ∼ m
~ i × (~z × m
~ i )].
 97
FTI (m~ = 0. 2Aŷ) G − AFTI (m~ = 0. 2Aŷ)
(a) (b)
19 0.06 19 0.02

Sx (10 −2 )

Sx (10 −4 )
10 0.00 10 0.00

W
00 10 20 30 39 0.06 00 10 20 30 39 0.02
19 1.05 19 0.02

Sy (10 −2 )

Sy (10 −2 )
10 0.00 10 0.00
W

W
00 10 20 30 39 1.05 00 10 20 30 39 0.02
19 0.71 19 0.72

Sz (10 −2 )

Sz (10 −2 )
10 0.00 10 0.00
W

W
00 10 20 30 39 0.71 00 10 20 30 39 0.72
L L
1.5
(c) 1.5 (d)
τD (10 −3 A) τF (10 A)

τD (10 −6 A) τF (10 −3 A)
1.0
−3

0.5 0.0
0.0 1.5
0.2 0.5
0.0 0.0
0.2 10 20 30 39 0.50 10 20 30 39
L L

Figure 5.4: Non-equilibrium spin density for (a) FTI and (b) G-AFTI. We use a
magnetization strength m = 0.2A along ŷ and the spin densities are evaluated at
EF = 0.25A with a bias voltage (µL − µR ) = 0.02A. (c,d) shows the corresponding
field like (τF , magenta) and antidamping (τD , green) SOT evaluated at the top edge
for FTI and G-AFTI respectively.

Fig. 5.4(c,d) represent the spatial profile of the field-like (τF ) and antidamping torques
(τD ) at the top edges of the FTI and AFTI, respectively. To understand how these
torques evolve in the presence of disorder, we further study the robustness of Sx
and Sz in presence of scalar disorder (see Fig. 5.5). We define the (i) uniform spin
density (Sux,z = hSix,z i) and (ii) staggered spin density (Sstx,z = hsign(mi )Six,z i) where
the average is over the lattice sites. Since the spin density is localized at the edges
we calculate the robustness for the top edge only (Fig. 5.5). Similar results can also
be obtained for the bottom edge.
From Fig. 5.5 we can see that, correspondingly with conductance, the non-equilibrium
 98
0.8 (a)FTI 2.0 0.8 (b)G − AFTI 2.0
G G
0.6 1.5 0.6 Sux 1.5

Su , Sst (10 −2 )
Su (10 −2 )
0.4 Sux 1.0 0.4 Sstx 1.0

G/G0

G/G0
Suz Suz
0.2 0.5 0.2 Sstz 0.5
0.0 0.0 0.0 0.0
0 1 2 3 0 1 2 3
V0 (A) V0 (A)

Figure 5.5: Variation of uniform (Sux,z ) and staggered (Sstx,z ) spin densities as a function
of disorder strength with an in plane magnetic order |m ~ i | = 0.2A for (a) FTI and (b)
G-AFTI. The green dot-dashed line shows the corresponding conductance.

spin densities also fall down faster in FTI compared to AFTI. Due to its periodic mod-
ulation, Sux is initially zero for both FTI and AFTI. When increasing the disorder,
two different effects take place: (i) a progressive smearing of the edge wave function
accompanied by a reduction in Suz ; (ii) an increase of disorder-induced spin-dependent
scattering resulting in enhanced spin mixing. This disorder-induced spin mixing is at
x
the origin of the Sst,u component observed in Figs. 5.5(a) and (b). This mechanism
has been originally established in metallic spin-valves[207] and domain walls [208].
~0 ,
In a disordered ferromagnetic device submitted to a non-equilibrium spin density S
spin dephasing and relaxation produce an additional corrective spin density of the
form ∼ m ~0 . In the case of FTI, the magnetization is uniform so that a uniform
~ ×S
Sux ∼ Suz m
~ × ~z is produced [Fig. 5.5(a) and Fig. 5.6(a)]. In the case of AFTI, the
magnetization is staggered so that a staggered Sstx ∼ Suz m
~ i ×~z is generated [Fig. 5.5(b)
and Fig. 5.6(b)]. In the latter, no uniform Sux emerges. Notice that the build-up of
x
Sst,u upon disorder is a non-linear process as disorder increases spin mixing and re-
duces Suz , at the same time. Hence, one can identify three regimes of disorder. In the
case of AFTI displayed in Fig. 5.5(b):

• From V0 =0 to V0 ≈ A, Suz , remains (mostly) unaffected, while Sst,u

x
vanishes on
average.
 99
• From V0 ≈ A to V0 ≈ 2A, topological protection breaks down progressively
and Suz is reduced upon disorder due to increased delocalization of the edge
wavefunction. During this process, disorder enhances spin mixing and thereby
Sstx increases moderately. During this moderate increase the reduction of Suz is
compensated by the increase in spin mixing, thereby producing a finite Sstx .

• For V0 > 2A, Suz is further reduced and correspondingly Sstx decreases too, as the
disorder-driven spin mixing cannot compensate the reduction of Suz anymore.

Notice that although Suz  Sstx in the weak disorder limit, both spin density com-
ponents tend towards a similar value for large disorder (V0 > 2A), Suz ≈ Sstx . In the
case of FTI displayed in Fig. 5.5(a), the three regimes appear at different disorder
strengths due to the weaker topological protection of the edges states.
x
To illustrate the progressive build-up of Sst,u upon disorder, the spatial profile of
the spin density at the edge of the sample is reported on Fig. 5.6 for (a) FTI and
(b) AFTI. These calculations confirm that the overall increase in Sstx is a direct con-
sequence of spin-dependent scattering upon disorder. In the absence of disorder, the
component Sx displays a smooth oscillation (green dots), as discussed above. Follow-
ing the process described above, in FTI with positive magnetization, the scattering
creates mostly positive Sx along the top edge [Fig. 5.6(a)], while in AFTI due to the
staggered magnetization Sx acquires a staggered nature [Fig. 5.6(b)].
It is worth mentioning that since the staggered Sstx emerges as a correction to
Suz upon scattering, its magnitude is not only sensitive to disorder but also to the
dimension of the channel. As a matter of fact, Fig. 5.7 shows that for a given amount
of disorder, the magnitude of Sstx decreases with increasing the channel length L,
and (slightly) decreases when reducing the channel width W . We remind that the
magnitude of Sx depends on the amplitude of the edge wavefunction as well as on
the strength of scattering potential, as mentioned above. Due to finite size effect,
the edge localization increases and gradually reaches a saturation value as the width
 100
(a)FTI 0.1 (b)G − AFTI

Sx (10 −2 )

Sx (10 −2 )
0.2
0.1
0.0 0.0
0.1 0.10

®
0 10 20 30 39 10 20 30 39

­
SxV = 0­. 0 ® SxV = 1. 5 SxV = 0. 0­× 100
® SxV = 1. 5
−2 −2
Sx (10 ) Sx (10 )
0.5 0.0 0.5 0.2 0.0 0.2
3.0 3.0
2.0 2.0
V0 (A)

V0 (A)
1.0 1.0
0.00 10 20 30 39 0.00 10 20 30 39
L L

Figure 5.6: Evolution of Sx at top layer with disorder strength for (a)FTI and
(b)AFTI. The magnetic order and Fermi level is |m
~ i | = 0.2A and EF = 0.25A respec-
tively. The spin density is averaged over 1280 configurations and calculated with a
bias voltage 0.02A.

is increased. Therefore for a given disorder strength, the edge states of a wider
AFTI undergo smaller delocalization resulting an enhanced Suz and thereby larger
Sstx . Increasing the length L favors destructive interferences and therefore reduces Sstx
progressively.

5.0 m~ = 0. 2Am̂ y , V0 = 1. 5A
4.5 20
4.0
30
Sstx (10 −4 )

40
3.5
3.0
2.540 45 50 55 60 65 70 75 80
L

Figure 5.7: Staggered Sstx at V0 =1.5A for different length (L) and width (W , given in
legend).

It is quite instructive to analyze our results in the light of the latest developments
of spin torque studies on antiferromagnets [143, 146, 145]. As a matter of fact, it
is well known that an external uniform magnetic field only cants antiferromagnetic
moments and is unable to switch the direction of antiferromagnetic order parameter.
 101
Notwithstanding, time-dependent uniform magnetic fields (e.g. a magnetic pulse) can
induce inertial antiferromagnetic dynamics [146, 145]. In addition, it was recently
~ i × (~p × m
proposed that a spin torque possessing an antidamping symmetry [i.e. m ~ i ),
where p~ = ~z in our case] can manipulate the antiferromagnetic order parameter of
a collinear antiferromagnet [205, 150]. Applied to the AFTI studied in the present
work, these considerations imply that a current pulse can exert a torque on the
antiferromagnetic order parameter through the uniform spin density Suz , while a dc
current can exert a torque via the staggered spin density Sstx . Therefore, the staggered
Sstx computed in Figs. 5.5(b) and 5.6(b) can in principle be used to control the
antiferromagnetic order of an AFTI.
Note that in case of AFTI, we can choose EF very close to zero where the topo-
logical protection is stronger, without significantly affecting the magnitude of the
current-driven spin densities [see Fig. 5.8(a,b)]. By tuning the parameter M one can
weaken the topological protection and turn the AFTI into a trivial antiferromagnet
(TAF). In this regime, the non-equilibrium spin density is more distributed within
the bulk of the TAF and does not have any topological protection. As a result and
in spite of the strong spin-orbit coupling, Suz and Sstx remain both very small [see Fig.
5.8(c)].

(a) G − AFTI (EF = 0. 25A) (b) G − AFTI (EF = 0. 05A) (c) TAF (EF = 0. 25A)
0.8 2.0 0.8 2.0 0.8 2.0
G G G
0.6 Sux 1.5 0.6 Sux 1.5 0.6 Sux 1.5
Su , Sst (10 −2 )

Su , Sst (10 −2 )

Su , Sst (10 −2 )

0.4 Sstx 1.0 0.4 Sstx 1.0 0.4 Sstx 1.0

G/G0

G/G0

G/G0

Suz Suz Suz

0.2 Sstz 0.5 0.2 Sstz 0.5 0.2 Sstz 0.5
0.0 0.0 0.0 0.0 0.0 0.0
0 1 2 0 1 2 0 1 2
V0 (A) V0 (A) V0 (A)

Figure 5.8: The uniform and staggered order parameter for (a) AFTI at EF = 0.05A,
(b) AFTI at EF = 0.25A and (c) for TAF at EF = 0.25A. For TAF we use B =
1.0A, M = −2.2A. |m ~ i | = 0.2A for all three cases.
 102
5.4 Conclusion

In this work we present a detailed analysis of spin transport in two dimensional FTI
and AFTI. We show that topological transport in AFTI is more robust compared
to FTI in presence of both out of plane and in plane magnetic order. An in plane
magnetic order opens a gap in a FTI but preserves the gapless states in an AFTI
when the antiferromagnetic order is along the direction of transport, which allows an
AFTI to operate at a much smaller energy. We also study the robustness of the non-
equilibrium spin density and SOT against scalar disorder and find that the in-plane
spin densities get mixed up due to scattering. In the clean limit, this mixing is two
orders of magnitude smaller in AFTI compared to FTI, which suggests that AFTI
has stronger topological protection against scalar disorder. The SOT possesses two
components, a field-like torque arising from the spin-momentum locking at the edges
and an antidamping torque arising from scattering. This antidamping torque linearly
decreases when increasing the length of the sample due to destructive interferences.
 103

Chapter 6

Topological insulator-ferromagnet heterostructure

The conventional way to record information on magnetic elements exploits spin trans-
fer torque (STT), a mechanism that transfers spin angular momentum from a ref erence
to a f ree magnetic layer via a spin polarized current [209]. This effect has been used
successfully in the context of magnetic random access memories [210], offering both
scalability and fast switching rate. Yet, its efficiency remains limited by the polariza-
tion of the reference layer and reading/writing operations must be performed through
the same channel. These limitations are overcome in a spin-orbit torque (SOT) de-
vice, where the role of the polarizer is replaced by the spin-orbit coupling (SOC) of the
material [211, 84, 85, 107, 90]. Since its first observation in (Ga,Mn)As [93], SOT has
drawn significant attention for its ability to promote fast switching [94, 96, 212], very
high domain wall motion [213, 214, 215] and GHz excitations [216, 217]. Several se-
tups have been proposed to improve the device efficiency [100, 218, 99, 219, 220]. Most
SOT devices are made of a magnetic layer (FM) deposited on a substrate with strong
SOC. The SOC in the substrate generates a non-equilibrium spin density that exerts
a torque on the adjacent FM layer and thus can manipulate its order parameter. Two
types of torques are usually observed in such systems [91, 221]: (i) a field-like torque,
TF ∼ m × (z × je ), and (ii) an (anti)damping-like torque, TD ∼ m × [(z × je ) × m].
Here, m is the magnetization direction of the magnet, z is the normal to the interface
and je is the injected current. Two main effects are usually considered to be at the
origin of the SOT: the spin Hall effect (SHE) taking place in the bulk of the substrate
induces (mostly) a damping-like torque [104, 106, 105] while the inverse spin galvanic
 104
effect (ISGE - also called Rashba-Edelstein effect [222, 52]) present at the interface
produces (mostly) a field-like torque [84, 85, 211, 107, 108]. However, several exper-
imental [110, 112, 109, 111, 113] and theoretical [223] clues indicate that the debate
is not settled yet and additional mechanisms, such as intrinsic magnetoelectric effect
[224, 225, 161] and spin swapping [226, 227], have been suggested to take place in
ultrathin magnetic multilayers.
Due to their inherent strong SOC, topological insulators display large spin-charge
conversion efficiency [122, 123, 124] and are now viewed as suitable candidates for
spintronics applications [121]. A three-dimensional time-reversal symmetric topolog-
ical insulator (TI) is characterized by an insulating bulk and surface spin-momentum
locked single Dirac cones with well-defined chirality [61]. Recently TIs have been
found to be a powerful source of SOT [117, 120, 98, 121, 228] that can be further
controlled by gate fields[121]. Most importantly for memory and logic applications,
switching current as low as ∼ 105 A/cm2 at room temperature [127, 128, 141] has
been reported in Bi2 Se3 -based bilayers, two to three orders of magnitude smaller than
their heavy metal counterpart [95, 96]. A parameter conventionally used to evaluate
the SOT efficiency in experiments is the dimensionless “effective spin Hall angle”.
This angle (expressed in percent) quantifies the overall efficiency of the spin-charge
conversion processes taking place in the heterostructure. In FM-TI heterostructures
a gigantic effective spin Hall angle ranging from 160% [127] to 42,500% [98] has been
reported. To date, such enormous efficiencies remain largely unexplained.
The nature of SOT in FM-TI heterostructures has been investigated within the
linear response formalism using the simplistic two-dimensional Dirac Hamiltonian
[229, 131, 230, 133], possibly supplemented with spin diffusion in adjacent layers [117,
132]. Spin transport in FM-TI has also been investigated within time-dependent non-
equilibrium Green’s function method implemented on a two terminal device [231, 139],
where spin-flip scattering was shown to generate large damping-like torque. Despite
 105
these substantial efforts, the physical origin of the huge SOT efficiencies reported
experimentally remains unclear. What makes FM-TI heterostructures unique and
subtle is the major role played by the topological characteristics of the bulk bands
[63, 72, 232]. Although transport through the bulk states has been neglected in the
above theories, experimental evidence shows that realistic TIs are conductive and
suggests that bulk states substantially contribute to SOT. Another subtlety comes
from the very nature of the orbital hybridization between FM and TI layers. Density
functional theory reveals that not only the spin texture gets modified in presence of
hybridization with a magnetic overlayer [233], but the TI bands are also pushed below
Fermi level [137, 234], which can significantly modify the features of SOT. Till now,
there is no clear explanation of how these modifications of band structure and spin
texture affect the spin density and SOT. A proper description of the SOT efficiency in
surface-dominated and bulk-dominated transport regimes is therefore highly solicited.
In this chapter we build a tight-binding model for the TI and magnetic mate-
rial heterostructure that describes bulk and surface spin transport on equal footing.
The chapter is divided into three sections. First section describes the tight binding
model and non-equilibrium Green’s function formalism. The second section deals
with the TI-FM heterostructure and in the third section describes the TI-AFM het-
erostructure. Since we we are working with a model with finite thickness, interface-
driven ISGE and bulk-driven SHE are computed simultaneously. By investigating
the layer-resolved conductivity, non-equilibrium spin density and SOT over a wide
range of transport energies, we uncover a crossover between interface-dominated and
bulk-dominated regimes, associated with a substantial variation of the field-like and
damping-like torque components. We show that the SOT is maximal when surface
transport dominates, while the SHE arising from the bulk of the TI has a very small
contribution to SOT, even in the bulk-dominated regime. We observe that the largest
contribution to the damping-like torque is attributable to the intrinsic magnetoelec-
 106
tric effect[229, 224], not to the SHE.

6.1 Model and Method

6.1.1 Tight Binding Model

A TI surface state is characterized by a Dirac cone with strong spin-momentum

locking and a well-defined chirality. When a FM material is brought to the vicinity
of a TI surface, three main effects take place [137]: (i) The TI acquires an induced
magnetization by proximity effect which opens a gap, (ii) the FM receives an induced
(Rashba-like) SOC resulting in a modification of the spin texture in both layers, and
(iii) the TI Dirac cone is pushed down in energy due to the increased carrier density
at the surface. The tight-binding model described in the present section aims at
accounting for these three effects through a simple hybridization scheme.
We define our TI motif with a 4 × 4 effective Hamiltonian regularized on a cubic
lattice [138]. The onsite Hamiltonian (H0 ) for each layer reads

 
ĥk + ûk m̂k 
Ĥ0 =  , (6.1)
m̂k −ĥk + ûk
ĥk = A[σ̂y sin(kx a0 ) − σˆx sin(ky a0 )], (6.2)

ûk = [c − d(cos(kx a0 ) + cos(ky a0 ))] Î2 , (6.3)

m̂k = [M − B(cos(kx a0 ) + cos(ky a0 ))] Î2 , (6.4)

whereˆdenotes an operator, σ̂i is the i-th Pauli spin matrix and Î2 is the 2×2 identity
matrix. The individual layers are connected by the matrix ĤT
 
0 Î2 (A1 − B1 )/2
ĤT =  . (6.5)

Î2 (−A1 − B1 )/2 0
 107
Here A, B, M , c, d, A1 and B1 are the tight binding parameters of the model and a0
is the lattice constant. The basic idea behind this model is to create a pair of chiral
states at each layer and then connect each Dirac node to a state with opposite chirality
in the adjacent layer in such a way that each surface only has a single Dirac node (see
Fig. 6.1). Such a scheme can be adopted to model chiral TIs and superconductors
[235]. Depending on the mode of connection one can create a positive or negative
spin Hall current in the bulk, which in turns determines the surface spin texture and
spin density [232].

Figure 6.1: Schematic of the interlayer coupling procedure and origin of the bulk spin
Hall current and surface texture.

The real hybridization scheme between the FM (d and p) orbitals and the TI (p)
orbitals is quite complex [137, 234]. In the present work, we limit ourselves to a
minimal, spin-independent coupling that is sufficient to promote the effects we are
interested in, i.e., induced Rashba spin splitting in FM and induced magnetization in
TI. The magnetic layers are defined with a 2 × 2 Hamiltonian ĤM and are coupled
together by the 2 × 2 Hamiltonian T̂M . The coupling between the topmost TI and
the bottommost FM layers is governed by a 2 × 4 connection matrix T̂T M .
 108

ĤM = (0 − t0 (cos kx a0 + cos ky a0 )) Î2 + ∆σ̂z , (6.6)

 
1 0 1 0
T̂M = t0 Î2 , T̂T M = tTM  , (6.7)
0 1 0 1

where 0 , t0 and ∆ are the onsite, hopping and Zeeman energy of the FM and tTM is
the coupling between TI and FM layers. For simplicity we define the FM layer with
a cubic lattice with lattice parameter a0 . The complete Hamiltonian for the FM-TI
heterostructure thus looks like,

...
 
T̂M 0 0 0
 
 † 
T̂M ĤM T̂T M 0 0 
 
 
Ĥ(kx , ky ) =  † . (6.8)
 0 T̂TM Ĥ 0 ĤT 0 
 
†
 0 0 ĤT Ĥ0 ĤT 
 
 
† ..
0 0 0 ĤT .

A system with n1 FM layers and n2 TI layers is therefore defined by a (2n1 +4n2 )×

(2n1 + 4n2 ) matrix. For our study we choose 20 layers of TI and 5 layers of FM. We
choose the parameter A as our unit of energy and set the other parameters for the
TI with respect to A as displayed in Table 6.1. We keep the Zeeman splitting (∆)
and coupling strength (tTM ) as free parameters for now. Unless mentioned otherwise
we choose 0 , the onsite energy of the magnetic layers, such a way that the decoupled
magnetic bands starts from energy 0.3A so that we can identify the contributions
coming from the FM layers and FM-TI coupling easily. The actual values of these
parameters can be determined from a DFT calculation for the bulk material. For
example in case of Bi2 Se3 the bulk band gap is ∼ 0.4 eV which can be obtained with
A=0.27 eV.
 109
A B M c d A1 B1 t0 tM
1 1.5 3.5 1.5 0.75 1.5 1.5 0.5 0.5

Table 6.1: Tight-binding parameters used in the main text.

The band structure of a system with 20 TI layers and 5 FM layers is shown in

Fig. 6.2 for various coupling parameters. At tTM = 0, one distinguishes the surface
Dirac cones (overlapping blue and green), the TI bulk states (black) and the uncoupled
magnetic states (red). Upon turning on the interlayer coupling, tTM , the Dirac cone
of the top surface (blue) is pushed downward and acquires a gap due to proximity
effect. The magnetic states progressively hybridize with the surface states away from
the Γ-point and acquire a chiral spin texture. Notice that the Dirac cone at the
bottom surface remains unaffected.
(a) (b) (c) (d) (e) (f)
1.5

1.0

0.5
E/A

0.0

-0.5

-1.0
X← Γ →M X← Γ →M X← Γ →M X← Γ →M X← Γ →M X← Γ →M

Figure 6.2: Band structure for 5F M +20T I with ∆ = 0.25A and for different coupling
parameter : (a) tTM =0, (b) 0.1A, (c) 0.2A, (d) 0.3A, (e) 0.4A, (f) 0.5A. The red,
blue and green colors correspond to contributions from FM layers, top TI layer and
bottom TI layer. The black lines correspond to the bulk TI bands.

The modification of spin texture due to coupling in a typical FM-TI is demon-

strated schematically in Fig. 6.3, which is instrumental to understand the modified
spin texture in a multilayer system in the next section. In order to provide a compre-
hensive picture of the complex spin-momentum locking taking place in this structure,
we choose three different momenta denoted by vertical dashed lines in Fig. 6.3(a) and
the corresponding spin texture is reported on Figs. 6.3(b, c, d), respectively. In a
nutshell, one can notice that bands with a dominant TI character (blue) display a
Dirac spin-momentum locking, S ∼ z × k, while bands with a dominant FM char-
 110
acter (red) display a spin angular momentum S ∼ z. In general, the spin texture
lies in-between these two cases. Notice that the induced chirality of the FM bands
changes sign with the magnetization of the bands (denoted by red arrows).

Figure 6.3: (a) Schematic of the band structure of a typical FM-TI heterostructure.
The blue and red colors correspond to contributions from TI and FM layers, re-
spectively. (b, c, d) Corresponding non-collinear spin texture computed at different
positions in the band structure, as denoted by the vertical lines in (a).

6.1.2 Non-equilibrium transport formalism

The SOT is calculated within the linear response framework [223, 161, 236]. We start
by defining the retarded (advanced) Green’s function at energy E,

R(A)
ĜE,k = [(E ± iη)În − Ĥ(k)]−1 (6.9)

where η is the impurity broadening and In is an n × n identity matrix, n being the

dimension of Ĥ. In this work, disorder is modeled by a constant broadening η which is
found to be a fair approximation for calculating SOT in Co-Pt heterostructure [223].
In order to keep the computation time reasonable, vertex corrections are not taken
into account in this calculation. Such an approximation is not expected to qualita-
tively affect our results (see e.g. Ref. [237]). To compute the SOT, we first calculate
the non-equilibrium spin density caused by an applied electric field E. Without any
 111
loss of generality we assume the the electric field to be along x̂ direction and incorpo-
rated as an interaction term evx E. The non-equilibrium spin density per unit electric
field is described by velocity-spin correlation functions [238, 161] and possesses two
contributions

S = Ssea + Ssur (6.10)

where Ssea and Ssur correspond contribution coming from Fermi sea and Fermi surface
and are defined by,

e~ d2 k
Z
Ssur = Re(Tr[ŝĜR k A R
E,k v̂x (ĜE,k − ĜE,k )])EF
2π (2π)2
(6.11)
EF 2
e~ dk
Z Z 
Ssea = dE Re Tr[ŝĜR k R
E,k v̂x ∂E ĜE,k
2π −∞ (2π)2

R k R
−ŝ∂E ĜE,k v̂x ĜE,k ] (6.12)
 
..
 . 
  k
ŝn =  σ̂n  , v̂x = ∂~kx Ĥ(k). (6.13)
 
..
 
.

where σ̂n is the spin operator for the nth layer (σ̂ for FM layers and σ̂ ⊗ Î2 for TI
layers). The integration over k goes over the first Brillouin zone [±π/a0 , ±π/a0 ].
Re takes the real part and Tr is the trace on both spin and orbital spaces. In the
case of our slab geometry, we find that the Fermi sea contribution, Ssea , is negligible
compared to the Fermi surface contribution Ssur . The spin-orbit torque is defined
T = (2∆/~)z × S, and therefore Sx and Sy produce the damping, TD , and field-like
torque, TF , respectively.
One can similarly calculate the longitudinal charge conductance using the velocity-
 112
velocity correlation function,

σxx = σsea + σsur (6.14)

e2 ~ d2 k
Z
σsur = Re(Tr[v̂xk ĜR k A R
E,k v̂x (ĜE,k − ĜE,k )])EF
2π (2π)2
(6.15)
EF
e2 ~ d2 k
Z Z 
σsea = dE Re Tr[v̂xk ĜR k R
E,k v̂x ∂E ĜE,k
2π −∞ (2π)2

k R k R
−v̂x ∂E ĜE,k v̂x ĜE,k ] (6.16)

Finally, it is also useful to determine the spin Hall current flowing in the bulk of
the TI as SHE is considered as a source for large damping-like SOT. The bulk TI
Hamiltonian ĤB and spin current operator ĵz are given by

ĤB (k) = Ĥ0 + ĤT e−ikz a0 + ĤT† e+ikz a0 , (6.17)

ĵz = (~/4){v̂zB , ŝB }, (6.18)

where v̂zB = ∂~kz ĤB , and ŝB = σ̂ ⊗ Î2 . The bulk spin Hall conductivity, σzi , for a
spin current polarized along i and flowing along z can be calculated by replacing
ŝ with ĵz in Eqs. (6.11)-(6.12) and performing a three-dimensional integration over
the Brillouin zone. Unlike the slab geometry we find that the Fermi sea term, Eq.
(6.12), does not vanish for the bulk σzy and rather provides the main contribution to
the quantized Hall conductance within the bulk gap region. σzx and σzz , on the other
hand, vanish for the whole energy range.
 113
6.2 Spin Hall current, non-equilibrium spin density and SOT
in topological insulator - ferromagnet heterstructure

As briefly mentioned in the introduction, in magnetic multilayers SOT can be caused

by two mechanisms, (i) SHE arising from the bulk of the heavy metal and (ii) ISGE
induced by the interfacial (Rashba, Dirac) SOC, as illustrated in Fig. 6.4. These
two mechanisms are a priori distinct from each other. SHE generates a spin current
in the bulk of the heavy metal that is injected into the adjacent FM layer; the re-
sulting non-equilibrium spin density penetrates inside the FM layer, precesses about
the magnetization vector and generates a spatially oscillating spin density. If spin
dephasing is strong enough (in the case of a strong FM for instance), Sx component
(i.e., the damping-like torque TD ) dominates [106].

Figure 6.4: Schematic of the origin of different non-equilibrium spin components

in an FM-TI heterostructure. In the bulk of the TI away from the interface, SHE
spatially separates the flowing electrons with spin oriented along ±y (large blue and
red arrows). Electrons polarized along +y penetrates into the FM layer, generating
an effective spin density oriented along +x (small green arrows). In addition, ISGE at
the interface directly generates a spin density oriented along +y (small blue arrows).
Both spin densities exert a torque on the magnetization of the FM layer (yellow
arrows).

In contrast, ISGE generates a spin density at the interface between the heavy
metal and the ferromagnet. While interfacial SOC alone generates an Sy component
(i.e., TF ) through the extrinsic Rashba-Edelstein effect[52], the coexistence of SOC
and magnetic exchange generates an Sx component (i.e., TD ) through the intrinsic
magnetoelectric effect[161, 229, 133]. This effect is related to the Berry curvature in
 114
mixed spin-momentum space [224, 225]. In TIs, the surface properties are caused by
their bulk topology [239], so one can expect a tight connection between these SHE
and ISGE. We now aim at understanding the interplay between SOT arising from
transport in the bulk TI and SOT arising from interfacial SOC.

6.2.1 Spin Hall conductance in the bulk

(a) (b)
1.5 1.5
σxz
1.0 1.0
TItop
0.5 σyz 0.5
E/A

0.0 0.0
TIbot
-0.5 -0.5
σzz
-1.0 -1.0
X Γ M 0 1 2 3 4 5
σsz (ℏ/2e)104 Ω-1 .m -1

Figure 6.5: (a) Band structure of bulk TI (thick green lines) and a 20-layer TI slab
(black lines). The corresponding spin texture of the top and bottom layers of the
slab at E = 0.25A is represented on the insert of (b). (b) Spin Hall conductances as
a function of the energy computed in the bulk of the same TI. The spin conductance
is constant as long as the energy lies in the bulk gap and decreases when bulk states
become conductive.

To clarify these aspects, we first calculate the bulk spin Hall conductivity and the
surface spin texture of a TI slab with 20 layers in the absence of magnetic overlayer.
The band structures of the bulk TI (thick green) together with the one of the equiva-
lent 20-layer slab (black) are displayed on Fig. 6.5(a); the spin texture at E = 0.25A
for both top and bottom surfaces is reported as an inset on Fig. 6.5(b). The cor-
responding bulk spin Hall conductance is displayed in the main panel of Fig. 6.5(b).
For −0.5 < E/A < 1, the TI is bulk insulating and only conducts through its surface
states. Hence, the spin Hall conductance is constant, σzy ≈ 5 × 104 (~/2e) Ω−1 ·m−1 ,
i.e., an electric field applied along x creates a spin current polarized along y and
propagating along z. For E/A > 1, the TI is conductive, which results in a pro-
 115
gressive decrease of the spin conductance σzy . For the sake of comparison, the spin
Hall conductivity of 5d transition metals such as Pt, Ta or W has been calculated to
be in the range of 103 to 104 (~/2e) Ω−1 ·m−1 (Ref. [240]), while a maximum of 105
(~/2e) Ω−1 ·m−1 was computed for Bi0.83 Sb0.17 [241]. The value of the intrinsic spin
Hall conductivity reported in Fig. 6.5 is therefore quite large and should generate
large damping-like SOT.

6.2.2 Surface versus Bulk Transport

(a) (b) (c) (d)

1.5

1.0
E/A

0.5

0.0
X← Γ →M

-0.37 0. 0.37 -7. 0. 7. 0 12.

Figure 6.6: Layer-resolved non-equilibrium Sx , Sy and longitudinal charge conduc-

tivity σxx across the FM-TI heterostructure, with tTM = 0.5A and ∆ = 0.25A. The
FM-TI interface is marked by a vertical dashed line. The color scale gives Sx,y in
units of 1010 V−1 ·m−1 and σxx in 10−4 Ω−1 .

Figure 6.6 shows the spatial profile of Sx , Sy and σxx for a system with 5 FM and
20 TI layers, while tuning the transport energy through the band structure. Within
the bulk gap region (−0.5 < E/A < 1), Sy and σxx are mostly coming from TI
surfaces. The magnetic bands are crossing the surface states at E ∼ 0.3A, which
is denoted by a progressive rise in Sx , Sy and σxx within the magnetic layer and a
drop in Sy in the top TI surface (close to FM layer). Interestingly, the rise of Sx and
collapse of Sy at the top TI surface are correlated with the onset of bulk conduction,
when E > A.
 116

(a) (b) (c)

1.2 0.6 12
M5 0.6 0.3 8
0.0 0.0
-0.6 -0.3 4
-1.2 -0.6 0
1.2 0.6 12
0.6 0.3 8
M4

0.0 0.0
-0.6 -0.3 4
-1.2 -0.6 0
1.2 0.6 12
0.6 0.3 8
M3

0.0 0.0
-0.6 -0.3 4
-1.2 -0.6 0
1.2 0.6 12
0.6 0.3 8
M2

0.0 0.0
-0.6 -0.3 4
-1.2 -0.6 0
1.2 0.6 12
0.6 0.3 8
M1

0.0 0.0
-0.6 -0.3 4
-1.2 -0.6 0
1.2 1.6 12
0.6 1.2 8
TItop

0.0 0.8
-0.6 0.4 4
-1.2 0.0 0
1.2 0.0 12
0.6 -0.4 8
TIbot

0.0 -0.8
-0.6 -1.2 4
-1.2 -1.6 0
0.0 0.4 0.8 1.2 1.6 0.0 0.4 0.8 1.2 1.6 0.0 0.4 0.8 1.2 1.6
E/A E/A E/A
tTM : 0. 0.1 A 0.2 A 0.3 A 0.4 A 0.5 A
1 2 3 4 5 6 25

M5 M4 M3 M2 M1 TItop Bulk TI TIbot

Figure 6.7: (a) Sx , (b) Sy and (c) σxx for a 5FM+20TI heterostructure for different
coupling constant (tTM ) with exchange coupling ∆ = 0.25A and broadening η =
0.01A. The vertical dashed lines show the beginning of decoupled magnetic bands
(E = 0.3A) and bulk TI bands (E = 1.0A). Sx is given in units of 109 V−1 ·m−1 , Sy
in 1010 V−1 ·m−1 and σxx in 10−4 Ω−1 . The bottom panel shows the color code for
different layers.
 117
To better understand this behavior, we investigate in more details the layer-
resolved spin density and conductivity as a function of the energy in selected layers.
Figure 6.7 displays the transport energy dependence of Sx , Sy and σxx , at the top and
bottom TI surfaces, TItop and TIbot , as well as in the different FM layers, M1, M2,
M3, M4 and M5, for different coupling strengths (tTM ). Based on the band structure
shown in Fig. 6.6(a) and on the conductivity map of Fig. 6.6(d), we define three
transport regimes. When 0 < E/A < 0.3, the transport solely occurs through the
surface states of the TI and the FM behaves like a magnetic insulator. This situation
is comparable to the magnetic Dirac gas studied in Refs. [229, 131, 133]. Then, for
0.3 < E/A < 1, the FM layer becomes progressively conductive while the TI remains
bulk insulating. Finally, when E/A > 1 conduction occurs throughout the entire het-
erostructure, and the higher the transport energy the more bulk transport dominates
over surface transport.
Let us first consider the uncoupled situation, i.e., the TI and FM layers are discon-
nected (tTM = 0, blue line in Fig. 6.7). When varying the transport energy across the
band structure, Sy progressively increases at both TItop and TIbot , while Sx remains
exactly zero. Sy reaches a maximum close to the bulk conduction edge (E/A ≈ 1)
and decreases monotonously for E/A > 1, consistently with the behavior of the bulk
spin Hall conductivity displayed on Fig. 6.5. Obviously, the FM layers do not show
any non-equilibrium spin density because tTM = 0.
When the FM-TI coupling is turned on, the FM bottom band acquires a Rashba-
like SOC, while a gap opens at the Dirac cone at TItop . To apprehend the physics at
stake, the spin texture in three different energy regimes has been plotted on Fig. 6.8,
for tTM = 0.5. As mentioned above, when E/A < 0.3 the transport is dominated by
the Dirac states of the TI surfaces, while the FM is insulating. Turning on the FM-TI
coupling pushes the Dirac cone downwards [see Fig. 6.8(a1)] and thereby enhances
the density of states at Fermi level resulting in an increase in both Sy and σxx at TItop
 118
(Fig. 6.7). The first two magnetic layers, M1 and M2, become weakly conductive by
proximity effect and acquire a small spin texture aligned on the one of TItop [see
Fig. 6.8(b1,c1,d1)]. As a result, Sy penetrates into the FM layers by proximity and
the larger the FM-TI coupling, the stronger the induced spin density. From the gap
of TItop bands, we can estimate that for tTM = 0.5 the induced magnetic exchange
is roughly 25% of the FM exchange coupling (∆). Consequently, an Sx component
progressively appears in both TItop and M1 via the magnetoelectric effect[224, 229],
due to the coexistence of magnetism and SOC. Yet, this component remains extremely
small.
In the intermediate regime, 0.3 < E/A < 1, the TI layer is still in a topologically
non-trivial state (bulk insulating, conductive chiral surface states), while the FM
layers become more conductive and acquire a complex spin texture whose chirality
depends on the magnetization, as displayed in Fig. 6.8(b2,c2). On the other hand,
the strength of the spin-momentum locking at TItop decreases upon increasing the
transport energy [Fig. 6.8(d2)]. Consequently, the competition between the different
spin chiralities produces an oscillating behavior of Sy as a function of the energy: the
dips correspond to FM-TI band crossing. Notice that increasing the coupling results
in a decrease of Sy at TItop and an increase in the FM layers (Fig. 6.7). Indeed, upon
increasing the coupling, the FM layers acquire more SOC and because the FM is
now conducting, Sy penetrates deeply into the FM layers producing a Sx component
upon spin precession. Sx also increases in the FM layers upon increasing the FM-TI
coupling.
Finally, when E/A > 1 the transport is progressively dominated by the bulk states
of the TI. The central portion of the texture is coming from bulk TI and depending
on whether states are dominated by the top TI layer or the hybridization of the FM
layer, it can have either positive or negative chirality. As shown in Fig. 6.8(d3),
the TI bands possess weaker spin-momentum locking, causing a fall in Sy in TItop
 119
and a deeper penetration inside the FM layers. This penetration is associated with
spin precession in the FM, and therefore a significant increase in Sx , as displayed in
Fig. 6.7. Because the wavefunctions are now delocalized, they expand throughout the
structure, from M5 to TIbot [Figs. 6.6(b,c)].

Figure 6.8: (a1, a2, a3) Band structure and Spin texture of (b1, b2, b3) M1 and (c1,
c2, c3) M2, and (d1, d2, d3) TItop at (1) E = 0.25A, (2) E = 0.75A and (3) E = 1.25A
for tT M = 0.5A, ∆ = 0.25A and η = 0.01A. The texture colors correspond the Sz
component of the spin density. The band colors have the same meaning as Fig. 6.2.

In summary, while the Sy component dominates in the surface-dominated regime,

in agreement with all previous theories on TI [229, 131, 133], a crossover appears upon
increasing the transport energy and in the bulk-dominated regime, the Sx component
is significantly enhanced inside the FM layer, displaying a large oscillation across the
thickness. This is consistent with the onset of the SHE in the bulk TI. It is worth
noticing that the magnitude of the Sx component is about the same as Sy , so at this
stage it is unclear whether these results can explain the experimental observations.
 120
One possible reason could be the fact that we are modeling each FM layer with
only a pair of parabolic bands, whereas in practice a real FM material has many
more bands crossing Fermi level. Therefore our results are expected to be up to an
order of magnitude smaller than reported experimentally. Nevertheless, the physical
arguments are still valid for realistic heterostructures. To this end, we now address
the impact of disorder on the magnitude of the SOT components, in order to identify
their physical origin.

6.2.3 Intrinsic versus Extrinsic Spin-Orbit Torque

To better understand the spin-charge conversion mechanisms at stake in the FM-TI

heterostructure, we investigate the dependence of the spin density as a function of
the impurity broadening η. As a matter of fact, it was shown recently that in the
context of the Rashba two-dimensional electron gas or in the bulk of (Ga,Mn)As,
both interband and intraband processes participate to the SOT and give rise to two
classes of contributions [224, 161, 133]: (i) extrinsic contributions that depend on the
amount of disorder in the system and (ii) intrinsic contributions that are independent
on the amount of disorder. ISGE is an extrinsic mechanism and is expected to exhibit
a 1/η dependence, like the conductivity, while the magnetoelectric effect and SHE are
both intrinsic contributions and should not vary as a function of η.
FM
P
Figure 6.9 displays the dependence of the total spin density Sx,y = FM Sx,y ,
TI
P
summed over the FM layers, as well as the total conductivity of the TI, σxx = TI σxx .
Both SyFM and σxx
TI
display the 1/η-dependence expected for extrinsic mechanisms. It
means that Sy comes from the interfacial (Rashba- or Dirac-driven) ISGE, namely
from the spin-momentum locked interfacial states of the heterostructure. The change
of sign of SyFM for E/A = 1.2 can be ascribed to the competition between Rashba-
and Dirac-driven ISGE, as they have opposite sign. The behavior of SxFM is richer
and depends on the transport regime. As long as the transport is purely interfacial,
 121
(a) (b)

SyFM 109 V -1 .m -1
1.0 20

-4 -1
xx 10 Ω
0.5 15
0.0 10
-0.5 5

σTI
-1.0 0
0 2 4 6 8 10 0 2 4 6 8 10
SxFM 109 V -1 .m -1 (c)
0.0
E/A
-0.3 0.0 0.2 0.4
-0.6 0.6 0.8 1.0
1.2 1.4 1.6
-0.9
0 2 4 6 8 10
η/A 10-2

Figure 6.9: non-equilibrium (a) SyF M , (c) SxF M in FM and TI conductance (b) σxx
TI
with
tTM = 0.5A and ∆ = 0.25A for different disorder broadening η. The blue lines corre-
spond to the surface-dominated/magnetic insulator regime, the red lines correspond
to the surface-dominated/magnetic metal regime, and the green lines correspond to
the bulk-dominated/magnetic metal regime.

SxFM ∼ 1+O(η), i.e., this component is independent on the disorder. Since the system
is in the quantum SHE regime, the physical origin of Sx is attributed to the interfacial
magnetoelectric effect. In the intermediate and bulk regimes, when the FM and TI
layers are conductive, SxFM displays a small dependence as a function of the disorder,
which can be attributed to the diffusion of the spin density inside the metallic FM.
Yet, SxFM converges to a constant value when η → 0 indicating its intrinsic origin.

6.2.4 Understanding the torque efficiency

Let us now complete this study by quantitatively comparing the magnitude of the
SOT computed with our model to the one observed in experiments. Assuming A ∼
0.27 eV and a0 = 4.2 Å gives a group velocity (a0 A/~) ∼ 1.7 × 105 m/s near the Dirac
cone, and a bulk band gap 1.5A = 0.4 eV. Near E ∼ 0.7A, uncoupled (tTM = 0)
TI surfaces display a non-equilibrium two dimensional spin density per unit field
Sy ∼ 1.5 × 1010 V−1 ·m−1 (Fig. 6.7). Assuming the thickness of a single TI layer to
 122
be 1 nm and an applied electric field E = 2 × 104 V/m [117], we obtain a total spin
density 0.3 × 1024 m−3 which is in good agreement with the spin density calculated
from density functional theory [242]. This estimation ensures that the TI parameters
chosen in the present study are realistic. The non-equilibrium spin density on the FM
layers on the other hand depends on the coupling strength (tTM ) as well as on the
magnetic exchange. In the strong coupling limit (tTM = 0.5A), within the previous
parameter settings, the first FM layer (M1) can acquire a spin density Sy that amounts
up to 20% of the value of the decoupled TI surface.
In experiments, the SOT is measured as a magnetic field HSOT (in Oe or T)
per unit of current density flowing in the heavy metal je (in A/cm2 ). As briefly
explained in the introduction, it is now conventional to quantify the torque in terms
the spin conductivity σs = HSOT Ms d/E [in (~/2e) Ω−1 ·m−1 ] and the dimensionless
SOT efficiency θH = (2e/~)σs /σxx , where Ms is the saturation magnetization, d is
the FM thickness, and σxx is the conductivity of the adjacent heavy metal. It is
understood that θH is equal to the spin Hall angle if and only if SHE is the only
spin-charge conversion mechanism present in the system - which is clearly not the
case of transition metal bilayers or TI-FM heterostructures, as demonstrated in the
present work.
k k
Overlayer Ref. σxx σs σs⊥ θH ⊥
θH
Bi2 Se3 (8)/NiFe(8) [117] 5.7 16 20 3.5 2.8
Bi2 Se3 (7.4)/CoTb(4.6) [127] 9.4 1.5 0.16
Bi2 Se3 (5)/CoFeB(7) [128] 2.4 3.89 1.6
Bi2 Se3 (4)/CoFeB(5) [141] 0.78 15 18.83
Dirac gas [133] 0.19 1.48 2.35 7.66 12.1
Table 6.2: Room temperature bulk conductivity, spin conductivity and SOT effi-
ciency measured in various Bi2 Se3 (t)/FM(d) systems, where t and d are the thick-
nesses in nm. The bulk conductivity is in 104 Ω−1 ·m−1 and the spin conductivity is
in (~/2e) 104 Ω−1 ·m−1 . The numbers in the last column are computed using Eqs.
(6.19) and (6.20).

We selected four experimental works characterizing SOT on Bi2 Se3 -based het-
 123
erostructures, and whose results are reported on Table 6.2. Overall, the spin con-
ductivity ranges from 104 to 105 Ω−1 ·m−1 , while the SOT efficiency spans over two
orders of magnitude depending on the estimated bulk TI conductivity. These data
need to be taken with sane care, considering the difficulty in accurately estimating
the various materials parameters (due to large interfacial roughness, magnetic dead
layers [120], inhomogeneous TI conductivity etc.). In our tight-binding model, the
torque density is T = (2∆/~)z × S, and the corresponding spin conductivity is de-
k,⊥ P P
fined σs = ±(2e/~)∆ FM Sx,y . Here FM denotes the summation over the FM
layers and the superscript k, ⊥ denotes the damping-like and field-like SOT compo-
P
nents. The effective bulk conductivity of the TI layer is then σbulk = TI σxx /W ,
where W is the thickness of the TI and the summation is only performed on the
TI layers. Figure 6.10 displays these four observables as a function of the transport
energy for different magnetic exchange energies. The impurity broadening is taken as
η = 0.1A (≈ 27 meV), such that the TI conductivity corresponds to the one observed
experimentally, close to the bulk conduction regime [see insert of Fig. 6.10(d)].
We observe that the spin conductivity monotonously increases with the trans-
port energy, reaching a maximum around E/A ≈ 0.6 − 0.8, which corresponds to
the surface-dominated transport. Beyond this point, the spin conductivity decreases
[Figs. 6.10(a) and (b)]. In other words, the maximum SOT magnitude is attained
before reaching the bulk transport regime despite the large bulk spin conductivity of
our system [see Fig. 6.5]. These results indicate that SHE is not the main mechanism
for SOT in TIs (as the largest torque magnitude is obtained in the surface-dominated
regime), but rather the interfacial ISGE. As a matter of fact in the bulk-dominated
transport regime, E/A > 1, the SOT efficiency decreases dramatically due to the
increasingly large conductivity of the bulk TI states. Within our set of parameters,
k
the computed σs⊥ and σs are comparable to the ones observed in Refs. [128] and
[127], and one order of magnitude smaller than Refs. [117] and [141], see Table 6.2
 124
(a) (b)

σ⊥s (ℏ/2e)104 Ω-1 .m-1

σ∥s (ℏ/2e)104 Ω-1 .m-1
0.0 3.0
-0.5 2.0
-1.0 1.0
0.0
-1.5
-1.0
-2.0
-2.0
0.0 0.4 0.8 1.2 1.6 0.0 0.4 0.8 1.2 1.6
(c) (d)
0 60 1.0
50
σ TI
-8 40 0.5
xx
104 Ω-1 .m -1
30
θH

θH
⊥
∥

20 0.0
-16 0.3 1.0 1.6
10
0
-24 -10
0.0 0.4 0.8 1.2 1.6 0.0 0.4 0.8 1.2 1.6
E/A E/A
Δ: 0.25 A 0.50 A 0.75 A 1.00 A

Figure 6.10: Non-equilibrium spin density summed over the magnetic layers and the
conductance summed over TI layers of the full system for 5 FM + 20 TI layers with
tTM = 0.5 and η = 0.1A.

(probably due to the details of the band structure and scattering processes). Inter-
estingly, the damping-like SOT efficiency reaches up to ∼ 24 close to the onset of
the bulk-dominated transport, a magnitude comparable to the one reported in Ref.
[141]. Notice though that this quantity is very sensitive to the overall bulk conductiv-
ity σbulk , which can be easily tuned by changing the disorder broadening η. While we
do not intend to quantitatively fit the experimental data, these calculations compare
favorably with those reported in Table 6.2 and clearly suggest that SHE from bulk
states are inefficient to generate large SOTs in TI-FM heterostructures.
To complete this discussion, we also report the results obtained within the Dirac
model that assumes a induced magnetic exchange at the surface. This model gives[133]

σs⊥ = −eτ ∆ind εF /(~2 vF π), (6.19)

σsk = 2e∆2ind /(~εF vF π), (6.20)

 125
where ∆ind is the induced magnetic exchange (about 25 % of the magnetic exchange
for tTM = 0.5A), εF is the Fermi energy, τ is the scattering time and vF is the
2D
Fermi velocity. The surface conductivity of the Dirac gas reads σxx = (e2 /h)(τ εF /~),
2D
giving an effective three-dimensional conductivity σxx = σxx /W (where W ≈ 20 nm).
For this estimation, we took the same parameters as for our FM-TI heterostructure,
vF = 1.7×105 m/s, τ = ~/2η ≈ 10−14 s, εF = 0.2A = 54 meV, and ∆ind = 17 meV. We
obtain a spin conductivity that is comparable to ones observed experimentally, i.e., in
the range (~/2e) 105 Ω−1 ·m−1 and a small longitudinal conductivity, ∼ 104 Ω−1 ·m−1 ,
comparable to Ref. [141]. In this experiment, Bi2 Se3 is not grown epitaxially but
by sputtering, which results in a reduced bulk conductivity and enhanced surface
transport. All these considerations rule out SHE as a source of giant SOT in TI-FM
heterostructures.

6.3 Conclusion

In summary, we have developed a tight-binding model for FM-TI heterostructure that

accounts for surface and bulk transport on equal footing. In addition, our minimal
hybridization scheme reproduces properly the induced magnetic exchange, induced
Rashba spin-orbit coupling and energy shift of the band structure observed using ab
initio calculations[137]. Our study unveils the momentum-dependent spin texture
across the band structure when varying the hybridization strength. This approach
enables us to compute the SOT emerging from the coexistence of spin Hall effect,
inverse spin galvanic effect, magnetoelectric effect, as well as spin precession inside
the ferromagnet. Our analysis shows that SOT increases steadily when increasing the
transport energy and reaches a maximum before the bulk states start contributing
to the transport. This result indicates that the spin Hall effect from bulk states is
unlikely to explain the large damping-like SOTs observed experimentally. In contrast,
large damping-like torque is achieved through the interfacial magnetoelectric effect
 126
and is therefore very sensitive to the nature of interfacial orbital hybridization.
 127

Chapter 7

Topological insulator-antiferromagnetic heterostructure

Spin-orbit torque (SOT) has recently become a viable candidate mechanism for the de-
velopment of magnetic memory devices [210, 243, 244]. A typical SOT device mainly
consists of two elements, a source of strong spin-orbit coupling (SOC) and a magnetic
material [95, 96]. When a charge current is passed through the material with strong
SOC it produces a non-equilibrium spin density which is utilized to manipulate the
magnetic order [156]. In recent years, the field of SOT devices has been revolution-
ized by two major breakthroughs: one is the discovery of topological insulator (TI)
[60, 61] and the other is the introduction of antiferromagnet (AF) [149, 143, 146, 145].
The strong interfacial SOC in TIs provides a high charge-spin conversion efficiency
and thus can be used as an efficient source of SOTs [117, 98, 120, 121]. Recent
experiments show that TI-based SOT devices can be operated with current of the
order of 105 A/cm2 even in room temperature [127, 128, 141], which is two orders
of magnitude smaller than the switching current density required to operate heavy
metal-based SOT devices. Antiferromagnetic materials, on the other hand, have been
known for decades and used extensively as passive exchange bias layers in spintronics
spin-valves [245]. Ten years ago, it was proposed that the antiferromagnetic order
parameter could be manipulated electrically using spin transfer torque [149, 205].
The recent prediction that SOT could be used to control collinear antiferromagnet
[150, 206, 246, 247] confirmed shortly after in CuMnAs, opened appealing avenues
as antiferromagnets are immune to external magnetic fields and host ultrafast (THz)
dynamics [114, 115]. To date, SOT-driven switching has been observed in the non-
 128
centrosymmetric CuMnAs [151] and Mn2 Au crystals [248, 249], but also in Pt/NiO
[250] and compensated Pt/CoGd bilayers [251]. The switching is not only robust
against external magnetic field but takes place at much faster rate compared to the
ferromagnet based SOT devices [252, 153, 116].
Due to their individual strength, these two fields of research have stimulated
substantial amount of theoretical and experimental research. However there are very
little attempt to combine these two fields. Considering the promises born by these
two classes of materials, it is natural to investigate the nature of SOTs in AF-TI
bilayers and identify its main features. One major difficulty in this regard is to
determine the right material combination, as interfacing TIs with transition metal
layers is known to significantly affect the topological surface states (e.g., [137, 233].
Nonetheless, recent experimental progress has been achieved towards the fabrication
of AF-TI heterostructures, revealing complex magnetic configuration at the interface
and suggesting viable routes towards the observation of SOTs [253, 254]. A proper
theoretical understanding of these systems is therefore highly solicited at this moment.
In this work, we present a systematic study of spin transport in AF-TI heterostruc-
tures based on a tight-binding model. We calculate the non-equilibrium spin density
using linear response theory and investigate its behavior with respect to the differ-
ent system’s parameters. We show the existence of a non-equilibrium staggered spin
density, localized near the interface and immune to scalar impurity at low energy.
Similar behavior is observed for the longitudinal conductivity at energy close to the
Dirac cone of the TI. The non-equilibrium spin densities show an angular dependence
similar to that of a Rashba ferromagnet which can be distorted by the interfacial
AF-TI coupling.
 129
7.1 Method

We use a tight binding model to describe the AF-TI heterostructure. The TI is

modeled following the method described in Ref. [140] with the same set of parameters.
To define the G-type antiferromagnet we have to double the unit cell (Fig. 7.1b) and
define lattice vectors r1 = (ex + ey ) and r2 = (ex − ey ) where ex and ey are the unit
vector along the x and y axis. The corresponding first Brillouin zone and reciprocal
lattice vectors (L1 , L2 ) are shown in Fig. 7.1c.

(a) (b) (c)

Figure 7.1: (a) Schematic of an AF-TI heterostructure. Red and blue spheres denotes
up and down magnetic moment of the AF layer and the green spheres denotes the
TI. (b) Top view of an AF lattice. The gray boundary denotes the unit cell. a is
the interatomic distance. (c) Corresponding first Brillouin zone (gray region) with
lattice vector L1 and L2 and high-symmetric points M , Γ and X for the AF-TI
heterostructure.

The Hamiltonian for the TI layers can be written as

X X
HT I = a†i (h0 )ai + a†i (hij )aj , (7.1)
i hi,ji

where a†i /ai is the creation/annihilation operator for the TI at ith site, hi denotes the
summation over the nearest neighbors and h0 and hij denote the onsite and hopping
 130
elements. For a TI the onsite and hopping elements are 4 × 4 matrices defined as

h0 = Γ1 M + I4 c,

h†−x = hx = (−iΓ2 A − Γ1 B − I4 d)/2,

h†−y = hy = (iΓ3 A − Γ1 B − I4 d)/2,

h†−z = hz = (−iΓ4 A1 − Γ1 B1 )/2, (7.2)

where I4 is the identity matrix of rank 4, while M , A, B, c, d, A1 and B1 are the

material parameters [140]. The Γ matrices are defined as

Γ1 = σ1 ⊗ I2 , Γ2 = σ3 ⊗ σ2 ,

Γ3 = σ3 ⊗ σ1 , Γ4 = σ2 ⊗ I2 , (7.3)

where I2 is the identity matrix of rank 2 and σ’s are the Pauli matrices for spin 1/2.
Similarly the Hamiltonian for the AF layers is given by

X X
HAF = c†i (∆i m.σ + ε0 I2 )ci − c†i (tij I2 )cj , (7.4)
i hi,ji

where c†i /ci is the creation/annihilation operator for the AF at ith site, ∆i = ∆(−1)xi +yi +zi
is the staggered onsite exchange energy, and ε0 is the uniform onsite energy which
can tune the position of the AF bands. In case of AF we use isotropic real hopping
as tx,y,z = t−x,−y,−z = tAF . The TI and AF layers are connected by

X
HAF T I = a†i (TC )cj + H.C.,
hi,ji

TC = tC (I2 , I2 ), (7.5)

where H.C. denotes the Hermitian conjugate. The typical parameter values we use
 131
here are A = 1.0, B = 1.5, M = 3.5, c = 1.5, d = 0.75, A1 = 1.5, B1 = 1.5, tAF = 0.15
and ∆ = 0.2. Unless explicitly mentioned otherwise, we use ε = 0.0 and tC = 0.5 and
set the magnetic moment of the AF layer out of plane. We consider here a system
with 10 layers (20 sites) of TI and 5 layers (10 sites) of AF. Corresponding band
structures for different coupling strengths are shown in Fig. 7.2. Note that the Dirac
cone coming from the interfacial TI layer is still preserved [159] and, depending on the
AF-TI coupling strength, is shifted to slightly higher energy. In addition, we notice
that the AF gap reduces upon increasing the coupling with the TI surface states.

(a) (b) (c)

1.5

1.0
E/A

0.5

0.0

-0.5
M Γ XM Γ XM Γ X

Figure 7.2: Band structure of a heterostructure with 10 layers of TI and 5 layers of

AF with a coupling strength (a) tC = 0, (b) tC = 0.5 and (c) tC = 0.75. The red,
green, blue and black colors correspond to the contributions from AF layers, bottom
TI layer, interfacial TI layer and bulk TI layers, respectively. The gray region shows
the gap due to exchange splitting for decoupled AF.

The nonequilibrium spin density and conductivity are calculated within the linear
response framework [223, 161, 236, 140]. We start by defining the retarded/advanced
Green’s function GR,A (E, k) at energy E and momentum k,

GR,A (E, k) = [(E ± iη)In − H(k)]−1 . (7.6)

For simplicity we write it as GR,A in the following section omitting the explicit depen-
dence of E and k. The expectation value of an observable O due to a perturbation
 132
P consists of two parts

hÔP̂ i = hÔP̂ isur + hÔP̂ isea , (7.7)

where Osur and Osea correspond to contributions from Fermi surface and Fermi sea,
respectively, and are defined by,

d2 k
Z
P̂ 1
hÔ isur = Re(Tr[ÔGR P̂(GA − GR )])EF ,
2π (2π)2
(7.8)
EF 2
 R
dk ∂G
Z Z
1
hÔP̂ isea = dE 2
Re Tr[ÔGR P̂
2π −∞ (2π) ∂E
∂GR

−Ô P̂GR ] . (7.9)
∂E

In this article we are interested in the non-equilibrium spin density and conduc-
tivity due to an applied electric field applied along x direction. Therefore the pertur-
bation term is given by P̂ = eEvx , where vx = ∂H/∂(~kx ) and E is the amplitude of
the applied electric field and e is the electrical charge. Since we are interested in the
response function, the computed quantities are normalized with respect to E. The
spin operator for individual site is given by

Ŝi = s ⊗ |iihi|, (7.10)

where s = σ for the AF and s = σ ⊗ I2 for TI sites. |iihi| is the projection operator
for ith site. Since we are using a bipartite lattice, instead of site resolved velocity
operator, we use an average velocity operator defined as

v̂xj = (1/~[∂H/∂kx ]...[layerj ]... )/2, (7.11)

where ...[layerj ]... refers to the fact that we take the block corresponding to the layer
 133
containing j th site from the full matrix. Since each layer contains two sites, we divide
this matrix by 2 to obtain the average contribution coming from a single site. The
response function for non-equilibrium spin density and the conductivity are given by

Si = hŜievˆx i, (7.12)
j
σxx = he(v̂xj )ev̂x i. (7.13)

7.2 Result and discussion

7.2.1 Spatial distribution of spin density and conductivity

First we calculate the site resolved spin density and conductivity for the AF layer
and the interfacial TI layer (Fig. 7.3). For simplicity we use natural units, i.e. e =
~ = a = 1 which does not affect the qualitative behavior of the observables.
From Fig. 7.2 we see that the coupling between TI and AF isolates one pair of
AF bands from the conduction bands and brings it lower in the exchange gap region.
This pair is dominated by the interfacial AF layer, which is also reflected in the fact
that the spin densities and conductivity are stronger in the interfacial layer at this
energy. Note that the spin densities show maxima at energies where the coupling
between the AF and TI layers are strongest, which breaks the continuity of the TI
bands (horizontal dashed lines in Fig. 7.3a). Correspondingly, the spin densities in
the TI layers reduce at these energies. Also note that at lower energy, the spin density
has opposite sign. This is because the coupling moves the Dirac cone from top TI
layer higher in energy and since the upper and lower halves of the Dirac cone possess
opposite texture, the induced non-equilibrium spin density also switches sign. To
understand the connection between the equilibrium spin texture and non-equilibrium
spin density, we calculate the site-resolved spin texture at different energies.
Figure 7.4 shows the band structure and corresponding spin texture for site
 134
(a) (b) (c) (d)
1.5

1.0

E/A 0.5

0.0
M Γ X

-0.1 0 0.1 -0.7 0 0.7 -1.7 0 1.7

AFM TI
B AFB5 AFB4 AFB3 AFB2 AFB1 TIB1 TIB2 TIB9 TIB10
A AFA5 AFA4 AFA3 AFA2 AFA1 TIA1 TIA2 TIA9 TIA10
1 2 3 4 5 6 7 14 15

Figure 7.3: (a) Band structure, non-equilibrium spin components (b) Sx (c) Sy and
(d) conductivity for sites 1 to 12 with out-of-plane magnetization. The red, blue,
green and black colors in (a) show the contribution from AF, interfacial TI, bottom
TI and bulk TI bands. The horizontal dashed lines denote the peaks of spin densities.
In (b), (c) and (d) the vertical dashed green lines show the AF layers and the thick
green line the position of the interface. Corresponding configuration is shown in
the bottom panel where the blue and red boxes denote AF sites with up and down
magnetic moments, and the green boxes correspond to TI sites.

8(AFB B B
2 , ↓), 10(AF1 , ↑) and 12(TI1 ) for different energies. At E/A = 0.156, the degen-

eracy of the AF bands have been lifted due to coupling with TI layer and is manifested
as two isolated rings in the Brillouin zone. Note that these two rings have opposite
textures and in case of a degenerate band they cancel each other. This is why the
non-equilibrium Sy , and therefore Sx , is zero in most of the regions. We further see
that due to confinement effect, AF states at different energies are dominated by differ-
ent layers, for example E/A = 0.582 is dominated by the 10th site (5th layer) whereas
E/A = 0.724 is dominated by the 8th site (4th layer). This explains the variation of
Sy component at different layers (Fig. 7.3). Note that at low energy the AF bands
are well separated in momentum from the interfacial TI Dirac cone, which allows the
 135
(a1) (b1) (c1) (d1)
1.0
0.8 M M M
0.6

E/A
Γ Γ Γ
0.4 X X X
0.2
0.0
M Γ X
(a2) (b2) (c2) (d2)
1.0
0.8 M M M
0.6
E/A

Γ Γ Γ
0.4 X X X
0.2
0.0
M Γ X
(a3) (b3) (c3) (d3)
1.0
0.8 M M M
0.6
E/A

Γ Γ Γ
0.4 X X X
0.2
0.0
M Γ X

↓ 0 ↑

Figure 7.4: (a) Band structure and spin texture at site (b) AFB B B
2 , (c) AF1 and (d) TI1
(see bottom panel in Fig. 7.3) at different energies [(1) E/A = 0.156, (2) E/A = 0.582,
(3)E/A = 0.724]. The arrow indicates the in-plane component and the color of the
arrow represents the out of plane component. The dotted lines correspond to the
degenerate bands.

TI layer to induce the SOC in AF layer and also to retain its own texture. From the
slope of the band structure one can see that the TI and the AF layers have opposite
group velocity within the intermediate energy range (0.15 < E/A < 0.85, which is
roughly the region between the second and sixth horizontal dashed lines in Fig. 7.3a).
As a result, within this regime the AF layer and the TI layer possess opposite Sy
component in spite of having similar spin texture.
 136
7.2.2 Fermi surface and sea contributions to Sx

The non-equilibrium Sx component is particularly important because it enables the

electrical control of the AF magnetic order. As a matter of fact, its magnitude de-
pends both on the induced SOC as well as the magnetization of the AF sublattices. As
a result one can obtain a staggered Sx that can be utilized to switch the antiferromag-
netic order parameter [150, 255]. Interestingly, unlike Sy , Sx has finite contribution
from both Fermi surface [Eq. (7.8)] and Fermi sea [Eq. (7.9)] (see Fig. 7.5) which
indicates its topological origin.
(a) (b) (c)
1.0
AFB3 AFB2 AFB1
0.8

0.6
E/A

0.4

0.2

0.0
-0.1 0. 0.1 0.2 -0.1 0. 0.1 0.2 -0.1 0. 0.1 0.2
Sx Sx Sx
Surface Sea Total

Figure 7.5: Contribution from Fermi surface (orange) and Fermi sea (green) for the B
sublattice of first three AF layers. The equilibrium Sz is shown in the boxed regions
where red and blue color correspond negative and positive values. The maximum
value in red box is 0.35 and maximum value in the blue box in 0.1.

One can readily see that both of the sea and surface terms dominate in the region
where the degeneracy of the AF band is lifted. Note that the Fermi sea contribution
attains a maximum value in the middle of the gap where the surface contribution
shows a minima. This is because while the Fermi sea term depends on the strength
of Berry curvature, the Fermi surface term depends on the induced SOC. To under-
stand the topological origin of the non-equilibrium Sx one needs to study the relevant
topological invariant. However, constructing a proper topological invariant for the
 137
interfacial states of such antiferromagnetic heterostructure is mathematically quite
challenging and is beyond the scope of present work. Instead, here we present a
heuristic argument to determine the Berry curvature contribution. We note that the
Berry curvature is analogous to a magnetic field whose strength is maximum when the
spin texture encounters a singularity marked by a zero in-plane and finite out-of-plane
component. To demonstrate that we consider the Rashba-Ferromagnet Hamiltonian

HR = ~2 k 2 /2m0 + αẑ.(σ × p) + ∆σz . (7.14)

Following Ref. [256] the Berry curvature reads

1 α2 ∆
Ω± (k) = ∓ , (7.15)
2 (∆2 + α2 k 2 )3/2

which can readily be compared with the momentum-dependent z-component of the

spin density,

∆
Sz± (k) = ± , (7.16)
(∆2 + α2 k 2 )1/2

where ± corresponds two different bands. It clearly appears that the maximum in
Berry curvature coincides with a maximum in Sz . Corresponding band structure
along with its spin projection and Berry curvature is shown in Fig. 7.6. One can
readily see that the Berry curvature is localized in a region where the out-of-plane
component of spin is maximum, consistently with Eq. (7.15,7.16) above. At these
points the in-plane texture vanishes creating a singularity. Therefore we track the
equilibrium Sz as a smoke signal for Berry curvature. From Fig. 7.5, we see that
most of the equilibrium Sz is localized at the bottom of the positive AF bands and
therefore most of the non-equilibrium Sx component is also generated in this region.
Since the bands with opposite texture have same Sz component (Fig. 7.4c1), the
 138
(a) (b) (c) (d)

E 0 0 0 0.5

Figure 7.6: Band structure of a Rashba-ferromagnet system (7.14) along kx with

projected (a) σy , (b) σz , (c) Berry curvature and (d) integrated Berry curvature or
the anomalous Hall coefficient. The dashed lines show the exchange gap. Here we
use ~ = 1, m0 = 0.5, α = 1 and ∆ = 0.05.

Fermi surface component switches sign as we move from the first AF band to the
second (E/A ∼ 0.156). Corresponding Berry curvature is localized at the bottom of
the bands where the Fermi sea term changes sign. At higher energy (E/A ∼ 0.724)
opposite textures possess opposite Sz (see the blue box in Fig. 7.5) and therefore
the Fermi surface term does not change sign anymore. The Fermi sea term being
dependent on the Berry curvature still oscillates with the sign of Sz . For a decoupled
AF band, there is no equilibrium Sz component near the band maxima. In an AF-TI
heterostructure this component, and therefore its corresponding Berry curvature is
produced due to the interaction between the TI and AF and therefore is localized
close to the interface. As a result, for E/A > 0.3, the non-equilibrium Sx component
is finite only near the interface (Fig. 7.3b) although the Sy component is visible deep
inside (Fig. 7.3c). This is also reflected in the fact that the Fermi sea term dominates
in the bottom of the positive AF band for the first AF layer (AFB
1 , Fig. 7.5c), whereas

for AFB B
2 onwards, the surface term dominates. As a result, while in AF1 the Sx

component has the same sign as the local magnetic moment, AFB B
2 and AF3 have

opposite signs.
 139
7.2.3 Effect of AF-TI coupling

To understand the cumulative behavior we introduce the staggered x-component of

tot
the spin density, Sxstg , the total y-component, Sytot , and the total conductivity, σxx ,
defined as

↑ ↓
X X
Sxstg = Sxi − Sxj (7.17)
i j
↑ ↓
X X
Sytot = Syi + Syj (7.18)
i j
↑ ↓
X X
tot j i
σxx = σxx + σxx (7.19)
i j

where the indices i and j run over the AF sites with up and down spin. The Sxstg
component induces the damping like torque that enables the electrical control of the
Neél order parameter [205, 150, 255], while the Sytot induces an effective field that
does not contribute to the torque. First we look at the effect of coupling between the
AF and TI which is the only way to imprint SOC on the AF layer and produce the
non-equilibrium spin density.
tot
Figure 7.7 shows the variation of Sxstg , Sytot and σxx as a function of energy and
with respect to the coupling strength between AF and TI layers (tC ). For better
understanding we divide the energy range into three regions. Region I corresponds
to the exchange gap of the decoupled AF bands. This region can be occupied by an
AF band only upon turning on the coupling between AF and TI (Fig. 7.2). Region II
spans the energy range corresponding to the bottom of the positive AF bands which
corresponds to the maximum density of states and also contains strong equilibrium
Sz component. The coupling opens a gap between different bands in this region
(Fig. 7.2). Region III contains all the energies above region II. The coupling mainly
distorts the TI bands rather than the AF bands here. Sytot increases in both region I
 140
(a) (b)
2 2

1 1

Sxstg × 10

Sytot
0 0

-1 I II -1 I II
III III
-2 -2
0. 0.2 0.4 0.6 0.8 1. 0. 0.2 0.4 0.6 0.8 1.
E/A E/A
(c) (d)
16 1.00 1.0

12 0.75 0.8
III
σxxtot

8 0.6
0.50

E/A
tC
4 I 0.4
II 0.25
III II
0 0.2
0. 0.2 0.4 0.6 0.8 1. 0.00 I
0.0
E/A M Γ X

tot
Figure 7.7: (a) Sxstg , (b) Sytot and (c) σxx of the AF layer as a function of coupling
strength whose value is shown in the colorbar. (d) The band structure of decoupled
(red) AF-TI and with coupling tC = 0.5A (blue). The dashed lines show three
different regions I, II and III.

and region II as it depends only on the induced SOC. Note that the Sx component
switches sign with respect to the local magnetization in the first two consecutive
layers (Fig. 7.3), which reduces Sxstg . This effect is more prominent in region I as it
is dominated both by the first (AFA,B A,B
1 ) and second (AF2 ) interfacial as well as bulk

AF layer (Fig. 7.3). This is also reflected in the equilibrium spin texture (Fig. 7.4). As
a result, in region I, Sxstg starts decreasing for stronger coupling strength (Fig. 7.7a).
Region II is less dominated by the interfacial AF layer and therefore the cancellation
due to opposite texture is minimized. Moreover stronger coupling opens larger gaps
which provides more Sz component (i.e. stronger Berry curvature) giving rise to
larger value of Sxstg . In region III the coupling breaks the degeneracy of the AF bands
which ensures that only one type of texture dominates. For smaller coupling two
opposite textures have almost same contribution and cancel each other. For stronger
coupling the TI bands undergo more distortion resulting in reduced SOC and weak
 141
spin texture. Therefore we see that in region III both Sytot and Sxstg attain a maximum
value for some intermediate coupling strength. The conductivity on the other hand
depends on the density of states and increases in region I and decreases in region II
and III with coupling strength as the coupling pulls down the AF states from region
II and III to region I (Fig. 7.7).

7.2.4 Coupling to surface vs bulk TI states

So far we have considered the AF bands coupled to the surface TI bands only. In
reality the AF bands can be connected to the bulk TI bands as well or can be coupled
to both surface and bulk. To understand these different scenarios we change the
position of the AF bands by varying the onsite energy [ε0 , Eq. (7.4)] of the AF layer.

(a) (b) (c)

E/A

ε0 /A ε0 /A ε0 /A
-0.4 0. 0.4 -4. 0. 4. -16. 0. 16.

tot
Figure 7.8: Energy dependence of (a) Sxstg , (b) Sytot and (c) σxx of the AF layer as a
function of the onsite energy of the AF layer with tC = 0.5A. The inclined dashed
line shows the region spanned by the decoupled AF bands.

Fig. 7.8 shows the variation of Sxstg , Sytot and σxx

tot
for different onsite energies ε0 .
tot
The total conductivity (σxx ) is confined within the region marked by the dashed line
denoting the energy range occupied by the AF bands. For Sxstg and Sytot , one can
clearly see five branches corresponding to the intersections of AF bands with the TI
bands at lower energies. Careful observation reveals a faint branch with opposite sign
within the exchange gap region which corresponds to the interfacial AF band. For
larger values of ε0 the AF bands are coupled to more number of bulk TI bands which
increases both Sy and Sx . However one should note that in this case more current
 142
flows through the bulk TI, which effectively reduces the spin Hall angle and hence
the efficiency of the system [140]. Besides, the bulk bands are not immune to the
scattering which makes the coupling to the bulk bands inefficient against impurity
scattering.

7.2.5 Effect of impurity

Next we study the effect of impurity scattering modelled as a constant broadening,

an approach that is found to be sufficient in realistic systems as well [223]. Figure
7.9 shows the variation of Sxstg , Sytot and σxx
tot
with the broadening parameter (η). We
have already seen that the non-equilibrium Sx and Sy are maximum only at some
specific values of energies (Figs. 7.3,7.7). Here we consider the energies marked by
the dashed line in Fig. 7.3 and Fig. 7.5.

(a) (b)
SyFM 109 V -1 .m -1

1.0 20
-4 -1
xx 10 Ω

0.5 15
0.0 10
-0.5 5
σTI

-1.0 0
0 2 4 6 8 10 0 2 4 6 8 10
(c)
SxFM 109 V -1 .m -1

0.0
E/A
-0.3 0.0 0.2 0.4
-0.6 0.6 0.8 1.0
1.2 1.4 1.6
-0.9
0 2 4 6 8 10
η/A 10-2

Figure 7.9: (a) Sxstg , (b) Sytot and (c) σxx

tot
as a function of broadening. Different line
colors correspond to different energies as shown in panel (d) along with the band
structure.

From Fig. 7.9 we see that for energies coming from region III (Fig. 7.7), Sxstg , Sytot
tot
and σxx show a 1/η decay which is the characteristic of a metallic system. However
the behavior is completely different when we choose an energy from the region I
tot
(see Fig. 7.9, blue line). We see that σxx remains constant as the effect of impurity
 143
is quenched by the strong spin polarization. Sytot also falls down but with a much
slower rate which indicates the weakening of the induced SOC. Sxstg shows an initial
increase and then becomes almost constant which also points towards its topologically
protected origin. Note that the enhancement of Sxstg component due to scattering has
also been observed in two dimensional antiferromagnetic TIs [163] as well as TI-
ferromagnet heterostructures [140].

7.2.6 Angular dependence of torques

Finally we present the angular dependence of field like and damping like torques with
respect to the polar (θ) and azimuthal (φ) angles (Fig. 7.10) for different planes. Note
that in the case of out-of-plane magnetization the field like and damping like torques
are directly proportional to Sytot and Sxstg . Therefore one can easily understand the
nature of field like and damping like torques from the Sxstg and Sytot components. In
general the torque components at each site can be expressed as

T i = mi × S i

= τFi (mi × ey ) + τDi mi × (mi × ey ), (7.20)

ex is direction of current flow and ez is normal to the surface. The general expressions
i
for τF,D for any arbitrary angle are given by

τFi = (mi × S i ) · (mi × ey )/|mi × ey |2

sin φ cos φ sin2 θ i i sin φ sin θ cos θ i
= − 2 Sx + Sy − S,
2
1 − sin φ sin θ 1 − sin2 φ sin2 θ z
(7.21)

τDi = (mi × S i ) · (mi × (mi × ey ))/|mi × (mi × ey )|2

cos θ i sin θ cos φ i
= − 2 2 Sx + 2 2 Sz . (7.22)
1 − sin φ sin θ 1 − sin φ sin θ
 144

Plane τFi τDi

zx(φ = 0) Syi -Sxi cos θ + Szi sin θ
xy(θ = π/2) Syi − Sxi tan φ Szi / cos φ
i i
xy(φ = π/2) Sy − Sz tan θ Sxi / cos θ

Table 7.1: Torque coefficients τFi and τDi for zx, xy and yz plane.

Note that each layer contains two sites with opposite magnetization. Since these
two sites have same sign of induced SOC, they would have same sign of field like
coefficient (τF ) and opposite sign of damping like coefficient (τD ). Therefore we
define the total field like and damping like torque coefficients as

↑ ↓
X X
τF = τFi + τFj (7.23)
i j
↑ ↓
X X
τD = τDi − τDj . (7.24)
i j

From Fig. 7.10 one can readily see that both τF and τD do not show any angular
dependence for all energies dominated by the bulk AF layers (region II and III in Fig.
7.7). When the magnetization is in the xz plane (Fig. 7.10a,d), it does not affect the
non-equilibrium Sy component. As a result τF remains constant for all energies. Since
Sx,z is generated through the interaction with z, x components of magnetization, they
change as cos θ, sin θ respectively and therefore τD remains constant. The situation is
different for xy and yz planes. In these cases, the non-equilibrium Sy component also
has a contribution from the y component of magnetization. In addition the texture
itself is modified resulting in a complex angular dependence. Note that for xy, yz
planes, the periodicity of Sy and Sx,z seems to be double compared to the periodicity
of Sz,x . Careful observation shows that the variation is not strictly sinusoidal and
depends on the strength of the induced SOC. As a result the blue lines undergo less
distortion due to their stronger SOC compared to the other lines. The Sz,x compo-
 145
(a) (b) (c)
2 2 2

τF × 10
0 ϕ=0 0 θ=π/2 0 ϕ=π/2
-2 -2 -2
-4 -4 -4
(d) (e) (f)
8 8 8
τD × 1000 4 4 4
ϕ=0 θ=π/2 ϕ=π/2
0 0 0
-4 -4 -4
(g) (h) (i)
9 1
Sz × 10 Sx × 10

1
0 0 0
-1 -9 θ /2 -1
1 1 9
0 0 0
-1 -1 -9 /2
1 1 1
0 0 0
Sy

-1 -1 -1
0. 0.5 1. 0. 0.5 1. 0. 0.5 1.
θ/π ϕ/π θ/π

Figure 7.10: Variation of τ̄F L (a,b,c) and τ̄DL (d,e,f) with respect to θ (a,c;d,f) and φ
(b;e). The inset value shows the fixed angle. (g), (h) and (i) shows Sx , Sz and Sy
where solid and dashed line denotes the contribution from AF sites with up and down
spin. The colors represent different energies as described in Fig. 7.9d.

nent, on the other hand, is generated through the interaction with x, z component
of magnetization and therefore shows a cos φ, θ dependence. Note that in case of τD ,
there is small deviation from constant value for the blue line in all three cases. This
effect comes from the “breathing” of the band structure due to the interplay between
magnetism and SOC [206]. For τF this effect is nullified by the contribution from Sy
and Sx,z . This is further clarified by showing the angular dependence of τD and Sz in
xy plane for different coupling strengths (Fig. 7.11).

7.3 Conclusion

In this article we present a systematic study of non-equilibrium spin density and SOT
for an AF-TI heterostructure using a tight binding model within the framework of
 146
(a) (b) (c)
8 8 8

τDL × 1000
4 4 4
θ=π/2 θ=π/2 θ=π/2
0 0 0

-4 -4 -4
(d) (e) (f)
2 2 2

z × 10
1 1 1
0 0 0
Sstg

-1 -1 -1
-2 -2 -2
0 0.5 1 0 0.5 1 0 0.5 1
ϕ/π ϕ/π ϕ/π

Figure 7.11: Variation of τD in xy plane for (a) tC = 0.25, (b) tC = 0.50 and (c)
tC = 0.75 and corresponding non-equilibrium Sz (d,e,f). Different colors correspond
the energies at which Sz have peaks. The band structures are shown in Fig. 7.2.

linear response theory. From the site resolved spin density we show the existence of a
staggered Sx component which can be utilized to switch the antiferromagnetic order
parameter. We explain the behavior of different spin components at different energies
and show their connections with the band structure and equilibrium spin texture.
We show that for an out-of-plane magnetization, the Sx component is dominated by
both Fermi sea and Fermi surface contributions. The contribution from the Fermi
sea points towards its topological origin, which is later verified from the impact of
impurity. We further study the effect of the coupling between the AF and TI layers
and show that the Sx component attains its maximum value for an intermediate
coupling strength whereas Sy component shows different behavior at different energies.
The non-equilibrium spin density also depends on the total amount of overlap which
we demonstrate by coupling the AF bands with both the surface and the bulk TI
bands. However the spin density produced away from the Dirac cone of the TI are
not robust against impurity and falls down rapidly with growing impurity strength.
Interestingly we find that near the Dirac cone the Sx component is slightly enhanced
by the impurity. At this energy, the conductivity is not affected by the impurity.
Finally we show the angular dependence of the non-equilibrium spin densities as well
 147
as the different torque coefficients, which have the same behavior as a two dimensional
Rashba gas for the bulk AF bands. However within the AF gap, the damping like
torque shows a complex angular dependence due to the breathing of band structure
which can be enhanced with the coupling strength between AF and TI.
 148

Chapter 8

Concluding Remarks

8.1 Summary

In this thesis, we present a systematic study of the non-equilibrium transport prop-

erties of topologically nontrivial systems. The nontrivial topology manifests itself
as an edge or surface state. We start by demonstrating the effect of interference
between the bulk and the edge states on the dynamics of the wavepacket in a quasi-
one-dimensional system. We study a relativistic phenomenon called zitterbewegung,
which is the oscillation of wavepacket due to the interference of different energy states.
These oscillations not only provide the information about the bulk and edge states
but also can be utilised to detect the topological phases of the system. Then we
move to the transport properties of the two-dimensional magnetic topological insu-
lator. We use Bernevig-Hughes-Zhang model to model the topological insulator and
study the impact of magnetic ordering on its conductance as well as non-equilibrium
spin density. We show that the antiferromagnetic topological insulators are more
robust compared to the ferromagnetic one by calculating their conductance for differ-
ent impurity strengths. Interestingly, we find that the non-equilibrium spin density
corresponding to the damping like torque is enhanced by impurity when the system
is in the topologically nontrivial regime. Then we study the topological insulator-
ferromagnet heterostructure and show how the non-equilibrium spin density changes
at different layers with the coupling strength. We show how this coupling affects the
band structure and spin texture and also show their connection to the non-equilibrium
 149
spin density. We find that close to the damping like torque is robust against the scalar
impurity. The resulting torques are more efficient within the energy range dominated
by the surface states of the topological insulator. Finally, we study the topological
insulator-antiferromagnet heterostructure. Unlike the ferromagnetic heterostructure,
the Dirac cone coming from the interfacial topological insulator states are preserved
in this case offering stronger spin-orbit coupling. We find that the non-equilibrium
spin density corresponding to the damping like torque is mostly localised at the inter-
face and have a finite contribution from all the bands below the Fermi level. Within
the antiferromagnetic gap region, this component is robust against scalar impurity
and therefore can offer robust spin-orbit torque. These studies provide a qualita-
tive understanding of topological insulators as well as their interfaces with magnetic
materials. Our studies indicate that antiferromagnets are better suited with topolog-
ical insulators compared to ferromagnets for generating non-equilibrium spin density
which would be beneficial for designing next-generation magnetic memory devices.

8.2 Future Research Work

Although the use of simplified models offers comprehensive transparency, they can
not offer significant quantitative accuracy. A more rigorous approach to overcome
this limitation is to merge these methods with first principle calculation which can
provide more accurate results for different materials. So far this work has been focused
on one class of topological insulators only. These studies can be easily extended
to other classes of emergent material like topological crystalline insulators or Weyl
fermions which can reveal their pertinent features that can be instrumental for new
applications.
 150

REFERENCES

[1] C. Kittel. “Introduction to Solid State Physics”. Willey, 8th Edition, (2004).

[2] E. H. Hall. “On a New Action of the Magnet on Electric Currents”. Am. J.
Math. 2, 287 (1879).

[3] B. J. van Wees, et. al. “Quantized conductance of point contacts in a two-
dimensional electron gas”. Phys. Rev. Lett. 60, 848 (1988).

[4] D. Ferry and S. M. Goodnick. “Transport in Nanostructures”. Cambridge

University Press, Cambridge, (1997).

[5] Y. Onose, et. al. “Observation of the Magnon Hall Effect”. Science (80-. ).
329, 297 (2010), [arXiv:1008.1564].

[6] X. Yin, Z. Ye, J. Rho, Y. Wang, and X. Zhang. “Photonic Spin Hall Effect at
Metasurfaces”. Science (80-. ). 339, 1405 (2013).

[7] L. Zhang, J. Ren, J.-S. Wang, and B. Li. “The phonon Hall effect: theory
and application”. J. Phys. Condens. Matter 23, 305402 (2011), [arXiv:1101.
5229].

[8] K. von Klitzing. “Quantum Hall Effect: Discovery and Application”. Annu.
Rev. Condens. Matter Phys. 8, 13 (2017).

[9] K. V. Klitzing, G. Dorda, and M. Pepper. “New Method for High-Accuracy

Determination of the Fine-Structure Constant Based on Quantized Hall Resis-
tance”. Phys. Rev. Lett. 45, 494 (1980).

[10] D. C. Tsui, H. L. Stormer, and A. C. Gossard. “Two-Dimensional Magneto-

transport in the Extreme Quantum Limit”. Phys. Rev. Lett. 48, 1559 (1982).

[11] M. A. Paalanen, D. C. Tsui, and A. C. Gossard. “Quantized Hall effect at low

temperatures”. Phys. Rev. B 25, 5566 (1982).

[12] T. Ando and Y. Uemura. “Theory of Quantum Transport in a Two-Dimensional

Electron System under Magnetic Fields. I. Characteristics of Level Broadening
and Transport under Strong Fields”. J. Phys. Soc. Japan 36, 959 (1974).
 151
[13] T. Ando. “Theory of Quantum Transport in a Two-Dimensional Electron Sys-
tem under Magnetic Fields II. Single-Site Approximation under Strong Fields”.
J. Phys. Soc. Japan 36, 1521 (1974).

[14] T. Ando. “Theory of Quantum Transport in a Two-Dimensional Electron Sys-

tem under Magnetic Fields. III. Many-Site Approximation”. J. Phys. Soc. Japan
37, 622 (1974).

[15] R. B. Laughlin. “Quantized Hall conductivity in two dimensions”. Phys. Rev.

B 23, 5632 (1981).

[16] B. I. Halperin. “Quantized Hall conductance, current-carrying edge states,

and the existence of extended states in a two-dimensional disordered potential”.
Phys. Rev. B 25, 2185 (1982).

[17] A. Rauh, G. H. Wannier, and G. Obermair. “Bloch Electrons in Irrational

Magnetic Fields”. Phys. status solidi 63, 215 (1974).

[18] D. R. Hofstadter. “Energy levels and wave functions of Bloch electrons in ra-
tional and irrational magnetic fields”. Phys. Rev. B 14, 2239 (1976).

[19] D. J. Thouless, M. Kohmoto, M. P. Nightingale, and M. den Nijs. “Quantized

Hall Conductance in a Two-Dimensional Periodic Potential”. Phys. Rev. Lett.
49, 405 (1982).

[20] R. Kubo. “Statistical-Mechanical Theory of Irreversible Processes. I. General

Theory and Simple Applications to Magnetic and Conduction Problems”. J.
Phys. Soc. Japan 12, 570 (1957).

[21] R. Kubo, M. Yokota, and S. Nakajima. “Statistical-Mechanical Theory of Irre-

versible Processes. II. Response to Thermal Disturbance”. J. Phys. Soc. Japan
12, 1203 (1957).

[22] M. Kohmoto. “Topological invariant and the quantization of the Hall conduc-
tance”. Ann. Phys. (N. Y). 160, 343 (1985).

[23] Y. Hatsugai. “Chern number and edge states in the integer quantum Hall effect”.
Phys. Rev. Lett. 71, 3697 (1993).

[24] E. Hall. “XXXVIII. On the new action of magnetism on a permanent electric

current”. Philos. Mag. Ser. 5 10, 301 (1880).
 152
[25] E. Hall. “XVIII. On the Rotational Coefficient in nickel and cobalt”. Philos.
Mag. Ser. 5 12, 157 (1881).

[26] R. Karplus and J. M. Luttinger. “Hall Effect in Ferromagnetics”. Phys. Rev.

95, 1154 (1954).

[27] M. V. Berry. “Quantal Phase Factors Accompanying Adiabatic Changes”. Proc.

R. Soc. A Math. Phys. Eng. Sci. 392, 45 (1984).

[28] J. Smit. “The spontaneous hall effect in ferromagnetics I”. Physica 21, 877
(1955).

[29] J. Smit. “The spontaneous hall effect in ferromagnetics II”. Physica 24, 39
(1958).

[30] L. Berger. “Side-Jump Mechanism for the Hall Effect of Ferromagnets”. Phys.
Rev. B 2, 4559 (1970).

[31] M. I. D’yakonov and V. I. Perel’. “Possibility of Orienting Electron Spins with

Current”. JETP Lett. 13, 467 (1971).

[32] M. Dyakonov and V. Perel. “Current-induced spin orientation of electrons in

semiconductors”. Phys. Lett. A 35, 459 (1971).

[33] J. E. Hirsch. “Spin Hall Effect”. Phys. Rev. Lett. 83, 1834 (1999).

[34] E. M. Hankiewicz, L. W. Molenkamp, T. Jungwirth, and J. Sinova. “Manifesta-

tion of the spin Hall effect through charge-transport in the mesoscopic regime”.
Phys. Rev. B 70, 241301 (2004), [arXiv:0409334].

[35] Y. K. Kato, R. C. Myers, A. C. Gossard, and D. D. Awschalom. “Observation

of the spin hall effect in semiconductors”. Science (80-. ). 306, 1910 (2004).

[36] J. Wunderlich, B. Kaestner, J. Sinova, and T. Jungwirth. “Experimental Obser-

vation of the Spin-Hall Effect in a Two-Dimensional Spin-Orbit Coupled Semi-
conductor System”. Phys. Rev. Lett. 94, 047204 (2005), [arXiv:0410295].

[37] S. Zhang. “Spin Hall Effect in the Presence of Spin Diffusion”. Phys. Rev.
Lett. 85, 393 (2000).

[38] S. O. Valenzuela and M. Tinkham. “Direct electronic measurement of the spin

Hall effect”. Nature 442, 176 (2006), [arXiv:0605423].
 153
[39] T. Kimura, Y. Otani, T. Sato, S. Takahashi, and S. Maekawa. “Room-
Temperature Reversible Spin Hall Effect”. Phys. Rev. Lett. 98, 156601 (2007),
[arXiv:0609304].

[40] C. Brüne, et. al. “Evidence for the ballistic intrinsic spin Hall effect in HgTe
nanostructures”. Nat. Phys. 6, 448 (2010), [arXiv:0812.3768].

[41] G. Chen. “Skyrmion Hall effect”. Nat. Phys. 13, 112 (2017).

[42] N. Nagaosa and Y. Tokura. “Topological properties and dynamics of magnetic

skyrmions”. Nat. Nanotechnol. 8, 899 (2013).

[43] W. Jiang, et. al. “Direct observation of the skyrmion Hall effect”. Nat. Phys.
13, 162 (2017), [arXiv:1603.07393].

[44] G. Dresselhaus. “Spin-Orbit Coupling Effects in Zinc Blende Structures”. Phys.

Rev. 100, 580 (1955).

[45] E. I. Rashba and V. I. Sheka. “Symmetry of Energy Bands in Crystals of

Wurtzite Type II. Symmetry of Bands with Spin-Orbit Interaction Included”.
Fiz. Tverd. Tela Collect. Pap. 2, 162 (1959).

[46] D. Stein, K. V. Klitzing, and G. Weimann. “Electron Spin Resonance on GaAs−

Alx Ga1−x As Heterostructures”. Phys. Rev. Lett. 51, 130 (1983).

[47] H. L. Stormer, et. al. “Energy Structure and Quantized Hall Effect of Two-
Dimensional Holes”. Phys. Rev. Lett. 51, 126 (1983).

[48] F. T. Vas’ko. “Spin splitting in the spectrum of two-dimensional electrons due

to the surface potential”. Pis’ma Zh. Eksp. Teor. Fiz. 30, 574 (1979).

[49] P. Gambardella and I. M. Miron. “Current-induced spin-orbit torques”. Philos.

Trans. R. Soc. A Math. Phys. Eng. Sci. 369, 3175 (2011).

[50] M. Blume and R. E. Watson. “Theory of Spin-Orbit Coupling in Atoms. I.

Derivation of the Spin-Orbit Coupling Constant”. Proc. R. Soc. A Math. Phys.
Eng. Sci. 270, 127 (1962).

[51] M. Marco, A. Credi, L. Prodi, and M. T. Gandolfi. “Handbook of Photochem-

istry”. CRC Press, third Edition, (2006).
 154
[52] V. M. Edelstein. “Spin polarization of conduction electrons induced by electric
current in two-dimensional asymmetric electron systems”. Solid State Commun.
73, 233 (1990).

[53] S. LaShell, B. A. McDougall, and E. Jensen. “Spin Splitting of an Au(111)

Surface State Band Observed with Angle Resolved Photoelectron Spectroscopy”.
Phys. Rev. Lett. 77, 3419 (1996).

[54] E. Rotenberg, J. W. Chung, and S. D. Kevan. “Spin-Orbit Coupling Induced

Surface Band Splitting in Li/W(110) and Li/Mo(110)”. Phys. Rev. Lett. 82,
4066 (1999).

[55] M. Hochstrasser, J. G. Tobin, E. Rotenberg, and S. D. Kevan. “Spin-Resolved

Photoemission of Surface States of W(110)-(1×1)H”. Phys. Rev. Lett. 89,
216802 (2002).

[56] M. Hoesch, et. al. “Spin structure of the Shockley surface state on Au(111)”.
Phys. Rev. B 69, 241401 (2004).

[57] C. Tusche, A. Krasyuk, and J. Kirschner. “Spin resolved bandstructure imag-

ing with a high resolution momentum microscope”. Ultramicroscopy 159, 520
(2015).

[58] G. Bihlmayer, Y. M. Koroteev, P. M. Echenique, E. V. Chulkov, and S. Blügel.

“The Rashba-effect at metallic surfaces”. Surf. Sci. 600, 3888 (2006).

[59] M. Z. Hasan and C. L. Kane. “Colloquium : Topological insulators”. Rev. Mod.

Phys. 82, 3045 (2010), [arXiv:1002.3895].

[60] M. Z. Hasan and J. E. Moore. “Three-Dimensional Topological Insulators”.

Annu. Rev. Condens. Matter Phys. 2, 55 (2011), [arXiv:1011.5462].

[61] X.-L. Qi and S.-C. Zhang. “Topological insulators and superconductors”. Rev.
Mod. Phys. 83, 1057 (2011), [arXiv:1008.2026].

[62] L. Fu and C. L. Kane. “Time reversal polarization and a Z2 adiabatic spin

pump”. Phys. Rev. B 74, 195312 (2006), [arXiv:0606336].

[63] L. Fu, C. L. Kane, and E. J. Mele. “Topological Insulators in Three Dimen-

sions”. Phys. Rev. Lett. 98, 106803 (2007), [arXiv:0607699].
 155
[64] J. E. Moore and L. Balents. “Topological invariants of time-reversal-invariant
band structures”. Phys. Rev. B 75, 121306 (2007), [arXiv:0607314].

[65] F. D. M. Haldane. “Model for a Quantum Hall Effect without Landau Levels:
Condensed-Matter Realization of the ”Parity Anomaly””. Phys. Rev. Lett. 61,
2015 (1988).

[66] G. Jotzu, et. al. “Experimental realization of the topological Haldane model with
ultracold fermions”. Nature 515, 237 (2014).

[67] C. L. Kane and E. J. Mele. “Quantum Spin Hall Effect in Graphene”. Phys.
Rev. Lett. 95, 226801 (2005), [arXiv:0411737].

[68] C. L. Kane and E. J. Mele. “Z2 Topological Order and the Quantum Spin Hall
Effect”. Phys. Rev. Lett. 95, 146802 (2005), [arXiv:0506581].

[69] S. Murakami, N. Nagaosa, and S.-C. Zhang. “Spin-Hall Insulator”. Phys. Rev.
Lett. 93, 156804 (2004), [arXiv:0406001].

[70] B. A. Bernevig, T. L. Hughes, and S.-C. Zhang. “Quantum Spin Hall Effect
and Topological Phase Transition in HgTe Quantum Wells”. Science (80-. ).
314, 1757 (2006).

[71] M. Konig, et. al. “Quantum Spin Hall Insulator State in HgTe Quantum Wells”.
Science (80-. ). 318, 766 (2007), [arXiv:0710.0582].

[72] R. Roy. “Topological phases and the quantum spin Hall effect in three dimen-
sions”. Phys. Rev. B 79, 195322 (2009), [arXiv:0607531].

[73] H. Zhang, et. al. “Topological insulators in Bi2 Se3 , Bi2 T e3 and Sb2 T e3 with a
single Dirac cone on the surface”. Nat. Phys. 5, 438 (2009).

[74] D. Hsieh, et. al. “A topological Dirac insulator in a quantum spin Hall phase”.
Nature 452, 970 (2008).

[75] Y. L. Chen, et. al. “Experimental Realization of a Three-Dimensional Topo-

logical Insulator, Bi2Te3”. Science (80-. ). 325, 178 (2009), [arXiv:arXiv:
0904.1829v1].

[76] Y. Zhang, et. al. “Crossover of Three-Dimensional Topological Insulator of

Bi2Se3 to the Two-Dimensional Limit”. Nat. Phys. 6, 584 (2009), [arXiv:
0911.3706].
 156
[77] Y. Xia, et. al. “Observation of a large-gap topological-insulator class with a
single Dirac cone on the surface”. Nat. Phys. 5, 398 (2009).

[78] D. Hsieh, et. al. “Observation of Unconventional Quantum Spin Textures in

Topological Insulators”. Science (80-. ). 323, 919 (2009).

[79] Y. Ando. “Topological Insulator Materials”. J. Phys. Soc. Japan 82, 102001
(2013), [arXiv:1304.5693].

[80] S.-Y. Lin, et. al. “Theoretical search for half-Heusler topological insulators”.
Phys. Rev. B 91, 094107 (2015), [arXiv:1405.1305].

[81] D. Hsieh, et. al. “A tunable topological insulator in the spin helical Dirac trans-
port regime”. Nature 460, 1101 (2009), [arXiv:0908.0564].

[82] J. Slonczewski. “Current-driven excitation of magnetic multilayers”. J. Magn.

Magn. Mater. 159, L1 (1996).

[83] L. Berger. “Emission of spin waves by a magnetic multilayer traversed by a

current”. Phys. Rev. B 54, 9353 (1996).

[84] A. Manchon and S. Zhang. “Theory of nonequilibrium intrinsic spin torque in

a single nanomagnet”. Phys. Rev. B 78, 212405 (2008).

[85] A. Manchon and S. Zhang. “Theory of spin torque due to spin-orbit coupling”.
Phys. Rev. B 79, 094422 (2009).

[86] M. D. Stiles and A. Zangwill. “Anatomy of spin-transfer torque”. Phys. Rev.

B 66, 014407 (2002), [arXiv:0202397].

[87] D. Ralph and M. Stiles. “Spin transfer torques”. J. Magn. Magn. Mater. 320,
1190 (2008), [arXiv:0711.4608].

[88] A. Matos-Abiague and R. L. Rodrı́guez-Suárez. “Spin-orbit coupling mediated

spin torque in a single ferromagnetic layer”. Phys. Rev. B 80, 094424 (2009).

[89] K. M. D. Hals and A. Brataas. “Phenomenology of current-induced spin-orbit

torques”. Phys. Rev. B 88, 085423 (2013), [arXiv:1307.0395].

[90] A. Brataas and K. M. D. Hals. “Spinorbit torques in action”. Nat. Nanotechnol.

9, 86 (2014).
 157
[91] K. Garello, et. al. “Symmetry and magnitude of spinorbit torques in ferromag-
netic heterostructures”. Nat. Nanotechnol. 8, 587 (2013), [arXiv:1301.3573].

[92] A. Manchon, H. C. Koo, J. Nitta, S. M. Frolov, and R. A. Duine. “New

perspectives for Rashba spinorbit coupling”. Nat. Mater. 14, 871 (2015),
[arXiv:1507.02408].

[93] A. Chernyshov, et. al. “Evidence for reversible control of magnetization in a

ferromagnetic material by means of spinorbit magnetic field”. Nat. Phys. 5, 656
(2009), [arXiv:0812.3160].

[94] I. Mihai Miron, et. al. “Current-driven spin torque induced by the Rashba effect
in a ferromagnetic metal layer”. Nat. Mater. 9, 230 (2010).

[95] I. M. Miron, et. al. “Perpendicular switching of a single ferromagnetic layer

induced by in-plane current injection”. Nature 476, 189 (2011).

[96] L. Liu, et. al. “Spin-Torque Switching with the Giant Spin Hall Effect of Tan-
talum”. Science (80-. ). 336, 555 (2012), [arXiv:1203.2875].

[97] M. Jamali, et. al. “Spin-Orbit Torques in Co/Pd Multilayer Nanowires”. Phys.
Rev. Lett. 111, 246602 (2013), [arXiv:1311.3033].

[98] Y. Fan, et. al. “Magnetization switching through giant spinorbit torque in a
magnetically doped topological insulator heterostructure”. Nat. Mater. 13, 699
(2014).

[99] Y.-c. Lau, D. Betto, K. Rode, J. M. D. Coey, and P. Stamenov. “Spinorbit

torque switching without an external field using interlayer exchange coupling”.
Nat. Nanotechnol. 11, 758 (2016).

[100] S. Fukami, C. Zhang, S. DuttaGupta, A. Kurenkov, and H. Ohno. “Magneti-

zation switching by spinorbit torque in an antiferromagnetferromagnet bilayer
system”. Nat. Mater. 15, 535 (2016).

[101] C. Ciccarelli, et. al. “Room-temperature spinorbit torque in NiMnSb”. Nat.

Phys. 12, 855 (2016).

[102] N. Tesaová, et. al. “Experimental observation of the optical spinorbit torque”.
Nat. Photonics 7, 492 (2013), [arXiv:1207.0307].
 158
[103] R. Mikhaylovskiy, et. al. “Ultrafast optical modification of exchange interactions
in iron oxides”. Nat. Commun. 6, 8190 (2015).

[104] K. Ando, et. al. “Electric Manipulation of Spin Relaxation Using the Spin Hall
Effect”. Phys. Rev. Lett. 101, 036601 (2008).

[105] L. Liu, T. Moriyama, D. C. Ralph, and R. A. Buhrman. “Spin-Torque Ferro-

magnetic Resonance Induced by the Spin Hall Effect”. Phys. Rev. Lett. 106,
036601 (2011), [arXiv:1011.2788].

[106] 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), [arXiv:1301.4513].

[107] I. Garate and A. H. MacDonald. “Influence of a transport current on mag-

netic anisotropy in gyrotropic ferromagnets”. Phys. Rev. B 80, 134403 (2009),
[arXiv:0905.3856].

[108] P. M. Haney, H.-W. Lee, K.-J. Lee, A. Manchon, and M. D. Stiles. “Current-
induced torques and interfacial spin-orbit coupling”. Phys. Rev. B 88, 214417
(2013), [arXiv:1301.4513v1].

[109] J. Kim, et. al. “Layer thickness dependence of the current-induced effective
field vector in Ta—CoFeB—MgO”. Nat. Mater. 12, 240 (2013), [arXiv:1207.
2521].

[110] X. Fan, et. al. “Observation of the nonlocal spin-orbital effective field”. Nat.
Commun. 4, 1799 (2013).

[111] J. Kim, et. al. “Anomalous temperature dependence of current-induced torques

in CoFeB/MgO heterostructures with Ta-based underlayers”. Phys. Rev. B 89,
174424 (2014).

[112] C.-F. Pai, et. al. “Enhancement of perpendicular magnetic anisotropy and
transmission of spin-Hall-effect-induced spin currents by a Hf spacer layer in
W/Hf/CoFeB/MgO layer structures”. Appl. Phys. Lett. 104, 082407 (2014),
[arXiv:1401.4617].

[113] M.-h. Nguyen, D. C. Ralph, and R. A. Buhrman. “Spin Torque Study of the
Spin Hall Conductivity and Spin Diffusion Length in Platinum Thin Films with
Varying Resistivity”. Phys. Rev. Lett. 116, 126601 (2016).
 159
[114] R. Cheng, M. W. Daniels, J.-G. Zhu, and D. Xiao. “Ultrafast switching of
antiferromagnets via spin-transfer torque”. Phys. Rev. B 91, 064423 (2015),
[arXiv:1503.00076].

[115] R. Cheng, D. Xiao, and A. Brataas. “Terahertz Antiferromagnetic Spin Hall

Nano-Oscillator”. Phys. Rev. Lett. 116, 207603 (2016), [arXiv:1509.09229].

[116] N. Bhattacharjee, et. al. “Néel Spin-Orbit Torque Driven Antiferromagnetic

Resonance in M n2 Au Probed by Time-Domain THz Spectroscopy”. Phys. Rev.
Lett. 120, 237201 (2018), [arXiv:1802.03199].

[117] A. R. Mellnik, et. al. “Spin-transfer torque generated by a topological insulator”.

Nature 511, 449 (2014), [arXiv:1402.1124].

[118] C.-F. Pai, et. al. “Spin transfer torque devices utilizing the giant spin Hall effect
of tungsten”. Appl. Phys. Lett. 101, 122404 (2012).

[119] Y. Fan, et. al. “Magnetization switching through giant spinorbit torque in a
magnetically doped topological insulator heterostructure”. Nat. Mater. 13, 699
(2014).

[120] Y. Wang, et. al. “Topological Surface States Originated Spin-Orbit Torques in
Bi2 Se3 ”. Phys. Rev. Lett. 114, 257202 (2015), [arXiv:1505.07937].

[121] Y. Fan, et. al. “Electric-field control of spinorbit torque in a magnetically doped
topological insulator”. Nat. Nanotechnol. 11, 352 (2016), [arXiv:1511.07442].

[122] Y. Shiomi, et. al. “Spin-Electricity Conversion Induced by Spin Injection into
Topological Insulators”. Phys. Rev. Lett. 113, 196601 (2014).

[123] M. Jamali, et. al. “Giant Spin Pumping and Inverse Spin Hall Effect in the Pres-
ence of Surface and Bulk Spin-Orbit Coupling of Topological Insulator Bi2 Se3 ”.
Nano Lett. 15, 7126 (2015), [arXiv:1407.7940].

[124] J.-C. Rojas-Sánchez, et. al. “Spin to Charge Conversion at Room Temperature
by Spin Pumping into a New Type of Topological Insulator: α-Sn Films”. Phys.
Rev. Lett. 116, 096602 (2016), [arXiv:1509.02973].

[125] Y. Fan and K. L. Wang. “Spintronics Based on Topological Insulators”. SPIN

06, 1640001 (2016).
 160
[126] M. DC, et. al. “Room-temperature perpendicular magnetization switching
through giant spin-orbit torque from sputtered BixSe(1-x) topological insulator
material”. Unpublished, [arXiv:1703.03822].

[127] J. Han, et. al. “Room-Temperature Spin-Orbit Torque Switching Induced by

a Topological Insulator”. Phys. Rev. Lett. 119, 077702 (2017), [arXiv:1703.
07470].

[128] Y. Wang, et. al. “Room temperature magnetization switching in topological

insulator-ferromagnet heterostructures by spin-orbit torques”. Nat. Commun.
8, 1364 (2017), [arXiv:1709.02159].

[129] T. Yokoyama, J. Zang, and N. Nagaosa. “Theoretical study of the dynamics

of magnetization on the topological surface”. Phys. Rev. B 81, 241410 (2010),
[arXiv:1003.3769].

[130] T. Yokoyama. “Current-induced magnetization reversal on the surface of a

topological insulator”. Phys. Rev. B 84, 113407 (2011), [arXiv:1107.0116].

[131] A. Sakai and H. Kohno. “Spin torques and charge transport on the surface of
topological insulator”. Phys. Rev. B 89, 165307 (2014).

[132] M. H. Fischer, A. Vaezi, A. Manchon, and E.-A. Kim. “Spin-torque generation

in topological insulator based heterostructures”. Phys. Rev. B 93, 125303 (2016),
[arXiv:1305.1328].

[133] P. B. Ndiaye, et. al. “Dirac spin-orbit torques and charge pumping at the surface
of topological insulators”. Phys. Rev. B 96, 014408 (2017), [arXiv:1509.
06929].

[134] L. Fu. “Hexagonal Warping Effects in the Surface States of Topological Insulator
Bi2 Te3 ”. Phys. Rev. Lett. 103, 266801 (2009), [arXiv:0908.1418].

[135] W.-Y. Shan, H.-Z. Lu, and S.-Q. Shen. “Effective continuous model for surface
states and thin films of three dimensional topological insulators”. New J. Phys.
12, 043048 (2010), [arXiv:1001.0526].

[136] C.-X. Liu, et. al. “Model Hamiltonian for topological insulators”. Phys. Rev. B
82, 045122 (2010), [arXiv:1005.1682].
 161
[137] J. Zhang, J. P. Velev, X. Dang, and E. Y. Tsymbal. “Band structure and spin
texture of Bi2 Se3 3d ferromagnetic metal interface”. Phys. Rev. B 94, 014435
(2016), [arXiv:1606.00763].

[138] D. J. J. Marchand and M. Franz. “Lattice model for the surface states of a
topological insulator with applications to magnetic and exciton instabilities”.
Phys. Rev. B 86, 155146 (2012), [arXiv:1209.4055].

[139] F. Mahfouzi, B. K. Nikolić, and N. Kioussis. “Antidamping spin-orbit torque

driven by spin-flip reflection mechanism on the surface of a topological insulator:
A time-dependent nonequilibrium Green function approach”. Phys. Rev. B 93,
115419 (2016), [arXiv:1506.01303].

[140] S. Ghosh and A. Manchon. “Spin-orbit torque in a three-dimensional topolog-

ical insulatorferromagnet heterostructure: Crossover between bulk and surface
transport”. Phys. Rev. B 97, 134402 (2018), [arXiv:1711.11016].

[141] M. DC, et. al. “Room-temperature high spinorbit torque due to quantum con-
finement in sputtered BixSe(1x) films”. Nat. Mater. (2018).

[142] A. H. MacDonald and M. Tsoi. “Antiferromagnetic metal spintronics”. Philos.

Trans. R. Soc. A Math. Phys. Eng. Sci. 369, 3098 (2011), [arXiv:0510797].

[143] E. V. Gomonay and V. M. Loktev. “Spintronics of antiferromagnetic systems

(Review Article)”. Low Temp. Phys. 40, 17 (2014).

[144] X. Marti, et. al. “Room-temperature antiferromagnetic memory resistor”. Nat.

Mater. 13, 367 (2014).

[145] V. Baltz, et. al. “Antiferromagnetic spintronics”. Rev. Mod. Phys. 90, 015005
(2018), [arXiv:1606.04284].

[146] T. Jungwirth, X. Marti, P. Wadley, and J. Wunderlich. “Antiferromagnetic

spintronics”. Nat. Nanotechnol. 11, 231 (2016).

[147] T. Jungwirth, et. al. “Focused issue on antiferromagnetic spintronics: An

overview (Part of a collection of reviews on antiferromagnetic spintronics)”.
Unpublished, [arXiv:1705.10489].

[148] O. Gomonay, T. Jungwirth, and J. Sinova. “Concepts of antiferromagnetic

spintronics”. Phys. status solidi - Rapid Res. Lett. 11, 1700022 (2017), [arXiv:
1701.06556].
 162
[149] A. S. Núñez, R. A. Duine, P. Haney, and A. H. MacDonald. “Theory of spin
torques and giant magnetoresistance in antiferromagnetic metals”. Phys. Rev.
B 73, 214426 (2006).

[150] J. Železný, et. al. “Relativistic Néel-Order Fields Induced by Electrical Current
in Antiferromagnets”. Phys. Rev. Lett. 113, 157201 (2014), [arXiv:1410.
8296].

[151] P. Wadley, et. al. “Electrical switching of an antiferromagnet”. Science (80-.

). 351, 587 (2016), [arXiv:1503.03765].

[152] Y.-W. Oh, et. al. “Field-free switching of perpendicular magnetization through
spinorbit torque in antiferromagnet/ferromagnet/oxide structures”. Nat. Nan-
otechnol. 11, 878 (2016).

[153] K. Olejnı́k, et. al. “Terahertz electrical writing speed in an antiferromagnetic

memory”. Sci. Adv. 4, eaar3566 (2018), [arXiv:1711.08444].

[154] X.-L. Li, X. Duan, Y. G. Semenov, and K. W. Kim. “Electrical switching of

antiferromagnets via strongly spin-orbit coupled materials”. J. Appl. Phys. 121,
023907 (2017).

[155] N. Thielemann-Kühn, et. al. “Ultrafast and Energy-Efficient Quenching of Spin

Order: Antiferromagnetism Beats Ferromagnetism”. Phys. Rev. Lett. 119,
197202 (2017), [arXiv:1703.03689].

[156] A. Manchon, et. al. “Current-induced spin-orbit torques in ferromagnetic and

antiferromagnetic systems”. Unpublished, [arXiv:1801.09636].

[157] L. Šmejkal, Y. Mokrousov, B. Yan, and A. H. MacDonald. “Topological anti-

ferromagnetic spintronics: Part of a collection of reviews on antiferromagnetic
spintronics”. Unpublished, [arXiv:1706.00670].

[158] L. Šmejkal, T. Jungwirth, and J. Sinova. “Route towards Dirac and Weyl anti-
ferromagnetic spintronics”. Phys. status solidi - Rapid Res. Lett. 11, 1700044
(2017), [arXiv:1702.07788].

[159] R. S. K. Mong, A. M. Essin, and J. E. Moore. “Antiferromagnetic topological

insulators”. Phys. Rev. B 81, 245209 (2010), [arXiv:1004.1403].
 163
[160] R. A. Müller, et. al. “Magnetic structure of GdBiPt: A candidate antifer-
romagnetic topological insulator”. Phys. Rev. B 90, 041109 (2014), [arXiv:
1406.6663].

[161] Z. Li, H. Su, X. Yang, and J. Zhang. “Electronic structure of the antiferro-
magnetic topological insulator candidate GdBiPt”. Phys. Rev. B 91, 235128
(2015).

[162] P. Baireuther, J. M. Edge, I. C. Fulga, C. W. J. Beenakker, and J. Tworzydlo.

“Quantum phase transitions of a disordered antiferromagnetic topological insu-
lator”. Phys. Rev. B 89, 035410 (2014), [arXiv:1309.5846].

[163] S. Ghosh and A. Manchon. “Spin-orbit torque in two-dimensional antiferro-

magnetic topological insulators”. Phys. Rev. B 95, 035422 (2017), [arXiv:
1609.01174].

[164] S. Pancharatnam. “Generalized theory of interference, and its applications, Part

I. Coherent Pencils”. Proc. Indian Acad. Sci. - Sect. A 44, 247 (1956).

[165] S. Pancharatnam. “Generalized thoery of interference and its applications, Part

II. Partially Coherent Pencils”. Proc. Indian Acad. Sci. - Sect. A 44, 398
(1956).

[166] M. Berry. “Pancharatnam, virtuoso of the Poincaré sphere: an appreciation”.

Curr. Sci. 67, 220 (1994).

[167] R. Resta. “Theory of the electric polarization in crystals”. Ferroelectrics 136,

51 (1992).

[168] R. D. King-Smith and D. Vanderbilt. “Theory of polarization of crystalline

solids”. Phys. Rev. B 47, 1651 (1993).

[169] D. Xiao, M.-C. Chang, and Q. Niu. “Berry phase effects on electronic proper-
ties”. Rev. Mod. Phys. 82, 1959 (2010), [arXiv:0907.2021].

[170] G. D. Mahan. “Many-Particle Physics”. Springer, Boston, MA, 3rd Edition,

(2000).

[171] A. Bastin, C. Lewiner, O. Betbeder-matibet, and P. Nozieres. “Quantum oscil-

lations of the hall effect of a fermion gas with random impurity scattering”. J.
Phys. Chem. Solids 32, 1811 (1971).
 164
[172] L. Smrcka and P. Streda. “Transport coefficients in strong magnetic fields”. J.
Phys. C Solid State Phys. 10, 2153 (1977).

[173] B. K. Nikolic, et. al. “First-principles quantum transport modeling of spin-

transfer and spin-orbit torques in magnetic multilayers”. Unpublished, [arXiv:
1801.05793].

[174] R. Peierls. “Zur Theorie des Diamagnetismus von Leitungselektronen”.

Zeitschrift fur Phys. 80, 763 (1933).

[175] G. H. Wannier. “Dynamics of Band Electrons in Electric and Magnetic Fields”.

Rev. Mod. Phys. 34, 645 (1962).

[176] P. R. Wallace. “The Band Theory of Graphite”. Phys. Rev. 71, 622 (1947).

[177] G. W. Semenoff. “Condensed-Matter Simulation of a Three-Dimensional

Anomaly”. Phys. Rev. Lett. 53, 2449 (1984).

[178] Z. Qiao, H. Jiang, X. Li, Y. Yao, and Q. Niu. “Microscopic theory of quantum
anomalous Hall effect in graphene”. Phys. Rev. B 85, 115439 (2012), [arXiv:
1201.0543].

[179] M. Ezawa. “Valley-Polarized Metals and Quantum Anomalous Hall Effect in

Silicene”. Phys. Rev. Lett. 109, 055502 (2012), [arXiv:1203.0705].

[180] C.-C. Liu, H. Jiang, and Y. Yao. “Low-energy effective Hamiltonian involving
spin-orbit coupling in silicene and two-dimensional germanium and tin”. Phys.
Rev. B 84, 195430 (2011), [arXiv:1108.2933].

[181] M. Ezawa. “Monolayer Topological Insulators: Silicene, Germanene, and

Stanene”. J. Phys. Soc. Japan 84, 121003 (2015), [arXiv:1503.08914].

[182] S. Murakami, N. Nagaosa, and S. C. Zhang. “Dissipationless quantum spin

current at room temperature”. Science (80-. ). 301, 1348 (2003), [arXiv:
0308167].

[183] S. Ghosh and U. Schwingenschlögl. “Hexagonal graphene quantum dots”. Phys.

status solidi - Rapid Res. Lett. 11, 1600226 (2017).

[184] J. A. Lock. “The Zitterbewegung of a free localized Dirac particle”. Am. J.

Phys. 47, 797 (1979).
 165
[185] J. Cserti and G. Dávid. “Unified description of Zitterbewegung for spintronic,
graphene, and superconducting systems”. Phys. Rev. B 74, 172305 (2006).

[186] L. J. LeBlanc, et. al. “Direct observation of zitterbewegung in a BoseEinstein

condensate”. New J. Phys. 15, 073011 (2013).

[187] R. Gerritsma, et. al. “Quantum simulation of the Dirac equation”. Nature 463,
68 (2010).

[188] L. Brey and H. A. Fertig. “Electronic states of graphene nanoribbons studied

with the Dirac equation”. Phys. Rev. B 73, 235411 (2006).

[189] K. Wakabayashi, K.-i. Sasaki, T. Nakanishi, and T. Enoki. “Electronic states

of graphene nanoribbons and analytical solutions”. Sci. Technol. Adv. Mater.
11, 054504 (2010).

[190] K. Nakada, M. Fujita, G. Dresselhaus, and M. S. Dresselhaus. “Edge state in

graphene ribbons: Nanometer size effect and edge shape dependence”. Phys.
Rev. B 54, 17954 (1996).

[191] Y.-W. Son, M. L. Cohen, and S. G. Louie. “Half-metallic graphene nanorib-

bons”. Nature 444, 347 (2006).

[192] A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov, and A. K. Geim.

“The electronic properties of graphene”. Rev. Mod. Phys. 81, 109 (2009).

[193] T. M. Rusin and W. Zawadzki. “Transient Zitterbewegung of charge carriers in

mono- and bilayer graphene, and carbon nanotubes”. Phys. Rev. B 76, 195439
(2007).

[194] J. Schliemann, D. Loss, and R. M. Westervelt. “Zitterbewegung of Electronic

Wave Packets in III-V Zinc-Blende Semiconductor Quantum Wells”. Phys.
Rev. Lett. 94, 206801 (2005).

[195] T. M. Rusin and W. Zawadzki. “Zitterbewegung of electrons in graphene in a

magnetic field”. Phys. Rev. B 78, 125419 (2008).

[196] S. Ghosh, U. Schwingenschlögl, and A. Manchon. “Resonant longitudinal Zit-

terbewegung in zigzag graphene nanoribbons”. Phys. Rev. B 91, 045409 (2015).
 166
[197] D. Sun, et. al. “Ultrafast Relaxation of Excited Dirac Fermions in Epitaxial
Graphene Using Optical Differential Transmission Spectroscopy”. Phys. Rev.
Lett. 101, 157402 (2008).

[198] S. Winnerl, et. al. “Carrier Relaxation in Epitaxial Graphene Photoexcited Near
the Dirac Point”. Phys. Rev. Lett. 107, 237401 (2011).

[199] S. Ulstrup, et. al. “Ultrafast Dynamics of Massive Dirac Fermions in Bilayer
Graphene”. Phys. Rev. Lett. 112, 257401 (2014).

[200] I. Gierz, et. al. “Phonon-Pump Extreme-Ultraviolet-Photoemission Probe in

Graphene: Anomalous Heating of Dirac Carriers by Lattice Deformation”.
Phys. Rev. Lett. 114, 125503 (2015).

[201] B. A. Bernevig and S.-C. Zhang. “Quantum Spin Hall Effect”. Phys. Rev. Lett.
96, 106802 (2006), [arXiv:0504147].

[202] X. Dang, J. D. Burton, A. Kalitsov, J. P. Velev, and E. Y. Tsymbal. “Complex

band structure of topologically protected edge states”. Phys. Rev. B - Condens.
Matter Mater. Phys. 90, 1 (2014).

[203] C. W. Groth, M. Wimmer, A. R. Akhmerov, and X. Waintal. “Kwant: a

software package for quantum transport”. New J. Phys. 16, 063065 (2014),
[arXiv:1309.2926].

[204] D. S. Fisher and P. A. Lee. “Relation between conductivity and transmission

matrix”. Phys. Rev. B 23, 6851 (1981).

[205] H. V. Gomonay and V. M. Loktev. “Spin transfer and current-induced switching

in antiferromagnets”. Phys. Rev. B 81, 144427 (2010), [arXiv:0909.0234].

[206] J. Železný, et. al. “Spin-orbit torques in locally and globally noncentrosymmetric
crystals: Antiferromagnets and ferromagnets”. Phys. Rev. B 95, 014403 (2017),
[arXiv:1604.07590].

[207] S. Zhang, P. M. Levy, and A. Fert. “Mechanisms of Spin-Polarized Current-

Driven Magnetization Switching”. Phys. Rev. Lett. 88, 236601 (2002), [arXiv:
0202363].

[208] S. Zhang and Z. Li. “Roles of Nonequilibrium Conduction Electrons on the

Magnetization Dynamics of Ferromagnets”. Phys. Rev. Lett. 93, 127204 (2004),
[arXiv:0407174].
 167
[209] A. Brataas, A. D. Kent, and H. Ohno. “Current-induced torques in magnetic
materials”. Nat. Mater. 11, 372 (2012).

[210] A. D. Kent and D. C. Worledge. “A new spin on magnetic memories”. Nat.

Nanotechnol. 10, 187 (2015).

[211] B. A. Bernevig and O. Vafek. “Piezo-magnetoelectric effects in p -doped semi-

conductors”. Phys. Rev. B 72, 033203 (2005), [arXiv:0408476].

[212] K. Garello, et. al. “Ultrafast magnetization switching by spin-orbit torques”.

Appl. Phys. Lett. 105, 212402 (2014), [arXiv:1310.5586].

[213] I. M. Miron, et. al. “Fast current-induced domain-wall motion controlled by the
Rashba effect”. Nat. Mater. 10, 419 (2011), [arXiv:1302.2257].

[214] S. Emori, U. Bauer, S.-M. Ahn, E. Martinez, and G. S. D. Beach. “Current-

driven dynamics of chiral ferromagnetic domain walls”. Nat. Mater. 12, 611
(2013), [arXiv:1302.2257].

[215] S.-H. Yang, K.-S. Ryu, and S. Parkin. “Domain-wall velocities of up to 750
ms−1 driven by exchange-coupling torque in synthetic antiferromagnets”. Nat.
Nanotechnol. 10, 221 (2015).

[216] L. Liu, O. J. Lee, T. J. Gudmundsen, D. C. Ralph, and R. A. Buhrman.

“Current-Induced Switching of Perpendicularly Magnetized Magnetic Layers Us-
ing Spin Torque from the Spin Hall Effect”. Phys. Rev. Lett. 109, 96602 (2012),
[arXiv:1110.6846].

[217] V. E. Demidov, et. al. “Magnetic nano-oscillator driven by pure spin current”.
Nat. Mater. 11, 1028 (2012).

[218] S. Fukami, T. Anekawa, C. Zhang, and H. Ohno. “A spin-orbit torque switch-

ing scheme with collinear magnetic easy axis and current configuration”. Nat.
Nanotechnol. 11, 621 (2016).

[219] A. van den Brink, et. al. “Field-free magnetization reversal by spin-Hall effect
and exchange bias”. Nat. Commun. 7, 10854 (2016), [arXiv:1509.08752].

[220] X. Qiu, et. al. “Enhanced Spin-Orbit Torque via Modulation of Spin Current
Absorption”. Phys. Rev. Lett. 117, 217206 (2016), [arXiv:1610.06989].
 168
[221] X. Qiu, et. al. “Angular and temperature dependence of current induced spin-
orbit effective fields in Ta/CoFeB/MgO nanowires”. Sci. Rep. 4, 4491 (2015),
[arXiv:1404.1130].

[222] E. L. Ivchenko and G. E. Pikus. “New photogalvanic effect in gyrotropic crys-

tals”. JETP Lett. 27, 604 (1978).

[223] F. Freimuth, S. Blügel, and Y. Mokrousov. “Spin-orbit torques in Co/Pt(111)

and Mn/W(001) magnetic bilayers from first principles”. Phys. Rev. B 90,
174423 (2014).

[224] H. Kurebayashi, et. al. “An antidamping spinorbit torque originating from the
Berry curvature”. Nat. Nanotechnol. 9, 211 (2014), [arXiv:1306.1893].

[225] A. Manchon. “Spinorbitronics: A new moment for Berry”. Nat. Phys. 10, 340
(2014).

[226] M. B. Lifshits and M. I. Dyakonov. “Swapping Spin Currents: Interchanging

Spin and Flow Directions”. Phys. Rev. Lett. 103, 186601 (2009), [arXiv:
0905.4469].

[227] H. B. M. Saidaoui and A. Manchon. “Spin-Swapping Transport and Torques

in Ultrathin Magnetic Bilayers”. Phys. Rev. Lett. 117, 36601 (2016), [arXiv:
1511.03454].

[228] K. Yasuda, et. al. “Current-Nonlinear Hall Effect and Spin-Orbit Torque Mag-
netization Switching in a Magnetic Topological Insulator”. Phys. Rev. Lett.
119, 137204 (2017).

[229] I. Garate and M. Franz. “Inverse Spin-Galvanic Effect in the Interface between
a Topological Insulator and a Ferromagnet”. Phys. Rev. Lett. 104, 146802
(2010), [arXiv:0911.0106].

[230] C. S. Ho, et. al. “Effect of surface state hybridization on current-induced spin-
orbit torque in thin topological insulator films”. Sci. Rep. 7, 792 (2017), [arXiv:
1611.08116].

[231] F. Mahfouzi, N. Nagaosa, and B. K. Nikolić. “Spin-to-charge conversion

in lateral and vertical topological-insulator/ferromagnet heterostructures with
microwave-driven precessing magnetization”. Phys. Rev. B 90, 115432 (2014),
[arXiv:1312.7091].
 169
[232] X. Peng, Y. Yang, R. R. P. Singh, S. Y. Savrasov, and D. Yu. “Spin gener-
ation via bulk spin current in three-dimensional topological insulators”. Nat.
Commun. 7, 10878 (2016).

[233] J. M. Marmolejo-Tejada, et. al. “Proximity Band Structure and Spin Tex-
tures on Both Sides of Topological-Insulator/Ferromagnetic-Metal Interface
and Their Charge Transport Probes”. Nano Lett. 17, 5626 (2017), [arXiv:
1701.00462].

[234] Y.-T. Hsu, K. Park, and E.-A. Kim. “Hybridization-induced interface states in
a topological-insulatorferromagnetic-metal heterostructure”. Phys. Rev. B 96,
235433 (2017), [arXiv:1707.06319].

[235] P. Hosur, S. Ryu, and A. Vishwanath. “Chiral topological insulators, supercon-

ductors, and other competing orders in three dimensions”. Phys. Rev. B 81,
45120 (2010).

[236] S. Wimmer, K. Chadova, M. Seemann, D. Ködderitzsch, and H. Ebert. “Fully

relativistic description of spin-orbit torques by means of linear response theory”.
Phys. Rev. B 94, 54415 (2016), [arXiv:1604.02798].

[237] A. Dyrdal and J. Barnaś. “Current-induced spin polarization and spin-orbit

torque in graphene”. Phys. Rev. B 92, 165404 (2015), [arXiv:1505.02530v1].

[238] N. A. Sinitsyn, A. H. MacDonald, T. Jungwirth, V. K. Dugaev, and J. Sinova.

“Anomalous Hall effect in a two-dimensional Dirac band: The link between
the Kubo-Streda formula and the semiclassical Boltzmann equation approach”.
Phys. Rev. B 75, 45315 (2007), [arXiv:0608682].

[239] L. Isaev, Y. H. Moon, and G. Ortiz. “Bulk-boundary correspondence in three-

dimensional topological insulators”. Phys. Rev. B 84, 75444 (2011), [arXiv:
1103.0025].

[240] T. Tanaka, et. al. “Intrinsic spin Hall effect and orbital Hall effect in 4d and
5d transition metals”. Phys. Rev. B 77, 165117 (2008).

[241] C. Sahin and M. E. Flatté. “Tunable Giant Spin Hall Conductivities in a

Strong Spin-Orbit Semimetal: Bi1−x Sbx ”. Phys. Rev. Lett. 114, 107201 (2015),
[arXiv:1410.7319].
 170
[242] P.-H. Chang, T. Markussen, S. Smidstrup, K. Stokbro, and B. K. Nikolić.
“Nonequilibrium spin texture within a thin layer below the surface of current-
carrying topological insulator Bi2 Se3 : A first-principles quantum transport
study”. Phys. Rev. B 92, 201406 (2015), [arXiv:1503.08046].

[243] S.-W. Lee and K.-J. Lee. “Emerging Three-Terminal Magnetic Memory De-
vices”. Proc. IEEE 104, 1831 (2016).

[244] N. Sato, F. Xue, R. M. White, C. Bi, and S. X. Wang. “Two-terminal spinorbit

torque magnetoresistive random access memory”. Nat. Electron. 1, 508 (2018).

[245] J. Nogués and I. K. Schuller. “Exchange bias”. J. Magn. Magn. Mater. 192,
203 (1999).

[246] A. Manchon. “Spin diffusion and torques in disordered antiferromagnets”. J.

Phys. Condens. Matter 29, 104002 (2017).

[247] H. Watanabe and Y. Yanase. “Symmetry analysis of current-induced switching

of antiferromagnets”. Phys. Rev. B 98, 220412 (2018), [arXiv:1812.11277].

[248] S. Y. Bodnar, et. al. “Writing and reading antiferromagnetic Mn2Au by Néel
spin-orbit torques and large anisotropic magnetoresistance”. Nat. Commun. 9,
348 (2018), [arXiv:1706.02482].

[249] M. Meinert, D. Graulich, and T. Matalla-Wagner. “Electrical Switching of

Antiferromagnetic M n2 Au and the Role of Thermal Activation”. Phys. Rev.
Appl. 9, 064040 (2018), [arXiv:1706.06983].

[250] X. Z. Chen, et. al. “Antidamping-Torque-Induced Switching in Biaxial An-

tiferromagnetic Insulators”. Phys. Rev. Lett. 120, 207204 (2018), [arXiv:
1804.05462].

[251] T. Moriyama, W. Zhou, T. Seki, K. Takanashi, and T. Ono. “Spin-Orbit-

Torque Memory Operation of Synthetic Antiferromagnets”. Phys. Rev. Lett.
121, 167202 (2018).

[252] P. E. Roy, R. M. Otxoa, and J. Wunderlich. “Robust picosecond writing of a

layered antiferromagnet by staggered spin-orbit fields”. Phys. Rev. B 94, 014439
(2016), [arXiv:1604.05918].
 171
[253] Q. L. He, et. al. “Tailoring exchange couplings in magnetic topological-
insulator/antiferromagnet heterostructures”. Nat. Mater. 16, 94 (2017),
[arXiv:1605.04854].

[254] Q. L. He, et. al. “Topological Transitions Induced by Antiferromagnetism in a

Thin-Film Topological Insulator”. Phys. Rev. Lett. 121, 096802 (2018), [arXiv:
1612.01661].

[255] A. Manchon and A. Belabbes. “Spin-Orbitronics at Transition Metal Inter-

faces”. In Solid State Physics, Vol. 68, pages 1–89. (2017).

[256] D. Culcer, A. MacDonald, and Q. Niu. “Anomalous Hall effect in paramagnetic

two-dimensional systems”. Phys. Rev. B 68, 045327 (2003), [arXiv:0311147].
 172

List of publications

Published

1. S. Ghosh and A. Manchon, “Spin-orbit torque in 3D topological insulator-

ferromagnet heterostructure: crossover between bulk and surface transport”,
Phys. Rev. B 97, 134402 (2018), [arXiv:1711.11016].

2. S. Ghosh and A. Manchon, “Spin-orbit torque in two-dimensional antiferro-

magnetic topological insulators”, Phys. Rev. B 95, 035422 (2017), [arXiv:
1609.01174].

3. S. Ghosh and A. Manchon, “Signature of Topological Phases in Zitterbewegung”,

SPIN 06, 1640004 (2016), [arXiv:1605.02207].

4. S. Ghosh and U. Schwingenschlögl, “Hexagonal graphene quantum dots”, Phys.

status solidi - Rapid Res. Lett. 5, 1600226 (2016).

5. S. Ghosh, U. Schwingenschlögl and A. Manchon, “Resonant longitudinal Zitter-

bewegung in zigzag graphene nanoribbons”, Phys. Rev. B 91, 045409 (2015).

Unpublished

6. S. Ghosh and A. Manchon, “Non-equilibrium spin density and spin-orbit torque

in three dimensional topological insulators-antiferromagnet heterostructure”, [Un-
published, arXiv:1901.08314].

7. S. Ghosh and A. Manchon, “Non-equilibrium spin density in 5d transition met-

als: A first principle study”, [In preparation].
 173
8. D. Yudin, S. Ghosh, M. Titov and A. Manchon, “Symmetry of spin-orbit torques
in two-dimensional antiferromagnets on a hexagonal lattice”, [In preparation].

9. S. Laref, S. Ghosh, E. Y. Tsymbal, and A. Manchon, “Induced Spin-textures at

3d Transition Metals/Topological Insulator Interfaces”, [In preparation].

10. S. Ghosh and A. Manchon, “Signature of topological transition in persistent

current in a Dirac Ring”, [Unpublished, arXiv:1607.05052].