Professional Documents
Culture Documents
Saurav Thesis
Saurav Thesis
insulators
A Thesis
Submitted For the Degree of
Doctor of Philosophy
in the
Faculty of Science
by
Saurav Islam
Department of Physics
Indian Institute of Science
Bangalore 560 012
July 2018
Declaration
The work reported in this thesis is original and was carried out by me during my tenure
as a PhD student at the Department of Physics, Indian Institute of Science, Bangalore
under the supervision of Prof. Arindam Ghosh.
This thesis has not formed the basis for award of any degree, diploma, associateship,
membership or similar title of any university or institution.
(Saurav Islam)
January, 2014
Department of Physics,
Indian Institute of Science,
Bangalore,
India
i
To
My family
Acknowledgements
The work presented in this thesis would not have been possible without the encourage-
ment of several people. In the beginning itself, I must thank my advisor Prof Arindam
Ghosh. He is phenomenal. His dedication and hard working nature is something that we
should all learn from. He gave me freedom to work on my own, and his timely advice have
benefited me immensely. I have been extremely benefitted from the various collabora-
tions that I have been involved in. I express my sincere gratitude to Prof Suja Elizabeth,
Prof Nitin Samarth, Prof N Ravishankar, Dr N Hariharan, Dr Anthony Richardella, Dr
Jayantha Mishra, Abinash Kumar, and Dipanwita Chatterjee.
I express my sincere gratitude towards my parents, who have sacrificed a lot for my
upbringing, and have never imposed anything on me, which has made me the person I
am today, and to my family members for always being there.
I have been extremely lucky to have fabulous lab seniors who have taught me almost
everything. Foremost amongst them is Dr Semonti Bhattacharyya. She has been the
guiding light during my stay here at Indian Institute of Science. She has essentially
taught me most of the things I know, and I shall be forever grateful to her. She is
the only one aware of all my misadventures, and yet surprisingly, still talks to me. I
am indebted to Dr Saquib Shamim for a lot of help with my experiments, teaching me
how to handle the dilution refrigerator, and various other things. It has been a pleasure
watching and learning from Dr Kallol Roy. You are awe-inspiring. I have spent a lot of
time with Kallol in the lab, from learning hetero-structure fabrication to opto-electronic
measurements. Our activities were not limited to lab only, and we actively took part in
Students’ Council related activities as well. Another person who has helped me a lot is
Dr Mohammad Aamir Ali. And I thank him sincerely. Dr Paritosh Karnatak is another
person, from whom I have benefited a lot. Then there is Dr Phanindra Sai, who is a
walking encyclopedia, and has had the solution to every problem that I have ever come
across. Another person who has helped me a lot is Dr Mitali Banerjee who will be always
v
remerbered very fondly. There are other seniors like Dr R Koushik, Dr Vidya Kochat,
Dr Subhamoy Ghatak, Dr Anindita Sahoo, Dr Avradip Pradhan, Dr Ranjit Kashid, who
have taught and assisted me with a lot of things. I have to especially thank Anindita and
Avradip from proof checking the thesis, inspite of their extremely busy schedule. I have
spent a lot of time outside the lab well with Paritosh, Anindita, Tathagata, Kimberly,
Phani, Tanweer, Aditya. It has been a great pleasure to be around them. Bhaskar and
Saloni are two very enthusiastic juniors, and have been very helpful during my stay here.
I have been involved with a lot of activities in IISc, especially mess related work. I
express my sincere gratitude for the mess superintendent Bheemanna, who made my life
much easier, and took care of a lot of things. Also Ramchandra, Anothony and Hanu-
mantha for keeping food separately for me, on every occasion. The physics office staff has
been extremely helpful. Meena Mam, Srivatsa Sir, Rekha, Sumitra have always helped
me. Swati and Rajeswari has also done a fantastic job of handling all the administrative
duties of the lab.
Outside lab, my life would not be the same without Anindita Brahma, Souvik Mon-
dal, Subhradeep Mistry, Priyadarshini, Rudra, Rituparna di, and I am extremely grateful
for all the love and affection that I have received from them. I have had a lot of en-
joyable moments with Soumyabrata, Sohini, Syamantak, Somananda, Arpan, Ananya,
Chandradish, Arka, and will always cherish these memories. I have been blessed to
have some great seniors in IISc when I joined, who made me feel at home: Nirmalendu,
Debarghya, Rupamanjari, Ananyo, Arpan, Wrick, Saroj, Tirthankar, Sandeep, Shaon,
Abhiruk. It has aslo been great to have Kazi, Hemanta, Sumit, Jit, Saienten, Tanujit
around in IISc. I have known them for over a decade now, and it has been extremely
beneficial to have such old friends around. I have made a lot of friends as well here as
well. Ishanu, Manuj, Chandan, Abhishek, Manab, Aabhas, Puspander, Rakesh, Soumi,
Debasmita, Adhip, Mit, Tathagata, Arup, Prakriti Subham, Marsha, and they will be
always remembered. I thank my integrated PhD batchmates and juniors, Ranjani, Ab-
hiram, Gopinath, Suman, Sangram, Anjali for being great friends. I express my sincere
gratitude to Prof Jainendra Jain, Manju Aunty, Nitin Samarth, Sanil, Sahana for their
amazing hospitality when I visited USA. My old friends Soumangsu, Pritha, Bodhisatya,
Aniket, Shrabasti, Dibya, Shaili, Anomitra have always been there for me, no matter
what, and I shall be forever grateful to them. I shall also miss the beautiful, green cam-
pus at IISc. Finally, I thank Samarpita. My life would have been quite different without
her presence, and whose memories I shall carry with me forever.
vi
Publications
The work described in this thesis has resulted in the following publications:
• Published manuscripts
vii
Arindam Ghosh; Sonu Gandhi, and Arindam Ghosh
Sci. Rep., Volume 9, Article number: 276 (2019)
• Book chapter:
viii
1. Title: 2D Inorganic Materials beyond Graphene
Chapter: 2D van der Waals Hybrid: Structures, properties and devices
Publisher: World Scientific (2017)
Authors: M. A. Aamir, T. Ahmed, K. Hsieh, S. Islam, P. Karnatak, R. Kashid, P.
S. Mahapatra, J. Mishra, T. Paul, A. Pradhan, K. Roy, A. Sahoo, and A. Ghosh
• Conference Proceeding:
• Under preparation:
2. Dynamic tuning of the stiffness and the intensities of first-order Raman modes in
WSe2 using van der Waals hybrids
S. P. Suman, T. Ahmed, S. Islam, R. Debnath, and A. Ghosh
ix
Abstract
xi
scanning the Fermi energy, provide an unambiguous proof of the intrinsic preservation of
time reversal symmetry in topological insulators. The magnetic-field dependence of the
fluctuations clearly indicate that they originate from universal conductance fluctuations
in mesoscopic samples, and also shows a transition from the ground state symplectic to
unitary symmetry class. Quantum transport measurements in a variety of device configu-
rations, in both exfoliated and epitaxially grown samples, have revealed the saturation of
phase coherence length at low temperature, which may originate due to electron-magnetic
fluctuations in the bulk of the TIs. In the second part, we focus on opto-electronic in-
vestigations, and demonstrate that TIs, when integrated with graphene, can lead to
extremely high detectivity, and can be used as sensors in the near infra-red regime with
a resolution of ∼ 10 photons.
xii
Contents
Declaration i
Acknowledgements v
Publications vii
Abstract xi
1 Introduction 1
1.1 Basics of topological insulators1 . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 What is topology? . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.2 Berry phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.3 Topological invariant, Chern number and TKNN invariant . . . . 4
1.1.4 Bulk-boundary correspondence . . . . . . . . . . . . . . . . . . . 5
1.1.5 Key experimental discoveries . . . . . . . . . . . . . . . . . . . . 7
1.1.5.1 Integer quantum Hall effect . . . . . . . . . . . . . . . . 8
1.1.5.2 Quantum spin Hall effect . . . . . . . . . . . . . . . . . 8
1.2 3D topological insulator . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.2.1 Z2 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.2.2 Methods for experimentally confirming TI materials . . . . . . . . 12
1.2.3 Properties of surface states . . . . . . . . . . . . . . . . . . . . . . 13
1.2.3.1 Topological protection of surface states . . . . . . . . . . 14
1.2.4 3D Topological insulator materials . . . . . . . . . . . . . . . . . 14
1.2.5 Synthesis of TI Materials . . . . . . . . . . . . . . . . . . . . . . . 19
1.2.5.1 Bulk single crystals . . . . . . . . . . . . . . . . . . . . . 19
1
Discussions of this part inspired by Ref [1, 2]
xiii
Contents
2
Inspired by Ref [3]
xiv
Contents
1.7.5.2 Photo-responsivity . . . . . . . . . . . . . . . . . . . . . 55
1.7.5.3 Noise Equivalent Power . . . . . . . . . . . . . . . . . . 55
1.7.5.4 Specific detectivity . . . . . . . . . . . . . . . . . . . . . 56
1.8 Motivation for the present work . . . . . . . . . . . . . . . . . . . . . . . 56
1.9 Thesis layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2 Experimental Techniques 61
2.1 Material properties and device fabrication . . . . . . . . . . . . . . . . . 61
2.1.1 Bulk single crystals of topological insulators . . . . . . . . . . . . 61
2.1.1.1 Growth procedures . . . . . . . . . . . . . . . . . . . . . 61
2.1.1.2 Structural and compositional analysis . . . . . . . . . . 63
2.1.2 Epitaxially grown topological insulators . . . . . . . . . . . . . . . 64
2.1.2.1 Growth procedures . . . . . . . . . . . . . . . . . . . . . 64
2.1.2.2 Structural and compositional analysis . . . . . . . . . . 64
2.1.3 Nanowires of bismuth telluride . . . . . . . . . . . . . . . . . . . . 65
2.1.3.1 Growth procedures . . . . . . . . . . . . . . . . . . . . . 65
2.1.3.2 Structural and compositional analysis . . . . . . . . . . 67
2.2 Fabrication of field effect transistors (FETs) . . . . . . . . . . . . . . . . 67
2.2.1 Characterization techniques . . . . . . . . . . . . . . . . . . . . . 70
2.2.1.1 Optical microscopy . . . . . . . . . . . . . . . . . . . . . 70
2.2.1.2 Atomic force microscopy . . . . . . . . . . . . . . . . . . 70
2.2.1.3 Raman spectroscopy . . . . . . . . . . . . . . . . . . . . 70
2.2.2 Hetero-structure assembly . . . . . . . . . . . . . . . . . . . . . . 73
2.2.2.1 Electron beam lithography . . . . . . . . . . . . . . . . . 76
2.2.2.2 Metallization . . . . . . . . . . . . . . . . . . . . . . . . 77
2.2.2.3 PMMA protection . . . . . . . . . . . . . . . . . . . . . 77
2.2.2.4 Packaging and bonding . . . . . . . . . . . . . . . . . . . 77
2.2.2.5 Fabrication of FET of epitaxially grown sample . . . . . 79
2.2.2.6 Fabrication of graphene-bismuth telluride hybrids for opto-
electronics response . . . . . . . . . . . . . . . . . . . . . 79
2.3 Measurement setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
2.3.1 Variable temperature cryostat . . . . . . . . . . . . . . . . . . . . 80
2.3.2 He refrigerator . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3
81
2.3.3 He3 -He4 dilution refrigerator . . . . . . . . . . . . . . . . . . . . . 81
xv
Contents
xvi
Contents
xvii
Contents
xviii
Contents
xix
Chapter 1
Introduction
The discovery of graphene in 2004, a single layer of carbon atoms, has led to the extrac-
tion of several other van der Waal’s materials such as transition metal dichalcogenides,
boron nitride, and bismuth-based topological insulators. Topological insulators (TIs)
are a class of materials exhibiting gapless linear surface states in the bulk band gap, due
to non-trivial topology of the bulk bands. These surface states have various interest-
ing properties like spin-momentum locking, protection against time reversal symmetry
breaking from non-magnetic impurities, π Berry phase and exhibit a plethora of fun-
damentally important properties and applications. In this chapter, we delve towards
understanding the fundamental properties of TI, and also review different aspects re-
lated to quantum transport measurements in TIs, which shall be essential to understand
the outcome of the investigations performed in the subsequent chapters.
1
Discussions of this part inspired by Ref [1, 2]
1
Chapter 1. Introduction
ˆ
κdA = 2(2 − g) (1.1)
where κ and A are the radius of curvature and area of a surface respectively. In case of
2D surfaces, the genus denotes the number of perforations through the surface. For an
ellipsoid or sphere, g = 0, while for a donut or coffee mug g = 1. In condensed matter
systems, the Brillouin zone corresponds to the geometrical surface, and Berry curvature
is analogous to radius of curvature. The Chern number is an integral of the Berry
curvature, and is used to define the topological invariant corresponding to a particular
class of systems.
2
1.1. Basics of topological insulators2
If the parameter R changes adiabatically from 0 to R0 , the time evolution of the eigen-
state is given by,
∂
H[R(t)]|n, ti = i~ |n, ti (1.3)
∂t
and |n, ti after a time t is given by,
ˆ t
i
|n, ti = exp 0 0
dt Ln [R(t )] |n, R(t)i (1.4)
~ 0
where Ln [R(t)] = i~Ṙ(t) · hn, R(t)|∇R |n, R(t)i − En [R(t)]. Using the expression
for Ln [R(t)], the time-dependence of the state can be expressed as,
ˆ t ˆ
i t
|n, ti = exp 0
En [R(t )] × exp − Ṙ(t ) · hn, R(t )|∇R |n, R(t )i |n, R(t)i (1.5)
0 0 0
0 ~ 0
The first exponential factor is the trivial term called the dynamical term, while the
second term is a quantum mechanical effect, and corresponds to the phase accumulated
during the time evolution. When R(t) traverses in a closed loop between t = 0 to t = T ,
the Berry phase can be defined as,
ˆ t
γn [C] = dt0 Ṙ(t0 ) · ihn · R(t)|∇R |n · R(t)i (1.6)
0
˛
= dR · ihn · R|∇R |n · Ri (1.7)
C
3
Chapter 1. Introduction
˛
=− dR · An (R) (1.8)
C
The Berry phase has three main properties [5]. Firstly, it is gauge invariant. Since the
eigen-function is defined by a set of homogeneous linear equations, one can multiply an
overall parameter-dependent phase factor. This makes the Berry phase physical, and the
initial studies were focused on measuring it directly through interference phenomena.
Second, the Berry phase is geometrical, which makes it possible to express the Berry
phase in terms of local geometrical quantities in the parameter space. Third, the Berry
phase has close analogies to gauge field theories and differential geometry [6]. Berry
phase is like the Aharonov-Bohm phase of a charged particle traveling a loop including
a magnetic flux, while the Berry curvature is like the magnetic field. The integral of the
Berry curvature over closed surfaces, such as a sphere or torus, is known to be topological
and quantized as integral Chern numbers.
The key difference between the quantum Hall state and an ordinary insulator is related
to topology, as was first explained by Thouless, Kohmoto, Nightingale, and den Nijs
(TKNN) in 1982 [7]. A 2D band-structure consists of mapping the crystal momentum
to the Bloch Hamiltonian. When the band structure is gapped, it can be grouped
into equivalent classes according to topology, and can be distinguished by a topological
invariant, n ∈ Z, referred to as the Chern number. The topological invariant relevant
for the quantum Hall effect, which was probably the first topological insulator, is the
4
1.1. Basics of topological insulators3
TKNN invariant. It can be shown that, for a periodic potential with Bloch states |unk i,
the Hall conductivity reduces to,
e2
σxy = ν (1.12)
h
P ´ d2 k
∂an,y
where ν = − ∂a∂kn,x . Here an (k) = −ihunk |∇R |unk i and νn is
P
n νn = n BZ 2π ∂kx y
the summation over each band. We can easily see that the Hall conductivity is related
to the Berry phase, since
ˆ
d2 k
∂an,y ∂an,x
νn = − (1.13)
BZ 2π ∂kx ∂ky
˛
1
= dk · an (k) (1.14)
2π
∂BZ
1
=− γn [∂BZ] (1.15)
2π
Since the wave-function is single-valued, the change of phase, while traversing the Bril-
lioun zone can only be integer multiples of 2π. ν can only take integer values, which
makes the Hall conductivity quantized. ν is referred to as the TKNN invariant, which
plays the role of a topological invariant in integer quantum Hall effect, which is a time
reversal symmetry breaking topological insulator. It is also called the first Chern num-
ber or the winding number, and is equal to the Berry phase of the Bloch wave function
calculated around the Brillouin zone (BZ) boundary divided by 2π. The Chern num-
ber cannot change as long as the Hamiltonian is varied smoothly, which leads to robust
quantization of the Hall conductance. In case of 2D and 3D TIs, it is referred to as the
Z2 index.
5
Chapter 1. Introduction
(a) (b)
Conduction band Conduction band
𝐸𝐹 𝐸𝐹
E E
Г𝑎 k Г𝑏 Г𝑎 k Г𝑏
Figure 1.2: Electronic dispersion: (a) Schematic of the electronic dispersion between
two boundary Kramer’s degenerate points a = 0 and b = π/a. The number of surface
states crossing the Fermi energy (EF ) is even. (b) The number of surface states crossing
EF is odd.
4n = NR − NL (1.16)
where NR and NL are the right and left moving modes respectively, and 4n is the
difference of the Chern numbers across the interface. By modifying the Hamiltonian near
the surface, the dispersion of the edge states can be changed. However, the difference in
the number of right and left moving modes remain unchanged and is determined by the
topology of the bulk bands.
The schematic of the electronic states along the edge of a time reversal invariant 2D
insulator as a function of the crystal momentum is shown in Fig 1.2a-b. Only half of the
Brillouin zone a = 0 < kx < b = π/a is depicted, since time reversal symmetry requires
that the other half −π/a < kx < 0 is a mirror image. According to Kramer’s theorem,
there can be two-fold degeneracy at the Γ -invariant momenta at kx = 0 and kx = π/a (
same as −π/a). Away from these special points, the degeneracy is lifted due to spin-orbit
interaction. There are two possible ways that the states at kx = 0 and kx = π/a can be
joined.
They connect pairwise as depicted in Fig. 1.2a. In this particular scenario, the bands
intersect EF an even number of times and the edge modes can be eliminated by pushing
all of the bound states out of the gap. However, in Fig. 1.2b, the edge states cannot
6
1.1. Basics of topological insulators4
(a) (b)
g (E) ℏ𝜔𝑐 ℏ𝜔𝑐 ℏ𝜔𝑐
Ef
(c) Skipping orbits
B
Localised electrons
Figure 1.3: Quantum Hall effect: (a) Schematic depiction of Landau level forma-
tion in a two-dimensional electron gas due to application of a perpendicular magnetic
field. The difference in energy between each level is ~ωC . (b) Experimental curve show-
ing quantized Hall resistance as different Landau Levels are populated with increasing
perpendicular magnetic field. Quantized Hall plateaus are accompanied by vanishing
longitudinal resistance. (c) Schematic illustration of an insulating bulk, while the edge
modes conduct current in the quantum Hall state. (Fig. 1.3c reproduced with permission
from Ref. [8]).
be eliminated since the bands intersect EF an odd number of times. These are non-
trivial edge states. The particular scenario taking place in a material is dependent on
the topological class of the bulk band-structure. Since each band which intersects the
Fermi energy at kx has a Kramer’s pair at −kx , the number of Kramer’s pairs (NK )
intersecting EF is related to the change in the Z2 invariants across the interface, and
can be calculated using bulk-boundary interference which states,
NK = 4ν mod 2 (1.17)
Here, we provide a brief review of some of the topologically invariant systems, before
proceeding to three dimensional topological insulators.
7
Chapter 1. Introduction
The integer quantum Hall effect (QHE) was discovered by von Klitzing et al. in 1980 in
a high-mobility silicon metal-oxide-semiconductor device at a temperature (T ) = 1.5 K,
under a high magnetic field (18 T) [9]. When electrons are subjected to perpendicular
magnetic fields, their energy levels become quantized to En = (n+ 21 )~ωc , which are called
Landau levels, where ωc is the cyclotron frequency, as shown in Fig. 1.3a. This leads to
vanishing longitudinal conductivity (σxx ), along with quantized Hall conductivity (σxy )
νe2
of h
(ν is an integer), when the Fermi energy is tuned in between Landau levels (Fig.
1.3b) [8]. The bulk in such cases is insulating, but the edges are conducting due to
skipping orbits which form chiral 1D modes (Fig. 1.3c), and are prohibited from back-
scattering. It was shown by TKNN in 1982 that the quantum Hall effect is topological
in nature. When EF is tuned into the gap, dissipationless edge states appear, which are
topologically protected from back-scattering, which results in the forward and backward
channels remaining spatially separated. The topological invariant for the quantum Hall
effect is given by the integer ν, which is known as the TKNN invariant or the first Chern
number, and equal to the Berry flux enclosed by the Brillouin zone.
Before the discovery of the QHE in 1982, all states of matter, like solids, liquids, super-
conductor, magnets were described using principles of broken symmetry. The QHE was
the first system in condensed matter, where the notion of topology was used, in which
chiral one-dimensional, back-scattering prohibited edge states are present. The quantum
spin Hall effect (QSHE) is essentially two copies of the quantum Hall effect, superim-
posed on each other, under an equal and opposite magnetic field. States with opposite
spins propagate in opposite directions, and are referred to as helical edge states, due to
spin-momentum locking. One key difference with the QH effect, is that the time-reversal
symmetry (TRS) is protected in the QSH effect, with spin-orbit interaction playing the
role of the magnetic field. The two modes form a time-reversed pair to recover the overall
TRS.
Kane and Mele in their seminal paper, were the first to propose that graphene with
spin-orbit coupling, can be utilized to realize the QSHE [10]. Graphene’s band-structure
consists of two linearly-dispersed Dirac cones at two equivalent points (K and K 0 ) of
the Brillouin zone (Fig. 1.4a). The degeneracy at the Dirac point is protected by both
8
1.1. Basics of topological insulators5
(a) (b)
Figure 1.4: Kane-Mele model of graphene with spin-orbit coupling: (a) The
dispersion relation for a single layer of graphene, showing linear gapless dispersion near
the K and K 0 points. (b) Dispersion relation of a strip of graphene (shown in the inset)
with spin-orbit coupling, which leads to dissipationless edge-modes in the band-gap.
(Reproduced with permission from Ref. [10]).
time reversal and inversion symmetry. Spin-orbit interaction allows a new mass term,
which opens up a gap at the Dirac point and can lead to dissipationless edge transport
(Fig. 1.4b). However, the gap due to spin-orbit coupling is of the order ∼ 10−3 meV,
which makes it difficult to probe the edge states experimentally. Bernevig and Zhang
independently proposed the QSHE in semiconductors in the presence of strain gradients
in Ref [11], which also could not be realized experimentally.
In 2006, Bernevig, Hughes and Zhang (BHZ) in their seminal paper [12], proposed
sandwiching a thin layer of HgTe between two sheets of CdTe. The band structure
of CdTe is similar to other semiconductors, wherein the conduction-band-edge states
(Γ6 ) and valence-band-edge states (Γ8 ) have a s and p-like symmetry respectively. In
HgTe however, it is opposite as the p-level resides above the s-level, thus leading to an
inverted band-structure (Fig. 1.5a). The difference in energy between the Γ6 and Γ8 band
is ∼ −300 meV at T = 1.5 K. BHZ proposed that the QSH effect can be observed in
inverted type-III HgTe/CdTe quantum wells. They presented a quantum phase transition
between a trivial insulator phase and a QSH phase depending on the thickness d of the
sandwiched HgTe layer, forming the quantum well, which can be verified experimentally
by electrical measurements (Fig. 1.5b). When the gate-voltage is swept to vary the Fermi
energy (EF ) in the insulating regime, the two-terminal conductance drops to zero when
EF is in the bulk band gap. However, in the inverted regime, in the bulk gap, due to
9
Chapter 1. Introduction
(a)
(c)
(b)
Figure 1.5: Band inversion in HgTe/CdTe quantum wells and schematic for
experimental verification: (a) Bulk energy bands of HgTe and CdTe near the Γ
point. (b) The CdTe-HgTe-CdTe quantum well in the normal regime E1 > H1 with
d < dC and in the inverted regime H1 > E1 with d > dC . The red line indicates
Γ8/H1 symmetry, while Γ6/E1 symmetry is indicated in blue color. (c) Schematic
for experimental verification of spin Hall effect in HgTe/CdTe quantum well structures.
(Reproduced with permission from Ref. [12]).
the presence of back-scattering free edge modes, the conductance is quantized to 2e2 /h,
as shown in Fig. 1.5c. When the thickness of the quantum well is more than 6.3 nm,
the 2D electronic states have the normal band order. However, for dC ≥ 6.3 nm, the 2D
bands invert, leading to a quantum phase transition and the appearance of edge states.
Indeed, this was experimentally verified within a year by the group of Laurens
Molenkamp [13]. Transport measurements in an inverted HgTe/CdTe system were per-
formed at zero magnetic field, by sweeping the Fermi energy (EF ) from the bulk con-
duction to the bulk valence band. When the EF is in the bulk band gap, a quantized
conductance of 2e2 /h was observed due to one pair of helical edge states, conclusively
establishing the quantum spin Hall effect in HgTe/CdTe hetero-structures, when the
thickness of the HgTe layer was more than 6.3 nm. When the quantum well thickness
was 5.5 nm, such as in sample 1, the conductance was ∼ 0.01e2 /h, indicating the absence
of the QSHE as predicted (Fig. 1.6).
10
1.2. 3D topological insulator
1.2.1 Z2 Index
3D topological insulators are characterized by four Z2 topological invariants (ν0 ; ν1 ν2 ν3 )
[14–16]. There are four time reversal invariant points Γ1,2,3,4 at the surface of the Bril-
lioun zone, which are Kramer’s degenerate. The degeneracy is lifted due to spin-orbit
interaction away from these points. The 2D Dirac point is formed by these Kramer’s
degenerate points in the surface band structure. There are two possible ways in which
these points can connect to each other, either according to Fig 1.7a or Fig 1.7b. If the
Fermi surface intersects the line joining Γa and Γb points an even number of times, then
it is topologically trivial, whereas if it intersects an odd numbers of times, then the states
are topologically protected. The key difference between Fig 1.7a and Fig 1.7b is that,
in the first case the Fermi surface encloses an odd number of Kramer’s degenerate Dirac
point, whereas in the second case it encloses an even number of points. The former case,
which is characterized by ν0 = 0, is a weak TI, and can be constructed by stacking QSH
layers on top of each other, as shown by Fu, Kane, and Mele in Ref [10]. However, in
the presence of disorder, these states tend to localize. The latter, which is characterized
by ν0 = 1, is a strong TI, and in the simplest scenario, it encloses a single Dirac point
(Fig 1.7c). The indices ν1 , ν2 and ν3 can be interpreted as Miller indices categorizing the
orientation of the constituent layers. The Z2 index can be calculated using the formula
11
Chapter 1. Introduction
Figure 1.7: (a) Fermi circles in the surface Brillouin zone for a weak topological insulator
(b) Representaion of Fermi circles for a strong topological insulator. (c) In strong topo-
logical insulators, the Fermi circle encloses a single Dirac point, in the simplest scenario.
(Reproduced with permission from Ref. [2]).
[1],
4
Y P f [w(Λi )]
(−1) =ν
p (1.18)
i=1 det[w(Λi )]
Here, wαβ (k) = huα,−k |Θ|uβ,k i is the matrix representation of the time-reversal operator
Θ in the Bloch wave function basis, and α, β are the band indices.
In the case of 2D TIs, the smoking gun proof is to prove the existence of helical edge
states. In the presence of edge modes, the longitudinal conductance is quantized, which
can be verified by scanning the Fermi energy across the bulk band gap. The spin polarized
edge states can also be detected using the quantum spin Hall effect [13]. For 3D TIs,
the most compelling proof for existence of surface states is through angle resolved photo-
emission spectroscopy (ARPES) measurements, ideally using spin-resolved ARPES to
confirm that the Dirac cone is non-degenerate and helically spin-polarised [17, 18]. For
materials where ARPES is difficult, the Dirac Fermionic nature of the surface carriers
can be established by measuring the oscillations in conductivity called Shubnikov–de
Haas oscillations (SdH) [19, 20]. In SdH oscillations, conductivity oscillates periodically
with 1/B, and can be quantitatively expressed as
12
1.2. 3D topological insulator
F 1
4σxx ∼ cos 2π − +β (1.19)
B 2
where F is the oscillation frequency and β is the phase shift (0 ≤ β ≤ 1) and is equal
to the Berry phase divided by 2. The Berry phase is 0 for spin-less Fermions and π for
Dirac Fermions. The phase factor can be extracted from the Landau level fan diagram,
by extrapolating 1/B to zero. If the value of the intercept is half, then the oscillations
are due to Dirac Fermions. Another way to confirm Dirac Fermionic nature is to perform
scanning tunneling spectroscopy measurements under the influence of magnetic fields.
The energy level spacing for Dirac Fermions ∼ 1/N , and the zero-energy Landau level is
pinned to the Dirac point, all of which can be confirmed from the bias-voltage dependence
of the Landau quantization peaks.
p
E = ± p2 + m2 c2 (1.21)
For 2D massless Dirac Fermions, Eq. 1.20 is reduced to a 2 × 2 matrix, and is written as
e−iθ(k)/2
!
1
ψ± (r) = √ eik·r = u± (k)eik·r (1.24)
2 ±eiθ(k)/2
13
Chapter 1. Introduction
where θ(k) = tan−1 (ky /kx ) and the energy eigenvalues are E± = ±~vf · k, thus having
a linear dispersion. The typical Fermi velocity in these materials is 105 m/sec, which
has been confirmed through ARPES measurements as well [21, 22]. Though TIs have a
band-structure which to similar to graphene, there are key differences. Unlike graphene,
which has both valley and spin degeneracy at the K and K 0 points in the BZ, TIs are
non-degenerate, implying the degrees of freedom are 1/4 th of graphene. One of the most
prominent characteristics of topological surface states is spin-momentum locking. The
spin of the surface state carriers is non-degenerate and the direction of the spin is perpen-
dicular to the k, and is confined in the surface plane. This property leads to robustness
against back-scattering in the presence of non-magnetic disorder and also immunity from
Anderson localization, since time reversal symmetry prevents back-scattering between k
and k 0 , unless the direction of spin flips as well.
When the wave-vector k adiabatically traverses a closed loop in an anti-clockwise
¸
direction, the Berry phase acquired is γc = C dk · ihu± (k)|∇k |u± (k)i = π. Due to
π Berry phase, charge carriers interfere destructively, leading to weak-anti-localization.
Another interesting property of Dirac Fermions is the Landau level formation under
magnetic fields, which occurs according to the equation, E± (N ) = ± (2e2 vf ~B/c)N ,
p
√
where N = 0, 1, 2.... This implies that the level spacing ∝ 1/ N , unlike normal metals
where it is constant and equal to ~ωc = ~eB/mc c, where mc is the cyclotron mass.
1. The presence of gap-less surface states are preserved as long as time reversal symme-
try is maintained in the systems, which can be broken by application of a magnetic
field or ferromagnetic doping.
3. π Berry phase prevents weak localization, and gives rise to weak anti-localization
in these materials.
14
1.2. 3D topological insulator
Figure 1.8: Band-structure of Bi1−x Sbx : Schematic representation of the band struc-
ture of Bi1−x Sbx , which evolves from semi-metallic behavior for x < 0.07 to semiconduct-
ing behavior for 0.07 < x< 0.22, and back to semi-metallic behavior for x > 0.22. The
conduction and valence bands Ls,a invert when x ∼ 0.04. (Reproduced with permission
from Ref. [16]).
Bi1−x Sbx was the first experimentally verified 3D topological insulator, following predic-
tions by Fu and Kane [16]. Bismuth (Bi) is a high spin-orbit coupled semi-metal, with a
band-structure as illustrated in Fig. 1.8. The conduction and valence bands overlap at
the T point, which results in pockets of holes near the T point, and pockets of electrons
at the L points. While in pure Bi the conduction band has symmetric orbitals (Ls ) and
the valence band has anti-symmetric (La ) orbitals, in Antimony (Sb) it is opposite, with
the conduction band having anti-symmetric orbitals (La ) and the valence band having
symmetric orbitals (Ls ) orbitals with a band gap, as shown in Fig. 1.8. However, to
observe topologically protected states, the band gap needs to close. It was observed by
Kane and Fu [16], that by substituting Bi with Sb, the band gap can be closed, and
it would be possible to observe surface states. It was seen that for x = 0.04, where x
corresponds to Sb concentration, the gap between La and LS closes, and the massless
Dirac cones can be realized (Fig. 1.8). The 3D Z2 invariant for Bi1−x Sbx is [1; 111].
For x > 0.07, the band-structure is inverted, with the valence band becoming higher in
energy than the conduction band, and the system is an insulator. When x > 0.09, the
system is a direct band gap insulator, while for x > 0.22, the semi-metallic state is re-
stored. Angle resolved photo-emission spectroscopy measurements (ARPES) performed
by Hsieh et al. [22] have revealed the existence of surface states in Bi1−x Sbx , as shown
in Fig. 1.9. The surface bands cross the Fermi energy at five different points between Γ
and the M points, and has provided experimental verification of Bi1−x Sbx as a strong
topological insulator.
15
Chapter 1. Introduction
Figure 1.9: ARPES spectra of Bi0.9 Sb0.1 : ARPES data on the [111] surface of
Bi0.9 Sb0.1 , depicting the occupied surface states as a function of momentum (k) on the
line connecting the T invariant points Γ̄ and M̄ in the surface Brillouin zone. The sur-
face bands cross the Fermi energy five times. The ARPES data provides experimental
validation of Bi0.9 Sb0.1 as a strong topological insulator. (Reproduced with permission
from Ref. [22]).
The complex surface states, and along with impurity bands in the bulk band-gap due to
random substitutional disorder in Bi1−x Sbx led to the search of other 3D TI materials,
which are stoichiometric crystals with a well defined electronic structure. Zhang et al.
[21] in their seminal work predicted the existence of topologically protected surface states
in Bi2 Se3 , Bi2 Te3 , and Sb2 Te3 , with a single Dirac cone, at the k = 0 (Γ) point in the
surface Brilloiun zone. Of particular importance in this case is Bi2 Se3 , which has a band
gap of 0.3 eV, which is much larger than the energy scale at room temperature (KB T ∼ 25
meV). All these materials belong to rhombohedral crystal structure with the space group
D53d (R3̄m) with five atoms in a single unit cell. It consists of five atomic layers along
the z-direction, which is called one quintuple layer. The crystal structure of Bi2 Se3 is
illustrated in Fig. 1.10a. It has three-fold rotational symmetry along z direction, and
two-fold rotational symmetry along x-axis, and a bisectrix axis (reflection plane) in the
y-axis. While intra-quintuple layer coupling is strong, the inter-layer coupling is of Van
der Waal’s type, which makes them easily cleavable.
From ab-initio calculations, the band-gap of Bi2 Se3 is estimated to be ∼ 0.3 eV. The
band-structure of Bi2 Se3 without and with spin-orbit coupling is shown in Fig. 1.10b and
Fig. 1.10c respectively, which clearly exhibits that on the inclusion of spin-orbit coupling,
an anti-crossing feature appears at the Γ point. This indicates band inversion between
the valence and the conduction bands, suggesting Bi2 Se3 is a topological insulator. The
16
1.2. 3D topological insulator
(d) (e)
Figure 1.10: Origin of surface states in Bi2 Se3 , Bi2 Te3 , and Sb2 Te3 : (a) Crystal
structure of Bi2 Se3 with three primitive lattice vectors denoted as t1,2,3 . The red box
indicates one quintuple layer with Se1-Bi1-Se2–Bi10 -Se10 . (b) Band-structure for Bi2 Se3
without SOC. (c) Band-structure of Bi2 Se3 with SOC, showing a gap at the Γ point. The
dashed line indicates the Fermi level. (d) The parities of fourteen occupied bands (five
s bands and nine p bands, and the lowest unoccupied band) at the Γ point for the four
different materials Sb2 Te3 , Sb2 Se3 , Bi2 Se3 and Bi2 Te3 . The product of the parities for
the fourteen occupied bands is included in brackets on the right of each row. (e) Energy
and momentum dependence of the LDOS for Bi2 Se3 . The warmer color indicates higher
density of states. (Reproduced with permission from Ref. [21]).
17
Chapter 1. Introduction
(a) (b)
Figure 1.11: ARPES spectra of Bi2 Se3 : (a) ARPES measurements for Bi2 Se3 reveal-
ing the existence of surface states with a single spin-polarized Dirac cone along Γ̄ − M̄ .
(b) APRES along Γ̄ − K̄.(Reproduced with permission from Ref. [23] ).
parity of the 14 fourteen occupied band (five s bands and nine p bands) and the lowest
unoccupied band is shown in Fig. 1.10d for Sb2 Se3 , Bi2 Se3 , Bi2 Te3 , and Sb2 Te3 . On
switching the spin-orbit coupling, the parity eigen-value of one of the occupied bands
is changed at the Γ point, while the eigen-values are unchanged for all other momenta.
From Fig. 1.10d, we can see that the parity eigen-values of the highest occupied band
and the lowest unoccupied bands of Bi2 Se3 , Bi2 Te3 , and Sb2 Te3 are opposite compared
to Sn2 Se3 , similar to that obtained from band-structure calculations. This provides an
unambiguous proof that Bi2 Se3 , Bi2 Te3 , and Sb2 Te3 are topologically non-trivial, while
Sb2 Se3 is not. In the context of Bi2 Se3 , band inversion occurs when large spin-orbit
coupling pushes the| P 1+
z , ()i state below | P 2z , ()i, and since the levels are of
−
opposite parity, the inversion drives the system into a topologically non-trivial state.
The band inversion is qualitatively similar in Bi2 Te3 , and Sb2 Te3 as well.
However, in Sb2 Se3 the spin-orbit coupling strength is not strong enough for band
inversion, and the system is a trivial insulator. Zhang et al. also performed calculations,
which explicitly showed the presence of surface states in these systems, by constructing
the maximally localized Wannier function (MLWF), and subsequently obtaining the
surface Green’s function. The imaginary part of the surface Green’s function is the
local density of states (LDOS), from which the dispersion of the surface states can be
obtained. which is illustrated in Fig. 1.10e for Bi2 Se3 . The presence of surface states in
these materials have been confirmed by ARPES measurements [23], which is shown in
18
1.2. 3D topological insulator
Fig. 1.11.
In this method, the chalcogenides are sealed in evacuated quartz-glass tubes, capable
of sustaining temperatures up to 10000 C. The temperature of the melt is slowly de-
creased, while a gradient is maintained in the tube, so that the crystallization begins
at the cold end, and the crystal grows as the solidification proceeds from the cold end.
Bridgeman’s technique is particularly useful for growing layered TI like Bi2 Se3 , Bi2 Te3 ,
and Bi1.6 Sb.4 Te2 Se.
Another method of synthesizing bulk single crystals is vapor transport technique,
where a lump of poly-crystalline material is put at one end of a sealed quartz-glass tube.
The tube is kept within a furnace for a long duration, while maintaining a temperature
gradient, in which the poly-crystalline material is on the hotter end. The temperature
gradient ensures that the poly-crystal sublimates at the hotter end and crystallizes in
a colder part. When certain reagents such as Iodine gas is used as a transporter, the
technique is called chemical vapor transport. However, if the material sublimates easily,
the transporter is not required, and is known as physical vapor transport. Single crystals
of topological crystalline insulators such as SnTe, (Pb,Sn)Se, and (Pb,Sn)Te are usually
grown with the vapor transport technique.
Since these materials cleave easily, they can be used to obtain very thin flakes by
exfoliating them on SiO2 /Si wafer, by using scotch tape technique.
High quality thin films of TIs are usually grown by molecular beam epitaxy, where
materials are grown by co-evaporating the constituent elements with suitable flux ratios.
19
Chapter 1. Introduction
Figure 1.12: Dirac-cone engineering: Schematic band diagrams for Bi2−x Sbx Te3−y Sey
for various compositions, demonstrating the tunability of carrier density as well as the
evolution of the chemical potential for the different compositions. (Reproduced with
permission from Ref. [19]).
Because of the presence of Van der Waal’s gap in tetradymite TIs, lattice matching is
not very crucial and the sample and substrate are weakly coupled by Van der Waal’s
bond. However, controlling the substrate temperature is an important parameter for
obtaining high quality samples. Using epitaxial growth, 3D TI materials like Bi2 Se3 ,
Bi2 Te3 , Sb2 Se3 as well as the solid solution (BiSb)2 Te3 have been grown on a variety of
substrates such as Si(111), graphene terminated 6H-SiC(0001), SrTiO3 (111), GaAs(111),
sapphire (0001), CdS(0001),and InP(111). Other methods of growing thin films include
atomic layer deposition, chemical vapor deposition. However, the sample quality is not
as high as that obtained from molecular beam epitaxy.
20
1.2. 3D topological insulator
Inspite of the various available growth methods for synthesizing TIs, most materials
have a high residual doping in the bulk, which provides a parallel path for transport,
and impedes observation of surface transport properties. For examples, Bi2 Se3 crystals
are always electron-doped due to the presence of Se vacancies or anti-site defects that
have low formation energies. The bulk carrier number density as a result can be as high
as n3D ∼ 1019 cm−3 . The bulk number density can be reduced by addition of Ca2+ , which
leads to a lower number density n3D ∼ 1016 cm−3 . However, these defects lead to strong
scattering centers, and hence, reduced electron mobility. High-mobility p-type crystals
of Bi2 Se3 were grown by doping it with Cd in a Se-rich growth condition. These crystals
show ambipolar behavior as well as surface SdH oscillations. In contrast, Bi2 Te3 can be
grown in both n- and p-types. Te anti-site defects are formed in Te-rich environment,
while Bi anti-site defects form in Bi-rich conditions, which leads to high bulk number
density.
Higher insulating bulk can be achieved in a material like Bi2 Te2 Se, because of its
crystal structure (Te–Bi–Se–Bi–Te). Since Se is embedded in the middle of the quintuple
layer (QL), Se vacancy is more difficult to occur than in Bi2 Se3 . The higher electro-
negativity of Se compared to that of Te, ensures that Bi-Se is stronger than Te-Bi, which
reduces Bi/Te anti-site defects. All these, coupled to its ordered structure ensure high
mobility carriers and observation of topological properties in transport. Another method
to achieve a high insulating bulk it to use the solid solution Bi2−x Sbx Te3−y Sey . In this
material, Se resides in the middle of the QL, thus making it difficult for Se vacancies to
form, similar to Bi2 Te2 Se. The ratio of Bi/Sb and Te/Se ratio in the inner and outer
layers affect the levels of acceptors and donors, respectively, making it possible to achieve
a maximally compensated situation. By tuning the values of x and y, a high insulating
bulk as well as n or p-type carriers has been achieved. Such tuning of the Dirac-carrier
properties has been dubbed as “Dirac-cone engineering”, as shown in Fig. 1.12 [19]. A
summary of various experimentally verified topological insulators is provided in Fig. 1.13
[1].
21
Chapter 1. Introduction
22
1.3. Important phenomena in TI
1 1 α
L= (E 2 − B 2 ) + 2 θ(E · B) (1.25)
8π µ 4π
where is dielectric constant, µ is magnetic permeability, α = e2 /~c is the fine structure
constant and θ = π for Z2 TIs. The equations for D and H for TIs is given by
23
Chapter 1. Introduction
(a) (b)
(c) (d)
Figure 1.14: Quantum anomalous Hall effect: (a) Graphical illustration of the quan-
tum anomalous Hall effect in a ferromagnetically-doped TI, which breaks time reversal
symmetry and opens up a gap in the surface bands. (b) Schematic of the dependence of
σxx [σxx (0), in red] and σxy [σxy (0), in blue] on the chemical potential at zero magnetic-
field in the QAH effect. (c) Optical micrograph of a Cr0.15 (Bi0.1 Sb0.9 )1.85 Te3 thin film.
(d) ρxy (empty blue squares) and ρXX (0) (empty red circles) as a function of VG at
T = 20 mK, and B = 0 T, showing quantized Hall resistance, which provides exper-
imental verification of the QAH effect. (Reproduced with permission from Ref. [26]).
24
1.4. Quantum transport in topological insulators
αθ
D = E + 4πP − B (1.26)
π
αθ
H = B − 4πM + E (1.27)
π
where D, P , and M are the electric induction, electric polarization, and magnetization
respectively. This gives 4πM = αE and 4πP = αB. Thus electric field induces mag-
netism and the magnetic field induces electric polarization, both which are proportional
to a universal constant. This is the topological magneto-electric effect. Another intrigu-
ing consequence of the θ-term is the manifestation of an image magnetic monopole. If
the surface of a TI is gapped out, and a point charge is placed near the surface, the
response of the TI looks similar to a magnetic monopole in the TI.
1.4.1.1 Weak-localization
One of the most important effects of quantum mechanical interference is the enhanced
probability of back-scattering in a disordered metallic system, which leads to an anoma-
lous increase of resistivity at low temperature [29–33]. This effect is a precursor of strong
or Anderson localization in heavily disordered systems and is known as weak localiza-
tion (WL). It is purely a consequence of interference of the probability amplitudes of
the Feynman paths of the charge carriers. The probability P (a, b, t) for an electron to
traverse from point a to point b in a time t is given by:
X X
P (a, b) = |Ai |2 + Ai A∗j (1.28)
i i6=j
25
Chapter 1. Introduction
WL
WAL
The first term corresponds to the classical probability (corresponding to the Diffuson
channel), while the second term corresponds to quantum interference phenomenon (cor-
responding to the Cooperon channel). The interference term averages out in diffusive
systems when large number of paths contribute to the sum, provided the point a and b are
different, because different trajectories have uncorrelated phases. However, in the case
of coherent back-scattering, the probability amplitudes for clockwise and anti-clockwise
propagation around a closed loop are identical by virtue of time reversal symmetry (Fig.
1.15a) [33]. This gives P (a, b) = 4|A|2 , implying that the quantum mechanical proba-
bility of back-scattering is twice that of the classical probability. The weak localization
correction to conductivity ∆σW L for a 2D disordered conductor has a universal magni-
tude given by:
αe2
τφ
∆σW L = ln (1.29)
2h τ0
where α is a phenomenological pre-factor close to unity and τ0 and τφ are the momen-
tum relaxation time and phase breaking time respectively. For 2D systems, τφ ∝ T −1 for
electron-electron scattering, which is the dominant dephasing mechanism at low temper-
atures, which implies that the weak localization correction varies logarithmically with
26
1.4. Quantum transport in topological insulators
temperature.
The classical and the quantum diffusion probabilities are plotted qualitatively in Fig.
1.15b [34]. In the case of quantum diffusion, the dashed peak at the origin describes
the tendency to remain at or return to the origin, leading to enhanced resistivity. In
the presence of large spin-orbit coupling, the probability of back-scattering reduces (Fig.
1.15b). Under the application of a perpendicular magnetic field, the relative phase of the
two back-scattered waves change by 2eφ/~, where φ is the magnetic flux. The additional
phase introduced due to application of magnetic field, reduces the back-scattering prob-
ability, leading to positive magneto-conductance. In large spin-orbit coupled systems
like TI, there is an additional π phase between the time reversed paths, giving rise to a
negative magneto-conductance, referred to as weak-antilocalization (WAL) (Fig. 1.15c).
The magneto-conductance for WAL, which is observed in TI, can be fitted with the
Hikami-Larkin-Nagaoka (HLN) [29, 35] equation for diffusive metals with high spin orbit
coupling (τφ >> τso , τe ):
e2
1 Bφ Bφ
4σ = −α ψ + − ln (1.30)
πh 2 B B
where τφ , τso , τe are the phase coherence or dephasing time, spin-orbit scattering time
and elastic scattering time respectively, ψ is the digamma function and Bφ is the phase
breaking field. Here α and Bφ are fitting parameters. The phase coherence length
lφM R can be extracted using lφM R = ~/4eBφ . The value of α gives an estimate of the
p
27
Chapter 1. Introduction
(a) (a)
(b) (b)
(c)
aperodic yet reproducible, and the magnitude is independent of the specific material
1
properties or geometry (Fig. 1.16a). For T → 0, UCF magnitude hδG2 i 2 → e2 /h, while
at finite temperature,
2
e2
1 Ly
hδG i ' 2
α(kF δr) ns (T )lφ4 (1.31)
h kF l L3x
where ns (T ), kF , l, Lx and Ly are the density to active scatterers, Fermi wave-vector,
mean free path, and sample dimensions in x and y respectively. α(x) represents the
change of the phase of electron wave-function due to scattering by a moving impurity
at a distance δr. The magnitude of conductance fluctuations depends strongly on the
symmetries present in a system.
The field of random matrix theory was developed by mathematician Dyson, Wigner and
Mehta [43, 44], while working on the description of excited states of heavy nuclei in
nuclear reactions. It was argued that, since the Hamiltonian H is unknown, any specific
H should be regarded as a member of an ensemble of Hamiltonians, having the same
28
1.4. Quantum transport in topological insulators
properties, and the energy levels are related to the eigen-values of the large random
matrix. Wigner and Dyson in their seminal work, approximated H by an ensemble of
finite (n × n) Hermitian matrices Hn with a probability density,
2
e2 ks2
2
hδG i ∝ (1.33)
h β
Here β, s and k are the Wigner-Dyson parameter (dependent on the universality class
of the system), degeneracy of the system under investigation, and the number of inde-
pendent eigen modes of the Hamiltonian respectively. The dependence of the magnitude
of fluctuations on these parameters, makes it an important tool for identifying the vari-
ous symmetries present in the system. A factor of two reduction occurs when the time
reversal-symmetry is broken by the addition of a magnetic field, as shown in Fig. 1.16b
for phosphorous-doped Silicon systems, when B >> Bφ . The factor of two reduction
occurs due to the removal of time reversal symmetry, and a transition from orthogonal
to unitary symmetry classes. A further factor of two reduction can be observed on ap-
plication of B >> BZ , where BZ is the Zeeman field, due to removal of spin-degeneracy.
29
Chapter 1. Introduction
Conductance fluctuations has also been used to probe the valley symmetry in single-layer
graphene. As demonstrated in Fig. 1.16c, the magnitude of conductance fluctuations is
reduced by a factor of four, at higher number densities due to a transition from sym-
plectic to orthogonal symmetry classes. When the Fermi energy is tuned away from
the Dirac point, the valley degeneracy is removed, which eliminates the isospin triplet
channels, leading to the factor of four reduction [41, 42].
30
1.4. Quantum transport in topological insulators
Ag (6N)
Ag (5N)
Ag (5N)0.3Mn
Ag (5N)1Mn
Figure 1.17: Saturation of dephasing time: (a) The electron dephasing time τφ as
a function of temperature for 6000 Å thick Au–Pd films with different resistivities. The
√
solid line is drawn proportional to T −2 . (b) Normalized resistance as a function of T
for three Au–Pd films. (c) τφ as a function of T for Au and Ag thin films, made from
6N pure sources, none of which show any saturation. The dashed line is the behaviour
of τφ , which is expected for e-e interactions. (d) Comparison of τφ for pure Ag (6N)
and Ag ˘(5N) samples with 0.3 ppm and 1 ppm Mn-doped samples. The magnetically-
doped samples show a saturation, and a substantial reduction of τφ . (e) T -dependence
of several inelastic length, Lin , and spin-orbit length LSO for Q1D wires obtained from
magneto-conductance. (f) Variation of the ‘saturated’ dephasing time, τφ0 with diffusion
constant for several three-dimensional polycrystalline metals, showing a D−1 -dependence.
(Reproduced with permission from Ref. [47–50]).
31
Chapter 1. Introduction
over the past few decades [45, 49–62]. The saturation of lφ at low T raises questions about
the validity of the Fermi liquid theory itself in disordered conductors, which predicts an
increase in the phase coherence length with reducing T . This is due to reduced scat-
tering imposed by phase space constraints arising from Pauli exclusion principle. The
temperature for the onset of saturation depends on various parameters such as material,
the microscopic quality of the sample, and the sample geometry. The discrepancy be-
tween theory of weak-localization and experimental data has been attributed to the an
additional source of temperature independent inelastic scattering, which is dominant at
low T [51]. However, Mohanty et al. in Ref. [56] attributed the observed saturation to
the zero-point fluctuations of the phase coherent electrons and the environment. This
assumption faced criticism on thermodynamic considerations, since zero-point fluctua-
tions does not lead to dephasing because of the absence of energy at T → 0, which can
be exchanged between the electrons and the environment [63, 64]. Imry et al. in Ref.
[64] and Zawadowski et al. in Ref. [65] have proposed dephasing from two-level systems
(TLS) as a source of saturation. However, a TLS type distribution would lead to a large
anomaly in the specific heat of metals, which has not been observed [66]. Norman Birge’s
group showed the presence of minute concentration of magnetic impurities can lead to
the saturation [49], which were revealed through Aharonov-Bohm oscillations.
Some of the effects, which have been linked with the anomalous saturation of lφ are
discussed below.
(a) Heating effects:
One of the predicted reasons for the observed saturation is the presence of hot electrons,
having a temperature higher than that of the lattice, i.e. the electrons are not in thermal
equilibrium with the phonons. However, careful resistivity and magneto-resistance mea-
surements in metals films have shown that heating effects either due to external sources,
or sample current is not the cause of the saturation. Fig. 1.17a shows the phase breaking
time τφ , extracted for four sputtered Au-Pd films, which shows a saturation below T < 6
√
K. However, resistivity measurements, as shown in Fig. 1.17b, shows a T -dependence,
due to e-ph interactions in 3D, in the same T -range, where the saturation is observed
[48]. Similar conclusion gas also been reported in [47, 56, 67, 68].
(b) Effect of electro-magnetic field:
The effect of microwave on dephasing was considered by Altshuler in Ref. [69–71], by
considering small energy transfers (δE < KB T ) due to microwave noise, which can
32
1.4. Quantum transport in topological insulators
cause dephasing but not Joule heating. However, investigations by Webb et al. [61], and
Huibers et al. [72] in Au wires and quantum dots respectively, have revealed that heating
in the sample due to the presence of high frequency noise precedes dephasing. Burke et
al. studied the effect of fluctuating electric field on dephasing, which is similar to Nyquist
dephasing. However, no discernible effect was observed in these samples. Gersenshon et
al. (Ref. [73]) have argued that the saturation in silver wires can be caused by microwave
radiation with sub-pW power by measuring weak-localization in these wires.
The effect of magnetic impurities was investigated by Pierre et al. in Ref [49], which
showed that τφ extracted from magneto-resistance, did not saturate for the purest gold
and silver samples (purity ∼ 6 N), as demonstrated in Fig. 1.17c. To investigate the effect
of magnetic impurities, controlled implantation of manganese ions with a concentration
of 0.3 ppm and 1 ppm were done in two Ag-samples (5N purity). Mn atoms form Kondo-
type impurities with a Kondo temperature TK ∼ 40 mK. τφ in the Mn-doped samples
showed a substantial decrease compared to the pure sample, with an apparent saturation
for T < 1 K (Fig. 1.17d). In the presence of magnetic impurities, the phase breaking
time is given
1 1 1 1
= + + (1.34)
τφ τe−e τe−ph τsf
where τsf is the spin-flip scattering time. For T > TK , τsf is described by the Nagaoka-
Suhl expression, which is given by
1 cmag πS(S + 1)
= (1.35)
τsf π~vf π 2 S(S + 1) + ln2 (T /TK )
where S and TK are the spin and Kondo temperature of the magnetic impurities respec-
tively, and Cmag is the concentration of magnetic impurities. τsf reduces on decreasing
T , while τe−e and τe−ph are enhanced. A combination of these two effects leads to a
saturation of τφ in the samples with magnetic impurities.
The saturation in lφ was examined by Fukai et al. in Ref [50], on GaAs-AIGaAs on-facet
quantum wires from magneto-resistance and conductance fluctuations. These investiga-
tions revealed that the saturation sets in, when the spin-orbit coupling length, which is
33
Chapter 1. Introduction
Lin and Giordioano in Ref. [51] have studied several samples, both before and after
annealing, on glass and quartz substrates, which contains far fewer magnetic impurities
than glass. The nature of saturation does not change in these samples, indicating a non-
magnetic origin. Lin and Kao in Ref. [47] have investigated several three-dimensional
poly-crystalline disordered metals, fabricated using a variety of techniques (dc-sputtered,
RF-sputtered, thermal evaporation, arc-melting). Inspite of the differences in fabrication
methods, all the samples showed a saturation, with
where τφ0 is the saturation value of the phase breaking time, α ∼ 1, and D is the
diffusion constant. This is illustrated in Fig. 1.17f. Such a universal scaling implies that
the functional form of the dephasing rate might be universal with respect to disorder in
a given dimensionality.
• Electron-electron interactions
34
1.4. Quantum transport in topological insulators
(a) (b)
(b)
(c) (d)
(e) (f)
35
Chapter 1. Introduction
EF is in the bulk band-gap, which suppresses bulk conduction, and surface transport
predominates. The band-structure was verified by ARPES experiments as shown in Fig.
1.18a and Fig. 1.18b . The temperature and magnetic field-dependence of the quantum
correction to conductivity show that the combined effect of both weak-localization and
electron-electron interactions. The total correction to conductivity can be modeled as
where 4σEEI and 4σW L are the corrections due to electron-electron interaction and
weak-localization respectively. In the 2D limit, where the inelastic mean free path is
much larger than the sample thickness, both 4σW L and 4σEEI ∼ ln T , and it is essential
to incorporate both the contributions to explain the transport properties of doped and
undoped samples, as demonstrated in Fig. 1.18c-f.
Phase coherent transport in TI was investigated by Ockelmann et al. in Ref. [75] by
performing weak-antilocalization measurements in thin (6 − 30 nm) flakes of vapor phase
deposited TI Bi2 Se3 (Fig. 1.19a). The phase breaking length, obtained by fitting HLN
expression (Eq. 1.30) to magneto-resistance, shows a 1/T 0.5 scaling, at different number
densities (n), consistent with electron-electron interaction (Fig. 1.19b-d). At low T , lφ
saturates to various n-dependent values, which is attributed to additional dephasing due
to spin-flip scattering.
Liao et al. in Ref. [76] have carried out measurements in thin (15 − 30 nm) films of
(BiSb)2 Te3 to probe the dephasing mechanisms at low T , by measuring weak anti-
localization. Their results indicate that the dephasing field (Bφ ∝ 1/τφ ) evolves from
linear to sub-linear behavior, as the transport is varied from bulk to surface dominated
regime (Fig. 1.20a). The linear T -dependence arises due to Nyquist electron-electron
interaction, while the sub-linear behavior is due to coupling of the surface states to
the localized charged puddles in the bulk. The bulk layer is in the variable-range hop-
ping (VRH) regime, and the phase breaking length in VRH regime lφV RH ∝ T −p/2 with
p = 2/3 in 2D. The T -dependence of τφ for VRH is given by τφ−1 ∝ lφ−2 ∝ T p , while the
experimentally obtained values of p range between 0.45 − 0.6 (Fig. 1.20b). The coupling
of the surface states to the localized charged puddles, via tunneling leads to enhanced
dephasing in these systems compared to Nyquist dephasing (Fig. 1.20c).
36
1.4. Quantum transport in topological insulators
(a) (b)
(c)
Figure 1.20: Dephasing from charged defects in the bulk: (a) Evolution of the
phase breaking field (Bφ ) as a function of temperatures for various fixed gate-voltages.
The solid lines at fits using ∼ T p (b) α obtained by fitting Hikami-Larkin-Nagaoka
equation to WAL, which evolves from 0.5 to 1 for high positive gate voltages, signifying
decoupling of surface and bulk states. The exponent p, obtained from fitting Bφ vs T P for
the same gate voltages. (c) Schematic depiction of dephasing due to coupling of surface
states, and the charged puddles in the bulk. (Reproduced with permission from Ref.
[76]).
37
Chapter 1. Introduction
(a) (b)
Figure 1.21: Electron-phonon interaction effects: (a) WAL in Bi2 Te2 Se microflakes
at different temperatures. (b) lφ obtained from both magneto-resistance and conductance
fluctuations at different T . The solid lines are fits according to Eq. 1.38. (Reproduced
with permission from Ref. [77]).
• Electron-phonon interactions
Even though e-e interaction is the dominant mechanism in low T , the proximity of the
bulk to the surface states, leads to electron-phonon (e-ph) interactions, which cannot
be neglected. This was reported by Li et al. in Ref [77] by measuring both magneto-
resistance and conductance fluctuations in Bi2 Te2 Se micro-flakes. The dependence of
lφ on T can be adequately described by incorporating the effect of electron-phonon
interactions as well (Fig. 1.21a), and can be described by the expression,
1
lφ = (1.38)
(A0 + A1 T + A2 T 2 )0.5
Here, lφ ∝ T −0.5 , lφ ∝ T −1 are respective contributions from e-e and e-ph scattering. A1 ,
A2 are fitting parameters and A0 = 2 ,
1
lφ0
the saturation value of lφ (T (Fig. 1.21b).
Beidenkopf et al. in Ref. [78] have demonstrated from scanning tunneling microscopy
experiments, the susceptibility of topological surface states to scattering by charged
disorders in the bulk. These defects scatter the surface states, giving rise to an energy-
dependent spatial interference pattern in the conductance maps, which can be seen in
38
1.4. Quantum transport in topological insulators
(a) (b)
(c)
Fig. 1.22a. The Dirac point energy varies spatially due to fluctuations induced by poorly
screened charged dopants, and follows a Gaussian distribution with the width varying
between 20 meV to 40 meV (Fig. 1.22b). Near the Dirac point, the local electronic struc-
ture is dominated by electron-hole puddles, which has been previously seen in graphene
as well [79].
He et al. have investigated the effect of both non-magnetic and magnetic impuri-
ties on the surface of the TI Bi2 Se3 , by measuring magneto-resistance [80]. While the
WAL effect in TI is a consequence of π Berry phase, and remains unaffected after de-
position of non-magnetic Au impurities on the surface (Fig. 1.22c), the addition of Fe
impurities breaks the time-reversal symmetry in the system, leading to a crossover from
weak-antilocalization to weak-localization (Fig. 1.22c). This implies a topological phase
transition from a non-trivial to a trivial phase, and a crossover from symplectic to unitary
39
Chapter 1. Introduction
symmetry class. Magnetic doping leads to a gap at the Dirac point, and the competing
WAL and WL can also e explained due to the change in the Berry phase in the sample
[81]. Such crossver between topologically non-trivial to a trivial magnetic semiconductor
has also been observed by Liu et al. by depositing Cr on MBE-grown Bi2 Se3 [82].
40
1.5. Resistance fluctuations or noise
inelastic scattering, predominant at low T , and makes a compelling case for investigating
the different sources decoherence of surface carriers of TIs.
• Johnson’s Noise:
Johnson’s noise arises due to thermally generated random motion of charge carriers
in a system [89, 90]. The power spectral density for this type of noise is given by
SV = 4KB T R, where SV , KB , T and R are the power spectral density, Boltzmann
41
Chapter 1. Introduction
(a) (b)
constant, temperature, and resistance of the device under test respectively. Johnson’s
noise is independent of the frequency and is also referred to as white noise. Johnson’s
noise is used to accurately determine the electron temperature in a system.
• Shot Noise:
Shot noise arises due to discreteness of charge carriers [91, 92]. The power spectral
density of shot noise is given by SI = 2ehIi, where e andhIi are the electronic charge
and current through the sample respectively. Shot noise provides information about the
quasi-particle excitation of a system.
In electronic devices, 1/f noise in the drain current is a consequence of the slow random
change in the carrier number and/or the background disorder landscape, which in turn,
causes the mobility fluctuations of the carriers. Due to the two-dimensional nature and
diversity in the electronic structure of Van der Waal’s solids, which can depend on the
number of atomic layers [2, 93, 94], the screening of the Coulomb potential from the
traps can vary significantly from one system to the other. This makes the origin and
magnitude of noise between thin and thick topological insulators [95–97], and graphene
or TMDCs, quite different. Below, we briefly describe some of the key aspects of 1/f
noise, with an emphasis on the origin of 1/f noise in TIs.
42
1.5. Resistance fluctuations or noise
The omnipresence of 1/f -like power spectral density (PSD) of fluctuations in a solid state
system is generally attributed to slow reorganization of background potential landscape
[98–101]. The relaxation process is a superposition of fluctuators of characteristic time
τ with a distribution D(τ ) ∝ τ −1 . The PSD of due to a single time-scale is a Lorentzian
τ
S(f ) ∝ (1.39)
(2πf )2 τ 2 +1
The total power spectral density of noise is then a weighted sum of all relaxations over
a bandwidth (τ1 ,τ2 ),
ˆ τ2
τ
S(f ) ∝ D(τ )dt (1.40)
τ1 (2πf )2 τ 2 +1
which yields S(f ) ∝ 1/f , for τ2−1 f τ1−1 . Thus any random process with a wide
distribution of time scales shall exhibit a 1/f spectrum.
One of the earliest and most important frameworks of 1/f noise is the empirical Hooge
model [101–103], which is expressed as
SV (f ) γH
= (1.41)
V2 Nfα
where SV is the PSD of voltage fluctuations across an electrical resistor, in a constant
current mode. Here V is the average voltage drop across the sample, N is total number of
charge carriers, and γH is a phenomenological dimensionless parameter, called the Hooge
parameter, which is a figure of merit for noise in a system. The frequency exponent
α ≈ 1 ± 0.2 characterizes the 1/f noise. Experimentally, a 1/N -type decrease in the
noise magnitude is a characteristic feature of systems following the Hooge model. The
Hooge parameter varies from one system to another, and is often observed to be close
to 10−3 − 10−2 for thin films of simple metal. This model assumes that the conductance
noise originates from mobility fluctuations, arising due to scattering caused by number
fluctuations of phonon modes in the lattice. This model is applicable to homogeneous
and clean metals and semiconductors.
43
Chapter 1. Introduction
The Dutta-Horn model was first used to explain the origin of 1/f noise in the electrical
resistance of thin metal films [100, 101]. In this model, the effect of short and long range
diffusion of lattice defects and impurities is taken into account, where the activation
energy barrier E to diffusion has a distribution function D(E). Since the characteristic
time scale τ = τ0 exp(E/KB T ) is directly determined by the activation barrier, the noise
magnitude
ˆ ∞ τ0 exp( KEB T )
S(f ) ∝ D(E)dE (1.42)
0 1 + (2πf τ0 exp( KEB T ))2
naturally assumes the 1/f spectral dependence, when D(E) varies weakly with E in
the range KB T ln( ττ10 ) . E . KB T ln( ττ20 ). Here τ1−1 and τ2−1 are microscopic cut-off
frequency scales that are determined by, for example, finite system size or nature/spatial
correlation of the disorder itself, τ0−1 is the scale of phonon frequency, while KB is the
Boltzmann constant. The applicability of the Dutta-Horn model can be confirmed by
comparing the experimentally obtained α with that computed from the temperature
dependence of noise. In addition, Dutta-Horn model also predicts a collapse of f S(f )
plotted as a function of KB T ln(f /f0 ), where f0 ∼ 1013 Hz, is the phonon frequency, for
different defect density that can be confirmed experimentally.
The McWhorter carrier number fluctuation model was first introduced to explain the
1/f noise in silicon MOSFETs [101, 104], where both magnitude and spectral shape of
the noise was explained by the slow trapping-detrapping of charge carriers from the traps
at the channel and oxide interface. The charge fluctuation results in an effective change
in the potential landscape, leading to the change in carrier number in the channel.
The trapping/detrapping events are due to quantum tunneling of charge between the
channel and the trap states in the oxide. In such tunneling processes, the relaxation
time scale for a specific trap located at a distance x below the interface is given by,
τT = τ0 exp(2κx), τ0 being a microscopic time scale of the order of the phonon frequency,
and κ = 2m∗ φB /~2 ∼ 109 m−1 , where φB and m∗ are the oxide tunnel barrier height
p
and the effective electron mass, respectively. A uniform distribution of the trap states
close to the interface leads to a wide distribution of τs , which sum up to give 1/f noise
44
1.5. Resistance fluctuations or noise
One of the drawbacks of the McWhorter model was that it neglects the possibility of
mobility fluctuations, which is concomitant to the trapping of charge in the gate ox-
ide, because of Coulomb scattering. This leads to an over-estimation of Dit when the
McWhorter model is used whereas in reality noise contains explicit mobility and carrier
density dependence. The effect of Correlated number and mobility fluctuation was taken
into account in [105, 106].
From Ohm’s law and Drude model, the drain current can be expressed as
W
IDS = µN QN VDS (1.44)
L
where IDS is the source-drain current, µN is the mobility, QN = ne = C(VG − VT ), VDS
is the source-drain voltage, C is the capacitance, VG is the gate voltage, L is the length,
W is the width of the sample, and VT is the threshold voltage. In a channel element
4y4z, the fluctuations in the current due to fluctuations in the number density in a
volume element 4V and energy element 4E can be expressed as
W δQN 4y4z
4ID,4V 4E = VDS µN + QN δµN (1.45)
L WL
Since, the mobility fluctuations δµN are correlated with the fluctuation in the carriers
δQN , δµN can be expressed in terms of δQN . Using Matthiessen’s rule,
1 1 δQN 4y4z
= + S(x) (1.46)
µn µN WL
Here, µn is the instantaneous mobility, µN is the time average mobility, S(x) is the scat-
tering rate, and x is the distance into the oxide. The scattering rate can be approximated
45
Chapter 1. Introduction
δQN ∆y∆z
δµN = µ2N S(x) (1.47)
WL
Using Eq. 1.47 and Eq. 1.45, the fluctuations in the drain current can be expressed as
W δQN ∆y∆z
4ID,4V 4E = VDS µN + QN µ2N S(x) (1.48)
L WL
The normalized power-spectral density is given by
W 2 VDS
2
SI SQN ,∆y∆z
= µ2N 1 + 2µN QN S(x) + Q2N µ2N S 2 (x) (1.49)
I2 4V 4E
2
L I 2 W 2 L2
I 2 L2
Using QN = ne, and σ = 2 W2 ,
VDS
we have
µ2N
Sσ SQN ,∆y∆z
= 1 + 2σS(x) + σ 2 S 2 (x) (1.50)
σ2 4V 4E σ 2 W 2 L2
Since µ = 1 dσ
e dn
, we have
2
Sσ 1 dσ SQN ,∆y∆z
= 2 2 1 + 2σS(x) + σ 2 S 2 (x) (1.51)
σ2 4V 4E eσ dn W 2 L2
The spectral density of the fluctuation in the trapped charges is equal to the PSD of the
induced fluctuation in the channel charge, such that
τT
SNT ,4V 4E = N T (x, E)fT (1 − fT )4V 4E (1.53)
1 + ω 2 τ 2T
where NT is the trap density, ω is the frequency, and fT is the probability of a trap being
filled. The time
q constant associated with a trapping event is given by, τT = τ0 exp(2αx),
∗
where α = 2m~e2Φb . Thus, Eq. 1.51 can be re-written as
46
1.5. Resistance fluctuations or noise
2
SV Dit KB T dσ J1 J2
2
= + + J3 (1.54)
V dW L dn σ2 σ
where Dit is the areal-trap density, m∗e is the effective mass of electron in STO, and
Φb is the height of the potential barrier at the interface. J1 = 8α 1
represents a pure
´ 2 τT
number-fluctuation, J3 = A (x) 2 dx represents pure mobility-fluctuations and
1+(2πf τT )
´ τT
J2 = 2A(x) 2 dx represents combined number and mobility-fluctuations.
1+(2πf τT )
SV 1
2
∝ (1.55)
V 1 + (2πf τ )2
The generation-recombination noise generally appears in addition to the 1/f noise in the
channel, and exhibits a non-monotonic T dependence as the modal frequency lies within
the finite measurement bandwidth in the optimal temperature range. Experimentally, it
is often identified from the evolution of fmax with T , where fmax = τ0−1 exp(−ε/KB T ),
is the frequency at which the power spectral density becomes maximum.
Other models to describe various features of noise in 2D material-based field effect
devices have also been suggested and highlight that the effect of charge inhomogeneity
may last upto different Fermi energy values in different materials and is dependent on
their bandstructure [109].
The proximity, and possible hybridization[110? ] of metallic surface states and semi-
conducting bulk states, make noise characteristics in TI [95, 96, 111] differ significantly
from that in graphene. While the substrate plays a major role in determining the noise
magnitude in graphene [112], the 1/f noise in TI can have a very different origin even
though the basic device structure remains similar.
47
Chapter 1. Introduction
Early noise measurements [111] in exfoliated nanoflakes of Bi2 Se3 at room tempera-
ture, which clearly demonstrated noise power spectral density ∝ 1/f α did not differen-
tiate between the surface and bulk contributions. Bhattacharyya et al. [95] have carried
out extensive 1/f noise measurements in mechanically exfoliated Bi1.6 Sb0.4 Te2 Se (BSTS)
alloys on p++ SiO2 /Si substrate as function of temperature (T ∼ 7 − 300 K), film thick-
ness (d ≈ 10 nm to bulk), and gate voltage VG . The electrical transport in these systems
has been shown to be dominated by surface states for T < 50 K and d ≤ 1 µm. The
VG -dependence of noise in a typical exfoliated TI with d = 110 nm shows a monotonic
increase as the number density is tuned towards the charge neutrality point (CNP), or
the Dirac point, for different temperatures (Fig. 1.25a). It has been shown that the
mobility-fluctuation induced noise satisfactorily describes the VG -dependence of noise.
McWhorter type number density fluctuation model fails to describe the VG -dependence
both qualitatively as well as quantitatively. In thin exfoliated devices [95] of thickness
∼10 nm, the VG -dependence of noise shows a monotonic increase with a suppression
near Dirac point, which vanishes as the temperature is increased (Fig. 1.25b). Such
M-type behavior has been observed in graphene as well [113], and has been predicted to
originate from mesoscopic quantum interference effect in Dirac fermions in a spatially
inhomogeneous carrier distribution [88]. The T -dependence of noise in TI (Fig. 1.25c)
was found to be unique among different classes of 2D materials, and reveals a series
of unexpected peaks at characteristic temperatures, irrespective of the thickness. Such
peaks arise due to generation-recombination in the bulk of the TI, that manifest through
mobility fluctuations in the surface transport. 1/f noise has also been used as a tool to
48
1.6. Topological insulators as photo-detectors
49
Chapter 1. Introduction
Figure 1.27: Photo-response in Bi2 Se3 /Si hetero-structures: (a) Illustration of the
setup for studying the time response of the photo-detector. (b) Photo-current switching
performance of the photo-detector measured under 808 nm light illumination at zero bias
voltage. (c) Schematic illustration shows the transport of the photo-generated carriers
with the aid of surface states of the Bi2 Se3 film. (Reproduced with permission from Ref.
[127]).
50
1.6. Topological insulators as photo-detectors
(a) (b)
(d)
(c)
51
Chapter 1. Introduction
generates electron-hole pairs in the bulk. Due to band-bending, and the in-built electric-
field at the interface, the photo-generated the electrons get injected into the Si substrate,
while the holes are transferred from the Bi2 Se3 layer into electrode, leading to a photo-
current in the external circuit. These vertical hetero-structures exhibit a responsivity
of 24.28 A W−1 , detectivity of 4.39 × 1012 Jones, and a response time of ∼ 10−6 secs
(Fig. 1.27c). Not only that, the detectors respond to a wide range of wavelength, from
ultra-violet to infra-red, and upto 60 KHz frequency, which makes them suitable for a
variety of practical applications.
Opto-electronic characteristics of another TI Sb2 Te3 , grown on sapphire substrate by
molecular beam epitaxy were reported by Zheng et al. [128]. The temporal response
of the detectors is shown in Fig. 1.28a at different temperatures at a fixed VDS . The
current (resistance) switches from a “low” (“high”) state to a “high” (“low”) state when
optical illumination is switched on. However, the rise-time is very slow (∼ 100 secs),
which is a severe limitation. The photo-detectors also exhibit photo-response in the
near-IR regime for an incident illumination of 980 n.m. The photo-response is attributed
to two different factors: (a) the photo-generated carriers in the bulk due to incident
illumination, and (b) due to spin-current along the surface of the topological insulator
(Fig. 1.28b). The responsivity and photo-conductivity gain were estimated to be 21.7
AW−1 and 27.4 respectively, at 980 nm optical illumination.
Broadband photo-detectors based on graphene-Bi2 Te3 hetero-structures was reported
by Qiao et al. in Ref. [129]. Bi2 Te3 nano-crystals, which were grown on graphene, acts
as the light-sensitive material (Fig. 1.28c). The electric-field at the interface leads to
effective separation of electron-hole pairs after illumination, which are transferred onto
graphene, and shows a high responsivity due to high gain of graphene layer. The device
responds from visible to IR range (Fig. 1.28d), and shows a responsivity of 35 AW−1 at
a wavelength of 532 nm.
52
1.7. Photo-detection mechanisms:7
The comparison of the some of the TI-based opto-electronic devices is provided in Fig.
1.29 [127]. The responsivity is limited compared to other materials, such as graphene-
MoS2 , or graphene-Te nanowire hybrids [118, 126], due to high recombination rate in the
bulk.
6
Inspired by Ref [3]
53
Chapter 1. Introduction
The bolometric effect occurs due to heating of graphene associated with optical illu-
mination, which leads to a change in the transport properties. The thermal-resistance,
RH = dT /dP determines the sensitivity, while the heat capacity determines the response
time (τφ ) in these devices, since τφ = Rh Ch , where Ch is the heat capacity. The transport
properties are modified due to the change in the mobility of the carriers, or due change
in the number of carriers in the channel (Fig. 1.30c). This effect has been previously
observed in [119, 124].
In this effect, the electron-hole pairs are generated in another photo-active material in
the vicinity of the detector. One type of carrier remains trapped in the light sensitive
material, while the other is transferred onto another material due to an interfacial electric
field, which can arise due to work-function mismatch. This leads to an effective gating
of the channel, resulting in a change in its resistance. Photo-gating effect can give rise to
very high responsivity, and has been observed previously in samples involving graphene
[118, 123, 126].
54
1.7. Photo-detection mechanisms:8
1.7.5.1 Photo-current
Photo-current (Ip ) is defined as the difference between the on-state current when the
optical illumination is switched on (Ion ), and the off-state current or dark current (Iof f )
of the detector, when illumination is off, and is given by
Depending on the sign of Ip , the response can be positive (Ip > 0) or negative (Ip < 0).
1.7.5.2 Photo-responsivity
If Ip is the photo-current which has been generated under an illumination power P , then
the photo-responsivity is defined as
γ = IP /P (1.57)
N EP = IN /γ (1.58)
where IN is the total current noise of detector and γ is the photo-responsivity. The unit
of N EP is W Hz−1/2 . The total current noise is defined as
q
IN = IJ2 + IS2 + If2 + IP2 D (1.59)
Here IJ , IS ,If , and IP D are the contributions from Johnson’s noise, shot noise, flicker or
1/f noise, and the photo-detector noise respectively. The various contributions can be
55
Chapter 1. Introduction
(1.60)
p
IJ = 4KB T /R
√
IS = 2eI (1.61)
(1.62)
p
If = KI 2 /f
√
D∗ = A/N EP (1.63)
where A is the area of the sample. Since detectivity takes into consideration the area
of the sample as well, it is used at the parameter to compare the response in different
photo-detectors. The unit of detectivity is cmHz1/2 W−1 or Jones.
56
1.8. Motivation for the present work
computation, low power electronics, spintronics and thermoelectrics. Many of these ap-
plications require phase coherent electronic wave-functions, which manifests in quantum
corrections to conductivity. Quantum interference in these systems also leads to weak
anti-localization because of π Berry phase. Despite extensive transport and spectro-
scopic investigations in these systems, the understanding of conductance fluctuations in
the quantum transport regime is still in its incipient stages, and the sensitivity of the sur-
face states to different dephasing mechanisms is still a matter of debate. Understanding
these phenomena are critical, for applications of topological insulators in various quan-
tum technologies. From an application perspective, TIs are promising infra-red sensors,
due to their favorable bulk band gap, However, because of phonon-mediated scattering,
the recombination rate of photo-generated charge carriers are high, thereby posing a
great challenge in achieving ultra-sensitive opto-electronic detectors, especially at higher
temperatures. In this thesis, we have delved into exploring the role of quantum inter-
ference and external potential fluctuations in topologically protected materials, and also
enhancing the opto-electronic sensitivity of these materials by forming binary hybrids
with graphene.
In the first part, we explore the source of 1/f noise the TIs by studying time-
dependent conductance fluctuations. Earlier work on the origin of 1/f noise, could
not successfully establish the source of noise in thin TIs, because of the contribution
of universal conductance fluctuations (UCF) [95]. We use large-area, epitaxially grown
samples for reducing UCF contribution, to pin-point the origin of 1/f noise in these
materials.
In the next part, we probe the ground state symmetry class of topological insulators,
and the preservation of time-reversal symmetry in these systems. Experimentally, the
sensitivity of transport to magnetic impurities, or the saturation of the phase breaking
length at low temperatures are directly connected to the time-reversal symmetry. This
makes an explicit demonstration of the symplectic to unitary crossover an important
task, without the addition of any external impurities [80], which has not been realized
yet.
In the third part, we have probed the sources of inelastic scattering in TIs, which lead
to the saturation of phase coherence length at low temperatures. Topological protection
of the surface states make these materials a strong contender as the building blocks of
qubits, which require long phase coherence length for error tolerant quantum computa-
tion. Hence, it is critical to understand the mechanisms responsible for dephasing of the
57
Chapter 1. Introduction
surface electrons, which is equivalent to the loss of information. The dephasing mecha-
nisms in TIs at low T have been known to be electron-electron interaction [74, 75, 83],
electron-phonon interaction [85], and coupling of surface states and localized charged
puddles in the bulk [76]. All of these mechanisms lead to an increase of lφ with de-
creasing T [45, 75, 83, 132]. However, the enhancement is often followed by a saturation
[76, 84, 85]. A systematic approach to figure out the different avenues that lead to this
saturation remains unexplored.
The final part delves into opto-electronic applications of TIs, which are promising
candidates for infra-red photon detection. Earlier measurements have shown promise in
their use as photo-detectors, but their prospects have been limited due to poor sensitivity
[127–129]. We explore the possibility of achieving ultra-high sensitivity by combining TIs
with graphene.
58
1.9. Thesis layout
59
Chapter 2
Experimental Techniques
In this chapter, we outline the different methods that have been used to grow the topo-
logical insulators (TIs) investigated in this thesis, their the structural properties and
compositional analysis. This is followed by the detailed description of the techniques
used for fabrication of the field effect transistors (FETs). In the second part, we describe
the setups used for electrical and opto-electronic measurements at low temperatures. We
conclude this chapter with a review of the different measurement techniques that have
been used.
Bi2−x Sbx Te3−y Sey (BSTS) single crystals were synthesized using the modified horizontal
Bridgman technique. Elemental Bi, Se, Sb, and Te purity ≥ 4N were taken in stoichio-
metric ratio in an evacuated quartz tube and liquefied at 850 0 C. The charge was soaked
for 36 hours with intermittent shaking to ensure a homogenous mixture. The melt was
cooled slowly to 550 0 C, and held for 36 hours. Subsequently, when the melt cooled
down to room temperature, mirror-like shiny single crystals were obtained, which were
cleavable along [111] crystallographic plane.
61
Chapter 2. Experimental Techniques
(a) (b)
(c) (d)
Figure 2.1: Structural and compositional analysis of BSTS: (a) Powder XRD
pattern of crushed BSTS single crystals. Inset shows a magnified view of the (107)
and (0012) peaks (b) Single crystal XRD pattern of cleaved crystal of BSTS. Clear 00L
reflections (L = 3n) are observed. (c) Single crystal reflection Laue photograph of BSTS
using Cu K x-rays. Sharp spots and symmetry confirms good single crystalline nature.
(d) EDS spectrum acquired from the cleaved BSTS single crystal.
62
2.1. Material properties and device fabrication
Figure 2.2: Results of EDS quantitative composition analysis of three BSTS samples.
The single crystals of BSTS were characterized by different methods. Fig. 2.1a shows the
powder XRD pattern of the crystal. The pattern is refined and indexed in rhombohedral
structure with R-3m space group. The refined lattice parameters are a = 4.243(1) A0
and c = 29.812(8) A0 which are in agreement with the values reported in literature [133].
The inset of Fig. 2.1a shows a magnified view of (107) and (0012) peaks, which are
characteristic of BTS type chalcogen ordering [133, 134]. This implies that, chalcogen
ordering is preserved to some extent inspite of significant change in stoichiometry. The
XRD pattern along (111) crystallographic plane ((001) plane in the hexagonal setting) of
a cleaved single crystal is shown in Fig. 2.1b. Several 00L reflections are observed, obeying
the systematic absence for the rhombohedral space group in the hexagonal setting. Single
crystal reflection Laue photograph of BSTS using Cu K x-rays is shown in Fig. 2.1c.
Sharp spots and symmetry confirm the good single crystalline nature of the crystal.
63
Chapter 2. Experimental Techniques
The EDS spectrum acquired is shown in Fig. 2.1d. Composition of the as-grown
crystal was analyzed using JEOL JXA-8530F field emission electron probe micro analyzer
with incident beam voltage of 15 KV and current of 50 nA in the wavelength dispersion
mode. Three single crystal samples cleaved from different parts of the grown crystal
were used in these studies. Sample surfaces were cleaned thoroughly before the studies
and more than five different spots were scanned on each sample. Results of the studies
are summarized in Fig. 2.2. The composition is consistent with nominal values of each
sample, and no impurities were detected within the detection limit of electron probe micro
analyzer (EMPA). The single crystals were grown and characterized by Dr. Hariharan
Nhalil and Prof Suja Elizabeth in the Department of Physics, Indian Institute of Science,
Bangalore.
Thin films (∼ 10 nm) of the TI (Bi,Sb)2 Te3 were grown by molecular beam epitaxy
(MBE) on strontium titanate (SrTiO3 ) substrate under ultra-high vacuum [135]. The
purity of the source elements were ≥ 5N. The base pressure for growth was ∼ 2.5 × 10−10
Torr. The growth of the film is monitored by reflection high-energy electron diffraction
(RHEED). Using MBE, large-area, high quality films can be grown. The MBE films
were grown and characterized by Abhinav Kandala, Anthony Richardella, and Prof Nitin
Samarth in State College, Pennsylvania, United States of America.
The out-of plane x-ray diffraction pattern in Fig. 2.3a displays peaks corresponding to
the (000n) planes of the film in addition to the [111] SrTiO3 substrate, suggesting that
the films are grown along the c-axis, and do not form any impurity phases. The high
angle annular dark field (HAADF)-scanning transmission electron microscope (STEM)
image of the sample is displayed in Fig. 2.3b, which shows that the lattice is well ordered.
However, an inter-facial amorphous region of 1 nm thickness grows between the sample
and the SrTiO3 substrate, which can be seen from the TEM image. The chemical
composition was verified with X-ray photo-emission spectroscopy measurements.
64
2.1. Material properties and device fabrication
(a) (b)
SrTiO3
Figure 2.3: Structural analysis of epitaxially grown films: (a) X-ray diffraction
pattern of a (Bi,Sb)2 Te3 thin film grown on SrTiO3 substrate. (b) HAADF-STEM image
of the sample which shows that the lattice is well ordered.
The bismuth telluride (Bi2 Te3 ) nano-wires were grown in two steps:
0.3 mmol of Bi(NO3 )3 ·5H2 O was dissolved in 10 ml of ethylene glycol in a round bottom
flask, onto which 5 ml of Tellurium nano-wire solution was added. The next step was
to add 3 ml of hydrazine hydrate into the above solution, which leads to a reducing
atmosphere. Finally, the solution was placed inside a microwave for heating at 120 0 C
for 3 hours at 150 W power in a CEM DISCOVER-SP microwave.
65
Chapter 2. Experimental Techniques
66
2.2. Fabrication of field effect transistors (FETs)
Table 2.1: Core and shell composition in Te/Bi2 Te3 core-shell wire after 3 hours of
reaction.
Element Core composition (atomic %) Shell composition (atomic %)
Bismuth 43 37
Tellurium 57 63
To obtain flakes for field-effect devices, the exfoliation of crystals was done on Si/SiO2
substrate. The 285 nm SiO2 provides good contrast between the background and the
atomically thin materials, making it easier to identify flakes under an optical microscope
[138]. Prior to exfoliation, the wafers were first cut into small pieces of size ∼ 5 mm×5
mm from a large wafer of 5 inch diameter and cleaned. The following steps were adopted
for cleaning the substrates:
67
Chapter 2. Experimental Techniques
2. RCA1 cleaning: In this step, the wafers were immersed in a solution of NH4 OH
(ammonium hydroxide), H2 O2 (hydrogen peroxide), and DI water (DI = deionized)
in the ratio 1 : 1 : 5, which was preheated to 800 C. This not only removes organic
residues from the surface of the substrate, but also passivates the substrate, which
subsequently leads to a higher yield during exfoliation. The wafers are then rinsed
in DI water (resistivity > 1 MΩcm, from Millipore Milli-DI system), IPA, and blow
dried using a N2 gun.
3. Heating: The final step involves heating the wafer on a hot plate 1200 C. This
removes any moisture from the surface of the wafers.
• Exfoliation of materials
Since topological insulators are prone to oxidation and formation of Se defects [139,
140], exfoliation process was carried out inside a MBraun Glove box in a controlled
environment of 0.3 ppm oxygen and 5 − 10 ppm of water vapor concentration. The
essential ingredients for exfoliation are Si/SiO2 wafers, scotch tape, and the single crystal
of the material to be exfoliated (Fig. 2.5a). For exfoliation, bulk single crystal was first
cleaved on the sticky side of a scotch tape (manufactured by 3M) (Fig. 2.5b). This results
in part of the crystal getting attached tape, which is then cleaved for a few more times
(4 − 5 times) to thin down the attached layers (Fig. 2.5c). It is then made to touch the
SiO2 side of the substrate, and pressed with a teflon tweezer (Fig. 2.5d and Fig. 2.5e).
Because of van der Waal’s interaction between the substrate and the crystal, part of
the bulk crystal gets transferred from the scotch tape to the SiO2 substrate (Fig. 2.5f).
The bright spots on the substrate are large chunks of the crystal which got transferred
from the Scotch tape onto the substrate. To prevent degradation of the surface due
to contact with atmospheres, the wafers were spin-coated one layer of 495K PMMA
(Molecular weight of 495 K, 3 wt% in chlorobenzene) for 30 seconds at 5000 rpm and
baked at 110 0 C for 2 minutes to cure the resist inside the glove box before being taken
out. The entire process is described in Fig 2.5. However, exfoliation of samples required
for hetero-structure assembly were done in ambient conditions using the same process.
68
2.2. Fabrication of field effect transistors (FETs)
Substrate
Tweezer
2 cm 1 cm
2 cm 2 cm 1 cm
Figure 2.5: Steps used in exfoliation of two-dimensional materials: (a) The essen-
tial ingredients required for exfoliation are scotch tape, bulk crystal, cleaned substrate,
and tweezer for wafer handling. (b) The single crystal is first placed between the sticky
sides of two Scotch tapes, and pressed. (c) The tapes are then peeled off, and part of the
crystals get transferred onto the tape. The bright silver spots indicate the transferred
part of the crystal on the tape. (d) The SiO2 -side of the substrate is then placed on
the tape, which has a large coverage of the crystal. (e) The substrate is then pressed
with a Teflon tweezer, or eraser, or with fingers. (f) The substrate is then separated
from the tapes. The bright spots on the wafer indicates that some material has been
transferred from the tape onto the wafer. The wafer is then placed under a microscope
to find suitable flakes.
69
Chapter 2. Experimental Techniques
Suitable flakes of different materials can be identified after exfoliation by using a high
resolution optical microscope (Olympus BX51) with 100x to 1000x magnification. The
lateral dimension of these flakes vary from few microns to hundred microns (Fig. 2.6)
The optical contrast between the substrate and the flakes due to different reflection and
transmission co-efficient helps us in having an approximate idea of the height of the
flakes. The area can be measured by proper calibration for each magnification. It has
been observed that thin (height∼ 10 − 25 nm) flakes are usually a few microns in size.
Atomic force microscopy (AFM) is a type of scanning probe microscopy in which a tip
is raster-scanned over the surface of a sample, to reconstruct its surface topology by
measuring the force between the cantilever and the sample, as they are brought close to
each other [141]. The force can be calculated by measuring the deflection of the cantilever,
usually done by using a laser beam deflection system. The laser is reflected from the back
of the reflective AFM cantilever onto a position-sensitive detector. For making field effect
transistors of topological insulators, having an estimate of the thickness beforehand is
essential. However, since samples are always covered with PMMA to prevent degradation
of surface quality, measuring the thickness before making FETs is not preferred. To
circumvent this, various flakes of different sizes and optical contrast are chosen, using
an optical microscope. The thickness of the flakes are then measured in an AFM. The
samples are then covered in PMMA, and checked in a optical microscope for contrast.
Since these flakes have been subject to prolonged exposure to ambient conditions, they
are not used in devices. However, optical images before and after PMMA coating along
with AFM data gives an estimate of the thickness of other flakes, which are subsequently
used in actual devices. AFM has been used to measure height of samples upto 200 nm
(Fig. A.2). For higher thickness upto 1.5 µm, Dektak XTL stylus profiler has been used.
70
2.2. Fabrication of field effect transistors (FETs)
50x 100x
200x 500x
1000x
71
Chapter 2. Experimental Techniques
Height
10 μm
Scan distance
Figure 2.7: Atomic force microscopy of exfoliated flakes: (a) Optical image of a
typical thin TI (b) AFM image of the same TI flake (c) Line scan along the width of the
TI reveals a thickness of ∼ 12 nm.
72
2.2. Fabrication of field effect transistors (FETs)
500
400
2D
I (arb units)
300
200
G
100
(cm-1)
Figure 2.8: Raman spectra of single layer graphene: Typical Raman spectrum of
a single layer graphene. The ratio of the intensity of the 2D and G peak is ∼ 2 : 1.
ratio of the intersity of the 2D peak and the G-peak. The Raman spectroscopy were
performed in LabRAM HR, Horiba Scientific with confocal arrangement, with a 537 nm
wavelength laser, which was used to excite the sample.
1. Exfoliation of different materials: The desired materials were first exfoliated sepa-
rately on Si/SiO2 substrate. The flakes were then chosen in an optical microscope
and the relative orientation of the final assembly was decided (Fig. 2.9a).
73
Chapter 2. Experimental Techniques
(i) (j)
Figure 2.9: Different steps involved in fabrication of van der Waal’s hetero-
structures: (a) Exfoliation is performed on clean Si/SiO2 wafers, and the materials to
be transferred are chosen. (b) The top-most material in the stack is then alligned under
an optical microscope using a rotatable stage, and heated. (c) The transfer mask is
then contacted to the wafer, around the flake. (d) The temperature is reduced, and the
mask peels of the desired flake from the sample. (e) The second material in the stack
is then aligned under the microscope, below the same transfer mask, and the substrate
is heated (f) The mask is again contacted with the substrate, and the temperature is
reduced. This peels of the second material from the substrate. (g) The temperature
is again reduced, and the mask peels of the second layer of the stack. (h) The transfer
mask is then brought in contact to a RCA-cleaned pre-patterned wafer and heated to 700
C. The mask gets attached to the substrate, which is then dipped in acetone solution,
which dissolves the sacrificial layer in the polymer. (i) After dissolving in acetone, the
hetero-structure is left behind in the substrate. (j) Electrical contacts to the sample are
made by using electron-beam lithography, for performing measurements.
74
2.2. Fabrication of field effect transistors (FETs)
10 μm 10 μm 25 μm 25 μm 25 μm
1. Orient and attach technique: Once the transfer mask is ready, the top-most mate-
rial of the stack, which is on Si/SiO2 substrate, is placed under a microscope (Fig.
2.9b). The substrate is then heated using a local heater, and the temperature is
monitored using a temperature controller. The substrate is slowly brought close to
the mask, which is attached to a specially designed holder below the microscope
objective, and aligned (rotated) according to the final assembly plan. At 55 0 C,
the substrate and the mask are touched together (Fig. 2.9c). The area of contact
can be easily determined from the color contrast. The contact between the sub-
strate and the sacrificial layers is maintained for 2 minutes, after which the heater
is switched off. When the temperature of the stage comes below 490 C, the sac-
rificial layer gets detached on its own from the substrate, peeling of the material
75
Chapter 2. Experimental Techniques
with itself (Fig. 2.9d). The second layer of the hetero-structure is then kept below
the microscope (Fig. 2.9e), and the same process is repeated to pick up the second
and other subsequent layers, if required (Fig. 2.9f). The final assembly is then
contacted onto a pre-patterned SiO2 /Si substrate, at 700 C for 5 minutes, and the
LCC layer gets transferred onto the Si/SiO2 substrate (Fig. 2.9h). The LCC layer
is then dissolved in a solution of acetone and IPA in the ratio of 1 : 3. Due to van
der Waal’s interaction between the hetero-structure and the SiO2 /Si substrate, the
stack remains attached to it (Fig. 2.9i). The entire process takes around 2−3 hours.
Metal contacts can be depositied on the stacks by using electron beam lithography
(Fig. 2.9j).
Once the hetero-structure has been assembled, the next step is to define the contact pads
which are typically ∼ 300 nm wide, and hence electron-beam lithography is essential.
A focused beam of electrons is scanned over the substrate, which is coated with an e-
beam resist. The bombardment of high energy electrons breaks the polymer chains of
the positive resist in the exposed regions, thus changing the solubility of those regions,
which can be dissolved in certain chemicals. To perform e-beam lithography, the wafers
are spin-coated with one layer of 495K PMMA at 5000 RPM for 30 seconds and then
baked for 3 minutes, followed by spin-coating of one layer of 950K PMMA at 5000 RPM
for 30 seconds, and baking for 3 minutes. The lithography is performed in Carl Zeiss
SIGMA FESEM machine equipped with a Elphy Quantum software provided by Raith
GmbH. The wafers were scanned with an electron beam with energy 20 KeV and a
dosage of ∼ 300 C/cm2 at an aperture of 10 µm, 30 µm, and 120 µm, depending on
the feature size required. The beam current for the apertures were approximately ∼ 20
pA, ∼ 230 pA, and ∼ 5.4 nA respectively. After lithography, the exposed regions of
the samples were developed in a solution of isopropyl alcohol (IPA) and methyl isobutyl
ketone (MIBK) in the ratio of 4 : 1, then rinsed thoroughly in IPA and finally blow dried
with a nitrogen gun (Fig. 2.11f). Higher solubility of 495K PMMA than 950K PMMA
leads to an undercut, which improves reliability of lift-off after deposition and device
yield percentage.
76
2.2. Fabrication of field effect transistors (FETs)
2.2.2.2 Metallization
Metallization is the process of depositing metal contacts on the samples after lithogra-
phy. After developing, the samples are immediately mounted in a thermal evaporator
to minimze contact with air and pumped. Thermal evaporation of 5 nm chromium (Cr)
and 50 nm gold is performed at a base pressure of less than 10−6 mbar, at a current
value of ∼ 40 A and ∼ 220 A respectively. Before the actual deposition, the samples
are covered by a shutter and 2 nm Cr is deposited. This not only removes the oxide
layer that forms on top of the Cr pellets, but also improves the vacuum level. 99.99 %
pure gold wire and chromium pellets for thermal evaporation were procured from Kurt
J. Lesker Company (Pennsylvania, USA). After deposition, the samples are taken out
and kept inside a beaker with acetone for 6 hours for lift-off, during which the deposited
metal over the entire substrate, excluding the developed region, peels off (Fig. 2.11g).
The samples are then rinsed in IPA and blow-dried with a nitrogen gun. The different
steps in involved device formation, from hetero-structure fabrication to contact material
depostion is shown in Fig. 2.11a-g.
Devices involving TIs are then covered with a layer of PMMA 495K by spin coating it
at 4000 RPM for 30 seconds. The bonding pads (Fig. 2.12a) are then exposed to e-beam
lithography and developed, which removes PMMA over the metallic bonding pads, such
that the channel remains protected under PMMA.
The penultimate step device fabrication is to appropriately package the device for ease of
operation, for loading it in a cryostat for electrical measurements. The wafer containing
the device is first cut into approximately 4 mm×4 mm area, and placed on a leadless
ceramic chip carrier (sourced from Kyocera, Japan.), using silver epoxy paste. The final
step is to bond the sample by using 25 µm gold wire in a TPT wire bonding machine.
The method of wire bonding used is ball-bonding. The gold wire is placed inside a
ball-bonding tip, and the small length of the wire hangs outside the tip, which is first
melted using electrostatic discharge. The wire is then welded onto the bonding pads
using ultrasonication. Finally, the other end of the wire is contacted to the pads on
77
Chapter 2. Experimental Techniques
BN MoS2 TI
(e) (f) (g)
Figure 2.11: Different steps involved in device fabrication: (a)-(c) Individual flakes
of boron nitride (BN), MoS2 , and TI after exfoliation, which were selected for hetero-
structure fabrication after searching in an optical microscope. (d) Optical image after the
stack is prepared. (e) Optical micrograph of the sample after covering with PMMA. (f)
Optical image of the sample after electron-beam lithography, and subsequent developing
using the developer solution. The violet color in the image corresponds to the SiO2
substrate, which is now exposed due to electron bombardment and developing. (g) The
optical image device image after metallization.
78
2.2. Fabrication of field effect transistors (FETs)
(a) (b)
PCB
Chip carrier
STO
Bonding pads
Indium bonds
Si substrate
5 mm 5 mm
Figure 2.12: Packaging of devices: (a) Optical image of a typical sample fabricated on
Si/SiO2 substrate, which is placed on a chip carrier and ball-bonded. (b) Optical image
of large-area molecular beam epitaxy grown sample after bonding Indium and gold wires.
the chip carrier using ultra-sonication. The device is now ready for measurements. The
optical image of the sample after all these steps is shown in Fig. 2.12a.
Large area, 10 nm thin films of (BiSb)2 Te3 were grown by molecular-beam epitaxy on
SrTiO3 substrate. The devices were mechanically etched into Hall bars, which pre-
vent contribution of contact resistance. Mechanical etching prevents contamination and
degradation of the surface of the films, which can happen if electron-beam lithography
is performed, due to usage of various resists. The devices were stuck onto a specially
designed printed circuit board (PCB) using silver epoxy, and bonded by pressing gold
wires between Indium blobs. One such device after bonding is shown in Fig. 2.12b.
For fabrication of graphene-bismuth telluride (Bi2 Te3 ) hybrids for opto-electronic mea-
surements, a very simple drop-cast method was followed. At first, the graphene-FETs
were fabricated, using the techniques given above in Sec. 2. The electrical characteristics
of the bare graphene-FET was first checked. This is followed by drop-casting of Bi2 Te3
nanowires onto the device. Due to statistical nature of the drop-casting, the process is
repeated multiple times to ensure sufficient coverage of the graphene surface with the
nano-wires.
79
Chapter 2. Experimental Techniques
BNC Connections
to instrument
Pumping port
Electrical connections
(co-axial cables)
from sample
Liquid nitrogen/
Helium dewar Outer jacket
Copper block
for thermalization
Diode
Sample holder
A home-made variable temperature cryostat was used for electrical measurements be-
tween 4.2 K to 300 K, by dipping it directly in a liquid nitrogen dewar or in a liquid
helium dewar. The typical design of a such a dipstick is shown in Fig. 2.13. A long, hol-
low stainless steel cylinder forms the central core of the dipstick, through which stainless
steel co-axial cables (from Lakeshore cryogenics) run. These cables have a 50 ohm im-
pedence, and a low thermal conductivity, which is necessary to reduce the thermal load
80
2.3. Measurement setup
on the sample holder. The top end of the cables are terminated in a vacuum compatible
Fisher connector, which in turn is connected to a BNC connection box, for connecting
the sample to various instruments for electrical read-out. The bottom end of the co-axial
cables are anchored to a PCB, from which copper wires are soldered and wound around
a copper block and stuck with GE varnish, for proper thermal anchoring. The wires
are finally terminated on the connection pins of a sample holder (charntek), which is
attached to the copper block. The copper block has a 25 ohm nichrome wire heater,
which can be used to control the temperature of the sample. The temperature is read
using a DT-670 silicon diode placed at the backside of the chip carrier, which is then
loaded onto the sample holder.
81
Chapter 2. Experimental Techniques
(a) (b)
Figure 2.14: He-3 system: (a) Vapor pressure chart of He3 and He4 . (b) Optical image
of the He3 -refrigerator used for measurements.
is utilized to cool down the system. When He3 is pumped out from the D-phase, He3
from the concentrated phase crosses over the phase boundary to occupy the empty states
to restore the equilibrium. The energy required for restoring equilibrium is provided in
the form of heat from the walls of the mixing chamber, which cools down the system.
The cooling power (Q) is given by the enthalpy difference ∆H of He3 in the two phases,
multiplied by it’s flow rate (ṅ)
Q = ṅ3 ∆H = 84ṅ3 T 2
The design of the optical cryostat is shown in Fig. 2.16. It consists of a long stainless
steel tube, through which cold nitrogen gas is pumped into the upper part of the dipstick
using a diaphragm pump, while the dipstick is immersed in liquid nitrogen. The sample
is thermally connected to copper heat exchangers, through which the cold nitrogen gas
passes, and cools it down. The heat exchangers are zigzag in shape to increase the
effective area of contact between the gas and the exchangers. A radiation shield has
82
2.3. Measurement setup
1K pot
still
(700 mK)
heat
exchangers
mixing
chamber
Figure 2.15: Working principle of dilution refrigerator: (a) Phase diagram of He3
and He4 mixture as a function of He4 concentration. (b) Graphical illustration of the
working of a dilution refrigerator. (c) Optical image of a dilution refrigerator.
also been provided, which ensures low thermal coupling between the upper part of the
dipstick and the outer vacuum jacket. The minimum temperature of the cryostat is
∼ 80 K. The dipstick contains an optical port, onto which LEDs can be mounted for
illuminating the sample for opto-electronic measurements. The dipstick was designed
and made by Dr. Kallol Roy ([147]).
To accurately determine the power of illumination on the sample, a LED power control
unit was made. It consists of a voltage to current converter circuit, which utilizes a
n-p-n transistor in common emitter mode. The circuit diagram of the power control
unit is given in Fig. 2.17. A constant DC voltage source (VS ) is used to power both
the LED and the transistor (ST2N3904) at the collector end. The base voltage (VBE )
is usually provided using the lock-in auxiliary output. When VBE exceeds ∼ 0.7 V, the
transistor switches on, and current flows through it. The load resistance (RL ) determines
the current through the transistor as well as the LED, which is given by
VBE − 0.7
I=
RL
By controlling the lock-in auxiliary voltage (VBE ) as well as RL , the current, and hence
the power through the LED can be varied.
The LED power was measured with a Thorlabs S122C photo-diode, which is mounted
83
Chapter 2. Experimental Techniques
Figure 2.16: Schematic illustration of the optical cryostat. (Adapted from Ref. [147])
84
2.4. Electrical measurement techniques
VBE 10K
ST2N3904
VS 1K
1K 10K
Figure 2.17: Design of the electronic circuit used to control the power of optical illumi-
nation on the sample.
in front of the LED. The photo-diode is connected to a Thorlabs PM200 energy meter,
where the power output can be seen and recorded. The power density is calculated using
Ptotal
Pdensity = 2
π D4
2
Here Ptotal is the power measured by the detector, and π D4 is equal to the total area of
the photo-diode, where D = 9.5 mm is its diameter.
85
Chapter 2. Experimental Techniques
RL
I+
V+
V-
GND
VG
∼ 100 KΩ . The voltage bias across the samples are kept < KB T /e, to prevent any
heating in the sample.
In order to measure noise, a fluctuation quantity, like voltage (δV ) is measured as the
function of time (t), which contains all information about the processes taking place.
The power spectral density. Sv (f ) is defined as the square amplitude of the Fourier
transform of the signal over a time T , for which the signal is recorded, and is given by,
86
2.4. Electrical measurement techniques
-14
10
0.2 Hz
SV (V /Hz) 0.5 Hz
1/f
1 Hz
2
-15
10
2 Hz
5 Hz
10 Hz
-16
10
-1 0 1
10 10 10
f (Hz)
Figure 2.19: The power spectra of a single time-scale showing a Loreztian behavior. The
super-position of several such time-scales leads to the 1/f nature of the spectra.
ˆ−∞
1
SV (f ) = lim | dtδV (t)e−2πf t |2
T =∞ 2T
∞
It can also be defined as the Fourier transform of the auto-correlation function, C(τ )
(Weiner- Khintchine theorem), where
ˆ−∞
1
C(τ ) = lim dtδV (t)δV (t + τ )
T =∞ 2T
∞
The frequency dependence of the PSD arises due to the finite relaxation time of the
fluctuating variable. If time scale of the relaxation is given by the function φ(t), the
power spectral density can be calculated using the expression
ˆ−∞
SV (f ) = C(τ ) dtφ(t)e−2πf t
∞
For a single fluctuator with a characteristic time scale τ , the relaxation function is given
by the Debye relaxation function φ(t) ∼ e−t/τ , and the power spectrum for such a process
is given by the Lorentzian (Fig. 2.19a)
87
Chapter 2. Experimental Techniques
(b)
Resistance
(a)
LOCKIN AMPLIFIER
A B X Y VOUT
time
R1 R2
DAQ-16 BIT DIGITIZER (c)
I+ DSP - DECIMATION Sample noise
-13
V+ 10
SV (V /Hz)
V- POWER SPECTRUM
R3 WELCH PERIODOGRAM
2
-14
X Power Y Power 10
2τ
SV (f ) ∝
1 + (2πf τ )2
If the time-scales of the fluctuations has a distribution given by F (τ ), the PSD becomes,
ˆ∞
2τ
SV (f ) ∝ dτ F (τ )
1 + (2πf τ )2
0
In general, a distribution function of the form F (τ ) ∝ 1/τ α will give rise to a spectrum
SV (f ) ∝ 1/f α . Thus 1/f nature of the spectra can be obtained by adding a few well-
spaced Lorenztians (each corresponding to a different) time scale as shown in (Fig. 2.19b).
The shape of the power spectra and the exponent is dependent on the time-scales of the
physical mechanism which is responsible for the fluctuations, and provides information
about the underlying defect dynamics in a system.
(a) DC-technique
88
2.4. Electrical measurement techniques
In the DC method, a constant current is passed through the sample from a DC source,
and the time series is recorded, often using a pre-amplifier or a low noise transformer.
The PSD of the time series Sv (f ), can be obtained from the Fourier transform of the auto-
correlation function of the time series, using a commercially available spectrum analyzer.
However, the PSD contains contributions from both sample as well as extraneous sources.
To extract the sample noise, fluctuations are recorded at zero-bias, and the PSD of zero-
bias time series is subtracted from the total SV (f ). However, the sensitivity of the
DC-method is limited by the noise floor of the pre-amplifier itself. The background noise
floor can also sometimes shift due to long data collection times. In addition, noise in the
DC-method might require large bias, and can lead to Joule heating of the sample.
(b) AC-technique
Much of the drawbacks of the DC-technique can be overcome by using the lock-in or
AC-technique, which was proposed by Scofield in 1987 [148, 149]. The measurement
schematic is illustrated in Fig. 2.20a, which uses an AC-Wheatstone bridge configuration
to measure both the in-phase and the quadrature component using a dual channel lock-in
amplifier. A constant AC-current I0 sin(f0 t) is passed through the sample. The total
voltage across the sample is given by,
V (t) = VR + 4V (t)
Here, V0 is the voltage drop due to the average resistance (R) of the sample, and 4V (t) =
4VR + 4VBG is the total time-dependent fluctuations in the voltage output, and has
contributions from both sample (4VS ) and the background. (4VBG ). The power spectral
density of the output of the lock-in amplifier is given by
Here SR (f ) is the PSD due to resistance fluctuations in the sample, SBG (f ) is the PSD
due to fluctuations in the background, G0 is the product of gain of the pre-amplifier
and lock-in, f0 is the excitation frequency of the current through the sample, I0 is
the RMS value of the current through the sample and δ is the phase of the lock-in
amplifier detection with respect to the input current. When δ = 0, SV (f, 0) contains
both sample and background contributions. When δ = π/2, SV (f, π/2) = G20 SBG (f0 −f ),
and contains contributions only from the fluctuations in the background. Thus, by
89
Chapter 2. Experimental Techniques
SVA (f )
N F (f, R)(dB) = 10 log10 (2.2)
4kB T R
where SVA (f ) is the output noise power of the amplifier and R is the resistance at the
input of amplifier. Hence the choice of amplifiers is crucial to minimize the background
noise. The noise figure of the voltage amplifier SR560 is shown in Fig. 2.21. For SR560,
the region of the NF with lowest noise, also referred to as the eye of the amplifier, lies
90
2.4. Electrical measurement techniques
Figure 2.22: Filter roll-offs: The response function of the three stage digital filters that
were used. The dashed line represents the amplitude response function of the low pass
filter at the output of lock-in-amplifier having a roll-off of 24 dB/octave. The solid lines
represent the response functions of the intermediate digital anti-aliasing filters.
in the frequency range of 50 Hz-10 KHz, and the source resistance ranges from few KΩ
to few MΩ. Since the samples measured have a resistance in the range of a few KΩ,
and the noise measurement frequency used is ∼ 227 Hz, SR560 has been used as the
pre-amplifier in all our noise measurements.
The measurement set-up used for noise measurements is depicted in Fig 2.20a. It consists
of a Wheatstone bridge, where the sample, which is in series with a current-limiting
resistance (R2 ) forms one arm of the Wheatstone bridge. The other arm of the bridge
consists of a standard resistance (R1 ), and a variable resistance (R3 ). The voltage probes
of the sample are connected to the pre-amplifier (SR560), whose output is connected to
the inputs A of SR830 lock-in amplifier, whereas the voltage across R1 is connected to
the input B of the lock-in amplifier. The variable resistance R3 can be adjusted such
that the values of voltages at the inputs A and B are nearly equal, thereby reducing
the mean value of noise to zero. Typical values of R1 , R2 , and sample resistance are
∼ 10 MΩ, ∼ 1 MΩ, and ∼ 10 KΩ respectively, while R3 can be adjusted from ∼ 100 Ω
to ∼ 100 KΩ. The fluctuations at the lock-in outputs (X and Y ) are digitized using a
16-bit DAQ card from National Instruments (NI USB-6210) at a sampling rate of 1000
Hz. The card collects the data and stores it for a short duration in its internal memory,
91
Chapter 2. Experimental Techniques
Once the data has been acquired, the next two processes are decimation and fast Fourier
transform to calculate the power spectral density.
92
2.4. Electrical measurement techniques
involved downsampling the data by storing every Dth point to bring down the sampling
rate to fs , such thatfs >> 2ω to avoid aliasing, where D is called the decimation factor.
This process is called “decimation”, and is performed in three stages to optimize for
memory usage and speed. For a decimation factor of 64, it is done in three stages of 8,
4 and 2. The roll-off the digital filters were maximum at the third stage (Fig. 2.22). In
this way, we have a signal upto frequency fs /2 without any effects of aliasing.
After decimation, the power spectral density is calculated using “the method of averaged
periodogram” developed by Peter Welch [150]. The algorithm which is followed is given
below:
x(j)
0 x1(j) L-1
D x2(j) D+L-1
93
Chapter 2. Experimental Techniques
L−1
1X
Ak (n) = x(j)w(j) exp(−2kijn/L) (2.6)
L j=0
√
where i = −1. The modified periodograms are obtained for each segment as follows
L
Pk (fn ) = |Ak (n)|2 for k = 1, 2, . . . . . . , K (2.7)
U
where fn = n (2fb /L) for n = 0, 1, . . . , L/2 and U = (1/L) w2 (j) is the normaliza-
PL−1
j=0
tion factor. The frequency resolution in this method is given by 2fmax /L where fmax
represents the maximum frequency in the noise signal. The estimated PSD is given by
K
X
S(fn ) = Pk (fn ) (2.8)
k=1
The spectral length N is the trade-off between frequency resolution and number of av-
eraging.
94
Chapter 3
3.1 Introduction
95
Chapter 3. Bulk-impurity induced noise in large-area epitaxial thin films of
topological insulators
Figure 3.1: Device details: (a) Crystal structure of (Bi,Sb)2 Te3 : each atomic layer
is joined by covalent bonds, whereas each quintuple layer is bonded by van der Waal’s
interaction. Te atoms form the outer two and middle layers, while Bi and Sb atoms form
the other two layers. (b) Schematic of a topological insulator FET on STO substrate.
(c) Optical micrograph of a typical MBE grown sample. The devices are mechanically
etched into Hall bars and contacted using Indium bonding.
at both top and bottom surfaces [110, 163, 164], and also high enough to prevent any
hybridization between two surfaces, which can result in a gap at the Dirac point [? ].
Hence, it is important to evaluate and understand noise in 3D TI devices with thickness
∼ 10 nm.
96
3.2. Device details
(a) (b)
VG = 0 V
5.0 10
RS (K)
R (K)
1500
Rxy ()
4.5
5
4.0
0
0 50 100 150 200 250 300 0 50 100
T (K) VG (V)
Figure 3.2: Basic characteristics: (a) Sheet resistance (Rxx ) as a function of tem-
perature T at VG = 0 V, showing metallic behavior. (b) Gate-voltage dependence of
longitudinal (R) and transverse (Rxy ) resistance in device BST1 at 6.9 K. Rxy was ob-
tained at magnetic field B = −0.5 T.
The devices studied in this chapter were fabricated from thin (thickness, d = 10 nm)
films of (Bi,Sb)2 Te3 (BST) grown by molecular beam epitaxy on h111i SrTiO3 (STO)
substrates with a metallic coating of Indium at the back, that is used as back gate
electrode [135]. (Bi,Sb)2 Te3 is an alloy of Bi2 Te3 and Sb2 Te3 . While Bi2 Te3 tends to be
n-doped, Sb2 Te3 tends to p-doped, and an alloy composed of both of these materials,
leads to compensation doping, making the bulk insulating. The high dielectric constant
of STO substrate at low temperature (T ) makes it possible to tune the Fermi energy
(EF ) from surface to bulk-bands. The crystal structure of (Bi,Sb)2 Te3 is shown in Fig.
3.1a. Each quintuple layer consists of five atomic layers, which are joined by covalent
bonds, whereas each quintuple layer is joined by van der Waal’s interaction. While
Te atoms form the outer two and middle layers, Bi and Sb atoms form the other two
layers. All the measurements, except T -dependence of 1/f -noise, were performed in the
device BST1. The as-grown thin film was mechanically etched into a Hall bar with a
channel length (L) of 1 mm and width (W ) 0.5 mm as shown in Fig. 3.1b. While non-
invasive Hall bar geometry minimizes contribution of contacts in electrical transport and
noise measurements [165, 166], the mechanical etching process prevents the film from
being exposed to chemicals, such as resists and solvents associated with electron-beam
lithography, resulting in a superior surface quality. The large dielectric constant of the
97
Chapter 3. Bulk-impurity induced noise in large-area epitaxial thin films of
topological insulators
(10 mho)
4
-4
6.9 K
2
-2 0 2
13 -2
n (10 cm )
STO substrate at cryogenic temperatures allows effective electrical back gating for tuning
the chemical potential in the sample [164, 167, 168]. The actual device image is shown
in Fig. 3.1c.
−1
RS = (R0 + AT )−1 + (ρ exp(∆/KB T ))−1 (3.1)
Here R0 , A, ρ, and ∆ are the fitting parameters respectively. The first term in the
contribution due to the metallic surface states and the exponential term is due the bulk,
which is activated in nature. The dashed green line in Fig. 3.2a is the fit to the data
98
3.3. Basic characterization
l (nm)
2V
2
10V
26V
46V
66V 100
82V
-0.5 94V
-0.5 0.0 0.5 -50 0 50
B (T) VG-VD (V)
according to Eq. 3.1, which qualitatively explains the nature of R vs T in the sample.
The R-VG characteristics at T = 6.9 K, as shown in Fig. 3.2b, demonstrates graphene
like ambipolar transport, with Dirac point or charge neutrality point at VG = 66 V,
indicating that the sample is intrinsically hole-doped. Such high tunability of EF is
possible due to high dielectric constant of the STO substrate at low T . The transverse
resistance Rxy , as demonstrated in Fig. 3.2b, shows a change in sign across the Dirac
point, firmly establishing the tunability from hole-type to electron-type conduction in
the sample with VG .
n 2
σ∼E e /h for n > n∗ (3.2)
ni
and
99
Chapter 3. Bulk-impurity induced noise in large-area epitaxial thin films of
topological insulators
(a) 1.6
(b) (c)
VG = 72 V 50 0V 1 17.5 K
n (10 m )
6.9 K
-2
20V
Rxy (K)
42V
Rxy (K)
14
1.4 60V
64V
0
66V
80V
90V 0
1.2
100V
-50
-2 -1 0 1 2 -0.2 0.0 0.2 -100 -50 0 50
B (T) B (T) VG -VD (V)
n∗ 2
σ∼E e /h for n < n∗ (3.3)
ni
Here n∗ is the residual carrier density in electron and hole puddles, and E is a constant
that depends on the Wigner–Seitz radius rs . The extracted value of the number density
of Coulomb traps is ni = 1 × 1014 cm−2 , which is quite high compared to typical den-
sity of Coulomb traps at oxide-channel interface, such as in graphene-FETs [170], but
matches well with the number density of charged disorders in TI-devices [164], and can
be attributed to the bulk charged-defects in the TI.
e2
1 Bφ Bφ
4σ = −α ψ + − ln (3.4)
πh 2 B B
100
3.3. Basic characterization
where τφ , τso , τe are the phase coherence or dephasing time, spin-orbit scattering time
and elastic scattering time respectively, ψ is the digamma function, and Bφ is the phase
breaking field. Here, α and Bφ are fitting parameters. The phase coherence length
lφ can be extracted using lφ = ~/4eBφ . The value of α gives an estimate of the
p
number of independent conducting channels in the sample. The phase breaking length
shows an increase away from the Dirac point, as shown in Fig. 3.4b. As the number
density of carriers increases away from the Dirac point, it leads to enhanced screening of
electromagnetic fluctuations, leading to reduced dephasing, thereby showing an increase
of the phase breaking length (lφ ). Such increase in lφ has been observed in other Dirac
Fermionic systems such as graphene as well [31, 79, 167].
The carrier number density has been extracted by measuring the transverse Hall
resistance (or voltage) at different back-gate voltages. The raw data is shown in Fig. 3.5a,
where the transverse resistance has been plotted as a function of B. However, since the
Hall bars are mechanically etched, the measured resistance also contains a longitudinal
contribution due to imperfect nature of the etching and needs to be corrected for such
asymmetry. The transverse resistance Rxy as a function of B for different VG , is shown
in Fig. 3.5b, which demonstrates a linear behavior after it is corrected for asymmetry
using Rxy = R(B)−R(−B)
2
. The Hall data clearly indicates a change in sign of the slope as
the VG is tuned from the hole side to the electron side. The Hall number density (nHall ),
extracted using RH = − nHall
1
e, is shown in Fig. 3.5c, as a function of (VG −VD ), where RH
is the Hall co-efficient. The measured intrinsic Hall number density is nHall = 0.9 × 1014
cm−2 (at T = 6.9 K and VG = 0 V), which is high compared to the typical number
density of surface charge carriers in TI [167], indicating that the Fermi energy is located
within the valence band, and holes in the bulk of the material are mainly responsible for
such metallic behavior [164]. This intrinsic high doping leads to a high value of the gate
voltage (VG = 66 V) at which the charge neutrality point is observed, which is identified
by a maximum in R and a change of sign in transverse resistance Rxy as shown in Fig.
3.2b.
101
Chapter 3. Bulk-impurity induced noise in large-area epitaxial thin films of
topological insulators
5 =
-2
n (10 cm )
-2
C = 7.78e-4 F/m
13
= 144000
0
-2
C = 2.5e-3 F/m
20 40 60 80
VG (V)
ne = CV . In the hole-side, the linear fit of this data gives us the value of capacitance
C = 2.5 × 10−3 Fm−2 , which corresponds to a dielectric constant value ε = 144000. In
the electron-side, we obtained C = 7.78 × 10 −4 Fm−2 and ε = 44000. As ε = 144000
is unphysical, and ε = 44000 is close to experimentally reported values of dielectric
constant in STO, we have used C = 7.78 × 10 −4
Fm−2 for our calculations. The
non-linearity in the n vs VG data at higher gate voltages can be due to the non-linearity
of the capacitance STO substrate at high electric fields [167].
An AC four-probe Wheatstone bridge technique was used for noise measurements [149,
171]. The voltage fluctuations, as shown in Fig. 3.7a, were recorded as a function of time
using a 16-bit digitizer. The time-series data were then digitally processed to obtain the
power-spectral-density (PSD, SV ) as a function of frequency (f ) (Fig. 3.7b). In both
devices, we found SV ∝ 1/f α , where the frequency exponent α ≈ 1–1.2. The SV shows
a quadratic dependence with bias (V ) (Fig. 3.7c), which ensured that all measurements
are in the Ohmic regime, and no heating effects were present.
102
3.4. 1/f noise measurements
SV (10-15 V2Hz-1)
10
SV (V /Hz)
-15
V (10 V)
10
0.2
-8
2
5
1
0 10
-17
1/f 65 V
0V 0V
65 V 65 V
-5 0.0 0
0 50 100 0.01 0.1 1 10 0 2 2 10
V (V )
time (s) f (Hz)
Figure 3.7: 1/f noise measurements: (a) Voltage fluctuations (δV ) in time domain in
sample BST1 at 6.9 K and VG = 0 V and 65 V, obtained with a fixed biasing current. (b)
Typical voltage power spectral density (SV ) obtained from time- dependent fluctuations
shown in (a), indicating 1/f type characteristics. (c) SV at 1 Hz for as a function of V 2 ,
showing linear characteristics at the same values of VG .
of effective gate voltage (VG − VD ) at 6.9 K and 17.5 K is shown in Fig. 3.9c. It is
minimum close to the Dirac point (VG − VD = 0), but increases rapidly (∼ 10 times)
with increasing |VG − VD |, and reaches a maximum (at (VG − VD ) = −33 V, at T = 6.9
K). With further increase in |VG − VD |, noise magnitude reduces. When correlated with
the number density as shown in Fig. 3.9d, it is clearly seen that the noise magnitude is
enhanced for |nHall | & 1013 cm−2 on the hole side, which is the number density associated
with the edge of the bulk valence band [164, 167], demonstrating the enhanced impact
103
Chapter 3. Bulk-impurity induced noise in large-area epitaxial thin films of
topological insulators
(a) (b) 12
SV / V (10 Hz )
-1
-12
R (K)
9 25 10
R (K)
2
6 8
0
-100 0 100 0 150 300
VG (V) T (K)
3.5 Discussion
104
3.5. Discussion
(a) (i)
(iii)
t b t
b
b t
(ii)
(b) 6
RS (K / sq)
BST1
2
(c) 6.9 K
Hz )
17.5 K
-1
-12
5
SV / V (10
2
0
(d) 1
nHall (10 cm )
-2
nHall (10 cm )
-2
1
13
0
14
-1
0 50
0
17.5 K
6.9 K
-100 -50 0 50
VG-VD (V)
Figure 3.9: Gate-voltage dependence of noise: (a) Schematic showing Fermi energy
at top (t) and bottom (b) surfaces of TI film at different ranges of VG . Region (i)
indicates dominant hole contribution from surface and bulk states. Region (ii) indicates
the charge puddle dominated regime close to the Dirac points for both the surfaces.
Region (iii) indicates the surface-electron dominated regime for both the surfaces. (b)
Sheet-resistance (RS = WLR , where W , L and R are width, length and resistance of the
channel respectively), (c) Normalized voltage power spectral density V 2 and (d) Hall
SV
number density of charge carriers (nHall ) as a function of effective gate voltage (VG − VD )
at 6.9 K and 17.5 K. Inset of (d) highlights into the low number-density area, the arrows
highlight the extrema. The dashed lines indicate the onset of bulk conduction where
nHall = 1013 cm−2 .
105
Chapter 3. Bulk-impurity induced noise in large-area epitaxial thin films of
topological insulators
(a) (b)
E Surface
Bulk
∆𝐸
Valence EF
Band Surface
Figure 3.10: Mechanism of noise: Schematic showing two different noise mechanisms.
(a) Generation-recombination of charge-carriers between valence band and defect state.
(b) Reorganization of the charge carriers in the bulk of the material, which leads to
mobility fluctuations.
106
3.5. Discussion
(a)10-15 (b)
42 K
46 K
1
50 K
fSV/V2(a.u)
fmax (Hz)
0.1
10-16
0.01
0.01 0.1 1 0.020 0.022 0.024
f (Hz) 1/T (K-1)
In order to unravel the microscopic framework of the noise, we have fitted the VG -
dependence of SV
V2
data (Fig 3.12) using the framework of correlated mobility-number
density fluctuations model [105]. According to this model, the noise in the channel is
affected by the trapping-detrapping of charges in gate dielectric-channel interface (Fig.
3.10b). The main components of the noise are,
1. The fluctuations of the number of charge carriers in the channel due to tunneling to
and from the oxide traps at the channel-gate dielectric interface.
2. The mobility fluctuations caused by the Coulomb scattering induced by fluctuating
trapped charges (Fig. 3.10b).
The total noise can be expressed as,
2
SV Dit kB T dσ J1 J2
2
= + + J3 (3.5)
V dW L dn σ2 σ
´ τT
where J1 = represents a pure number fluctuation, J3 = A2 (x)
1
8α 2 dx represents
1+(2πf τT )
´ τT
pure mobility fluctuations and J2 = 2A(x) 2 dx represents combined number
1+(2πf τT )
and mobility fluctuations. Here, α is the decay constant for the spatially decaying time
constant τT of a typical trapping event, and A(x) is the scattering constant, which can
be evaluated using phenomenological values [105]. Dit , kB , W , L, σ, n, x are the areal
107
Chapter 3. Bulk-impurity induced noise in large-area epitaxial thin films of
topological insulators
6.9 K
Hz ) 17.5 K
-1
6.9 K
17.5 K
-12
5
SV / V (10
2
0
-100 -50 0 50
VG - VD (V)
VD ) at 6.9 K and 17.5 K. The solid lines show fits according to Eq. 3.5.
trapped charge density per unit energy, Boltzmann constant, width of the channel, length
of the channel, conductance and number density of charge carriers, axis in the direction
perpendicular to the channel respectively, f = 1 Hz frequency and d = 1 nm is the
distance over which the tunneling is effective. The data at T = 6.9 K and 17.5 K were
fitted independently to extract the values of Dit .
The time constant associated with
q a trapping event is given by, τT = τ0 exp(2αx).
∗
Typically, τ0 = 10−10 s and α = 2m~e2Φb , where m∗e is the effective mass of electron in
STO, and Φb is the height of the potential barrier at the interface. Using, m∗e = 9.8m0
(m0 is the free electron mass), and Φb = 0.25 eV, we find α = 8 × 109 m−1 . The
scattering rate S(x) can be expressed as S(x) = 2S0 ln(rmax ⁄ x), where the scattering
rate constant is S0 = 2.56 × 102 V.s/C (the value experimentally obtained for SiO2 ),
rmax is the distance above which the scattering charges seem to form a continuum and
the image terms cause the Coulomb scattering potential to fall off more rapidly than
1 ⁄ r. Using the relation rmax ≈ εox +εchannel
COX
, we obtain rmax = 5 × 10−4 m. The calculated
values of J1 , J2 , J3 in S. I. unit are listed in Table. 3.1. We found Dit = 3.2 × 1017
108
3.5. Discussion
-13 T
10 -7 T
0T
0
-50 V -V (V) 0
G D
cm−2 eV−1 at T = 6.9 K, which is several orders of magnitude higher compared to typical
trap density at oxide surface [105, 175]. However, considering an energy window of KB T ,
the number of activated trapped states, nI = 1.3 × 1014 cm−2 , which matches with the
number density of Coulomb traps ni = 1 × 1014 cm−2 , extracted from the fitting of the
σ-n (Fig. 3.3), showing that the extracted value of Dit is not an artifact, and hence
this high density of charge impurities can be attributed to intrinsic bulk defect states
in (Bi,Sb)2 Te3 [164]. Similar Dit (8 × 1017 cm−2 eV−1 ) was found for 17.5 K. The origin
of the slight elevation of noise magnitude at 17.5 K can be attributed to the thermal
activation of the charged-defects.
109
Chapter 3. Bulk-impurity induced noise in large-area epitaxial thin films of
topological insulators
1.5 -4.2 V
-19.2 V
Hz )
-1
-29.2 V
1.0 -59.2 V
-11
SV /V (10
2
0.5
0.0
-10 -5 0
B (T)
SV SV
(B) = (0) 1 + (βµ H B)2
(3.6)
V2 V2
Here β is a fitting parameter. Although µH B 1 in our devices, the extracted value of
β = 7.5 is quite high compared to the theoretical value expected for charge or mobility
fluctuation noise [177].
In the absence of a theory which accounts for the magnetic field dependence of cor-
related mobility-number density fluctuation noise for a doped semiconductor, we cannot
comment whether this strong dependence on magnetic field results from simple geomet-
ric effects or is a consequence of the topological nature of the charge-carriers. However,
we note that the behavior does not change as the chemical potential is tuned through
surface and bulk bands of topological insulators.
Sv γH
2
=
V Nfα
110
-9 -8 -7 -6 -5 -4 -3 -2 -1 1
10 10 10 10 10 10 10 10 10 1 10
111
HOOGE PARAMETER
Graphene
SV ( f ) H 1D Si:P Wires Si-MOSFET Single layer
= Bi2Se3 MoS2 Bilayer
2 Nf
V CVD
MBE TI: Bi2Se3 / (BiSb2)Te3 Metal films
Black Phosphorus CNT Suspended
TOTAL NUMBER OF CARRIERS
Multi-layer
3.5. Discussion
Here γH and N are the Hooge parameter and the total number of carriers respectively.
The Hooge parameter for MBE-grown (BiSb)2 Te3 , γH ∼ 1, which is very high compared
to some of the other known two-dimensional materials, and metal films, consistent with
the high density of defects obtained from conductivity or noise measurements.
3.6 Summary
1. The results of our experiments clearly indicate that even in thin films of topological
insulator (BiSb)2 Te3 with thickness as low as ∼ 10 nm, the intrinsic bulk chalco-
genide defects are mainly responsible for resistance fluctuations for both surface
and bulk electronic states. This is in contrast to other two-dimensional materials,
where the noise originates either due to trapping-detrapping at the channel-oxide
interface or at the contacts.
2. When the Fermi energy is in the bulk valence band, generation-recombination pro-
cesses between the valence band and the impurity states are the origin of resistance
fluctuations. The peak in noise at T = 50 K is a consequence of this.
3. When the Fermi energy is tuned towards the surface bands, the noise is caused
by correlated mobility-charge density fluctuations, due to trapping-detrapping of
surface charge carriers by the charged-defects in the bulk of the crystal. This
scenario is different from our earlier study of noise in thicker exfoliated topological
insulator devices where generation-recombination of charge carriers at the bulk
gives rise to a fluctuations of the potential landscape at the surface, which results
in a mobility fluctuation noise.
112
Chapter 4
4.1 Introduction
Topological insulators (TIs) [2, 12, 13, 178] at zero magnetic field are time reversal-
invariant systems, which are characterized by surface states with a linear band struc-
→
−
ture. The Hamiltonian (H) for such surface states is given by H = ~vF → −
σ · k , which
belongs to the AII/symplectic universality class, where vF , ~σ , and ~k are the Fermi ve-
locity, Pauli spin-matrices and momentum respectively (Fig. 4.1). This is also known as
the Anderson universality class for non-relativistic particles in the presence of a random
spin-orbit coupling, where time reversal symmetry (TRS) is preserved [179]. One of the
key features of TIs is that the symplectic symmetry can be changed to unitary symmetry,
when time reversal symmetry (TRS) is lifted, and the topological phase transition oc-
curs. The addition of an external magnetic field or ferromagnetic impurities introduces a
Zeeman/orbital term in the Hamiltonian, and breaks the TRS which results in this topo-
logical to trivial phase transition in the bulk states [12], and manifests as a gap opening
in the linear surface states (Fig. 4.1). The magnitude of the gap is ∼ gµB B, where g,
113
Chapter 4. Universal conductance fluctuations and direct observation of
crossover of symmetry classes in topological insulators
SYMPLECTIC UNITARY
µB , and B are the Lande g-factor, Bohr magneton, and magnetic field respectively. Ex-
perimentally, the sensitivity of the surface states to magnetic impurities [26, 180] or the
saturation of the phase breaking length at low temperatures can be directly connected
to the TRS in TI systems [26, 35, 49, 76]. This makes an explicit demonstration of the
symplectic to unitary crossover an important task, which has however, not been achieved
yet.
One direct method to probe such crossover of symmetry classes is universal conduc-
tance fluctuations (UCF) [37–39]. The magnitude of UCF is strongly determined by the
physical symmetries of the Hamiltonian, and within the framework of random matrix
theory, the magnitude is proportional to [39]
2
e2 ks2
2
hδG i ∝ (4.1)
h β
Here β, s and k are the Wigner-Dyson parameter, Kramer’s degeneracy and the number
of independent eigen modes of the Hamiltonian respectively. The various parameters
relevant for the two classes important for TI is depicted in Table 4.1. The UCF-based
technique has been previously used as an experimental probe in mesoscopic samples of
graphene [94, 136], where a factor of four reduction was observed in the magnitude as
a function of number density due to crossover from symplectic to orthogonal classes
114
4.2. Device details
Table 4.1: Values of symmetry parameters for the two classes relevant for TI. k, s, β and
Hij are the number of independent eigenvalues of the Hamiltonian, Kramer’s degeneracy,
Wigner-Dyson parameter and the elements of the Hermitian matrix representing the
Hamiltonian respectively.
2 2 2
Ensemble TRS k s β Hij hδG2 i ∝ eh ksβ
Symplectic Yes 1 2 4 real quarternion 1
Unitary No 1 1 2 complex 0.5
[42]. Similar reduction of noise with magnetic field were observed in metal films [32, 40,
181, 182], metallic single crystals of silicon [183], and also in δ-doped silicon-phosphorous
systems [41, 184]. Though a symmetry class crossover in TI has been induced by addition
of ferromagnetic impurities and inferred from weak anti-localisation [80], a more direct
observation of the symmetry class and its crossover on breaking TRS, which does not
require any addition of exterrnal impurities remains experimentally elusive.
In this chapter, we have probed the intrinsic symmetry class of topological insulators
by measuring the mesoscopic conductance fluctuations in the TI Bi1.6 Sb0.4 Te2 Se, which
shows an exact factor of two reduction on application of a magnetic field due to crossover
from symplectic to unitary symmetry classes. The reduction provides an unambiguous
proof that the fluctuations arise from the universal conductance fluctuations (UCF), due
to quantum interference and persists from T ∼ 22 mK to 4.2 K. We have also compared
the phase breaking length lφ extracted from both magneto-conductivity and UCF, which
agree well within a factor of two in the entire temperature and gate voltage range. Our
experiment confirms UCF as the major source of fluctuations in mesoscopic disordered
topological insulators, and the intrinsic preservation of time reversal symmetry in these
systems.
115
Chapter 4. Universal conductance fluctuations and direct observation of
crossover of symmetry classes in topological insulators
R (K)
BN 5.0
Si02/Si BN
VBG 10 μm
-30 0 30
VG (V)
Figure 4.2: Device details: (a) Schematic of a topological insulator FET on boron
nitride. (b) Optical micrograph of a typical topological insulator on boron nitride field-
effect transistor device. (c) Resistance as a function of gate-voltage of the device at
different temperatures.
TI channel [145, 165]. The details of hetero-structure fabrication are given in Chapter 2,
Sec. 2. The quaternary alloy Bi1.6 Sb.4 Te2 Se offers reduced bulk number density due to
compensation doping, which results in enhanced surface transport, and a more insulating
bulk, with experiments showing ∼ 99 % surface transport in this material [19, 133].
After hetero-structure fabrication, the contact pads were defined by standard e-beam
lithography. This was followed by metallization using thermally evaporated 5/40 nm
Cr/Au. The device is schematically shown in Fig. 4.2a. The optical micrograph of an
actual TI on BN field-effect transistor device is shown in Fig. 4.2b. The sample was
covered with PMMA (poly(methylmethacrylate)) during the entire measurement cycle.
This protects the sample from coming in contact with air, and prevents the subsequent
degradation of surface quality. The length (L) and width (W ) of the sample are both
∼ 2 µm. All measurements from T = 22 mK to T = 4.2 K were done in a dilution
refrigerator.
116
4.4. Weak anti-localization in TIs
(a) (b) 1
22 mK 20 mK
0.0 4.2 K
-60 V
0.8
(e2/h)
-25 V
60 V
0.6
-0.5
CS (VG −VD )
number density calculated at VG = 0 V using the relation n = e
is −2.9 × 1016
m−2 , showing the sample is intrinsically n-doped. Here CS = 1.14 × 10 −4
F is the series
capacitance of 285n.m thick SiO2 , and the 14n.ṁ-thick hBN layers. The n-doping in the
sample arises due to Se vacancies [185].
e2
1 Bφ Bφ
4σ = −α ψ + − ln (4.2)
πh 2 B B
where τφ , τso , τe are the phase coherence or dephasing time, spin-orbit scattering time
and elastic scattering time respectively, ψ is the digamma function and Bφ is the phase
breaking field. Here α and Bφ are fitting parameters. The phase coherence length
lφM R can be extracted using lφM R = ~/4eBφ . The value of α gives an estimate of the
p
117
Chapter 4. Universal conductance fluctuations and direct observation of
crossover of symmetry classes in topological insulators
9.5
1. Polynomial fit
118
4.5. Universal conductance fluctuations
R (K)
600 mK
R (K)
10 100 mK
25 mK
9 8.8
Figure 4.5: Features of conductance fluctuations: (a) R-VG in a small gate voltage
window of 4 V for four different temperatures used to calculate hδG2 i by using a smooth
polynomial fit. The curves have been offset for clarity. (b) R-VG for two different
runs, showing the reproducibility of conductance fluctuations arising due to quantum
interference.
of the residual of the fit gives the value of hδR2 i and the mean value corresponds to
hRi. hδG2 i is then obtained using the relation hδG2 i = hδR2 i/hRi4 . The different steps
involved in the evaluation of conductance fluctuations magnitude is given in Fig. 4.4b.
119
Chapter 4. Universal conductance fluctuations and direct observation of
crossover of symmetry classes in topological insulators
(a) 6 (b)
20 mK -60 V
G2 /G mK
4.2 K ~ T-1
G2 10-3 (e2/h)2
4 1
0.1
0 30 60 0.1 1 10
VG - VD (V) T (K)
of scattering due to bulk defects, which are known to be the dominant sources of noise
in TIs [95–97, 190, 191].
2
e2
1 Ly
2
hδG i ' α(kF δr) ns (T )lφ4 (4.3)
h kF l L3x
where kF , l, Lx and Ly are the Fermi wave-vector, mean free path and sample dimen-
sions in x and y directions respectively. α(x) represents the change of the phase of
120
4.6. Crossover of symmetry classes
8 20 mK
R (K)
-60 0 60
VG (V)
121
Chapter 4. Universal conductance fluctuations and direct observation of
crossover of symmetry classes in topological insulators
(a)
1.0
-25 V
N(B)/N(0)
-60 V
22 mK 22 mK
0.5
-60 V -25 V
N(B)/N(0)
4.2 K 4.2 K
0.5
122
4.7. Discussion
(a) (b)
1.0 1.0
N(B)/N(0)
N(B)/N(0)
60 V 60 V
20 mK 4.2 K
0.5 0.5
field scales for quantum interference effect (Bφ ), reported for TIs [76, 87, 192]. Since
Bφ ∼ ~/4elφ2 , the increase in field scales with T can be readily attributed to the decrease
in lφ with increasing T . Although lφ ∝ T −1/2 expected from electron-electron scattering
would lead to much larger change in Bφ in the experimental temperature range, we
believe a saturation in lφ [50, 55] limits the decrease in Bφ at low temperatures. The
factor of two reduction proves that the time reversal symmetry is intrinsically preserved
in topological insulators. While at VG = −60 V and VG = −25 V, we see a clear reduction
of the conductance fluctuations magnitude, VG = 60 V is marked by the absence of this
reduction, as shown in Fig. 4.9, for both T = 20 mK and 4.2 K.
4.7 Discussion
• Physical origin of the factor of two reduction
The physical origin of the factor of two reduction is related to the number of indepen-
dent low-energy independent eigen-modes contributing to transport. In the symplectic
class, there are two modes: (a) the Diffuson (particle-hole channels), and (b) Cooperon
(particle-particle channels) modes, which correspond to the direct and self-intersecting
paths respectively between two points (Fig. 4.10). The Diffuson modes are the classical
trajectories, while the Cooperon modes arise due to quantum mechanical interference.
123
Chapter 4. Universal conductance fluctuations and direct observation of
crossover of symmetry classes in topological insulators
Diffuson Cooperon
Figure 4.10: Physical origin of factor of two reduction: Graphical illustration of the
direct Diffuson channels and the self-intersecting Cooperon channels. In the presence of a
magnetic field, the Cooperon modes are suppressed, leading to the factor of two reduction
in the conductance fluctuations magnitude.
The introduction of a magnetic field leads to an additional phase between the original
´
and the time-reversed path ∝ A · dl, which is zero for Diffusons, but non-zero for
Cooperons. Here A is the magnetic-vector potential and l corresponds to an infinites-
imal path length. Thus the Cooperon contribution gets eliminated due to incoherent
superposition between different loops. This reduction by half in the number of transport
modes, is also manifested in the universal conductance fluctuations in the sample as a
reduction by a factor of two.
The absence of the factor of two reduction at high positive voltages is not due to
spontaneous removal of time reversal symmetry, which has been previously seen in
phosphorous-doped silicon systems [184]. The absence of the factor of two reduction
is most likely due to additional noise contributions in the system. As EF is tuned to-
wards the bulk bands, trapping-detrapping processes from the charged impurities in the
bulk, which are independent of B, and are known to be dominant source of noise in TI
increases [95–97, 190, 191], which diminishes the factor of two reduction.
The reduction is independent of T -scales that are either large (KB T ∼ 0.362 meV
at 4.2 K) or very small (KB T ∼ 0.00172 meV at T = 20 mK) compared to the Zeeman
energy scale (4E = gµB B = 0.156 meV at B = 100 mT), where g = 27 is the Lande
g-factor and µB is the Bohr magneton [193].
124
4.7. Discussion
∞
1 1 X 1
v(B, T ) = + 2 (4.4)
2 b n=0 [ n + 2 + 1b ]3
1
1 1 00 1 1
= − 2Ψ + (4.5)
2 2b 2 b
00
Here b = 8πB(lφ )2 /(h/e) and Ψ is the double derivative of the digamma function and
lφ is the fitting parameter. The solid lines in Fig. 4.8 are fits according to Eq. 4.5, which
captures the variation of conductance fluctuations magnitude with B well, especially at
gate voltage values close to the Dirac point. The corresponding lφ , extracted from fitted
value of b, are plotted in Fig. 4.11a (solid circles).
For gate-voltages, where a full factor of two reduction is not observed, a modified
crossover function ν 0 (B, T ), given by
Here c is the total fraction of the noise originating from UCF. The decrease in the reduc-
tion allows us to estimate ≈ 28 % of the observed conductance fluctuations magnitude to
arise from UCF at T = 20 mK (Fig. 4.9a) at VG = 60 V but becomes negligibly smaller
at T = 4.2 K (Fig. 4.9b).
Finally to ascertain the consistency of the analysis, we have evaluated the VG -
dependence of lφ from three different methods:
(c) magneto-conductance.
We have however, restricted the calculation of the phase breaking length between VG =
−60 V and 0 V, where UCF is the dominant source of noise. lφ extracted directly from
hδG2 i at B = 0 T using the expression [179, 189, 196]
2 2
e2
3 lφ
2
hδG i ' (4.7)
π h L
125
Chapter 4. Universal conductance fluctuations and direct observation of
crossover of symmetry classes in topological insulators
300
(a) 20 mK
100 mK
(b) 300
1K
200 4.2 K
(nm)
200
l (nm)
100
lMR
20 mK
100 mK
100 1K
4.2 K
0 30 60 0 30 60
VG - VD (V) VG - VD (V)
Figure 4.11: Comparison of phase breaking length: (a) lφ extracted from conduc-
tance fluctuations magnitude as a function of gate-voltage (VG ) at different temperatures
(T ) from (i) magnetic field-dependence using Eq. 4.5 (solid circles), and (ii) directly at
B = 0 T using Eq. 4.7 (hollow circles) (b) lφM R extracted from magneto-conductance
using Eq. 4.2 as a function of VG for different T . The phase breaking length obtained
from three different methods agree well within a factor of two.
is shown in Fig. 4.11a (hollow circles). lφM R extracted from magneto-conductance using
Eq. 4.2, is shown in Fig. 4.11b. We find that lφM R and lφU CF are similar in magnitude within
a factor of two, rendering validity to the factor of two reduction and the corresponding
analysis. lφM R increases away from the Dirac point, similar to the trend of lφ obtained
from UCF (Fig. 4.11a). Near the Dirac point, the system is highly inhomogeneous with
the presence of electron-hole puddles. As the gate voltages is tuned away from the charge
neutrality point, the carrier concentration increases leading to enhanced screening, which
suppresses dephasing due to electromagnetic fluctuations. This leads to a larger phase
breaking length away from the Dirac point [31, 79, 167]. The factor of two difference in
lφ obtained from these two methods can arise because this expression is valid in the case
of Lφ << L,LT [179, 196], where LT is the thermal length. The phase breaking time τφ
and hence, the phase breaking length lφ = Dτφ (D is the electron diffusivity) relevant
p
for weak localization is related to the Nyquist dephasing rate [46]. In case of UCF, the
relevant scattering time for 2D is the out-scattering time, which maybe related to the
inverse of the inelastic collision frequency and differ by a logarithmic factor compared
to the WL rate [30, 32, 41, 197–199]. While both lφM R and lφU CF are of the same order,
suggesting they are governed by the same mechanism, the noise in the data prevents a
conclusive claim in case of topologically protected systems.
126
4.8. Summary
4.8 Summary
1. We have measured the Fermi energy dependent aperiodic and reproducible fluctu-
ations in mesoscopic topological insulator systems, by scanning the Fermi energy
using a gate-voltage.
3. The factor two of reduction occurs due to a transition from symplectic symmetry
class to unitary symmetry class. Physically, the application of a perpendicular
magnetic field suppresses the Cooperon contribution, leading to the reduction.
127
Chapter 5
5.1 Introduction
Topological protection of the surface states make topological insulators (TIs) a strong
contender for the building blocks of qubits, which require long phase coherence length (lφ )
for error free quantum computation. Hence, it is critical to understand the mechanisms
responsible for dephasing or decoherence, which is equivalent to loss of information, in the
surface states of TIs. The most common dephasing mechanism in TIs at low temperature
(T ) has been known to be electron-electron interaction [74, 75, 83]. However, Li et
al. have demonstrated that electron-phonon interaction is also required to explain the
dependence of lφ on T [85]. Liao et al. have shown the origin of dephasing to be the
coupling of surface states and localized charged puddles in the bulk [76]. All of these
mechanisms lead to a diverging lφ with decreasing T [45, 75, 83, 132]. However, this
increase is often followed by a saturation of lφ for T ≤ 2 − 5 K [76, 84, 85].
In this chapter, we have inspected the factors that can lead to the saturation of lφ in
TIs by measuring both VG -dependent conductance fluctuations and magneto-resistance
(MR) [29, 37, 41, 42, 59, 148, 181, 183, 195]. The saturation of lφ at low T is often
attributed to spin-flip scattering processes. The normalized magnitude of conductance
fluctuations hδG2 i/G2 , however, shows a factor of two reduction on application of a
129
Chapter 5. Low temperature saturation of phase coherence length in
topological insulators
10 μm 10 μm 15 μm 1 mm
d = 7 nm d = 100 nm d = 11 nm d = 10 nm
Figure 5.1: Devices measured: Optical image of different types of devices measured:
(a) Thin (∼ 10 nm) TI on Si/SiO2 substrate (b) Thick (∼ 100 nm) TI on Si/SiO2
substrate (c) Thin (∼ 11 n.m) TI on hBN substrate. (d) Molecular beam epitaxy grown
10 nm thick TI. The sample is mechanically patterned into a Hall bar.
perpendicular magnetic field (B), implying time reversal symmetry is intrinsically pre-
served in these systems. Additionally, at all gate voltages, hδG2 i/G2 shows a saturation
for T < 2 K, even in the presence of a large B, which freezes the spin-flip scattering,
implying magnetic impurities are not responsible for the observed saturation. Our anal-
ysis suggests that the saturation can originate due to dephasing from electromagnetic
fluctuations from the charged puddles in the bulk of these samples.
130
5.3. Methodology for investigation of dephasing
Table 5.1: Details of measured devices: The thickness, substrate, composition, sat-
uration value of temperature (T sat ), and saturation value of phase breaking length (lφsat )
for various devices that been investigated is provided in the table.
Sample Thickness Substrate Composition T sat lφsat
F7 7 SiO2 /Si++ Bi1.6 Sb.4 Te2 Se ∼2K 200 nm
F100 100 SiO2 /Si++ Bi1.6 Sb.4 Te2 Se ∼2K 150 nm
F9 9 SiO2 /Si++ Bi1.6 Sb.4 Te2 Se ∼2K NA
TBN11 11 Boron nitride Bi1.6 Sb.4 Te2 Se ∼2K 190 nm
M10 10 STO (Bi,Sb)2 Te3 ∼ 300 mK 2000 nm
5.1.
e2
1 Bφ Bφ
4σ = −α ψ + − ln (5.2)
πh 2 B B
where τφ , τso , τe are the phase coherence or dephasing time, spin-orbit scattering time
and elastic scattering time respectively, ψ is the digamma function and Bφ is the phase
breaking field. Here α and Bφ are fitting parameters. The phase coherence length lφM R
can be extracted using lφM R = ~/4eBφ .
p
The predominant dephasing mechanisms can then be inferred from the temperature-
dependence of lφ . In 2D, for electron-electron type-interaction, lφ ∝ T −0.5 , while for
electron-phonon interaction, lφ ∝ T −1 .
131
Chapter 5. Low temperature saturation of phase coherence length in
topological insulators
10 0.0
(a) 10 (b)
-0.2
RMBE (K)
(e2/h)
REXF (K)
8 8
-0.4
F7
F9
F 100
-0.6 M10
6 6 TBN11
-0.8
-60 0 60 120 -0.2 0.0 0.2
VG (V) B (T)
Figure 5.2: Basic device characteristics: (a) R - VG for exfoliated TI (REXF ) TBN11
and epitaxially grown TI (RM BE ) M10 at T = 20mK. (b) Magneto-conductance for
different samples at T = 300 mK, showing weak-antilocalization. The solid black lines
are fits to the data using Eq. 5.2.
Fig. 5.2b shows weak anti-localization (WAL) for different samples, which is charac-
terized by a cusp in the quantum correction to conductivity 4σ at B = 0 T, and is a
result of π Berry phase in topological insulators. The magneto-conductance data can be
fitted with the HLN equation (Eq. 5.2), from which lφ can be estimated.
132
5.5. Gate-voltage dependent conductance fluctuations
(10 e /h)
-2 2
R (K)
-0.5
~T
2
G
4.4 320 mK
700 mK
1.5 K
3K
11 K
0.1
-2 0 2 1 10
VG (V) T (K)
133
Chapter 5. Low temperature saturation of phase coherence length in
topological insulators
(a) (b)
200
100
150
(nm)
l (nm)
100
MR
50
lUCF
F7
F100 F100
10 50
1 10 1 10
T (K) T (K)
Figure 5.4: Saturation of phase breaking length: (a) Phase breaking length (lφ )
vs temperature (T ) extracted from T -dependence of hδG2 iusing Eq. 5.1 for device F100.
The solid line is fit according to Eq. 5.3. (b) lφ vs T extracted from T -dependence of
WAL data using Eq. 5.2 for exfoliated samples F7 and F100 which shows a saturation
below T < 2 K. The solid lines are fit according to Eq. 5.3.
are saturation below T < 2 K at a value ∼ 130 nm. Since hδG2 i ∝ lφ2 , any saturation
obtained from VG -dependence should also be reflected in the saturation of lφ , obtained
directly from MR measurements. The values of lφ extracted from MR data similar to
Fig. 5.2b, as a function of T for the exfoliated samples F100 and F7 is displayed in Fig.
5.4b. We find that lφ obtained from two different methods, lφU CF and lφM R (extracted
from Eq. 5.1 and Eq. 5.2 respectively) show similar trend with T , first increasing with
decreasing T , followed by a saturation below T ∼ 2 K, thus discarding the possibility of
the saturation to be an artifact. The discrepancies in the values of lφM R and lφU CF can
arise because Eq. 5.1 is strictly valid in the case of Lφ << L,LT [179, 189, 196]. The
higher value of lφ for F7 compared to F 100 can be due to enhanced dephasing, arising
from electromagnetic fluctuations due to trapping-detrapping processes in the bulk, since
the thickness of F100 is much larger than that of F7.
134
5.7. Discussion
(a) 0.32 K
0.5 K
(b) m* = .065 m
e
= 1.9 ps
0.7 K
2.3 K 0.01
3K
600 4K
0.001
300 1E-4
5.7 Discussion
For a quantitative understanding, lφ -T data (Fig. 5.4a-b) has been fitted with the ex-
pression commonly used in 2D diffusive systems [45, 85].
1
lφ = (5.3)
(A0 + A1 T + A2 T 2 )0.5
135
Chapter 5. Low temperature saturation of phase coherence length in
topological insulators
Table 5.2: Fitting parameters for obtained by fitting lφ vs T data with Eq. 5.3 for the
exfoliated sample F7 and F100.
Fit parameter F7 F100 F100 (from UCF)
13 13
A0 2.3 × 10 4.4 × 10 6.1 × 1013
A1 2 × 1012 1 × 1012 4 × 1012
11 11
A2 7.5 × 10 6 × 10 7 × 1012
1
(m)
0.5
lMR
Figure 5.6: Saturation of phase breaking length for large-area TI (a) lφ vs T for
different VG −VD extracted from T -dependence of WAL data using Eq. 5.2 for epitaxially
grown TI M10 which shows a saturation below T < 300 mK. The solid lines are fits to
Eq. 5.3. (b) α extracted from Eq. at different VG − VD values.
(a) saturation of electron temperature due to heating from external sources [45]
(b) finite size effects, where lφ becomes comparable to L, the length of the sample [203]
(c) spin-flip scattering due to local magnetic moments [49, 55, 57]
(d) spin-orbit coupling length becoming comparable to the phase breaking length [50].
We have systematically explored the possibility of saturation arising from any of the
above reasons.
136
5.7. Discussion
(a) 0 (VG-VD)
(b) 1.0 0 (VG-VD)
30 (VG-VD) 30 (VG-VD)
60 (VG-VD) 60 (VG-VD)
300
l (nm)
0.5
100
0.0
0.1 1 10 0.1 1
T (K) T (K)
• Thermalization:
An important factor which may lead to the saturation is improper thermalization, where
the electron temperature Te can be much higher than the lattice temperature TL . For
extracting Te accurately in our He-3 system, we have measured the T -dependence of
Shubnikov-de Haas oscillations using GaAs/Al0:33 Ga0:66 As hetero-structure (Fig. 5.5a).
The temperature-dependence of the amplitudes of the SdH-oscillations can be expressed
using the equation [204, 205]
∆ρ 8π 2 KB Te /~ωC π
= 2
exp(− ) (5.4)
ρ0 sin(2π KB Te /~ωC ) ωC τ
Here τ is the quantum scattering time, ∆ρ is the amplitude of the oscillations, ρ0 is the
average between maximum and minimum value of one oscillation, KB is the Boltzmann
constant, Te is the electron temperature and ωC the cyclotron frequency. The electron
temperature (Te ) has been extracted using Eq. 5.4, and shows that even down to 0.3 K
Te matches TL , as shown in Fig. 5.5b. Saturation due to external microwave noise has
also been discarded in Ref [61, 72, 73], which leads to heating before dephasing.
137
Chapter 5. Low temperature saturation of phase coherence length in
topological insulators
Table 5.3: Fitting parameters for sample M10 obtained by fitting lφ vs T data with Eq.
5.3 for different (VG − VD ).
Fit parameter −60 V −30 V 0V 60 V
11 11 12
A0 2.5 × 10 8.3 × 10 1.1 × 10 8.57 × 1011
A1 2.5 × 1011 1.1 × 1012 3.5 × 1012 1.8 × 1012
A2 3.5 × 1010 9 × 1010 2.5 × 1011 5 × 1010
1 mm, which is three orders of magnitude more than the saturated value of lφ . The lφ
extracted from MR data using Eq. 5.2 also shows a saturation for T < 300 mK for all
gate voltages (Fig. 5.6a). The value of the pre-factor α also remains fixed in the entire
temperature range (Fig. 5.6b). The saturation obtained in the large area sample, which
has also been reported in Ref [76] indicates that finite size effects at low T , where lφ
becomes comparable to L, the length of the sample, is not the cause of the observed
behavior of lφ . Even for exfoliated samples, L ∼ 4 − 5 times of lφ . However, we note
the temperature where lφ saturates in M10 is ∼ 300 mK, which is an order of magnitude
lower compared to the exfoliated samples (F7 and F100). This is possibly because the
temperature for the onset of saturation depends of various parameters like material, size
and microscopic quality of the sample [45, 61] . Since, the purity of the starting elements
for synthesizing BSTS (Bi, Te, Sb, Se) was ≥ 4N, while for BST it was 5N for Bi, 5 N
for Sb and 6 N for Te, there can be a difference in the T for the onset of saturation. The
difference in the saturation T can also arise due to screening from STO, since it has a
very high dielectric constant ( ∼ 44000 at 5 K) [97]. lφ obtained for M10 can be fitted
with Eq. 5.3 at all gate-voltages. The fitting parameters are provided in Table. 5.3.
• Presence of local-moments
The most common reason for the saturation of lφ is the presence of magnetic impurities
[49, 55, 57] which leads to the saturation, even in extremely pure systems. In the presence
of magnetic impurities, the total phase breaking time τφ is given by
1 1 1 1
= + + (5.5)
τφ τe−e τe−ph τsf
where τe−e , τe−ph , and τsf are the e-e, e-ph, and spin-flip scattering time respectively.
For T > TK , τsf is described by the Nagaoka-Suhl expression, which is given by
1 cmag πS(S + 1)
= (5.6)
τsf π~vf π S(S + 1) + ln2 (T /TK )
2
138
5.7. Discussion
450 mK
Fit
1.00
-60 V
TBN11
N(B)/N(0)
0.75
0.50
where S and TK are the spin and Kondo temperature of the magnetic impurities respec-
tively, and Cmag is the concentration of magnetic impurities. τsf reduces with decreas-
ing T , while τe−e and τe−ph increase with decrease in T . The inter-play of these two
time-scales leads to the saturation of τφ in the samples with magnetic impurities. Ex-
perimental and theoretical studies have reported the presence of localized spins [206] or
intrinsic magnetic instabilities in the TI surface [207]. To probe the presence of magnetic
impurities, we have used an exfoliated TI on BN substrate (TBN11). The atomically flat
boron nitride (thickness d = 14 nm) prevents trapping-detrapping between the channel
and the SiO2 substrate, thus reducing any dephasing due to electromagnetic fluctuations
induced by potential traps present in SiO2 . lφ and α for the TBN11 is shown in Fig. 5.7a
and Fig. 5.7b respectively. Both lφ and α extracted from magneto-resistance, show a
saturation, at all measured gate voltages, similar to that observed in the other samples,
signifying that electromagnetic fluctuations originating from charge exchange between
the channel and the SiO2 substrate is not the cause of the saturation.
One manifestation of the presence of magnetic impurities is the presence of long-
range correlation among the impurities, which can lead to the removal of time reversal
symmetry (TRS) in the system. To probe this, we have measured hδG2 i as a function
of transverse magnetic field (B) in the sample TBN11. The magnitude of the conduc-
tance fluctuations, plotted as NG (B)/NG (0) , where NG = hδG2 i/hG2 i is the normalized
variance, is depicted in Fig. 5.8. We observe a clear factor of two reduction in NG when
139
Chapter 5. Low temperature saturation of phase coherence length in
topological insulators
(a) (b)
0T
2T
G2(10-4)/G2
TBN 11
G2(10-4)/G2
1 -60 V
1
0V
60 V
0.1
0.1
0.1 1 10 0.1 1
T (K) T (K)
B >> Bφ , the phase breaking field, due to the suppression of the Cooperon contribu-
tion in transport, and the crossover of the system from symplectic to unitary symmetry
class. This indicates that the TRS is intrinsically intact in these systems. For a quan-
titative understanding, the normalized magnitude has been fitted with the expression
[41, 194, 195]
∞
1 1 X 1
N (B)/N (0) = + 2
2 b n=0 [ n + 2 + 1b ]3
1
1 1 00 1 1
= − 2Ψ + (5.7)
2 2b 2 b
00
Here b = 8πB(lφ )2 /(h/e) and Ψ is the double derivative of the digamma function and lφ
is the fitting parameter. The solid line in Fig. 5.8 is the fit of the data according to Eq.
5.7, which captures the variation with normalized magnitude with B. The lφ obtained
from the fit is 220 nm, similar to lφ magnitude obtained from conductance fluctuations
and magneto-resistance, which confirms the validity of the analysis and the factor of two
reduction. To investigate the cause of saturation further, we have extracted hδG2 i as
a function of T for different VG at a fixed magnetic field of 0 T and 2 T respectively.
The normalized magnitude hδG2 i/G2 shows a saturation for both B = 0 T and 2 T for
140
5.7. Discussion
Saturation of lφ has been observed in GaAs-AlGaAs systems at low T , when the spin-
orbit length (lso ) becomes comparable to the inelastic scattering length [50]. To verify
this, we have extracted lSO by using the full HLN equation [192, 208] by using the
expression,
e2
2α Be Bφ + 2BSOz + 2BSOX
4σ = − η −η
π h B B
2α e2
1 Bφ 1 Bφ + 4BSOX
− η − η (5.8)
π h 2 B 2 B
Here, η(x) = ψ 1
+ x − ln (x), Bn = ~/4elφ2 , n = e is the momentum relaxation
2
field, and n = SOx and SOy denote spin-orbit interaction field in x and z-direction
respectively. The fit of our experimental data over the full magnetic field range using
Eq. 5.8 has been shown in Fig. 5.10. The extracted values of Be and Bsox are 0.4 T and
0.04 T respectively, while Bsoz ∼ 9 T, which is much higher than both Be , BSOx , and
similar to values reported in Ref. [208]. The extracted lSO is much smaller compared to
lφ ruling out this possibility as well.
141
Chapter 5. Low temperature saturation of phase coherence length in
topological insulators
0.0 180mK_-60V
Model Lso (User)
-(2*A/pi)*[real_polygamma((.5+ B/a
bs(x)),0)-ln(B/abs(x))] + (2*A/pi)*[re
al_polygamma((.5+ (C+(2*D)+(2*E
))/abs(x)),0)-ln((C+(2*D)+(2*E))/abs
Equation
(x))]-(A/pi)*[real_polygamma((.5+ C
(e2/h)
-0.2
/abs(x)),0)-ln(C/abs(x))] +(A/pi)*[rea
l_polygamma((.5+ (4*E+c)/abs(x)),
0)-ln((4*E+C)/abs(x))]
Plot ?$OP:F=1
A 0.5 ± 0
B 0.40437 ± 0.01252
C 0.0045 ± 2.75786E-5
D 9.2732E15 ± 0
E 0.04392 ± 0.00197
-0.4
Reduced Chi-Sqr 2.66393E-4
R-Square(COD) 0.98491
Adj. R-Square 0.9849
-0.6
-2 0 2
B (T)
Figure 5.10: Estimation of spin orbit coupling length: Fit to the WAL data by
using the full HLN expression, given by Eq. 5.8. The solid green line represents the fit.
The increase in resistance at low T in bulk ∼ 100 nm TIs can be attributed to variable-
range hopping type transport, which has been previously reported in the materials [209,
210]. The resistance can be fitted with Efros-Shklovskii variable range hopping type
transport, as demonstrated in Fig. 5.11a [211–213], where
−1/2
T0
R = R exp (5.9)
T
2.8e2
Here T0 = 40 KB lξ
gives an estimate of the localization length (lξ ) . The localization
length comes out to be ∼ 1 nm, similar to other reports in TI [210]. Variable range
hopping in topological insulators at low T has been predicted theoretically as well [212,
214, 215], due to the presence of charged puddles in the bulk. It is possible that dephasing
by charged disorders in the bulk at low temperatures can lead to the saturation of lφ
[76, 216]. The effective inelastic scattering rate, using Matthieson’s rule at low T , can
be written as
1 1 1
= + (5.10)
τ τss τbulk
Here 1/τss refers to the dephasing rate due to interaction of the surface electrons, whereas
1/τbulk refers to dephasing from charged disorders in the bulk. The dephasing rate due
to the bulk disorders, can be estimated using Nyquist dephasing mechanism [45, 70, 217]
and is given by
1 KB T
= (5.11)
τbulk 2πg()D~2
142
5.7. Discussion
0.9
(sec)
R (K)
0.6
1E-12
0.3
0.1 0.2 1 10
T-1/2 (K-1/2) T (K)
where g() is the density of states at the Fermi level, and D is the diffusivity. The
density of states for 2D for a random potential in a localized system at T = 0 is given
as [211, 213, 215]
2α2 κ2
g() = || (5.12)
πe4
where α = e2 /κ~vf is the effective fine structure constant, and κ is the dielectric constant,
which gives g() ∼ 1
2π~2 vf2
| |. For a finite temperature, we assume ∼ KB T . So from
Eq. 5.11 and Eq. 5.12, we have
1 vf2
∼
τbulk D
Typical values of vf and D are 105 m/sec and 10−2 m2 s−1 for TIs respectively [76]. This
gives τbulk ∼ 10−12 secs, which is similar to the dephasing rates obtained from MR, by
using the relation τ = lφ2 /D (Fig. 5.11b). It is possible that the dephasing rate from
charged impurities in the bulk becomes more dominant at low T , when the uncom-
pensated charges are strongly localized, leading to reduced screening of electromagnetic
fluctuations. This may lead to the saturation of the phase breaking length in topolog-
ical insulators at low temperatures. The electron-holes puddles have been modelled as
quantum dots in literature [218–220]. These puddles can absorb energy, get excited to
143
Chapter 5. Low temperature saturation of phase coherence length in
topological insulators
the different levels, and relax back to the ground state state by emitting photons. Mea-
surements of the lifetimes of the excited states by intercepting the emitted photons or
by measuring time-resolved photo-luminescence may shed light on this. Further analysis
and measurements of length dependent dephasing rates are also required to pin-point
the exact source of saturation of lφ at low temperatures.
5.8 Summary
• We have investigated the decoherence in topological insulator surface states.
144
Chapter 6
6.1 Introduction
The detection of electromagnetic radiation in the infra-red regime (0.7–100 µm) is ex-
tremely crucial for technological applications, since infra-red (IR) detection provides a
variety of advantages, compared to other wavelengths. According to Wien’s displace-
ment law, every object at a temperature T (K) emits electromagnetic radiation with a
peak value at λ = 2900/T µm, which implies that objects within a temperature range
of ∼ 100 − 4000 K emits radiation in the range in the IR range. Thus detecting IR
photons can give information about the source itself, such as its T , without any external
illumination. This method is also known as passive detection. Secondly, since Rayleigh
scattering intensity I ∝ 1/λ4 , IR radiation can travel through different mediums with-
out significant scattering in comparison to visible or UV radiation. These properties
make detection of infra-red radiation extremely valuable, since they can be utilized in
night-vision cameras, thermal imaging, medical diagnosis, astronomy, satellite communi-
cations, and also in various military applications. Conventional IR detectors are usually
made from materials such as HgCdTe, InAs, GaAs, and quantum dots [115, 221–224].
However, these photo-detectors typically show low responsivity and often need to be
cooled for practical purposes.
145
Chapter 6. Ultra-high detectivity of near infra-red radiation in bismuth
telluride-graphene hybrids
In this chapter, we have tried to circumvent the problem of low optical response, by
employing graphene-Bi2 Te3 nano-wire hybrids. The Bi2 Te3 nanowires, having a band
gap of 0.17 eV, are an ideal candidate as the light absorbing layer, and were drop-cast
on graphene channel which acts as the charge collector in this case [1, 231, 232]. Our
devices show an extremely high sensitivity of the order 105 –106 AW−1 in the wavelength
range of 920 nm – 1720 nm. The calculated specific detectivity of our devices is also
1010 –1011 Jones, which is comparable to the best commercially available IR detectors.
146
6.2. Device fabrication
(a) (b)
before dropcast VD= 14.7 V
R (K)
10 μm 2
-15 0 V (V) 15
G
Figure 6.1: Basic device characteristics: (a) Optical micrograph of a typical graphene
FET fabricated on Si/SiO2 substrate. (b) Resistance (R) as a function of gate-voltage
(VG ) of the FET, showing ambipolar transport. The maxima in the resistance indicates
the Dirac point or the charge neutrality point. The graphene layer is intrinsically hole-
doped.
147
Chapter 6. Ultra-high detectivity of near infra-red radiation in bismuth
telluride-graphene hybrids
I (nA)
I (nA)
2
0
16
-2
-10 0 10 30 60 90 120
VDS (V) t (sec)
Figure 6.2: Properties of Bi2 Te3 nanowires: (a) Scanning electron micrograph of
a nanowire field effect transistor. (b) Output characteristics of the nano-wire at vari-
ous temperatures, showing non-linear behavior, and negligible dependence on the gate-
voltage at 82 K. (c) Opto-electronic measurements in the nanowires, showing the absence
of photo-response in bare nano-wires.
using
σ = neµ
where σ is the conductivity, n is the number density, and µ us the mobility of the
sample. The typical mobility of these devices is around ∼ 1000 cm2 V−1 s−1 . The presence
of charged traps and dangling bonds in the 285 nm SiO2 substrate leads to trapping-
detrapping between the channel and the substrate, and limits the mobility of these
devices [145, 165].
Before measuring the photo-detection capabilities of the hybrid devices, we have mea-
sured photo-response in single nano-wire field-effect transistors (Fig. 6.2a). The nanowires
show non-linear I-V at all gate-voltages, possibly due to non-Ohmic contacts (Fig. 6.2b).
The single nano-wire devices did not show any measurable photo-response, as demon-
strated in Fig. 6.2c.
148
6.3. Shift in Dirac point
(a) (b)
Drain
Bi2Te3
Source
Graphene
VG Si/SiO2
5 μm
149
Chapter 6. Ultra-high detectivity of near infra-red radiation in bismuth
telluride-graphene hybrids
VD (V)
VD= 14.7 V
4 -20
2
0
-15 0 15 1 2 3 4
VG (V) Device No
Figure 6.4: Shift in graphene Dirac point: (a) The resistance vs gate-voltage char-
acteristics of the graphene channel before and after hybrid formation shows a shift in the
Dirac point from 14.7 V to 11.7 V, implying that the graphene channel has been doped
with electrons after contact with the nanowires, which has shifted the Dirac point to the
left. (b) The shift in the Dirac point voltage for different devices is shown. The shift
of the Dirac point towards more negative voltages denotes that the graphene is getting
electron-doped.
150
6.4. Photo-response in graphene-Bi2 Te3 hybrids
(a) 1720 nm
5145 off on off on off on off on off
5130
1550 nm (b)
6.105
R ( )
180 K
6.090 6 940 nm
-6
R (K)
R ()
6.075
1060 nm
6220 4
-12
6210
2 -18
940nm -20 0 20
VG (V)
6240
6225
t (sec)
Figure 6.5: Photo-response of the hybrid devices: (a) The optical-response of the
device at fixed gate-voltage (VG ) for different wavelengths (1720 nm, 1550 nm, 1060 nm,
and 940 nm). The resistance of the sample changes periodically from a high-resistance
state (Rof f ) to a low resistance state (Ron ) when the optical illumination is switched on.
(b) VG -dependence of photo-response for 940 nm wavelength at 180 K, showing that the
opto-electronic response in maximum at the Dirac point.
151
Chapter 6. Ultra-high detectivity of near infra-red radiation in bismuth
telluride-graphene hybrids
Figure 6.6: Mechanism of photo-response: (a) Electron (blue)-hole (green) pairs are
generated in the nano-wires under optical illumination. When the graphene is electron-
doped for VG > VD , electrons are transferred from Bi2 Te3 to the graphene layer due to
the in-built electric field. (b) When the graphene is hole-doped for VG < VD , holes are
transferred to the channel due to the inter-facial electric field.
since the photo-gating shows a maxima at the points where dR/dVG is extrema [126, 234].
The photo-response is also free from persistence, as the resistance recovers to its original
state (before optical illumination), when the light is switched off.
152
6.6. Time-scale of photo-response
(a) (b) 1
940 nm
6.240 82 K
6.235
OFF ON
R (K)
OFF ON
0.1
6.230
6.225
0.01
40 60 80 100 40 60 80
t (sec) t (sec)
d(4n) 4n
= ηΓ −
dt τ
Here 4n is the difference of carrier densities between on and off state, η is the quantum
efficiency, Γ is the charge generation rate, and τ is the recombination time scale. The
first term in the right-hand side of the equation corresponds to the net photo-carrier
generation, and the second term gives the recombination of carriers. This gives,
4n = ηΓτ (1 − exp(−t/τ ))
Since the change of resistance (R) is proportional to 4n, the time-scales for photo-
response can be determined by modeling the resistance with an exponential function.
We have fitted the resistance with a function of the form,
t − t0
R(t) = R0 + 4R 1 − exp − (6.1)
τ
Here R0 is the off-state resistance, ∆R is the change in the resistance between off and
on state, and t0 is the time when optical illumination is switched on. Such a fit is shown
in Fig. 6.7a. The time-scales corresponding to the photo-response, τ range between
153
Chapter 6. Ultra-high detectivity of near infra-red radiation in bismuth
telluride-graphene hybrids
(a) (b)
OFF ON OFF ON OFF ON OFF ON OFF ON OFF
T = 83 K
3.95
100
R (K)
Ip (nA)
P
3.94
940nm
10 1060nm
1720nm
0 60 120 180 240 300 360 420 480 540 600 660
100 1000
t (sec) 2
P (fW/µm )
2 − 4 secs for different devices. Alternatively, τ can be estimated from the semi-log
graph of η = 1 − R(t)−R
4R
0
, plotted as a function of t − t0 . The straight line fit can be used
to estimate the value of τ as shown in Fig. 6.7b. This time-scale of photo-response is
much slower compared to charge transfers within van der Waal’s hetero-structure, which
is of the order of picoseconds.
VDS
IP = − 2
4R (6.2)
Rof f
154
6.7. Photo-response figure of merit
(a) (b)
940 940 nm
6 -16
10 1060 10 1060 nm
1300 1300 nm
)
-1/2
1550 1550 nm
1720 1720 nm
NEP (WHz
(AW )
-1
5 -17
10 10
4 -18
10 1 10 1 2 3 4
10 10
2
10
3 10 10 10 10
2 2
P(fW/µm ) P (fW/µm )
(c)
11
D* (Jones)
10
940 nm
10 1060 nm
10 1300 nm
1550 nm
1720 nm
2 3
10 2
10
P (fW/µm )
Figure 6.9: Photo-response figure-of-merit: (a) Responsivity (γ) of the device for
different wavelengths at 82 K. The responsivity is almost constant because of the linear-
dependence of IP on power. (b) Noise equivalent power,(N EP ) for different wavelengths
at 82 K. (c) Specific detectivity (D∗ ) for different wavelengths at 82 K.
155
Chapter 6. Ultra-high detectivity of near infra-red radiation in bismuth
telluride-graphene hybrids
Bi2Se3/Si Heterostructure
Polymer Photodetector
Polymer Photodetector
1E10
D*(Jones)
HgTe Nanocrystals
PbS Quantum Dot
GaAs/AlGaAs
This work
Figure 6.10: Comparison of specific detectivity with some of the other devices reported
literature (Ref. [126, 127, 222, 235–237]).
high value of ∼ 106 AW−1 for excitation wavelength = 1060 nm, which is comparable
to some of the best detectors reported in literature. The high mobility of graphene,
responsible for large carrier lifetime, leads to a very high photo-conducting gain, which
is responsible for the high responsivity in these devices. The noise equivalent power
(N EP ) of these devices has been estimated using
N EP = IN /γ
q
IN = IJ2 + IS2
√
Here IJ = 4KB T /R and IS = 2eI are the contributions from Johnson’s noise and
p
shot noise respectively. N EP for our devices show a very low magnitude of 2 × 10−18
WHz−1/2 , as demonstrated in Fig. 6.9b. The N EP for a single photon is 1.6 × 10−19
WHz−1/2 . This shows that our devices can be used to estimate photons in the IR regime
with a resolution of ∼ 10 − 100. The specific detectivity, which is defined as D∗ =
√
A/N EP is shown in Fig. 6.9c, where A is the area of the detector. Since specific
detectivity is area normalized, it is used as the yardstick for comparing various detectors.
156
6.8. Temperature dependence of photo-response
940 nm
1060 nm
VG = -10 V
1300 nm
2e4*exp(230*x)
1550 nm
1720 nm
(AW )
-1
5
10
5e3*exp(250*x)
3e3*exp(210*x)
3
10
6 8 10 12
-3 -1
1/T (10 K )
The specific detectivity is also very high for our devices ∼ 1011 Jones, which is comparable
to some of the other devices reported in literature, as illustrated in Fig. 6.10 [126, 127,
222, 235–237].
The photo-conducting gain (G) can be estimated from using the following relation
ηext
G=
η
Here ηext is the external quantum efficiency, and is defined as the ratio of the number
of photo-generated charge carriers for each incident photon, whereas η is the internal
quantum efficiency, and is defined as the ratio of the number of photo-generated charge
carriers for each absorbed photon of a given energy. The external quantum efficiency is
given by
γhc
ηext =
λe
where h, c, and e are the Planck’s constant, velocity of light, and electronic charge
respectively [118]. Assuming γ = 106 AW−1 for λ = 1060 nm, we get ηext = 1.17 × 106 .
Using η = 0.3, we get G ∼ 3.9 × 106 .
157
Chapter 6. Ultra-high detectivity of near infra-red radiation in bismuth
telluride-graphene hybrids
6.9 Drawbacks
One major hindrance of our devices seem to be the limited-duration of photo-response,
which lasts around a few days. Even though all the measurements are done in vacuum ∼
10−6 mbar, the photo-response is probably limited due to the formation of an amorphous
layer on the wires. This was verified by transmission electron microscope (TEM) images,
as shown in Fig. 6.13a and Fig. 6.13b. The TEM image in Fig. 6.13a shows a typical
158
6.9. Drawbacks
OFF
1296
1280
R ()
R ()
1284
1270
R ()
1430 1420
1420
1400
0 120 240 360 480 0 120 240 360 480 600 720 840 960
t (sec) t (sec)
159
Chapter 6. Ultra-high detectivity of near infra-red radiation in bismuth
telluride-graphene hybrids
(a) (b)
5 n m 5 n m
Before After
nano-wire from the stock solution, which shows perfect crystalline nature of the wires.
However, after measurements, as can be seen from Fig. 6.13b, the wires develop an
amorphous layer due to oxidation, which might lead to the absence of photo-response
in these hybrids devices. This can be bypassed by use of some capping element such as
Te, which may prevent the oxidation of the surfaces, or by transferring the graphene on
top of the wires, which will prevent the wires from coming in contact with air. Another
possible method is to dropcast PMMA on the sample after the nano-wires have been
dropcast to prevent the wires from coming in contact with oxygen.
6.10 Summary
• We have measured photo-response graphene-bismuth telluride nano-wire hybrid
devices.
• The photo-response occurs due to electron-hole pair generation in the Bi2 Te3 . The
inter-facial electric field aids the transfer of charges from one layer to the other
leading to the observed photo-response.
• The photo-response is free any persistence, signifying trap-free interface. The time-
scale of charge transfer is ~2 − 4 seconds.
160
6.10. Summary
• The responsivity of the devices ∼ 106 AW−1 . The specific detectivity of the devices
∼ 1011 Jones, which is comparable to some of the best known detectors.
161
Chapter 7
Concluding remarks
7.1 Conclusions
In this thesis, we have explored the origin of fluctuations in the quantum transport
regime, as well as opto-electronic performances of topological insulators (TIs). One
of the important tools for our investigation has been time and gate-voltage dependent
conductance fluctuations, which we have used extensively to probe the various scattering
as well as dephasing mechanisms in TIs in a variety device configurations. We have also
used conductance fluctuations to investigate the symmetries of the topological insulator
Hamiltonian. We have employed binary hybrids of TIs and graphene, for detecting
infra-red radiation.
In Chapter 3, we have investigated the origin of low frequency 1/f noise in large-
area, thin (10 nm) topological insulators. We have demonstrated from gate voltage
and temperature-dependence of noise, that even for thin topological insulators, which
are ideal for electronic applications, the bulk defects are the source of 1/f noise. This is
unlike other two-dimensional materials, where the source of noise is either at the contacts
or due to trapping-detrapping at the sample-oxide interface.
In Chapter 4, we have explored by the intrinsic symmetries of the TI Hamiltonian
by measuring gate-voltage dependent conductance fluctuations. The magnetic-field de-
pendence of conductance fluctuations shows a factor of two reduction, which provides an
unambiguous proof of the intrinsic preservation of time-reversal symmetry in topological
insulators. The factor of two reduction also proves beyond doubt that the fluctuations
originate from universal conductance fluctuations, especially at gate-voltages near the
163
Chapter 7. Concluding remarks
Dirac point.
In Chapter 5, we have carried out systematic investigation in TIs in a variety of de-
vice configurations, to unravel the sources of dephasing in these materials. We have used
magneto-resistance and conductance fluctuations to probe the sources inelastic scattering
in these materials. Our experiments have revealed that below T = 2 K, the phase coher-
ence length (lφ ) saturates. Above 2 K, the dephasing can be attributed to a combination
of electron-electron and electron-phonon type interactions. The saturation of lφ is often
attributed to spin-flip scattering mechanism. However, the magnetic-field dependence
of conductance fluctuations prove unambiguously, that spin-flip scattering processes are
not responsible for the saturation. Our analysis suggests that fluctuations from the bulk
of the TIs can be the source of the observed saturation of lφ .
In Chapter 6, we have harnessed the properties of topological insulators for photo-
detection in the near infra-red regime by forming binary hybrids with graphene. These
hybrid devices show ultra-high sensitivity ∼ 106 AW−1 , under an optical illumination
from 920 nm to 1720 nm. The noise equivalent power in these devices show that they can
be used to detect ∼ 10 − 100 photons. The specific detectivity of these hybrids devices
is ∼ 1011 Jones, which is comparable to some of the best known detectors.
In this section, we suggest a few possible directions in which the research presented in
this thesis may be extended.
Quantum anomalous Hall effect is the perfect playground to investigate 1D edge trans-
port, which can be achieved in magnetically-doped topological insulators. However, the
sensitivity of these edge modes to disorder, and the scattering mechanisms affecting the
quantization still remain unknown. In addition, these samples also show metal-insulator
transition, and their scaling behavior can be probed by noise measurements.
164
7.2. Scope for future work
• The time-scale of photo-response in Bi2 Te3 -graphene hybrid devices is ∼few sec-
onds, whereas in pump-probe measurements, the typical time-scale of photo-response
∼picoseconds. This naturally calls for investigating the origin of the time-scales
associated with the photo-response. The speed at which these photo-detectors can
function, also need to be explored. The photon resolution is ∼ 10 photons in these
devices. It remains to be seen, where by reducing the dark current, and optimizing
the area coverage between graphene-TI, the photon resolution can reduced, and
whether they hybrids can resolve a single photon.
Even though the bulk disorder is responsible for the origin of noise in topological insu-
lators, the substrate, surface adsorbates, and contact regions can also play a significant
role in mesoscopic samples. For further understanding of the origin of noise, fully encap-
sulated TIs, with a top and bottom gate, which can independently tune both the surface
number densities are required.
165
Chapter 7. Concluding remarks
166
Appendix A
A.1 Introduction
Human immunodeficiency virus (HIV) has transformed into a global epidemic, and
emerged as a multi-system disorder, primarily involving cardiovascular and musculoskele-
tal system [241–243]. HIV-induced drug interactions causing cardiovascular disorders
(CVD) in patients, are the one of leading causes of morbidity and mortality. HIV-CVD
association often remains clinically undiagnosed or wrongly attributed to other non-
cardiac diseases such as chronic obstructive pulmonary disease (COPD), and obstructive
sleep apnea (OSA) [244]. Increased awareness and routine screening for cardiovascular
symptoms in HIV patients can aid in early diagnosis and subsequent treatment. Mul-
tiple observational cohort studies have been carried out, the results of which showed
an increased risk of acute myocardial infarction in HIV affected patients as compared
to control individuals [245–247]. Moreover, a systematic review and meta-analysis con-
ducted by Islam et al.[248], with an aim to estimate the risk of CVD in HIV-infected
patients (PLHIV) compared to healthy individuals, found HIV infection to increase the
risk of CVD by 61 %. Examples of some HIV associated CVDs are infective endocarditis,
pulmonary hypertension, pericardial effusion, and premature atherosclerosis[249–251].
Patients having HIV can develop arthritis at any stage of the disease. Moreover,
167
Appendix A. Biosensing using graphene: detecting HIV, cardiac troponin,
arthiritis, and chlorpyrifos
(a)
(b)
(c)
(d)
168
A.1. Introduction
serological markers of arthritis can be found in HIV infected patients [252]. A number
of immunological and electro-chemical assays have been developed for the detection of
HIV, and its associated diseases (CVD and RA) [253–255]. Diagnosis of HIV (primarily
associated with the detection of p24 antigen), RA, and CVD is primarily performed us-
ing immunological assays such as enzyme linked immunosorbent assay (ELISA), Western
blot analysis, and immunofluorescence assay (IFA) [256–258]. Immunosensors provide a
promising alternative approach for the detection of various biomolecules such as pesti-
cides, narcotic drugs, cancer, HIV, CVD, and RA [259–262]. Advancements in electro-
chemical immunosensing have paved the way for the development of better detection
methods that have simple design, low cost, high sensitivity and field applicability [263–
265]. Most of the known methods, however, require sophisticated instrumentation, a
high analytical time, and may not be suitable for real-time, onsite applications.
In this Appendix, we report the development of graphene-FET based sensor for the
detection of multiple bio-markers such as HIV, cTn1 and RA. The FETs were fabricated
from both single and multilayer graphene [136, 137], which work at very low noise lev-
els [101, 112, 165], and were fabricated using the method given in Chapter 2 Section
2. The specific antibodies [HIV (anti-p24), cardiac troponin 1 (anti-cTn1), and Cyclic
Citrullinated Peptide (CCP) for RA] were covalently conjugated to amine functional-
ized graphene via carbodiimide activation chemistry. The sensors show extremely high
sensitivity, and can be used for onsite applications.
Fig. A.1 illustrates the processes followed to fabricate the sensors. At first, the graphene-
FET is fabricated using standard exfoliation technique, followed by electron-beam lithog-
raphy. The antibodies are then immobilized on the graphene surface by dropcasting.
Then the antibody solution is washed away, and antigens of various concentrations are
added, while the resistance of the sample is monitored continually. Fig. A.2a shows the
optical micrograph of a typical graphene-FET, fabricated using exfoliation method. Fig.
A.2b(i)-(ii) shows the SEM image shows the presence of Abs separately on the graphene
surface. The bright spots in the AFM image in Fig. A.2c(i)-(iii) demonstrate the presence
of antibody-antigen binding on the graphene surface.
169
Appendix A. Biosensing using graphene: detecting HIV, cardiac troponin,
arthiritis, and chlorpyrifos
(a)
(b)
Figure A.2: Device details: (a) Optical micrograph of a typical graphene FET. (b)
Scanning electron microscope images of the FET’s after measurement. The light patches
between the contacts show antibody agglomeration. (c) Atomic force microscopy in three
different devices. The brighter spots indicate antibody-antigen binding on the graphene
surface.
170
A.1. Introduction
(a) (b)
Figure A.3: Detection of HIV: (a) Kinetic response of the graphene biosensor for
several HIV antigen concentrations for HIV. (b) Percentage change in resistance of the
biosensor for the different antigen concentrations. The device shows a linear response in
the range 1 fg/mL to 1 μg/mL and a limit of detection of 100 fg/mL.
171
Appendix A. Biosensing using graphene: detecting HIV, cardiac troponin,
arthiritis, and chlorpyrifos
(a) (b)
Figure A.4: Detection of cTn1: (a) Kinetic response of the graphene biosensor for
several antigen concentrations for cTN1. (b) Percentage change in resistance of the
biosensor for the different antigen concentrations. The device shows a linear response in
the range 1 fg/mL to 1 µg/mL and a detection limit of 10 fg/mL.
(a) (b)
Figure A.5: Detection of rheumatic arthritis: (a) Kinetic response of the graphene
biosensor for several antigen concentrations for rheumatoid arthritis. (b) Percentage
change in resistance of the biosensor for the different antigen concentrations. The device
shows a linear response in the range 1 fg/mL to 1 µg/mL with limit of detection as 10
fg/mL.
172
A.1. Introduction
I(nA)
6
0
4
-2
2 -4
-20 0 20 40 1000 1200 1400 1600
VG (V) t(sec)
Figure A.6: (a) The change in the Dirac point voltage before and after measurements,
which indicates the doping of the graphene channel due to antibody-antigen binding.
(b) Binding studies with a bare gold-FET showing no measurable change, confirming
the role of graphene in a biosensing measurements.
173
Appendix A. Biosensing using graphene: detecting HIV, cardiac troponin,
arthiritis, and chlorpyrifos
(a) AuNP
Alexa Fluor 647 donkey antimouse IgG
(b) HRP-labeled p24 Ab2
DNA modified gold nanoparticles
0.01 Magnetic Nanoparticles 10 Au/EV-p24 Ab2
Au Nanoparticles
Peptide epitope graphene (present work)
0.001 Graphene (present work)
L.O.D (ng/ml)
0.01
L.O.D (ng/mL)
1E-4
1E-5 1E-5
1E-6
1E-8
1E-7
1E-8 1E-11
Calorimetric Flourescence Paramagnetic Electrochemical SPR Electrical ELISA Electrochemical Electrochemical Electrical
immunoassay
Figure A.7: Comparison of limit of detection for (a) cTn1 (b) HIV using different tech-
niques.
the graphene, e is the electronic charge, ∆VCN P is the change in charge neutrality point
or Dirac point of bare graphene and graphene channel at the end of the measurement,
and C = 1.2×10−4 F is the capacitance of 285 nm silicon dioxide. The change in number
density was calculated to be 1.95×1016 m−2 for cTn1 antibodies, thus reaffirming binding
between the graphene channel and the bio-molecules. For field applicability, the stability
of the device functionalized with antibodies was monitored up to three weeks. For further
confirmation of antibody binding on the graphene surface, 0.01 μg/mL antibodies were
drop-cast directly on a gold-FET device. However, the absence of any measurable change
further confirmed that the signal originated from antibody bound with graphene (Fig.
A.6b). The limit of detection for cTn1 and HIV antibodies, obtained from different
methods is illustrated in Fig. A.7a and Fig. A.7b respectively [266, 267], showing that
they are comparable to the best reported sensors till date.
174
A.2. Detection of chlorpyrifos pesticide
(a) (b)
Figure A.8: Chlorpyrifos detection: (a) Kinetic response from the graphene FET at
each step and after exposure of free chlorpyrifos at different concentrations: (i) 1 fM
(ii) 10 fM (iii) 100 fM (iv) 1 pM (v) 10 pM, and (vi) 100 pM in PB buffer pH 7.4. (b)
The percentage change of resistance as a function of time for different concentration of
antigens.
175
Appendix A. Biosensing using graphene: detecting HIV, cardiac troponin,
arthiritis, and chlorpyrifos
because all the sites have been reacted. The graphene-FET based immuno-sensor for
chlorpyrifos detection displayed a limit of detection of 1.8 fM using antigen-antibody
complex, and is one of the most sensitive detectors till date [270–273].
A.3 Summary
• Our investigation reveals that graphene can be an extremely sensitive platform for
detection of HIV and its related cardiovascular diseases and arthritis.
• The developed sensor is highly sensitive and showed a linear response to HIV, cTn1,
and RA from 1 fg/mL to 1 μg/mL. The limit of detection for p24 is100 fg/mL,
while for cTn1, it is10 fg/mL.
• The developed graphene-based FET immuno-sensor has a clear edge over the con-
ventional techniques because of its quick response time and increased selectivity
and sensitivity. Therefore, graphene-based device displayed excellent performance,
and has a high point-of-care testing potential for onsite detection of HIV, cTn1,
and RA in real samples.
176
Appendix B
Opto-electronic response in
graphene-tin selenide hybrids
The importance of detection of infra-red radiation is well known because they can be
used in the various applications such as night vision cameras, satellite communications,
security systems, and medical purposes [115, 221–224]. Two-dimensional materials like
graphene and transition metal dichalcogenides (TMDCs) offer a variety of advantages in
their application as photo- detectors [3, 93, 119, 136, 225]. While single layer graphene
exhibits fascinating electrical properties, it suffers from poor optical sensitivity due to
low absorption co-efficient (∼ 2.3 %) [226]. This can be bypassed by utilizing smart
designs, such as hybrids of graphene with other light-sensitive materials [3, 116–118,
120, 123, 125]. Here we have used hybrids of graphene-SnSe2 nano-platelets to detect
infra-red radiation (Fig. B.1). SnSe2 having a band-gap of 1 eV, acts as the optically
active layer. The photogenerated electrons and holes get transferred from SnSe2 due to
an interfacial electric field to graphene, leading to an extremely long carrier lifetime and
a high optical gain.
177
Appendix B. Opto-electronic response in graphene-tin selenide hybrids
Drain
Source
SnSe2
VG Graphene
a micro-pipette, and left for few minutes to let the ethanol evaporate. The coverage of
SnSe2 on graphene was examined using optical microscope, and the dropcasting process
was repeated until there was sufficient coverage of the graphene channel with SnSe2 . The
scanning electron micrograph of a typical device is shown in Fig. B.2a.
The resistance (R) vs gate voltage (VG ) demonstrated in Fig. B.2b of the hybrid device
shows ambipolar transport with a Dirac point at VD = −4.4 V at 300 K. To perform
opto-electronic measurements, the device was cooled down to 82 K. The resistance was
monitored as a function of time for various on-off cycles continuously, to obtain the photo-
response (Fig. B.3a). The photo-response is plotted as ∆R
Rof f
vs VG , where ∆R is given by
Rof f − Ron . Here Rof f and Ron are the graphene channel resistance in the absence and
presence of optical illumination respectively. The device responds from 470 nm to 1300
nm wavelength till 125 K. sThe fractional change in resistance for different gate-voltages
and wavelengths in shown in Fig. B.3b-d. The hybrid-device shows photo-response in
the near infra-red regime from 940 nm to 1300 nm.
178
B.2. Results and discussions
(a) (b)
VD = -4.4 V
300 K
4
Graphene
R (K)
3
SnSe2
2
-10 0 10
VG (V)
As evident from Fig. B.3b-d, the hybrid devices show a maximum photo-response near
gate-voltages where dR/dVG is maximum. The sign of the photo-response also changes
across the Dirac point, which are signatures of the photo-gating effect [118, 126]. The
work function of SnSe2 is 5.3 eV, while graphene has a work function 4.56 eV. Because
of work-function mismatch, there will be hole-transfer from SnSe2 to the graphene, and
an in-built electric field will prevent further charge transfer from occurring. When the
optical illumination is switched on, there will be electron-hole pair generation in SnSe2 .
Because of bend-bending at the interface, the electrons will get transferred from SnSe2
to the graphene layer, while the holes get trapped at the interface, which acts as an
effective gate-voltage. The change in resistance is given by ∆R = dR tσ
dVG o r
, where t and r
are the depletion layer-thickness and relative permittivity respectively. The sign of the
photo-response depends on the doping of the graphene, which can be tuned with a gate-
voltage. For VG > VD , the graphene is electron-doped, and for VG < VD , the graphene
channel is hole-doped. During optical illumination, only electrons are transferred to the
channel, so depending on the doping of the graphene, the resistance will increase (when
hole-doped), or reduce (when electron-doped). This is explained schematically in Fig.
B.4.
179
Appendix B. Opto-electronic response in graphene-tin selenide hybrids
(a)4.764 640 nm
(b) 20
920 nm delR/Roff
0.4
OFF ON OFF ON OFF ON dR/dVg
R(10-4)/Roff
4.760
dR(103)/dVG
R (K)
0
4.756 0.0
4.752
-20
4.748
-0.4
0 60 120 180 240 300 360 -20 -10 0 10 20
t (sec) VG (V)
(c) 10 1060 nm
0.4 (d) 1300 nm delR/Roff
4
dR/dVg
dR(103)/dVG
dR(10-4)/dVG
R(10-4)/Roff
R(10-4)/Roff
0
0
0.0 0
-10 -5
-20 -10
-0.4 -4
-20 -10 0 10 20 -20 -10 0 10 20
VG (V) VG (V)
Figure B.3: Opto-electonic response: (a) Resistance as a funtion of time for different
on-off cycles, showing a change in the resistance when optical illumination is switched
on. (b) The fractional change in resistance of the sample for 920 nm wavelength as a
function of wavelength. The change is maximum near where dR/dVG is maximum, and
shows a change in sign across the Dirac point, a signature of photo-gating effect. (c) The
fractional change in resistance of the sample for 1060 nm incident illumination. (d) The
fractional change in resistance of the sample for 1300 nm.
180
B.2. Results and discussions
CB CB
VB VB
Figure B.4: Photogating effect: (a) Mechanism of charge transfer when the graphene
channel is electron-doped. Due to band-bending at the interface, the photo-generated
electrons get transferred to the graphene, while the holes are trapped, leading to an
effective gate-voltage, which is manifested as change in resistance. (b) Charge transfer
mechanism in the hole-doped regime.
640nm
8 920nm
1060nm
D (1012 Jones)
1300nm
6
0
1000 10000
2
P (fW/m )
Figure B.5: Detectivity for different wavelengths.
181
Appendix B. Opto-electronic response in graphene-tin selenide hybrids
Bi2Se3/Si Heterostructure
D*(1010 Jones)
Polymer Photodetector
1
HgTe Nanocrystals
PbS Quantum Dot
GaAs/AlGaAs
Graphene-Te
1E-10
This work
VDS
IP = − 2
4R (B.1)
Rof f
B.3 Summary
• We have measured opto-electronic properties of graphene-SnSe2 hybrids from visi-
ble to near infra-red regime (640 nm to 1300 nm).
182
B.3. Summary
• The photo-response is also persistence free. The resistance comes back to its orig-
inal value after optical illumination is switched off.
• These hybrids devices show a very high responsivity∼ 104 AW−1 , and a very high
specific detectivity ∼ 1012 Jones.
183
Appendix C
• The liquid top gate was prepared by using solid polymer electrolyte consisting of
lithium perchlorate (LiClO4 ) and polyethylene oxide (PEO) in the ratio 0.12 : 1
by weight in methanol.
185
Appendix C. Ambipolar transport in topological insulators
200 μm
Figure C.1: Optical image of a liquid top-gated topological insulator field-effect transistor
450 T = 300 K
R (ohm)
400
350
186
C.2. Device characteristics
C.3 Summary
1. We have observed ambipolar transport in topological insulators at room tempera-
ture.
2. A large effective capacitance due to the liquid gate leads to a large capacitance,
which has been utilized to tune the number density across the Dirac point.
187
Bibliography
[1] Y. Ando, “Topological insulator materials,” J. Phys. Soc. Jpn., vol. 82, no. 10, p.
102001, 2013.
[5] F. Wilczek and A. Shapere, Geometric phases in physics. World Scientific, 1989,
vol. 5.
[6] B. Simon, “Holonomy, the quantum adiabatic theorem, and Berry’s phase,” Phys.
Rev. Lett., vol. 51, no. 24, p. 2167, 1983.
[9] K. v. Klitzing, G. Dorda, and M. Pepper, “New method for high-accuracy deter-
mination of the fine-structure constant based on quantized Hall resistance,” Phys.
Rev. Lett., vol. 45, no. 6, p. 494, 1980.
189
Bibliography
[10] C. L. Kane and E. J. Mele, “Quantum spin Hall effect in graphene,” Phys. Rev.
Lett., vol. 95, no. 22, p. 226801, 2005.
[11] B. A. Bernevig and S.-C. Zhang, “Quantum spin Hall effect,” Phys. Rev. Lett.,
vol. 96, no. 10, p. 106802, 2006.
[12] B. A. Bernevig, T. L. Hughes, and S.-C. Zhang, “Quantum spin Hall effect and
topological phase transition in HgTe quantum wells,” Science, vol. 314, no. 5806,
pp. 1757–1761, 2006.
[15] R. Roy, “Topological phases and the quantum spin Hall effect in three dimensions,”
Phys. Rev. B, vol. 79, no. 19, p. 195322, 2009.
190
Bibliography
[21] H. Zhang, C.-X. Liu, X.-L. Qi, X. Dai, Z. Fang, and S.-C. Zhang, “Topological
insulators in Bi2 Se3 , Bi2 Te3 and Sb2 Te3 with a single Dirac cone on the surface,”
Nat. Phys., vol. 5, no. 6, p. 438, 2009.
[24] E. H. Hall, “On a new action of the magnet on electric currents,” Am. J. Math.,
vol. 2, no. 3, pp. 287–292, 1879.
[26] C.-Z. Chang, J. Zhang, X. Feng, J. Shen, Z. Zhang, M. Guo, K. Li, Y. Ou, P. Wei,
and L.-L. Wang, “Experimental observation of the quantum anomalous Hall effect
in a magnetic topological insulator,” Science, vol. 340, pp. 167–170, 2013.
[28] X.-L. Qi and S.-C. Zhang, “Topological insulators and superconductors,” Rev. Mod.
Phys., vol. 83, no. 4, p. 1057, 2011.
191
Bibliography
[31] S.-P. Chiu and J.-J. Lin, “Weak antilocalization in topological insulator Bi2 Te3
microflakes,” Phys. Rev. B, vol. 87, no. 3, p. 035122, 2013.
[33] T. Ando, T. Nakanishi, and R. Saito, “Berry’s phase and absence of back scattering
in carbon nanotubes,” J. Phys. Soc. Jpn., vol. 67, no. 8, pp. 2857–2862, 1998.
[35] L. Bao, L. He, N. Meyer, X. Kou, P. Zhang, Z.-G. Chen, A. V. Fedorov, J. Zou,
T. M. Riedemann, and T. A. Lograsso, “Weak anti-localization and quantum os-
cillations of surface states in topological insulator Bi2 Se2 Te,” Sci. Rep., vol. 2, p.
726, 2012.
[36] J. Liao, Y. Ou, X. Feng, S. Yang, C. Lin, W. Yang, K. Wu, K. He, X. Ma, and Q.-K.
Xue, “Observation of anderson localization in ultrathin films of three-dimensional
topological insulators,” Phys. Rev. Lett., vol. 114, no. 21, p. 216601, 2015.
192
Bibliography
[43] F. J. Dyson, “Statistical theory of the energy levels of complex systems,” J. Math.
Phys., vol. 3, no. 1, pp. 140–156, 1962.
[45] J.-J. Lin and J. Bird, “Recent experimental studies of electron dephasing in metal
and semiconductor mesoscopic structures,” J. Phys. Condens. Matter, vol. 14,
no. 18, p. R501, 2002.
[48] J.-J. Lin, Y. Zhong, and T. Li, “Effect of annealing on electron dephasing in three-
dimensional polycrystalline metals,” EPL, vol. 57, no. 6, p. 872, 2002.
[51] J. Lin and N. Giordano, “Electron scattering times from weak localization studies
of Au-Pd films,” Phys. Rev. B, vol. 35, no. 3, p. 1071, 1987.
193
Bibliography
[54] D. Pivin Jr, A. Andresen, J. Bird, and D. Ferry, “Saturation of phase breaking in
an open ballistic quantum dot,” Phys. Rev. Lett., vol. 82, no. 23, p. 4687, 1999.
[58] P. Mohanty and R. A. Webb, “Decoherence and quantum fluctuations,” Phys. Rev.
B, vol. 55, no. 20, p. R13452, 1997.
[59] C. Chuang, L.-H. Lin, N. Aoki, T. Ouchi, A. M. Mahjoub, T.-P. Woo, R. K. Puddy,
Y. Ochiai, C. Smith, and C.-T. Liang, “Mesoscopic conductance fluctuations in
multi-layer graphene,” Appl. Phys. Lett., vol. 103, no. 4, p. 043117, 2013.
[60] R. Webb, P. Mohanty, and E. Jariwala, “Intrinsic decoherence and loss of Fermi liq-
uid behavior in mesoscopic systems,” Fortschritte der Physik: Progress of Physics,
vol. 46, no. 6-8, pp. 779–789, 1998.
194
Bibliography
[63] D. Cohen and Y. Imry, “Dephasing at low temperatures,” Phys. Rev. B, vol. 59,
no. 17, p. 11143, 1999.
[66] I. Aleiner, B. Altshuler, and Y. M. Galperin, “Experimental tests for the relevance
of two-level systems for electron dephasing,” Phys. Rev. B, vol. 63, no. 20, p.
201401, 2001.
[72] A. Huibers, J. Folk, S. Patel, C. Marcus, C. Duruöz, and J. Harris Jr, “Low-
temperature saturation of the dephasing time and effects of microwave radiation
on open quantum dots,” Phys. Rev. Lett., vol. 83, no. 24, p. 5090, 1999.
195
Bibliography
[74] J. Wang, A. M. DaSilva, C.-Z. Chang, K. He, J. Jain, N. Samarth, X.-C. Ma, Q.-
K. Xue, and M. H. Chan, “Evidence for electron-electron interaction in topological
insulator thin films,” Phys. Rev. B, vol. 83, no. 24, p. 245438, 2011.
[76] J. Liao, Y. Ou, H. Liu, K. He, X. Ma, Q.-K. Xue, and Y. Li, “Enhanced electron
dephasing in three-dimensional topological insulators,” Nat. Commun., vol. 8, p.
16071, 2017.
[77] Z. Li, T. Chen, H. Pan, F. Song, B. Wang, J. Han, Y. Qin, X. Wang, R. Zhang,
J. Wan et al., “Two-dimensional universal conductance fluctuations and the
electron-phonon interaction of surface states in Bi2 Te2 Se microflakes,” Sci. Rep.,
vol. 2, p. 595, 2012.
[80] H.-T. He, G. Wang, T. Zhang, I.-K. Sou, G. K. Wong, J.-N. Wang, H.-Z. Lu, S.-Q.
Shen, and F.-C. Zhang, “Impurity effect on weak antilocalization in the topological
insulator Bi2 Te3 ,” Phys. Rev. Lett., vol. 106, no. 16, p. 166805, 2011.
[81] H.-Z. Lu, J. Shi, and S.-Q. Shen, “Competition between weak localization and
antilocalization in topological surface states,” Phys. Rev. Lett., vol. 107, no. 7, p.
076801, 2011.
[82] M. Liu, J. Zhang, C.-Z. Chang, Z. Zhang, X. Feng, K. Li, K. He, L.-l. Wang,
X. Chen, X. Dai et al., “Crossover between weak antilocalization and weak local-
ization in a magnetically doped topological insulator,” Phys. Rev. Lett., vol. 108,
no. 3, p. 036805, 2012.
196
Bibliography
[85] Z. Li, T. Chen, H. Pan, F. Song, B. Wang, J. Han, Y. Qin, X. Wang, R. Zhang,
J. Wan et al., “Two-dimensional universal conductance fluctuations and the
electron-phonon interaction of surface states in Bi2 Te2 Se microflakes,” Sci. Rep.,
vol. 2, p. 595, 2012.
[86] X. Hong, K. Zou, B. Wang, S.-H. Cheng, and J. Zhu, “Evidence for spin-flip scat-
tering and local moments in dilute fluorinated graphene,” Phys. Rev. Lett., vol.
108, no. 22, p. 226602, 2012.
[87] Z. Li, Y. Meng, J. Pan, T. Chen, X. Hong, S. Li, X. Wang, F. Song, and
B. Wang, “Indications of topological transport by universal conductance fluctu-
ations in Bi2 Te2 Se microflakes,” APEX, vol. 7, no. 6, p. 065202, 2014.
[89] J. B. Johnson, “Thermal agitation of electricity in conductors,” Phys. Rev., vol. 32,
no. 1, p. 97, 1928.
[91] Y. M. Blanter and M. Büttiker, “Shot noise in mesoscopic conductors,” Phys. Rep.,
vol. 336, no. 1-2, pp. 1–166, 2000.
[92] A. Van der Ziel, Noise in solid state devices and circuits. Wiley-Interscience, 1986.
197
Bibliography
[98] M. Weissman, “1/f noise and other slow, nonexponential kinetics in condensed
matter,” Rev. Mod. Phys, vol. 60, no. 2, p. 537, 1988.
[100] P. Dutta and P. Horn, “Low-frequency fluctuations in solids: 1/f noise,” Rev. Mod.
Phys., vol. 53, no. 3, p. 497, 1981.
[101] P. Karnatak, T. Paul, S. Islam, and A. Ghosh, “1/f noise in Van der Waals mate-
rials and hybrids,” Adv. Phys. X, vol. 2, no. 2, pp. 428–449, 2017.
[102] F. N. Hooge, “1/f noise,” Physica B+ C, vol. 83, no. 1, pp. 14–23, 1976.
[103] F. Hooge, “1/f noise sources,” IEEE Trans. on Electron Devices, vol. 41, no. 11,
pp. 1926–1935, 1994.
198
Bibliography
[105] R. Jayaraman and C. G. Sodini, “A 1/f noise technique to extract the oxide trap
density near the conduction band edge of silicon,” IEEE Trans. Electron Devices,
vol. 36, no. 9, pp. 1773–1782, 1989.
[106] K. K. Hung, P. K. Ko, C. Hu, and Y. C. Cheng, “A unified model for the flicker
noise in metal-oxide-semiconductor field-effect transistors,” IEEE Trans. on Elec-
tron Devices, vol. 37, no. 3, pp. 654–665, 1990.
[110] Y. Zhang, K. He, C.-Z. Chang, C.-L. Song, L.-L. Wang, X. Chen, J.-F. Jia, Z. Fang,
X. Dai, and W.-Y. Shan, “Crossover of the three-dimensional topological insulator
Bi2 Se3 to the two-dimensional limit,” Nat. Phys., vol. 6, no. 8, p. 584, 2010.
[112] A. N. Pal and A. Ghosh, “Resistance noise in electrically biased bilayer graphene,”
Phys. Rev. Lett., vol. 102, no. 12, p. 126805, 2009.
199
Bibliography
[115] A. Rogalski, “History of infrared detectors,” Opto-Electron. Rev., vol. 20, no. 3, pp.
279–308, 2012.
[120] C.-H. Liu, Y.-C. Chang, T. B. Norris, and Z. Zhong, “Graphene photodetectors
with ultra-broadband and high responsivity at room temperature,” Nat. Nanotech-
nol., vol. 9, no. 4, pp. 273–278, 2014.
[125] F. Bonaccorso, Z. Sun, T. Hasan, and A. Ferrari, “Graphene photonics and opto-
electronics,” Nat. Photonics, vol. 4, no. 9, p. 611, 2010.
200
Bibliography
[127] H. Zhang, X. Zhang, C. Liu, S.-T. Lee, and J. Jie, “High-responsivity, high-
detectivity, ultrafast topological insulator Bi2 Se3 /Silicon heterostructure broad-
band photodetectors,” ACS Nano, vol. 10, no. 5, pp. 5113–5122, 2016.
[128] K. Zheng, L.-B. Luo, T.-F. Zhang, Y.-H. Liu, Y.-Q. Yu, R. Lu, H.-L. Qiu, Z.-
J. Li, and J. A. Huang, “Optoelectronic characteristics of a near infrared light
photodetector based on a topological insulator Sb2 Te3 film,” J. Mater. Chem. C,
vol. 3, no. 35, pp. 9154–9160, 2015.
[129] H. Qiao, J. Yuan, Z. Xu, C. Chen, S. Lin, Y. Wang, J. Song, Y. Liu, Q. Khan, and
H. Y. Hoh, “Broadband photodetectors based on graphene-Bi2 Te3 heterostructure,”
ACS Nano, vol. 9, no. 2, pp. 1886–1894, 2015.
[131] X. Xu, N. M. Gabor, J. S. Alden, A. M. van der Zande, and P. L. McEuen, “Photo-
thermoelectric effect at a graphene interface junction,” Nano Lett., vol. 10, no. 2,
pp. 562–566, 2009.
[132] J. G. Checkelsky, Y. S. Hor, R. J. Cava, and N. Ong, “Bulk band gap and surface
state conduction observed in voltage-tuned crystals of the topological insulator
Bi2 Se3 ,” Phys. Rev. Lett., vol. 106, no. 19, p. 196801, 2011.
201
Bibliography
[138] P. Blake, E. Hill, A. Castro Neto, K. Novoselov, D. Jiang, R. Yang, T. Booth, and
A. Geim, “Making graphene visible,” Appl. Phys. Lett., vol. 91, no. 6, p. 063124,
2007.
[141] G. Binnig and H. Rohrer, “Scanning tunneling microscopy from birth to adoles-
cence,” Rev. Mod. Phys., vol. 59, no. 3, p. 615, 1987.
[144] T. F. Scientific, “Raman spectroscopy,” Beverly, MA, USA, pp. 33–37, 2015.
202
Bibliography
[146] O. Lounasmaa, “Dilution refrigeration,” J. Phys. E: Sci. Instrum., vol. 12, no. 8,
p. 668, 1979.
[147] K. Roy, “Optoelectronic properties of graphene based Van der Waals hybrids,”
2017.
[149] J. H. Scofield, “AC method for measuring low-frequency resistance fluctuation spec-
tra,” Rev. Sci. Instrum, vol. 58, no. 6, pp. 985–993, 1987.
[150] P. Welch, “The use of fast Fourier transform for the estimation of power spectra: a
method based on time averaging over short, modified periodograms,” IEEE Trans.
Signal Process., vol. 15, no. 2, pp. 70–73, 1967.
[151] D. Kong and Y. Cui, “Opportunities in chemistry and materials science for topo-
logical insulators and their nanostructures,” Nat. Chem., vol. 3, no. 11, p. 845,
2011.
[154] D.-X. Qu, S. K. Roberts, and G. F. Chapline, “Observation of huge surface hole
mobility in the topological insulator Bi0.91 Sb0.09 (111),” Phys. Rev. Lett., vol. 111,
no. 17, p. 176801, 2013.
203
Bibliography
across the topological phase transition in (Bi1−x Inx )2 Se3 ,” Nat. Phys., vol. 9, no. 7,
p. 410, 2013.
[161] Y. Fan, P. Upadhyaya, X. Kou, M. Lang, S. Takei, Z. Wang, J. Tang, L. He, L.-
T. Chang, M. Montazeri et al., “Magnetization switching through giant spin–orbit
torque in a magnetically doped topological insulator heterostructure,” Nat. Mat.,
vol. 13, no. 7, p. 699, 2014.
[163] J. Chen, X. He, K. Wu, Z. Ji, L. Lu, J. Shi, J. Smet, and Y. Li, “Tunable sur-
face conductivity in Bi2 Se3 revealed in diffusive electron transport,” Phys. Rev. B,
vol. 83, no. 24, p. 241304, 2011.
204
Bibliography
[167] J. Tian, C. Chang, H. Cao, K. He, X. Ma, Q. Xue, and Y. P. Chen, “Quantum
and classical magnetoresistance in ambipolar topological insulator transistors with
gate-tunable bulk and surface conduction,” Sci. Rep., vol. 4, p. 4859, 2014.
[168] X. He, T. Guan, X. Wang, B. Feng, P. Cheng, L. Chen, Y. Li, and K. Wu, “Highly
tunable electron transport in epitaxial topological insulator (Bi1−x Sbx )2 Te3 thin
films,” App. Phys. Lett., vol. 101, no. 12, p. 123111, 2012.
[170] J.-H. Chen, C. Jang, S. Adam, M. Fuhrer, E. Williams, and M. Ishigami, “Charged-
impurity scattering in graphene,” Nat. Phys., vol. 4, no. 5, p. 377, 2008.
[173] S. Ghatak, A. N. Pal, and A. Ghosh, “Nature of electronic states in atomically thin
MoS2 field-effect transistors,” ACS Nano, vol. 5, no. 10, pp. 7707–7712, 2011.
205
Bibliography
[177] P. V. Voorde and W. Love, “Magnetic effects on 1/f noise in n-InSb,” Phys. Rev.
B, vol. 24, no. 8, p. 4781, 1981.
[178] J. E. Moore, “The birth of topological insulators,” Nature, vol. 464, no. 7286, p.
194, 2010.
[180] L. Bao, W. Wang, N. Meyer, Y. Liu, C. Zhang, K. Wang, P. Ai, and F. Xiu, “Quan-
tum corrections crossover and ferromagnetism in magnetic topological insulators,”
Sci. Rep., vol. 3, p. 2391, 2013.
[181] N. Birge, B. Golding, and W. Haemmerle, “Conductance fluctuations and 1/f noise
in Bi,” Phys. Rev. B, vol. 42, no. 5, p. 2735, 1990.
206
Bibliography
[190] J. Pelz and J. Clarke, “Quantitative "local-interference" model for 1/f noise in
metal films,” Phys. Rev. B, vol. 36, no. 8, p. 4479, 1987.
207
Bibliography
[202] M. Liu, C.-Z. Chang, Z. Zhang, Y. Zhang, W. Ruan, K. He, L.-l. Wang, X. Chen,
J.-F. Jia, S.-C. Zhang et al., “Electron interaction-driven insulating ground state in
Bi2 Se3 topological insulators in the two-dimensional limit,” Phys. Rev. B, vol. 83,
no. 16, p. 165440, 2011.
[203] K. R. Amin, S. S. Ray, N. Pal, R. Pandit, and A. Bid, “Exotic multifractal con-
ductance fluctuations in graphene,” Commun. Phys., vol. 1, no. 1, p. 1, 2018.
208
Bibliography
[207] Y. Baum and A. Stern, “Magnetic instability on the surface of topological insula-
tors,” Phys. Rev. B, vol. 85, no. 12, p. 121105, 2012.
[209] R. Du, H.-C. Hsu, A. C. Balram, Y. Yin, S. Dong, W. Dai, W. Zhao, D. Kim, S.-Y.
Yu, and J. Wang, “Robustness of topological surface states against strong disorder
observed in Bi2 Te3 nanotubes,” Phys. Rev. B, vol. 93, no. 19, p. 195402, 2016.
[211] A. Efros and B. Shklovskii, “Coulomb gap and low temperature conductivity of
disordered systems,” J. Phys. C: Solid State Phys., vol. 8, no. 4, p. L49, 1975.
[212] A. Efros, “Coulomb gap in disordered systems,” J. Phys. C: Solid State Phys.,
vol. 9, no. 11, p. 2021, 1976.
[213] A. Efros, B. Skinner, and B. Shklovskii, “Coulomb gap in the one-particle density of
states in three-dimensional systems with localized electrons,” Phys. Rev. B, vol. 84,
no. 6, p. 064204, 2011.
[214] B. Skinner, T. Chen, and B. Shklovskii, “Effects of bulk charged impurities on the
bulk and surface transport in three-dimensional topological insulators,” J. Exp.
Theor. Phys., vol. 117, no. 3, pp. 579–592, 2013.
209
Bibliography
[215] B. Skinner and B. Shklovskii, “Theory of the random potential and conductivity
at the surface of a topological insulator,” Phys. Rev. B, vol. 87, no. 7, p. 075454,
2013.
[216] F. Van Keuls, X. Hu, H. Jiang, and A. Dahm, “Screening of the Coulomb interaction
in two-dimensional variable-range hopping,” Phys. Rev. B, vol. 56, no. 3, p. 1161,
1997.
[217] I. Aleiner and Y. M. Blanter, “Inelastic scattering time for conductance fluctua-
tions,” Phys. Rev. B, vol. 65, no. 11, p. 115317, 2002.
210
Bibliography
[228] X. Zhang, J. Wang, and S.-C. Zhang, “Topological insulators for high-performance
terahertz to infrared applications,” Phys. Rev. B, vol. 82, no. 24, p. 245107, 2010.
[231] Y. Chen, J. G. Analytis, J.-H. Chu, Z. Liu, S.-K. Mo, X.-L. Qi, H. Zhang, D. Lu,
X. Dai, and Z. Fang, “Experimental realization of a three-dimensional topological
insulator, Bi2 Te3 ,” Science, vol. 325, no. 5937, pp. 178–181, 2009.
[233] D. Park, S. Park, K. Jeong, H.-S. Jeong, J. Y. Song, and M.-H. Cho, “Thermal
and electrical conduction of single-crystal Bi2 Te3 nanostructures grown using a one
step process,” Sci. Rep., vol. 6, p. 19132, 2016.
211
Bibliography
[239] B. Yan and C. Felser, “Topological materials: Weyl semimetals,” Ann. Rev. Cond.
Mat. Phys., vol. 8, pp. 337–354, 2017.
[240] S. Jia, S.-Y. Xu, and M. Z. Hasan, “Weyl semimetals, Fermi arcs and chiral anoma-
lies,” Nat. Mat., vol. 15, no. 11, p. 1140, 2016.
212
Bibliography
[248] F. Islam, J. Wu, J. Jansson, and D. Wilson, “Relative risk of cardiovascular disease
among people living with HIV: a systematic review and meta-analysis,” HIV Med.,
vol. 13, no. 8, pp. 453–468, 2012.
213
Bibliography
[256] I.-H. Cho, E.-H. Paek, Y.-K. Kim, J.-H. Kim, and S.-H. Paek, “Chemiluminometric
enzyme-linked immunosorbent assays (ELISA)-on-a-chip biosensor based on cross-
flow chromatography,” Anal. Chim. Acta., vol. 632, no. 2, pp. 247–255, 2009.
[261] Y. Zhao, Y. Liu, X. Li, H. Wang, Y. Zhang, H. Ma, and Q. Wei, “Label-free ecl
immunosensor for the early diagnosis of rheumatoid arthritis based on asymmetric
214
Bibliography
215
Bibliography
216