New Approaches To The Realization and Identification of Majorana Qubits in Solid State Quantum Devices

Als pdf oder txt herunterladen
Als pdf oder txt herunterladen
Sie sind auf Seite 1von 147

New approaches to the realization and

identification of Majorana qubits in solid


state quantum devices

Inaugural-Dissertation

zur

Erlangung des Doktorgrades

der Mathematisch-Naturwissenschaftlichen Fakultät

der Universität zu Köln

vorgelegt von

Jan Manousakis

Köln 2020
Berichterstatter:
Prof. Dr. Alexander Altland
Prof. Dr. Simon Trebst

Tag der letzten mündlichen Prüfung: 29.05.2020


Abbreviations

The following abbreviations will be used throughout this dissertation:

TI: Topological insulator

TSC: Topological superconductor

SC: Superconductor

MBS: Majorana bound state

MZM: Majorana zero-mode

PHS: Particle hole symmetry

ABS: Andreev bound state

FCS: Full counting statistics

BdG: Bogoliubov-de Gennes

BCS: Bardeen-Cooper-Schrieffer

1D: one-dimensional

ZBCP: zero-bias conductance peak

SM: Semiconductor
Abstract

Majorana bound states in topological superconductors exhibit exotic non-Abelian braid-


ing statistics and hold promise for particularly robust qubits with natural built-in mech-
anisms against decoherence. The theme of this dissertation is the theory of novel ap-
proaches to realization and identification of such Majorana qubits.
Towards the realization of Majorana qubits, we present architectures based on topolog-
ical insulator nanoribbons, e.g. made of Bi2 Se3 , and proximitized by an s-wave super-
conductor. Piercing of proximitized nanoribbons with an axial uniform magnetic flux of
suitably adjusted strength has been previously predicted to give rise to one-dimensional
topological superconductors with robust Majorana bound states. We propose qubit de-
signs that incorporate two such topological superconductors connected by a constricted
topological nanoribbon segment. This constriction is non-proximitized and its lesser
cross section results in a local gap opening. We present theoretical results showing
the possibility to conveniently tune the coupling of a pair of Majorana states localized
across the constriction via gating. Moreover, we discuss proof-of-principle experiments
for initialization, manipulation, and readout of the floating version of the device, which
is dominated by charging effects. We compare the platform to other Majorana qubit
proposals and give an outlook on applications such as the Majorana surface code.
The experimental identification of Majorana bound states represents one of the out-
standing goals of contemporary condensed matter physics. Towards identification, we
present the theory of novel transport spectroscopic approaches geared to qubits in the
Coulomb blockade regime. In particular, we propose a scheme in which three normal-
conducting leads are weakly coupled to three different Majorana bound states of the
qubit. The protocol relies on the simultaneous continuous weak measurement of two
noncommuting, nonlocal Pauli operators of the Majorana qubit and results in a phe-
nomenon of surprisingly strong current cross-correlations. This is the prime signature
containing information that enables to identify the nonlocal Pauli algebra, which is

3
intimately related to the celebrated non-Abelian braiding statistics. The latter is a
property notoriously hard to demonstrate and of large attractiveness from the fun-
damental as well as applied perspective. The conditions under which the pronounced
current cross-correlations are observable depend on the device configuration, a fact that
leads to several experimentally verifiable predictions that allow to test the authenticity
of the Majorana qubit. Beyond that, we put forward two further detection methods in
this thesis. First, a shot noise scheme which is viable for a single floating topological
Majorana wire. Second, a protocol relying on projective current measurements. Beyond
the usefulness of these protocols, we identify the aforementioned protocol of monitoring
a nonlocal Pauli algebra as the scheme accessing the most information related to the
constitutive nature of Majorana bound states.
Zusammenfassung

Gebundene Majorana-Zustände in topologischen Supraleitern weisen exotische nicht-


abelsche Statistik auf und versprechen besonders robuste Qubit-Realisierungen mit
Schutzmechanismen gegen Dekohärenz. Das Thema der vorliegenden Dissertation ist
die Theorie neuer Methoden zur Realisierung und Identifizierung solcher Majorana-
Qubits.
Im Hinblick auf die Realisierung von Majorana-Qubits stellen wir Designarchitek-
turen vor, die auf topologischen Isolator-Nanodrähten beruhen. Letztere können etwa
aus dem Material Bi2 Se3 bestehen. Vorangegangene Arbeiten haben derartige Nano-
drähte in Proximität zu s-Wellen-Supraleitern untersucht und vorhergesagt, dass ein
gleichmäßiger axialer magnetischer Fluss von der Stärke eines halben Flussquantums
einen eindimensionalen topologischen Supraleiter mit gebundenen Majorana-Zuständen
entstehen lässt. In unseren Qubit-Designentwürfen werden zwei solche topologische
Supraleiter durch einen verengten Abschnitt aus topologischem Nanodraht verbun-
den. Der geringe Querschnitt dieser Verengung führt lokal zu einer Energielücke.
Wir präsentieren theoretische Resultate, die die Möglichkeit einer bequem manipulier-
baren Kopplung der Majorana-Zustände demonstrieren. Des Weiteren diskutieren wir
Proof of Principle Experimente zur Initialisierung, Manipulation und Auslesung un-
serer Qubit-Platform unter Bedingungen der Coulomb-Blockade. Wir vergleichen die
Plattform mit anderen Ansätzen der Majorana-Qubit Realisierung und geben einen
Ausblick auf weitergehende Anwendungen wie etwa den Majorana-Oberflächencode.
Die experimentelle Identifizierung von Majorana-Zuständen gehört zu den herausragen-
den Zielen der gegenwärtigen Festkörperphysik. Zu diesem Zweck stellen wir neuartige
transportspektroskopische Methoden vor, die auf Majorana-Qubits unter Bedingungen
der Coulomb-Blockade ausgelegt sind. Es ist Teil der Methode, dass drei normallei-
tende Elektroden schwach an drei verschiedene Majorana-Zustände des Qubits gekop-
pelt werden. Das Protokoll stützt sich auf die simultane kontinuierliche schwache Mes-

5
sung zweier nicht-kommutierender, nicht-lokaler Pauli-Operatoren des Majorana-Qubits
und führt zu einem Phänomen überraschend starker Strom-Kreuzkorrelationen. Diese
experimentell messbare Größe enthält Informationen, die es ermöglichen, die nicht-
lokale Pauli-Algebra zu identifizieren, welche eng mit der nichtabelschen Statistik ver-
wandt ist. Die nichtabelsche Statistik ist eine Eigenschaft, die berüchtigt dafür ist, dass
sie schwer nachzuweisen ist, und sowohl aus fundamentaler als auch aus angewandter
Perspektive eine große Attraktivität besitzt. Die Bedingungen, unter denen die aus-
geprägten Kreuzkorrelationen vorherrschen, hängen von der Gerätekonfiguration ab;
eine Tatsache, die zu mehreren experimentell nachprüfbaren Vorhersagen führt. Diese
Vorhersagen erlauben es, die Authentizität des Majorana-Qubits zu testen. Darüber
hinaus schlagen wir in dieser Arbeit zwei weitere Methoden vor, um Majorana-Zustände
zu detektieren. Erstens Schrotrausch-Messungen an einem nicht geerdeten, topolo-
gischen Majoranadraht und zweitens ein Protokoll, das sich auf projektive Strommes-
sungen stützt. Über die Nützlichkeit dieser beiden Protokolle hinaus macht das oben
beschriebene Monitoring einer nicht-lokalen Pauli-Algebra mehr Information über die
Natur der Majorana-Zustände zugänglich.
Contents

1 Introduction 11
1.1 This thesis and its structure . . . . . . . . . . . . . . . . . . . . . . . . 12

2 Majorana bound states in solid state quantum devices 15


2.1 Topological Majorana nanowires . . . . . . . . . . . . . . . . . . . . . . 15
2.1.1 Majorana wires in the framework of topological matter . . . . . 15
2.1.2 Kitaev’s p-wave superconducting lattice model . . . . . . . . . . 17
2.1.3 Physical realization of 1D p-wave SC . . . . . . . . . . . . . . . 19
2.2 Topological Majorana qubits . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2.1 Majorana qubit interferometric readout . . . . . . . . . . . . . . 23
2.3 The non-Abelian nature of Majorana zero-modes . . . . . . . . . . . . 25
2.3.1 Non-Abelian braiding statistics . . . . . . . . . . . . . . . . . . 25
2.3.2 Non-Abelian fusion rules . . . . . . . . . . . . . . . . . . . . . . 27
2.3.3 Topological quantum computation . . . . . . . . . . . . . . . . . 28
2.4 Experimental identification of Majorana bound states . . . . . . . . . . 30
2.4.1 Next generation of experiments beyond local probes . . . . . . . 32

3 Majorana qubits in proximitized TI nanoribbon device architectures 35


3.1 TI nanoribbons and the emergence of MBSs in proximity to an s-wave SC 35
3.1.1 TI nanoribbons and surface Dirac theory . . . . . . . . . . . . . 35

7
CONTENTS

3.1.2 Dirac fermion model of proximitized TI nanoribbons . . . . . . 38


3.1.3 MBS at the interface of TI wire segments of different width . . . 41
3.2 Gate tunable coupling of Majorana states across a constriction . . . . . 44
3.3 Majorana box qubits from TI nanoribbons . . . . . . . . . . . . . . . . 50
3.3.1 Floating box qubit and elementary quantum operations . . . . . 50
3.3.2 Devices with switchable grounding . . . . . . . . . . . . . . . . 52
3.3.3 Majorana qubit comparison: TI nanoribbon vs SM platform . . 53
3.4 Conclusions and outlook . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.4.1 Outlook: Majorana surface code and beyond . . . . . . . . . . . 57

4 Majorana qubit detection via simultaneous weak measurement of its


nonlocal Pauli operators 61
4.1 Experimental setting and qualitative discussion . . . . . . . . . . . . . 62
4.1.1 Device geometries . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2 Effective Hamiltonian for the weak measurement settings . . . . . . . . 67
4.3 Formalism for weak measurement protocols applied to Majorana devices 70
4.3.1 Method of full counting statistics . . . . . . . . . . . . . . . . . 70
4.3.2 Modified Liouville-von Neumann equation . . . . . . . . . . . . 73
4.3.3 Derivation of the quantum master equation . . . . . . . . . . . 74
4.3.4 Resulting quantum master equation for weak measurement setup 79
4.4 Phenomenology of the Majorana box qubit . . . . . . . . . . . . . . . . 81
4.4.1 Derivation of the generating function for the Majorana qubit . . 82
4.4.2 Pronounced shot noise cross-correlations and qubit evolution . . 84
4.4.3 The effects of finite temperature . . . . . . . . . . . . . . . . . . 87
4.4.4 Outcome distribution and extreme value statistics . . . . . . . . 88
4.4.5 The effects of finite MBS coupling on the box qubit . . . . . . . 91
4.5 Counting statistics for Andreev bound states . . . . . . . . . . . . . . . 92

8
CONTENTS

4.5.1 Counting statistics for source lead coupled to ABS . . . . . . . . 93


4.5.2 Counting statistics for drain leads coupled to ABSs . . . . . . . 95
4.6 Weak measurement protocols for MBS detection . . . . . . . . . . . . . 97
4.7 Conclusions and outlook . . . . . . . . . . . . . . . . . . . . . . . . . . 100

5 Further new approaches to MBS detection 103


5.1 Current cross-correlations for a single Majorana quantum wire . . . . . 103
5.1.1 Three-terminal device and theoretical model . . . . . . . . . . . 103
5.1.2 Signatures of current cross-correlations . . . . . . . . . . . . . . 106
5.2 MBS detection via number of projective current outcomes . . . . . . . 108
5.3 Conclusions and outlook . . . . . . . . . . . . . . . . . . . . . . . . . . 112

6 Overall conclusion and outlook 115

Appendices 117
Appendix A: Analysis of spinor wave functions . . . . . . . . . . . . . . . . . 117
Appendix B: Effective tunnel Hamiltonian of three-terminal device . . . . . 120
Appendix C: Further details on the derivation of the master equation . . . . 123
Appendix D: Counting statistics for Majorana box qubit . . . . . . . . . . . 125
Appendix E: Counting statistics for Andreev bound states . . . . . . . . . . 127
Appendix F: Further new approaches to MBS detection . . . . . . . . . . . . 130

Bibliography 135

Acknowledgements 145

9
CONTENTS

10
Chapter 1

Introduction

In 1928, Paul Dirac published the relativistic quantum wave equation [1], which in his
honor is known as the Dirac equation and constitutes one of the pillars of our modern
scientific world view. This fundamental contribution enabled a unified view of quantum
mechanics and the special theory of relativity, laid the groundwork of quantum field
theory, and inspired Dirac himself to make an important prediction about nature
the existence of antimatter. Ettore Majorana made a landmark contribution in 1937
proving the existence of a representation of the Dirac equation with a purely real wave
function, Ψ = Ψ∗ , thus, postulating a neutral fermionic excitation [2]. Ever since, the
notion of a neutral fermion being its own antiparticle became reputable from the vantage
point of established theoretical physics. Whether nature chooses implementation of this
scenario in the form of an elementary particle is still experimentally undecided with the
neutrino being the most prominent suspect [3].
Interestingly, the time-dependent quantum fields Ψ(r, t) describing Bogoliubov quasi-
particles in superconductors within the framework of the well established Bardeen-
Cooper-Schrieffer (BCS) theory share the mathematical properties of the neutral
fermions studied by Majorana [3]. This becomes less surprising when we recall that
these quasiparticles constitute superpositions of electrons and electron holes, which play
a role analogous to that of matter and antimatter particles in the discussion above. This
thesis is centered around the closely related concept of the Majorana zero-mode (MZM)
[4], which is the cause for great activity and interest in the field of condensed matter
physics. The particle hole symmetry (PHS) of the superconductor is described by an
anti-unitary operator P which satisfies Φ−E (r) = P ΦE (r) when we express the quasi-

11
1.1. THIS THESIS AND ITS STRUCTURE

particle as Ψ(r, t) = e−iEt ΦE (r). Hence, the PHS dictates that the spectrum exhibits
symmetry around zero energy. A MZM Φ0 (r) = P Φ0 (r) in the presence of a finite en-
ergy gap ∆ and remote from other MZMs is therefore protected by symmetry: it cannot
move away from zero energy. This protection is very robust and a consequence of non-
trivial topology of the bulk of the special host systems, which are known as topological
superconductors (TSCs) [5]. A paradigmatic example of TSCs is the Kitaev model [6].
The second quantized operator creating a MZM localized at the domain walls or vor-
tices of TCSs is identical to the operator annihilating it, i.e. γ † = γ. Disappointingly,
topological insulators are rarely occurring in nature [4], but inventive ideas to engineer
them by forming heterostructures have been put forward. This approach was initialized
by an innovative proposal of Fu and Kane [7]. The latter led to prominent proposals
involving proximitized semiconducting nanowires with spin-orbit coupling [8, 9] making
the above mentioned Kitaev model experimentally accessible. We emphasize that, in
contrast to the neutral fermions studied by Majorana in 1937, MZMs in TSCs come
into existence as an emergent phenomenon caused by the constituent electrons of the
material. The most noteworthy difference, however, is that MZMs bound to defects
adhere to a special kind of statistics known by the scholars of theoretical physics as
non-Abelian anyonic. The exchange of identical non-Abelian anyons is radically differ-
ent from that of identical fermions or bosons it can result in a different quantum state
at the same energy which may pave the way to topological quantum computation [10].
Hence, such exotic properties are not only intriguing from the point of view of funda-
mental physics, but also because of the prospect of technological advances in quantum
information processing hardware.

1.1 This thesis and its structure

The theme of this dissertation is the theory of novel approaches to realize and de-
tect Majorana qubits, which have been envisioned and deemed attractive due to their
topological protection as the building blocks of quantum information processing im-
plementations. This is founded on the hope that at low energies Majorana qubits
may provide for natural built-in mechanisms against detrimental decoherence and de-
cay [10, 11, 12]. Towards realization of such systems, we propose device designs of
Majorana qubits relying on previously proposed MBS platforms based on topological
insulator (TI) nanowires proximitized by an s-wave superconductor (SC) in the pres-

12
CHAPTER 1. INTRODUCTION

ence of a longitudinally aligned magnetic field [13, 14]. Towards detection, we note that
platforms that are predicted to contain unpaired MBSs are studied in several labora-
tories around the world, including but not limited to superconductor–semiconductor
heterostructures in Copenhagen and Delft [15]. This in itself is a testament to the
intense efforts in the condensed matter community in recent years. Although a wealth
of accumulated measured data consistent with MBS predictions exists, the unambigu-
ous demonstration of topological MBSs has proven to be difficult as we will discuss in
more detail in Section 2.4. The three main aspects which have to be investigated to
demonstrate MBSs are broadly speaking the following: (a) verification of non-Abelian
braiding properties, (b) corroboration that two MBSs constitute a vastly nonlocal Dirac
fermion, and (c) demonstration of localized Majorana bound states in TSCs. Property
(a) is the ultimate goal but also the most demanding. In this thesis, we provide theoret-
ical blueprints for a next generation of experiments to detect the Majorana qubit which
can as well be understood as detecting the sought-after unpaired MBS itself. First, we
are guided by the requirement that the protocols should substantiate aspects (a) and
(b) by going beyond local probes. Second, reliance on presently available hardware is
important for experiments that can be implemented realistically in the short term to
avoid stalling of progress in the field. We make the argument that a good balance of
information gain and experimental feasibility is found in transport spectroscopy where
we propose protocols that lie within the scope of present day experimental physics. The
protocols aim at the core property of the Majorana qubit, its nonlocal Pauli algebra, by
accessing information closely related to the celebrated non-Abelian braiding statistics.
The thesis is organized in six chapters:

ˆ Chapter 2: In this chapter, we review several aspects of the physics of MBSs


in solid state quantum devices aiming to make this thesis self-contained. We
briefly explain the broader context of topological matter and discuss the quest for
MBSs in 1D condensed matter systems. We go on to discuss Majorana box qubits
and their associated nonlocal Pauli algebra as well as the celebrated non-Abelian
braiding statistics. Finally, we discuss the state of the presently accumulated evi-
dence for MBSs as well as existing proposals for a next generation of experiments.

ˆ Chapter 3: In this chapter, we propose new architectures for the realization of


Majorana qubits based on platforms of topological insulator (TI) nanowires in
proximity to s-wave superconductors (SCs). We show quantitatively that the

13
1.1. THIS THESIS AND ITS STRUCTURE

coupling of the robust MBSs in these architectures is conveniently manipulable.


We then go on to discuss proof-of-principle experiments and applications for the
proposed platforms. Moreover, we analyze design advantages and drawbacks in
comparison to other approaches to Majorana qubit realization.

ˆ Chapter 4: This chapter is central to the thesis and details novel detection pro-
tocols for MBSs. The protocols are based on simultaneous weak measurement of
the nonlocal Pauli operators of the Majorana qubit. We discuss in great detail a
predicted strong effect of current cross-correlations, which allows the identifica-
tion of genuine Majorana qubits. The protocols yield information similar to that
of a full braiding protocol.

ˆ Chapter 5: In this chapter, we discuss two further new variants of MBS detection
protocols. The discussion at the end of the chapter reveals that the aforemen-
tioned protocol presented in Chapter 4 accesses the most information related to
the fundamental nature of MBSs.

ˆ Chapter 6: In the last chapter, we provide an overall conclusion and outlook.

14
Chapter 2

Majorana bound states in solid


state quantum devices

2.1 Topological Majorana nanowires

The field of topological insulators (TIs) and topological superconductors (TSCs) is


one of the most vibrant and active fields in contemporary condensed matter physics
[16, 17]. In this section, we put 1D Majorana wires in the context of topological matter,
discuss an important model of 1D p-wave superconductivity and subsequently discuss
its implementation. For further reading, we note that numerous informative review
articles are available [3, 4, 5, 18, 19, 20].

2.1.1 Majorana wires in the framework of topological matter

Topological Majorana wires are quasi one-dimensional superconducting representatives


of a celebrated form of matter known as symmetry-protected topological matter, which
provides a theoretically unified view on topological insulators (TIs) and topological
superconductors (TSCs) [21]. Conceptually, topological matter is founded on the topo-
logical classification of noninteracting gapped fermionic Hamiltonians based on their
symmetry. According to a fundamental result of Altland and Zirnbauer, there are
generically precisely ten symmetry classes that the system can belong to [22]. This
result is also known as the “tenfold way” for disordered fermions. In the presence of
a gap, these free fermion systems are topologically classified in the “periodic table for

15
2.1. TOPOLOGICAL MAJORANA NANOWIRES

topological insulators and superconductors” [23]. An excerpt of this table is shown in


Table 2.1. For given spatial dimension and symmetry class, the table predicts whether
a topological material with bulk invariant either of Z or Z2 type is possible or not.
Every TI or TSC generically features gapless surface states that are robust against
perturbations which are not gap closing.

AZ Class TRS PHS 1D 2D 3D


BDI +1 +1 Z 0 0
D - +1 Z2 Z 0
DIII −1 +1 Z2 Z2 Z

Table 2.1: Excerpt of the “periodic table for topological insulators and superconduc-
tors” [21, 23] showing TSCs in up to three dimensions with PHS squaring to the identity.
The entry “0” for a given Altland-Zirnbauer (AZ) symmetry class and dimension im-
plies that every ground state is a member of the same phase, which is topologically
trivial.

Based on the discussion so far, one might wonder how superconductivity fits into a
scheme of noninteracting fermions. After all, superconductivity is rooted in attractive
phonon mediated quantum interactions of electrons near the Fermi surface. This can
be understood by recalling that within a standard mean field approximation a SC is
described as a noninteracting system of fermionic Bogoliubov quasiparticles which is
gapped due to the superconducting order parameter [5]. The BCS mean field Hamilto-
nian (up to a constant) is given by [24]
X Xh i
HBCS = ξk c†kσ ckσ + ∆k c†k↑ c†−k↓ + ∆∗k c−k↓ ck↑ , (2.1)
σ,k k

where ∆k is the superconducting order parameter and ckσ creates an electron with
2 k2
spin σ, momentum k and dispersion ξk = ~2m − µ. Introducing the spinor Ψk† =
(c†k↑ , c†k↓ , c−k↑ , c−k↓ ), we can write the Hamiltonian in the first quantized form up to a
constant as
1X †
HBCS = Ψ HBdG (k)Ψk , (2.2)
2 k k

which is well known as the Bogoliubov-de Gennes (BdG) Hamiltonian [24]. The anti-
unitary particle hole symmetry P = UP K, with UP a unitary operator and K the
complex conjugation operation, is a generic feature of superconducting BdG Hamilto-

16
CHAPTER 2. MAJORANA BOUND STATES IN QUANTUM DEVICES

nians [5],

UP HBdG (k)UP−1 = −HBdG (−k). (2.3)

In the absence of time-reversal symmetry and for P 2 = 1, a SC is placed in the Altland-


Zirnbauer symmetry class D. In spatial dimension d = 1 there exists a Z2 topological
invariant (see Table 2.1) also referred to as the “Majorana number” M = ±1 [6].
The principle of bulk-boundary correspondence states that nontrivial bulk topology,
M = −1, manifests itself holographically in protected states on the surface. The
“surface” of a finite wire is defined by the two end points, and the protected surface
states are the Majorana bound states γα which this thesis is focused on. Creation and
annihilation are achieved with the same operator γα† = γα , the Majorana operators obey
the Clifford algebra [20]
γα γα0 + γα0 γα = 2δαα0 . (2.4)

Furthermore, γα squares to the identity, i.e. γα2 = 1.

2.1.2 Kitaev’s p-wave superconducting lattice model

Kitaev’s paradigmatic chain model [6] is discussed in great detail in many excellent
reviews, e.g. Refs. [3, 4, 19, 20]. Hence, we discuss it here only briefly. The one-
dimensional lattice contains N sites and a spinless fermion is created on the site i by
the operator c†i . The Hamiltonian of the 1D SC reads

N
X −1 h    i
H= −µc†i ci − t c†i ci+1 + c†i+1 ci + ∆ c†i c†i+1 + ci+1 ci , (2.5)
i=1

and includes an onsite chemical potential µ as well as a nearest neighbor hopping


term with amplitude t. The pairing is of p-wave type coupling nearest neighbors with
amplitude ∆. Furthermore, we assume that a global phase rotation has been performed
to achieve ∆ ∈ R. How can we see that this model is indeed an example of a 1D
topological superconductor? To this end, it is instructive to consider the special point

17
2.1. TOPOLOGICAL MAJORANA NANOWIRES

t = ∆ and µ = 0 in parameter space, where the Hamiltonian takes on the form

N
X −1 h    i
H(t = ∆, µ = 0) = −∆ c†i ci+1 + c†i+1 ci + ∆ c†i c†i+1 + ci+1 ci
i=1
N
X −1   
= ∆ c†i + ci c†i+1 − ci+1 . (2.6)
i=1

We take note that two Majorana operators represent the Hermitian and anti-Hermitian
part of a conventional Dirac fermion ci since we may define them as γi = c†i + ci and
γ̃i = i(c†i − ci ). The Hamiltonian then takes on the form

N
X −1
H(t = ∆, µ = 0) = −i∆ γi γ̃i+1 . (2.7)
i=1

This means that two of the Majorana operators represent zero-modes, i.e. [H, γ̃1 ] =
[H, γN ] = 0, that are PHS protected and locked in at zero energy. Thus, the Majoranas
γ̃1 and γN are unpaired and can be arbitrarily far apart depending on the wire length.
We can define a highly nonlocal Dirac fermion created by

1
f † = (γ̃1 − iγN ) (2.8)
2

which can be occupied at vanishing energy cost. The ground state of topologically
trivial SCs is unique and accommodates all electrons in the form of Cooper pairs. The
presence of the fermion (2.8) changes this picture: consider that |0i with f |0i = 0 is a
ground state. Then, we have a two-fold degenerate ground state because |1i ≡ f † |0i
is another ground state with different fermion parity. The Majorana number for the
simple model can then be derived [6] to be

M = sign µ2 − 4t2 .

(2.9)

A nontrivial TSC configuration (µ, t, ∆) in parameter space is characterized by a finite


gap and M = −1. All nontrivial TSC configurations, of which the special point (t =
∆, µ = 0) is merely an example, can be smoothly deformed into each other without
closing the gap (as long as PHS is preserved). For general parameters in the topological
phase, a MBS γ1 localized at one wire end satisfies [H, γ1 ] ∼ exp (−L/ξ), with ξ the
parameter dependent size of the MBS and L the length of the wire [4]. It follows that

18
CHAPTER 2. MAJORANA BOUND STATES IN QUANTUM DEVICES

γ1 becomes an exact zero-mode in the limit L → ∞ of an infinite wire. The associated


topological phase is signified by weak pairing. In order to reach a topologically trivial
configuration by smooth deformation of the Hamiltonian, then requires necessarily to
close and reopen the gap and to undergo a topological phase transition.

2.1.3 Physical realization of 1D p-wave SC

The lesson learned in the previous subsection can be summarized as follows: in order to
realize the Kitaev chain model an effectively spinless SC with p-wave pairing is required.
Unfortunately, intrinsic p-wave SCs are exceedingly scarce in nature [4]. However,
Fu and Kane came up with the idea to synthetically engineer TSCs by putting TIs
and s-wave SCs in proximity [7]. This important contribution led to many further
proposals, two of which are particularly important in this thesis. One of them proposes
to proximitize TI nanowires to give rise to 1D TSCs [13, 14], which we will review and
apply to realize qubits in Chapter 3.
Another particularly promising approach involves spin-orbit coupled semiconductor
(SM) nanowires [8, 9], which we are going to discuss in the remainder of this sub-
section. In the presence of a strong magnetic field, such systems can be tuned to a
regime where only one longitudinal band is important and the system is effectively
spinless. In proximity to an s-wave superconductor, the spin polarized electrons are
endowed with a pairing term allowing them to be driven into the topological phase.
The experimental activity based on these proposals led to the probably most advanced
Majorana platform [15]. The noteworthiness of this approach stems for the fact that
it provides a way to accomplish a Majorana wire in realistic systems with the help of
well studied components such as semiconductors and s-wave SCs. The Hamiltonian
describing such systems is given by [4, 15]

H = HSM + HSC , (2.10)


ˆ L
~2 2
X  
HSM = dxψσ† (x) − ∂ − µ + i~αR σ2 ∂x + Γ σ1 ψσ0 (x), (2.11)
σ,σ 0 0
2m x σσ 0
ˆ L
HSC = dx (∆ψ↑ (x)ψ↓ (x) + H.c.) . (2.12)
0

Here Γ = gµB B is the external magnetic field with g the effective Landé g-factor, B
the applied magnetic field and µB the Bohr magneton. Furthermore, µ is the chemical

19
2.1. TOPOLOGICAL MAJORANA NANOWIRES

potential, m the effective mass and αR the Rashba spin-orbit coupling constant. The
field operator ψσ† (x) creates an electron at position x with spin σ. We assume that the
magnetic field and the spin-orbit field are oriented orthogonally to each other. In that
case, the energy eigenvalues for momentum k and ∆ = 0 are given by [4]

~2 k 2
q
2 2
Ek = −µ± Γ2 + αR k (2.13)
2m

as shown in Fig. 2.1. If the chemical potential µ is located within the gap opened by
Γ, the fermions are effectively spin-polarized. Proximity coupling of a s-wave SC to the
SM endows the effectively spinless fermions with Cooper pairing. For Zeemann field
Γ > Γc with q
Γc = |∆|2 + µ2 , (2.14)

the Refs. [8, 9] have ingeniously predicted that the system is in a topological supercon-
ducting phase. The experimental implementation of this approach will be discussed in
more detail in Section 2.4.

μ
Ε so

Figure 2.1: Schematic form of the energy spectrum as a function of momentum k


of a SM wire in the presence (black) and absence (red) of a TRS-breaking Zeemann
field, respectively. The spin-orbit-coupling is characterized by the energy scale Eso =
m 2
α [4, 15]. For suitable chemical potential values µ, the system displays effective
2 R
spinlessness.

20
CHAPTER 2. MAJORANA BOUND STATES IN QUANTUM DEVICES

2.2 Topological Majorana qubits

The Majorana qubit in the most elementary incarnation can be conceptualized as a


topological superconductor of mesoscopic size hosting four MBSs γ0 , . . . , γ3 [25, 26].
Two MBSs are not sufficient for a useful qubit since the quantum information would be
stored in the parity of a single fermionic mode and could be read out by the Coulomb
interaction, which is long ranged. Two nonlocal fermionic modes on the other hand
enable to perform parity-protected quantum computation.
Now, we would like to think about the mesoscopic SC island or “box” as being not
grounded a situation that we will also refer to as “floating”. The N island electrons
affect each other by means of the Coulomb interaction giving rise to an energy HC =
EC N 2 . The effective interaction strength is set by the charging energy EC = 2e2 /C
with C the linear capacitance of the device regions in good electrical contact with each
other. A capacitively coupled gate can regulate the electrostatic potential locally on
the island. This modifies the ground state energy according to HC = EC N 2 +eN V [27].
Up to a constant and with N̂ the electron number operator we obtain the Hamiltonian

HC = EC (N̂ − ng )2 , (2.15)

with the dimensionless back-gate parameter ng = eV /2EC controlling the energetically


ideal island charge. This charging Hamiltonain has been used in the context of topolog-
ical systems e.g. in Refs. [25, 26, 28, 29]. We point out that due to the MBSs an odd
number of electrons can also be accommodated. Under Coulomb valley conditions, or
more generally far away from the charge degeneracy point ng = 1/2, the fermion parity
is fixed [25],
γ0 γ1 γ2 γ3 = ±1. (2.16)

The charging physics provides a mechanism alleviating the harm caused by quasiparticle
poisoning events from outside the island in the regime of low energy [29, 30], because it
assigns an energy cost to these detrimental processes. The QP poisoning time has been
studied in the SM devices and was found to be of order & 1µs [31]. According to Eq.
(2.16), the degeneracy of the ground state is broken down from four-fold to two-fold,
which we identify with our Hilbert space. The topologically protected Majorana qubit

21
2.2. TOPOLOGICAL MAJORANA QUBITS

operating in this Hilbert space is characterized by a nonlocal Pauli algebra [25]

σ1 = iγ1 γ0 , σ2 = iγ2 γ0 , σ3 = iγ2 γ1 . (2.17)

These Pauli operators are “fractionalized” into Majorana bilinear operators.


A practical implementation scheme is based on two one-dimensional (1D) proximitized
nanowires which are driven in the topological phase by an applied magnetic field with
MBSs at their terminations [11, 12]. Parallely aligned, the wires are joined with the
help of a conventional s-wave superconducting bridge, see Fig. 2.2. Here, we assume
the charging energy of each wire to be small compared to the Josephson coupling so
that the composite object forms a single island with a single charging energy typically
of the order of 1meV [15].

γ2 γ0
SC
γ3 γ1
Figure 2.2: The “Majorana box qubit” or “tetron” [11, 12] is made from two one-
dimensional, proximitized nanowires (gray-green) hosting MBSs γα at their ends. The
linear dimension of the two wires is assumed to vastly exceed the superconducting
coherence length ξ so that the MBSs effectively represent zero-modes. A trivial s-wave
SC (gray) connects the wires to form a superconducting “Cooper pair box” that is
effectively characterized by a single charging energy EC .

The geometry displayed in Fig. 2.2 has been referred to in the literature as “tetron” [12]
or “Majorana box qubit” [11]. Closely related and also important in the context of this
thesis is the “hexon” [12] structure with six MBSs, which is analogously constructed
from three parallel 1D TSCs forming a single island.
In Chapter 3 of this thesis we describe another Majorana box qubit construction based
on topological insulator nanowires which are coated with a superconductor.

22
CHAPTER 2. MAJORANA BOUND STATES IN QUANTUM DEVICES

2.2.1 Majorana qubit interferometric readout

The step of readout and initialization of quantum information encoded in the Majo-
rana qubit is indispensable for its functionality. Extracting the occupation from a pair
of completely decoupled MZMs is not possible [6]. Therefore, one requires a tunable
coupling term that can be strongly suppressed to protect the information and turned
on during the readout phase enabling an observable to couple to it. There are multiple
different approaches to reading out Majorana qubits, e.g. involving the inductive cou-
pling to a flux qubit [30, 32]. Another approach requires to couple the Majorana qubit
capacitively to quantum dots [11, 12].
Now, we review another approach founded on Majorana interferometry, where the ob-
servable is the electrical conductance, see Ref. [11]. The approach relies on a striking
nonlocal transport phenomenon, which forms the basis for the weak measurement pro-
tocols that we will discuss in Chapter 4. The Majorana qubit is strongly Coulomb
blockaded, and tunable tunneling barriers connect two electrodes with two different
spatially separated MBSs. The strong charging effects hinder the electrons from tun-
neling onto the SC island. However, quantum charge fluctuations permit cotunneling
through the island [33]. At energy scales small compared to EC and the superconduct-
ing gap ∆, the box qubit device is equivalent to two fermionic levels with support at
the four ends of the two TSC wires. The nonlocal transport through the box from lead
α to lead α0 due to Coulomb charging is described by the Hamiltonain [29, 30]

H = t1 (iγα γα0 )c†α,k cα0 ,k0 + H.c., (2.18)

with the fermionic creation operators c†α,k for lead α and momentum k. The phase and
absolute value of the transmission amplitude t1 is independent of the tunneling distance
a fact that Fu has labeled as “teleportation” of electrons [29]. We will refer to the
transport as phase coherent and note that there is experimental evidence consistent
with such a long distance phenomenon [34]. The foundation of various interferometric
schemes is the observation that a flip of the parity iγαi γαj 0 causes a π phase shift in the
transmission amplitude. This mechanism allows to obtain a parity dependent conduc-
tance (current) by devising a second path for the electrons to tunnel through, see Fig.
2.3. This interference path serves as a reference and may be realized using a low density
SM or a second Majorana island with fixed parity [11]. The interferometric purpose
of the link dictates that its length should be short enough to enable phase coherent

23
2.2. TOPOLOGICAL MAJORANA QUBITS

V
γ2 γ0
Φ
γ3 γ1

Figure 2.3: Electron interferometric scheme for Majorana qubit [11, 12, 30] readout of
the occupancy of the nonlocal Dirac fermion (γ0 + iγ1 )/2 by measuring the tunneling
conductance G = dI/dV . The first path goes through the Coulomb blockaded Majorana
qubit, while the second path is a tunneling link, e.g. realized by a low-density semi-
conductor with a sufficient phase-coherence length (vertical dashed line). Furthermore,
there is an external magnetic flux Φ present which is enclosed by the two paths.

electron transport. When we further assume that there is a magnetic flux Φ piercing
in between the two paths, the conductance depends on the parity according to [11, 30]

G(Φ) = g0 + iγ0 γ1 g(Φ), (2.19)

with g(Φ) = g(Φ + h/e), i.e. exhibiting Aharonov-Bohm oscillations. We emphasize


that without the interference path, the conductance is not parity dependent. The
conductance measurement hence amounts to a projective readout of the parity iγ0 γ1 .

24
CHAPTER 2. MAJORANA BOUND STATES IN QUANTUM DEVICES

2.3 The non-Abelian nature of Majorana zero-


modes

We will now discuss the physical meaning of the non-Abelian properties of MZMs [10],
highlighting the connection to the nonlocal Pauli algebra of a Majorana qubit, which
will be of crucial importance in Chapter 4.

2.3.1 Non-Abelian braiding statistics

It was first shown by Ivanov in 2001 that MZMs bound to vortices in 2D p-wave
Superconductors behave as non-Abelian anyons when the vortices are exchanged [35].
It has been put forward that 1D TSCs can in principle be used in branched Y- or T-
geometries by manipulating the MBS carrying domain walls [36, 37]. To understand
the term anyon, we recall that in three dimensions, quantum particles are either bosons
or fermions. In two dimensions, a quantum particle can be neither a fermion nor a
boson but a third option called anyon [10]. Under the exchange of identical particles α
and β, the wavefunction can pick up a general phase factor, i.e. |. . . , ψα , . . . , ψβ , . . .i =
exp (iθ) |. . . , ψβ , . . . , ψα , . . .i. For real statistical angle θ ∈
/ {0, π}, such particles are
called Abelian anyons [10]. If the many body ground state has multiple degeneracies,
the exchange of identical particles can even cause a change of the quantum state, a
scenario which is known as non-Abelian statistics. This term is motivated by the fact
that the phase factor exp (iθ) is replaced by a unitary operator Tαβ in the space spanned
by the degenerate ground states [10],

|. . . , ψα , . . . , ψβ , . . .i = Tαβ |. . . , ψβ , . . . , ψα , . . .i . (2.20)

In general, unitary operators are noncommutative, implying that the final quantum
state can depend on the order in which the indistinguishable particles have been ex-
changed. MZMs bound to defects represent a special type of non-Abelian anyon known
as Ising anyon and we now discuss how braiding implements a unitary evolution in
the degenerate ground state manifold. Drawing on our discussion of Majorana based
qubits, we again consider four MZMs γ0 , . . . , γ3 and we may think of the associated
Hilbert space as our computational space. For a start, we determine the form of the
unitary operator corresponding to the adiabatic exchange of two of the Majoranas in

25
2.3. THE NON-ABELIAN NATURE OF MAJORANA ZERO-MODES

counter-clockwise direction following the heuristic discussion given in Ref. [38]. De-
† †
noting this transformation γ0 → γ̃0 = T01 γ0 T01 and γ1 → γ̃1 = T01 γ1 T01 , we conclude
γ̃1 = λ0 γ0 and γ̃0 = λ1 γ1 from the fact that the Majoranas are simply swapped. The new
operators γ̃α are evidently required to obey γ̃α† = γ̃α and γ̃α2 = 1, which implies λα = ±1.
Since the adiabatic exchange cannot result in a change of the parity P = iγ0 γ1 , we know
that iγ0 γ1 = iγ̃0 γ̃1 . This implies two possible choices for the sign:

γ0 → ±γ1 , (2.21)

γ1 → ∓γ0 . (2.22)

This is a gauge freedom and we choose to work with the lower sign choice in the
following. It is straightforward to check that the unitary transformation of counter-
clockwise exchange of the defects α and β is given by [36]

π iπ

 e4
Tαβ = e 4 exp γα γβ = √ (1 + γα γβ ) . (2.23)
4 2
The second equation can be shown using the properties of the Majorana operators.
The collection of all braiding operations forms a group the so called braid group [10].
One can identify the Pauli operators σ1 = iγ1 γ0 , σ2 = iγ2 γ0 and σ3 = iγ2 γ1 (see Ref.
[25] and Subsection 2.2) as the generators of the braid group, because according to Eq.
(2.23) we may write


 π  iπ
 π  iπ
 π 
T10 = e 4 exp − σ1 , T20 = e 4 exp −i σ2 , T21 = e 4 exp −i σ3 . (2.24)
4 4 4

This reveals a deep connection between the nonlocal Pauli operators and the braiding
statistics. In Fig. 2.4, we see an example of the graphical representation of the exchange
operation T12 . Using the Clifford algebra relations, it is straightforward to prove that
p
Tαβ = iγα γβ [36]. Thus, by exchanging the Ising anyons twice, which is equivalent to
one Ising anyon encircling the other, we obtain the Pauli operators:

2 2 2
T10 = σ1 , T20 = σ2 , T21 = σ3 . (2.25)

An exchange of e.g. 1 and 2 transforms a state Ψ, defined in the ground state manifold,
according to Ψ → exp i π4 σ2 Ψ, while the encircling operation implements Ψ → σ2 Ψ


[18]. As an example, we may consider the state Ψ = |001 , 123 i with nij the occupancy

26
CHAPTER 2. MAJORANA BOUND STATES IN QUANTUM DEVICES

γ0 γ1 γ2 γ3
Figure 2.4: Geometrical representation of the elementary braiding operation of adia-
batic and counter-clockwise exchange of the defects which bind the MZMs γ1 and γ2 .
The horizontal and vertical axes denote the spatial and temporal direction respectively.
Braidings are equivalent if they can be smoothly deformed into each other without
moving the start and endpoints (black dots) or crossing the strands [38].

of the fermion number operator n̂ij ≡ 12 (1 + iγi γj ). The exchange of MBSs belonging to
different fermions creates a maximally entangled Bell state in the given basis [3], e.g.

e4
T12 |001 , 123 i = √ [|001 , 123 i + |101 , 023 i] , (2.26)
2
2
whereas the qubit state (with fixed odd fermion parity) is flipped: T12 |001 , 123 i =
|101 , 023 i. The exchange of Majoranas constituting the same fermion on the other hand
results in a parity dependent phase factor being multiplied to the state.

2.3.2 Non-Abelian fusion rules

Braiding implies a time evolution in the degenerate ground state space due to the mo-
tion of indistinguishable anyons. To understand Majorana fusion rules, consider two
Majorana zero-modes γ1 and γ2 , bound e.g. to vortices of 2D TSCs or the termina-
tions of 1D TSCs wires, which due to large separation have exponentially suppressed
overlap. When the spatial distance between the associated MZMs carrying defects gets
reduced, a hybridization term gradually appears. The degenerate ground state splits
and the fusion may or may not give rise to an unpaired fermion. The case of a fermion

27
2.3. THE NON-ABELIAN NATURE OF MAJORANA ZERO-MODES

corresponds to an excited state commonly denoted ψ, while the case of no fermion


corresponds to the ground stated denoted 1. Hence, one writes [10, 36]

γ1 × γ2 → 1 + ψ. (2.27)

One can view this as the projective measurement of the fermionic particle state that
forms when the two MBSs are brought together, which can be empty or occupied at the
cost of a finite energy. The two options illustrated in Eq. (2.27) are also known as fusion
channels. Detection of the fusion rule (2.27), amounts to establishing the non-Abelian
nature of MZMs, because the fusion of Abelian particles always exhibits just a single
fusion channel [10]. In Chapter 3, we briefly mention a fusion rule detection scheme
based on charge sensing [39] applied to Majorana qubit architectures in proximitized
topological insulator nanowires.

2.3.3 Topological quantum computation

The idea to build topological quantum computers based on non-Abelian anyons gen-
erated great excitement in the quantum information community [10, 40], partially ex-
plaining the interest in MBSs. 2N + 2 mutually far separated MZMs form a degenerate
ground state manifold of dimension 2N corresponding to N qubits if we restrict our-
selves to a well defined charge parity. However, drawing on the Majorana box qubit
discussion in subsection 2.2, we rather would like to think of 4N MZMs encoding N
qubits with N constraints on the parities γ1 γ2 γ3 γ4 , γ5 γ6 γ7 γ8 ,... which is known as
“sparse” encoding [41].
A quantum computation is defined as a sequence of unitary operations called quantum
gates. A set of gates is esteemed to be “universal” if an arbitrary computation can
be performed. The unitary operations performed by braiding of Ising anyons, like e.g.
π
T01 = ei 4 (1+σ1 ) (see Eq. (2.24)), do not meet this requirement. In fact they are elements
of the Clifford group. Remarkably only a single gate, e.g. the so called T-gate
!
1 0
T= iπ (2.28)
0 e4

is missing to make it universal [41]. The latter would however be unprotected unlike
the braiding unitary operations which are completely specified by the topology of the

28
CHAPTER 2. MAJORANA BOUND STATES IN QUANTUM DEVICES

braid. Thus, by employing MBSs an element of the Clifford group can in principle
be applied exactly to all decimal places while the T-gate can only be approximated.
The finite excitation pairing gap ∆ causes further limitations [42]. The timescale on
which the braiding is performed is important. It has to be slower than ~/∆0 because
otherwise quasiparticles could be excited. On the other hand, the coherence time of
the Majorana qubit clearly represents an upper limit on operation times.

29
2.4. EXPERIMENTAL IDENTIFICATION OF MAJORANA BOUND STATES

2.4 Experimental identification of Majorana bound


states

MBSs are neutral and spinless objects that do not couple to electric and magnetic
fields. Due to electron non-conservation of the U (1) → Z2 broken superconducting
state, electrons provide a way to couple to the MBSs by using tunneling spectroscopy
[40]. Probing the density of states of one end of a topological wire locally by means
of conductance spectroscopy is therefore the most straightforward approach to obtain
MBS signatures. To understand this point, first recall what happens for a trivial SC: the
application of a voltage between a normal contact and the superconducting part leads to
Andreev processes [27]. In such processes, the electron incoming from the normal metal
is reflected as a hole from the tunnel barrier, while the current in the superconductor
is carried by Cooper pairs. For a topological SC wire, in contrast, there is a resonant
Andreev process at zero bias voltage when the energy of incoming electrons matches
that of the Majorana zero-mode at the interface. In the zero temperature limit, the
MBS gives rise to perfect Andreev reflection resulting in a robust zero-bias conductance
peak (ZBCP) with a quantized maximum height of G = 2e2 /h [43]. By investigating SM
nanowires with proximity induced superconductivity, a 2012 Delft experiment provided
the first experimental ZBCP evidence consistent with Majorana bound states [44]. In
these experiments, a NbTiN superconductor was used to proximitize an InSb nanowire
and the device as a whole was grounded. A recent success was that the height of the
ZBCPs was confirmed to be 2e2 /h quantized [45, 46].
Naively, one might expect that a genuine topological superconducting wire reveals itself
through the robustness and persistence of the quantized ZBCP signature for sufficiently
strong magnetic field. In a similar spirit, the early experiments where partially cele-
brated as MBS verification implying that proximitized nanowires have been driven into
the topologically superconducting regime. However, in this context, the results of fur-
ther research implied that different alternative scenarios could also explain the same
signatures. In light of these results that we will summarize now, the ZBCP is a neces-
sary but not yet conclusive observation of MZMs.
First, disorder has been shown to be a possible origin of quantized zero-bias conductance
peaks with unit spectral weight and robust to magnetic fields [47, 48]. This highlights
the need to achieve fabrication of cleaner materials and devices and concerted endeavors
already led to substantial progress in this regard [15].

30
CHAPTER 2. MAJORANA BOUND STATES IN QUANTUM DEVICES

Second, topologically trivial Andreev bound states (ABS) have been shown to be able
to cause the same phenomenology as genuine (single) MBSs when they are detected on
the basis of zero-bias conductance peaks [49, 50]. Andreev bound states are fermionic
in nature implying that their creation operator c = (γ1 + iγ2 )/2 can be split into a pair
of Majorana operators γ1† = γ1 , γ2† = γ2 . The corresponding MBS component wave
functions may actually have only a small overlap due to spatial displacement [50]. In
the same reference it is shown that these so-called “partially separated Andreev bound
states” (denoted ps-ABS in the same Ref.) are generic for low-energy spin-orbit cou-
pled SM-SC heterostructures in the presence of a suitably oriented magnetic field [50].
These generic midgap ABSs in nanowires originate in inhomogeneities of the chemical
potential or individual impurities behaving as quantum dots. Also the combination of
spin-orbit coupling and Zeeman splitting permits Andreev bound states to reside close
to the band center for finite values of the Zeemann field [51]. These ps-MBSs mimic
topological unpaired MBSs so accurately that it has been pointed out that they may
be useful for quantum computation because their non-Abelian braiding properties may
be accessible [52]. We also point out that for the experimentalist it is not possible to
know the value of the critical magnetic field independently, making it difficult to decide
whether a measured ZBCP is seen on the topological or trivial side of the topological
phase transition [51].
In summary, it is fair to say that the advances so far consist to a large extent of lo-
cal probes and are consistent with the predictions for MBSs made by theory. But
a false-positive interpretation is not yet ruled out. Nevertheless, the existing experi-
mental observations represent crucial steps bringing us closer to topological quantum
computation. Non-observation of a zero-bias peak in experiments would have falsified
the Majorana wire hypothesis already some time ago. We also mention that there
are many further experiments that have been performed which we have not mentioned
here. E.g. another noteworthy breakthrough was achieved on Coulomb blockaded is-
lands in 2016 at the University of Copenhagen [53]: the first systematic measurement of
the finite-size exponential suppression of the ground-state degeneracy associated with
overlapping Majorana bound states.

31
2.4. EXPERIMENTAL IDENTIFICATION OF MAJORANA BOUND STATES

2.4.1 Next generation of experiments beyond local probes

As briefly stated in the introduction, the main aspects that must be investigated in
order to demonstrate Majorana physics are broadly speaking the following: (a) verifi-
cation of non-Abelian braiding properties, (b) corroboration that two MBS constitute
a vastly nonlocal Dirac fermion, and (c) demonstration of unpaired Majorana bound
states localized in TSCs. As discussed above, there is mostly a consensus in the con-
densed matter community that signatures of local probes such as ZBCP measurements
(belonging to category (c)) are not able to unambiguously identify genuine MBSs. It
thus seems clear that a next generation of experiments is needed to clarify the situation.
There are various theoretical proposals available on how to achieve this goal [54]:

ˆ For SC islands with more than two MBSs, a non-perturbative topological Kondo
effect as discussed in Refs. [25, 26, 28, 55] is present. This could provide a
smoking gun signature, also in light of related low-energy transport proposals for
two-box setups [56]. The characteristic nonlocal transport features are expected
at low temperatures T < TK and are predicted to disappear when a far-away lead
is decoupled. However, reaching this regime is very challenging in experiments
since the Kondo temperature TK is predicted to be extremely low.

ˆ In trijunctions formed from three TSCs, a single MZM is expected at the junction.
Ref. [57] has shown that this state causes giant shot noise peaks when choosing
commensurate voltages, like for instance V1 = −V /2, V2 = 0, V3 = +V /2, on
the TSC terminals. This effect disappears in the absence of such a protected
zero-energy state.

ˆ Another proposal made in Ref. [58] is to use interferometry by embedding a


Coulomb-blockaded Majorana wire into a ring and then probe Aharanov-Bohm
h/e-periodic conductance oscillations with the magnetic flux piercing the ring,
see also Ref. [59]. For Andreev states, interference is argued to be suppressed,
while for genuine MBSs one expects h/e oscillations, see also Refs. [29]. A related
proposal is to study the Josephson current phase relation across a Majorana box
with two MBSs [60]. For ABSs, the authors of this reference predict a phase shift
when the gate parameter is shifted according to ng → ng + 1, which in fact is
absent for authentic MBSs.

32
CHAPTER 2. MAJORANA BOUND STATES IN QUANTUM DEVICES

ˆ The detailed time dependence of the transient current after switching on the
tunnel couplings in the standard ZBCP setting has been proposed as a signature
allowing to distinguish genuine MBSs and ABSs in Ref. [61]. The fact that short-
time transients can be affected by various other effects could make implementation
of this proposal challenging in practice.

ˆ Measuring the zero bias peaks at both ends of a TSC wire by studying the end-
to-end correlations could provide a signature for authentic MBSs because one
expects the ZBCPs to appear at both ends simultaneously [51]. In practice,
this suggestion might be difficult to implement because it requires different gate
settings at both ends of the wire. Furthermore, it is challenging to probe the same
fermionic zero-mode.

ˆ Thermoelectric noise due to the interplay of electric fields and temperature gra-
dients has been proposed in Ref. [62] as a means to distinguish genuine MBSs
and ABSs.

ˆ Microwave spectroscopy of a Majorana wire could in principle also prove useful


for distinguishing genuine MBSs and ABSs as discussed in Ref. [63]. The system
requires spectroscopic measurement in the microwave regime and is thus more
complicated to implement than a DC setup.

In light of the challenging nature of most of these experiments, the need to move on to
a next generation of experiments on Majorana devices that go beyond local probes and
are as simple as possible is quite substantial. This serves as an important motivation
for this thesis. We believe that the simplicity can be found in transport spectroscopy
concretely, by measurements of currents and shot noise as will be discussed in Chapters
4 and 5.

33
2.4. EXPERIMENTAL IDENTIFICATION OF MAJORANA BOUND STATES

34
Chapter 3

Majorana qubits in proximitized TI


nanoribbon device architectures

This chapter contains a detailed study of the realization of Majorana based qubits built
in platforms of TI nanoribbons in proximity to s-wave SCs [13, 14]. The chapter is
structured as follows: Section 3.1 reviews TI nanoribbons and the emergence of MBSs
in proximity to an s-wave SC. In Sections 3.2 and 3.3, we model and analyze the
novel Majorana box qubit device setups which allow for gate tunable hybridization of
MBSs and proof-of-principle experiments to test the devices. Finally, in Section 3.4, we
give an outlook on long-term applications of the proposed qubit devices. Our research
presented in this chapter has previously been published in Ref. [64].

3.1 TI nanoribbons and the emergence of MBSs in


proximity to an s-wave SC

3.1.1 TI nanoribbons and surface Dirac theory

Let us start by introducing TI nanowires with a focus on their transport properties and
reviewing the well established description in terms of Dirac surface models, which we
will work with subsequently. In this subsection, we will see that topological insulator
nanoribbons exhibit a finite-size gap. This gap can be closed to generate a single gapless
helical 1D mode in the presence of an axial magnetic flux Φ = Φ0 /2, where Φ0 is the

35
3.1. TI NANORIBBONS AND THE EMERGENCE OF MBSS

magnetic flux quantum [65, 66, 67, 68, 69, 70]. A reader familiar with this subject may
move on the next subsection.
The main ingredient are three-dimensional topological insulators (TIs) [16, 17], a type
of matter which, as discussed in Chapter 2, exhibits the celebrated bulk-boundary
correspondence. The nontrivial topology of the bulk matter goes hand in hand with
a robust gapless Dirac-type surface state circumventing the fermion doubling theorem
[71]. In this chapter, we consider nanowires made out of such topological insulator
materials. The term ‘wire’ indicates effective one-dimensionality, i.e. the typical (nano)
length scale of the width is small compared to the length of such a wire. Because of their
layered structure, the nanowires of 3D TI materials commonly grow in a tape-like shape
with rectangular cross sections of order 40 × 100nm2 [72, 73], explaining why the wires
are referred to as nanoribbons. The Dirac fermion surface state governs the transport of
three-dimensional topological insulator materials [16, 17] for chemical potential values
within the bulk band gap Eg . The latter sets a length scale on which the surface Dirac
fermion decays into the bulk, which is typically of the order of several nanometers.
Quantum transport of bulk insulating TI nanowires with a width exceeding this length
scale is therefore approximately accounted for by relying on surface theories (see e.g.
Refs. [65, 66, 67, 68, 69, 70]). As explained in Refs. [65, 74], the surface theory can be
obtained e.g. by starting from a massive Dirac Hamiltonian in three dimensions
!
−Eg σ · p
HBulk = , (3.1)
σ · p Eg

with Eg being the bulk band gap, p = −i~∇ the momentum operator and σ ≡
(σ1 , σ2 , σ3 ) the vector of Pauli spin matrices. This Hamiltonian is defined in the space
of the spin and pseudospin (sublattice) degrees of freedom and neglects other bands of
the 3D TI energy spectrum. Let us assume that the interface of TI and the vacuum is
defined by a position dependent normal vector n̂. Formally, one can derive the surface
Hamiltonian by sending the mass parameter Eg to infinity away from the surface. In
this way, one obtains the Hamiltonian [65]

~υ υ
HSurface = ∇ · n̂ + (n̂ · [p × σ] + [p × σ] · n̂) , (3.2)
2 2

with υ the Dirac velocity. This Hamiltonian describes a single species of a two-
dimensional massless Dirac fermion. In the topology dominated transport of TI

36
CHAPTER 3. MAJORANA QUBITS IN PROXIMITIZED TI NANORIBBONS

nanowires, accounting for the precise details of the wire geometry is often inessen-
tial in predicting transport properties. Indeed, the obtained low-energy band structure
is very similar when the idealization of a cylindrical nanowire is invoked to describe a
nanoribbon with the same cross section [14, 69, 70]. Assuming a cylindrical nanowire
with radius R and in alignment with the z-axis, the Eq. (3.2) yields the TI wire
Hamiltonian
~υ2
H0 = −i~υ1 σy ∂z + σz i∂θ , (3.3)
R
with the coordinate θ parametrizing the wire circumference [14, 75, 76]. The parameters
υ1 and υ2 are the Fermi velocities in the longitudinal and circumferential direction. In
deriving this Hamiltonian, a unitary transformation (spin rotation) has been performed,
which implies antiperiodic boundary conditions for the spinor wave function, i.e.

ψ(z, 0) = −ψ(z, 2π). (3.4)

An immediate consequence of the antiperiodic boundary condition (3.4) is a gap in


the spectrum, physically originating in the necessary finite transverse momentum. The
energy scale ∼ ~υ2 /R characterizes the transversal confinement.1
Now, we imagine that a homogeneous axial magnetic field B is being turned on. Ac-
cording to the standard procedure of minimal substitution, the dimensionless magnetic
flux
Φ
ϕ≡ (3.5)
Φ0
longitudinally threading the wire appears as an azimuthal vector potential in Eq. (3.3),
yielding
~υ2
H0 = −i~υ1 σy ∂z − σz (−i∂θ + ϕ). (3.6)
R
It is instructive to understand the qualitative features of the corresponding spectrum
for a translationally invariant wire. We may perform a Fourier transformation of the
spinor wave function, i.e.
ˆ X
ψ(z, θ) = dz eikz e−ijθ ψk,j , (3.7)
Z+ 12
j∈( )
1
In Ref. [76], a TI nanowire with arbitrary smooth, constant cross section perpendicular to the
z-axis was considered. Hence, the surface of such a wire has no intrinsic Riemannian curvature. The
corresponding Hamiltonian can be transformed to coordinates (s, z) in which it takes on the flat space
form [76], H = −iυ~ [σz ∂z + σy ∂s ]. Consequently, the use of the Hamiltonian (3.3) is not as restrictive
as it may seem.

37
3.1. TI NANORIBBONS AND THE EMERGENCE OF MBSS

where we introduced the axial momentum k and the quantized angular momentum
j [76]. The fact that the summation is performed over half-integers accounts for the
antiperiodic boundary conditions. The Hamiltonian separates into blocks of the form
Hk,j = ~υ1 kσy + ~υR2 (j + ϕ)σz . The spectrum is obtained by squaring Hk,j , which results
in [14] r
υ2
Ek,j = ±~ υ12 k 2 + 22 (j + ϕ)2 . (3.8)
R
This spectrum has been plotted in Fig. 3.1.1, where the qualitative differences for
different values of the dimensionless flux are being displayed. One of the key takeaways
from the spectrum is that for ϕ = 1/2 there is a non-degenerate gapless linear band.
This is precisely the condition for the underlying normal state of a TSC [6, 7]. The
spectrum makes it intuitive to anticipate that the system in proximitiy to an s-wave
SC will give rise to MBSs.

E E E E

k k k k
φ=1/2 φ=0.35 φ=0.15 φ=0

Figure 3.1: Energy spectrum (schematic) of a TI nanoribbon as a function of the


longitudinal momentum k for different values of the threading dimensionless flux ϕ.
For ϕ = 1/2 (left), the system is effectively time reversal symmetric and there is a non-
degenerate linear gapless band. All the other bands have twofold degeneracy. Hence, for
any value of the chemical potential there is an odd number of Fermi points for k > 0. In
the absence of magnetic flux ϕ = 0 (right), there is a gap in the spectrum as discussed
in the main text and every band has a twofold degeneracy. In the intermediate cases
ϕ = 0.35 (center left) and ϕ = 0.15 (center right), the system is gapped and every band
is nondegenerate. (The plot has been created with Mathematica.)

3.1.2 Dirac fermion model of proximitized TI nanoribbons

In the previous Subsection 3.1.1, we have seen that TI nanowires with half-integer axial
flux quantum ϕ = 1/2, e.g. made of the strong TI Bi2 Se3 , exhibit an odd number of

38
CHAPTER 3. MAJORANA QUBITS IN PROXIMITIZED TI NANORIBBONS

Fermi points for k > 0 for any value of the chemical potential within the bulk band
gap. According to Kitaev [6], in proximity to an s-wave SC, MBSs should emerge in
such a situation if the superconductivity gaps the spectrum. For the proximitized TI
nanowire system, topological superconductivity was first predicted by Cook and Franz
in Refs. [13, 14] and subsequently analyzed e.g. in Refs. [77, 78, 79, 80]. We now
discuss the proximitized system again using a surface Dirac fermion model, which is
wrapped around the nanowire. For a cylindrical wire, due to its rotational symmetry
with respect to the z-axis, the total angular momentum is conserved with half-integer
eigenvalue j. According to the discussion in the previous subsection (see Eq. (3.6)),
the effective Dirac surface Hamiltonian of the j-branch is given by

~υ2
H0 = −i~υ1 σy ∂z − (j + ϕ(z)) σz − µ(z)σ0 , (3.9)
R(z)

where we now explicitly included a chemical potential µ. For given angular momentum
quantum number j ∈ Z + 1/2, the surface state in spin space is generally of the form
[14] !
eijθ e−iθ/2 fj (z)
ψj (z, θ) = √ , (3.10)
2π eiθ/2 gj (z)
where θ is again the azimuthal coordinate parametrizing the circumference of the cylin-
der. The functions fj and gj are subject to the normalization condition
ˆ
dz(|fj (z)|2 + |gj (z)|2 ) = 1. (3.11)

Expressed in this way, the Hamiltonian effectively acts on spinor states (fj (z), gj (z))T
in a reduced 1D description. We now add the superconducting term needed to account
for the proximity effect with induced pairing ∆, which yields the description of the
system via the Bogoliubov-de Gennes (BdG) Hamiltonian in Nambu space [14]
!
H0 (z) iσ2 ∆(z)
HBdG = ∗
. (3.12)
−iσ2 ∆ (z) −H0 (z)

Surface states inside the gap of the bulk TI are then found as eigenstates of the BdG
Hamiltonian. In Subsection 3.1.3 and Section 3.2, we will apply the Hamiltonian (3.12)
to describe proximitized nanowire systems where ∆, ϕ and µ are functions of the coor-
dinate z defined along the wire axis.

39
3.1. TI NANORIBBONS AND THE EMERGENCE OF MBSS

For an infinite, translationally invariant system with a uniform flux ϕ, the BdG solu-
tions are plane waves with axial momentum k. Under these assumptions, the dispersion
relation is given by [14]
q
Ek,j,σ,σ0 = σ (~υ1 k)2 + (Mj + σ 0 ∆)2 , (3.13)

where σ, σ 0 ∈ {±1}. The parameter characterizing the size quantization gap is defined
as
~υ2
Mj = |ϕ + j| . (3.14)
R
The gap associated with the mode j = −1/2 is denoted

~υ2
M− 1 ≡ M (ϕ) ≡ |ϕ − 1/2| (3.15)
2 R

and is zero for half a flux quantum, i.e. for ϕ = 1/2. We consider TI nanoribbon
cross sections that are sufficiently small such that ∼ ~υ2 /R is a large energy scale
and modes with angular momentum j 6= −1/2 can be neglected within a low-energy
approximation scheme. We discuss possible causes and effects of flux mismatch away
from ϕ = 1/2 in our discussion in Subsection 3.3.3. Note that a gapless branch exists
for Mj = ±∆, signaling a topological phase transition [4]. We will see in more de-
tail in the next subsection how an interface of both types of gaps gives rise to MBSs
localized at the ensuing domain walls. These MBSs emerge under the circumstances
of Altland-Zirnbauer symmetry class D with broken spin SU (2) symmetry and broken
time reversal symmetry (TRS). This is despite the fact that the infinite proximitized TI
nanoribbon with flux ϕ = 1/2 gives rise to an effective time reversal invariance placing
it in symmetry class DIII, because the non-quantized magnetic flux at the wire ends
breaks TRS [14].

40
CHAPTER 3. MAJORANA QUBITS IN PROXIMITIZED TI NANORIBBONS

3.1.3 MBS at the interface of TI wire segments of different


width

A novel idea that will play a prominent role in our discussion of qubit architectures is
to consider an interface separating TI nanowire segments of different radius such that
the thicker right half with flux Φ = Φ0 /2 is proximitized and the thinner left half is not
(see Fig. 3.2). We derive now the Jackiw-Rebbi zero-mode solution corresponding to
a localized MBS at the interface. We treat the problem analogously to Refs. [13, 14],
where a MBS arising at the interface with a magnetically gapped domain has been
considered. In this subsection, we follow the presentation in these references using the
effective BdG Hamiltonian (3.12) discussed in the previous subsection. The respective
energy gaps M and ∆ on the two sides are rooted in the size quantization2 and the
superconducting pairing. A radius smoothly varying in the interface region leads to a

Figure 3.2: TI nanowire (gray) divided into two halves of different radius such that
only the thicker right half is in proximity to an s-wave SC (gray-green). An axially
aligned magnetic field B is adjusted in strength to generate a flux Φ = Φ0 /2 in the
proximitized half (right). Consequently, the flux in the left half is Φ < Φ0 /2, implying
the occurrence of a size quantization gap. This gap as well as the pairing gap may
be approximated by real, smooth functions M (ϕ(z)) and ∆(z) describing the effective
gaps of different origin in the effective BdG Hamiltonian given in Eq. (3.12). This leads
to a Jackiw-Rossi Majorana zero-mode (red) localized at the interface z = z0 , where
the two kinds of gaps coincide, i.e. ∆(z0 ) = M (ϕ(z0 )).
2
As discussed in Subsection 3.1.2, we assume that modes with angular momentum j 6= −1/2 can
be neglected at low energies.

41
3.1. TI NANORIBBONS AND THE EMERGENCE OF MBSS

smooth flux profile3 ϕ(z) and we assume ∆(z0 ) = M (ϕ(z0 )) at the interface point
z = z0 , which is the condition for a gapless branch in the dispersion relation of Eq.
(3.13). Furthermore, we assume the flux to vary only in the vicinity of the interface
z = z0 such that ϕ(z  z0 ) ' ϕ0 < 1/2 and ϕ(z  z0 ) ' 1/2. Moreover, the gap
function
~υ2
M (ϕ(z)) = |ϕ(z) − 1/2| (3.16)
R
and the (assumed to be real) pairing gap ∆(z) vary only around z = z0 as well. On
the non-proximitized half we assume ∆(z  z0 ) ' 0 and M (ϕ(z  z0 )) ' M0 > 0.
On the other hand, for the proximitized half it holds that ∆(z  z0 ) ' ∆0 > 0 and
M (ϕ(z  z0 )) ' 0.
In accordance with general principles, under such conditions a localized Jackiw-Rossi
zero-mode comes into being in between the two domains. In the present case it corre-
sponds to a localized Majorana zero-mode and can be approximated as the zero energy
solution of the Hamiltonian (3.13) with the smooth, real gap functions discussed above.
In the language of second quantization this solution can be expressed as [13, 14]
ˆ
1
dz f (z) − g(z) + f † (z) − g † (z) u(z),

Ψ̂0 (z) = (3.17)
2

which is localized at z = z0 with

ˆz
 
dy
u(z) = u0 exp  [M (ϕ(y)) − ∆(y)] . (3.18)
νF
z0

Hence, 1/M0 and 1/∆0 define length scales of exponential decay of u(z) in the left and
right region respectively. The operators f and g correspond to the functions defined in
Eq. (3.10). Due to the reality property u(z) = u∗ (z), the field operator is Hermitian

Ψ̂0 = Ψ̂†0 . (3.19)

Thus, Ψ̂0 represents a Majorana zero-mode, because creation and annihilation are rep-
resented by the same operator. The topological nature of this Majorana zero-mode is
reflected in the fact that the precise form of M (ϕ(z)) − ∆(z) does not matter, as long
as the sign of this expression changes at z = z0 . Since the total number of unpaired
MBSs in an electronic system must be even, there must exist another unpaired MBS in
3
We neglect flux channelling effects due to the smallness of the magnetic susceptibility.

42
CHAPTER 3. MAJORANA QUBITS IN PROXIMITIZED TI NANORIBBONS

a finite system. It can neither be on the non-proximitized half nor in the gapped bulk
of the proximitized half. Hence, it must exist on the other end independently of the
details of the boundary condition [14].

43
3.1. GATE TUNABLE COUPLING OF MAJORANA STATES

3.2 Gate tunable coupling of Majorana states


across a constriction

In this section, we quantitatively analyze the following idea: A weak link between two
proximitized TI nanoribbons is engineered by means of a narrower, non-proximitized
nanowire segment (see Fig. 3.3). For suitably adjusted flux, MBSs form at both sides
of this central constriction, which is characterized by a local gap due to the lowered
local flux. As will be quantitatively analyzed in detail below, this allows creation
and annihilation of MBSs localized across the constriction, because two TSCs can be
electrically disconnected by means of gating. In contrast to the discussion in Subsection
3.3.3, where the narrow wire segment was infinite, it now has finite length W .

Gate

γ1 γ2 γ3 γ4

B
Φ < Φ0 / 2 Φ = Φ0 / 2

Figure 3.3: A narrow non-proximitized TI nanoribbon segment of length W is fabricated


in between two thicker outer sections of the TI nanoribbon. These outer sections are
proximitized by an s-wave SC (gray-green) and enclose a flux Φ = Φ0 /2 due to a fine
tuned magnetic field B giving rise to four MBSs γα (red dots). The central constriction
results in a local gap opening in the one-dimensional surface mode due to the enclosed
magnetic flux ϕ < 1/2. As is quantitatively analyzed below, the top gate (center) allows
to manipulate the hybridization of the MBSs γ2 and γ3 localized across the constriction.
We assume the proximitized sections are long compared to the length scale ξ∆ = ~υ1 /∆
such that the outer red dots are effective MZMs.

For convenience, we model the device depicted in Fig. 3.3 with system parameters that
exhibit a step-like dependence on the z coordinate along the wire. To derive the finite
wavefunction overlap of the inner MBSs γ2 and γ3 (see Fig. 3.3), we assume for the
moment the length L of the two outer TSCs to be infinite. Hence, we model the radius
of the cylindrical wire as a Heaviside step function,

R(z) = R0 + (R − R0 )Θ(|z| − W/2) (3.20)

44
CHAPTER 3. MAJORANA QUBITS IN PROXIMITIZED TI NANORIBBONS

with R the radius of the outer part and R0 < R the smaller radius of the inner part. It
assumes interfaces with extension over a few lattice sites which are of order of 3nm [17].
The axial magnetic field B is fine tuned such that the dimensionless flux ϕ(z) ≡ Φ(z)/Φ0
takes on the value 1/2 in the outer parts. Thus, as a function of the z parameter, we
model ϕ as

 
1
ϕ(z) = ϕ0 + − ϕ0 Θ(|z| − W/2). (3.21)
2
The dimensionless flux through the constriction ϕ0 is calculated to be ϕ0 ≡ R02 /(2R2 ) <
1/2. The associated size quantization gap is denoted by

~υ2
M0 ≡ M (ϕ0 ) = |ϕ0 − 1/2| . (3.22)
R

The relation (3.21) neglects the effect of magnetic screening (flux channeling) in the con-
striction, which is justified because the magnetic susceptibility is small, especially since
the surface state spectrum is gapped. For the superconducting gap ∆(z) induced by the
proximitized s-wave SC in the outer regions of the device, we make the assumption that
its absolute value is the same on the left and right segments, i.e. |∆(|z| > W/2)| = ∆,
with ∆ ∈ R. We include a difference between the superconducting phases on both sides
denoted by φ, i.e.

∆(z) = ∆e−sgn(z)iφ/2 Θ(|z| − W/2). (3.23)

This relation neglects the breaking of rotational symmetry due to the s-wave supercon-
ductors, which has been discussed in Ref. [77]. We mention that φ will be a dynamical
quantity in a Coulomb blockaded device. We also include an electrochemical potential
term µ in the constriction which is induced by means of the gate electrode on top of
the constriction (see Fig. 3.3)

µ(z) = µΘ(W/2 − |z|). (3.24)

Despite this convenient choice, we note that finite µ in the outer region |z| > W/2 is
not predicted to result in qualitatively different physics.

45
3.2. GATE TUNABLE COUPLING OF MAJORANA STATES

To construct the solution of the 1D BdG equation [5],

HBdG Ψ = EΨ, (3.25)

we take note that the interface points, z = ±W/2, define three regions in which the
problem is effectively uniform. For large energies, the solution is obtained via a plane-
wave ansatz. But we are interested in the low-energy subgap regime |E| < min (∆, M0 ),
where angular momentum modes with j 6= 1/2 are negligible for moderate values of
the chemical potential µ < M0 . Making an appropriate evanescent state ansatz in the
three regions, we can write the requirement of continuity of the spinor wave function
Ψ(z) in the form of a corresponding condition of a vanishing determinant

D(E) = 0. (3.26)

Further details on the derivation and the specific form of the determinant D(E) for
|E| < min (∆, M0 ) are provided in Appendix A. The relation (3.26) contains the infor-
mation about the low-energy spectrum, which for general parameter values is obtained
via numerical solution. We find the robust existence of subgap states at E = ±ε, which
can be confirmed to constitute the expected pair of MBSs. With the self-consistent
assumption |ε| < min (∆, M0 ), a formula for ε can be derived from Eq. (3.26) by
second-order expansion of D(E) in E. The Majorana hybridization energy associated
with the inner two MBSs γ2 and γ3 is then given by
 
φ
ε(φ) = ε(0) cos , (3.27)
2

where the energy at vanishing phase difference, φ = 0, is given by

2∆ (~υ1 /ξ)2 −W/ξ


ε(0) = e . (3.28)
M0 ∆ + ~υ1 /ξ
Here, we have introduced the localization length scale

~υ1
ξ (µ) = p (3.29)
M02 − µ2
of wave function decay into the non-proximitized part. The central constriction of width
W realizes an insulating tunneling barrier between two 1D TSCs with MBSs at their

46
CHAPTER 3. MAJORANA QUBITS IN PROXIMITIZED TI NANORIBBONS

terminations. The dependence of the splitting ε(φ) of the ground state degeneracy on
the difference in the SC phases φ exhibits 4π-periodicity, according to Eq. (3.27). This
is a consequence of the fractional Josephson effect [4, 40] and, hence, in agreement
with the theoretical predictions for a Josephson junction with MBSs. The exactness
of the 4π-periodicity is an artifact of the fact that only the j = −1/2 mode has been
included in the derivation. The surface states with j 6= −1/2 cause small conventional
2π-periodic contributions.
The size of the hybridization ε is mainly determined by the length of the constriction W
in units of the length scale ξ (µ). In particular, the splitting is exponentially suppressed,
ε ∼ exp (−W/ξ), and thus negligible for sufficient spatial distance W  ξ between the
MBSs. In the limit W → ∞, a pair of exact Majorana zero modes is present and the
approximate ground state degeneracy becomes exact. The Bogoliubov-de Gennes solu-
tion (see Appendix A) Ψ(z) describing the Majorana states localized at the interfaces
decays exponentially into the outer segments |z| > W/2 on the length scale

~υ1
ξ∆ = . (3.30)

We now consider the example of a Bi2 Se3 nanowire with radius R = 35nm in the outer
regions and an inner radius of R0 = R/2. The Fermi velocities of Bi2 Se3 along the
ez direction is given by ~υ1 = 226meV × nm and along the azimuthal direction it is
υ2 = 1.47υ1 [68]. The size quantization gap M0 ' 7.14meV reigning in the constriction
is large compared to the proximity gap, which we assume to be ∆ = 0.18meV in
the outer regions. In Fig. 3.4, we plot the Majorana hybridization energy ε = ε(0)
depending on the constriction length W for different fixed values of the electrochemical
potential µ. This plot has been obtained by numerical solution of Eq. (3.26) and is in
agreement with the exponential suppression ε ∼ exp (−W/ξ) for W  ξ predicted in
formula Eq. (3.28). Moreover, the length scale ξ = ξ(µ) of decay in Fig. 3.4 is consistent
with Eq. (3.29). We infer from Fig. 3.4 that in the limit of a short constriction, W → 0,
we find ε → ∆. This behavior is expected since the MBSs increase their overlap and
eventually reach the quasiparticle continuum of the spectrum.

47
3.2. GATE TUNABLE COUPLING OF MAJORANA STATES

200

100
ε [μeV ]

50

μ=6meV, ξ=58.4nm
20
μ=3meV, ξ=34.9nm
μ=0meV, ξ=31.7nm
10
0 20 40 60 80 100
W [nm]
Figure 3.4: Hybridization ε (in units of µm) of the two inner Majorana states γ2 and
γ3 for φ = 0 plotted (semi-logarithmically) as a function of the length W (in nm) of
the unproximitized central part for the device displayed in Fig. 3.3. The proximity gap
and the size quantization gap are assumed to be ∆ = 0.18meV and M0 ' 7.14meV
respectively. Furthermore, we consider the TI nanoribbon to be made of Bi2 Se3 . We
display the result for several values of the electrochemical potential µ in the central
region. The corresponding length scale ξ is shorter for smaller values of µ. In the case
that W  ξ, the plots are in agreement with the prediction ε ∼ e−W/ξ of Eq. (3.28).
On the other hand, for ε → ∆ we obtain W → 0 as expected. The plot was taken from
our publication Ref. [64].

In practice, the constriction width W is constant, but the experimentalist can tune the
effective length scale ξ (µ) in Eq. (3.29) by changing the local chemical potential using
a local gate, see Fig. 3.3. Hence, the gating allows to manipulate the effective distance
between the two inner Majorana states. Increasing the electrochemical potential µ
makes the weak link more transparent and the coupling ε is enhanced. The coupling
is at its minimum for µ = 0. In Fig. 3.4, the hybridization ε is numerically plotted
as a function of µ. These numerical plots are consistent with the analytical formulas
stated in Eq. (3.28). Thus, the results displayed in Fig. 3.5 indicate that ε can be
conveniently controlled.

48
CHAPTER 3. MAJORANA QUBITS IN PROXIMITIZED TI NANORIBBONS

5 W =200nm
ε [μeV] 4 W =250nm
3 W =300nm

2
1
0
0 1 2 3 4 5 6 7
μ [meV]
Figure 3.5: Hybridization ε (in units of µm) between the inner Majorana states γ2
and γ3 for φ = 0 plotted as a function of the electrochemical potential µ (in units of
meV). The plots are displayed for the constriction widths W = 200nm, W = 250nm and
W = 300nm. Like in Fig. 3.4, we consider Bi2 Se3 and use the parameters ∆ = 0.18meV
and M0 ' 7.14meV. The hybridization ε is minimal for µ = 0 and remains close to this
value for µ  M0 . However, at some point ε begins to grow at a bigger rate upon the
increase of µ. The plot shows that the coupling of the MBSs can be manipulated via
local gating in the constriction. The plot was taken from our publication Ref. [64].

49
3.3. MAJORANA BOX QUBITS FROM TI NANORIBBONS

3.3 Majorana box qubits from TI nanoribbons

The Subsection 3.3.1 is inspired by the Refs. [11, 12] and analyzes the Coulomb block-
aded version of the proposed device presented in the previous section to perform read-
out and initialization. In Subsection 3.3.2, we study the options to realize devices with
switchable grounding of the TI nanoribbon platform following Ref. [39]. Finally, in
Subsection 3.3.3, we compare the TI nanoribbon based approach to Majorana qubits
to the SM based approach.

3.3.1 Floating box qubit and elementary quantum operations

To encode topologically protected qubits, we consider the floating version of the TI


nanoribbon device, as shown in Fig. 3.6, with four MBSs on the box defining the
corresponding Hilbert space. We assume the outer MBSs γ1 and γ4 to be effective zero-
modes. This means that the length L of the proximitized segments is large compared to
ξ∆ = ~υ1 /∆ such that the wave functions of the outer MBSs γ1 and γ4 have negligible
overlap with all other MBSs hosted by the floating island.

γ1 γ2 γ3 γ4
t0

σ1
Figure 3.6: Floating version of the quantum device shown in Fig. 3.3 with four localized
MBSs γα (red dots). The s-wave SCs (gray-green) coating the TI nanoribbon (gray)
on the outer regions have a SC bridge in between them so that the Majorana island
is characterized by a single charging energy EC . Normal leads (thick black lines) are
tunnel coupled to MBSs, and coupled amongst themselves by an interference link t0 .
This is to make use of interferometric readout schemes [11, 12] of the nonlocal Pauli
operators, here σ1 = iγ1 γ2 .

Note that the two superconducting halves are connected by a superconducting bridge
(see Fig. 3.6) such that the box is characterized by a single charging energy EC . In

50
CHAPTER 3. MAJORANA QUBITS IN PROXIMITIZED TI NANORIBBONS

this case, the effective Hamiltonian of the qubit device capturing the physics on energy
scales small in comparison to ∆ and EC is dominated by the hybridization energy ε
given in Eq. (3.27). It reads
 2
Heff = ε(iγ2 γ3 ) + EC N̂ − ng , (3.31)

where a side-gate controls the charging term via the back gate parameter ng . The
operator N̂ = 2N̂s + n̂12 + n̂34 counts the number of electrons on the island, where
N̂s counts the island Cooper pairs, and n̂12 + n̂34 yields the number of fermions in
the Majorana domain. Here, n̂ij ≡ d†ij dij is the occupation of the nonlocal fermion
d†ij = γi − iγj . Physically, the low-energy Hamiltonian Heff neglects quasiparticles
above the topological gap ∆ as well as the high lying surface states with j 6= −1/2.
The degree to which the inner Majoranas at the two sides of the constriction hybridize
sets the qubit lifetime due to dephasing. On time scales below this dephasing time
and below the characteristic poisoning time, the state of the qubit is restricted to the
almost degenerate low-energy ground state manifold. For ng close to an integer value,
the charge quantization on the box implies that global fermion parity is fixed [25]. For
concreteness, we assume the parity to be odd, i.e.

P ≡ γ1 γ2 γ3 γ4 = −1. (3.32)

Analogously to the box qubits based on SM nanowires discussed in Section 2.2, the box
degrees of freedom are given by the fractionalized Majorana bilinears [25]

σ1 = iγ1 γ2 , σ2 = iγ3 γ2 , σ3 = iγ3 γ1 . (3.33)

Since fermion parity is conserved, the identification of a bilinear, e.g. formed from the
pair γ2 and γ3 , always automatically determines the bilinear formed from the remaining
two Majoranas, e.g. iγ1 γ4 = −iγ2 γ3 . Labeling eigenstates based on the eigenvalues nij ,
the nonlocal logical qubit state |ψi can be encoded in the 2D Hilbert space spanned
by the states |↑i ≡ |012 , 134 i and |↓i ≡ |112 , 034 i. Qubit states can now be defined
as |ψi = α |↑i + β |↓i with α, β ∈ R. Hence, it holds e.g. that σ1 |↑i = |↑i and
σ1 |↓i = − |↓i. To readout and initialize the qubit, the interferometric projective con-
ductance measurement scheme [11, 12] reviewed in Chapter 2 can be applied. In this
way, all bilinears in Eq. (3.33) can be addressed via electrodes tunnel-coupled to the

51
3.3. MAJORANA BOX QUBITS FROM TI NANORIBBONS

corresponding pair of MBSs. We point out that the interference link required to read
out σ2 = iγ3 γ2 does not have to be realized as an additional structure, because the
superconducting bridge (see Fig. 3.6) can be used for this purpose [12].

3.3.2 Devices with switchable grounding

We now discuss the option to switch the devices from the grounded to the non-grounded
regime and vice versa. This has been discussed to be useful for “parity-to-charge”
conversion protocols, as well as for fusion rule or braiding confirmation, by Aasen et
al. in Ref. [39]. The discussion in this reference is geared to the semiconductor based
platform but can be applied to the TI nanoribbon based platform as well.
The switchable grounding requires to connect the proximitized nanoribbon which hosts
MBSs, to a grounded superconducting bulk reservoir. To implement such a coupling,
we again employ a non-proximitized TI nanoribbon with a lesser cross section than the
proximitized nanoribbon, as shown in Fig. 3.7. The principle is the same as for the
weak link connecting the two poximitized TI nanowire segments discussed in Section
3.2: Due to geometric confinement, the connecting section is gapped and, therefore,
constitutes a tunnel junction.

Bulk SC γ1 γ2

Φ < Φ0 / 2 Φ = Φ0 / 2

Figure 3.7: Switchable grounding [39] can be achieved by once again employing a non-
proximitized TI wire segment of smaller cross section. In the present case, this segment
is used to couple a proximitized TI nanoribbon realizing a TSC to a grounded bulk
superconducting reservoir. A local top gate is installed at the narrowed segment and
allows to control the Josephson coupling between the reservoir and the TSC. In this way,
the ratio of charging energy EC and Josephson energy EJ can be tuned. Switching from
the grounded regime (EJ  EC ) to the floating regime (EC  EJ ) makes it possible to
achieve “parity-to-charge conversion” and to read out the corresponding charge states
using a charge sensor [39].

52
CHAPTER 3. MAJORANA QUBITS IN PROXIMITIZED TI NANORIBBONS

The Hamiltonian of the device shown in Fig. 3.7 contains not only the charging energy
EC , but also the Josephson energy EJ describing the coupling to the bulk supercon-
ductor due to Cooper pair tunneling [25, 28, 39],
 2
Heff = ε(iγ2 γ3 ) + EC 2N̂s + n̂12 + n̂34 − ng − EJ cos φ̂s . (3.34)

The hybridization energy ε of the inner MBSs γ2 and γ3 has already been stated in Eq.
(3.27). Furthermore, N̂s and n̂12 +n̂34 are defined as in 3.3.1. Note that N̂s is canonically
conjugate to the phase difference φ̂s between the left and right superconductors. The
tunneling strength and, thus, the Josephson energy EJ can now be influenced by a
gate placed on top of the narrow TI wire segment. This makes it possible to tune
from the grounded regime (EJ  EC ), where two SCs behave essentially like a single
SC, to the floating regime (EC  EJ ) and vice versa. The tuning of the ratio EJ /EC
provides a way to read out iγ1 γ2 by converting the corresponding parity eigenstates into
charge states [39]. The latter can be read out by a capacitively coupled single electron
transistor acting as a charge sensor. For nanowires of length of a few µm, a typical
charging energy is EC ' 0.1K. We then expect a tunable parameter range of order
0.1 . EJ /EC . 10 [64].

3.3.3 Majorana qubit comparison: TI nanoribbon vs SM plat-


form

In this section, we critically discuss and evaluate the proximitized TI nanoribbon plat-
form and compare it to other approaches to Majorana qubits. Moreover, we discuss po-
tential challenges of the devices e.g. due to harmful physical mechanisms. The MBSs in
the TI nanoribbon construction derive from the protected surface states of a topological
insulator. Those surface states are endowed with robust protection from pair-breaking
disorder as well as from elastic impurity scattering [75]. The device cleanliness of the
semiconducting nanowire platform is presently more pure [15, 64]. Thus, advances in
the experimental physics of TI nanoribbons are required for proof-of-principle demon-
stration of the platforms practicability, e.g. towards QIP applications.
An important property of the MBSs is their typical localization length scale. For the
TI nanoribbon platform the decay of the Majorana wavefunction into the proximitized
segment is characterized by the length scale ξ∆ = ~υ1 /∆, which is vastly longer than

53
3.3. MAJORANA BOX QUBITS FROM TI NANORIBBONS

that for the decay into the constriction. If we assume a Bi2 Se3 nanowire and a proximity
gap ∆ = 0.18meV, we obtain ξ∆ = 1.25µm. This estimate reveals that the proximitized
TI nanowires have to be of substantial length & 5µm in order to have sufficiently small
overlap of γ1 and γ2 (γ3 and γ4 ). We note in this context that nanoribbons of such
length are already available [72, 73]. Nevertheless, the required wire length is a slight
disadvantage in comparison to the semiconductor nanowires, where a spin-orbit coupling
energy of order ~α ≈ 20eV × nm corresponds to a localization length of 125nm [81].
The MBS localization length scale reigning in the inner, non-proximitized segment is
p
given by ξ (µ) = ~υ1 / M02 − µ2 and is typically shorter than ξ∆ for realistic system
parameters, as discussed in Section 3.2.
Experimentally, it may be challenging to arrange the magnetic flux in such a way that
it realizes the desired value ϕ = 1/2 in the proximitized segments. Despite the fact that
magnetic fields can be fine tuned with great precision, a mismatch of the flux could be
caused for several reasons:

1. The area of cross section is not completely homogeneous. Such imperfections


imply that the flux will display small variations as a function of z.

2. The spatial orientation of the TI nanoribbon has to be completely parallel to the


applied magnetic field. This is another source of error which also causes slight
breaking of the rotational symmetry leading to admixtures of the higher-energy
states with j 6= −1/2.

3. The different segments of the device have to be aligned along a common axis,
which constitutes another source of error.

Despite these factors, we expect that the discrepancy in magnetic flux can be limited
to be a rather weak perturbation. A flux ϕ = 21 + δϕ with small constant mismatch
δϕ  1 results in a finite mass parameter M (ϕ) = ~υR2 |δϕ| in the proximitized regions.
We note that in this case a finite electrochemical potential µS = µ(|z| > W/2) in the
proximitized regions is required, i.e. Eq. (3.24) has to be modified. More precisely,
M (ϕ) < |µS | < ~υ2 /R is necessary for well-defined 1D surface states for ∆ = 0, which is
the condition for MBSs at finite pairing. In summary, we anticipate that the inevitable
flux discrepancy is not detrimental to the robustness of the MBSs. The electrochemical
potential µS (focusing on the case ϕ = 1/2) does not change the qualitative results: the
quantity µS manifests itself mainly in the difference µ − µS which enters the formulas.

54
CHAPTER 3. MAJORANA QUBITS IN PROXIMITIZED TI NANORIBBONS

This is expected since the dispersion of the j = −1/2 states is linear in the absence of
a proximitized superconductor.
In view of quantum hardware applications, the maximal time scale τ0 on which the
Majorana box quit is operational is an important characteristic property. As discussed
above, in our proposed platform the qubit splitting due to the wave function overlap of
the two inner MBSs eventually dephases the state of the qubit. Thus, the time scale is
given by
τ0−1 = ε(0) (3.35)

with ε(0) as given in Eq. (3.28). For a given device this scale can be maximized
by designing the constriction section W as long as possible. For W = 300nm an
electrochemical potential within the window |µ| < 0.2meV keeps the hybridization
bounded as follows: ε(0) < 0.027µeV. This corresponds to coherence time scales longer
than τ0 > 2.4µs. Notably, the time scale for semiconductor nanowires is shorter [39],
i.e. this is a clear advantage of the proximitized nanoribbon architecture. While the
maximum of τ0 is achieved for µ = 0, there is reasonable range for the chemical potential
set by the parameter M0 in Eq. (3.22) in which the hybridization is not affected too
much. For the proximitized semiconducting nanowires on the other hand, the workable
interval is smaller. Indeed, the chemical potential needs to be tuned to the bottom of
the band. This makes the states more sensible to disorder.
Semiconducting nanowires currently represent the experimentally most advanced plat-
form toward MBS realization. In such devices the QP poisoning time has been studied
and was found to be & 1µs [31]. Similar investigations still have to be performed for the
proximitized TI devices. This is another important aspect in a comprehensive platform
comparison. Regarding the links that provide a way to interferometrically read out and
initialize the qubit, the TI nanoribbon platform offers the perspective to forge these
links from the TI itself (see Fig. 3.3). This may be another advantage compared to SM
devices, where the need for separate links has proven to be challenging [82].

55
3.4. CONCLUSIONS AND OUTLOOK

3.4 Conclusions and outlook

In this chapter, we have presented new design architectures for Majorana qubits that
are based on TI nanoribbons, e.g. made of Bi2 Te3 or Bi2 Se3 . In proximity to an s-wave
SC and in the presence of a fine tuned axial magnetic field, TI nanoribbons realize a 1D
TSC. In the proposed layouts, shown e.g. in Fig. 3.3, two linearly aligned TSCs of this
kind are connected by a non-proximitized narrow TI constriction, effectively dividing
the TSC into two halves. Due to lowered magnetic flux in the TI element connecting
the two halves, a size quantization gap is opened locally. As discussed in Sections
3.1 and 3.2, MBSs are localized at the two domain walls that separate regions either
dominated by the pairing gap ∆ or by the size quantization gap M0 . In Section 3.2, we
have calculated the coupling of the two MBSs (γ2 and γ3 in Fig. 3.3) localized across a
constriction of finite length W , with the result being stated in Eqs. (3.27) and (3.28).
p
Importantly, we identified the length scale ξ (µ) = ~υ1 / M02 − µ2 , with υ1 being the
Fermi velocity in longitudinal direction. This length scale sets the effective distance
between the inner MBSs γ2 and γ3 . Tuning of the local electrochemical potential µ
in the constriction therefore enables to control the hybridization of the pair of MBSs.
This tuning can be conveniently performed via gating and allows to switch from a
regime where the Majoranas are coupled, to a regime where the coupling is suppressed
and the quantum information is (better) protected from decoherence. Furthermore, the
derived result for the hybridization exhibits 4π-periodicity in the superconducting phase
difference φ, which is in agreement with the theoretical predictions for the fractional
Josephson effect [4, 40].
Throughout this analysis, several approximations have been made. First, the results
have been derived under the assumption that the two outer MBSs (γ1 and γ4 in Fig. 3.3)
are far enough away to have negligible wavefunction overlap. Moreover, we worked in
a low-energy approximation scheme throughout this chapter. Thus, we have neglected
modes with angular momentum j 6= −1/2 under the assumption that the energy scale ∼
~υ2 /R is large. Perturbative inclusion of the bands with j 6= −1/2 leads to corrections,
e.g. 2π-periodic admixtures are expected in the result for the Majorana coupling stated
in Eq. (3.27). Moreover, we neglected any contributions from the bulk of the TI
nanoribbon. The effects of flux mismatch were discussed in Subsection 3.3.3. Despite
the simplifications, we are confident that the most essential low-energy features were
captured in the modeling.
We note that an alternative technique to describe the device consists in the use of

56
CHAPTER 3. MAJORANA QUBITS IN PROXIMITIZED TI NANORIBBONS

k · p-theory. This approach is based on the low-energy bulk Hamiltonian and requires
imposition of Dirichlet boundary conditions on the nanoribbon surface (see e.g. Ref.
[68]). The problem posed in this way can be approached numerically. The general
expectation is that the essential theoretical results described above can be reproduced
with this method.
As long as functional Majorana qubits are not at our disposal, it is important that
different promising proposals are being actively pursued, because it is difficult to predict
a priori which approach will work best. The comparison to the semiconducting quantum
wire platform, performed in Subsection 3.3.3, has shown that, in fact, TI nanoribbon
Majorana qubits represent a promising and viable alternative. The reasons for this
include but are not limited to the fact that the TI nanoribbon surface states can be
anticipated to yield MBSs of reasonably high robustness, possibly more so than for
SM wire platforms, because of the linear spectrum of the 1D modes. Moreover, the
only requirement for the chemical potential is that it lies within the bulk band gap.
In contrast to other TI based proposals, a ferromagnetic insulator is not necessary to
realize an edge localizing Majorana states.
Overall, we believe the platform could prove to be versatile and accommodating for
the purpose of demonstrating and ultimately using Majorana qubits. However, further
experimental progress has to be made regarding the material platform. TI nanoribbons
of sufficient length are already available [72, 73]. And moreover, experiments have
reported Andreev reflection from the surface states in a TI nanowire Josephson junction
based on the TI material BiSbSeTe2 [83, 84]. This gives further credence to the prospect
that the achievement of topological superconductivity in these platforms is realistic (as
was also pointed out in Ref. [76]). A proximitized nanowire could initially be checked
by making point contacts at the ends so that the ZBCPs and their dependence on the
flux can be measured. In the next subsection, we discuss applications of the device
platforms that can be envisioned once they are under experimental control.

3.4.1 Outlook: Majorana surface code and beyond

We now turn to more ambitious long-term applications of the TI nanoribbon qubit


devices. We start by a discussion of a particularly attractive application, namely, the
Majorana surface code. This code holds promise for efficient quantum information pro-
cessing and its theory has been developed in the Refs. [85, 86, 87, 88]. Majorana surface

57
3.4. CONCLUSIONS AND OUTLOOK

code implementations require to manufacture extended 2D network structures compris-


ing many tunnel coupled Majorana box qubits, see Fig. 3.8. Hence, a precondition for
the Majorana surface code is that the qubits are established and under experimental
control.

Figure 3.8: Majorana surface code (see Refs. [86, 87, 88]) relying on Majorana box
qubits made of two parallel proximitized TI nanoribbons connected by a SC bridge.
Each of the boxes is therefore described in terms of its own charging energy, and hosts
four MBSs (red dots). Neighboring qubits are tunnel coupled, where the tunneling can
be formed by using the TI material itself. In the two-dimensional network structure,
stabilizer operators of two types, denoted A and B, are defined as products of eight
Majorana operators belonging to MBSs (red dots) going around minimal plaquettes of
type A or B as indicated. The figure was taken from our publication Ref. [64].

For the purpose of serving as the building block in a Majorana surface code, we expect
that it is convenient to realize the TI based Majorana qubit in a “H”-shaped geometry,
see Fig. 3.8. This is achieved by having two parallel, proximitized TI nanoribbons
connected by a superconducting bridge. In contrast, in the Sections 3.2 and 3.3, we
considered two 1D TSCs aligned on the same axis (see Figs. 3.3 and 3.6). Moreover, the
necessary tunneling links to connect MBSs on neighboring boxes (as well as interference
links for the interferometric readout) could be realized from the TI material itself.
The network fabrication itself is a challenging task that may be achieved using refined

58
CHAPTER 3. MAJORANA QUBITS IN PROXIMITIZED TI NANORIBBONS

lithographic and wet etching techniques.


The operational principle of the Majorana surface code is founded on the so-called
stabilizer operators of which there are two types denoted A and B. These operators are
defined as the products of the eight Majorana operators corresponding to the minimal
loops of type A and B displayed in Fig. 3.8. The mutually commuting stabilizer
operators have eigenvalues ±1, defining the so called physical qubits of the system. All
stabilizer operators are now repeatedly measured, hence, projecting the code onto a
well-defined highly entangled simultaneous eigenstate of the system. The logical qubits
are formed by a few qubits that are not being measured. The access elements for
initialization, readout and manipulation can be realized by tunnel conductance probes
and charge pumping by means of single-electron transistors [87].

Another perspective for the TI nanoribbon platform is the confirmation of Majorana


fusion rules using the protocol of Aasen et al. [39]. It may be possible to apply this
protocol to a proximitized TI nanoribbon device as shown in Fig. 3.9. This qubit
device is connected on both ends with a bulk superconducting reservoir as described
in Subsection 3.3.2. Additionally to the two non-proximitzed constrictions serving for
this purpose, the device contains a third narrow TI constriction connecting the two
proximitized halves of the nanoribbon as discussed in Section 3.2. As outlined before,
the manipulation of the three junctions can be achieved via gates placed at the three
constrictions.

Bulk γ1 γ2 γ3 γ4
SC

Figure 3.9: The TI nanoribbon qubit device (see Fig. 3.3) is connected with bulk
superconducting reservoirs at both ends, as described in Subsection 3.3.2. The three
non-proximitized constrictions constitute barriers that can be regulated via gates. Fol-
lowing the protocol of Aasen et. al may make it possible to confirm Majorana fusion
rules [39] by reading out the charge on the two floated halves after the necessary steps
of the protocol are performed. The time-dependent steps of the protocol involve the
lowering and raising of the three barriers in a certain order.

Following the steps of the aforementioned protocol, could make it possible to confirm
Majorana fusion rules. At the end of this protocol, the two fusion channels of the Ising

59
3.4. CONCLUSIONS AND OUTLOOK

anyons manifest themselves in a probabilistic readout of the charge in the two floated
halves of the device [39].

Another perspective is that as soon as lithographically defined TI nanoribbons are


demonstrated, fabrication efforts could concentrate on T-junction geometries [37] to
carry out braiding experiments [39]. In summary, there are multiple exciting directions
to pursue once the platform is under experimental control.

60
Chapter 4

Majorana qubit detection via


simultaneous weak measurement of
its nonlocal Pauli operators

In this chapter, we introduce a novel weak measurement approach to Majorana qubit


detection based on the shot noise of tunneling current probes. We formulate protocols
that lie within the scope of present day experimental physics and provide experimen-
tally verifiable, quantitative predictions that allow to test the genuineness of MBSs.
The protocols aim at a core property of the Majorana qubit, its nonlocal Pauli algebra,
which is closely related to the non-Abelian braiding.
The remainder of this chapter is structured as follows: In Section 4.1, we give an expla-
nation and qualitative discussion of the envisioned experiments. The Sections 4.2 and
4.3 describe the theoretical modeling of the corresponding devices. The phenomenology
of the Majorana box qubit is presented in detail in Section 4.4, and to draw a contrast
the corresponding signatures for Andreev bound states (ABSs) are discussed in Section
4.5. These results are condensed into experimental protocols that allow for the clear
identification of MBSs and Majorana qubits in Section 4.6. Throughout this chapter,
we work in units of
~ = e = kB = 1. (4.1)

Most results presented in this chapter are published in Ref. [89]. The present chapter
also contains additional details and results in comparison to the published material,
e.g. in Subsections 4.4.2, 4.4.3, 4.4.4 and 4.4.5.

61
4.1. EXPERIMENTAL SETTING AND QUALITATIVE DISCUSSION

4.1 Experimental setting and qualitative discussion


According to fundamental quantum physics, it is impossible as a matter of princi-
ple to simultaneously perform projective measurements of noncommuting observables.
Nevertheless, simultaneous measurements are, in fact, possible using weak continuous
quantum measurements [90]. It is at the heart of our proposal to study the Majorana
box qubit to simultaneously realize continuous weak measurement of its two nonlocal
Pauli components σ1 and σ2 . This approach is illustrated in the schematic of Fig. 4.1
where two tunneling currents I1 and I2 are sensitive to σ1 and σ2 , respectively. For
an introduction to weak continuous measurement we refer to Refs. [91, 92, 93]. The
simultaneous weak measurement of orthogonal qubit components was previously con-
sidered in Refs. [90, 94, 95]. The observables σ1 and σ2 cannot be known sharply at

Figure 4.1: Bloch sphere for a qubit described by the density matrix ρ. The variables
Iα (t) with α = 1, 2 are the outputs of detectors weakly continuously measuring the
qubit component σα yielding a time continuous noisy signal (e.g. voltage but in the
realizations discussed below it is current). The interaction between the qubit and the
detectors is assumed to be sufficiently weak so that it takes time to accumulate infor-
mation and distort the qubit. For the simultaneous monitoring of the noncommuting
pseudospin components the different imprecise (weak) readouts are incompatible. This
figure was inspired by Ref. [95].

the same time. Instead, there is a competing tendency to partially collapse the system
state over time. This implies a random diffusion of the density matrix in the Bloch
sphere [95]. Furthermore, the information about the initial state of the qubit is lost

62
CHAPTER 4. MAJORANA QUBIT DETECTION

on the timescale of the dephasing rate [94]. However, as we will see the statistics of
the detector outcomes encode valuable information about the underlying Pauli algebra.
The latter is a unique property of the Majorana qubit, which is alternatively verified
by means of a more elaborate braiding protocol.
How can the situation abstractly illustrated in Fig. 4.1 be realized in terms of concrete
devices? The experimental setup we propose contains a Coulomb blockaded topolog-
ical superconductor island with three normal-conducting leads weakly coupled with
amplitudes λα to different Majorana states, see Fig. 4.2. If the neighboring leads
are coupled amongst themselves with weak tunnel links t0 , a two-sided interferomet-
ric setting is realized. As discussed in Section 2.2, the effective low-energy tunnel-

I1
λ1
γ1
t0 gate
λ0
γ0
t0 V
γ2 λ2
I2

Figure 4.2: Floating Majorana island realized from three parallel Majorana quantum
wires (gray-green) connected by a topologically trivial SC backbone (gray). The three
Majorana states γα localized at the right ends of the wires (red dots) are tunnel coupled
to normal-conducting electrodes (thick black lines) with amplitudes λα . To realize a
two-sided interferometer, direct tunneling links t0 are installed in between neighboring
leads. The central lead (α = 0) is operated as a source of electrons by means of
an applied bias voltage V driving the currents I1 and I2 to ground, which couple to
σ1 = iγ1 γ0 and σ2 = iγ2 γ0 respectively. The MBSs localized at the left wire ends
may or may not be fused into a single Majorana, and are not displayed since they are
uncoupled and do not influence the transport.

ing between the leads 0 and 1 is known to be described by the effective Hamiltonian
H̃T = (t0 + t1 (iγ1 γ0 )) c†1 c0 +H.c. with cα = k cα,k the lead fermions [11]. For decoupled
P

lead 2 and applied bias V , the measurement of the tunneling current I1 (t) effectively
constitutes a continuous weak measurement of iγ1 γ0 . The latter becomes projective
after a time of order ∼ 1/V . Depending on the eigenstate of iγ1 γ0 that the system gets

63
4.1. EXPERIMENTAL SETTING AND QUALITATIVE DISCUSSION

projected to, one finds two outcomes (in units of e2 /~) [11]

hI1 i = 2πν 2 V |t0 |2 + |t1 |2 + 2Re(t∗0 t1 )(iγ1 γ0 ) .



(4.2)

Because of the symmetry of the setting, the same thing can be said about the average
current hI2 i if lead 1 is decoupled instead of lead 2. When all three leads are coupled
and the bias is applied, the two readouts of the tunneling current Iα=1,2 (t) are therefore
incompatible. No simultaneous eigenstate of σ1 = iγ1 γ0 and σ2 = iγ2 γ0 exists and the
qubit is conflicted and cannot approach a pure state. Strong Coulomb blockade is a
crucial ingredient for this physics, because in a grounded device an electron tunneling
into γ0 would be approximately uncorrelated with an electron tunneling out of the
island via γ1 or γ2 . Our strategy is to look for signatures in the shot noise of the two
currents that reflect the continuous monitoring of the nonlocal Pauli operators [25],

σ1 = iγ1 γ0 , σ2 = iγ2 γ0 , σ3 = iγ2 γ1 . (4.3)

Note that the algebra is not of the Pauli type anymore in the competing case of ABSs,
which, as we will show, has profound consequences that allow to distinguish such bound
states. The prime observable of interest is the current cross-correlation amplitude,

ˆt
S12 = dt hhI1 (t)I2 (0)ii , (4.4)
0

where we define the second cumulant in the standard way as hhABii = hABi − hAi hBi.
In this chapter, we study the observable S12 in the long time limit, both in the presence
and absence of interference links (see the vertical dashed lines in Fig. 4.2) connecting
the electrodes coupled to the island amongst themselves. The structure defined by these
links can be controlled via gate electrodes during an experiment and makes a detailed
noise profile of MBSs clearly accessible as we will see. The signatures are qualitatively
unique to the fluctuating current for a device with either MBSs or Andreev bound
states.

64
CHAPTER 4. MAJORANA QUBIT DETECTION

4.1.1 Device geometries

I2
γ3 γ2
γ4 SC γ0
V
γ5 γ1
I1
Figure 4.3: The envisioned experiment can be carried out on a wide range of possible
device layouts with the hexon geometry [12] being a viable and advantageous option.
This system consists of three parallel Majorana wires (gray-green) joined by a trivial
SC (gray) to form a single island. Of the six MBSs present in the system, only those
localized on the right ends are being addressed by coupled leads (here: γ0 , γ1 and γ2 ).
The three MBSs on the other side (here: γ3 , γ4 and γ5 ) are far enough away to be
decoupled and irrelevant for the transport. In this geometry as well as in the geometry
displayed in Fig. 4.2, it is possible to have sufficiently short interference paths (vertical
dashed lines) to maintain phase coherence.

The envisioned effects and experiments can be carried out on a wide range of de-
vice geometries as well as different condensed matter platforms that give rise to zero-
dimensional MBSs. For concreteness in this chapter, we have the semiconducting wire
platform in mind. The geometry displayed in Fig. 4.2 represents a one-sided hexon.
The topologically trivial SC backbone connects three parallel SM wires that are driven
into the topological phase. The MBSs to the left are either fused into a single un-
coupled Majorana or not, but are in any case uncoupled and inconsequential for the
physics of interest. Instead, one can also use the hexon geometry, see Fig. 4.3. Of
the six Majoranas present in this system, we are only addressing three of them that
are localized on one side. Another possible option is to use geometries where only two
topological superconducting wires are connected by a trivial superconductor to form a
single island. For a genuine Majorana qubit, the charge quantization on such islands
implies the parity constraint γ0 γ1 γ2 γ3 = ±1 [25]. The first example of such a device
is shown in Fig. 4.4 where one uses the “tetron” or “Majorana box qubit” geometry
[11, 12] discussed in 2.2. The second option displayed in Fig. 4.4 arranges the two SM
wires on the same axis connected by a U-shaped s-wave superconducting bridge. This
can also be seen as a close relative of the device discussed in Chapter 3. The drawback

65
4.1. EXPERIMENTAL SETTING AND QUALITATIVE DISCUSSION

I2

γ2 γ0
γ3 SC γ1 V
I1

Figure 4.4: The schematic gives another example for a realistic geometry where only
two topological superconducting wires (gray-green) are connected by a trivial super-
conductor (gray) to form a single island, in this case a “tetron” or “Majorana box
qubit” [11, 12]. Compared to the geometries based on three parallel Majorana wires
forming a single island the displayed geometry has an important drawback: one of the
interference links has to transport electrons phase coherently along the entire length of
the Majorana wire. It has been proposed to use another Coulomb blockaded Majorana
wire to realize a long phase coherent interference link [11].

of the geometries shown in Fig. 4.4 and 4.5 is the fact that one of the interference links
has to transport electrons along the entire length of the Majorana wire. The latter have
to be of the order of one micrometer for the MBSs to have sufficiently small overlap,
which clearly exceeds typical phase coherence lengths of e.g. a semiconductor. It has
therefore been proposed, e.g. in Ref. [11], to use another Coulomb blockaded SM wire
in the topological phase to realize a sufficiently long phase coherent interference path.

γ1 γ0 γ2 γ3
I1 V I2

Figure 4.5: The schematic gives a second example for a realistic geometry where only
two topological superconducting wires (gray-green) are connected by a trivial super-
conductor (gray) to form a single island and are positioned on the same axis. One can
also think of the qubit geometry based on proximitized TI nanoribbons proposed in
Chapter 3. Once again, this geometry makes it necessary to have a reference arm of
length of the 1D TSC itself.

66
CHAPTER 4. MAJORANA QUBIT DETECTION

4.2 Effective Hamiltonian for the weak measure-


ment settings

In this section we discuss the low-energy effective Hamiltonian description of the three-
terminal Majorana devices (see Figs. 4.2, 4.3, 4.4 and 4.5). Assuming the parallel
proximitized nanowires are driven into the topological regime by a sufficiently strong
magnetic field (Γ > Γc ), MBSs emerge at the wire terminations. The topological gap
∆ separating the corresponding (almost) zero energy states and the quasicontinuum of
extended states as well as the charging energy EC shall define much larger energy scales
than temperature and the applied bias, i.e.

V, T  ∆, EC . (4.5)

The charge conservation on the mesoscopic island implies that the conjugate variables
of island electron number N and superconducting phase φ are described by operators
with commutation relation [φ, N ] = 2i. Exponentiation of the phase operator e±iφ/2
yields an operator which adds (removes) an electron charge [29]

e±iφ/2 |N i = |N ± 1i , (4.6)

where |N i is the charge state of the island. Now, consider the field operator Ψα which
annihilates an electron on the SC close to the nearby lead α. In general, Ψα is ex-
panded in terms of all MBSs on the island as well as the Bogoliubov quasiparticles of
the SC. However, we are now interested in the low-energy physics where the quasipar-
ticles above the gap are not important. Furthermore, in the considered geometries the
MBSs localized near leads α0 6= α have exponentially suppressed amplitudes and do not
contribute. Thus, the field operator Ψα affords the low-energy representation solely in
terms of the Nα MBSs localized close to lead α [29, 30]


X

Ψα = ξα,j e−iφ/2 γαj + . . . , (4.7)
j=1


where ξα,j is related to the wavefunction belonging to γαi . This relation encapsulates the
break-up of the electron statistics and its charge degree of freedom after it has tunneled
into the superconductor. Its fermionic anticommutation relations are maintained within

67
4.2. EFFECTIVE HAMILTONIAN

the neutrally charged localized MBSs while its charge is spread over the whole SC island.
By inserting the low-energy approximation (4.7) of the electron field operators into the
tunneling Hamiltonian HT = k 2α=0 λ̃α c†α,k Ψα +H.c. between the leads and Majorana
P P

island [30], we obtain the low-energy tunnel Hamiltonian

Nα X X
X 2
HT ' λjα c†α,k γαj e−iφ/2 + H.c. (4.8)
j=1 k α=0


with λiα ≡ λ̃α ξα,i and cα,k the respective lead fermions. According to (4.8), the zero mode
operators γα are tunnel coupled with amplitude λiα to the respective lead fermions cα,k .
i

Such a Hamiltonian is by now standard and has been considered e.g. in the following
Refs. [25, 26, 29, 28].
Hence, we model the three-terminal Majorana devices (see Figs. 4.2, 4.3, 4.4 and 4.5)
using the Hamiltonian
H = HLeads + HT + HC , (4.9)

with the charging term  2


HC = Ec N̂ − ng , (4.10)

and the backgate parameter ng . The metallic leads are described by the Hamiltonian
HLeads to be specified later. We assume that we are far away from the charge degeneracy
point. Throughout this chapter, the charging energy is assumed to be a large energy
scale. The significant size of the charging energy scale means that a low-energy electron
will be merely virtually present in the island and has to tunnel out again immediately.
As discussed in Chapter 2, the MBSs enable a highly nonlocal electron transfer pro-
cess where an electron tunneling into γαi tunnels out from γαj 0 while sustaining phase
coherence even when the pair of MBSs is separated by a sizable distance [29]. We
expand around the unique charge ground state |N0 i taking into account fluctuations to
the next higher charge states |Q0 ± 1i with energy of order of the charging energy EC .
Neglecting the frequency dependence of the self-energy, we can write a series expansion
of the effective Hamiltonian [96]

X (n)
Heff = Heff , (4.11)
n=1

68
CHAPTER 4. MAJORANA QUBIT DETECTION

with the n-th order term


 n−1
(n) 1
Heff = P0 HT HT P0 . (4.12)
−HC

Here P0 is the projection operator onto the charge ground state |N0 i. This proce-
dure yields the effective low-energy Hamiltonian (see e.g. Refs. [87, 88] for similar
derivations)
HC + HT → H̃T ≡ Heff . (4.13)

As shown in Appendix B, the leading order contribution to the effective tunneling in


our case is given by
1X X †
H̃T = Oαα0 cα,k cα0 ,k0 + H.c., (4.14)
2 α6=α0 k,k 0

where the operator Oαα0 is a linear combination of Majorana bilinears

Nα0
Nα X
X 0 0
Oαα0 = tjj j j
1,αα0 (iγα γα0 ). (4.15)
j=1 j 0 =1

Here we have introduced the cotunneling amplitudes


0
0 2iλjα (λjα0 )∗
tjj
1,αα0 ' . (4.16)
EC

The next higher terms contributing to the Hamiltonian in the expansion (4.11) are of
−2 j 3
order O EC |λα | and thus strongly suppressed in the relevant limit of weak tunnel-
ings and large charging energy. We further note that from the definition (4.15) we read
0 jj 0
off that tjj ∗
1,αα0 = −(t1,α0 α ) and therefore

Oαα0 = Oα† 0 α . (4.17)

This operator will play the role of a jump operator in a quantum master equation later
in this chapter.

69
4.2. FORMALISM FOR WEAK MEASUREMENT PROTOCOLS

4.3 Formalism for weak measurement protocols ap-


plied to Majorana devices

In this subsection, we describe an efficient formalism to access the signatures of electron


transport in the three-terminal Majorana devices introduced in subsection 4.1, see Figs.
4.2, 4.3 and 4.4. To this end we start by introducing some necessary background on the
methodology of full counting statistics (FCS). In particular, we derive a variant of the
Liouville-von Neumann equation, which is modified by counting fields [27]. Based on
this equation as well as on the effective Hamiltonian derived in Subsection 4.2, we go
on to derive quantum master equations augmented with counting parameters. Finally,
we take a Markovian limit and explicitly state the resulting equation for the conditions
of practical importance for us, namely the limit in which temperature is low compared
to the applied bias voltage V .

4.3.1 Method of full counting statistics

The experimental scenario introduced in section 4.1 realizes a three-terminal structure.


We argued that the crucial information regarding MBS detection is contained in the
noise, or more generally the statistics of electron transport. To prepare for the de-
tailed analysis of this statistics, we introduce the necessary elements of the theory of
full counting statistics (FCS). For a general M -terminal structure we introduce the
probability [97]
Pτ (Q) ≡ Pτ (Q1 , . . . , QM ) (4.18)

that Qα electrons transfer to terminal α ∈ {1, . . . , M } during the time interval [0, τ ].
Hence, a negative value of Qα implies that charges have left the terminal α during this
time interval. The probability distribution can be written as the Fourier transformation
ˆ π ˆ π
dχ1 dχM
Pτ (Q) = ... Z(χ, τ )e−iχQ (4.19)
−π 2π −π 2π

of the characteristic function Z(χ, τ ) [27, 98], also called moment generating function,
which is the central quantity of interest to us. For the Fourier variables we introduced
the shorthand notation χ ≡ (χ1 , . . . , χM )T with the parameter χα counting electrons
in lead α. Moreover, we have defined Q ≡ (Q1 , . . . , QM )T . Now let ρ(t) be the density

70
CHAPTER 4. MAJORANA QUBIT DETECTION

operator of the system which is governed by the Liouville-von Neumann equation [24]


ρ(t) = −i [H, ρ(t)] , (4.20)
∂t

where H is the Hamiltonian. Quantum mechanically, the probability P τ (Q) can be



expressed using the projection postulate which states that Pτ (Q) = Tr ρ(t)P̂ (Q) ,
where P̂τ (Q) is the projection operator onto the subspace with eigenvalues {Q1 , Q2 , . . .}.
This can be rewritten in the following way [98]

∞ ∞
!
X X
Pτ (Q) = Tr ρ(t) ... δn1 Q1 . . . δnM QM P̂ (n)
n1 =0 nM =0
ˆ π ˆ π ∞ ∞
!
dχ1 dχM −iχQ X X
= ... e Tr ρ(t) ... eiχn P̂ (n)
−π 2π −π 2π n1 =0 nM =0
ˆ π ˆ π
dχ1 dχM −iχQ  
= ... e Tr ρ(t)eiχQ̂ , (4.21)
−π 2π −π 2π

with n ≡ (n1 , . . . , nM )T . In the third equality we have used the spectral represen-
P∞ P∞
tation eiχQ̂ = n1 =0 . . . nM =0 e
iχn
P̂ (n). We assume ρ(t = 0) ≡ ρ0 to be the ini-
tial time density operator describing a state where each terminal α is in a definite
charge state Qα,0 . The operators Q̂α count the net difference of electrons added and
hence satisfy Q̂α ρ(0) = 0. The moment generating function, which is inferred to be
Z(χ, t) = Tr(ρ(t)eiχQ̂ ) by comparing (4.21) and (4.19), in this case yields the statistics
of transported electrons during the time interval [0, τ ] [99]. We now use the cyclic in-
variance of the trace, the fact that ρ(0) = e−iχQ̂ ρ(0) and [Q̂α , ρ(0)] = 0, to rewrite the
moment generating function in a more symmetric form [98]:
 
Z(χ, t) = Tr U (t)ρ(0)U † (t)eiχQ̂
 i 
χQ̂ − 2i χQ̂ − 2i χQ̂ † i
χQ̂
= Tr e 2 U (t)e ρ(0)e U (t)e 2

 

≡ Tr Uχ (t)ρ(0)U−χ (t) . (4.22)

Here we defined the time evolution operator [27, 98]

ˆt
 
i i i
Uχ (t) ≡ e 2 χQ̂ U (t)e− 2 χQ̂ = exp − dsHχ (s) (4.23)
~
0

71
4.3. FORMALISM FOR WEAK MEASUREMENT PROTOCOLS

with the Hamiltonian


i i
Hχ ≡ e 2 χQ̂ He− 2 χQ̂ (4.24)

modified by the counting parameters. Note that Eq. (4.22) implies that the forward and
backward time evolution is generated by a different Hamiltonian. We also mention in
passing that Hχ is Hermitian. We now define a generalized density operator according
to

ρ(χ, t) = Uχ (t)ρ(0)U−χ (t). (4.25)

Unlike for a real density operator its trace is not unity, but in fact yields the moment
generating function, i.e. Z(χ, τ ) = Tr (ρ(χ, τ )). Moreover, its time evolution obeys the
following modified Liouville-von Neumann equation [27, 98],


ρ(χ, t) = −iHχ ρ(χ, t) + iρ(χ, t)H−χ . (4.26)
∂t
The equation is solved with the initial condition ρ(χ, 0) = ρ(0). The generating function
of irreducible moments is given by the logarithm of the moment generating function
evaluated at the final time τ after the measurement, F(χ, τ ) = ln Tr (ρ(χ, τ )). By
differentiation with respect to the counting fields all cumulants (irreducible moments)
of the charge Qα in lead α = 1, 2 can be obtained via
 n  m
n m ∂ ∂
hh(Qα ) (Qα0 ) . . .ii = −i −i . . . ln Tr (ρ(χ, τ )) . (4.27)
∂χα ∂χα0 χ=(0,0,...)

P
The fact that the probability adds to unity {Qα } Pτ (Q) = 1 implies that F(χ =
0, τ ) = 0. Moreover, the cumulant generating function generally is 2π periodic in each
χα [97],
F(. . . , χα + 2π, . . . , τ ) = F(. . . , χα , . . . , τ ). (4.28)

This is related to the fact that the transferred charge physically is an integer multiple of
the charge of the electron. The conservation of charge now implies that the moment and
the cumulant generating function only depend on counting field differences [97]. More
concretely, the difference χα − χα0 is associated with the charge transfer between the
terminals α and α0 . Therefore, there is a redundancy in the counting fields χ1 , . . . , χM
and we may choose a gauge with χM ≡ 0 and keep M − 1 independent counting fields
which we denote by χ ≡ (χ1 , . . . , χM −1 ) [97].

72
CHAPTER 4. MAJORANA QUBIT DETECTION

4.3.2 Modified Liouville-von Neumann equation

As argued in section 4.1 the correlations of the currents coupled to the Majorana Pauli
algebra contain considerably more information than is contained in simple current aver-
ages. To formalize this we now draw on our discussion of full counting statistics in the
previous subsection 4.3.1 and apply it to our present problem. The setup introduced in
subsection 4.1 is a three-terminal structure that can be described using two counting
fields χ ≡ (χ1 , χ2 ) by setting χ0 ≡ 0. We want to work with the generalized density
matrix governed by the modified Liouville-von Neumann equation in the interaction
picture. To this end we split the effective Hamiltonian of the full system including the
interference links into two contributions

H = H0 + HI . (4.29)

The unperturbed part is the Hamiltonian,

2 X
X
H0 = (ξk − V δα0 )c†α,k cα,k (4.30)
α=0 k

k 2
describing metallic leads and has a continuum of states with dispersion ξk = 2m . Here
m is an effective mass and k the momentum quantum number. The voltage bias V in
the central lead α = 0 is applied such that it acts as the source of electrons that are
transmitted either to the leads α = 1 or α = 2. The interaction part of the Hamiltonian
is given by
HI = Href + H̃T , (4.31)

with H̃T as defined in Eq. (4.14) and Href the interference link Hamiltonian defined as

2 X
X
Href = (t0,α c†α,k c0,k0 + H.c.). (4.32)
α=1 k,k0

Thus, the gate tunable tunnel couplings t0 introduce a direct link between the source
and drain leads. The modified Liouville-von Neumann equation [27] discussed in the
previous subsection 4.3.1 now stated for the interaction picture density matrix ρI (χ, t)
takes on the form

73
4.3. FORMALISM FOR WEAK MEASUREMENT PROTOCOLS


ρI (χ, t) = −iHI,χ (t)ρI (χ, t) + iρI (χ, t)HI,−χ (t) (4.33)
∂t
and is solved with the initial condition ρ(χ, 0) = ρ(0). With the time evolution operator
U0 (t) = e−itH0 generated by the lead Hamiltonian Eq. (4.30) we define the interaction
picture operators
HI,χ (t) = U0† (t)HI,χ U0 (t). (4.34)

The Hamiltonian modified with the counting fields (4.24) is given by

1 X i (χα −χα0 ) X †
HI,χ = e2 (t0,α δ0α0 + Oαα0 ) cα,k cα0 ,k0 + H.c. (4.35)
2 α6=α0 k,k 0

where the operators Oαα0 were defined in Eq. (4.15).

4.3.3 Derivation of the quantum master equation

We use the quantum master equation to study the electronic transport through the
weakly coupled mesoscopic Majorana island with Coulomb blockade. In this section,
we will provide a detailed weak coupling limit derivation of the Bloch Redfield equation,
which we use to model the Majorana devices and from which we obtain the full counting
statistics (FCS). The microscopic derivation of quantum master equations in the weak
coupling limit is a well established method in the theory of open quantum systems
[24, 100]. In this subsection, we will apply this method. In our context, the role of
the environment, which will be traced out, is played by the weakly coupled fermionic
reservoirs. The starting point is the modified Liouville-von Neumann equation (4.33)
from the previous section. We continue to work in the interaction picture and reiterate
the equation in its integrated form

ˆt
ρ̃total (χ, t) = ρ̃total (0) + dsLs [ρ̃total (χ, s)], (4.36)
0

where ρ̃total is the interaction picture density matrix (we omit the index I) of the
total system. The total system consists of the Majorana island, the metallic leads
and the tunnel couplings between the two. In Eq. (4.36) we have defined a Liouville-

74
CHAPTER 4. MAJORANA QUBIT DETECTION

superoperator by its action on an arbitrary density operator ρ according to

Lt [ρ] := −iHI,χ (t)ρ + iρHI,−χ (t). (4.37)

By plugging this relation into Eq. (4.33) we obtain

ˆt

ρ̃total (χ, t) = Lt [ρ̃total (0)] + dsLt [Ls [ρ̃total (χ, s)]] , (4.38)
∂t
0

which governs the time evolution of the density operator of the full system ρ̃total (χ, t).
We define the generalized reduced density matrix in the low-energy subspace defined
by the Majorana operators γαi via the partial trace TrL over the fermionic reservoirs

ρ(χ, t) := TrL (ρ̃total (χ, t)) . (4.39)

The integro-differential evolution equation obeyed by the reduced density operator reads

ˆt

ρ(χ, t) = TrL (Lt [ρ̃total (0)]) + dsTrL (Lt [Ls [ρ̃total (χ, s)]]) . (4.40)
∂t
0

This exact equation can be treated within the framework of perturbation theory. In
the case of weak coupling between the quantum system and the fermionic reservoirs,
the Born approximation of a factorized total density matrix

ρ̃total (χ, s) = ρ(χ, s) ⊗ ρL (4.41)

(for s > 0) is permissible. This applies to our setting since we envision the reservoirs
to be weakly coupled to the mesoscopic superconducting island. The master equation
in Born approximation is then given by

ˆt

ρ(χ, t) = TrL (Lt [ρ(0) ⊗ ρL ]) + dsTrL (Lt [Ls [ρ(χ, s) ⊗ ρL ]]) . (4.42)
∂t
0

Physically, the Born approximation is well justified if the back action of the system
on the fermionic reservoirs is negligible such that the latter effectively remain in the
state ρL . Naively, it may seem a bit contradictory to neglect the back action on the

75
4.3. FORMALISM FOR WEAK MEASUREMENT PROTOCOLS

fermionic reservoir, given that we are interested in the changes of particle number in
those reservoirs. However, it is valid since Eq. (4.42) does not describe the fermionic
reservoirs but the reduced density matrix in the low-energy subspace of the mesoscopic
island.
We now apply a Markovian approximation which involves two steps. First, we assume
that the density matrix varies on a slower time scale than the typical decay time of the
correlation functions of the fermionic leads. This assumption justifies the replacement

ρ(χ, s) → ρ(χ, t), (4.43)

which makes the master equation time-local. For χ = (0, 0) the equation preserves
Hermiticity as well as unit trace of the density matrix. The coefficients still exhibit a
time dependence. Thus, we substitute s → t − s and extended the time integration
to infinity. This is justified by the assumption that the correlation functions of the
fermionic reservoirs decay quickly. We arrive at the Markovian master equation

ˆ∞

ρ(χ, t) = TrL (Lt [ρ(0) ⊗ ρL ]) + dsTrL (Lt [Lt−s [ρ(χ, t) ⊗ ρL ]]) , (4.44)
∂t
0

where according to (4.37) the operator on the right hand side reads

Lt [Lt−s [ρ]] = HI,χ (t)ρHI,−χ (t − s) + HI,χ (t − s)ρHI,−χ (t)


−HI,χ (t)HI,χ (t − s)ρ − ρHI,−χ (t − s)HI,−χ (t). (4.45)

The next step is to trace out the fermionic reservoirs. We approximate the latter to be
in thermal equilibrium,
1
ρL ∼ e− T H0 (4.46)

with respect to the lead Hamiltonian (4.30) since the coupling due to Href and H̃T is
weak. One finds that the following term vanishes, TrL (Lt [ρ(0) ⊗ ρL ]) = 0. Hence, we
arrive at the equation

ˆ∞

ρ(χ, t) = dsTrL (Lt [Lt−s [ρ(χ, t) ⊗ ρL ]]) . (4.47)
∂t
0

76
CHAPTER 4. MAJORANA QUBIT DETECTION

Using Eq. (4.14), we write the integrand on the right hand side in the following form:

2
X
TrL (Lt [Lt−s [ρ(χ, t) ⊗ ρL ]]) = L̃α,s [ρ(χ, t)] . (4.48)
α=0

With the shorthand notation


X
Qαα0 = c†α,k cα0 ,k0 (4.49)
k,k0

the Liouvillians L̃α,s [ρ] are given by


   
† †
L̃0,s [ρ] = ei(χ2 −χ1 ) O12 ρO12 − O12 O12 ρ TrL ρL Q12,t Q†12,t−s (4.50)
   
† †
+ ei(χ2 −χ1 ) O12 ρO12 − ρO12 O12 TrL ρL Q12,t−s Q†12,t
   
† †
+ e−i(χ2 −χ1 ) O12 ρO12 − ρO12 O12 TrL ρL Q†12,t−s Q12,t
   
−i(χ2 −χ1 ) † † †
+ e O12 ρO12 − O12 O12 ρ TrL ρL Q12,t Q12,t−s

and

 

L̃α=1,2,s [ρ] = t0,α [Oα0 , ρ]TrL ρL Qα0,t−s Q†α0,t − ρL Q†α0,t Qα0,t−s
 
+t∗0,α [Oα0 , ρ]TrL ρL Q†α0,t−s Qα0,t − ρL Qα0,t Q†α0,t−s
   
−iχα † † †
+ e Oα0 ρOα0 − Oα0 Oα0 ρ TrL ρL Qα0,t Qα0,t−s
   
† †
+ e−iχα Oα0 ρOα0 − ρOα0 Oα0 TrL ρL Qα0,t−s Q†α0,t
   
† † †
+ eiχα Oα0 ρOα0 − ρOα0 Oα0 TrL ρL Oα0,t−s Qα0,t
   
† † †
+ eiχα Oα0 ρOα0 − Oα0 Oα0 ρ TrL ρL Oα0,t Qα0,t−s
h i

+(e−iχα − 1) t∗0,α ρOα0 + t0,α Oα0 ρ + |t0,α |2 ρ
 
† †
×TrL ρL Qα0,t−s Qα0,t + ρL Qα0,t Qα0,t−s
h i

+(eiχα − 1) t∗0,α Oα0 ρ + t0,α ρOα0 + |t0,α |2 ρ
 
×TrL ρL Q†α0,t Qα0,t−s + ρL Q†α0,t−s Qα0,t . (4.51)

77
4.3. FORMALISM FOR WEAK MEASUREMENT PROTOCOLS

Furthermore, we make the assumption that the bandwidth of the reservoirs Λ is large
compared to the energy scales V and T . Exemplarily, this yields relations that can be
obtained by standard techniques [101] like e.g.

ˆ∞ D E
ds Q†α0,t−s Qα0,t − Qα0,t Q†α0,t−s = ν 2 (iΛ + πV ), (4.52)
0

where Λ is the bandwidth of the leads and ν the density of states and h. . .i ≡ TrL (ρL . . .).
Analogously, we obtain, for instance,

ˆ∞ D E
ds Qα0,t−s Q†α0,t + Qα0,t Q†α0,t−s = 2πν 2 V nB (V ), (4.53)
0

where nB (V ) is the Bose function which is negligible in the limit T  V . The terminals
α = 1, 2 are at the same potential such that thermal cotunneling processes are relevant
here,
ˆ∞ D E
ds Q12,t−s Q†12,t + Q12,t Q†12,t−s = 2πν 2 T. (4.54)
0

Further technical details concerning the tracing out of the leads are found in Appendix
C. We now go on to discuss the master equation, which emerges from the outlined
procedure in the next subsection.

78
CHAPTER 4. MAJORANA QUBIT DETECTION

4.3.4 Resulting quantum master equation for weak measure-


ment setup

The resulting quantum master equation is a Lindblad equation modified with counting
parameters. It is valid for arbitrary number of Majoranas Nα coupled to reservoir α.
The equation governs the time evolution of the reduced density matrix ρt in the Hilbert
space corresponding to the Majorana operators γαi . Without the assumption T  V ,
it reads
2
∂ 2
X
iχα

∗ † 2

ρt = −i[Hq , ρt ] + 2πν V
e (e − 1) t0,α Oα0 ρt + t0,α ρt Oα0 + |t0,α | ρt
∂t α=1
2
X
2 2
+2πν T (D12 (ρt ) + D21 (ρt )) + 2πν V nB (V ) D0α (ρt )
α=1
2
X
2
+2πν V (1 + nB (V )) Dα0 (ρt ). (4.55)
α=1

Here we have defined the superoperator

† 1 †
Dαα0 (ρ) ≡ ei(χα −χα0 ) Oαα0 ρOαα 0 − {O 0 Oαα0 , ρ}. (4.56)
2 αα

The jump operator Oαα0 is again defined as

Nα0
Nα X
X 0 0
Oαα0 = tjj j j
1,αα0 (iγα γα0 ), (4.57)
j=1 j 0 =1

and describes the tunneling of an electron from the terminal α0 to α. The coherent part
of the evolution in Eq. (4.55) is described by a Hamiltonian H e q which we will specify
below in the limit T  V .
Throughout this chapter, we mostly focus on the low temperature limit T  V  Λ,
with Λ the lead bandwidth. In this limit, the Bloch Redfield equation (4.55) takes on
the form
2
∂ 2
X
iχα

∗ † 2

ρt = −i[Hq , ρt ] + 2πν V (e − 1) t0,α Oα0 ρt + t0,α ρt Oα0 + |t0,α | ρt
∂t α=1
2
X
2 2
+2πν T (D12 (ρt ) + D21 (ρt )) + 2πν V Dα0 (ρt ), (4.58)
α=1

79
4.3. FORMALISM FOR WEAK MEASUREMENT PROTOCOLS

where the Hamiltonian part of the dynamics in the limit V  Λ (see Appendix C) is
generated by

2
X  ν 2Λ X †
Hq = −Λν 2
t∗0,α Oα0 + H.c. − {O , Oαβ }
α=1
2 α<β αβ
2
X †
+ν 2 V ln(Λ/2V ) [Oα0 , Oα0 ]. (4.59)
α=1

The Eq. (4.58) is of the Lindblad form but modified with counting parameters. The
equation is time-local and for zero counting parameters it preserves Hermiticity and unit
trace. Similar equations have been studied in Refs. [94, 102, 103]. We are interested
in obtaining the statistics of charge transported during the time interval t ∈ [0, τ ].
To this end, we solve (4.58) with the initial condition ρ(χ, 0) = ρ0 and obtain the
generating function of the charge variables [27, 98] by performing the trace in the
reduced generalized qubit space Z(χ, τ ) = Trq (ρτ (χ1 , χ2 )). The equation is valid in
0
the weak coupling limit defined by νt0,α , νtjj
1,αα0  1. Its Markovian property implies
that we are neglecting memory effects.

80
CHAPTER 4. MAJORANA QUBIT DETECTION

4.4 Phenomenology of the Majorana box qubit

In this section, we present the resulting weak measurement phenomenology of the Ma-
jorana box qubit. We start by providing instructive details on the derivation of the
generating function in Subsection 4.4.1. In Subsection 4.4.2, we analyze the obtained
current statistics and describe a remarkable effect of strong current cross-correlations.
We present results on the effects of finite temperature in Subsection 4.4.3 and then
go on to study the outcome distribution and the statistics of rare current outcomes in
Subsection 4.4.4. Finally, we consider finite hybridizations of MBSs on the box qubit
in Subsection 4.4.5.
This entire section is founded on the evolution equation for genuine MBSs governing
the time evolution of the qubit density matrix ρt ≡ ρt (χ1 , χ2 ). To obtain the equation,
we have to set Oαα0 = t1,αα0 (iγα γα0 ) in Eq. (4.58). Where possible, we use the simplified
notation t0,α = t0 and t1,αα0 = t1 throughout this chapter. The resulting equation reads

∂ X Γαα0
ρt = −i[H̃q , ρt ] + (zαα0 (iγα γα0 )ρt (iγα γα0 ) − ρt )
∂t α>α0
2
2
X
+2πV (eiχα − 1) (t∗0 t1 (iγα γ0 )ρt + t0 t∗1 ρt (iγα γ0 )) (4.60)
α=1

with the rates defined as Γα0 = 4πV ν 2 |t1 |2 , Γ21 = 4πν 2 T |t1 |2 and with zα0 = eiχα ,
z21 = 2 cos(χ1 − χ2 ). The Hamiltonian evolution is generated by

2
X
H̃q = −2ν 2
(ΛRe(t∗0 t1 ) + πV Im(t∗0 t1 )) (iγα γ0 ). (4.61)
α=1

Notice that the splitting (4.61) of the qubit energy is due to the presence of the interfer-
ence paths. The splitting term leads to precession physics and will play an important
role for the counting statistics.

81
4.4. PHENOMENOLOGY OF THE MAJORANA BOX QUBIT

4.4.1 Derivation of the generating function for the Majorana


qubit

To solve the evolution equation (4.60) and derive the generating function, we
reparametrize the time variable, ρt = eθt ρ̃t , with

2
X
2 2 2
θ = 2πν (|t0 | + |t1 | )V (eiχα − 1)
α=1
+4πν 2 |t1 |2 T (cos(χ1 − χ2 ) − 1). (4.62)

Furthermore, we define the rates as Γ̃α0 = 4πV ν 2 |t1 |2 eiχα and Γ̃21 = 8πν 2 T |t1 |2 cos(χ1 −
χ2 ). To solve the Bloch Redfield equation, we parametrize the density matrix as

3
X
ρt = ρµ,t σµ . (4.63)
µ=0

Thus, we obtain a first order system of four coupled equations for ρ̃ given by

3
∂ X
ρ̃µ,t (χ1 , χ2 ) = Ωµν ρ̃ν,t (χ1 , χ2 ). (4.64)
∂t µ=0

The matrix Ω of coefficients reads


 
0 a1 a2 0
 
 a1 −Γ̃20 − Γ̃21 0 h2 
Ω=  (4.65)
 a
 2 0 −Γ̃10 − Γ̃21 −h1 

0 −h2 h1 −Γ̃10 − Γ̃20

with
aα = 4πV ν 2 Re(t∗0 t1 )(zα − 1), (4.66)

hα = −4ν 2 ΛRe(t∗0 t1 ) − 4πν 2 V Im(t∗0 t1 )eiχα . (4.67)

The Eq. (4.60) is solved with the initial condition ρ(χ, 0) = ρ0 where ρ0 is the initial
reduced density matrix of the qubit [27, 94, 98]. Moreover, we assume the counting
fields χ1 and χ2 to be constant in the time interval (0, τ ) of measurement. The solution

82
CHAPTER 4. MAJORANA QUBIT DETECTION

is given by a matrix exponential

3
X
θτ
ρµ,τ (χ1 , χ2 ) = e exp(τ Ω)µν ρν,0 , (4.68)
ν=0

where ρ0 is the initial reduced density matrix of the qubit and thus in particular ρ0,0 = 12 .
We obtain the statistics of the currents by considering the logarithm of the generating
function
ln Z(χ, τ ) ≡ ln Trq (ρτ (χ1 , χ2 )) . (4.69)

In the long time limit Γ τ  1 the cumulant generating function reads

ln Z(χ) = τ θ(χ1 , χ2 ) + τ λ0 (χ1 , χ2 ). (4.70)

Here θ is defined as in (4.62) and λ0 (χ1 , χ2 ) is the (unique) solution of the characteristic
polynomial of Ω

λ4 + c1 (χ)λ3 + c2 (χ)λ2 + c3 (χ)λ + c4 (χ) = 0, (4.71)

which satisfies
λ0 (0, 0) = 0. (4.72)

Here the ci (χ) are counting parameter dependent coefficients. Such a solution always
exists because c4 (χ1 = 0, χ2 = 0) = 0. The other eigenvalues λi=1,2,3 of Ω have a
negative real part at (χ1 , χ2 ) = (0, 0). In the limit T  V  Λ, we expand Eq.
(4.70) to first order in V to obtain the counting statistics. This results in the cumulant
generating function
 s 
2
Γτ  2|t0 | + |t1 |2
8 [Re(t∗0 t1 )]2 2 
ln Trρτ (χ1 , χ2 ) = z−1+ (z + 1)2 + z (4.73)
2 |t1 |2 |t1 |4

with
1 1
z ≡ eiχ1 + eiχ2 − 1. (4.74)
2 2

83
4.4. PHENOMENOLOGY OF THE MAJORANA BOX QUBIT

The two physical observables can now be obtained by taking derivatives according to
Eq. (4.27). E.g. the current α = 1, 2 is calculated via

−1 ∂
hIα i = −iτ ln Trq (ρτ (χ1 , χ2 )) . (4.75)
∂χα (χ1 ,χ2 )=(0,0)

The cross-correlations are obtained by means of the relation

∂2

−1

S12 = − τ ln Trq (ρτ (χ1 , χ2 )) . (4.76)
∂χ1 ∂χ2 (χ1 ,χ2 )=(0,0)

4.4.2 Pronounced shot noise cross-correlations and qubit evo-


lution

In this subsection, we present the results for the shot noise as well as the evolution of
the Majorana qubit density matrix. In particular, we discuss below that positive cross-
correlations S12 = F hIi with a Fano factor F = O(1) characterize the Majorana system
in the limit of low temperature T  V and weak tunnel couplings. First, however, it
is insightful to take a closer look at the dynamics of the qubit density matrix in the
absence of the counting parameters. Completely independent of the initial qubit state,
the qubit density matrix ρt approaches the maximally mixed state

1
ρtΓ−1 ' σ0 (4.77)
2

on the time scale Γ−1 with Γ ≡ 4πν 2 |t1 |2 V . We take note that this time scale is set
by voltage and the dimensionless conductance contribution due to cotunneling. The
associated loss of quantum information regarding the qubit state is reflected in the
averages of the currents passing through the outer barriers as well. In fact, on time
scales larger than Γ−1 , these averages (in units of e2 /~)

hIα i = hIi ≡ 2πν 2 V (|t0 |2 + |t1 |2 ) (4.78)

do not depend on the qubit state anymore. This is a striking feature since it is in sharp
contrast to the individual readout of a single current Iα (see Eq. (4.2)), which couples to
the parity iγα γ0 [11] if the other drain terminal is completely decoupled. Nevertheless,
it is expected for the simultaneous measurement of noncommuting operators that the

84
CHAPTER 4. MAJORANA QUBIT DETECTION

quantum information of the initial state of the qubit gets wiped out. Now we turn our
´t
attention to the study of the second cumulant Sαα0 = 0 dt hhIα (t)Iα0 (0)ii for α, α0 ∈
{1, 2} and hhABii = hABi − hAi hBi, in the long time limit Γτ  1, and under the
assumption T  V  Λ. The shot noise in the current signals α = 1, 2 can then be
derived to be

4πV ν 2 [Re(t∗0 t1 )]2


Sαα = hIα i + . (4.79)
|t1 |2

The first term is the standard Schottky contribution with hIα i ≡ 2πν 2 V (|t0 |2 + |t1 |2 )
being the average current as discussed above. The second term shows that the noise
level is enhanced whenever the currents I1 and I2 couple to the underlying Pauli algebra,
because Re(t∗0 t1 ) 6= 0 is precisely the condition for Pauli dependent current averages, see
Eq. (4.2). The most noteworthy observable signature is the current cross-correlation
amplitude S12 , which is given by

4πV ν 2 [Re(t∗0 t1 )]2


S12 = . (4.80)
|t1 |2

We point out that this result is derived under the assumption that t1 6= 01 . We may
conveniently reexpress Eq. (4.80) by defining the Fano factor F which characterizes the
correlations via the relation
S12 = F hIi , (4.81)

with I¯ as given in (4.78). The corresponding Fano factor again for T  V is then given
by
2 [Re(t∗0 t1 )]2
F = = O(1). (4.82)
|t1 |2 (|t0 |2 + |t1 |2 )
This result is interesting for a multitude of reasons, one of them being that the cur-
rents are positively correlated. The most important aspect of this result, however, is
the largeness of the shot noise correlations, F = O(1). To appreciate this point, it
is instructive to draw the contrast with noninteracting electrons in a T -shaped junc-
tion [104]. In this case, negative correlations are observed due to processes involving
two electrons and the Fano factor is parametrically suppressed in the dimensionless
tunnel conductances2 . This makes it clear that rather different physical mechanisms
1
The maximally mixed state 12 σ0 is reached after the timescale Γ−1 . Hence, for t1 = 0 this state is
never reached and the contribution (4.80) does not arise.
2
Consider e.g. three leads with a central lead α = 0 and drain leads α = 1, 2 tunnel coupled

85
4.4. PHENOMENOLOGY OF THE MAJORANA BOX QUBIT

must be active to give rise to the striking result (4.82). Indeed, the competing weak
measurements of the Pauli operators σ1 = iγ1 γ0 and σ2 = iγ2 γ0 cause a frustration in
the system. The large positive cross-correlations in the two incompatible currents I1
and I2 coupled to the underlying Pauli algebra are the observable consequence of this
frustration. Consistent with this interpretation is the fact that the cross-correlation
contribution (4.82) vanishes only for Re(t∗0 t1 ) = 0, in which case the currents do not
measure the Pauli operators on average. Since Re(t∗0 t1 ) = 0 implies a considerable
amount of fine tuning generically one will observe F (V  T ) = O(1). The fact that in
the absence of the reference links, t0 = 0, the Fano factor is parametrically suppressed,
i.e. F  1, will turn out to be useful for the formulation of measurement protocols.
A complementary physical interpretation of (4.82) is that an electron incoming from
the source which contributes to I1 helps to project iγ1 γ0 onto a definite state. This,
in turn, introduces more uncertainty in the expectation value of iγ2 γ0 and, therefore,
affects the current I2 . This physical picture conveys how an electron can exert influence
on a subsequent electron that might be incoming much later.
according to Href = α=1 k,k0 (t0 c†α,k c0,k0 + H.c.). With a bias V applied to operate lead 0 as source,
P2 P

the average current is given by hIα i = 2πνα ν0 |t0 |2 V with α = 1, 2. In a symmetric situation where
each lead has the density of states ν, the cross-correlations are given by Fnon−interacting = −ν 2 |t0 |2
[104] and thus |Fnon−interacting |  1. The minus sign is a result of the fact that a single electron is
either transmitted to lead 1 or 2.

86
CHAPTER 4. MAJORANA QUBIT DETECTION

4.4.3 The effects of finite temperature

For the most part of this chapter we neglect finite temperature contributions assuming
that temperature is much smaller than the applied bias. For completeness, we now take
a look at corrections due to a small3 finite temperature T > V in the all-important result
(4.82) for the Fano factor F . As before, we still assume V, T  EC , ∆. Physically, the
corrections originate in direct cotunneling processes back and fourth from the two drain
terminals 1 to 2 with the amplitude t1,12 . For clarity, we distinguish this parameter from
the cotunneling amplitudes from the source to the drains, which we again assume to be
equal, t1,10 = t1,20 . The finite temperature corrected current shot noise (4.79) is given
by
2 2 4πV 2 ν 2 [Re(t∗0 t1,10 )]2
Sαα = hIα i + 4πν |t1,12 | T + . (4.83)
V |t1,10 |2 + 2T |t1,12 |2
The current cross-correlation formula in the long time limit is given as a sum of two
contributions
4πV 2 ν 2 [Re(t∗0 t1,10 )]2
S12 = −4πν 2 |t1,12 |2 T + . (4.84)
V |t1,10 |2 + 2T |t1,12 |2
The first term is negative and proportional to the temperature T . This term originates
in the above mentioned processes where electrons thermally cotunnel via the Majorana
states on the box from lead 1 to lead 2 or vice versa. The second term can be identified
with the finite temperature version of the formula (4.82). It shows that finite temper-
ature has the effect of weakening the positive F = O(1) Fano factor, but confirms that
in the regime of interest T  V the positive second term in Eq. (4.84) is the dominant
contribution and results in pronounced positive cross-correlations.
3
To be precise we keep terms linear in temperature in the Bloch Redfield equation, but approximate
coth(V /2T ) ' 1.

87
4.4. PHENOMENOLOGY OF THE MAJORANA BOX QUBIT

4.4.4 Outcome distribution and extreme value statistics

In this section we examine the joint probability distribution P (I1 , I2 ) of the current
outcomes. In particular, we address the statistics of extreme values and show that there
are large fluctuations in the currents. Furthermore, we will plot the joint distribution
for several parameters. The probability for the current values (I1 , I2 ) is given as the
Fourier transformation of the generating function

ˆπ ˆπ
dχ1 dχ2
P (I1 , I2 ) = Trq (ρτ (χ1 , χ2 )) e−i(χ1 I1 +χ2 I2 )τ . (4.85)
2π 2π
−π −π

In the limit of a measurement duration τ  Γ −1 , where again Γ ≡ 4πν 2 |t1 |2 V , we


can make a saddle point approximation and write the logarithm of the probability
distribution in the form [89]
" 2
#
ln P (I1 , I2 ) ln Trq (ρτ (iµ1 , iµ2 )) 1 2 X
¯ .
' min + |t0 /t1 | + 1 µα (Iα /I) (4.86)
Γτ {µα } Γτ 2 α=1

According to (4.73), we here have


s
ln Trq (ρτ (iµ1 , iµ2 )) 1 8 [Re(t∗0 t1 )]2 −µ1
= (e−µ1 + e−µ2 )2 + (e + e−µ2 − 2)2
Γτ 4 |t1 |4
 
1 2 1
(e−µ1 + e−µ2 ) − |t0 /t1 |2 + 1 . (4.87)

+ |t0 /t1 | +
2 2

For simplicity of discussion, we again choose system parameters such that I1 and I2 have
the same average value I¯ = 2πν 2 V (|t0 |2 +|t1 |2 ) for long times Γ τ  1. According to the
relation (4.86), the probability distribution is found by minimization. For atypically
large current outcomes Iα  I, ¯ this minimization can be done analytically. In fact, the
¯ α and as a result, the probability for these rare

minimum is located at µα ' ln I/I
events is given by [89]

2  ¯ Iα τ
Iα I¯
Y I
P (I1 , I2 ) ' . (4.88)
α=1

We thus see that that there are large fluctuations and rare events occur more often than
a Gaussian model would predict. The formula (4.88) derived under the assumption

88
CHAPTER 4. MAJORANA QUBIT DETECTION

(Γτ)-1ln P(I1,I2) 0.0


-0.5
-1.0
I2 = 1 I
-1.5
I2 = 2 I
-2.0
I2 = 3 I
-2.5
0 1 2 3 4
I1 / I
Figure 4.6: Logarithm of the outcome distribution for the Majorana box qubit in the
long time limit Γτ  1 as a function of I1 /I¯ for several the fixed values of I2 . We choose
¯ the maximum is
the parameters |t0 /t1 |2 = 1 and Re(t∗0 t1 ) = |t0 t1 |. For fixed I2 = I,
located at I1 = I¯ as expected. For fixed I2 > I,
¯ the maximum shifts to I1 > I¯ consistent
with the predicted strong positive cross-correlations stated in Eq. (4.82)

Iα  I¯ factorizes, which means that the distributions are independent in this case.
In general, the two tunnel current outcomes are strongly conditioned to each other,
because of the simultaneous incompatible measurement. This is visualized in Fig. 4.6
where the probability distribution is plotted based on Eq. (4.86) as a function of I1
for several fixed values of I2 . There it can be seen that for fixed I2 > I¯ the maximum
of the distribution as a function of I1 shifts to a value bigger than its average as well.
This is consistent with the positive cross-correlations, see Eq. (4.82). In contrast,
at the fine tuned point Re(t∗0 t1 ) = 0 the probability distribution (4.86) factorizes in
fact for all values of the currents I1 and I2 . Accordingly, all maxima are located at
I1 /I¯ = 1 independent of the value of I2 . This is because in this case the current Iα does
not couple to σα on the average and the strong cross-correlation effect (4.82) vanishes.
Finally, in Fig. 4.7 we plot the distribution Eq. (4.86) for a larger range of outcomes to
illustrate the probability distribution for rare events. This plot is consistent with the
formula (4.88) and also goes beyond it by showing the full regime in which e.g. only I1
is large and I2 = I¯ fixed.

89
4.4. PHENOMENOLOGY OF THE MAJORANA BOX QUBIT

(Γτ)-1ln P(I1,I2) 0

-5

-10 I2 = 1 I
I2 = 4 I
-15
I2 = 7 I
-20
0 2 4 6 8 10 12
I1 / I
Figure 4.7: Logarithm of the outcome distribution in the case of authentic Majorana
states as a function of I1 /I¯ for several fixed values of I2 in the limit Γτ  1. The
parameters are chosen such that |t0 /t1 |2 = 1. Furthermore, we choose arg(t∗0 t1 ) = 0,
which means that the strong positive cross-correlations, Eq. (4.82), characterizing the
Majorana box qubit are present. For fixed I2 = I, ¯ the maximum is located at I1 /I¯ = 1
as expected. For fixed I2 > I, the maximum shifts to I1 /I¯ > 1, which is consistent
¯
¯ the probability for rare events can be
with the latter result. For bigger values of I1 /I,
inferred.

90
CHAPTER 4. MAJORANA QUBIT DETECTION

4.4.5 The effects of finite MBS coupling on the box qubit

The overlap of the Majorana wave functions results in additional finite coupling terms
in the Hamiltonian H̃q given in Eq. (4.61),

3
X
H̃q → H̃q + εα σ α . (4.89)
α=1

Finite ε1 6= 0 or ε2 6= 0 does not affect the main result for F given in formula (4.82)
as long as it does not approximately cancel the splitting due to the interference links,
see Eq. (4.61). Consequently, there is a high degree of robustness with respect to
finite hybridizations ∼ iγα γ0 coupling source and drain MBSs. However, the cross-
correlations exhibit sensitivity with respect to the addition of the term ε3 6= 0 to the
Hamiltonian H̃q . More precisely, such hybridizations of the MBSs γ1 and γ2 coupled
to the drain leads cause a modification of the Fano factor F in the formula (4.82). As
shown in Appendix D, this modification is given by

2 [Re(t∗0 t1 )]2
F0 = η (4.90)
|t1 |2 (|t0 |2 + |t1 |2 )

with
2
Λ2 [ν 2 Re(t∗0 t1 )]
η= . (4.91)
Λ2 [ν 2 Re(t∗0 t1 )]2 + ε23
Hence, F = O(1) is generically maintained as long as the finite coupling ε3 fulfills
2
ε23 > Λ2 [ν 2 Re(t∗0 t1 )] with Λ the bandwidth of the leads. We note that in our proposed
geometries γ1 and γ2 are localized on different topological wires such that it is reasonable
to expect that the corresponding hybridization |ε3 | is sufficiently small.

91
4.5. COUNTING STATISTICS FOR ANDREEV BOUND STATES

4.5 Counting statistics for Andreev bound states

λ1 I1
γ1
t0 gate
λ0
γ0
t0 V
λ2
γ2
I2
Figure 4.8: Illustration of the device setting where Andreev states are represented as
pairs of MBSs γαi with i = 1, 2. The respective couplings are denoted by λiα and faded
red dots represent the i = 2 states in each wire α. The cross-correlation shot noise
Fano factor F of the currents I1 and I2 enables to distinguish whether a single MBS or
multiple MBSs are coupled to the leads.

In this section, we analyze the full counting statistics assuming that low-energy Andreev
bound states (ABSs) are coupled to the nearby leads. The latter are equivalent to a
pair of Majorana components whose overlap can be generically quite small, see e.g. Ref.
[50]. We adopt the simplified notation

0 0
tjj jj
1,αα0 ≡ tαα0 , (4.92)

and distinguish the two MBSs γαj=1,2 which may be coupled to lead α. To this end, the
effective amplitude (4.92) is endowed with indices which according to Eqs. (4.14) and
(4.15) refer to cotunneling processes from the i-th Majorana operator in lead α to the
j-th Majorana operator in lead α0 . Generally, we assume that both MBSs couple to the
adjacent lead with a strength of the same order of magnitude, i.e. O(|λiα |) = O(|λjα0 |),
such that the same is true for the effective cotunneling parameters, i.e. O(|tij αα0 |) =
O(|tkl
δδ 0 |). These assumptions are discussed in detail in Section 4.7. In the analysis that
follows, we are again interested in the low temperature regime defined by T  V . It is
not important, though, that voltage is small compared to lead bandwidth.
In terms of the Majorana operators γαi and the effective tunneling amplitudes tij
αα0 ,
the general evolution equation (4.58) governing the generalized density matrix ρt ≡

92
CHAPTER 4. MAJORANA QUBIT DETECTION

ρt (χ1 , χ2 ) takes on the form

2 X
2
∂ X
ρt = −i[H̃q , ρt ] + 2πν 2 V |tij 2 iχα
α0 | [e (iγαi γ0j )ρt (iγαi γ0j ) − ρt ]
∂t α=1 i,j=1
2
X X
+ eiχα [tij kl ∗ i j k l kl ij ∗ k l i j
α0 (tα0 ) (iγα γ0 )ρt (iγα γ0 ) + tα0 (tα0 ) (iγα γ0 )ρt (iγα γ0 )]
α=1 (i,j)6=(k,l)
2 X
X 2
i2 ∗ 1i 2i ∗
Im(ti1 1 2 1 2
 
+ α0 (tα0 ) ){iγ0 γ0 , ρt } + Im(tα0 (tα0 ) ){iγα γα , ρt }
α=1 i=1
2
!
X
22 ∗ 12 ∗
t11 t21 {γα1 γα2 γ01 γ02 , ρt }

− Re α0 (tα0 ) − α0 (tα0 ) . (4.93)
α=1

Here the Hamiltonian evolution is generated by

2
X
∗ l2 1l ∗ 2l
H̃q = 4ν 2 V (1 + ln(Λ/2V )) Im((tl1 1 2 1 2
 
α0 ) tα0 )iγ0 γ0 + Im((tα0 ) tα0 )iγα γα
l,α=1
X
2 ∗ 22 ∗ 12
(t11 − (t21
 1 2 1 2
−2ν Λ Re αα0 ) tαα0 αα0 ) tαα0 γα γα γα0 γα0 . (4.94)
α<α0

4.5.1 Counting statistics for source lead coupled to ABS

In this subsection, we will show that the shot noise cross-correlation amplitude is ex-
tremely sensitive to an ABS coupled to the source lead. In fact, the cross-correlations
are characterized generically by a Fano factor |F | = O(1) whenever more than a single
MBS is coupled to the source lead, i.e. N0 > 1. For this statement it does not matter
whether the drain leads are coupled to MBSs or ABSs and whether the interference
links are present or not.
We demonstrate this by first considering the case N0 = 2, N1,2 = 1 in the absence
of the interference links, i.e. for t0 = 0. After times Γ̃α τ  1 with the rates Γ̃α =
2πν 2 (|t11 2 12 2
α0 | + |tα0 | )V and at low temperature, T  V , we obtain (see Appendix E) the
cumulant generating function τ −1 ln Z as the eigenvalue of θ̃I + M̃ which vanishes at
(χ1 , χ2 ) = (0, 0). Here we have defined

2 X
X 2
θ̃ = 2πν 2 V |t1i 2 iχα
α0 | (e − 1) (4.95)
α=1 i=1

93
4.5. COUNTING STATISTICS FOR ANDREEV BOUND STATES

and the matrix


2
!
2
X 0 −ηα (eiχα − 1)
M̃ = 4πν V (4.96)
α=1
ηα (eiχα + 1) −(|t11 2 12 2 iχα
α0 | + |tα0 | )e

12 ∗
as well as an interference parameter ηα = Im(t11 α0 (tα0 ) ). By performing the derivatives
we obtain the second cumulant
"
∗ 12 2 11 ∗ 12 2
[Im((t1110 ) t10 )] + [Im((t20 ) t20 )]
S12 = 2πν 2 V
|t11 2 12 2
10 | + |t10 |
#
12 ∗ 11 12 ∗ 11 12 ∗ 11 12 ∗ 2
2Im(t11 (t
10 10 ) )Im(t (t
20 20 ) ) [Im(t (t
10 10 ) ) + Im(t (t
20 20 ) )]
− 11 2 12 2 3
. (4.97)
(|t10 | + |t10 | )

This contribution is of the same order of magnitude as the current and hence the Fano
factor is of order unity,
|F | = O(1), (4.98)

despite the absence of the interference links. This is in stark contrast to the case of the
Majorana box qubit where interference links t0 (of the order of t11 1,α0 ) were essential to
the observation of |F | = O(1). We can see from the formula (4.97) how the gradual
decoupling of just one of the two MBS components belonging to the ABS causes this
0
contribution to diminish. We stress that |F | = O(1) as long as the couplings tjj αα0 are of
the same order of magnitude, even when the couplings are made arbitrarily small. Let
us further analyze the formula (4.97): It is straightforward to see that the first term is
always positive, while the second term can have either sign. The second term vanishes
∗ 12 11 ∗ 12
when Im((t111,10 ) t1,10 ) = −Im((t1,20 ) t1,20 ) in which case the Fano factor characterizing
the cross-correlations takes on the simple form
 ∗ 12
2
Im((t11 1,10 ) t1,10 )
F =
12 2 2
 = O(1). (4.99)
|t11 2
1,10 | + |t1,10 |

Moreover, it follows from Eq. (4.97) that |F | = O(1) is predicted to persist upon the
inclusion of interference links, t0 6= 0, or the coupling of additional MBSs to either one of
the leads as long as more than a single MBS is coupled to the source lead, N0 > 1. This
can be understood because additional MBSs or interference links result in additional
parameters in the problem which are independent and hence generically cannot cancel
the O(1) contribution displayed in Eq. (4.97).

94
CHAPTER 4. MAJORANA QUBIT DETECTION

Physically, despite being robustly energetically low-lying, the ABSs typically are are
not exactly at zero energy. Rather, due to small overlap of the MBS components a
finite term of the form
X 3
H̃q → H̃q + ε̃α (iγα1 γα2 ) (4.100)
α=1

arises. The parameters ε̃α are small compared to the induced gap. We show in Ap-
pendix E that the inclusion of such a term does not affect the results discussed in this
subsection. In fact, e.g. the formulas (4.97) and (4.99) are unaffected by it.

4.5.2 Counting statistics for drain leads coupled to ABSs

In this subsection we will show that the shot noise cross-correlation amplitude is unaf-
fected if ABSs are coupled only to the drain leads, but not to the source lead. Therefore,
we now analyze the case N1,2 = 2 and N0 = 1 first in the absence of the interference
links, t0 = 0. We find that τ −1 ln Trρτ is given by the unique eigenvalue of θ0 I + M 0
which vanishes at (χ1 , χ2 ) = (0, 0), with

2 X
X 2
0 2
θ = 2πν V |ti1 2 iχα
1,α0 | (e − 1) (4.101)
α=1 i=1

and the matrix  


0 c1,− c2,− 0
 
−c1,+ −d1 0 c2,−
M0 = 
 
. (4.102)

 −c2,+ 0 −d2 c1,− 

0 −c2,+ −c1,+ −d1 − d2
The quantities cα,± are defined as


cα,± = 4πν 2 V Im(t11 21
1,α0 (t1,α0 ) )(e
iχα
± 1). (4.103)

Furthermore, dα is defined as

2
X
dα = 4πν V2
|tj1 2 iχα
1,α0 | e . (4.104)
j=1

95
4.5. COUNTING STATISTICS FOR ANDREEV BOUND STATES

At zero temperature we obtain to the leading order of the current

S12 = 0. (4.105)

In conclusion, this result shows that without interference links (t0 = 0) one has a
strongly suppressed Fano factor if the ABSs are only coupled to the drain leads. Hence,
the key requirement for the observation of the signature F = O(1) for t0 = 0 is that
multiple MBSs couple to the source lead, i.e. N0 > 1. On the other hand, for N0 = 1 and
more than a single MBS coupled to either or both of the drain leads, the predicted order
of magnitude of F coincides with the Majorana qubit case. As we will discuss below,
this fact can be effectively addressed by variation of the bias voltage configurations. We
also remark that for finite ABS energy (4.100) the result stated in Eq. (4.105) remains
unchanged.

96
CHAPTER 4. MAJORANA QUBIT DETECTION

4.6 Weak measurement protocols for MBS detec-


tion

The effect of remarkably pronounced current cross-correlations is an observable con-


sequence of the underlying Pauli algebra of the Majorana box qubit. The conditions
under which this effect occurs depend on the chosen device configuration. This leads
to several qualitative and experimentally verifiable predictions that allow to test the
genuineness of the Majorana qubit. Regarding the device configuration we have two
choices in mind: First, the way in which the bias is applied and, second, whether the
device is interferometric or not meaning whether electrons have a second direct tun-
neling bridge t0 available to reach either of the two leads. The prime observable is the
Fano factor F , which characterizes the cross-correlations via the relation

¯
S12 = F I, (4.106)

with the average current I¯ = hIα i. The distinctive predictions for setups containing
genuine MBSs and local fermionic states are summarized in Table 4.1.

t0 6= 0 t0 = 0
Majorana bound states (Nα = 1) |F | = O(1) |F |  1
Andreev bound states (N0 = 2, N1,2 = 1, 2) |F | = O(1) |F | = O(1)
Andreev bound states (N0 = 1, N1,2 = 1, 2) |F | = O(1) |F |  1

Table 4.1: The results are shown for the Fano factor, |F |, depending on the number of
MBSs coupled to each terminal α. A functional Majorana qubit is characterized by its
underlying Pauli algebra and by single MBSs localized at the ends of the 1D TSCs. By
measuring the Fano factor, both in the presence and absence of interference links t0 , the
Majorana box qubit is clearly identified. It is important that for the non-interferometric
configuration (for t0 = 0) each terminal serves as source lead at least once, see Fig. 4.9.
This addresses the fact that the Fano factor is sensitive only to Andreev states coupled
to the source lead, α = 0, which can be seen by comparing the first and the third row.

The continuous monitoring of a Majorana Pauli operator algebra reflects itself in strong
positive cross-correlations with F = O(1) for the interferometric setting in which t0 6= 0.
If a single Majorana state is coupled to the source lead (and the drain terminals either to
MBSs or ABSs), the Fano factor is parametrically suppressed |F | = O(T /V, ν 2 |tij 2
αα0 | ) 
1 in the absence of these interference paths, t0 = 0. If an ABS is coupled to the source

97
4.6. WEAK MEASUREMENT PROTOCOLS FOR MBS DETECTION

lead (and either MBSs ore ABSs to the drain leads), there is a strong level of cross-
correlations independent of the absence or presence of the interference links t0 . It
is this insensitivity of the cross-correlation profile to the presence of the interference
link which indicates the presence of an Andreev state coupled to the source electrode.
Thus, for pinched off interference links, t0 = 0, the cases of ABSs and MBSs coupled
to the source lead differ by a strong observable effect. All of this is to be contrasted
with noninteracting electrons in the limit T  V in a T-shaped junction. In this
case, the currents are approximately uncorrelated and the Fano factor is parametrically
suppressed in the dimensionless tunnel conductance [104] as previously mentioned.

I1 I1
γ1 γ1
t0
a) γ0 b) γ0
t0 V V
γ2 γ2
I2 I2

γ1 γ1
V I1
c) γ0 d) γ0
I0 I0
γ2 γ2
I2 V

Figure 4.9: The four device configurations for which the Fano factor F has to be
measured to identify the nonlocal Pauli algebra. In a) the interferometric setting with
t0 6= 0 is shown, which allows to confirm the F = O(1) prediction for genuine MBSs.
In the non-interferometric (t0 = 0) settings b), c), d), each lead serves as source once,
while the currents flowing to the drains are being measured. Here in each case F  1
is predicted for a genuine Majorana box qubit.

Since the sensitivity of the Fano factor is limited to distinguish single and multiple
MBSs that are coupled to the source terminal, it is important to exchange the role
of the drain and source to clearly identify the Majorana box qubit with Nα = 1 for
α = 0, 1, 2. This results in the following necessary measurement configurations, shown
in Fig. 4.9, for which the Fano factor has to be determined:

98
CHAPTER 4. MAJORANA QUBIT DETECTION

t0 6= 0 : Measurement of |F | with the central lead α = 0 as source to confirm a


positive F = O(1).

t0 = 0 : Measurement of |F | with the bias V applied such that each lead α (for
α = 0, 1, 2) serves as source at least once to confirm |F |  1. As soon as
one observes strong cross-correlations |F | = O(1) for lead α operated as
source, one can conclude that Nα > 1.

In addition, there are two control measurements that can be carried out to avoid mis-
interpretation: First, accidental fine tuning may be the origin of a |F |  1 signature,
e.g. if arg(t∗0 t1 ) = π/2 in Eq. (4.82). However, the suppression of the Fano factor
|F |  1 for ABSs requires substantial fine tuning and is non-generic. Therefore, this
scenario can be ruled out by a modest number of measurement repetitions for different
strengths of the gate potentials regulating the tunneling amplitudes as well as magnetic
field values to change the microscopic details. Second, since all results hold to leading
order in the tunnel conductances, one can repeat the protocol for a sequence of several
decreasing conductance strengths. Contributions with entry |F | = O(1) in Table 4.1
remain unchanged parametrically, while the entries |F |  1 are further suppressed.

99
4.7. CONCLUSIONS AND OUTLOOK

4.7 Conclusions and outlook

The weak measurement protocols proposed in Section 4.6 constitute a new method to
detect MBSs. For a given Majorana qubit candidate, several qualitative predictions can
be tested to unveil the genuineness of the device. The observable to be measured is the
current cross-correlation amplitude, which is characterized by a remarkably large Fano
factor |F | = O(1), see Eq. (4.82). The Fano factor is of order unity even for arbitrarily
small dimensionless tunnel conductances 2πν 2 t0,α and 2πν 2 t1,α0 as long as they are of
the same order of magnitude4 .
These pronounced cross-correlations for a Majorana qubit arise due to the simultaneous
measurement of the nonlocal Pauli operators σ1 = iγ1 γ0 and σ2 = iγ2 γ0 , which are cou-
pled to the two currents I1 and I2 respectively. The physical mechanism leading to this
effect is intimately related to the incompatibility of the two current readouts and the en-
suing frustration and conflictedness within the system. To better understand this point,
consider an electron which is incoming from the source and contributes to the current
I1 . This electron helps to project iγ1 γ0 onto a definite state. This, in turn, increases the
uncertainty in the expectation value of the operator iγ2 γ0 , which is noncommuting with
iγ1 γ0 . In this way, the aforementioned electron affects the transmission phase shift in
the cotunneling amplitude for a subsequent electron that contributes to the current I2 .
This physical interpretation accounts for two important features: First, the occurrence
of correlations even when the subsequent electron is incoming much later. This is the
case in the weak tunneling limit of interest, where electrons tunnel one-by-one through
the barriers. And second, it explains that the observation of |F | = O(1) for the Majo-
rana qubit requires the setting to be interferometric, i.e. t0,α 6= 0. For completeness,
we also performed an analysis on how the result for the Fano factor is affected by finite
MBS hybridization on the box as well as finite temperature in the Subsections 4.4.3
and 4.4.5. There we found that both effects can cause a weakening of the Fano factor.
However, we anticipate that the Fano factor |F | = O(1) will remain identifiable despite
these effects due to the strength of the cross-correlations.
Several approximations have been made in our theoretical analysis. Generally, the
results were derived perturbatively in the dimensionless tunnel conductances 2πν 2 t0,α
4
The dimensionless tunnel conductances are defined in the Hamiltonians given in Eqs. (4.14) and
(4.32). According to these definitions, ν is the density of states in the leads and t0,α describes the
tunneling link connecting lead 0 and lead α. On the other hand, t1,α0 describes the effective cotunneling
form lead 0 to lead α.

100
CHAPTER 4. MAJORANA QUBIT DETECTION

and 2πν 2 t1,α0 . Furthermore, the Lindblad equation augmented with counting param-
eters that we used to derive the results was subject to a Markovian approximation.
The latter amounts to neglecting possible memory effects. Moreover, the low-energy
Hamiltonian Eq. (4.14) does not include several features which are present in a real
experiment. Such features include the Zeemann field, spin-orbit coupling and the detri-
mental effects of quasiparticles above the gap (more on this point below). However, the
effective Hamiltonian accounts accurately for the essential physics on low energy scales
compared to charging energy EC and pairing gap ∆. And most importantly, the Fano
factor of order unity represents such a strong effect that it will remain identifiable also
upon the inclusion of further correction terms. Hence, we can draw the conclusion that
the qualitative properties of the predicted signatures will persist when more sophisti-
cated microscopic models are used.
We now turn in more detail to the in- and out-tunneling of quasiparticles through the
MBSs. The noise and decoherence caused by these processes could potentially com-
promise the interpretation of the cross-correlation amplitudes, which were derived in
the long time (zero-frequency) limit. The F = O(1) effect, Eq. (4.82), for genuine
MBSs appears after the time scale Γ−1 with Γ ≡ 2πν 2 |t1 |2 V. Therefore, we summarize
a protocol [11] geared to the characterization of the typical timescale of quasiparticle
poisoning processes. Consider both t0,2 = λj2 = 0 such that lead 2 remains decoupled.
The current I1 then depends on the state of the MBSs through the expectation value
of σ1 (or, more generally, that of an operator O10 if ABSs are present). Beyond a time
scale τproj ∼ 1/V , the measurement of I1 becomes projective, and a weakly fluctuating
result defined by one of the two values hσ1 i → ±1 is approached. However, quasiparticle
tunneling accidentally switching the state σ1 → −σ1 will cause discrete jumps in the
readout. Monitoring of the current signal therefore allows to study the quasiparticle
poisoning processes.
We have shown that for Andreev bound states the Fano factor signatures generically
differ in a drastic way. This is due to the underlying operator algebra, which in this
case is different from the nonlocal Pauli algebra. We emphasize again that observation
of the ABS signatures discussed in Section 4.6 requires that both Majorana components
forming the ABS actually couple to the nearby lead. If one of the MBS components
of the ABS is invisible to the nearby electrode, MBSs and ABSs are in principle indis-
tinguishable. Note that there are studies stating that for smooth confining potentials,
in fact, only one MBS component effectively couples while the coupling of the other

101
4.7. CONCLUSIONS AND OUTLOOK

component is exponentially smaller (see e.g. Refs. [49, 50]). It has also been discussed,
however, that the coupled “fake” MBS could potentially still be used for braiding [52].
We reiterate that the protocol presented in Section 4.6 is tailor made to distinguish for
each lead α a single coupled MBS, Nα = 1, and multiple coupled MBSs, Nα > 1 . Hence,
the protocol does not distinguish the topological (Nα odd) from the non-topological (Nα
even) superconductor. However, we believe our protocol is highly useful in practice,
because it is desirable to identify a functional Majorana qubit with Nα = 1. This is
because a topologically nontrivial system in which a genuine MBS coexists with ad-
ditional low-lying Andreev states causes a reduced effective gap rendering the device
computationally less useful.
In summary, the proposed protocols rely on available experimental hardware and, there-
fore, may allow to detect the Pauli algebra of a Majorana box qubit in the near future.
The experiments have to be performed in the Coulomb blockade regime at low temper-
atures, T  V , and in the tunneling limit, i.e. 2πν 2 t0,α  1 and 2πν 2 t1,α0  1. Here,
one must not underestimate the task of shot noise measurement in nanoscale devices
where the electrons tunnel one-by-one through the barriers. Despite that, the weak
measurement protocols may represent the operationally and practically most viable op-
tion to detect the Pauli algebra of the Majorana box qubit and the nonlocality of MBSs.
The successful completion of these experiments would constitute notable progress that
could erase the doubts about the MBS. Nevertheless, at some point in the future actual
braiding of MBSs has to be performed, which would constitute a landmark achievement
in condensed matter physics.

102
Chapter 5

Further new approaches to MBS


detection

The theory of MBS detection is a major theme in this thesis and has been addressed
at length in Chapter 4. In the present chapter, we investigate two further transport
spectroscopic methods. First, in Section 5.1, we discuss detection based on current
cross-correlations for a single Coulomb blockaded Majorana quantum wire with three
tunnel coupled electrodes. Second, in Section 5.2, we study MBS identification based
on two terminal projective conductance measurements by means of a protocol that
accesses the number of distinct current outcomes. Beyond the usefulness of these de-
tection protocols, the discussion in Section 5.3 reveals that the aforementioned protocol
of simultaneous monitoring a nonlocal Pauli algebra of Chapter 4 accesses the most in-
formation related to the fundamental nature of MBSs.

5.1 Current cross-correlations for a single Majo-


rana quantum wire

5.1.1 Three-terminal device and theoretical model

We now explore current cross-correlation experiments based on a single TSC wire in the
Coulomb blockade regime. In particular, we consider that one end of the Majorana wire
is coupled to a source lead while two separate drain leads are coupled to the other end

103
5.1. CROSS-CORRELATIONS FOR A SINGLE MAJORANA WIRE

I1
λ1
λ0
γL γR
V λ2
I2

Figure 5.1: Experimental setting of a topological superconducting nanowire (green)


coupled to three normal metal leads (thick black lines). A bias voltage V is applied to
operate the lead 0, which is tunnel coupled to the MBSs γLi localized at the left wire end
(red dots), as the source. At the right end of the wire the two drain leads 1 and 2, which
are electrically isolated from each other, are tunnel coupled to the MBSs γRi . The two
currents I1 and I2 are flowing to ground and the cross correlation amplitude S12 = F hIi
gets measured with hIα i = hIi. The quantity F contains qualitative information on the
i
number of MBSs γL/R that are coupled to the respective leads. The observation of
F  1 is consistent with single genuine MBSs. On the other hand, F = O(1) implies
that ABSs are being present.

(see Fig. 5.1). The experiments are envisioned to be carried out at low temperature T
in comparison to the applied voltage bias V . The charging Hamiltonian is again given
by HC = Ec (N − ng )2 with a large charging energy EC . Let the equilibrium charge on
the floated wire be the integer N0 in units of the elementary charge with the tunable
backgate parameter ng = N0 + ∆ng that is weakly detuned, ∆ng  1. The microscopic
tunneling Hamiltonian [30] for the system depicted in Fig. 5.1 reads

NL X h
X i
H = HLeads + HC + λj0 c†0,k γLj e−iφ/2 + H.c.
j=1 k
NR X h
X i
+ (λj1 c†1,k + λj2 c†2,k )γRj e−iφ/2 + H.c. , (5.1)
j=1 k

i
with NL/R MBSs γL/R being localized on the left and right end of the wire respectively.
As before, the parameters λjα are tunnel couplings, the operator e−iφ/2 lowers the island
charge and HLeads describes the metallic leads. Using perturbation theory for the case
NL/R = 1 yields the effective tunneling Hamiltonian for single MBSs (see Appendix F)

104
CHAPTER 5. FURTHER NEW APPROACHES TO MBS DETECTION

" 2
#
X X
HMBS,eff = η 1 c†1,k c2,k0 + iγR1 γL1 t11 †
α0 cα,k c0,k0 + H.c. , (5.2)
k,k0 α=1

where the direct hopping amplitude from lead 1 to 2 is given by

j 4∆ng λj1 (λj2 )∗


η '− (5.3)
EC
0 0
and tjj j j ∗ j
αα0 ' 2iλα (λα0 ) /EC . Note that η is proportional to ∆ng = ng − N0  1 and
originates in the tunneling process c2 → γRj → γRj → c1 with cα=1,2 the drain lead
fermions. For the case of ABSs NL/R = 2, we obtain the effective Hamiltonian (see
Appendix F)

2 X
2 X
2 Xh i
jj 0 j j0 †
X
HABS,eff = HLeads + tα0 (iγR γL )cα,k c0,k0 + H.c.
α=1 j=1 j 0 =1 k,k0
Xh  † i
+ t12
12 (iγ 1 2
γ
R R ) + η 1
+ η 2
c c
1,k 2,k 0 + H.c. . (5.4)
k,k0

In order to derive the counting statistics, we follow the same strategy as in Chapter 4.
The equation governing the time evolution of the reduced density matrix ρt (χ1 , χ2 ) in
j
the Hilbert space corresponding to the Majorana operators γL/R is given by

2
∂ X
ρt = −i[Hq , ρt ] + 2πν 2 V D̃α0 (ρt )
∂t α=1
 
2
+2πν T D̃12 (ρt ) + D̃21 (ρt ) . (5.5)

The superoperator is defined as

† 1 †
D̃αα0 (ρ) ≡ ei(χα −χα0 ) Παα0 ρΠαα 0 − {Π 0 Παα0 , ρ} (5.6)
2 αα

with the counting parameters χα defined as in Chapter 4 and again χ0 ≡ 0. The


operators Παα0 are defined according to

2
X 0 0
Πα0 = tjj j j
α0 (iγR γL ) (5.7)
j,j 0 =1

105
5.1. CROSS-CORRELATIONS FOR A SINGLE MAJORANA WIRE

and
Π12 = η 1 + η 2 + t12 1 2
12 (iγR γR ) (5.8)

with Π21 = Π12 . The Hamiltonian part of the dynamics for T  V  Λ is generated
by
2
2
X † ν 2Λ X †
Hq = ν V ln(Λ/2V )) [Πα0 , Πα0 ]− {Παα0 , Παα0 }. (5.9)
α=1
2 α<α 0

In the next subsection, we will discuss the resulting current cross-correlations.

5.1.2 Signatures of current cross-correlations


´t
The cross-correlation amplitude S12 = 0 dt hhI1 (t)I2 (0)ii can be derived from the evo-
lution equation augmented with counting parameters stated in Eq. (5.5), which is
shown in Appendix F. In the case of genuine single MBSs, NL/R = 1, we obtain thermal
anticorrelations as the leading order contribution to the cross-correlations:

S12 = −4πν 2 |η 1 |2 T. (5.10)

Assuming that the couplings are of the same order, i.e. O(λα ) = O(λα0 ), we find that
the Fano factor for MBSs is small
 
T
|F | = O ∆ng  1. (5.11)
V

In fact, it is parametrically suppressed in temperature T /V  1 and the small detuning


∆ng  1. Experimental variation of the back gate parameter can, therefore, tune the
strength of the thermal anticorrelations.
In stark contrast to the result (5.11), in the ABS case there are generically strong cross-
correlations characterized by a Fano factor |F | = O(1). To demonstrate this point it
is sufficient to discuss the case NR = 2, NL = 1. Physically, this situation may be
realized when only one of the MBSs localized at the left wire end couples effectively to
the nearby lead, a situation that has been discussed e.g. in Ref. [50]. By solving the
corresponding master equation augmented with counting parameters (see Appendix F),

106
CHAPTER 5. FURTHER NEW APPROACHES TO MBS DETECTION

we obtain the cross-correlation amplitude


  
P2 12 ∗ 2 1j 2 11 12 ∗ 2 1j 2
j=1 [Im(t11
10 (t10 ) )] |t20 | + [Im(t20 (t20 ) )] |t10 |
S12 = 8πν 2 V 

P 2
2 P2 1j 2
α=1 j=1 |tα0 |

12 ∗ 11 12 ∗ 11 12 ∗ 12 ∗ 2
4Im(t11
10 (t10 ) )Im(t20 (t20 ) ) [Im(t10 (t10 ) ) + Im(t11
20 (t20 ) )] 
− P 3 . (5.12)
2 P2 1j 2
α=1 |t
j=1 α0 |

0
Hence, assuming again that the couplings are of the same order, i.e. O(λjα ) = O(λαj 0 ),
we predict a generic Fano factor of order unity,

|F | = O(1). (5.13)

We point out that |F | = O(1) is generally the case whenever NR > 1 and NL ≥
1 is satisfied. The situation may also be realized in a topological Majorana wire,
where the genuine MBSs coexist with ABSs. Thus, the cross-correlations in this setting
provide another scheme to distinguish single MBSs, which are characterized by the
strong suppression F ∼ ∆ng T /V , from ABSs. This is summarized in Table 5.1.

Cross-correlation Fano factor


Majorana states |F | = O (∆ng T /V )  1
Andreev bound states |F | = O(1)

Table 5.1: Qualitative behavior of the Fano factor, |F |, characterizing the cross-
correlations for single vs multiple MBSs localized at the ends of the nanowire. The
key observable distinguishing between the two cases is the large value |F | = O(1)
for Andreev bound states, which is to be contrasted with the strong suppression
|F | ∼ ∆ng T /V for MBSs. Observing the first line is a result consistent with MBSs
and the experimentalist can conclude that not more than single MBSs are coupled to
the respective terminals.

The main advantage of the protocol is its ability to rule out ABSs based on the distinc-
tive large cross-correlations that are present in this case. However, observing F  1
(first line of Table 5.1) does not contain a direct signature of the MBSs. This can be
1
seen by replacing γR/L by complex numbers in the Hamiltonian (5.2) in which case F
is suppressed as well. Thus, observation of F  1 is consistent with single MBSs and
one learns from it that not more than single MBSs are coupled to the respective leads.

107
5.1. NUMBER OF PROJECTIVE CURRENT OUTCOMES

5.2 MBS detection via number of projective cur-


rent outcomes

In this section, we consider projective current measurements on a Majorana qubit as


studied in Ref. [11]. As in this reference, we consider a floating Majorana qubit with
two normal-conducting leads coupled to two different Majorana states in the presence
of an applied bias, see Fig. 5.2. The idea is to investigate the extent to which MBSs
and ABSs can be distinguished based on the number of distinct projective current out-
comes.
The charging Hamiltonian of the Majorana box qubit is again given by HC =
Ec (N − ng )2 . In order to practically identify different current outcomes (stationary
currents), one has to measure the current multiple times projectively. To this end, we
propose the option to run two currents simultaneously to reset the qubit by temporarily
coupling the box qubit to a third lead as well. Let Nα be the number of MBSs coupled

1.) λ0 V
γ3 γ0
SC t0,1
γ2 γ1 λ1
I1

2.) λ0 V
γ3 γ0
λ2 SC t0,1
γ2 γ1 λ1 I
I2 1

Figure 5.2: The two alternating stages of the protocol are displayed. Top: Stage of
projective measurement of the current hI1 i [11] in the box qubit geometry. Only the
source lead 0 and a single drain lead 1 (thick black lines on the right) are being coupled
to the island, a bias is applied and an interference link t0,1 serves as direct tunneling
link. Bottom: Resetting stage of the box qubit, which prepares for the next projective
measurement iteration. Three leads are being coupled such that due to the biasing on
average the two currents I1 and I2 flow simultaneously. The dynamics during this stage
brings the qubit density matrix in the maximally mixed state as discussed in Subsection
4.4.2. After a certain time lead 2 is detached again and step 1.) is repeated.

108
CHAPTER 5. FURTHER NEW APPROACHES TO MBS DETECTION

to lead α. According to the discussion of Section 4.2, the effective Hamiltonian during
the projective measurement stage (upper half of Fig. 5.2) is given by

X
H̃T = (t0,1 + O10 ) c†1,k c0,k0 + H.c., (5.14)
k,k0

with O10 = N
P 0 PN1 ij i j
i=1 j=1 t1 (iγ1 γ0 ) and an interference link t0,1 between lead 1 and 0. For
a Majorana qubit (N0 = N1 = 1) with interference links t0 the current can take on two
values [11]
∗ 11
hI1 i = 2πν 2 V (|t0,1 |2 + |t11 2 1 1
1 | + 2Re(t0 t1 )(iγ1 γ0 )) (5.15)

defined by the eigenvalue of iγ11 γ01 that can take on the values ±1. Note that the bias
V needs to be finite in order to have a Pauli dependent current average. In the absence
of the interference link, i.e. t0,1 = 0, the same current average hI1 i = 2πν 2 |t11 2
1 | V is
measured in each iteration, irrespective of the state of the Majorana qubit. Now we
have to address the fact that during the measurement the qubit has been projected to
the corresponding eigenstate of iγ11 γ01 . Thus, a reset has to be carried out by changing
the qubit state such that the subsequent projective measurement has a finite chance
to yield all possible outcomes again. We achieve this by coupling lead 2 to one of the
TSC ends on the other side of the box qubit (see Fig. 5.2), which is far separated.
We assume lead 2 to be at the same potential as lead 1, such that a simultaneous
average flow of the two currents I1 and I2 sets in. For genuine MBSs, the central island
constitutes a Majorana qubit and the coupling of lead 2 implies that the evolution
equation ∂t ρt = Γ [(iγ11 γ01 )ρt (iγ11 γ01 ) − ρt ] + . . . for the reduced qubit density matrix ρt
gets modified according to

2
X
 1 1 1 1
Γ (iγα1 γ01 )ρt (iγα1 γ01 ) − ρt
  
Γ (iγ1 γ0 )ρt (iγ1 γ0 ) − ρt → (5.16)
α=1

with Γ ≡ 4πν 2 |t1 |2 V . The additional decoherence term in Eq. (5.16) ensures that
the state of the qubit cannot remain in an eigenstate of σ1 = iγ11 γ01 . In fact, after a
timescale τreset ≡ Γ −1 the qubit density matrix has approached the maximally mixed
state ρ0 = 12 σ0 , as discussed in Subsection 4.4.2. This means that after both currents
have been flowing for some time, the observed average current is given by hI1 i =
2πν 2 V (|t0,1 |2 + |t11 2
1 | ), as discussed in Subsection 4.4.2. Moreover, the shot noise S11 is
increased according to Eq. (4.79), when both currents flow simultaneously. After some

109
5.2. NUMBER OF PROJECTIVE CURRENT OUTCOMES

time, lead 2 gets decoupled and the projective measurement of I1 has a finite chance of
either of the two possible outcomes (5.15) again. In summary, we propose the following
two-step protocol:

1. Measurement: The tunnel coupling λ2 connecting drain 2 and the MBS γ2 is


turned off (e.g. controlled via a gate) and hI1 i gets measured projectively resulting
in a well defined (projective) average current (upper half of Fig. 5.2).

2. Reset of qubit state: To enable a next measurement which has again a finite
probability to yield all the possible current outcomes, the drain 2 gets coupled
such that the currents I1 and I2 flow simultaneously on average for a sufficient
time τ > Γ −1 (bottom half of Fig. 5.2).

Hence, multiple repetitions of the two-step protocol for t0,1 6= 0 allow to identify two
different stable current averages for a genuine Majorana qubit, which do not change
because the protocol leaves the relevant tunnel couplings unaffected.
On the other hand, in the case of ABSs the dimension of the Hilbert space spanned
by the MBSs coupled to the two leads 0 and 1 is bigger. The projective current mea-
surement can then result in more possible outcomes. Let us illustrate this for the case
N0 = N1 = 2 where we simplify the discussion by assuming t0,1 , tij
1 ∈ R. In this case
the current operator is given by

2
!
X
hI1 i = 2πν 2 V t20,1 + (tij 2 ij i j 21 12 11 22 1 2 1 2
 
1 ) + 2t0,1 t1 (iγ0 γ1 ) + 2(t1 t1 − t1 t1 )γ0 γ0 γ1 γ1 .
i,j=1
(5.17)
This current operator generically gives rise to four different current outcomes. If the
interference link is switched off, t0,1 = 0, we have

hI1 i = V (g1 + g2 γ01 γ02 γ11 γ12 ), (5.18)

where we define the dimensionless tunnel conductances g1 = 2πν 2 2i,j=1 (tij 2


P
1 ) and g2 =
4πν 2 (t21 12 11 22 1 2 1 2
1 t1 − t1 t1 ). Since the quartic product of Majorana operators γ0 γ0 γ1 γ1 has the
two possible eigenvalues ±1 there are still two different outcomes, hI1 i = V (g1 ± g2 ).
Hence, we can conclude that for ABSs we can have four outcomes in the presence
of the interference path and two outcomes in its absence. Now we look at the cases
N0 = 1, N1 = 2 and N0 = 2, N1 = 1, i.e. we ask whether a single additionally coupled

110
CHAPTER 5. FURTHER NEW APPROACHES TO MBS DETECTION

MBS can be detetced. We find that there is a single outcome for t0,1 = 0 and two
outcomes for t0,1 6= 0, just as in the case Nα = 1 (see Appendix F). Thus, summarizing
this discussion we arrive at the Table 5.2.

Number of MBSs coupled to lead α t0,1 6= 0 t0,1 = 0


N0 = N1 = 1 2 current outcomes 1 current outcome
N0 = 1, N1 = 2 or N0 = 2, N1 = 1 2 current outcomes 1 current outcome
N0 = N1 = 2 > 2 outcomes > 1 outcome

Table 5.2: Projective current measurement of I1 as a way to distinguish genuine MBSs


from ABSs: By repeating the steps 1.) and 2.) outlined in the main text and Fig.
5.2, the number of outcomes can be obtained. There are qualitative differences in this
number, which enable the distinction. The first row shows the result consistent with
single MBSs coupled to the leads 0 and 1. If more different outcomes are observed, the
presence of ABS can be inferred (third row) and the device is not a Majorana box qubit.
However, the number of outcomes is blind to a single additional MBS, see second row.

The number of current outcomes differs qualitatively, when step 1.) and 2.) are iterated
many times, depending on whether the source lead 0 and the drain lead 1 are each
coupled to single MBSs or ABSs. In particular, one should repeat the two stages
multiple times for a single device with and without the interference link connecting
leads 0 and 1. When more different current outcomes than predicted for MBSs can
be identified, the presence of ABSs can be inferred (see Table 5.2). Importantly, the
couplings λj0 and λj1 (see Fig. 5.2) stay unchanged throughout the protocol since only
the amplitudes λj2 get tuned e.g. via gating, which means that the value of the current
average for a given state is indeed unaffected.
The time scale for which the projective measurement has to be conducted to identify
an outcome can be estimated by the time

hI1 i
τproj ' (5.19)
(∆I)2

it takes to reach a signal to noise ratio of order unity. Here we made use of the fact
that the noise S11 is given by the Schottky formula, S11 = hI1 i. The quantity ∆I
represents the difference in the current outcomes, which can be nearly degenerate in
highly fine tuned situations. For the example given in Eq. (5.18), we have ∆I = 2g2 V .
For a Majorana qubit, N0 = N1 = 1, the difference in currents is ∆I = 8πν 2 V Re(t∗0,1 t1 )
2
leading to the timescale τproj ' (|t0,1 |2 + |t1 |2 )/(32π Re(t∗0,1 t1 ) V ).


111
5.3. CONCLUSIONS AND OUTLOOK

5.3 Conclusions and outlook

In this chapter, we introduced two further protocols geared to identify genuine MBSs
in topological SC wires. The first is a current cross-correlation measurement protocol
for a Coulomb blockaded Majorana quantum wire with three tunnel coupled terminals.
The second is a projective conductance measurement protocol based on the number of
distinct current outcomes. In the following, we discuss these two experimental protocols
and compare them with each other, as well as with the weak simultaneous measurement
protocol of noncommuting Pauli operators presented in Chapter 4. We argue that the
latter is more desirable to be executed, because it yields more information related to
the fundamental nature of MBSs.
The first detection protocol of Majorana qubits presented in Section 5.1 leads to strong
observable differences distinguishing genuine MBSs and ABSs that are coupled to the
respective leads. The main advantage of the protocol is its ability to rule out ABSs
based on the observation of |F | = O(1). However, observing the signature F  1 is
merely consistent with MBSs but does not contain a direct signature of the Majorana
states. This is because for a setting of non-interacting electrons, F is suppressed as
well. Thus, observation of F  1 has to be interpreted as a result that is consistent
with MBSs, which implies that not more than single MBSs are coupled to the respective
leads. In this sense, the information gained is not comparable to the protocol presented
in Chapter 4. Nevertheless, it is useful to rule out multiple MBSs coupled to the drain
leads.
A new feature of the second protocol discussed in Section 5.2 is the alternation of
two stages as displayed in Fig. 5.2. This addresses the fact that after a projective
measurement of a single current a reset of the qubit state is necessary. We achieve
this reset by temporarily coupling a second drain lead to another MBS on the box,
resulting in the simultaneous flow of the two currents I1 and I2 . We emphasize that the
experimental demonstration of this reset procedure reveals evidence of noncommuting
degrees of freedom on the box. This is because of the way that the average of the
current I1 and its noise level is affected depending on whether I2 is flowing or not.
The main shortcoming of the protocol is that the number of outcomes for the cases
N0 = 1, N1 = 2 and N0 = 2, N1 = 1 is the same as for N0 = N1 = 1. In contrast, the
protocol outlined in Chapter 4 can clearly identify single MBSs. We also emphasize
that the current outcomes are obtained under the same conditions (density of states,

112
CHAPTER 5. FURTHER NEW APPROACHES TO MBS DETECTION

shape of the contacts, etc.) and that constant contributions to the current due to
quasiparticles are always present in each measurement iteration. For the protocol of
Section 5.2, the timescale τproj (see Eq. (5.19)), on which the current measurement can
be considered to be projective, must be shorter than the typical time scale after which
a quasiparticle poisoning event takes place. Those events could either occur due to
thermal excitation or due to a tunneling event from outside the island. Quasiparticle
poisoning events would cause a flip of the Pauli operator and hence a jump in the
average of the noisy current signal. Therefore, the applied bias must be large enough
such that τproj is shorter than the characteristic timescale of both of these quasiparticle
poisoning processes.
Finally, we can draw the following conclusion regarding the MBS detection: When the
three detection protocols presented in this dissertation are compared, the simultaneous
weak measurement approach of noncommuting operators described in Chapter 4 yields
the most information about the fundamental nature of MBSs. Hence, it is the most
desirable approach when it comes to practical implementation.

113
5.3. CONCLUSIONS AND OUTLOOK

114
Chapter 6

Overall conclusion and outlook

Improvements in the ability to more accurately manufacture and control quantum de-
vices are likely to lead to disruptive technological developments. The Majorana bound
state has been pursued in topological SCs for a number of years now, but its unam-
biguous confirmation has proven to be a particularly challenging task up to this day.
An attractive application of the Majorana state is the Majorana qubit, but, so far,
no functional device of this kind has been realized in the laboratory. In this thesis,
we have addressed this situation and studied the Majorana qubit in a two-fold way:
Firstly, by a thorough investigation of new TI nanoribbon architectures of Majorana
qubits as alternatives to the qubit architectures based on semiconductors in Chapter 3.
And secondly, by proposing new experimental methods and corresponding observable
predictions to perform the identification of Majorana qubits in Chapters 4 and 5.
As shown in Chapter 3, Majorana qubit architectures building on proximitized TI
nanowire segments of differing width promise to represent a suitable and flexible plat-
form with robust Majorana states. The manipulation of local electrochemical potentials
in these systems allows to control the coupling of pairs of MBSs. In Chapters 4 and 5,
three methods have been put forward and studied theoretically in detail. The protocol
of simultaneous weak measurement of noncommuting operators presented in Section 4.6
enables to identify the Majorana qubit devices with single MBSs and allows to extract
a clear signature of the nonlocal Pauli algebra. This is a clear advantage and makes
this protocol more efficacious than the two methods put forward in Chapter 5. In that
chapter, we found that measuring the shot noise in a setting with three leads and a
single Majorana quantum wire can be used to rule out more than a single MBS at

115
the wire ends. However, for a genuine Majorana wire no direct signature of the MBSs
is accessed in this way. In contrast, the protocol of simultaneous weak measurement
of noncommuting operators presented in Section 4.6 clearly allows to extract a direct
signature of genuine MBSs. Hence, the latter represents the more desirable goal that
would clearly constitute experimental progress in the pursuit of the Majorana state.
Due to the accessibility of the necessary hardware, the milestone of establishing that
a given quantum device is a genuine Majorana qubit is realistically achievable in the
near term future. This quantum device could be engineered using TI nanoribbons as
outlined in Chapter 3, it could be made of the semiconducting wires that are studied
in several laboratories around the world, or it could be based on any other physical
platform in which zero-dimensional Majorana states are predicted. The experiments
have to be carried out in the Coulomb blockade regime at low temperatures and in the
tunneling limit. Here, one must not underestimate the task of shot noise measurement
in nanoscale systems where electrons pass the tunneling barriers individually. Never-
theless, one could argue that the protocol of Section 4.6 represents the most viable
option to identify Majorana states and the nonlocal Pauli algebra of the Majorana
qubit from the operational and practical standpoint. Successful observation of the pre-
dicted Majorana signature would provide further impetus to move on to realize full
braiding protocols, whose completion would constitute a remarkable breakthrough for
fundamental science.
Once under full control, the Majorana qubit becomes a building block for higher-level
applications and further experiments. For instance, it would be noteworthy to realize
a Majorana surface code [86, 87, 88], which is a two-dimensional network structure
as briefly discussed in Section 3.4. In conclusion, we anticipate that this work will
stimulate much needed experimental and theoretical progress in the field and bring us
closer to Majorana based quantum computation.

116
Appendices

Appendix A: Analysis of spinor wave functions

This part of the appendix provides additional details on the derivation of the results
presented in Chapter 3. To keep the formulas compact, we work in units with ~υ1 = 1.
We start by providing the explicit form of the eigenstates of the BdG Hamiltonian
(3.12), 
Ψ (z) , z ≤ −W/2,
 L


Ψ(z) = ΨC (z) , |z| ≤ W/2, (6.1)


ΨR (z) , z ≥ +W/2,

for the Majorana qubit platform shown in Fig. 3.3. Since we are interested in con-
structing Majorana bound states, we only consider energies below the superconducting
gap, |E| < ∆. We shall first write down general solutions of the BdG equation in each
(±)
of the three regions defined by the constriction length W . Using parameters A1,2 , the
solution for |z| > W/2 vanishing exponentially at |z| → ∞ reads

    √ 
E ∆2 − E 2
  √   
√  (−)  ∆2 − E 2 −E
− ∆2 −E 2 |z|  + A(−)
  
ΨL (z) = e A1  2 

−iφ/2
 , (6.2)



 0 
  ∆e


∆e−iφ/2 0

    √ 
E − ∆2 − E 2
  √   
√  (+)  − ∆2 − E 2 −E
− ∆2 −E 2 |z| (+)
  
ΨR (z) = e A1   + A2   , (6.3)



 0 


 ∆eiφ/2 

∆eiφ/2 0

117
Appendices

where the first and second (third and fourth) component refers to the spin structure of
the particle (hole) part of the Nambu spinor. In the central region |z| < W/2 and with
(±)
coefficients B1,2 , the solution is given by

√ √
M02 −(E+µ)2 z
(+) 2 2 (−)
ΨC (z) = e B1 u+ (E) + e− M0 −(E+µ) z B1 u− (E)
√ 2 √ 2
2 (+) 2 (−)
+e M0 −(E−µ) z B2 v + (E) + e− M0 −(E−µ) z B2 v − (E), (6.4)

with
   
± (M0 + E + µ) 0
 q
 M02 − (E + µ)2  0
  
 
u± (E) = 
 , v ± (E) =  . (6.5)
 0



 ± (M0 − E + µ) 

q
0 M02 − (E − µ)2

Now one has to demand continuity of the spinor wave functions at the points z =
±W/2 , which translates into the conditions ΨL (−W/2) = ΨC (−W/2) and ΨC (W/2) =
ΨR (W/2). For eigenenergies with |E| < min(∆, M0 ) these conditions can be cast into
a zero-determinant condition
D(E) = 0. (6.6)

Here the corresponding determinant reads


X
(2E 2 − ∆2 ) a+ a− ± µ2 ∓ E 2 ± M02 ∆2 cosh[(a− ± a+ )W ]
 
D(E) =
±
√ X
+2E ∆2 − E 2 [(E − µ)a+ ± (E + µ)a− ] sinh[(a− ± a+ )W ]
±
2
−2∆ a+ a− cos φ, (6.7)

with q
a± (E) ≡ M02 − (E ± µ)2 . (6.8)

The determinant D(E) is a symmetric function of energy, i.e. D(E) = D(−E), as


required by the particle hole symmetry of the superconductor. A robust feature found
by solving the Eq. (6.6) is the presence of subgap state solutions at

E = ±ε. (6.9)

118
Appendices

Under the self consistent condition |ε|  min(∆, M0 ), we expand D(E) up to second
order in E. For this we use the relation

M02 + µ2 2
a+ (E)a− (E) = M02 − µ2 − E + O E4

2 2
(6.10)
M0 − µ
and obtain

D(E) = − 4(~υ1 /ξ)2 ∆2 cos2 (φ/2) + 2E 2 (~υ1 /ξ)−2 M04 + ∆2 (µ2 + 2W 2 µ4 )




− M02 (3 + 2W 2 ∆2 + 4W ∆)µ2 + 2(1 + 2W ∆)µ4 + ∆2 (M02 + µ2 ) cos (φ)


+M02 (∆2 + (~υ1 /ξ)2 ) cosh(2W/ξ) + 2M02 ∆(~υ1 /ξ) sinh(2W/ξ) + O E 4 .
 

(6.11)

Assuming W  ξ we obtain the following explicit expressions for the energies of the
above mentioned subgap states as

2(~υ1 /ξ)2 ∆ cos (φ/2)
ε = p
M0 (∆2 + (~υ1 /ξ)2 ) cosh(2W/ξ) + 2∆(~υ1 /ξ) sinh(2W/ξ)
φ 2∆ (~υ1 /ξ)2 −W/ξ
 
' cos e . (6.12)
2 M0 ∆ + ~υ1 /ξ

This concludes the derivation of the hybridization formula Eq. (3.28).

119
Appendices

Appendix B: Effective tunnel Hamiltonian of three-


terminal device

This part of the appendix provides details on the derivation of the effective Hamilto-
nian (4.14) presented in Section 4.2. The series expansion for the effective tunneling
Hamiltonian is given by [96]
∞ ∞  n−1
X (n)
X 1
Heff ≡ Heff = P0 HT HT P0 , (6.13)
n=1 n=1
−HC

where HC is the charging Hamiltonian and P0 is the projection operator onto the
definite charge ground state |N0 i (see below). Furthermore, the microscopic tunneling
is described by

2 XX
X Nα
HT = λjα c†α,k γαj e−iφ/2 + H.c.. (6.14)
α=0 k j=1

This perturbative procedure HC + HT → Heff is valid in the weak tunneling limit λiα 
∆, EC . Similar Hamiltonians have been previously derived, see Refs. [29, 30, 87, 88].
The Hamiltonian HC has the eigenvalues E(N ) = Ec (N − ng )2 with N the number
of island electrons and we assume to be far away from the charge degeneracy points
ng ∈ Z + 21 . Then, the ground state has a definite charge which we denote by N0 .


We take into account fluctuations to the sectors of higher energy defined by the charge
states N0 ± 1. In the same way as in Ref. [30], we define the energy differences
E± = E(N0 ± 1) − E(N0 ). Since the operator eiφ/2 causes the addition of an electron
on the island, it follows that the first order contribution in the series 6.13 vanishes, i.e.
(1) (2)
Heff = P0 HT P0 = 0. The first non-vanishing contribution is Heff and gives
 
(2) 1
Heff = P0 HT HT P0
−HC
2 N α0
Nα X j j0 ∗
j −iφ/2 λα (λα0 ) iφ/2 j 0
X X X
j †
= − P0 λα cα,k γα e e γα0 cα0 ,k0 P0
k,k0 α,α0 =0 j=1 j 0 =1
E−
0
!
j0 (λj 0 )∗ λjα †
+P0 eiφ/2 γα0 cα0 ,k0 α cα,k0 γαj e−iφ/2 P0 + H.c.. (6.15)
E+

120
Appendices

We may write 
(2)
3 
Heff = Heff + O EC−2 λjα . (6.16)

Therefore, the higher order terms are strongly suppressed in the limit of interest. We
split the effective Hamiltonian (6.15) into three different types of terms,

(2)
Heff ≡ H̃T + H̃L + H̃ABS . (6.17)

The first type arises from contributions with α 6= α0 and leads to tunneling terms
connecting the leads α and α0 . The corresponding Hamiltonian term reads

N α0
Nα X 0
XXX iλj (λj 0 )∗ (E+ + E− ) 0
H̃T = α α
(iγαj γαj 0 )c†α,k cα0 ,k0 + H.c.. (6.18)
k,k0 α6=α0 j=1 j 0 =1
E+ E−

The effective amplitude for the single electron tunneling processes from α0 to α is given
by

0 0 E+ + E−
tjj j j ∗
1,αα0 = iλα (λα0 )
E+ E−
0
2iλjα (λjα0 )∗
=
EC (1 − 4∆ng )
0
λjα (λjα0 )∗
' 2i , (6.19)
EC

where we have used the assumption |N0 − ng |  1 and thus arrive at the Hamiltonian
stated in Eq. (4.14). These second order cotunneling processes, where electrons tunnel
in and out at different terminals, are the most relevant for our purposes. For complete-
ness, we mention the two other types of processes: First, for α = α0 and j = j 0 the
term H̃L up to a constant reads

2 X
Nα 2
XX |λi | (E− − E+ )
H̃L = α
c†α,k cα,k0 + H.c. (6.20)
k,k0 α=0 i=1
E+ E−

and renormalizes the lead potentials. This renormalization of the leads vanishes for
ng ∈ Z. Physically, this term originates in the process in which an electron tunnels
from the lead into γαi and subsequently from the same MBS back into the same lead.
Second, there is a type of term that is only present for Andreev bound states. It
originates in processes where an electron tunneling into an island MBS is accompanied

121
Appendices

by an electron tunneling back into the same lead from a different MBS, i.e. contributions
with α = α0 and j 6= j 0 :

2 X h i
0 0
XX † −1 †
H̃ABS = λjα (λjα )∗ γαj γαj −1
E+ cα,k0 cα,k − E− cα,k cα,k0 + H.c.. (6.21)
k,k0 α=0 j6=j 0

After tracing out the leads, these terms weakly renormalize the energy of the ABSs,
which does not affect the qualitative conclusions drawn in the counting statistics cal-
culations as shown in Appendix E.

122
Appendices

Appendix C: Further details on the derivation of the


master equation

In this part of the appendix, we provide further details on the derivation of the Bloch
Redfield equation supplementing the Subsections (4.3.3) and (4.3.4). In the procedure
of integrating out the lead degrees of freedom, one is confronted with expressions like

ˆ∞   ¨ ˆ∞
dsTr ρL Qα0 (t − s)Q†α0 (t) =ν 2
dεdωnF (ε) [1 − nF (ε + ω − V )] dseiωs ,
0 0
(6.22)
P †
with Qαβ = k,k0 cα,k cβ,k0 . To make progress, we use the relation

ˆ∞
i
dseiωs = πδ(ω) + p.v.( ), (6.23)
ω
0

where “p.v.” refers to the Cauchy principal value. The finite temperature integrals
ˆ
dεnF (ε) [1 − nF (ε + ω)] = ωθ(ω) + |ω| nB (|ω|) (6.24)

and ˆ
dε [nF (ε) − nF (ε + ω)] = ω (6.25)

are useful as well for dealing with expressions like the one stated in Eq. (6.22). This
results in the exemplary relations (4.52), (4.53) and (4.54) stated in the main text.
Moreover, one obtains the following term which contributes to the “Lamb shift” (4.59)
in the Bloch Redfield Eq. (4.58):
ˆ  
dsTr ρLeads (Q†α0 (t − s)Qα0 (t) − Qα0 (t)Q†α0 (t − s)) = ν 2 (iΛ + πV ). (6.26)

The terms proportional to the bias in the Bloch Redfield Eq. (4.58) arise from
ˆ  
dsTr ρLeads (Q†α0 (t)Qα0 (t − s) + Q†α0 (t − s)Qα0 (t)) = 2πν 2 V (1 + nB (V )), (6.27)

123
Appendices

where nB (V ) is the Bose function which is negligible in the limit T  V . Terms


proportional to temperature in the Bloch Redfield equation arise from
ˆ  
dsTr ρLeads (Q†12 (t)Q12 (t − s) + Q†12 (t − s)Q12 (t)) = 2πν 2 T. (6.28)

The Hamiltonian Hq generating the unitary evolution term in the Markovian Master
Eq. (4.58) in the limit T  V  Λ is derived by noting that we can neglect the
expression

ˆ ˆ∞  
|ω − V | ω ω
p.v. dω nB (|ω − V |) ' p.v. dω + nB (ω)
ω V +ω V −ω
0
ˆ∞
2T 2 x T2
' dx ' 3.29 . (6.29)
V ex − 1 V
0

Therefore, the contributions to the Hamiltonian evolution term in the Bloch Redfield
equation for t0 = 0 arise from terms such as
¨ 
∂ † nF (ε) [1 − nF (ε + ω − V )]
ρ = iν 2 dεdω [Oα0 Oα0 , ρ]
∂t ω

† nF (ε + ω − V ) [1 − nF (ε)]
−[Oα0 Oα0 , ρ] + ...
ω
ˆ 
2 † ω−V
= iν dω [Oα0 Oα0 , ρ] θ(ω − V )
ω

† ω−V
+[Oα0 Oα0 , ρ] θ(−ω + V ) + . . .
ω
2
ν 2Λ † 2
X †
= i [{Oα0 , Oα0 }, ρ] − iν V (1 + ln(Λ/2V )) [[Oα0 , Oα0 ], ρ] + . . .(6.30)
2 α=1

Thus, the Hamiltonian evolution in the master equation is generated by

2
X ν 2Λ X †
Hq = −ν 2
((Λ − iπV )t∗0 Oα0 + H.c.) − {O , Oαβ }
α=1
2 α<β αβ
2
X
2 †
+ν V (1 + ln(Λ/2V )) [Oα0 , Oα0 ], (6.31)
α=1

which in the limit V  Λ leads to the Hamiltonian (4.58) stated in the main text.

124
Appendices

Appendix D: Counting statistics for Majorana box


qubit
Appendix D provides additional details on the derivation of Eq. (4.64) from the evolu-
tion Eq. (4.60) stated in Section 4.4. Starting from the parametrization ρ = 3µ=0 ρµ σµ ,
P

we rewrite the right hand side of Eq. (4.60). The decoherence term can be cast in the
form
X Γ̃αα0
((iγα γα0 )ρ̃t (iγα γα0 ) − ρ̃t ) = −(Γ̃20 + Γ̃21 )ρ1 σ1 − (Γ̃10 + Γ̃21 )ρ2 σ2
α>α0
2
−(Γ̃10 + Γ̃20 )ρ3 σ3 . (6.32)

Using
t∗0 t1 σα ρ̃t + t0 t∗1 ρ̃t σα = Re(t∗0 t1 ) {σα , ρ̃t } + iIm(t∗0 t1 ) [σα , ρ̃t ] , (6.33)

we see that the remaining terms are commutator and anticommutator terms. The
former can be written as
3
X 3
X
−i [hk σk , ρ] = 2 klm hk ρl σm , (6.34)
k=1 k,l,m=1

while the latter obey


X Ωk X X
{σk , ρ} = ρ0 Ωk σk + Ωk ρk σ0 . (6.35)
k
2 k k

By combining these relations, we obtain the matrix elements Ωµν as given by Eq. (4.65).
We now derive the Eqs. (4.90) and (4.91) showing the effect of the inclusion of a finite
hybridization
H̃q → H̃q + ε3 σ3 (6.36)

on the current cross-correlations. The inclusion of this term leads to the modification
Ω → Ω̃ of the matrix given in Eq. (4.65) with
 
0 a1 a2 0
 
 a1 −Γ̃20 − Γ̃21 2ε3 h2 
Ω̃ =  . (6.37)

 a2 −2ε3 −Γ̃10 − Γ̃21 −h1 

0 −h2 h1 −Γ̃10 − Γ̃20

125
Appendices

Here, ε3 is defined as in Eq. (6.36) and aα = 4πV ν 2 Re(t∗0 t1 )(zα − 1), hα =


−4ν 2 ΛRe(t∗0 t1 ) − 4πν 2 V Im(t∗0 t1 )eiχα . The Fano factor derived from the corresponding
evolution equation defined by Ω̃ is given by

0 2Λ2 ν 4 [Re(t∗0 t1 )]4


F = , (6.38)
Λ2 [ν 2 Re(t∗0 t1 )]2 + ε23 |t1 |2 (|t0 |2 + |t1 |2 )


which coincides with the result stated in Eq. (4.90) in Subsection 4.4.5.

126
Appendices

Appendix E: Counting statistics for ABSs

This part of the appendix is focused on giving additional details on the derivation of
the results presented in Section 4.5. To solve the Bloch Redfield Eq. (4.93), we use
Pauli matrices to parametrize the density matrix as

3 X
X 3 X
3
ρ= ρµνλ σµ ⊗ σν ⊗ σλ . (6.39)
µ=0 ν=0 λ=0

The density matrix is defined in the eight-dimensional Hilbert space corresponding to


three fermions. There are two MBSs γα1 and γα2 coupled to each lead α. We choose the
following matrix representation of the resulting Majorana bilinears:

iγ11 γ01 = σ0 ⊗ σ1 ⊗ σ0 , iγ21 γ01 = σ2 ⊗ σ2 ⊗ σ0 , iγ12 γ11 = σ0 ⊗ σ2 ⊗ σ2 ,


iγ01 γ12 = σ0 ⊗ σ3 ⊗ σ2 , iγ01 γ22 = σ1 ⊗ σ2 ⊗ σ0 , iγ01 γ02 = σ3 ⊗ σ2 ⊗ σ0 ,
iγ11 γ02 = σ3 ⊗ σ3 ⊗ σ0 , iγ21 γ02 = σ1 ⊗ σ0 ⊗ σ0 , iγ22 γ21 = σ3 ⊗ σ0 ⊗ σ0 ,
iγ12 γ02 = σ3 ⊗ σ1 ⊗ σ2 , iγ22 γ02 = σ2 ⊗ σ0 ⊗ σ0 . (6.40)

The products of four and six Majorana operators are then expressed by the matrices

γ11 γ12 γ01 γ02 = σ3 ⊗ σ0 ⊗ σ2 ,


γ21 γ22 γ01 γ02 = σ0 ⊗ σ2 ⊗ σ0 ,
γ11 γ12 γ21 γ22 = −σ3 ⊗ σ2 ⊗ σ2 . (6.41)

For the product of all six Majorana operators we have

iγ01 γ02 γ11 γ12 γ21 γ22 = −σ0 ⊗ σ0 ⊗ σ2 . (6.42)

The counting statistics follows from the trace of the generalized density matrix, which
is given by Trq (ρτ (χ1 , χ2 )) = 8ρ000,τ (χ1 , χ2 ). This implies that soley the element ρ000,t
is important, which as we will see couples only to some of the other elements ρµνλ,t .
To express the Bloch Redfield equation in this parametrization of the density matrix,
we use relations of the following type (omitting the ⊗-symbols for brevity) for the

127
Appendices

anticommutators,

{γ11 γ12 , ρ} = 2(−ρ000 σ0 σ2 σ2 − ρ022 σ0 σ0 σ0 − ρ300 σ3 σ2 σ2 − ρ322 σ3 σ0 σ0 − ρ020 σ0 σ0 σ2


−ρ002 σ0 σ2 σ0 − ρ302 σ3 σ2 σ0 − ρ320 σ3 σ0 σ2 + . . . , (6.43)

{γ21 γ22 γ01 γ02 , ρ} = 2(ρ000 σ0 σ2 σ0 + ρ020 σ0 σ0 σ0 + ρ300 σ3 σ2 σ0 + ρ320 σ3 σ0 σ0 + ρ322 σ3 σ0 σ2


+ρ302 σ3 σ2 σ2 + ρ022 σ0 σ0 σ2 + ρ002 σ0 σ2 σ2 + . . . . (6.44)

Furthermore, we make use of relations like

(iγ21 γ01 )ρ̃(iγ21 γ02 ) − (iγ21 γ02 )ρ̃(iγ21 γ01 ) = 2i(−ρ000 σ3 σ2 σ0 + ρ320 σ0 σ0 σ0 + ρ300 σ0 σ2 σ0
−ρ020 σ3 σ0 σ0 + ρ322 σ0 σ0 σ2 − ρ002 σ3 σ2 σ2
−ρ022 σ3 σ0 σ2 + ρ302 σ0 σ2 σ2 + . . . . (6.45)

In analogy to our analytical treatment of the Majorana box qubit, it is convenient to


reparametrize the time variable according to ρt = eθt ρ̃t with

2 X
X 2
θ = 2πν V 2
|tij 2 iχα
α0 | (e − 1). (6.46)
α=1 i,j=1

For Nα = 2 without interference links, t0 = 0, we find that the element ρ000,t is coupled
only to seven of the other elements ρµνλ,t . The differential equation is given by


ξ(t) = (θI + M )ξ(t), (6.47)
∂t

where we have defined the quantity

ξ(t) = (ρ000,t , ρ022,t , ρ300,t , ρ320,t , ρ302,t , ρ020,t , ρ322,t , ρ002,t )T . (6.48)

Furthermore, we have defined

2 X
X 2
θ = 2πν V 2
|tij 2 iχα
α0 | (e − 1). (6.49)
α=1 i,j=1

128
Appendices

Furthermore, we obtain M/2 =

0 c1,− c2,− −b1,− − b2,− a1,− a2,− 0 0


 

 −c1,+ −d1 0 −a1,+ b1,+ − b2,− 0 c2,− a2,− 

−c2,+ 0 −d2 −a2,+ 0 −b1,− + b2,+ c1,− a1,−
 
 
 
 b1,+ + b2,+
 −a1,+ −a2,+ −d1 − d2 −c1,+ −c2,+ 0 0 


 a1,− −b1,− + b2,+ 0 c1,− −d2 0 −a2,+ −c2,+ 

b1,+ − b2,− −d1 −a1,+ −c1,+
 
 a2,− 0 c2,− 0 
 
 0 −c2,+ −c1,+ 0 −a2,+ −a1,+ −d1 − d2 b1,+ + b2,+ 
0 a2,− a1,− 0 c2,− c1,− −b1,− − b1,− 0

(6.50)

for the matrix defined in Eq. (6.47). Here, we have defined the following two quantities
for α = 1, 2
∗ ∗
aα,± = 2πν 2 V Re t11 22 21 12
 iχα
1,α0 (t1,α0 ) − t1,α0 (t1,α0 ) (e ± 1), (6.51)
2
X
bα,± = 2πν V 2
Im(tj1 j2 ∗
1,α0 (t1,α0 ) )(e
iχα
± 1). (6.52)
j=1

Furthermore, we have defined

2
X
cα,± = 2πν 2 V Im(t1j 2j ∗
1,α0 (t1,α0 ) )(e
iχα
± 1), (6.53)
j=1

2
X
dα = 2πν V 2
|tij 2 iχα
1,α0 | e . (6.54)
i,j=1

Taking derivatives of the resulting generating function Trq (ρτ (χ1 , χ2 )) generically yields
a Fano factor of order one as claimed in the main text.
In the special case N0 = 2 and N1,2 = 1 discussed in the main text, we find that
the Bloch Redfield equation couples the element ρ000,t only to ρ320,t . Concretely, this
is expressed by the relation ∂t ξ(t) ˜ ˜
= (θ̃I + M̃ )ξ(t), where we have defined ξ(t) ˜ =
T
(ρ000,t , ρ320,t ) and the matrix θ̃I + M̃ is stated in the main text in Eqs. (4.95) and
(4.96). The eigenvalue of θ̃I + M̃ which vanishes at zero counting parameters yields
the cumulant generating function. In the case N0 = 1 and N1,2 = 2 discussed in
the main text, we find ∂t ξ 0 (t) = (θ0 I + M 0 )ξ 0 (t), where we have defined the quantity
ξ 0 (t) = (ρ000,t , ρ022,t , ρ300,t , ρ322,t )T . The matrix θ0 I + M 0 is the one stated in Eqs. (4.101)
and (4.102) in the main text and its eigenvalue vanishing at zero counting parameters
once again yields the cumulant generating function.

129
Appendices

Appendix F: Further new approaches to MBS detec-


tion

In this part of the appendix, we provide details on derivations supplementing Chapter


5. First, we derive the effective tunnel Hamiltonians stated in Eqs. (5.2) and (5.4) for
the device shown in Fig. 5.1 in Section 5.1. We start from the microscopic tunneling
Hamiltonian
NL X h
X i
HT = λj0 c†0,k γLj e−iφ/2 + H.c.
j=1 k
NR X h
X i
+ (λj1 c†1,k + λj2 c†2,k )γRj e−iφ/2 + H.c. (6.55)
j=1 k

to derive the effective Hamiltonian along the lines of Appendix B. We again assume the
integer N0 in units of the elementary charge to be the equilibrium charge. The tunable
backgate parameter is given by ng = N0 + ∆ng with a small detuning, ∆ng  1.
Furthermore, we again define E± = EC (N0 ± 1 − ∆ng )2 − EC (N0 − ∆ng )2 . Compared
to the derivation of Appendix B, a qualitatively different term arises due to the fact
that a single wire end is being coupled to two different leads. The new type of term is
denoted η j (see below) and describes processes where an electron tunneling from lead 2
into γRi is accompanied by an electron tunneling out of γRi into lead 1. For the effective
tunneling Hamiltonian we obtain
 
1
Heff = P0 HT P0 + P0 HT HT P0 + . . .
−HC
Xh  † i
1 2 12 1 2
= η + η + t12 (iγR γR ) c1,k c2,k0 + H.c.
k,k0
NL X
2 X NR X h i
X 0 j j0 †
+ tjj
α0 (iγ γ
R L )c c
α,k 0,k 0 + H.c. + ... (6.56)
α=1 j=1 j 0 =1 k,k0

130
Appendices

with

E− − E+
η j = λj1 (λj2 )∗
E+ E−
j j ∗
λ (λ ) 4∆ng
= − 1 2
EC 1 − 4∆ng
4∆ng λj1 (λj2 )∗
' − (6.57)
EC

and tij i j ∗
αα0 ' 2iλα (λα0 ) /EC . Thus, we arrive at the effective tunneling Hamiltonians (5.2)
and (5.4).
We now go on to derive the counting statistics. The equation governing the time
evolution of the reduced generalized density matrix ρt (χ1 , χ2 ) in the Hilbert space cor-
j
responding to the Majorana operators γL/R has been stated in the main text in Eq.
(5.5). In the case of the genuine topological Majorana wire, we have NL/R = 1 and the
evolution equation takes on the form

2
∂ 2
X 11 2  iχ
tα0 e α (iγR1 γL1 )ρt (iγR1 γL1 ) − ρt

ρt = 2πν V
∂t α=1
2 1 2

+4πν T η (cos (χ1 − χ2 ) − 1) ρt . (6.58)

The corresponding cumulant generating function reads

2
X
−1 2
τ ln Trq ρτ (χ1 , χ2 ) = 2πν V |t11 2 iχα
α0 | (e − 1)
α=1
2
+4πν T η 1 (cos (χ1 − χ2 ) − 1)
2
(6.59)

and results in Eq. (5.10) as well as the Fano factor |F | = O ∆ng VT .




Now, we turn to Andreev bound states in which case there are generically strong cross-
correlations with the Fano factor |F | = O(1). To demonstrate this, we consider NR = 2,
NL = 1, as discussed in the main text. Then, the zero temperature evolution equation

131
Appendices

takes on the form


2 X
2
∂ X
|tj1
 iχα j 1
ρt = −i[Hq , ρt ] + 2πν 2 V 2
(iγR γL )ρt (iγRj γL1 ) − ρt

α0 | e
∂t α=1 j=1
2
X
21 ∗ 11 ∗ 21
+ eiχα [t11 1 1 2 1 2 1 1 1
α0 (tα0 ) (iγR γL )ρt (iγR γL ) + (tα0 ) tα0 (iγR γL )ρt (iγR γL )]
α=1
2
!
X
21 ∗
+ Im(t11 1 2
α0 (tα0 ) ){iγR γR , ρt } . (6.60)
α=1

The unitary part of the dynamics is generated by an effective Hamiltonian Hq ∼ iγR1 γR2 .
We represent the Majorana bilinears using the Pauli matrices, σ1 = iγR1 γL1 , σ2 = iγR2 γL1
and σ3 = iγR2 γR1 , and parametrize the density matrix via ρt = 3µ=0 ρµ,t σµ . Here, the
P

element ρ0,t , which determines the trace, is coupled only to ρ3,t according to

∂ b  
ξ(t) = θ I + M ξ(t)
b c b (6.61)
∂t

b = (ρ0,t , ρ3,t )T . The matrix M


with ξ(t) c is given by

2
!
21 ∗
c = 4πν 2 V
X 0 (eiχα − 1) Im(t11
α0 (tα0 ) )
M (6.62)
− (eiχα + 1) Im(t11 21 ∗
−eiχα 2j=1 |tj1 2
P
α=1 α0 (tα0 ) ) α0 |

and we have defined θb = 2πν 2 V 2α=1 2j=1 |tj1 2 iχα


P P
α0 | (e − 1). We can find the generating
function, Z ≡ ln Trq (ρτ (χ)), by solving the linear system of differential equations
defined by (6.61). The resulting cumulant generating function in the long time limit is
given by

2 X
X 2
τ −1
ln Trq (ρτ (χ)) = 2πν V 2
− |tj1
α0 |
2

α=1 j=1
v !2 " 2 #
u 2 2
u XX Y X
+t eiχα |tj1
α0 |
2 −4 (eiχα ± 1) ηα (6.63)
,
α=1 j=1 ± α=1

21 ∗
where we have defined the interference parameter ηα = Im(t11 α0 (tα0 ) ). By performing
derivatives with respect to the counting parameters, we arrive at the cross-correlation
formula (5.12).

132
Appendices

Finally, we provide details on the derivation of the number of possible stationary current
0
values for N0 = 2, N1 = 1 for general parameters tjj j1
αα0 . To this end, we set t20 = χ2 = 0
in Eq. (6.63). Calculating the stationary current hI1 i = −iτ −1 ∂χ1 ln Trq (ρτ )|χ=(0,0) ,
shows that in the absence of an interference link there is a single current outcome as
claimed in Section 5.2. The same statement can be shown analogously for N0 = 2,
N1 = 1. When a reference arm is present, t0 6= 0, there are two distinct current
outcomes.

133
Appendices

134
Bibliography

[1] P. A. M. Dirac, "The Quantum Theory of the Electron", Proc. Roy. Soc. A117:
610-624 (1928).

[2] E. Majorana, L. Maiani, “A symmetric theory of electrons and positrons”, In:


Bassani G.F., Council of the Italian Physical Society (eds) Ettore Majorana Sci-
entific Papers. Springer, Berlin, Heidelberg (2006).

[3] S. R. Elliott and M. Franz, “Colloquium: Majorana fermions in nuclear, particle,


and solid-state physics”, Rev. Mod. Phys. 87, 137 (2015).

[4] J. Alicea, “New directions in the pursuit of Majorana fermions in solid state
systems”, Rep. Prog. Phys. 75 076501 (2012).

[5] M. Sato and Y. Ando, “Topological superconductors: a review”, Rep. Prog. Phys.
80, 076501 (2017).

[6] A. Yu. Kitaev, “Unpaired Majorana fermions in quantum wires”, Usp. Fiz. Nauk
(Suppl) 171, 131 (2001).

[7] L. Fu and C. L. Kane, “Superconducting Proximity Effect and Majorana Fermions


at the Surface of a Topological Insulator”, Phys. Rev. Lett. 100, 096407 (2008).

[8] Y. Oreg, G. Refael, and F. von Oppen, “Helical Liquids and Majorana Bound
States in Quantum Wires”, Phys. Rev. Lett. 105, 177002 (2010).

[9] R. M. Lutchyn, J. D. Sau, and S. Das Sarma, “Majorana Fermions and a Topolog-
ical Phase Transition in Semiconductor-Superconductor Heterostructures”, Phys.
Rev. Lett. 105, 077001 (2010).

[10] C. Nayak, S. H. Simon, A. Stern, M. Freedman, and S. Das Sarma, “Non-Abelian


anyons and topological quantum computation”, Rev. Mod. Phys. 80, 1083 (2008).

135
BIBLIOGRAPHY

[11] S. Plugge, A. Rasmussen, R. Egger and K. Flensberg, “Majorana box qubit”,


New J. Phys 12, 012002 (2017).

[12] T. Karzig et. al., “Scalable designs for quasiparticle-poisoning-protected topolog-


ical quantum computation with Majorana zero modes”, Phys. Rev. B 95, 235305
(2017).

[13] A. Cook and M. Franz, “Majorana Fermions in Proximity-coupled Topological


Insulator Nanowires”, Phys. Rev. B 84, 201105(R) (2011).

[14] A. M. Cook, M. M. Vazifeh, and M. Franz, “Stability of Majorana Fermions in


Proximity-Coupled Topological Insulator Nanowires”, Phys. Rev. B 86, 155431
(2012).

[15] R. M. Lutchyn et. al., ”Majorana zero modes in superconductor–semiconductor


heterostructures”, Nature Reviews Materials 3, 52–68 (2018).

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

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


Phys. 82, 3045 (2010).

[18] C. W. J. Beenakker, “Search for Majorana fermions in superconductors”, Annu.


Rev. Con. Mat. Phys. 4, 113 (2013).

[19] M. Leijnse and K. Flensberg, “Introduction to topological superconductivity and


Majorana fermions”, Semicond. Sci. Technol. 27, 124003 (2012).

[20] R. Aguado, “Majorana quasiparticles in condensed matter”, Riv. Nuovo Cim, 40,
523 (2017).

[21] S. Ryu, A. P. Schnyder, A. Furusaki, and A. W. W. Ludwig, “Topological insula-


tors and superconductors: tenfold way and dimensional hierarchy”, New J. Phys
12, 065010 (2010).

[22] A. Altland and M. R. Zirnbauer, “Nonstandard symmetry classes in mesoscopic


normal-superconducting hybrid structures”, Phys. Rev. B 55, 1142 (1997).

[23] A. Kitaev, “Periodic table for topological insulators and superconductors”, AIP
Conf. Proc. 1134, 22–30 (2009).

136
BIBLIOGRAPHY

[24] A. Altland and B. D. Simons, “Condensed Matter Field Theory”, Cambridge


University Press, second edition (2010).

[25] B. Béri and N. R. Cooper, “Topological Kondo Effect with Majorana Fermions”,
Phys. Rev. Lett. 109, 156803 (2012).

[26] A. Altland, B. Béri, R. Egger, and A. M. Tsvelik, “Multichannel Kondo Impurity


Dynamics in a Majorana Device”, Phys. Rev. Lett. 113, 076401 (2014).

[27] Y. V. Nazarov and Y. M. Blanter, “Quantum transport”, Cambridge University


Press (2009).

[28] B. Béri, “Majorana-Klein Hybridization in Topological Superconductor Junc-


tions”, Phys. Rev. Lett. 110, 216803 (2013).

[29] L. Fu, “Electron Teleportation via Majorana Bound States in a Mesoscopic Su-
perconductor”, Phys. Rev. Lett. 104, 056402 (2010).

[30] S. Vijay and L. Fu, “Teleportation-based quantum information processing with


Majorana zero modes”, Phys. Rev. B 94, 235446 (2016).

[31] S. M. Albrecht, E. B. Hansen, A. P. Higginbotham, F. Kuemmeth, T. S. Jespersen,


J. Nygård, P. Krogstrup, J. Danon, K. Flensberg, and C. M. Marcus, “Transport
Signatures of Quasiparticle Poisoning in a Majorana Island”, Phys. Rev. Lett.
118, 137701 (2017).

[32] D. Pekker, C.-Y. Hou, V. E. Manucharyan, and E. Demler, ”Proposal for Coherent
Coupling of Majorana Zero Modes and Superconducting Qubits Using the 4π
Josephson Effect”, Phys. Rev. Lett. 111, 107007 (2013).

[33] D. V. Averin and Yu. V. Nazarov, “Virtual electron diffusion during quantum
tunneling of the electric charge”, Phys. Rev. Lett. 65, 2446 (1990).

[34] A. M. Whiticar, A. Fornieri, E. C. T. O’Farrell, A. C. C. Drachmann, T.


Wang, C. Thomas, S. Gronin, R. Kalla-her, G. C. Gardner, M. J. Man-
fra, C. M. Marcus, and F. Nichele, “Interferometry and coherent single-
electron transport through hybrid superconductor-semiconductor Coulomb is-
lands”, arXiv:1902.07085 (2019).

137
BIBLIOGRAPHY

[35] D. A. Ivanov, “Non-Abelian Statistics of Half-Quantum Vortices in p-Wave Su-


perconductors”, Phys. Rev. Lett. 86, 268 (2001).

[36] C. W. J. Beenakker, “Search for non-Abelian Majorana braiding statistics in


superconductors”, arXiv:1907.06497.

[37] J. Alicea, Y. Oreg, G. Refael, F. von Oppen, and M. P. A. Fisher , “Non-Abelian


statistics and topological quantum information processing in 1D wire networks”,
Nature Physics volume 7, pages 412–417 (2011).

[38] F. Hassler, “Majorana Qubits”, In "Quantum Information Processing. Lecture


Notes of the 44th IFF Spring School 2013", edited by D. P. DiVincenzo, Verlag
des Forschungszentrums Jülich (2013).

[39] D. Aasen, M. Hell, R. V. Mishmash, A. Higginbotham, J. Danon, M. Leijnse, T.


S. Jespersen, J. A. Folk, C. M. Marcus, K. Flensberg, and J. Alicea, “Milestones
Toward Majorana-Based Quantum Computing”, Phys. Rev. X 6, 031016 (2016).

[40] T. D. Stanescu, “Introduction to Topological Quantum Matter & Quantum Com-


putation,” CRC Press, Taylor & Francis Group (2016).

[41] S. Das Sarma, M. Freedman, C. Nayak, “Majorana Zero Modes and Topological
Quantum Computation”, npj Quantum Information 1, 15001 (2015).

[42] C. Knapp, M. Zaletel, D. E. Liu, M. Cheng, P. Bonderson, and C. Nayak, “The


nature and correction of adiabatic errors in anyon braiding”, Phys. Rev. X 6,
041003 (2016).

[43] K. T. Law, P. A. Lee, and T. K. Ng, “Majorana fermion induced resonant Andreev
reflection”, Phys. Rev. Lett. 103, 237001 (2009).

[44] V. Mourik, K. Zuo, S. M. Frolov, S. R. Plissard, E. P. A.Bakkers, and L. P.


Kouwenhoven, “Signatures of Majorana Fermions in Hybrid Superconductor-
Semiconductor Nanowire Devices”, Science 336, Issue 6084, pp. 1003-1007 (2012).

[45] F. Nichele, A. C. C. Drachmann, A. M. Whiticar, E. C.T. O’Farrell, H. J. Suomi-


nen, A. Fornieri, T. Wang, G. C.Gardner, C. Thomas, A. T. Hatke, P. Krogstrup,
M. J. Manfra, K. Flensberg, and C. M. Marcus, “Scaling of Majorana Zero-Bias
Conductance Peaks”, Phys. Rev. Lett. 119, 136803 (2017).

138
BIBLIOGRAPHY

[46] H. Zhang, C. X. Liu, S. Gazibegovic, D. Xu, J. A. Lo-gan, G. Wang, N. van Loo,


J. D. S. Bommer, M. W. A. deMoor, D. Car, R. L. M. Op het Veld, P. J. van
Veldhoven, S. Koelling, M. A. Verheijen, M. Pendharkar, D. J. Penna-chio, B.
Shojaei, J. S. Lee, C. J. Palmstrom, E. P. A. M.Bakkers, S. Das Sarma, and L.
P. Kouwenhoven, “Quantized Majorana conductance”, Nature 556, 74 (2018).

[47] D. Bagrets and A. Altland, “Class D Spectral Peak in Majorana Quantum Wires”,
Phys. Rev. Lett. 109, 227005 (2012).

[48] J. Liu, A. C. Potter, K. T. Law, and P. A. Lee, “Zero-Bias Peaks in the Tunneling
Conductance of Spin-Orbit-Coupled Superconducting Wires with and without
Majorana End-States”, Phys. Rev. Lett. 109, 267002 (2012).

[49] C.-X. Liu, J. D. Sau, Tudor D. Stanescu, and S. Das Sarma, “Andreev bound
states versus Majorana bound states in quantum dot-nanowire-superconductor
hybrid structures: Trivial versus topological zero-bias conductance peaks”, Phys.
Rev. B 96, 075161 (2017).

[50] C. Moore, T. D. Stanescu, and S. Tewari, “Two-terminal charge tunneling: Dis-


entangling Majorana zero modes from partially separated Andreev bound states
in semiconductor-superconductor heterostructures”, Phys. Rev. B 97, 165302
(2018).

[51] Y.-H. Lai, J. D. Sau, and S. Das Sarma, “Presence versus absence of end-to-
end nonlocal conductance correlations in Majorana nanowires: Majorana bound
states versus Andreev bound states”, Phys. Rev. B 100, 045302 (2019).

[52] A. Vuik, B. Nijholt, A.R. Akhmerov, M. Wimmer, “Reproducing topological


properties with quasi-Majorana states ”, SciPost Phys. 7, 061 (2019).

[53] S. M. Albrecht, A. P. Higginbotham, M. Madsen, F. Kuemmeth, T. S. Jespersen,


J. Nygård, P. Krogstrup, and C. M. Marcus, “Exponential protection of zero
modes in Majorana islands”, Nature volume 531, 206-209 (2016).

[54] R. Egger, Unpublished notes.

[55] A. Zazunov, A. Altland, and R. Egger, “Transport properties of the Coulomb–


Majorana junction”, New J. Phys. 16, 015010 (2014).

139
BIBLIOGRAPHY

[56] M. Gau, S. Plugge and R. Egger, “Quantum transport in coupled Majorana box
systems”, PRB 97, 184506 (2018).

[57] T. Jonckheere, J. Rech, A. Zazunov, R. Egger, A. Levy Yeyati, and T. Martin,


“Giant Shot Noise from Majorana Zero Modes in Topological Trijunctions”, Phys.
Rev. Lett. 122, 097003 (2019).

[58] M. Hell, K. Flensberg and M. Leijnse, “Distinguishing Majorana bound states


from localized Andreev bound states by interferometry”, Phys. Rev. B 97,
161401(R) (2018).

[59] S. Rubbert and A. R. Akhmerov, “Detecting Majorana nonlocality using strongly


coupled Majorana bound states”, Phys. Rev. B 94, 115430 (2016).

[60] C. Schrade and L. Fu, “Andreev or Majorana, Cooper finds out”,


arXiv:1809.06370 (2018).

[61] R. Tuovinen, E. Perfetto, R. van Leeuwen, G. Stefanucci, and M. A. Sentef,


“Distinguishing Majorana zero modes from impurity states through time-resolved
transport”, New J. Phys. 21 103038 (2019).

[62] S. Smirnov, “Universal Majorana thermoelectric noise”, Phys. Rev. B 97, 165434
(2018).

[63] K. Yavilberg, E. Ginossar and E. Grosfeld, “Differentiating Majorana from An-


dreev Bound States in a Superconducting Circuit”, Phys. Rev. B 100, 241408(R)
(2019).

[64] J. Manousakis, A. Altland, D. Bagrets, R. Egger, and Y. Ando, “Majorana qubits


in a topological insulator nanoribbon architecture”, Phys. Rev. B 95, 165424
(2017).

[65] P. M. Ostrovsky, I. V. Gornyi, and A. D. Mirlin, “Interaction-Induced Criticality


in Z2 Topological Insulators”, Phys. Rev. Lett. 105, 036803 (2010).

[66] Y. Zhang and A. Vishwanath, “Anomalous Aharonov-Bohm Conductance Oscil-


lations from Topological Insulator Surface States”, Phys. Rev. Lett. 105, 206601
(2010).

140
BIBLIOGRAPHY

[67] J. H. Bardarson and P. W. Brouwer, and J. E. Moore, “Aharonov-Bohm Oscil-


lations in Disordered Topological Insulator Nanowires”, Phys. Rev. Lett. 105,
156803 (2010).

[68] R. Egger, A. Zazunov, and A. L. Yeyati, “Helical Luttinger Liquid in Topological


Insulator Nanowires”, Phys. Rev. Lett. 105, 136403 (2010).

[69] A. Kundu, A. Zazunov, A. L. Yeyati, T. Martin, and R. Egger, “Energy spectrum


and broken spin-surface locking in topological insulator quantum dots,” Phys.
Rev. B 83, 125429 (2011).

[70] J. H. Bardarson and J. E. Moore, “Quantum interference and Aharonov–Bohm


oscillations in topological insulators”, Rep. Prog. Phys. 76, 056501 (2013).

[71] H. B. Nielsen and M. Ninomiya, “The Adler-Bell-Jackiw anomaly and Weyl


fermions in a crystal”, Physics Letters B, Volume 130, Issue 6, p. 389-396.

[72] S. Cho, B. Dellabetta, R. Zhong, J. Schneeloch, T. Liu, G. Gu, M. J. Gilbert, and


N. Mason, “Aharonov–Bohm oscillations in a quasi-ballistic three-dimensional
topological insulator nanowire”, Nature Comm. 6, 7634 (2015).

[73] L. A. Jauregui, M.T. Pettes, L. P. Rokhinson, L. Shi, and Y. P. Chen, “Magnetic


field-induced helical mode and topological transitions in a topological insulator
nanoribbon”, Nature Nanotech. 11, 345 (2016).

[74] V. A. Volkov and V. V. Enaldiev, “Surface states of a system of dirac fermions:


A minimal model”, JETP 149 702-716 (2016).

[75] J. H. Bardarson and R. Ilan, “Transport in Topological Insulator Nanowires”


(2018). In: D. Bercioux, J. Cayssol, M. Vergniory, M. Reyes Calvo (eds) Topo-
logical Matter. Springer Series in Solid-State Sciences, vol 190. Springer, Cham.

[76] F. de Juan, J. H. Bardarson, and R. Ilan, “Conditions for fully gapped topolog-
ical superconductivity in topological insulator nanowires”, SciPost Phys. 6, 060
(2019).

[77] P. Sitthison and T. D. Stanescu, “Robustness of topological superconductivity in


proximity-coupled topological insulator nanoribbons”, Phys. Rev. B 90, 035313
(2014).

141
BIBLIOGRAPHY

[78] F. de Juan, R. Ilan, and J. H. Bardarson, “Robust Transport Signatures of Topo-


logical Superconductivity in Topological Insulator Nanowires”, Phys. Rev. Lett.
113, 107003 (2014).

[79] R. Ilan, J. H. Bardarson, H.-S. Sim, and J. E. Moore, “Detecting perfect trans-
mission in Josephson junctions on the surface of three dimensional topological
insulators”, New J. Phys. 16, 053007 (2014).

[80] G.-Y. Huang and H. Q. Xu, “Majorana fermions in topological-insulator


nanowires: From single superconducting nanowires to Josephson junctions”,
Phys. Rev. B 95, 155420 (2017).

[81] S. Das Sarma, J. D. Sau, and T. D. Stanescu, “Splitting of the zero-bias conduc-
tance peak as smoking gun evidence for the existence of the Majorana mode in a
superconductor-semiconductor nanowire”, Phys. Rev. B 86, 220506(R) (2012).

[82] H. J. Suominen, M. Kjaergaard, A. R. Hamilton, J. Shabani, C. J. Palmstrøm, C.


M. Marcus, and F. Nichele, “Zero-Energy Modes from Coalescing Andreev States
in a Two-Dimensional Semiconductor-Superconductor Hybrid Platform”, Phys.
Rev. Lett. 119, 176805 (2017).

[83] L. A. Jauregui, M. Kayyalha, A. Kazakov, I. Miotkowski, L. P. Rokhinson, and


Y. P. Chen, “Gate-tunable supercurrent and multiple andreev reflections in a
superconductor-topological insulator nanoribbon-superconductor hybrid device”,
Appl. Phys. Lett. 112 (9), 093105 (2018).

[84] M. Kayyalha, M. Kargarian, A. Kazakov, I. Miotkowski, V. M. Galitski, V. M.


Yakovenko, L. P. Rokhinson, and Y. P. Chen, “Anomalous low-temperature en-
hancement of supercurrent in topological-insulator nanoribbon josephson junc-
tions: Evidence for low-energy andreev bound states”, Phys. Rev. Lett. 122,
047003 (2019).

[85] B. M. Terhal, F. Hassler, and D. P. DiVincenzo, “From Majorana fermions to


topological order”, Phys. Rev. Lett. 108, 260504 (2012).

[86] S. Vijay, T. H. Hsieh, and L. Fu, “Majorana Fermion Surface Code for Universal
Quantum Computation”, Phys. Rev. X 5, 041038 (2015).

142
BIBLIOGRAPHY

[87] L. A. Landau, S. Plugge, E. Sela, A. Altland, S. M. Albrecht, and R. Egger,


“Towards Realistic Implementations of a Majorana Surface Code”, Phys. Rev.
Lett. 116, 050501 (2016).

[88] S. Plugge, L. A. Landau, E. Sela, A. Altland, K. Flensberg, and R. Egger,


“Roadmap to Majorana surface codes”, Phys. Rev. B 94, 174514 (2016).

[89] J. Manousakis, C. Wille, A. Altland, R. Egger, K. Flensberg, and F. Hassler,


“Weak measurement protocols for Majorana bound state identification”, Phys.
Rev. Lett. 124, 096801 (2020).

[90] A. N. Jordan and M. Büttiker, “Continuous Quantum Measurement with Inde-


pendent Detector Cross Correlations”, Phys. Rev. Lett. 95, 220401 (2005).

[91] A. A. Clerk, M. H. Devoret, S. M. Girvin, Florian Marquardt, and R. J.


Schoelkopf, “Introduction to quantum noise, measurement, and amplification”,
Rev. Mod. Phys. 82, 1155 (2010).

[92] K. Jacobs and D. A. Steck, “A straightforward introduction to continuous quan-


tum measurement”, Contemporary Physics, 47:5, 279-303, (2007).

[93] H. M. Wiseman and G. J. Milburn, “Quantum Measurement and Control”, Cam-


bridge University Press (2010).

[94] H. Wei and Y. V. Nazarov, “Statistics of measurement of noncommuting quantum


variables: Monitoring and purification of a qubit”, Phys. Rev. B 78, 045308
(2008).

[95] R. Ruskov, A. N. Korotkov, and K. Mølmer, “Qubit State Monitoring by Mea-


surement of Three Complementary Observables”, Phys. Rev. Lett. 105, 100506
(2010).

[96] A. Kitaev, “Anyons in an exactly solved model and beyond”, Ann. Phys. 321, 2
(2006).

[97] W. Belzig, “Full Counting Statistics in Quantum Contacts”, arXiv:cond-


mat/0312180 (2003).

143
BIBLIOGRAPHY

[98] T. Brandes, “Quantensysteme im Nichtgleichgewicht: Einführung”, Vor-


lesungsskript TU Berlin, WS 2016/2017, viewed 24th of March 2020,
<http://www1.itp.tu-berlin.de/brandes/NG2016.pdf >

[99] M. Kindermann, Yu. V. Nazarov, “Full counting statistics in electric circuits”, In:
Yu. V. Nazarov (eds) Quantum Noise in Mesoscopic Physics. NATO Science Series
(Series II: Mathematics, Physics and Chemistry), vol 97. Springer, Dordrecht
(2003).

[100] H.-P. Breuer and F. Petruccione, “The Theory of Open Quantum Systems”, Ox-
ford University Press (2002).

[101] H. Bruus and K. Flensberg, “Many Body Quantum Theory in Condensed Matter
Physics: An Introduction”, Oxford University Press (2004).

[102] A. Franquet and Y. V. Nazarov, “Probability distributions of continuous mea-


surement results for two non-commuting variables and conditioned quantum evo-
lution”, Phys. Rev. A 100, 062109 (2019).

[103] A. Franquet, Y. V. Nazarov, H. Wei, “Statistics of continuous weak quan-


tum measurement of an arbitrary quantum system with multiple detectors”,
arXiv:1804.07639 (2018).

[104] T. Martin and R. Landauer, “Wave-packet approach to noise in multichannel


mesoscopic systems”, Phys. Rev. B 45, 1742 (1992).

144
Acknowledgements

Now the time has come to thank everyone who supported me during my doctoral
studies at the University of Cologne. First and foremost I wish to articulate my deepest
gratitude to my supervisor Alexander Altland. Being a part of his research group was
a truly great experience which taught me a lot. I would also like to pay my special
regards to the Collaborative Research Center 183 “Entangled states of matter” and
the Bonn-Cologne Graduate School of Physics and Astronomy (BCGS), both of which
supported me in many ways.
I gratefully acknowledge the opportunity of an eight months research stay at the center
for quantum devices of the Niels Bohr institute at the University of Copenhagen, which
was enabled and supported by the Collaborative Research Center 183 “Entangled states
of matter”. I wish to express my great thankfulness to Karsten Flensberg. During my
time in Copenhagen he was an amazing host and I enjoyed the collaboration with him
very much.
I am very indebted to Dmitry Bagrets, who has a vast knowledge and understanding of
theoretical physics, and from whom I have learned a lot. Furthermore I want to thank
my other collaborators Reinhold Egger, Carolin Wille, Fabian Hassler and Yoichi Ando.
Last but not least I wish to thank my family, my girlfriend and friends for their amazing
support.

145
Erklärung

Ich versichere, dass ich die von mir vorgelegte Dissertation selbständig angefertigt, die
benutzten Quellen und Hilfsmittel vollständig angegeben und die Stellen der Arbeit =
einschließlich Tabellen, Karten und Abbildungen =, die anderen Werken im Wortlaut
oder dem Sinn nach entnommen sind, in jedem Einzelfall als Entlehnung kenntlich
gemacht habe; dass diese Dissertation noch keiner anderen Fakultät oder Universität zur
Prüfung vorgelegen hat; dass sie = abgesehen von unten angegebenen Teilpublikationen
= noch nicht veröffentlicht worden ist, sowie, dass ich eine solche Veröffentlichung vor
Abschluss des Promotionsverfahrens nicht vornehmen werde. Die Bestimmungen der
Promotionsordnung sind mir bekannt. Die von mir vorgelegte Dissertation ist von Prof.
Dr. Alexander Altland betreut worden.

Datum, Ort Unterschrift

Teilpublikationen:

ˆ J. Manousakis, A. Altland, D. Bagrets, R. Egger, and Y. Ando, “Majorana qubits


in a topological insulator nanoribbon architecture”, Phys. Rev. B 95, 165424
(2017).

ˆ J. Manousakis, C. Wille, A. Altland, R. Egger, K. Flensberg, and F. Hassler,


“Weak measurement protocols for Majorana bound state identification”, Phys.
Rev. Lett. 124, 096801 (2020), [Selected as editors’ suggestion and featured as a
viewpoint in Physics].

Das könnte Ihnen auch gefallen