Lecture Notes in Physics

Alto Osada
Rekishu Yamazaki
Atsushi Noguchi

Introduction
to Quantum
Technologies
Lecture Notes in Physics
Founding Editors
Wolf Beiglböck
Jürgen Ehlers
Klaus Hepp
Hans-Arwed Weidenmüller

Volume 1004

Series Editors
Roberta Citro, Salerno, Italy
Peter Hänggi, Augsburg, Germany
Morten Hjorth-Jensen, Oslo, Norway
Maciej Lewenstein, Barcelona, Spain
Angel Rubio, Hamburg, Germany
Wolfgang Schleich, Ulm, Germany
Stefan Theisen, Potsdam, Germany
James D. Wells, Ann Arbor, MI, USA
Gary P. Zank, Huntsville, AL, USA
The series Lecture Notes in Physics (LNP), founded in 1969, reports new
developments in physics research and teaching - quickly and informally, but with a
high quality and the explicit aim to summarize and communicate current knowledge
in an accessible way. Books published in this series are conceived as bridging
material between advanced graduate textbooks and the forefront of research and to
serve three purposes:
• to be a compact and modern up-to-date source of reference on a well-defined
topic;
• to serve as an accessible introduction to the field to postgraduate students and
non-specialist researchers from related areas;
• to be a source of advanced teaching material for specialized seminars, courses
and schools.
Both monographs and multi-author volumes will be considered for publication.
Edited volumes should however consist of a very limited number of contributions
only. Proceedings will not be considered for LNP.
Volumes published in LNP are disseminated both in print and in electronic
formats, the electronic archive being available at springerlink.com. The series
content is indexed, abstracted and referenced by many abstracting and information
services, bibliographic networks, subscription agencies, library networks, and
consortia.
Proposals should be sent to a member of the Editorial Board, or directly to the
responsible editor at Springer:
Dr Lisa Scalone
Springer Nature
Physics
Tiergartenstrasse 17
69121 Heidelberg, Germany
[email protected]
Alto Osada · Rekishu Yamazaki ·
Atsushi Noguchi

Introduction to Quantum
Technologies
Alto Osada Rekishu Yamazaki
Komaba Institute for Science College of Liberal Arts
The University of Tokyo International Christian University
Meguro-ku, Tokyo, Japan Mitaka-shi, Tokyo, Japan

Atsushi Noguchi
Komaba Institute for Science
The University of Tokyo
Meguro-ku, Tokyo, Japan

ISSN 0075-8450 ISSN 1616-6361 (electronic)

Lecture Notes in Physics
ISBN 978-981-19-4643-1 ISBN 978-981-19-4641-7 (eBook)
https://doi.org/10.1007/978-981-19-4641-7

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Singapore Pte Ltd. 2022
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
For Yuki and Ray
Preface

What is quantum technology? Is it a fancy bandwagon fresh from the academics?

Definitely, this is not the case. Let us trace the history in reverse order from 2022.
Now the quantum technology is known to everybody, thanks to the many-qubit
quantum “computers” developed by several companies and laboratories. Some of
them are available on the cloud platform and are used for the educational and busi-
ness use, while quantum communication and quantum sensing/metrology are also
under intense investigation. Why does quantum computer gather such an attention?
An answer is that such a novel computer based on the unconventional working
principle—quantum mechanics—is expected to outperform conventional comput-
ers, or classical computers as a counterpart of the quantum computer, when a
certain computational task is given. A celebrated quantum algorithm of prime fac-
torization was invented by Peter Shor in 1994 [1]. In 1995, he invented another
landmark of the first proposal of a quantum error correction code [2]. We would
like to stress that this very work ignited the experimental exploration of quantum
computer which had been distinguished from a mere analog computer intolerable
to errors occurring in an analog manner as well. To date, a number of breathtaking
progress have been made in various quantum systems, especially in ion-trap and
superconducting-circuit systems.
Readers might have heard of another fancy word of quantum teleportation in
which a quantum state is transferred between remote parties by a bit tricky use
of the “spooky interaction” of quantum particles, or the quantum entanglement, a
concept coined back in 1935 by Einstein, Podolsky, and Rosen [3]. Quantum tele-
portation was proposed in 1993 by Bennet et al. [4], which was then followed by its
realizations in 1997–1998 [5, 6] using optical photons. These studies lay the basis
of quantum communication. Optical photons are indispensable for the quantum
technologies that demand the transfer of quantum states from one place to a distant
place, which is indeed made possible by the invention of optical fibers with excel-
lent property of extremely low-loss propagation at room temperature. This nature
can also be utilized for another quantum technology of quantum cryptography,
pioneered by Bennett and Brassard in 1984 [7].
Quantum technologies, or even quantum mechanics, have established their rel-
evance in the development of Atomic, Molecular, and Optical (AMO) physics
during the twentieth century. Furthermore, the spectroscopic and control tech-
niques frequently used in AMO physics originate in the Nuclear Magnetic
vii
viii Preface

Resonance (NMR) experiments since the late 1930s with celebrated pioneers such
as Rabi, Ramsey, Bloch, and Purcell. NMR technique is widely applied in our
daily life, for instance to Magnetic Resonance Induction (MRI) tomography and
to identification of materials using chemical shifts of NMR spectra, both of which
can be regarded as early forms of quantum sensing.
What we want to say is that quantum technology is not (or should not be) just a
short-lived trend, a scam business, nor a mathematical game in the Hilbert space,
but a technology that has been pursued substantially with physics and devices in
the history of understanding and controlling matters in real life, at least at the
current technological levels. Therefore, the name of the game is to know what
physical objects we want to address in quantum technologies and how to do that.
Abstracted, perfect (error-corrected) logical qubits are not at our hand—even for
a single logical qubit. Realization of them is a “holy grail,” at the time in 2022.
Various physical systems are used to aim it, and numerous experimental tech-
niques are developed for them. Readers might be overwhelmed when they think
about which quantum system to learn, since different quantum systems can have
different theoretical backgrounds and experimental techniques. However, we, the
researcher in this field, know that there are many theories and concepts in common
among the various quantum systems, such as quantum optics, quantum electronics,
and so on.
The purpose of this book is to summarize such common basics of quantum
technologies as a foundation for cultivating a comprehensive view of the state-of-
the-art quantum systems. The book is roughly divided into three parts.

• In Part I, we describe how the quantum states are represented based on

linear algebra, which is frequently used in quantum technology. We also
cover the fundamentals of quantum mechanics necessary for understanding the
basic quantum manipulations. It is an excerpt of what is to be collected and
summarized in a concise form.
• Part II starts with a description of two-level systems and electromagnetic
waves, and how they interact. We describe the interaction between the fun-
damental quantum systems, two-level systems, electromagnetic waves, and
harmonic oscillators. We also describe the theory relating it to the actual quan-
tum experiments and how they are realized. In particular, we cover details of
the “resonator”, which is an almost indispensable tool in the current quantum
technology.
• In Part III, we describe an overview of the quantum operations, quantum mea-
surements, quantum error corrections, and an introduction to the state-of-the-art
quantum technologies.

Readers who are familiar with analytical mechanics, statistical mechanics, quan-
tum mechanics, and electromagnetism should be able to read this book by using
those textbooks as references. In particular, Part I is a review of quantum mechan-
ics, so readers who are quite familiar with quantum mechanics may start reading
Preface ix

from Part II and continue reading while returning to Part I as necessary. How-
ever, Part I also includes some details of the methods and concepts that are often
seen in quantum technology, so it is recommended that you read them at some
point. In this sense, the subject of this book is suited for the third- and fourth-year
undergraduate students who have been studying physics. We also hope that some
ambitious first- and second-year students try reading. It may be a review for many
graduate students and researchers, but comprehensive coverage of the quantum
technologies compiled here may provide a new insight. In addition, this book may
be used by the researchers who wish to study quantum technologies, but currently
working outside the field. We also warn readers to be aware that there may be
some parts that lack theoretical rigor because we prioritize the description of the
quantum processes as a whole, instead of focusing on the details of the quantum
theories, which you may find in other literature. Lastly, problems are provided in
the last part of each chapter, except for Chaps. 1 and 11. The authors strongly
recommend the readers to tackle with them and check the solutions which are
electronically distributed on the Springer website.
One important topic that we are omitting in this book is the advancement of the
control and fabrication systems, forming a fundamental basis of the state-of-the-art
experimental quantum technologies. Digital controlling systems including Direct-
Digital-Synthesis (DDS) and Field-Programmable Gate Array (FPGA), low-noise
amplifier technologies, and single-photon detectors are only a few of the examples
that have been made in the lab by the researchers, but these high-end products
are now readily available commercially. Ultra-high vacuum systems and dilution
refrigerators to provide an environment with extremely low background noise are
also available. Furthermore, the advancement in the field of material sciences and
micro/nano fabrication technologies have allowed researchers to develop a “De-
signer’s Quantum”, such as Quantum dots, NV-center, and superconducting qubits,
with various novel materials and fabrication techniques. These fundamental bases
for the experiment might differ depending on the quantum system of the choice,
and we encourage the reader to refer to these details in the future.
Lastly, we would like to show our greatest appreciation to people who have
helped us preparing these materials. In writing this book, Masato Shigefuji, Yuki
Nakajima, and Genya Watanabe, members of Noguchi Laboratory in The Uni-
versity of Tokyo, Yoshimi Rokugawa, Sorato Nakano, Yutaro Nakai, and Tsubasa
Karino, members of Shu Lab in International Christian University, cooperated in
calculation check and proofreading of the text. Yuta Masuyama from National
Institute of Quantum Science and Technology gave us very useful advice on the
content. We would love to mention that Akiyuki Tokuno, Sridevi Purushothaman,
and Satish Ambikanithi from Springer worked really hard for the publication of
this book and Akiyuki in particular has made invaluable advice and comments.
We would like to express our gratitude here.

Meguro-ku, Japan Alto Osada

Mitaka-shi, Japan Rekishu Yamazaki
Meguro-ku, Japan Atsushi Noguchi
Contents

Part I Quantum States and Quantum Mechanics

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1 Common Language in Quantum Information . . . . . . . . . . . . . . . . . . 3
1.2 Various Quantum Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Electromagnetic Waves for Quantum Operations . . . . . . . . . . . . . . . 5
1.3.1 Development of Electromagnetic Wave Source . . . . . . . . 6
1.4 Concept of Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2 Linear Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1 Vector Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1.1 Vectors in Three-Dimensional Vector Space . . . . . . . . . . . 13
2.1.2 Inner Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.1.3 Orthonormal Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.1.4 Vector Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.5 Norm of a Vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.6 Outer Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.1.7 Expansion for Multidimensional System . . . . . . . . . . . . . . 16
2.2 Matrix and Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.1 Matrix Element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.2 Transpose Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.3 Matrix Multiplication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.4 Square Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3 Eigenvectors and Eigenvalues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4 Summary of Vector Characteristics in Index Notation . . . . . . . . . . 21
3 Wavefunction and Notations in Quantum Mechanics . . . . . . . . . . . . . . . 25
3.1 Equation of Motion in Classical and Quantum Mechanics . . . . . . 25
3.2 Wavefunction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.1 Inner Product of Wavefunction . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.2 Continuous and Discretized Wavefunction . . . . . . . . . . . . . 28
3.3 Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.4 Dirac Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.5 Matrix Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

xi
xii Contents

3.6 Properties of Wavefunction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3.7 Composite System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.7.1 Operator in Composite System . . . . . . . . . . . . . . . . . . . . . . . . 39
3.8 Examples of Notation for Frequently Used Quantum State . . . . . 40
3.9 Quantum State Representation: Qubit . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.9.1 One-Qubit state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.9.2 Bloch Sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.9.3 Two-Qubit State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.10 Eigenvalue Equation in Quantum Mechanics . . . . . . . . . . . . . . . . . . . 50
3.10.1 Example: Atom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.11 Operator Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.11.1 Operator Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.11.2 Hermitian Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.11.3 Projection Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.11.4 Unitary Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.11.5 Pauli Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.12 Density Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.12.1 Example of a Mixed State . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.12.2 General Properties of Density Operator . . . . . . . . . . . . . . . 63
3.12.3 Density Operator of Composite System and its
Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.13 Commutator and Anti-Commutator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.13.1 Heisenberg’s Uncertainty Principle . . . . . . . . . . . . . . . . . . . . 67
4 Time Evolution in Quantum System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.1 Time-Independent Schrödinger Equation . . . . . . . . . . . . . . . . . . . . . . . 73
4.2 Time Evolution in Terms of Unitary Operator . . . . . . . . . . . . . . . . . 76
4.3 Three Pictures in Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.3.1 General Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.3.2 Schrödinger Picture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.3.3 Heisenberg Picture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.3.4 Example: Harmonic Oscillator in Schrödinger
and Heisenberg Picture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.3.5 Dirac Picture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.4 Heisenberg Equation of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.5 von Neumann Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.6 Unitary Transformation to a Rotating Frame . . . . . . . . . . . . . . . . . . . 91
4.7 Driven Two-Level System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5 Perturbation Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.1 Time Independent Perturbation Theory . . . . . . . . . . . . . . . . . . . . . . . . 101
5.1.1 Zeeman Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.2 Treatment of Time-Dependent Perturbation . . . . . . . . . . . . . . . . . . . . 107
5.2.1 Time-Dependent Perturbation Expansion . . . . . . . . . . . . . . 114
Contents xiii

Part II Harmonic Oscillator, Qubit and Coupled Quantum

Systems
6 Harmonic Oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
6.1 Harmonic Oscillator and Its Hamiltonian . . . . . . . . . . . . . . . . . . . . . . 123
6.2 Electromagnetic Waves and Ladder Operators . . . . . . . . . . . . . . . . . 124
6.2.1 Quantization of Electromagnetic Waves . . . . . . . . . . . . . . . 124
6.2.2 Ladder Operators of a Harmonic Oscillator . . . . . . . . . . . . 125
6.3 Quantum States of a Harmonic Oscillator . . . . . . . . . . . . . . . . . . . . . . 126
6.3.1 Coherent State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.3.2 Schrödinger’s Cat State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
6.3.3 Squeezed State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
6.3.4 Thermal State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
6.4 Photon Correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
6.4.1 Amplitude Correlation/First-Order Coherence . . . . . . . . . 133
6.4.2 Intensity Correlation/Second-Order Coherence . . . . . . . . 134
6.5 Wigner Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
6.5.1 General Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
6.5.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7 Two-level System and Interaction with Electromagnetic Waves . . . . . 141
7.1 Two-level System, Spin, and Bloch Sphere . . . . . . . . . . . . . . . . . . . . 141
7.2 Interaction Between Two-level System
and Electromagnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.2.1 Spontaneous and Stimulated Emission . . . . . . . . . . . . . . . . 145
7.3 Two-level Systems and Electromagnetic Waves in Practice . . . . . 146
7.3.1 Two-level Systems in Nuclear Magnetic
Resonance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
7.3.2 Two-level Systems in Atomic Gases . . . . . . . . . . . . . . . . . . 147
7.3.3 Two-level Systems in Solid-state Quantum Defects . . . . 148
7.3.4 Two-level System in an Optically Controlled
Semiconductor Quantum Dot . . . . . . . . . . . . . . . . . . . . . . . . . 148
7.3.5 Two-level System in Superconducting Circuit . . . . . . . . . 149
7.4 Dynamics and Relaxations of Two-level Systems . . . . . . . . . . . . . . 149
7.4.1 Bloch Equations and Relaxations . . . . . . . . . . . . . . . . . . . . . 149
7.4.2 Rabi Oscillation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
7.4.3 Ramsey Interference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
7.4.4 Spin Echo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
8 Electromagnetic Cavities and Cavity Quantum
Electrodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
8.1 Properties of Cavities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
8.1.1 Q Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
8.1.2 Finesse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
xiv Contents

8.2 Measurement of a Cavity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162

8.2.1 Reflection and Transmission Measurements . . . . . . . . . . . 163
8.2.2 Actual Measurement Systems . . . . . . . . . . . . . . . . . . . . . . . . . 164
8.3 Input-Output Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
8.3.1 Propagating Mode and Input-Output Theory . . . . . . . . . . . 166
8.3.2 Single-Port Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
8.3.3 Dual-Port Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
8.4 Cavity Quantum Electrodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
8.4.1 Jaynes–Cummings Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
8.4.2 Tavis–Cummings Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
8.4.3 Weak and Strong Coupling Regimes . . . . . . . . . . . . . . . . . . 175
8.4.4 Dispersive Regime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
8.4.5 Waveguide-Coupled Cavity QED System . . . . . . . . . . . . . . 179
9 Various Couplings in Quantum Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
9.1 Interaction Hamiltonians . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
9.1.1 Jaynes–Cummings and Anti-Jaynes–Cummings
Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
9.1.2 Beam-Splitter and Two-Mode-Squeezing
Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
9.1.3 Interaction Between Qubits . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
9.1.4 Nonlinear Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
9.2 Atomic Ions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
9.2.1 Atom-Light Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
9.2.2 Sideband Transitions: Optomechanics with an Ion . . . . . 192
9.2.3 Phonon–Phonon Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
9.3 Superconducting Circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
9.3.1 Quantum-Mechanical Treatment of an LC
Resonator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
9.3.2 Superconducting Quantum Bit . . . . . . . . . . . . . . . . . . . . . . . . 197
9.3.3 Coupling Between a Transmon and an LC
Resonator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
9.4 Optomechanical Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
9.4.1 Brief Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
9.4.2 Optomechanical Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
9.4.3 Linearized Optomechanical Interaction . . . . . . . . . . . . . . . . 203
9.5 Hybrid Quantum Systems and Cooperativity . . . . . . . . . . . . . . . . . . . 204

Part III Quantum Information Processing and Quantum

Technologies
10 Basics of Quantum Information Processing . . . . . . . . . . . . . . . . . . . . . . . . 211
10.1 Quantum Gates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
10.1.1 Single-Qubit Gates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
10.1.2 Two-Qubit Gates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
10.1.3 Clifford and Non-Clifford Gates . . . . . . . . . . . . . . . . . . . . . . 218
Contents xv

10.2 Quantum Circuit Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219

10.3 Measurement and Imperfections of Quantum States . . . . . . . . . . . . 220
10.3.1 Projective and Generalized Measurement . . . . . . . . . . . . . . 220
10.3.2 State Tomography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
10.3.3 Fidelity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
10.3.4 Estimation of Gate Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
10.4 Essential Idea of Quantum Error Correction . . . . . . . . . . . . . . . . . . . 227
10.4.1 Stabilizer Formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
10.4.2 Three-Qubit Repetition Code . . . . . . . . . . . . . . . . . . . . . . . . . 229
10.4.3 Nine-Qubit Shor Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
10.4.4 Advanced Quantum Error Correction Codes . . . . . . . . . . . 231
10.5 DiVincenzo Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
11 Quantum Technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
11.1 Quantum Computer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
11.1.1 Grover’s Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
11.1.2 Phase Estimation Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . 237
11.1.3 Shor’s Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
11.2 Quantum Key Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
11.2.1 BB84 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
11.3 Quantum Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
11.3.1 Ramsey Interference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
11.3.2 Quantum Sensing with Entanglement . . . . . . . . . . . . . . . . . 241
11.4 Quantum Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
11.5 Quantum Internet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242

Appendix A: Position and Momentum Representations . . . . . . . . . . . . . . . . . 245

Appendix B: Unitary Transformation to a Rotating Frame . . . . . . . . . . . . . 247
Appendix C: Extraction of a Two-Level System from a Three-Level
System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
Appendix D: Quantum Theory of Hydrogen Atom . . . . . . . . . . . . . . . . . . . . . . 253
Appendix E: Master Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
Appendix F: Schrieffer–Wolff Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . 269
Appendix G: Derivation of the SWAP Gate from the Heisenberg
Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
Appendix H: Cavity Cooling of a Mechanical Mode . . . . . . . . . . . . . . . . . . . . . 275
Appendix I: Entangled States and Quantum Teleportation . . . . . . . . . . . . . 281
Appendix J: Quantum No-Cloning Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
 Part I
Quantum States and Quantum Mechanics
Introduction
1

1.1 Common Language in Quantum Information

Since the early 1900s, quantum mechanics has been constructed through numer-
ous discussions by renowned physicists such as Einstein, Dirac, Heisenberg, and
Schrödinger and has made significant progress. The seemingly non-trivial physical
phenomena predicted by the theory have plagued many physicists (and students),
but no observations contradicting quantum mechanics have yet to be found in this
long history.
The recent development of quantum mechanics over the past 50 years has been
greatly influenced by the maturity of experimental methods and new measurements.
Observation and manipulation of nuclear spin states by developing nuclear magnetic
resonance (NMR), spectroscopic experiments of atoms and ions in an ultra-high
vacuum, and verification of quantum properties of light through the development
of high-performance lasers and nonlinear optics have all contributed to form firm
foundations of quantum manipulation and technologies. Various quantum operations
have also been confirmed even in solid-state devices, which have been said to be diffi-
cult to guarantee their quantum properties. These experiments have made it possible
to verify quantum mechanics from various angles. Through these developments, the
quantum object, long been considered as a target of “observation” has been trans-
formed into that of “manipulation”. In other words, human beings have acquired a
way to freely manipulate the quanta, the smallest entity possible in this world, which
was previously only intended to be seen.
Meanwhile, the field of quantum information science has made great progress in
the last 20 years or so. With the development of experimental methods, the operation
of each physical system has improved, and the unique characteristics of these physical
systems have become more and more apparent. Information technology that makes
use of quantum properties, has been greatly developed. In quantum information,
the object holding the bit of information is the quantum entity called quantum bit

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 3
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7_1
4 1 Introduction

Fig. 1.1 Seemingly different physical systems can all be seen as qubits (quantum bits) from the
perspective of the quantum information

or qubit for short. The qubit allows the expansion of the computational capability
by incorporating the quantum features, such as superposition and entanglement.
This research on quantum information has given rise to some new findings. It has
become clear that even seemingly completely different physical systems have many
similarities by regarding quantum as the origin of information (Fig. 1.1). So far, each
quantum system has progressed independently. However, by taking a bird’s-eye view
of these various quantum systems in the framework of quantum information, it has
become possible to comprehensively understand different physical systems under a
unified common language.
This book provides an attempt to capture the characteristics of various quan-
tum physical systems within the framework of quantum information and to pro-
mote the development of quantum systems by providing a bird’s-eye view to young
researchers. It may still take some time before quantum information technology
advances into society. However, there are many proposals for the development of
quantum devices that greatly exceed existing technologies such as quantum comput-
ers, quantum communications/internet, quantum simulations, and quantum sensors.
Many scientists believe that these novel technologies will significantly change even
the way society should be. We believe that new ideas from young people will be the
key to the breakthrough of these quantum technologies, along with existing quantum
technologies. We would like to welcome all young students, demonstrating your
strength for the future development of the quantum technology.
1.3 Electromagnetic Waves for Quantum Operations 5

1.2 Various Quantum Systems

There are many quantum systems, such as ion traps, cold atoms, quantum dots, and
superconducting circuits, but of course, the strengths and weaknesses of the system
depends on the characteristics of each quantum system. For example, just as an
existing computer is composed of individual parts, CPU, memory, hard drive, and
Internet adapter, each quantum system could have a different “utility” suited to its
characteristics.
There are also many quantum systems that have some common elements. For
example, similarity in the manufacturing method (fabrication), the equipment used
for control/operation, and the environment (temperature, etc.) in which the device is
placed. In those similar systems, learning operations and techniques is often easier
than a completely new system. In addition, even quantum systems that look very
different at first glance are often said to be similar when the core elements are
considered. Also, as mentioned previously, in the framework of quantum information,
it is often easier to capture the characteristics and similarities of various quantum
systems by considering all the systems as different qubits. For these reasons, it is
expected that comprehensive understanding will be promoted by learning various
quantum systems.
With the maturation of quantum operations, the development of quantum systems
called “hybrid quantum systems” is being promoted worldwide. In hybrid quantum
systems, several quantum systems are combined in order to extend the strengths of
each quantum system and suppress its weaknesses. For example, photons in optical
fibers allow long-range quantum state transmission, but instead, it is very difficult to
keep photons (i.e., quantum states) in one place. By coupling an optical fiber with an
atom or ion trap, the quantum state is retained in the atom or trapped-ion, and when
necessary, it is taken out as a photon and sent to the fiber, thereby providing memory
and transmission functions. Understanding the pros and cons of each quantum system
to come up with complementary combination is a key for the development of a novel
hybrid quantum system.

1.3 Electromagnetic Waves for Quantum Operations

First of all, a physical phenomenon is the movement or influence of an object through

some kind of force, and we all learned that there are four forces in the world, weak
nuclear, strong nuclear, electromagnetic, and gravitational force, with corresponding
elementary particles called weak bosons, gluons, photons, and gravitons that are
responsible for the propagation of each force. Of these, the electromagnetic force
is the closest to our everyday life. Of course, without the nuclear force, the atoms
that make up the basic structure of the world would be disjointed and everything
falls apart, but everyday phenomena such as friction, sound propagation, and phase
transitions between gases, liquids, and solids, as well as the functions in the electrical
circuits, are all results of the electromagnetic interaction. Furthermore, nearly all
6 1 Introduction

physiological processes in the living body are phenomena that are manifested by
electromagnetic force. In other words, photons intervene in all these phenomena.
The quantum systems that we are manipulating are no exception. Quantum oper-
ations are performed using electric and magnetic fields and various types of electro-
magnetic waves. Therefore, “quantum optics”, which treats electromagnetic waves
themselves as a quantum object can be considered as the basis of quantum oper-
ations. Important concepts in the quantum optics appear in various forms of the
quantum research. There are many wonderful literature on quantum optics, so this
book will only introduce the basics, but we encourage students to explore these lit-
erature yourself. In addition, quantum optics has been studied mainly for “light”
among electromagnetic waves as its origin, but due to the maturation of dilution
refrigerators and the remarkable development of superconducting quantum circuits
in recent years, electromagnetic waves in the microwave region have become another
focus. These are called “microwave quantum optics” and have grown as a discipline
with features different from those in the optical domain. In addition, research and
development of electromagnetic waves in the THz region, which was difficult to
generate and measure, has been steadily progressing in recent years. There is still
little research in the context of quantum operations, but we look forward to future
developments.

1.3.1 Development of Electromagnetic Wave Source

The role of electromagnetic waves in quantum operations is described, and one can
see that the quality of quantum operations using electromagnetic waves is directly
related to the issue of “how to manipulate the electromagnetic waves precisely”. The
history of the quantum operation has always progressed with the development of
microwave oscillators and lasers as the electromagnetic wave source, so let’s touch
on the development of the electromagnetic wave source, although it is a bit technical
subject.
The state basis used for the qubit is often the energy states as shown in Fig. 1.2, and
the transition between the states is manipulated by electromagnetic waves. Common
electromagnetic waves used are the RF (MHz range), microwave (GHz range), and
the optical region (on the order of 100 THz). The frequency of the electromagnetic
E −E
wave corresponds to the energy difference of the transition as ω0 = e  g . To be
more specific with the physical system, NMR is mainly operated by RF, microwaves
are mainly used in superconducting circuits and spin systems, and optical light is
used in atoms and ions, for example.
One of the characteristics of electromagnetic waves, which is important exper-
imentally but hardly ever talked about in a textbook, is the “Linewidth” of elec-
tromagnetic waves. Considering an electromagnetic wave oscillator that radiates an
electromagnetic wave ω0 which is in resonance with an atomic transition in Fig. 1.2,
the output frequency of the ideal oscillator is a single frequency as shown on the
left side of Fig. 1.3. However, in an actual oscillator, there is a certain amount of
frequency or phase fluctuation, and the frequency staggers around ω0 . In the same
1.3 Electromagnetic Waves for Quantum Operations 7

Fig. 1.2 Electromagnetic waves are used for transitions between energy eigenstates

Fig. 1.3 (Left) Ideal single frequency oscillator output, (Right) Fluctuating oscillator output

figure on the right, the fluctuation is drawn as a histogram. The frequency spread
δω of the oscillator measured within a specific time is called the “Linewidth” of the
electromagnetic wave, and it is an index of how stable the electromagnetic wave is.
Of course, a stable electromagnetic wave oscillator can perform stable quantum oper-
ations. In other words, narrowing the linewidth of the electromagnetic wave source
can be a cornerstone of the higher operability and refined control of the quantum
systems.
The origin of the linewidth broadening of electromagnetic waves may be from a
variety of reasons, such as oscillator performance, the vibration of the experiment
table, and the Fourier limit (which spreads in frequency due to the short pulse time of
electromagnetic waves). In addition, the performance required for electromagnetic
waves varies greatly depending on the linewidth of the target energy level and the
operation time, so it is necessary to always consider what kind of oscillator perfor-
mance is required for each experiment.
RF and microwaves have long been used commercially for communications and
sensors, and various oscillators, filters, and related parts are well-developed. For the
electromagnetic wave generation, ultra-stabilized oscillators, arbitrary
8 1 Introduction

waveform generators, pulse generators, and various mixers and filters are used. Many
high-performance peripheral products such as high-speed oscilloscopes, spectrum
analyzers, and network analyzers are readily available in the market. In that sense, it is
probably safe to say that the RF and microwave oscillators have already accumulated
extremely high technological capabilities. In addition, with the recent evolution of
microwave quantum optics and the maturation of superconducting circuits with quan-
tum computers in mind, novel RF/microwave devices are rapidly developed. These
developments include ultra-low temperature circulators without magnetic field, arbi-
trary waveform generators making full use of FPGAs (Field Programmable Gate
Arrays) and DDSs (Direct Digital Synthesizers), noiseless quantum amplifiers using
parametric oscillations, and traveling wave parametric amplifiers that have a wider
bandwidth. Many of these novel devices are developed for quantum operations, par-
ticularly the large-scale quantum information processing being one of the strong
drives.
The linewidth of electromagnetic waves is very low, easily less than 1 Hz, in com-
mercialized RF/microwave oscillators, whereas it can be of several GHz in commonly
available lasers, such as semiconductor lasers. Since the frequency of light itself is
much higher than that of RF/microwave, one may think it is unfair to compare them
directly and better to compare the relative fluctuation. However, the relative fluctua-
tion δω/ω0 , normalized with the carrier frequency ω0 , is also overwhelmingly better
for common microwave oscillators than the lasers. The narrow linewidth and high
performance of RF/microwave oscillators are understandable because of the long
history of RF/microwave oscillators for their active commercial use.
Under such circumstances, it was necessary to make a home-made
high-performance laser suited for the precise control of the quantum states. Thus, the
evolution of high-performance lasers has always been cultivated with the develop-
ment of quantum technologies. Many laser developments such as ultra-stable laser
with a linewidth less than 1 Hz, ultra-wideband optical combs, and integrated laser
circuits on chip have emerged as the foundation of quantum operations. The ultimate
electromagnetic wave operation is indispensable for the future of quantum technol-
ogy, and these research and development will continue to be a part of the adventure
in the development of quantum technologies.

1.4 Concept of Temperature

We have mentioned that different quantum systems may have various similarities.
One of the important experimental parameters is the energy of the quantum systems
and the electromagnetic waves used to manipulate them. It is described below that
the importance comes from the temperature of the environment in which the quantum
system is installed at the time of the experiment.
Many quantum operations require a single quantum operation, where a sin-
gle quantum of energy is exchanged in various forms. The system can be eas-
ily explained by using the energy levels of an atom. For example, as shown in
Fig. 1.4, there is an atom with two different energy levels, the ground state |g and
1.4 Concept of Temperature 9

Fig. 1.4 Transition between the energy states. The origin of the energy state transition is not
limited to the signal light intended in the experiment, but also includes noise from the surrounding
environment such as phonons and spins in thermally excited solids

the excited state |e, with corresponding energy E g and E e , respectively. Normally,
the electron is in the ground state, but when the electromagnetic wave with energy
E 0 = E e − E g is absorbed, the electron makes a transition to the excited state
(Fig. 1.4). The electromagnetic wave here may be the electromagnetic wave intended
to use in the experiment, but if it is not, the atom may be mistakenly excited.
Let’s consider where this “unintended electromagnetic wave” possibly comes
from. We consider the temperature T of the environment surrounding the atoms and
bosonic particles (e.g., photons) that are in thermal equilibrium with the environment.
Using the frequency ω of the bosonic particle, the energy of the particle is written
as E Boson = ω, and the average number of particles n is given following the
Bose–Einstein distribution as

1
n = ω
, (1.4.1)
e −1
kB T

where  = h/2π and kB are the Dirac constant and the Boltzmann constant, respec-
tively. The Dirac constant is a variation of the Planck constant, and it is customary
to use the Dirac constant rather than the Planck constant in many quantum research
areas, including quantum optics and quantum information sciences.
When the effective energy of the thermal environment E env = kB T is sufficiently
higher than the boson particle energy E particle = ω, this equation is approximated
as
kB T
n  . (1.4.2)
ω
As can be seen from this equation, the average number of bosonic particles n,
proportional to T , is swelling in the environment surrounding the quantum system.
Elementary excited particles such as phonons in a solid, and even in a vacuum
10 1 Introduction

Fig. 1.5 Temperature of the environment and number of microwave photons inside a cavity. (Left)
At T = 300 K, average number of microwave photons is n  600, which intervene with the
signal microwave photon. (Right) At T = 10 mK, n  0.02 and hardly any microwave photons
are present except the signal microwave

Table 1.1 Relationship between the electromagnetic wave frequency and experimental environ-
ment temperature
EM wave Frequency Effective Temp. Environment Temp.
(ω0 /2π) (Teff ) (T )
Microwave 10 GHz 500 mK 10 mK
Optical light 200 THz 10,000 K 300 K

chamber, bosonic particles such as photons are zipping around with an average
number of n. The thermal bosonic particles that roar in these environments with
a temperature of T become “unintended electromagnetic wave”, an obstacle for the
quantum operation. To eliminate the bosonic particles from the environment at the
electromagnetic wave frequency ω0 = E 0 /, the environment temperature T needs
to be sufficiently low to satisfy n  1.
An example of the effect of the environment temperature is shown in Fig. 1.5. On
the left, there is a microwave cavity at temperature T = 300 K with a quantum sample
inside. In this cavity, the average number of microwave photons with frequency
ω0 /2π = 10 GHz is about n MW   600, and the signal microwave photon intended
to control the sample is buried under these background photons. On the other hand,
when the sample is installed in a dilution refrigerator where T = 10 mK, the number
of background microwave photons drops to n MW   0.02, enabling a quantum
control of the sample with the signal microwave.
From above, the requirement of the environmental temperature reads

ω0
T  Teff = . (1.4.3)
kB

This equation tells us that the effect of the thermal bosonic particles is mitigated
when the environmental temperature is sufficiently lower than the effective tem-
perature of electromagnetic waves (Teff = ω 0
kB ). In many of the quantum control
experiments, electromagnetic waves such as microwaves and optical light are often
used. In Table 1.1, we summarize the frequency, effective temperature, and common
environmental temperature for the microwave and optical light.
1.4 Concept of Temperature 11

As can be seen from this table, the effective temperature of a microwave is about
500 mK, requiring an extremely low-temperature environment often achieved using
a dilution refrigerator. For the optical light, in contrast, the effective temperature
is about 10,000 K, and many experiments can be performed at room temperature
300 K. One can intuitively understand that n  1 is fulfilled for optical light in
a room temperature from the fact that all the walls in our rooms are not “glowing”
(with optical light) from the heat.
As described above, the suited environmental temperature of the quantum device
greatly differs depending on the frequency of the electromagnetic wave used. Also,
it should be remembered that due to this stringent environmental condition for the
microwave experiment, the techniques and technologies related to experiments differ
greatly between microwaves and light despite the fact that both are electromagnetic
waves.
Linear Algebra
2

In the introductory quantum mechanics course, we all learn the behavior of wave-
function ψ(x) satisfying the Schrödinger equation. The wavefunction can actually be
represented as a vector in a vector space, the so-called Hilbert space. In the quantum
technology, we focus on the manipulation of the quantum state, where the vector in the
Hilbert space is transformed, measured, and manipulated through various physical
interactions. These dynamics can be represented simply by using the terminologies
and techniques of the vector space, using the language of linear algebra. Here, we
review the basics of the vector space by introducing frequently used terminologies.

2.1 Vector Space

2.1.1 Vectors in Three-Dimensional Vector Space

V and W
 are vectors in a three-dimensional vector space, represented as

V = Vx x̂ + Vy ŷ + Vz ẑ (2.1.1)
 = Wx x̂ + W y ŷ + Wz ẑ.
W (2.1.2)
A vector is formed with unit vectors x̂, ŷ, ẑ, called basis and its components
(scalar amplitudes) Vx , Vy , Vz and Wx , W y , Wz , respectively. In an actual calcu-
lation, it is convenient to use a matrix representation of a vector, column matrix, to
represent the vector.

Supplementary Information The online version contains supplementary material available at

https://doi.org/10.1007/978-981-19-4641-7_2.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 13
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7_2
14 2 Linear Algebra

Unit vectors can be represented using the column matrix as

⎛ ⎞ ⎛ ⎞ ⎛ ⎞
1 0 0
x̂ ≡ ⎝0⎠ , ŷ ≡ ⎝1⎠ , ẑ ≡ ⎝0⎠ (2.1.3)
0 0 1

and the vectors can be represented as

⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
1 0 0 Vx
V = Vx ⎝0⎠ + Vy ⎝1⎠ + Vz ⎝0⎠ = ⎝Vy ⎠ (2.1.4)
0 0 1 Vz
⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
1 0 0 Wx
 = Wx ⎝0⎠ + W y ⎝1⎠ + Wz ⎝0⎠ = ⎝W y ⎠ .
W (2.1.5)
0 0 1 Wz

2.1.2 Inner Product

Inner product of two vectors is defined as

V · W
 ≡ Vx Wx + Vy W y + Vz Wz . (2.1.6)

A scalar value is calculated from the inner product of two vectors. Using the matrix
representation, the inner product can be defined as a vector multiplication of the
transpose of the first vector
 
V T ≡ Vx Vy Vz , (2.1.7)
and the second vector as
⎛ ⎞
  Wx
V · W
 = V T W
 = Vx Vy Vz ⎝W y ⎠ = Vx Wx + Vy W y + Vz Wz . (2.1.8)
Wz

It is clear from the matrix representation that the two vectors, V and W
 , need to have
the same dimension to form an inner product.

2.1.3 Orthonormal Basis

When the inner product of two vectors is zero, V1 · V2 = 0. V1 and V2 are said to
be orthogonal. Also, a vector whose length is of unity is said to be normalized. In
many cases, the basis vectors such as x̂, ŷ, and ẑ in Fig. 2.1 (left) are chosen to be
orthogonal and normalized, referred to as orthonormal set of basis. Representing
the basis x̂, ŷ, and ẑ with x̂i , i = 1, 2, 3, the orthonormal set of basis follows the
condition
x̂i · x̂ j = δi j , (2.1.9)
2.1 Vector Space 15

Fig. 2.1 (Left) Orthonormal

set of basis x̂, ŷ, and ẑ.
(Right) Non-orthonormal set
of basis x̂  , ŷ  , and ẑ  . One
can choose any basis to
represent V ; however, it is
more convenient and
intuitive to use the
orthonormal set of basis in
most cases

where δi j is called the Kronecker delta and has the property


0, if i = j
δi j = (2.1.10)
1, if i = j.

Generally speaking, the basis does not require to be normalized or orthogonal as

shown in Fig. 2.1 (right). One can use a non-orthonormal set of basis x̂  , ŷ  , and ẑ  to
represent V ; however, it is much more intuitive to use the orthonormal set of basis,
and also it makes many of the calculations much simpler.

2.1.4 Vector Components

Given a vector V , one can obtain the amplitude of each component using the inner
product with the normalized basis. For example, the x-component of V , Vx , can be
calculated as
Vx = x̂ · V . (2.1.11)
Using the matrix representation, Vx is calculated as
⎛ ⎞
  Vx
Vx = 1 0 0 ⎝Vy ⎠ . (2.1.12)
Vz

2.1.5 Norm of a Vector

From each vector component Vi , one can calculate the “length” of the vector, called
Norm of the vector. Norm can be defined using the inner product as

norm(V ) ≡ V · V = Vx2 + Vy2 + Vz2 . (2.1.13)

Using this norm, one can calculate a unit vector v̂, which is pointing in the direction
of an arbitrary vector V as
V
v̂ = . (2.1.14)
norm(V )
16 2 Linear Algebra

Using this unit vector, the original vector V can easily be represented as

V = norm(V )v̂. (2.1.15)

2.1.6 Outer Product

For the case of inner product, a scalar is calculated from two vectors. Outer product
of vectors, on the other hand, calculates a matrix from two vectors. Using the same V
and W defined earlier, the matrix representation of the outer product of two vectors
is written as
⎛ ⎞
Vx  
V ⊗ W ≡ V W  T = ⎝ Vy ⎠ W x W y Wz (2.1.16)
Vz
⎛ ⎞
Vx Wx Vx W y Vx Wz
= ⎝ Vy W x Vy W y Vy Wz ⎠ . (2.1.17)
Vz Wx Vz W y Vz Wz

In this example, a matrix with a size 3 × 3 is calculated from two three-dimensional

vectors; however, in general, the vector dimensions do not have to be the same.
The outer product of m- and n-dimensional vectors results in a matrix of dimension
m × n.

2.1.7 Expansion for Multidimensional System

All the examples so far are shown with three-dimensional vectors, familiar to many
of us. In the Hilbert space, where a quantum state is represented, the space dimension
can be much larger than three, requiring systematic expansion methods that are simple
and easy to calculate.
By introducing the subscript index to expand the basis as (x̂, ŷ, ẑ) → (x̂i , i =
1, 2, 3, 4, . . .), a vector in a multidimensional space can be represented as

V = Vi x̂i = V1 x̂1 + V2 x̂2 + · · · (2.1.18)

i
 =
W Wi x̂i = W1 x̂1 + W2 x̂2 + · · · , (2.1.19)
i

where Vi and Wi are the components of the vector in the basis xi .

2.2 Matrix and Operator 17

2.2 Matrix and Operator

As shown earlier, the outer product of two vectors forms a matrix. m × n-dimensional
matrix  is represented as
⎛ ⎞
A11 A12 . . . A1n
⎜ A21 A22 . . . A2n ⎟
⎜ ⎟
 ≡ ⎜ . .. . . .. ⎟ , (2.2.1)
⎝ .. . . . ⎠
Am1 Am2 . . . Amn

where Ai j is the matrix element for the corresponding row i and column j.

2.2.1 Matrix Element

A matrix element, for example, A21 , can be calculated from matrix  with unit
vectors,
⎛ ⎞ ⎛ ⎞
1 0
⎜0⎟ ⎜1⎟
⎜ ⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎟
x̂1 = ⎜0⎟ , x̂2 = ⎜0⎟ (2.2.2)
⎜ .. ⎟ ⎜ .. ⎟
⎝.⎠ ⎝.⎠
0 0
as
⎛ ⎞
⎛ ⎞ 1
A11 A12 ... A1n ⎜ ⎟
 ⎜ ... A2n ⎟ ⎜0⎟
⎜ A21 A22 ⎟ ⎜0⎟
A21 = 0 1 0 ... 0 ⎜ . .. .. .. ⎟ ⎜ ⎟ . (2.2.3)
⎝ .. . . . ⎠⎜ .⎟
⎝ .. ⎠
Am1 Am2 ... Amn
0
Following the recipe above, matrix element Ai j can be obtained as a matrix
product of (row vector)·(matrix)·(column vector) as

Ai j = x̂iT Â x̂ j . (2.2.4)

2.2.2 Transpose Matrix

Transpose matrix  T of m × n-dimensional matrix  is

⎛ ⎞
A11 A21 . . . Am1
⎜ A12 A22 . . . Am2 ⎟
⎜ ⎟
 T ≡ ⎜ . .. . . .. ⎟ . (2.2.5)
⎝ .. . . . ⎠
A1n A2n . . . Amn
18 2 Linear Algebra

 T is n × m-dimensional and the matrix element AiTj satisfies

AiTj = A ji . (2.2.6)

2.2.3 Matrix Multiplication

Matrix multiplication results in a new matrix, for example, Ĉ = Â B̂. Matrix element
Ci j of this new matrix Ĉ satisfies

Ci j = Aik Bk j . (2.2.7)
k

As can be seen in this equation, when the product Ĉ is l × n-dimensional and Â

is l × m-dimensional, the other matrix B̂ has to be m × n-dimension to fulfill the
dimensional requirement of the product.
Matrix element of the transpose of Ĉ, Ĉ T , can be found by simply swapping the
indices as
CiTj = C ji = A jk Bki . (2.2.8)
k

2.2.4 Square Matrix

A matrix with the same number of rows and columns, m × m matrix, is called a
square matrix. Most of the matrices used in the quantum technologies are square
matrices. Here, we review some characteristics and special types of square matrices
often used.

Trace of a Matrix
Trace of a square matrix  is defined as

Tr  ≡ Aii = A11 + A22 + · · · (2.2.9)

i

and is a sum of the diagonal matrix elements Aii . When the product of matrices,
Ĉ = Â B̂, is a square matrix, the trace of Ĉ is given as

Tr Ĉ = Cii = Ai j B ji . (2.2.10)
i i, j
2.2 Matrix and Operator 19

Unit Matrix
One of the special square matrices is the unit matrix Iˆ, where all the matrix elements
in the diagonal are 1 with the rest of the elements being 0s. Three-dimensional unit
matrix is
⎡ ⎤
100
Iˆ ≡ ⎣0 1 0⎦ . (2.2.11)
001
The matrix element Ii j of unit matrix follows

Ii j = δi j . (2.2.12)

Given a set of orthonormal basis x̂i , the unit matrix can be obtained from outer
products of basis as

Iˆ = x̂i x̂iT (2.2.13)

i
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
1 0 ... 0 0 0 ... 0 0 0 ... 0
⎢0 0 . . . 0⎥ ⎢0 1 . . . 0⎥ ⎢0 0 . . . 0⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
= ⎢. .. . . .. ⎥ + ⎢ .. .. . . .. ⎥ + · · · + ⎢ .. .. . . .. ⎥ (2.2.14)
⎣ .. . . .⎦ ⎣. . . .⎦ ⎣. . . .⎦
0 0 ... 0 0 0 ... 0 0 0 ... 1
⎡ ⎤
1 0 ... 0
⎢0 1 . . . 0⎥
⎢ ⎥
= ⎢. .. . . .. ⎥ . (2.2.15)
⎣ .. . . .⎦
0 0 ... 1

Matrix as an Operator
An operator acting on a vector maps the vector to another vector in the same vector
space. The operator can be represented with a matrix, with the operation given as a
matrix multiplication with a vector.
For example, when a operator  maps vector V to another vector W
 , the operation
can be written as
 = Â V .
W (2.2.16)
Let us take a look at the operation focusing on the vector and matrix element. The
operator  and vector V are given as
⎛ ⎞
A11 A12 A13
 = ⎝ A21 A22 A23 ⎠ (2.2.17)
A31 A32 A33
⎛ ⎞
V1
V = ⎝V2 ⎠ . (2.2.18)
V3
20 2 Linear Algebra

 is calculated as a product
Resulting vector W
⎛ ⎞⎛ ⎞ ⎛ ⎞
A11 A12 A13 V1 A11 V1 + A12 V2 + A13 V3
 = Â V = ⎝ A21 A22 A23 ⎠ ⎝V2 ⎠ = ⎝ A21 V1 + A22 V2 + A23 V3 ⎠ . (2.2.19)
W
A31 A32 A33 V3 A31 V1 + A32 V2 + A33 V3

Using an index notation, the operation above can be given as

Wi = Ai j V j . (2.2.20)
j

2.3 Eigenvectors and Eigenvalues

There is an important set of vectors for an operator, called eigenvectors. When an

operator acts on a vector to create a new vector, where the new vector is a non-zero
scalar multiple of the original vector, it is called an eigenvector of the operator. It is
much simpler to see it in an equation form. Vector V is an eigenvector of an operator
 when the following equation is satisfied:

 V = a V , (2.3.1)

where a is a scalar factor. This scalar factor is called an eigenvalue of the operator Â,
corresponding to the eigenvector V . From the form of Eq. 2.3.1, where the resulting
vector has the same dimension as the original vector, we identify that the operator Â
needs to be a square matrix.
The eigenvalues of an arbitrary operator  can be found as a solution of a char-
acteristic function
c(λ) ≡ det( Â − λ Iˆ) = 0, (2.3.2)
where Iˆ is a unit matrix and det( Ô) represents determinant of matrix Ô. From the
expression of the characteristic function, there can be multiple eigenvalues λ, and it is
known that any operator has at least one eigenvalue and its corresponding eigenvector
as a consequence of the formal algebra. There is also a case, where one might find
multiple eigenvalues λ having the same value. For example, for a matrix
⎛ ⎞
1 0 2
 = ⎝0 −1 0⎠ , (2.3.3)
2 0 1
the characteristic equation is
⎛ ⎞
1−λ 0 2
c(λ) = det ⎝ 0 −1 − λ 0 ⎠ = 0, (2.3.4)
2 0 1−λ
2.4 Summary of Vector Characteristics in Index Notation 21

from which we can find the eigen values λ = 3, −1, −1 with corresponding eigen-
vectors
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
1 1 0
V1 = ⎝0⎠ , V2 = ⎝ 0 ⎠ , V3 = ⎝1⎠ . (2.3.5)
1 −1 0
The multiplets of the eigenvalues (here, λ = −1) are called degenerate eigenvalues.
It is important to notice that even if the eigenvalues are degenerate, the eigenvectors
can be orthogonal, which is easily checked as
⎛ ⎞
  0
V2 · V3 = 1 0 −1 ⎝1⎠ = 0. (2.3.6)
0

2.4 Summary of Vector Characteristics in Index Notation

Throughout the text, certain vector characteristics are used frequently. It is conve-
nient to use an index notation, which provides a compact representation, instead
of the bulky matrix representation. Following is the summary of the important and
frequently used vector characteristics using index notation.

Vectors in Multidimensional Space

Orhonormality x̂i x̂ j = δi j (2.4.1)

Vector component Vi = x̂i · V (2.4.2)

Norm of a vector norm(V ) = Vi2 (2.4.3)
i

Normalized vector Vi2 = 1 (2.4.4)

i

Inner product V · W
 = Vi Wi (2.4.5)
i
Outer product (V ⊗ W
 )i j = Vi W j . (2.4.6)

For the outer product, matrix element of the product V ⊗ W

 is given. When vectors
 
V and W are of dimension m and n, indices are given as i = 1 . . . m, j = 1 . . . n,
respectively.
22 2 Linear Algebra

Matrix element
 = Â V and Ĉ = Â B̂,
For vectors and matrices satisfying W

Matrix element and basis Ai j = x̂iT Â x̂ j (2.4.7)

Matrix element of products of matrices Ci j = Aik Bk j (2.4.8)
k

Transpose matrix CiTj = C ji = A jk Bki (2.4.9)

k

Trace Tr C = Cii = Ai j B ji (2.4.10)

i i, j

Unit matrix Iˆ = x̂i x̂iT (2.4.11)

i

Matrix-Vector multiplication Wi = Ai j V j . (2.4.12)

j

Problems

Problem 2-1 Normalize the following vector V to find the normalized vector V  .

(i) V = x̂ + ŷ.
(ii) V = 2 x̂ + 3 ŷ − 4ẑ.

Problem 2-2

(i) Show V1 = √1 x̂ + √1 ŷ and V2 = √1 x̂

are orthogonal. − √1 ŷ
2 2 2 2
(ii) Find a vector V̂2 = V2x x̂ + V2y ŷ which is orthogonal to V1 =
3x̂ − 2 ŷ and normalized.

⎛ ⎞
sin θ cos φ
Problem 2-3 Prove that a vector V = ⎝ sin θ sin φ ⎠ is a normalized vector for
cos θ
arbitrary chosen θ and φ.
Problem 2-4 Prove⎛that ⎞ three vectors
⎛ ⎞ ⎛ ⎞
1 −2 2
 1 ⎝−1⎠, and c = 1 ⎝−2⎠
a = 13 ⎝2⎠, b= 3 3
2 2 1
form a orthonormal set of basis.
2.4 Summary of Vector Characteristics in Index Notation 23

Problem 2-5 V1 ⎞

Find an outer product ⎛ ⊗ V2 of two vectors,
⎛ ⎞ 3
1 ⎜ 1⎟
V1 = ⎝0⎠ and V2 = ⎜ ⎟
⎝ 4 ⎠.
2
2
Problem 2-6 There is a “sibling” of outer product, called the Kronecker product.
For matrices  of m × n-dimension and B̂, the Kronecker product
is defined as
⎛ ⎞
A11 B̂ · · · A1n B̂
⎜ ⎟
 ⊗ B̂ = ⎝ ... . . . ... ⎠ . (2.4.13)
Am1 B̂ · · · Amn B̂
   
01 1 0
Find the Kronecker product, Â ⊗ B̂ for Â= and B̂= .
10 0 −1
Problem 2-7 For square matrices Â, B̂, and Ĉ, prove the following
 properties of
matrix trace using the index notation, e.g., Tr A = i Aii
 
(i) Tr  + B̂ = Tr  + Tr B̂.
 
(ii) Tr c  = c Tr Â.
   
(iii) Tr  B̂ = Tr B̂  .
     
(iv) Tr  B̂ Ĉ = Tr B̂ Ĉ  = Tr B̂ ÂĈ (Cyclic rule).

Problem 2-8 Use the property Ii j = δi j to prove Iˆ Â = Â Iˆ = Â.

Problem 2-9 Find the eigenvalues λi and corresponding eigenvectors Vi of fol-
lowing matrices:
 
22
(i) Â = .
13
 
42
(ii) B̂ = .
13
(iii) For a matrix Â, its eigenvalues λi and the trace of the matrix
has the following relations,

 =
(a) Tr(Ĉ)   i λi
(b) det Ĉ = i λ = i.
Show these relations are satisfied for matrix  and B̂.
24 2 Linear Algebra

Problem 2-10 As we shall see in details later, for a two-level system with a drive,
the Hamiltonian of the system can be written as
 

Ĥ = 2

2 . (2.4.14)
2 − 2

Find the energy eigenvalues λi of this Hamiltonian.

Wavefunction and Notations in
Quantum Mechanics 3

In this chapter, we review the properties of wavefunction and different representa-

tions. The introductory quantum mechanics courses usually start out with the wave-
mechanics based on wavefunction representation, while it is often more convenient
to use the matrix representation in the context of quantum information sciences.
However, there are always some exceptions, and we often switch back and forth
between two representations; thus, it is important for readers to be familiar with both
representations.
It is also important to be able to represent the composite system. For example, for
the case of two − qubit gate in a quantum computer, one wavefunction consisting
of two quantum bits of information is input to perform the gate. To do this, we need
a representation of a wavefunction, composed of several individual quantum states.
We also cover the representation for operator , as self-evident, which performs the
operation to a quantum state wavefunction for various tasks. Finally, we cover the
commutation relation, which is one of the unique and important properties of the
operators in quantum mechanics. We also relate the commutation relation to the
celebrated Heisenberg uncertainty principle.

3.1 Equation of Motion in Classical and Quantum Mechanics

In both classical and quantum mechanics, we are considering the motion of some
“object”; however, it is always the quantum mechanics which we stumble on. In

Supplementary Information The online version contains supplementary material available at

https://doi.org/10.1007/978-981-19-4641-7_3.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 25
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7_3
26 3 Wavefunction and Notations in Quantum Mechanics

Fig. 3.1 Classical motion (left) and Quantum motion (right)

classical mechanics, as shown in Fig. 3.1 (left), we often consider the motion of a
particle with mass m in a potential V (x). Applying Newton’s law, we can describe
the motion as
∂ V (x)
m ẍ = − , (3.1.1)
∂x
2
where ẍ = ddtẋ = ddt 2x .
On the other hand, in quantum mechanics, we are forced to deal with some liq-
uidy object so-called “wavefunction” as if it is spilled out of a cup as shown in
Fig. 3.1 (right). The motion of this wavefunction is governed by the celebrated
Schrödinger equation as

∂(x, t) 2 ∂ 2 (x, t)
i =− + V (x)(x, t). (3.1.2)
∂t 2m ∂x2

Form of the wavefuntion (x, t) can be determined by solving this equation of

motion.
There is a vast difference between these two mechanics. In classical mechan-
ics, the position of the object x(t) can be directly obtained from the equation of
motion. Once x(t) is found, other physical observables such as velocity v(t) = ddtx ,
p 2
momentum p = mv, and kinetic energy K E = 21 mv 2 = 2m are readily available
from simple calculation. In quantum mechanics, however, we learn that the wave-
function is analytically solvable in only a few special cases, and even when we are
super-lucky to be able to solve the actual wavefunction (x), we are still told that
“the particle is somewhere under the curve”. What a nightmare...
In order to extract useful information or some physical observables of interest,
one would have to use the “operators” on the wavefunction. For many students, this
quantum-mechanics-specific procedure and mysterious overall composition seem to
be an obstacle. The founders of the quantum mechanics were not ignorant about these
difficulties. They came up with numerous tools to make the quantum mechanics sim-
pler and easy to handle. One of the tools adopted is linear algebra. All the quantum
states are described as a vector in a quantum vector space, called the Hilbert space.
The operations such as time evolution and extraction of physical observables are
performed with linear algebraic operations on the state vector. In other words, seem-
ingly peculiar wavefunction can be systematically handled by using the framework
3.2 Wavefunction 27

of linear algebra. In this chapter, we apply the basics of the linear algebra covered in
the previous chapter and review the fundamentals of quantum mechanics.

3.2 Wavefunction

3.2.1 Inner Product of Wavefunction

An arbitrary one-dimensional wavefunction (x) can be considered as a vector in a

functional space. When states are described as vectors, first step is to define an inner
product of vectors within the vector space. With two wavefunctions  and , inner
product of two wavefunctions is defined as
 ∞
 ∗ (x)(x)d x, (3.2.1)
−∞

where  ∗ is a complex conjugate of .

Wavenfunction (x) can be written in terms of linearly independent basis ψi (x)
as

(x) = c1 ψ1 (x) + c2 ψ2 (x) + · · · + cn ψn (x)

n
= ci ψi (x), (3.2.2)
i=1

where ci is complex coefficient and basis ψi satisfies the orthonormal condition,

 ∞
ψi∗ (x)ψ j (x)d x = δi j . (3.2.3)
−∞

Using the relation above, the inner product of  with itself reads
 ∞  ∞ 
n
 ∗ (x)(x)d x = ci∗ c j ψi∗ (x)ψ j (x)d x
−∞ −∞ i, j=1


n 
n
= ci∗ c j δi j = |ci |2 ,
i, j=1 i=1

and the normalization condition of the wavefunction (x) now reads,


n
|ci |2 = 1. (3.2.4)
i=1
28 3 Wavefunction and Notations in Quantum Mechanics

Fig. 3.2 a Continuous wavefunction; b discretized wavefunction

3.2.2 Continuous and Discretized Wavefunction

Since the definition of the inner product and normalization is set, let’s take a look at an
example of wavefunction. In an introductory quantum mechanics course, a particle
wavefunction (x) in one-dimensional space is often described as in Fig. 3.2a.
Probability density of the particle is described as absolute square of the wavefunction
(||2 =  ∗ ), and probability of finding a particle in a ≤ x ≤ b, P(a ≤ x ≤ b) can
be calculated as
 b
P(a ≤ x ≤ b) = d x|(x)|2 . (3.2.5)
a
In the context of quantum information, the continuous wavefunction as shown
above may be used, but we often deal with discretized wavefunctions as in Fig. 3.2b.
In this example, we have a in the vicinity of x = a and another hump b around
x = b. We often describe these wavefunctions as separate states.

a = (x) for xa − δx < x < xa + δx (3.2.6)

b = (x) for xb − δx < x < xb + δx, (3.2.7)

where δx is a large enough distance to cover each wavefunction near x = a, b.

Each wavefunction can be normalized to define
a (x)
ψa =  (3.2.8)
a+δx
a−δx d x|a |
2

b (x)
ψb =  . (3.2.9)
b+δx
b−δx d x|b |
2

Since there is no spatial overlap between ψa and ψb ,

 ∞
d x ψi∗ ψ j = δi j (3.2.10)
−∞

is satisfied, allowing these states to form a orhonormal basis.

3.2 Wavefunction 29

Fig. 3.3 Discretized

wavefunction in energy
spectrum

Using these wavefunctions as basis, the original wavefunction  can be written

as
(x) = ca ψa (x) + cb ψb (x), (3.2.11)
where
 ∞
ca = d x ψa∗ (x)(x)
−∞
∞
cb = d x ψb∗ (x)(x)
−∞

and satisfies |ca |2 + |cb |2 = 1. If a continuous wavefunction can be segmented with

respect to some parameter (in this example, ψa and ψb were segmented in position
x), one could make them into a discrete set of wavefunctions.
In many of the quantum bits or quantum states shown later, wavefunctions are not
described as a spatial function, but of energy function as shown in Fig. 3.3 for the case
of an atom. In these cases, one could use a similar step to define the wavefunction
 in terms of its basis ψi as

 = cg ψg + ce ψe , (3.2.12)

where subscripts g and e refer to the ground and excited states. From the normaliza-
tion condition
||2 = |cg |2 + |ce |2 = 1, (3.2.13)
we see the state is in the ground and excited state with probability |cg and |ce |2 ,
|2
respectively. We can describe the state as if the particle is located in discrete states.
As shown above, many quantum systems have continuous energy structures often
separated into segments. If the segments are far enough that they do not have overlap,
one can easily describe these “bands” as discrete states by integrating each segment.
Discretized states are often defined as 0-state, ψ0 ≡ ψg , and 1-state ψ1 ≡ ψe to use
them as a basis for a quantum bit

 = c0 ψ0 + c1 ψ1 . (3.2.14)
30 3 Wavefunction and Notations in Quantum Mechanics

Fig. 3.4 Operators in action.

Various operations can be
performed on the quantum
state  for different tasks

3.3 Operator

Operators act on wavefunctions to perform linear mapping of wavefunctions. They

may perform different tasks, such as time evolution, measurement, and finding expec-
tation values of physical observables (Fig. 3.4). In the early part of this text, we use
the accent symbol  to write the operators. In the later part of the text, we often dis-
miss the accent symbol to avoid the clumsiness of the equations. We expect readers
to be familiarized enough with different operators, so you can identify the operators
without the accent symbol.
When an operator  is acted on a wavefunction , the new wavefunction  
emerged as a result of the operation can be written as

  = Â. (3.3.1)

The operator for a physical observable O is often denoted as Ô. The expectation
value of O for a state  is calculated as
 ∞
O =  ∗ (x) Ô(x)d x. (3.3.2)
−∞

As mentioned above, since the operator changes the state, one can see that the
state can be manipulated by selecting an appropriate operator. Also, using the same
effect, it can be seen that the operator is also used to extract a physical quantity from
a wavefunction.

3.4 Dirac Notation

In this section, we introduce an alternative representation for the wavefunctions,

known as the Dirac notation, used in the quantum information. The Dirac notation
often simplifies the equations and is also quite intuitive once you get a hang of it. We
only cover the notation briefly and skip some of the details and formal descriptions.
A quantum state described by a wavefunction  can be written as

 → |. (3.4.1)
3.4 Dirac Notation 31

Since the state is using the later part of the “braket” symbol  , it is called a ket-
vector. There exist a dual space vector for the ket-vector, called bra-vector denoted
as
 ∗ → |. (3.4.2)
When the wavefunction depends on specific parameters, such as quantum numbers
(n, l, m, ...), it is customary to include them in the notation as

nlm... → |nlm.... (3.4.3)

Using this Dirac notation, the inner product of two quantum states | and | is
denoted as
 ∞
 ∗ (x)(x)d x → |. (3.4.4)
−∞
Some careful readers may notice that the open-braket | and | are vectors, and
ones with closed-braket | are a scalar. We review common quantum concepts
using this convenient notation.
Wavefunction | can be written in terms of linearly independent basis |ψi  as

| = c1 |ψ1  + c2 |ψ2  + · · · + cn |ψn 

n
= ci |ψi . (3.4.5)
i=1

Also, a corresponding bra-vector can be written as

| = c1∗ ψ1 | + c2∗ ψ2 | + · · · + cn∗ ψn |

 n
= ci∗ ψi |. (3.4.6)
i=1

Orhonormality condition of the basis, the component or the probability amplitude

of a wavefunction, and the wavefunction normalization can be written in a simple
form as

ψi |ψ j  = δi j (3.4.7)

n 
n
ψi | = ψi | c j |ψ j  = c j ψi |ψ j 
j=1 j=1

n
= c j δi j = ci (3.4.8)
j=1
32 3 Wavefunction and Notations in Quantum Mechanics


n 
n
| = ci c∗j ψ j |ψi  = ci c∗j δ ji
i, j=1 i, j=1
 n
= |ci |2 = 1, (3.4.9)
i=1

respectively. Norm of the wavefunction is also given as

 
norm() = | = |ci |2 . (3.4.10)
i i

Similarly, the action of operator can also be written using the Dirac notation.
Upon operation of  on a state |, new state is given as

|   = Â|. (3.4.11)

Expectation value of an observable O of wavefunction  is

O = | Ô = | Ô|. (3.4.12)

The expectation value is (of course) a scalar and it can be confirmed with the closed-
braket as described previously. The last equality may seem obvious, but in | Ô, it
explicitly shows that the state is the resulting state of the operation Ô on |. Further
details are shown in Appendix A.

3.5 Matrix Representation

Following the Dirac notation, let us introduce the matrix representation of the quan-
tum state here. In the wavefunction representation, the function is regarded as a
state “vector”, which might not be intuitive at first. The quantum state can also be
described in the form of a matrix, which is much familiar to us as a “vector”.
The ket-vector | is usually represented as a vertical vector. Using the orthonor-
mal basis set |ψi , the wavefunction can be written as

| = c1 |ψ1  + c2 |ψ2  + · · · + cn |ψn 

⎛ ⎞ ⎛ ⎞ ⎛ ⎞
1 0 0
⎜0⎟ ⎜1⎟ ⎜0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
= c1 ⎜ . ⎟ + c2 ⎜ . ⎟ + · · · + cn ⎜ . ⎟
⎝ .. ⎠ ⎝ .. ⎠ ⎝ .. ⎠
0 0 1
⎛ ⎞
c1
⎜ c2 ⎟
⎜ ⎟
= ⎜ . ⎟. (3.5.1)
⎝ .. ⎠
cn
3.5 Matrix Representation 33

On the other hand, | is written as a horizontal vector with the vector elements
taking the complex conjugate, written as


| = c1∗ c2∗ · · · cn∗ . (3.5.2)
The inner product can be calculated by arranging bra and ket-vectors as a standard
matrix multiplication as
⎛ ⎞
c1
 ⎜ c2 ⎟
⎜ ⎟
| = c1∗ c2∗ · · · cn∗ ⎜ . ⎟ (3.5.3)
⎝ .. ⎠
cn

n
= |ci |2 = 1.
i=1

The last line shows the normalization condition of the wavefunction. We can also
write the general inner product as
⎛ ⎞
c1
 ⎜c  ⎟
⎜ 2⎟
|   = c1∗ c2∗ · · · cn∗ ⎜ . ⎟ (3.5.4)
⎝.⎠.
cn

n 
n
= ci∗ ci = (ci ci∗ )∗
i=1 i=1
 ∗
=  | ,

where |   =   |∗ is shown in the last line. The outer product of wavefunc-
tions is
⎛ ⎞
c1
⎜c  ⎟ 
⎜ 2⎟
|    | = ⎜ . ⎟ c1∗ c2∗ · · · cn∗ (3.5.5)
⎝ .. ⎠
cn
⎛  ∗  ∗ ⎞
c1 c1 c1 c2 · · · c1 cn∗
⎜c c∗ c c∗ · · · c cn∗ ⎟
⎜ 2 1 2 2 2 ⎟
=⎜ . .. . . . ⎟, (3.5.6)
⎝ . . . . .. ⎠
cn c1∗ cn c2∗ · · · cn cn∗
34 3 Wavefunction and Notations in Quantum Mechanics

whose matrix element |    | ij
is also calculated as

|    | ij
= ψi |  |ψ j 
⎛ ⎞
⎛ ⎞ 0
c1 c1∗ c1 c2∗ ··· c1 cn∗ ⎜ . ⎟
⎜ .. ⎟
⎜ c2 cn∗ ⎟
 ∗ c2 c2∗ ···
⎜ c2 c1 ⎟⎜ ⎟
= 0···1······ 0 ⎜ . .. .. ⎟ ⎜ ..
⎜1⎟
⎟
⎝ .. . . ⎠ ⎜.⎟ .
⎝ .. ⎠
cn c1∗ cn c2∗ · · · cn cn∗
0
= ci c∗j , (3.5.7)

where |ψi, j  are basis vectors. Operators are represented as matrices; for example,
an operator  acting on a three-basis space is written as
⎛ ⎞
A11 A12 A13
 = ⎝ A21 A22 A23 ⎠ . (3.5.8)
A31 A32 A33

The operation on a state is simply written as a matrix operating on a vector as

|   = Â|
⎛ ⎞⎛ ⎞
A11 A12 A13 c1
= ⎝ A21 A22 A23 ⎠ ⎝c2 ⎠
A31 A32 A33 c3
⎛ ⎞
c1 A11 + c2 A12 + c3 A13
= ⎝c1 A21 + c2 A22 + c3 A23 ⎠ . (3.5.9)
c1 A31 + c2 A32 + c3 A33

Also, from Eq. (3.5.2),

  | (3.5.10)
  
= c1∗ A∗11 + c2∗ A∗12 + c3∗ A∗13 c1∗ A∗21 + c2∗ A∗22 + c3∗ A∗23 c1∗ A∗31 + c2∗ A∗32 + c3∗ A∗33
⎛ ∗ ⎞
 A11 A∗21 A∗31
= c1∗ c2∗ c3∗ ⎝ A∗12 A∗22 A∗32 ⎠
A∗13 A∗23 A∗33
= | † . (3.5.11)

In the last line, † ≡ (  T )∗ is a Hermite conjugate, equivalent to the complex

conjugate of transpose of Â. Here, bra-vector of |   is obtained as

|   = Â| →   | = | † . (3.5.12)

3.6 Properties of Wavefunction 35

Using |   =   |∗ described previously, it is quite simple to obtain other

useful relations,

| = |∗ (3.5.13)

| Â| = | † |∗ . (3.5.14)

Since many operators are square matrices in quantum system analysis, it is impor-
tant to define the trace of operators. It is simply the same as a matrix trace,
 
Tr  = ψi | Â|ψi  = Aii . (3.5.15)
i i

3.6 Properties of Wavefunction

Below is a summary of the main properties of the wavefunction. We did not derive all
of them due to the limited space, but should be easy to derive. There is a one-to-one
correspondence with “Vectors in Multidimensional Space” in Sect. 2.4. If you are
new to the material below, it is quite helpful to  see the parallel between
 the two.
In the summary below, |   = Â|, | = i ci |ψi , |a  = i ai |ψi , |b  =
 
 
i bi |ψi , |  = i ci |ψi , Ĉ = Â B̂ are used.

Orthonormal basis ψi |ψ j  = δi j (3.6.1)

Component of wavefunction ci = ψi | (3.6.2)

Norm norm() = |ci |2 (3.6.3)
i

Wavefunction normalization |ci | = 1
2
(3.6.4)
i

Inner product of wavefunctions a |b  = b |a ∗ = ai∗ bi (3.6.5)
i
Outer product of wavefunctions (|a b |)i j = ψi |a b |ψ j  = ai∗ b j
(3.6.6)
Operator matrix element Ai j = ψi | Â|ψ j  (3.6.7)
Operator given by the product Ci j = ψi |Ĉ|ψ j 

= ψi | Â|ψk ψk | B̂|ψ j 
k

= Aik Bk j (3.6.8)
k
Hermitian conjugate of operator Ci†j = ψi |Ĉ † |ψ j  = ψ j |Ĉ|ψi ∗
36 3 Wavefunction and Notations in Quantum Mechanics

 ∗ 
= ψ j | Â|ψk ψk | B̂|ψi  = A∗jk Bki
∗

k k
(3.6.9)
 
Trace Tr  = ψi | Â|ψi  = Aii (3.6.10)
i

Completeness Iˆ = |ψi ψi | (3.6.11)
i

Operation ci = ψi |   = ψi | Â|ψ j ψ j |
j

= Ai j c j . (3.6.12)
j

3.7 Composite System

We have considered the state of a single particle so far, but when considering different
particles at the same time or how the particles interact with each other, it is necessary
to describe multiple quantum states as one wavefunction in a concise form. For
example, if there are two qubits as shown in Fig. 3.5, we can consider them as one
composite system by using the direct product symbol ⊗ as

Composite system | of ψ and φ : | = |ψ ⊗ |φ.

It is customary to use simplified notations, such as

| = |ψ ⊗ |φ = |ψ|φ = |ψφ.

Fig. 3.5 A composite

system  consists of two
qubits, ψ and φ
3.7 Composite System 37

The composite system can be expanded to many particles; for example, a system
containing ψ1 , ψ2 , . . . ψn can also be represented as any of the following notations

| = |ψ1  ⊗ |ψ2  ⊗ · · · |ψn 

= |ψ1 |ψ2  · · · |ψn 
= |ψ1 |ψ2 · · · |ψn
= |ψ1 ψ2 · · · ψn .

The notation could be cumbersome and care needs to be taken to keep it simple.
The composite system has the following linear characteristics:

α (|ψφ) = (α|ψ) |φ = |ψ (α|φ) (3.7.1)


|ψ + |ψ   |φ = |ψφ + |ψ  φ (3.7.2)

|ψ |φ + |φ   = |ψφ + |ψφ  . (3.7.3)

Two-Qubit State Examples

Let us show two simple examples.

Two-qubit state
| = |ψφ = |01 (3.7.4)
is a state in which the first qubit |ψ is in 0-state (|ψ = |0) and the second qubit
|φ is in 1-state (|φ = |1).
Next example is for two-qubit states,
√
1 3
|ψ = |0 + |1
2 2
1 1
|φ = √ |0 − √ |1,
2 2

and then the composite state | = |ψφ is

 √   
1 3 1 1
| = |0 + |1 ⊗ √ |0 − √ |1
2 2 2 2
√ √
1 1 3 3
= √ |0|0 + √ |0|1 + √ |1|0 − √ |1|1
2 2 2 2 2 2 2 2
1  √ √ 
= √ |00 + |01 + 3|10 − 3|11 ,
2 2

where we used different notations for each line. The last line is probably the most
frequently used notation, due to its simplicity. When the notation without indices is
used, ordering obviously matters and care needs to be taken.
38 3 Wavefunction and Notations in Quantum Mechanics

In general, two-particle state basis can be written, using orthonormal basis of each
state ψi and φ j , as

|i j  = |ψi |φ j 

|kl  = |ψk |φl .

The inner product of these states can be written as


i j |kl  = ψi |φ j | (|ψk |φl )
= ψi φ j |ψk φl 
= ψi |ψk φ j |φl 
= δik δ jl . (3.7.5)

It is important to note here that the inner product is only taken with the same physical
system (e.g., qubit 1 to qubit 1). In other words, if even one of the individual systems is
orthogonal, the composite system is also orthogonal. For example, two-qubit system
|a  = |01 and |b  = |11 is orthogonal since,

a |b  = 01|11 = 0|11|1 (3.7.6)

= δ01 δ11 = 0 · 1 = 0, (3.7.7)

even when the two qubits are both 1.

One can expand this idea to describe an arbitrary state as

|a  = ai j |ψi |φ j  (3.7.8)
i, j

|b  = bkl |ψk |φl . (3.7.9)
k,l

Then the inner product is written as


a |b  = ai∗j bkl ψi |ψk φ j |φl 
i, j,k,l

= ai∗j bkl δik δ jl
i, j,k,l

= ai∗j bi j . (3.7.10)
i, j

Now, let us introduce an example using the system in Fig. 3.6. When two qubits
are given as

|ψ = a0 |0 + a1 |1 (3.7.11)

|φ = b0 |0 + b1 |1, (3.7.12)
3.7 Composite System 39

then from normalization,


1
|ai |2 = |a0 |2 + |a1 |2 = 1 (3.7.13)
i=0

1
|b j |2 = |b0 |2 + |b1 |2 = 1. (3.7.14)
j=0

For the composite system of two qubits, the wavefunction can be expanded as

| = |ψ|φ
= (a0 |0 + a1 |1) ⊗ (b0 |0 + b1 |1)
= a0 b0 |00 + a0 b1 |01 + a1 b0 |10 + a1 b1 |11. (3.7.15)

As mentioned above, |00, |01, |10, and |11 are orthogonal to each other.
The inner product of this composite system wavefunction by itself is

|| = (a0∗ b0∗ 00| + a0∗ b1∗ 01| + a1∗ b0∗ 10| + a1∗ b1∗ 11|)
·(a0 b0 |00 + a0 b1 |01 + a1 b0 |10 + a1 b1 |11)
= |a0 |2 |b0 |2 + |a0 |2 |b1 |2 + |a1 |2 |b0 |2 + |a1 |2 |b1 |2
= (|a0 |2 + |a1 |2 )(|b0 |2 + |b1 |2 )
 
= |ai |2 |b j |2 = 1. (3.7.16)
i j

As one can see that if each state (|ψ and |φ) is normalized, it assures the normal-
ization condition for the composite state |.

3.7.1 Operator in Composite System

If two qubits are far enough apart and can be manipulated independently, as shown
in Fig. 3.6, then of course one operation has no effect on the other. That is, when
the quantum state of two qubits |ψa  and |ψb  is written as |ψ = |ψa  ⊗ |ψb , with
corresponding operator acting on each qubit  and B̂, respectively, the operators that
act on the composite system is written as

 ⊗ B̂, (3.7.17)

and it operate on the state as

( Â ⊗ B̂)|ψ = ( Â ⊗ B̂)|ψa  ⊗ |ψb 

= Â|ψa  ⊗ B̂|ψb . (3.7.18)
40 3 Wavefunction and Notations in Quantum Mechanics

Fig. 3.6 A composite

system  consists of two
qubits, ψa and φb

Fig. 3.7 A composite

system of two qubits, where
only one of the qubits is
under an operation

When  is applied only to |ψa  in the composite system, with |ψb  untouched as
Fig. 3.7, the operator acting on the system is  ⊗ Iˆ.
It is often required to use a Hermite conjugate of the composite operator  ⊗ B̂,
which takes a form
(  ⊗ B̂)† = † ⊗ B̂ † . (3.7.19)
Again, one can see that each system is treated independently.

3.8 Examples of Notation for Frequently Used Quantum State

In this book, the discussions cover various concepts in quantum technologies; how-
ever, the descriptions are given mainly with the two-level system and harmonic
oscillator. There is a qubit as a representative of the two-level system, and its states
are expressed as |0 and |1. In addition, the harmonic oscillator can be described
with the state |n, where n represents the number of particles or excitations.
As one starts reading journal papers, readers may encounter various kinds of
quantum states and notations introduced depending on the quantum system of inter-
est. Examples of various notations for quantum states in different physical systems
are shown in Fig. 3.8. For example, in atoms, the ground and excited state is often
referred to as |g and |e, respectively. In some experiments, higher states are denoted
as | f , |h, . . .. Also, |a may be used to represent an “A”uxiliary state (or Ancillary
3.8 Examples of Notation for Frequently Used Quantum State 41

Fig.3.8 Example of notations used to describe different quantum states. a Qubit, b atom, c harmonic
oscillator, d photon (vertical and horizontal), e photon (diagonal), and f auxiliary or ancillary state

states) in various quantum systems. Further, for photon states, the state basis may be
given in the polarization state of light, such as vertical, horizontal, . . ..
Notation and the choice of physical states to be used may vary depending on
the actual system. Typical state notations used in this book or in other literature are
summarized in Table 3.1.

Table 3.1 Notation for frequently used quantum states

Physical System State Notation
Atom Ground State |g
Excited State |e
Higher States | f , |h, . . .
Qubit 0-state |0
1-state |1
Harmonic Oscillator n-particle state (0, 1, 2, · · · , n) |0, |1, |2, . . . , |n
Photon Vertical Polarization State | 
Horizontal Polarization State | ↔
Right-diagonal Polarization | 
State
Left-diagonal Polarization | 
State
Other State Auxilary State |a
42 3 Wavefunction and Notations in Quantum Mechanics

3.9 Quantum State Representation: Qubit

Now we have the Dirac notation and matrix representation down. Let us take a look
at the representation of the qubit, the main workhorse for the quantum information
and quantum technology.

3.9.1 One-Qubit state

A qubit state |ψ consists of two physical states denoted as |0 and |1, and an
arbitrary qubit state with corresponding complex amplitudes c0 and c1 , respectively,
can be written as

|ψ = c0 |0 + c1 |1

     
1 0 c
≡ c0 + c1 = 0 , (3.9.1)
0 1 c1
   
1 0
where the matrix representations |0 ≡ and |1 ≡ are used in the second
0 1
line. Throughout this book, we may use either representations suited for the context.
Taking an inner product of the state itself, the normalization condition is given as
 
 c0
ψ|ψ = c0∗ c1∗ = |c0 |2 + |c1 |2 = 1. (3.9.2)
c1

It is often more intuitive to separate the complex probability amplitudes c0 and

c1 in terms of its amplitude and phase as c0 = |c0 |eiφ0 and c1 = |c1 |eiφ1 . The qubit
state now reads

|ψ = |c0 |eiφ0 |0 + |c1 |eiφ1 |1

 
= eiφ0 |c0 ||0 + |c1 |e−iφ |1 , (3.9.3)

where φ = φ0 − φ1 . Recall that the probability density |ψ|2 allows an arbitrary phase
factor for the wavefunction, meaning that the results will not change if we use |ψ
or eiθ |ψ. The phase factor eiθ is called Global Phase and can be ignored without a
loss of generality.
Ignoring the global phase φ0 , the qubit state is finally represented as

|ψ = c0 |0 + c1 e−iφ |1 , (3.9.4)

where |c0 | → c0 and |c1 | → c1 is used since the phase factor is explicitly written
and the amplitudes can be understood as real positive values.
Note here that the important phase factor is the difference of the state phase φ,
and, as we have seen, the global phase is ignored.
3.9 Quantum State Representation: Qubit 43

Fig. 3.9 Bloch sphere

3.9.2 Bloch Sphere

There is a geographical representation for a qubit, called the Bloch sphere representa-
tion, which is quite convenient and intuitive. We discuss the physics of the two-level
system in detail later, where the dynamics of the qubit are discussed using the Bloch
sphere. Here, we introduce the general concept of the Bloch sphere and show some
fundamental properties of the qubit state on the Bloch sphere.
The qubit state discussed above has three parameters, c0 , c1 , and φ, with one
constraint, the normalization condition, which means two parameters are enough to
define the state of the qubit.
The normalization condition can be conveniently rewritten as

θ θ
c02 + c12 = sin2 + cos2 = 1, (3.9.5)
2 2
which makes the qubit state,

θ θ
|ψ = sin |0 + cos e−iφ |1 . (3.9.6)
2 2

An arbitrary qubit state can be specified using two parameters, θ and φ. Using a
geometrical representation, Bloch sphere, the quantum state can be represented as an
arrow pointing at anywhere on the spherical surface of radius 1 as shown in Fig. 3.9.
Using this representation, the qubit state assignments for ±x, ±y, and ±z-
direction are as follows:

|+z = |1
44 3 Wavefunction and Notations in Quantum Mechanics

Fig. 3.10 Motion of the qubit state on the Bloch sphere for amplitude difference variation (left),
and phase difference variation (right) between |0 and |1 state

|−z = |0
1
|+x = √ (|0 + |1)
2
1
|−x = √ (|0 − |1)
2
1
|+y = √ (|0 − i|1)
2
1
|−y = √ (|0 + i|1) .
2

As one can see, the poles of the sphere represent |0 and |1 state, while the
equator of the sphere represents the superposition of |0 and |1 with equal amplitude
and varying phase difference. The reader may check these state assignments using
Eq. 3.9.6.
The role of two parameters representing the qubit can be understood as

• θ : Amplitude difference between |0 and |1.

• φ: Phase difference between |0 and |1.

If the relative occupancy between |0 and |1 changes, the qubit state will move up
and down in the polar direction (latitudinal direction), while the variation of the phase
difference between |0 and |1 moves the state in azimuthal direction (longitudinal
direction) as shown in Fig. 3.10.
3.9 Quantum State Representation: Qubit 45

Starting from Eq. 3.9.6, one may think it is possible to move an arbitrary qubit
state toward +z-direction (north pole) to make a quantum state,

θ θ
|ψ = sin |0 + cos e−iφ |1 → |ψ = e−iφ |1. (3.9.7)
2 2

At a first glance, this state is different from |1 (e−iφ |1 = |1); however, the phase
factor e−iφ here is considered as a global phase, which can be ignored. It is also easy
to see from the Bloch sphere representation that there is no phase angle (azimuthal
angle) that can be defined at the north pole, which makes e−iφ |1 ≡ |1.
Similarly, |+y-state can also be written as

1
|+y = √ (i|0 + |1) , (3.9.8)
2

which is equivalent as the |y-state shown previously since it is different just by the
global phase factor i.
The Bloch sphere representation provides a pictorial and intuitive understanding
of the qubit. Unfortunately, there is no two-qubit or many-qubit equivalent of the
Bloch sphere, and its use is limited only for one-qubit state.

3.9.3 Two-Qubit State

Composite qubit state consisting of two qubits,

 
a
|ψa  = a0 |0 + a1 |1 ≡ 0 (3.9.9)
a1
 
b
|ψb  = b0 |0 + b1 |1 ≡ 0 , (3.9.10)
b1

|ψ = |ψa |ψb  can be written using a matrix representation as

|ψ = |ψa |ψb 

   
a b
= 0 ⊗ 0
a1 b1
⎛  ⎞
b0
⎜a0 b1 ⎟
=⎜⎝
 ⎟
b ⎠
a1 0
b1
⎛ ⎞
a0 b0
⎜a0 b1 ⎟
=⎜ ⎟
⎝a1 b0 ⎠ . (3.9.11)
a1 b1
46 3 Wavefunction and Notations in Quantum Mechanics

The following are the examples of two qubit state basis using the matrix notation
⎛ ⎞
1
⎜0⎟
|00 = ⎜ ⎟
⎝0⎠ ,
0
⎛ ⎞
0
⎜1 ⎟
|01 = ⎜ ⎟
⎝0⎠ ,
0
⎛ ⎞
0
⎜0⎟
|10 = ⎜ ⎟
⎝1⎠ ,
0
⎛ ⎞
0
⎜0⎟
|11 = ⎜ ⎟
⎝0⎠ .
1

Similarly, when X̂ a acting on |ψa  and Ẑ b acting on ψb are written as

 
01
X̂ a = (3.9.12)
10
 
1 0
Ẑ b = , (3.9.13)
0 −1

THE MATRIX REPRESENTATION OF THE composite operator reads

⎛    ⎞ ⎛ ⎞
1 0 1 0 0 0 1 0
⎜0 · 1 · ⎟ ⎜0
0 −1 0 −1 0 0 −1⎟
Xa ⊗ Zb = ⎜
⎝
   ⎟ = ⎜ ⎟. (3.9.14)
1 0 1 0 ⎠ ⎝1 0 0 0⎠
1· 0·
0 −1 0 −1 0 −1 0 0

The reader should be careful with the nesting order of both states and operators when
dealing with the composite system.
3.9 Quantum State Representation: Qubit 47

Quantum Entangled State

Previously, the quantum states of two individual particles are represented in the form
of a direct product, but there are quantum states that cannot be described by a direct
product. For example,
1
| = √ (|00 + |11), (3.9.15)
2
this state cannot be represented as a direct product in a form |ψa  ⊗ |ψb . Contrarily,
an example of a quantum state which can be decomposed as a direct product is

1 1
| = √ (|00 + |01) = |0 ⊗ √ (|0 + |1). (3.9.16)
2 2

A state that cannot be expressed by a direct product like Eq. 3.9.15 is called a
quantum entangled state (or quantum entanglement state) and is an important state in
quantum information. These following four states, called Bell states, are frequently
used to describe entangled two-qubit states.

1
|+  = √ (|00 + |11) (3.9.17)
2
1
|−  = √ (|00 − |11) (3.9.18)
2
1
|+  = √ (|01 + |10) (3.9.19)
2
1
|−  = √ (|01 − |10). (3.9.20)
2

For those who have never studied the entanglement state before, it may sound mys-
terious and possibly very difficult to make. It actually is not. We show an example
of a quantum logic circuit that creates an entanglement state. We will discuss more
about the quantum logic circuit later, and this will be a preparatory example. (You
might not understand what the circuit is doing, but it is OK, details are covered later
in Sect. 9.2.)
The quantum logic circuit in Fig. 3.11 has two-qubit input and two-qubit output
with two quantum logic gates inside. The first gate is a one-qubit Hadamard gate (H)
only applied to the top qubit |ψa , followed by a two-qubit C-NOT gate. Initially,
two inputs are both 0-state, making the initial state
⎛ ⎞
1
⎜0⎟
|ψin  = |00 ≡ ⎜ ⎟
⎝0⎠ .
0
48 3 Wavefunction and Notations in Quantum Mechanics

Fig. 3.11 Quantum logic

circuit to make an
entanglement state. The
circuit consists of a
one-qubit Hadamard gate on
the first qubit, followed by a
two-qubit C-NOT gate

 
1
The input for the Hadamard gate is |ψa  = |0 ≡ , and the Hadamard gate in a
0
matrix representation is
 
1 1 1
Ĥ = √ . (3.9.21)
2 1 −1
The output
    
1 1 1 1 1 1
|ψa  = Ĥ |ψa  ≡ √ · =√
2 1 −1 0 2 1
1
= √ (|0 + |1)
2
is a superposition of |0 and |1 with equal weight. Now the state in the middle of
two gates is
⎛ ⎞
1
1 1 1 ⎜ 0 ⎟
|ψmid  = √ (|0 + |1) ⊗ |0 = √ (|00 + |10) ≡ √ ⎜ ⎝
⎟.
⎠ (3.9.22)
2 2 2 1
0

Next, C-NOT is applied to |ψmid , generating the output state,

⎛ ⎞ ⎛ ⎞ ⎛ ⎞
1 0 0 0 1 1
⎜0 1 0 0⎟ 1 ⎜0⎟ 1 ⎜0 ⎟
|ψout  = ÛCNOT |ψmid  ≡ ⎜
⎝0
⎟· √ ⎜ ⎟= √ ⎜ ⎟
0 0 1⎠ 2 ⎝1 ⎠ 2 ⎝0 ⎠
0 0 1 0 0 1
1
= √ (|00 + |11) , (3.9.23)
2
which is one of the Bell states discussed earlier. As shown above, it is quite simple
to generate an entangled state using two-qubit gate.
The entangled state, for example, |+  = √1 (|00 + |11), has a 50% probability
2
of measuring either |00 or |11. This is very strange when you think about it. As
3.9 Quantum State Representation: Qubit 49

Fig. 3.12 Measurement of Bell state |+ 

shown in Fig. 3.12, suppose that you give the first of the two qubits in |+  to a
friend, and then that friend goes to the moon. One day, you came up with the idea
of measuring your qubit and |1 was observed. Since it is confirmed that the state is
collapsed to |11 from your qubit, you know the qubit that your friend has is |1 with
100 % certainty. It is as if the information about the state of your friend’s qubit, which
is up in the moon, flew all the way to the earth instantaneously. Contrary, if the two
shared a state mentioned previously |ψa |ψb  = (a0 |0 + a1 |1) ⊗ (b0 |0 + b1 |1),
which is in a form of a direct product. Even if you find out the state of your qubit
|1, the qubit state of friend on the moon |0 cannot be determined.
This mysterious correlation (relationship between particles) that arises from an
entangled state is called a quantum correlation, which is a non-classical correlation
that occurs only in quantum states. This quantum correlation is considered as an
important source of its power in the quantum information processing; however, how
the entanglement is related to the computational power, nor in that matter, how to
quantify the degree of entanglement is still a hot topic in current research.
Although, we do not have a precise measure for the entanglement yet, there are
certain entanglement states that are believed to be more entangled than the others. For
example, consider a three-qubit system, |ψ = |ψ1 ψ2 ψ3 . The Greenberger–Horne–
Zeilinger state, also known as GHZ-state, and W-state, named after Wolfgang Dür,
are defined as
1
|GHZ = √ (|000 − |111) (3.9.24)
2
1
|W = √ (|001 + |010 + |100) , (3.9.25)
3
respectively.
Suppose, you measure the first qubit of this GHZ-state and found |ψ1  = |1,
then you would automatically know the rest of the qubits (2 and 3) being |1. The
50 3 Wavefunction and Notations in Quantum Mechanics

same kind of result is obtained for the measurement result |ψ1  = |0, where you
will know the rest of qubits are in |0. Now, try the same measurement on W-state.
When the measurement is |ψ1  = |1, again, you will know the rest of the qubits (2
and 3) are |0, but if you measure |ψ1  = |0, the resulting state is

1
|W → |ψ = |0 ⊗ √ (|01 + |10) , (3.9.26)
2

where there are still some uncertainty in the state of qubit 2 and 3.
The strong entanglement seen in GHZ-state is sometimes referred to as a “fragile
entanglement”. For example, if the measurement previously discussed is not the
intended one, but done by some anonymous person, or mistakenly measured by the
environmental noise, then the entanglement of the state, after the measurement, is
completely gone. W-state, on the other hand, still maintains the entanglement (50%
of time) between qubits 2 and 3 even after the measurement of qubit 1. Different
entangled states have different reactions to the measurement and therefore different
noise tolerance. Depending on the application, the use of entanglement and the choice
of entangled state may vary.
It is also important to note that, when two particles interact, in general, they will
make an entanglement. The degree of entanglement may be small, but this process
could be a serious obstacle for the quantum operations in some cases.
For example, let us consider a qubit |ψqubit  = |0 which we are interested in
performing quantum operations. Suppose, an extraneous two-level system |ψenv  =
|1 from environment is present in the vicinity of the qubit (such as background gas
and impurities in solids). As long as there is no interaction between the two, the
system stays as | = |ψqubit  ⊗ |ψenv  = |0 ⊗ |1. However, even with a small
interaction, the system could evolve to a new state, for example,
 
 999 1
| = |01 → |  = |01 + |10. (3.9.27)
1000 1000

The equation tells us that practically nothing happened as you can see the state is
nearly |01, but this state cannot be decomposed to a direct product, thus an entangled
state. Why is this entanglement a problem? If the two-level system |ψenv  is measured
by any means, we measure |01 most of the time, but there is a small chance of
measuring |10, changing our qubit state to |ψqubit  = |1! Many of the decoherence
mechanisms (deterioration of “quantumness” in the state) for the quantum system
are believed to be originated from this sort of “micro-entanglement” with a large
number of unintended particles in the environment.

3.10 Eigenvalue Equation in Quantum Mechanics

Previously, the eigenvalue equation in linear algebra was reviewed. The formulation
of the quantum mechanics is built upon these tools of linear algebra. In quantum
3.10 Eigenvalue Equation in Quantum Mechanics 51

mechanics, all physical observable A can be described as an operator  with corre-

sponding observable value a in a form,

Â|ψ = a|ψ, (3.10.1)

represented as an eigenvalue equation. In that matter, the time-independent

Schrödinger equation described later,

2 ∂ 2 |ψ
− + V̂ (x)|ψ = E|ψ, (3.10.2)
2m ∂ x 2
can also be seen as an eigenvalue equation, by recognizing the Hamiltonian operator

2 ∂ 2
Ĥ = − + V̂ (x), (3.10.3)
2m ∂ x 2
which makes
Ĥ|ψ = E|ψ. (3.10.4)
Here, we can see the energy E is an eigenvalue of the system Hamiltonian Ĥ, and
Schrödinger equation is just an eigenvalue equation of the system Hamiltonian.
Similarly, if the momentum operator p̂ is applied to the wavefunction | p , which
is the eigenstate of momentum, the momentum value p of the state can be retrieved
from the operation
p̂| p  = p| p , (3.10.5)
as an eigenvalue.
If an operator has multiple eigenvalues λi and the corresponding eigenvectors
|ψi  that can be orthonormalized, the operator can be represented as


n
 = λi |ψi ψi |
i
⎛ ⎞
λ1
⎜ λ2 ⎟
⎜ ⎟
=⎜ .. ⎟, (3.10.6)
⎝ . ⎠
λn

where off-diagonal elements of the matrix are all zeros. This representation of the
operator is called the diagonal representation, and the operator which can be repre-
sented in this form is called the diagonalizable operator. There are number of useful
characteristics in the diagonalized operator. It is simply easy for calculations and
have straight-forward correspondence between the eigenvalues and eigenvectors. It
is also useful for some contexts such as quantum measurement as discussed later.
52 3 Wavefunction and Notations in Quantum Mechanics

Fig. 3.13 Atomic states

3.10.1 Example: Atom

Let us consider a short example, where the Hamiltoinan of an atomic system is

expressed as a diagonal matrix. Consider a three-level atomic system shown in
Fig. 3.13. Let us use a simple state basis,
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
1 0 0
|ψ0  = ⎝0⎠ , |ψ1  = ⎝1⎠ , |ψ2  = ⎝0⎠ , (3.10.7)
0 0 1

and we would like ⎛to derive⎞the system Hamiltonian Ĥ. Solving the Schrödinger
ab c
equation with Ĥ = ⎝d e f ⎠ for state|ψ0 ,
gh i
⎛ ⎞⎛ ⎞ ⎛ ⎞ ⎛ ⎞
ab c 1 a 1
Ĥ|ψ0  = ⎝d e f ⎠ ⎝0⎠ = ⎝d ⎠ = E 0 ⎝0⎠
gh i 0 g 0
→ a = E 0 , d = 0, g = 0. (3.10.8)

The Hamiltonian can be found by repeating the procedure for |ψ1  and |ψ2  as
⎛ ⎞
E0 0 0
Ĥ|ψ0  = ⎝ 0 E 1 0 ⎠ . (3.10.9)
0 0 E2

The Hamiltonian is also expressed as


Ĥ = E i |ψi ψi |. (3.10.10)
i

In many cases in quantum technology, the system Hamiltonian describing the energy
eigenstates used for the manipulation is expressed as a simple diagonal matrix with
the energy eigenvalues E i .
3.11 Operator Classification 53

3.11 Operator Classification

There are various operators that act on the Hilbert space in which the quantum state is
manipulated. They can be classified according to the characteristics of the operators.
Here, we describe the definitions of Hermitian, unitary, and positive operators, which
will be important later, and their properties in quantum mechanics.

3.11.1 Operator Functions

Before going into details of the operator classification, here is a short review on the
operator functions. One may occasionally encounter a function of operator Â, such
as f ( Â). As any of the arbitrary function f (x) can be expanded using the Taylor
series as
1
f (x) = f (0) + f  (0)x + · · · f (n) (0)x n + · · · , (3.11.1)
n!
the function of operators can also be expanded as

1 (n)
f ( Â) = f (0) Iˆ + f  (0) Â + · · · f (0) Ân + · · · , (3.11.2)
n!

where Ân = Â · · · Â denotes the product of n operators Â. Of particular importance

is the exponential function of operators, which is represented as follows:

∞
 Ân Â2 Â3
e ≡
Â
= I + Â + + ··· . (3.11.3)
n! 2! 3!
n=0
It is also easy to derive from above
 † †
e  = e  . (3.11.4)

Both of the above calculations can be tedious, but some special operator
∞ functions
are easily derived. For example, for a polynomial function f (x) = i=0 ci x i , the
operator function can be written as
∞

f ( Â) = ci Âi . (3.11.5)
i=0

When |ψn  is one of the eigenvectors of operator  with an eigenvalue an , then the
operator function satisfies

Âi |ψn  = (an )i |ψn  (3.11.6)

⇒ f ( Â)|ψn  = f (an )|ψn , (3.11.7)
54 3 Wavefunction and Notations in Quantum Mechanics

for polynomial functions.

Another example similar to above is when the operator  is diagonal. For example,
when an operator has a form
⎛ ⎞
A1
 = ⎝ A2 ⎠, (3.11.8)
A3

where blank matrix elements are all zeros, the matrix satisfies
⎛ n ⎞
A1
Ân = ⎝ An2 ⎠, (3.11.9)
An3

and polynomial functions are simply

⎛ ⎞
f (A1 )
f ( Â) = ⎝ f (A2 ) ⎠. (3.11.10)
f (A3 )

In unitary operators, exponential functions of operators such as e  frequently

appear. Below are some useful formulas for the exponential operator. They become
particularly useful for changing the pictures (Schrödinger, Heisenberg, and Dirac
picture) or the frame rotation as discussed later.
The Baker–Campbell–Hausdorff identity
  α2   
eα Â B̂e−α Â = B̂ + α Â, B̂ + Â, Â, B̂ + · · · . (3.11.11)
2!
Glauber’s formula
 
1
Â, B̂
e  e B̂ = e Â+ B̂ e 2 . (3.11.12)
 
Here, we used the commutator symbol, Â, B̂ = Â B̂ − B̂ Â.

3.11.2 Hermitian Operator

For any operator Ĥ , the Hermitian conjugate operator Ĥ † can be defined as the
operator, which satisfies the following inner product of the wavefunction.
 
φ| Ĥ ψ = ψ ∗ Ĥ φd x = ψ( Ĥ † φ)∗ d x =  Ĥ † φ|ψ. (3.11.13)

With a matrix representation, Ĥ † can be found very easily as

Ĥ † = ( Ĥ T )∗ , (3.11.14)
3.11 Operator Classification 55

where Ĥ T is a transpose of Ĥ , thus the Hermitian conjugate of Ĥ is derived from

the transpose of the complex conjugate of Ĥ . Also, the matrix representation eases
the derivation of the following characteristics of Hermitian conjugate operators.
 ∗  T  ∗
( Ĥ † )† = Ĥ T = ( Ĥ T )T = Ĥ (3.11.15)

(c Ĥ )† = c∗ Ĥ † (3.11.16)
† †
( Ĥ1 + Ĥ2 ) = Ĥ1 + Ĥ2
†
(3.11.17)
† †
( Ĥ1 Ĥ2 )† = Ĥ2 Ĥ1 . (3.11.18)

If the Hermitian conjugate operator of an operator is the same as the original

operator, that is,
Ĥ † = Ĥ , (3.11.19)
then Ĥ is called the Hermitian operator.
The Hermitian operator has the following characteristics regarding the inner prod-
uct of wavefunctions.

ψ| Ĥ φ =  Ĥ ψ|φ. (3.11.20)

The Hermitian operator is closely related to the physical quantity, and all the oper-
ators corresponding to the physical quantity we measure are, in fact, the Hermitian
operator.
By its nature, all measured physical observables must be real numbers, so of course
the expectation value should also be real numbers. The following is a reverse proof
but shows that the expectation values and eigenvalues of the Hermitian operators are
real numbers.
Consider the conjugate complex of the expectation value of Hermitian operator
Ĥ ,

H ∗ = | Ĥ ∗ =  Ĥ | = | Ĥ = H . (3.11.21)

The expectation value is the same as that of Ĥ , which implies the expectation value
is a real number. Similarly, with the eigenvector |ψi  and its eigenvalue λi , H ∗ has
two solutions as
 ∗
H ∗ = ψi | Ĥ ψi  = (λi ψi |ψi )∗ = λi∗ (3.11.22)
 ∗  ∗
H ∗ = ψi | Ĥ ψi  =  Ĥ ψi |ψi  (3.11.23)
 ∗
= λi∗ ψi |ψi  = λi . (3.11.24)

The eigenvalue λi satisfying the conditions above is a real number. Thus, as stated,
the eigenvalue of Hermitian operators is real, and all the physical observables are
expressed as Hermitian operators.
56 3 Wavefunction and Notations in Quantum Mechanics

3.11.3 Projection Operator

One of the most important Hermitian operator is the projection operator. Suppose a
state is written in a basis |ψi  ≡ |i. Projection operator P̂m which projects the basis
to |ψm  ≡ |m can be written as

P̂m = |mm|. (3.11.25)

The state |i can be projected in basis |m as

P̂m |i = |mm|i = cm |m, (3.11.26)

where projected amplitude is cm = m|i.

The sum of the projection operator P̂m formed with orthonormal basis |m that
spans the vector space is
  
P̂m = |mm| = |ψm ψm | = Iˆ. (3.11.27)
m m m

This is called completeness relation and is quite practical. The identity Iˆ can be
thrown in just about anywhere, and this relation conveniently allows a transformation
of its basis.
For example, suppose a wavefunction | is expressed in basis |φi  that is different
from the basis above as

| = di |φi . (3.11.28)
i
This wavefunction
 can be transformed in terms of new basis using the completeness
relation Iˆ = m |ψm ψm | as

| = di Iˆ|φi  (3.11.29)
i

= di |ψm ψm |φi  (3.11.30)
i,m

= cm |ψm , (3.11.31)
m

where cm = i di ψm |φi . The projection operator allows to express or transform
the wavefunction into any basis of choice.
3.11 Operator Classification 57

Changing the Qubit basis

Let us try to change the basis of qubit from |0 and |1 to |+x and |−x using the
projection operator P̂. As derived previously,

1
|+x = √ (|0 + |1)
2
1
|−x = √ (|0 − |1) ,
2

these states form a orthonormal set of basis, satisfying the condition ψi |ψ j  = δi j
as
1
+x|+x = (0| + 1|) (|0 + |1) = 1
2
1
−x|−x = (0| − 1|) (|0 − |1) = 1
2
1
+x|−x = (0| + 1|) (|0 − |1) = 0.
2
To write |0(|1) in terms of |+x and |−x, the completeness relation,

Iˆ = |ψi ψi | = |+x+x| + |−x−x|, (3.11.32)
i

is used as

|0 = Iˆ|0 = (|+x+x| + |−x−x|) |0

1
= √ [|+x (0| + 1|) + |−x (0| − 1|)] |0
2
1
= √ [|+x (0|0 + 1|0) + |−x (0|0 − 1|0)]
2
1
= √ (|+x + |−x) .
2

Taking the same procedure for |1, we obtain

1 1 
|0 = √ (|+x + |−x) ≡ √ |0  + |1 
2 2
1 1 
|1 = √ (|+x − |−x) ≡ √ |0  − |1  .
2 2

Recognizing the new qubit basis |0  ≡ |+x and |1  ≡ |−x, the original state |0
and |1 is now written as a superposition of the new qubit basis (Fig. 3.14).
58 3 Wavefunction and Notations in Quantum Mechanics

Fig. 3.14 Changing the basis of Qubit

Another example is shown for the calculation of expectation value A, which is
transformed in terms of density matrix using the projection operator. One can expand
A = ψ| Â|ψ using the projection operator as

A = ψ| Â|ψ


= ψ| Â|ii|ψ
i
  
= i|ψψ| Â|i = |ψψ| Â
ii
i i
   
= Tr |ψψ| Â = Tr ρ̂ Â . (3.11.33)

ρ̂ ≡ |ψψ| is called Density matrix, which will be described in detail later in

Sect.3.12. Here, we only emphasize that it becomes quite simple to calculate the
expectation value using the density matrix.

3.11.4 Unitary Operator

When an operator Û satisfies the condition,

Û † Û = I , (3.11.34)

the operator Û is said to be unitary. For arbitrary vectors |ψ and |φ, transformed by
Û as (|ψ, |φ → Û |ψ ≡ |α, Û |φ ≡ |β), the inner product of the new vectors
are the same as that of the original vectors as

α|β = ψ|Û † Û |φ

= ψ|φ. (3.11.35)

The unitary transformation thus preserves the inner product of the vectors. In a
close quantum system, the time evolution of the system can be described by the
3.11 Operator Classification 59

unitary operator. Also, it is important to mention that Eq. (3.11.34) can be rewritten
as U −1 = U † , which implies that all unitary operators have its inverse. In other
words, all the manipulations using unitary operators are reversible. In the context of
quantum information, it is ideal to manipulate the system without loss, meaning the
system is closed and the operation is reversible. For this reason, the term “unitary
manipulation” is often used to describe the ideal quantum manipulation without loss.
Let us show an important example that the exponential operator of a form ei Ĥ ,
where Ĥ is a Hermitian operator, is a unitary operator. For a Hermitian operator Ĥ ,
define an exponential operator Û = ei Ĥ . The Hermitian conjugate of this operator
is
 †
Û † = ei Ĥ

= e−i Ĥ ,

where the property of the Hermitian operator Ĥ † = Ĥ is used. It is straight forward

to see Û Û † = ei Ĥ e−i Ĥ = Iˆ (and Û −1 = Û † ); therefore, Û is an unitary operator.
To show above, two identities e  e B̂ = e Â+ B̂ for commuting operators  and B̂
    † †
Â, B̂ = 0 and e  = e  are used.
This exponential form is used quite often in the later section, where different
pictures in quantum mechanics, as well as use of rotating frames, are discussed.
Also, the exponential operator of a system Hamiltonian Ĥ is later discussed as a
time-evolution operator.

3.11.5 Pauli Operators

The Pauli operators are important operators, not just for the quantum information, but
also for a wide range of physics. These operators, also known as the Pauli matrices,
are used to describe the angular momentum of spin-1/2 particles. As later shown
in detail, the qubit has the same structure as the spin-1/2 system, as they are called
“pseudo-spins”. In other words, we can use the Pauli matrices that describe the
spin-1/2 dynamics as the fundamental building blocks for the qubit operations.
They are simple but powerful tool in many areas and we recommend readers to
make it stick as a second nature. Different literature have slightly different notations
for the Pauli matrices, but usually one of the followings is used.

 
ˆ 1 0
σˆ0 ≡ I ≡ (3.11.36)
0 1
 
0 1
σ̂1 ≡ σ̂x ≡ X̂ ≡ (3.11.37)
1 0
 
0 −i
σ̂2 ≡ σ̂ y ≡ Ŷ ≡ (3.11.38)
i 0
60 3 Wavefunction and Notations in Quantum Mechanics
 
1 0
σ̂3 ≡ σ̂z ≡ Ẑ ≡ . (3.11.39)
0 −1

All Pauli matrices satisfy

σi† = σi (3.11.40)
σi† σi = I, (3.11.41)

and they are Hermitian operators. Also, they have the following characteristics.

σ12 = σ22 = σ32 = −iσ1 σ2 σ3 = I . (3.11.42)

In the description of the two-level system using the Bloch sphere, the rotation
operator R̂i (θ ), which is an exponential operator of Pauli matrices, becomes an
important tool. The rotation matrix performs the rotation of the state vector on the
Bloch sphere. Here, we briefly introduce the action of the rotation operator.
When an arbitrary operator  satisfies Â2 = Iˆ, the following equation is satisfied:

eiθ Â = cos θ Iˆ + i sin θ Â. (3.11.43)

Using this relation, the rotation operators are given as

 
θ cos θ2 −i sin θ2
R̂x (θ ) ≡ e−i 2 X̂ = (3.11.44)
−i sin θ2 cos θ2
 
θ cos θ2 − sin θ2
R̂ y (θ ) ≡ e−i 2 Ŷ = (3.11.45)
sin θ2 cos θ2
 −iθ/2 
θ e 0
R̂z (θ ) ≡ e−i 2 Ẑ = . (3.11.46)
0 eiθ/2

The rotation operator R̂i (θ ) is an operator that rotates the state vector by θ with
respect to the i-axis. Figure 3.15 shows an example, where the initial state vector
|ψ0 , which points at the south pole, is rotated at an angle θ by the x-axis.
The final state |ψ f  is written in terms of the initial state and the rotation operator
as

|ψ f  = R̂x (θ )|ψ0 
  
cos θ2 −i sin θ2 1
=
−i sin θ2 cos θ2 0
 
cos θ2
= (3.11.47)
−i sin θ2 .
3.12 Density Operator 61

Fig. 3.15 Rotation of the

state vector on the Bloch
sphere

3.12 Density Operator

The quantum state has been represented by the ket-vector |ψ. According to this
representation, it is possible to describe a superposition state as an element of the
complex vector space, and even for the quantum state of the composite system, the
tensor product of the basis stretches the vector space, enabling a new representation.
For example, in the Stern–Gerlach experiment, the time evolution of the quantum
state of a single electron with a upward spin |↑ incident on the apparatus can be
analyzed precisely by solving the Schrödinger equation of motion. A quantum state
that can be expressed by a ket-vector in this way is called a pure state.
In reality, when a single electron is emitted out of a hot tungsten wire (Fig. 3.16),
for example, it is a 50-50 chance that the spin is upward or downward. What kind
of quantum state is it in such a case? If we want to express it as a ket-vector, we
may try to express it as a superposition state of upward spin and downward spin, for
example,
1
|ψ = √ (|↑ + |↓) , (3.12.1)
2
which we recognize as |+x-state. However, if it can be written as a superposition
state in this way, the direction of the spin should be fixed when looking at it from

Fig. 3.16 State of the

thermally emitted electron
62 3 Wavefunction and Notations in Quantum Mechanics

some other direction (e.g., x-direction). In reality, thermal electrons are statistically
dispersed so that the spin direction is not fixed in any direction. So the pure state
representation cannot correctly represent these hot electrons.
Consider a more general quantum state, where the state is found in one of the pure
states with a certain probability, say 50% in |↑ and 50% in |↓. Also in a regular
superposition state, the phase relationship between the spin up and down states
are fixed. Let’s discard this constraint too and not worry about the phase relation
between different pure states. Such a quantum state is called a mixed state and is a
good representation of the quantum state of hot electrons out of tungsten. Here, we
introduce how these states can be represented.
First, we consider the statistical mixture of electrons with upward spin |↑ = |ψ1 
and downward spin |↓ = |ψ2  with probabilities p1 and p2 , respectively, and focus
on the expectation value of the observable O (for example, the z-component of spin)
in the experiment. The expectation value for the pure state |ψ is ψ| Ô|ψ, but for
later
 convenience, we change the form with a trace, by using a completeness relation
ˆ
m |mm| = I as


ψ| Ô|ψ = ψ|m m| Ô|n n|ψ
m,n

= n|ψψ|mm| Ô||n
m,n
 
= Tr |ψψ| Ô . (3.12.2)

Now, the expectation value for the mixed state can be written in terms of the expec-
tation value for each pure state that is statistically mixed as

O = p1 ψ1 | Ô|ψ1  + p2 ψ2 | Ô|ψ2  = pi ψi | Ô|ψi . (3.12.3)
i

As in the pure state, transforming into a form with a trace shows


O = pi ψi | Ô|ψi  (3.12.4)
i

= pi ψi |m m| Ô|n n|ψi  (3.12.5)
mn i

= pi n|ψi ψi |mm| Ô|n (3.12.6)
mn i
 

= Tr pi |ψi ψi | Ô . (3.12.7)
i

shown above, |ψψ| appearing in the calculation for the pure state is replaced
As
with i pi |ψi ψi | in the mixed state. For that matter, if only one pi is 1 and the
3.12 Density Operator 63

others are all 0, then the results for the mixed state include those in pure state. This
quantity,

ρ̂ = pi |ψi ψi |, (3.12.8)
i
is a natural extension of the quantum state description in that sense and called density
matrix or density operator. It should be noted that the statistical mixture weight
pi ≥ 0 that appears here is fundamentally different from the quantum probability
amplitudes used in the coefficients of the superposition state.

3.12.1 Example of a Mixed State

In the previous example of thermal electron spin states, the quantum states of the
upward spin and the downward spin are statistically mixed with equal probability,
and the density operator is given as
1 1
ρ̂ = |↑↑| + |↓↓|. (3.12.9)
2 2
For electron spin with only possible states |↑ and |↓, this leads to
1ˆ
ρ̂ = I. (3.12.10)
2
This is called a completely mixed state for a single spin-1/2 system.
Another example is the thermal state of the harmonic oscillator. As explained
later, the state of the harmonic oscillator can be specified with “how many” quantized
vibrations exist (number state); however, the thermal state has a statistical distribution
of this number state |n according to the Boltzmann distribution.
Specifically, the density operator ρ̂th of the harmonic oscillator in the thermal
state can be written as
∞

ρ̂th = e−nω/kB T (1 − e−ω/kB T )|nn|, (3.12.11)
n=0

where , ω, k B , and T shown in the above equation are the Dirac constant, the
intrinsic angular frequency of the harmonic oscillator, the Boltzmann constant, and
the temperature of the state, respectively.

3.12.2 General Properties of Density Operator

The density operator can describe any state, whether quantum or classical. It becomes
especially important when the noise, which tends to diminish the “quantumness” of
the states, is considered. Here, we list the general properties of the density operator
ρ̂, which may be useful in the later chapter.
64 3 Wavefunction and Notations in Quantum Mechanics

• The trace of density operator is always 1,


Tr ρ̂ = 1. (3.12.12)

This is immediately derived from the fact that i pi = 1 holds for statistical
mixing probabilities.
• The density operator is a Hermite operator.
• The density operator is positive-semidefinite, that is, φ|ρ̂|φ ≥ 0 holds for any
|φ. This can be seen from pi ≥ 0, and it also implies that the eigenvalues of the
density operator are non-negative.
• For a pure state, the diagonal elements of the matrix representation of the density
operator correspond to the occupancy probabilities in the corresponding basis
state.
• ρ̂ 2 = ρ̂ holds for the pure state.  
• For a general density operator Tr ρ 2 ≤ 1, but only in pure state Tr ρ̂ 2 =
Tr ρ̂ = 1.
• Since
 it can be used as a measure to distinguish between the pure and mixed state,
Tr ρ̂ 2 is often called “purity” of the state.  
• The expectation value of a physical observable O is given by Tr ρ̂ Ô .

3.12.3 Density Operator of Composite System and its Reduction

For two independently prepared quantum systems, their density operators can be
written as ρ̂1 and ρ̂2 . The density operator ρ̂ of the composite system of these two
quantum systems could be defined by the tensor product of ket-vectors if it is a
separable pure state. In general, even with the mixed state, if they are separable after
all by adding statistical weights by considering both the ket and bra vectors, we may
write
ρ̂ = ρ̂1 ⊗ ρ̂2 . (3.12.13)
The composite system like above will evolve in time according to the system
Hamiltonian and we would like to know the final state of each quantum system.
We consider a case where ρ̂2 is a system with many degrees of freedom as shown
in Fig. 3.17, that is, a density operator consisting of a large number of spins and
a collection of harmonic oscillators, and it is extremely difficult to reconstruct the
density operator from measurement. In such a case, we have no choice but to consider
only the quantum system of interest ρ̂1 . Then, what kind of operation would it take
to obtain the information about the system of interest without
 the knowledge  of ρ̂2 ?
Consider the density operators ρ̂1 and ρ̂2 and basis k1 |k1 k1 |, k2 |k2 k2 |.
After the time evolution, the density operator has evolved to ρ̂  as

ρ̂ = ρ̂1 ⊗ ρ̂2 → ρ̂  . (3.12.14)

3.12 Density Operator 65

Fig. 3.17 Density operator for composite system. Where ρ̂1 is the system of interest and ρ̂2
is a collection of two-level systems or harmonic oscillators, that could be thought of as a noisy
environment for ρ̂1

The expectation value of observable O on the system of interest (System 1) is written

as
 
O = Tr 12 ρ̂  Ô

= k1 | ⊗ k2 |ρ̂  Ô|k1  ⊗ |k2 
k1 ,k2
⎛ ⎞
 
= k1 | ⎝ k2 |ρ̂  |k2 ⎠ Ô|k1 
k1 k2
⎡⎛ ⎞ ⎤

= Tr 1 ⎣⎝ k2 |ρ̂ |k2 ⎠ Ô ⎦


k2
  
= Tr 1 Tr 2 ρ̂  Ô
 
= Tr 1 ρ̂1 Ô . (3.12.15)

Here, a trace operation with a subscript means to take a trace for either or both of
system ρ̂1 and ρ̂2 . A trace operation for only one system is also called a partial trace.
From the above derivation, the reduced density operator ρ̂1 = Tr 2 ρ̂  can be
thought of as the density operator of the system ρ̂1 after the time evolution. In
the last expression, the expectation value of the operator Ô is calculated from the
density operator ρ̂1 . As shown above, the reduced density operator comes in handy
when measuring the physical observable of only a partial system of the composite
system.
One thing to note here is that the density operator after time evolution ρ̂  cannot
be written as ρ̂  = Tr 2 ρ̂  ⊗ Tr 1 ρ̂  . This is only true when ρ̂  is separable. If there is
entanglement between the two systems, a partial trace operation on one system leads
to the loss of information on the other system.
For example, if the ρ̂1 and ρ̂2 system are a qubit and a collection of harmonic oscil-
lators, respectively, and they interact. After sometime, the qubit obtains an extremely
complex entanglement formed between the qubit and the group of harmonic oscil-
lators, and then the state of the quantum bit loses coherence.
66 3 Wavefunction and Notations in Quantum Mechanics

3.13 Commutator and Anti-Commutator

We have seen the matrix representation of the operators in the previous sections. As
we know from linear algebra, the matrix multiplication  B̂ is not always the same
as B̂ Â and the ordering matters. In quantum mechanics, this plays an important
role and we often check the commutation relation of the operators. Commutator and
anti-commutator of operators are each defined as

 
Â, B̂ ≡ Â B̂ − B̂ Â (3.13.1)
% &
Â, B̂ ≡ Â B̂ + B̂ Â. (3.13.2)
 
When operators  and B̂ satisfy Â, B̂ = 0 (  B̂ = B̂ Â), then they commute, while
 
if Â, B̂ = 0 ( Â B̂ = B̂ Â) is satisfied, the operators do not commute. One important
consequence is when the operators of physical observables A and B commute, then
those observables can be measured simultaneously. The two commuting observables
are called Simultaneous observables and can also have the same set of eigenfunctions.
On the contrary, in a non-commutative case, the measurement of one observ-
able changes the other observable. This is quite non-intuitive and arose as one of
the upsetting consequences of the quantum mechanics in the early development.
The commutator is strongly associated with measurement and sets a new type of
measurement constraint. In the next section, we derive the celebrated Heisenberg
uncertainty principle from the commutation relation. 
Many readers have probably seen the canonical commutation relation x̂, p̂ =
i. In quantum technology, some important commutation relations include that of
the ladder operators and Pauli operators, which are used repeatedly to describe the
dynamics of harmonic oscillators and two-level systems, respectively.
Annihilation and creation operators, â and â † , respectively, has the following
commutation relation:


â, â † = 1, (3.13.3)
 †  †
from which other commutation relations, such as â â, â and â â, â † , are derived.
We recommend readers to derive some of these commutation relations.
Pauli operators obey the following commutation relations:
  
σ̂x , σ̂ y = 2i σ̂z , σ̂ y , σ̂z = 2i σ̂x , σ̂z , σ̂x = 2i σ̂ y ,

which can be summarized in a simple form,


σ̂i , σ̂ j = 2ii jk σ̂k , (3.13.4)
3.13 Commutator and Anti-Commutator 67

where i jk is the Levi–Civita symbol defined by

⎧
⎪
⎨0 for i = j, j = k, or k = i
i jk = +1 for cyclic permutation, (i, j, k) = (1, 2, 3), (2, 3, 1), (3, 1, 2)
⎪
⎩
−1 for non-cyclic permutation, (i, j, k) = (1, 3, 2), (3, 2, 1), (2, 1, 3).

3.13.1 Heisenberg’s Uncertainty Principle

The commutation relation can be used to derive a generalized uncertainty prin-

ciple. When the result of repeated measurements of a physical observable A is
A1 , A2 , A3 , . . ., the classical mean value A, the deviation from the mean Ai ,
and the variance σ A are each defined as

1
n
A = Ai (3.13.5)
n
i
Ai = Ai − A (3.13.6)
σ A2 = (Ai )2  = (Ai − A)2 . (3.13.7)

Similarly, in the quantum system, the expectation value A for the wavefunction
|ψ, the deviation from the expectation value A, and the variance σ A are defined,
respectively, as

A = | Â| (3.13.8)

A = Â − A (3.13.9)
σ A2 = |( Â − A)2 |
= ( Â − A)|( Â − A) (3.13.10)

In the last line, we used that all physical observables are represented by the Hermitian
operator.
For two observables A and B, we can define |a and |b as

|a ≡ ( Â − A)| (3.13.11)

|b ≡ ( B̂ − B)| (3.13.12)

and using above, we may rewrite variance as

σ A2 = a|a (3.13.13)
σ B2 = b|b. (3.13.14)

Using Schwarz’s inequality, we may write

σ A2 σ B2 = a|ab|b ≥ |a|b|2 . (3.13.15)

68 3 Wavefunction and Notations in Quantum Mechanics

If we separate the real and imaginary part as a|b = α + iβ, one can derive
 2
a|b − b|a
|a|b| = α + β ≥ β =
2 2 2 2
. (3.13.16)
2i

Here,

a|b = |( Â − A)( B̂ − B)|

= |( Â B̂ − A B̂ − B Â + AB)|
= AB − AB, (3.13.17)

and similarly
b|a = B A − AB, (3.13.18)
and we get
+ ,
a|b − b|a = AB − B A = Â, B̂ . (3.13.19)

Finally, when Eqs. (3.13.15–3.13.19) are combined, we obtain

 ,2
1 +
σ A2 σ B2 ≥ Â, B̂ . (3.13.20)
2i

It describes the generalized Heisenberg uncertainty principle with the commutator

of the physical observable operators  and B̂. That is, given the commutator, the
Heisenberg uncertainty principle of those physical
 quantities can be derived.
Using the well-known commutation relation x̂, p̂ = i, the famous position and
momentum uncertainty principle can be derived as
 2

σx2 σ p2 ≥ (3.13.21)
2

σx σ p ≥ . (3.13.22)
2

As another example, we consider the spin commutation relation. For a spin-1/2

system, spin operators Ŝi are represented using Pauli operators σ̂i as


Ŝi = σ̂i . (3.13.23)
2

Commutation relation for spin operators Ŝi are given as


Si , S j = ii jk Sk . (3.13.24)
3.13 Commutator and Anti-Commutator 69

Fig. 3.18 State vector on a Bloch sphere

For i = x, j = y, and k = z, the spin commutator is given as

 
Ŝx , Ŝ y = i Ŝz (3.13.25)


σ Sx σ S y ≥  Ŝz  . (3.13.26)
2

Let us consider for a moment what this commutation relations mean. If there is a
spin pointing at +z-direction, then  Ŝz  = /2, so the uncertainty relation becomes

2
σ Sx σ S y ≥ . (3.13.27)
4
This means that spins pointing in the +z-direction have uncertainty in x and
y-directions. Since the spin equation permutates with respect to the subscript, the
spin directed in the x-direction (y-direction) has an uncertainty in the y and z-
directions (x- and z-directions). In other words, the uncertainty is spread in the
direction orthogonal to the pointing direction of the spin state.
Using the Bloch sphere, the state of qubits can be described similarly as shown in
Fig. 3.18a. The arrow pointing from the origin to the surface of the sphere represents
the state of the qubit wavefunction. Keep in mind that the uncertainty of the state
spreads in the direction perpendicular to the arrow (on the sphere) as shown in
Fig. 3.18b. A detailed description is given in the chapter covering the physics of
two-level systems.

Problems

Problem 3-1 Consider a wavefunction,

√ √
| = |ψ1  + 2|ψ2  − 2|ψ3  + i 3|ψ4 .

(i) Normalize |.

70 3 Wavefunction and Notations in Quantum Mechanics

(ii) Find the expectation value A for

⎡ ⎤
1 0 i 0
⎢0 1 0 −i ⎥
 = ⎢
⎣i
⎥. (3.13.28)
0 1 0⎦
0 i 0 1

(iii) Is operator  a Hermitian? Show it from (a) the matrix property

of Hermitian operator and (b) check with the expectation value
A calculated.

Problem 3-2

(i) Show |+y and |−y-state form a set of orthonormal basis for
a two-level system.
(ii) Use the completeness relation Iˆ = |+y+y| + |−y−y| to
express |0 and |1-state in terms of |+y and |−y.

Problem 3-3 For an arbitrary state |, use the completeness relation,
Iˆ = i |ψi ψi | to show
 
A = Tr || Â . (3.13.29)

(It is in the text, but don’t look!)

Problem 3-4 Consider a three-level system with energy eigenvalues⎛ ⎞E i = ωi
1
(i = 1, 2, 3) with corresponding eigenstates |ψ1  = ⎝0⎠, |ψ2  =
0
⎛ ⎞ ⎛ ⎞
0 0
⎝1⎠, |ψ3  = ⎝0⎠.
0 1

(i) Write the system Hamiltonian Ĥ using ket-bra notation, e.g.,

|ψ1 ψ1 |.
(ii) Write the system Hamiltonian Ĥ in a matrix representation.

Problem 3-5 Show that a general form of the eigenvalues of unitary operator Û
is λ = eiθ . (Meaning that the absolute value of eigenvalue is always
1.)
3.13 Commutator and Anti-Commutator 71
 
ab
Problem 3-6 Inverse of a 2×2 matrix  = is given by
cd
 
1 d −b
Â−1 = .
ad − bc −c a
  
01 0 −i
(i) Show Pauli matrices σ̂x = , σ̂ y = , and σ̂z =
10 i 0
 
1 0
are unitary operators by showing Û † = Û −1 .
0 −1
(ii) Find the eigenvalues and eigenvectors for Pauli matrices.

Problem 3-7 Show Eq. 3.11.43,

eiθ Â = cos θ Iˆ + i sin θ Â

when Â2 = Iˆ, using the definition of the exponential of a matrix,

∞
 Ân Â2 Â3
e  ≡ = I +  + + ··· . (3.13.30)
n! 2! 3!
n=0

Problem 3-8 For a two-qubit system, where the wavefunction of each qubit is

1 √ √ 
|ψ1  = √ 2|0 + 3|1
5
1 √ 
|ψ2  = 3|0 − i|1 ,
2
find | = |ψ1 |ψ2  = |ψ1 ψ2  in

(i) Braket notation;

(ii) Matrix notation;
(iii) Find density operator ρ̂ for this two-qubit state in matrix nota-
tion.
(iv) From above matrix, check Tr[ρ̂] = 1.

Problem 3-9 In classical computation, (three-bit) Toffoli gate shown below in

Fig. 3.19 is conceptually important, due to its (a) reversibility, and
(b) functional completeness, meaning that a combination of Toffoli
gates can be used to construct an arbitrary logic gate.
72 3 Wavefunction and Notations in Quantum Mechanics

Fig. 3.19 Toffoli gate

Toffoli gate T̂ can be represented in a matrix form as

⎡ ⎤
1 0 0 0 0 0 0 0
⎢0 1 0 0 0 0 0 0⎥
⎢ ⎥
⎢0 0 1 0 0 0 0 0⎥
⎢ ⎥
⎢0 0 0 1 0 0 0 0⎥
T̂ = ⎢
⎢0
⎥. (3.13.31)
⎢ 0 0 0 1 0 0 0⎥⎥
⎢0 0 0 0 0 1 0 0⎥
⎢ ⎥
⎣0 0 0 0 0 0 0 1⎦
0 0 0 0 0 0 1 0

(i) Show Toffoli gate is unitary (thus reversible) and find the inverse
of Toffoli gate T̂ −1 .
(ii) Write the input state |ψin  = |ψ1 |ψ2 |ψ3  = (a0 |0 + a1 |1) ⊗
|1 ⊗ |0 in matrix representation and calculate the output state
|ψout . Also, write out a braket representation of the output
state.
(iii) When you measure the output state calculated in (ii), |ψ1  and
|ψ3  should be identical (otherwise your answer in (ii) should
be rechecked!). In quantum mechanics, there is “No-cloning
theorem”, which prohibit the copying of a quantum state. Dis-
cuss if this gate violates the no-cloning theorem.”

Problem 3-10 In a quantum harmonic oscillator, the position and momentum oper-
ator can be expressed in terms of ladder operators (â and â † ) as

x̂ = xzp â † + â

p̂ = i pzp â † − â ,
 
 mω
where xzp = and pzp = â and â † are ladder opera-
2 .
2mω 
tors satisfying the commutation relation â, â † = 1. Using these
ladder-operator representation of position and momentum operators
to derive the commutation relation for the position and momentum,

x̂, p̂ = i.
Time Evolution in Quantum System
4

In quantum technology, the control of a quantum system is usually referred to as

the control of time-dependent dynamics of the system. For example, in quantum
computation, we apply sequences of quantum gates to qubits to perform calculations.
It is important to be able to derive the time evolution of the system for a given
system Hamiltonian and the external perturbations. In this chapter, we review the
fundamentals of quantum mechanics focusing on the time evolution of the quantum
system.

4.1 Time-Independent Schrödinger Equation

As shown in Fig. 4.1, in classical mechanics, we often consider a situation where a

point-like object is moving in a potential V (x). It is easy to see that when this potential
changes over time, V (x) → V (x, t), the problem suddenly becomes difficult.
It is the same in quantum mechanics. Schrödinger equation is given as

∂(x, t) 2 ∂ 2 (x, t)
i =− + V̂ (x, t)(x, t), (4.1.1)
∂t 2m ∂x2

but when the potential is time independent, V̂ (x, t) → V̂ (x), the solution for (x, t)
can be found relatively easy.
When the potential V̂ (x) is time independent, the wavefunction can be divided
into time and space functions using the method of separation of variables

Supplementary Information The online version contains supplementary material available at

https://doi.org/10.1007/978-981-19-4641-7_4.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 73
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7_4
74 4 Time Evolution in Quantum System

Fig. 4.1 Time-independent

potential V (x)

(x, t) = ψ(x)ϕ(t). The Schrödinger equation is then divided into two parts, spatial
and time equation as

2 d 2 ψ(x)
− + V̂ (x)ψ(x) = Eψ (4.1.2)
2m d x 2
dϕ(t) iE
= − ϕ(t). (4.1.3)
dt 
The solution for the time equation is easily found as

ϕ(t) = e−i Et/ , (4.1.4)

and the wavefunction (x, t) is expressed as

(x, t) = ψ(x)e−i Et/ , (4.1.5)

where E is the energy eigenvalue and ψ(x) is the eigenfunction associated with
the energy. Equation (4.1.2) is called Time-independent Schrödinger equation and
solving for a solution ψ(x) becomes the goal to identify the wavefunction (x, t)
of the system.
The wavefunction in Eq. (4.1.5) is called Stationary state. It can easily be checked
that when the probability distribution of this eigenstate is calculated

| 2 (x, t)| = ψ ∗ (x)ei Et/ ψ(x)e−i Et/ (4.1.6)

= |ψ(x)|2 , (4.1.7)

this wavefunction does not change in time, as it is called stationary. Also, the expec-
tation value Q of the physical observable Q for this eigenstate is calculated as
 ∞
Q = d xψ ∗ (x)ei Et/ Q̂ψ(x)e−i Et/ (4.1.8)
−∞
 ∞
= d xψ ∗ (x) Q̂ψ(x), (4.1.9)
−∞

and the time dependence cancels out, thus again, it is stationary in time.
4.1 Time-Independent Schrödinger Equation 75

Fig. 4.2 Rotation of the

wavefunction from the time
evolution in a phase space

There may be several energy eigenvalues E n and associated eigenfunctions ψn (x)

depending on the shape of the potential and the boundary conditions. The general
solution for the quantum state can be written as

(x, t) = cn ψn (x)e−i E n t/ , (4.1.10)
n

and one can see that the variation of the wavefunction in time is originated from the
phase evolution of the wavefunction with the angular frequency ωn = En associated
with each eigenfunction.
Figure 4.2 shows the idea of this phase evolution in a cartoon. Considering a phase
space with the vertical axis as the real part of the wave function ψ(x) and the hori-
zontal axis as the imaginary part, the wave function rotates at an angular frequency
ωn along the x-axis. This is the time evolution of each eigenstate ψn (x). Since it is
rotating in the phase-space, |(x, t)|2 = |ψ0 (x, t)|2 or |ψ1 (x, t)|2 is constant and
stationary as derived previously. Here, the ground state and the first excited state of
the harmonic oscillator, which may be familiar to many readers, are considered as
examples.
We mentioned that the probability distribution |(x, t)|2 is time independent,
when one of the eigenstate is the solution. However, the situation is quite different
in a superposition state. Let us consider the superposition of two harmonic oscillator
states used in Fig. 4.2

1 1
(x, t) = √ ψ0 (x)e−iω0 t + √ ψ1 (x)e−iω1 t , (4.1.11)
2 2

where ψ0 (x) and ψ1 (x) are normalized spatial wavefunction for two harmonic oscil-
lator states. In the example shown in Fig. 4.2, the rotation frequency ω0 and ω1 are
different. At a certain moment as shown in Fig. 4.3, the wave function (x, t) is
shifted to the left, while at other time it is shifted to the right. In other words, in
this superposition, the expectation value of the particle position is not x = 0, but
its center of mass oscillating at ω1 − ω0 , thus not stationary. This vibration is mani-
fested by the interference of multiple wavefunctions, while each wavefunction alone
is stationary in time.
Qubits are no exception in this regard. In many experiments, the energy eigenstates
discussed above are used in qubits. The ground state |g ≡ |0 and the excited state
76 4 Time Evolution in Quantum System

Fig. 4.3 Oscillation of the

probability distribution from
the superposition of the
wavefunctions

|e ≡ |1 forms the qubit state with the corresponding energies E 0 = 0 and E 1 = ω,
respectively. The time evolution of this qubit is written as

|(t) = c0 |0 + c1 e−iωt |1 . (4.1.12)

It should be remembered that the phase of the excited state rotates with the fre-
quency ω = E 1 / between the ground and the excited state. A careful reader might
remember that the change in the phase difference between |0 and |1 represents the
motion of the state vector in the azimuthal direction in the Bloch sphere. In the qubit,
where energy states are used as the qubit states, there is a perpetual rotation of the
state vector at frequency ω = E 1 /. In order to get rid of this “background rotation”,
the qubit state is often written in the rotating frame of the qubit energy as discussed
later.

4.2 Time Evolution in Terms of Unitary Operator

The time evolution of the quantum state | is described by the Schrödinger equation
as
∂
Ĥ | = i | . (4.2.1)
∂t
Here, we introduce a method, where the wave function at an arbitrary time t is
obtained by describing the time evolution as a unitary operator, acting on the initial
wavefunction at t = 0.
Describing the time evolution is closely related to the idea of quantum computa-
tion. In classical information processing, the computation is proceeded by sequen-
tially executing discrete digital logic gates (AND, XOR, etc.). For example, the NOT
gate in digital logic is 0 → 1, 1 → 0. While in quantum computation, quantum states
such as spin states are used to describe the qubit. When the downward spin |↓ ≡ |0
4.2 Time Evolution in Terms of Unitary Operator 77

is flipped to upward spin |↑ ≡ |1, the time evolution inverting the spin can be con-
sidered as a quantum equivalent gate of the classical NOT gate. In other words, the
time evolution operator itself can be thought of as a quantum gate.
Suppose, the initial wavefunction |ψ0  at time t = 0 is time-evolved to |ψ(t) by
the time evolution operator Û (t) as

|ψ(t) = Û (t) |ψ0  . (4.2.2)

Substituting into the Schrödinger equation, we obtain

∂(Û (t) |ψ0 ) ∂ Û (t)

ĤÛ (t) |ψ0  = i = i |ψ0  (4.2.3)
∂t ∂t
∂ Û (t)
→ ĤÛ (t) = i , (4.2.4)
∂t

and integrating the equation, the time evolution operator Û (t) is derived as

Û (t) = Û0 e−i Ĥt/

= e−i Ĥt/ . (4.2.5)

Here, we consider that the Hamiltonian Ĥ is time-independent. Also, in the last line,
Û0 = Iˆ is used since the time evolution starts at t = 0 (when nothing has happened
yet). In principle, one could start from a time-dependent Hamiltonian Ĥ = Ĥ(t),
and solve them slightly differently. Here, we chose a time-independent Hamiltonian
for simplicity.
The time evolution of the wave function |ψ(t) for arbitrary time t1 to t2 using
this time evolution operator Û (t) is written as

|ψ(t2 ) = Û (t2 − t1 ) |ψ(t1 ) = e−i Ĥ(t2 −t1 )/ |ψ(t1 ) . (4.2.6)

Previously, we mentioned that the rotation of the state vector in the Bloch sphere
θ
can be performed by using a rotation matrix such as R̂x (θ ) = e−i 2 X̂ , where in this
operator, the quantum state is rotated by θ with respect to the x-axis of the Bloch
sphere. Now, the question is how could we get the time evolution operator Û (t) act
as an R̂i (θ ), or in other words, what is necessary for U ˆ(t) to have a form of R̂i (θ )?
If the system Hamiltonian Ĥ contains operators, such as X̂ , Ŷ , Ẑ , it creates a
rotation matrix. For example, if the system Hamiltonian is

Ĥ = α X̂ , (4.2.7)

the time evolution operator reads

Û (t) = e−i Ĥt/ = e−iα X̂

= R̂x (2αt), (4.2.8)
78 4 Time Evolution in Quantum System

Fig. 4.4 Rotation operator as an time-evolution operator

which rotates the qubit around x-axis by an arbitrary angle θ = 2αt as shown in
Fig. 4.4. Note here that, as long as the Hamiltonian has a form Ĥ = α X̂ , the state
vector will keep rotating. What we need to do is to turn on the Hamiltonian just
momentary, like a pulse, so when the state vector reaches the target position, the
rotation will stop. For an arbitrary qubit rotation on the Bloch sphere, it is necessary
to be able to rotate the state vector by at least 2-axis. We later describe these controls
on the Bloch sphere.

4.3 Three Pictures in Quantum Mechanics

In the previous section, the time evolution of quantum state is described by the unitary
operator Û (t) = e−i Ĥt/ . However, what we can usually measure experimentally is
not the wavefunction itself, but some physical observable A. As the wave function
changes in time, the expectation value A also varies as

A(t) = ψ(t)| Â |ψ(t)

= ψ0 | ei Ĥt/ Âe−i Ĥt/ |ψ0  . (4.3.1)

Here, we consider the time variation of the physical observable A, relating it to

the variation of the states and operators. In quantum mechanics, there are so-called
“three pictures”, in which each picture has a different treatment of the time evolution
of state and operators to describe the system dynamics. There are pros and cons
for each picture, and we often select a picture that is suitable or easy to calculate
depending on the problem. Three pictures are Schrödinger, Heisenberg, and Dirac
picture. Here we describe the overview and distinctions between these pictures.
4.3 Three Pictures in Quantum Mechanics 79

Fig. 4.5 Difference between the Schrödinger and Heisenberg picture

4.3.1 General Overview

There are no definitive rules to pick which picture to be used for a particular situation,
but generally, these rules might be a good starting point.

Schrödinger picture
If the time evolution of the wavefunction ψ(t) is of interest,
e.g., calculating the electronic wavefunction in a quantum
well and observing its spatial dynamics.
Heisenberg picture
If the time evolution of the observable A(t) is of interest,
e.g., calculating the dynamics of electric field strength E(t)
in and out of an optical cavity for a laser spectroscopy.
Dirac picture
If you start out with one system and turn on an extra pertur-
bation. For example, a Hydrogen atom itself is a quantum
system consisting of an electron under the potential created
by the nucleus. External magnetic field (perturbation) is then
introduced to the system to investigate the dynamics under
the influence.

First, the difference between the Schrödinger and Heisenberg picture is the dif-
ference in the focus between the quantum state and its observable. In most of the
introductory quantum mechanics courses, we start out with the Schrödinger picture.
In this picture, the system dynamics are explained in terms of the state evolution as
shown in Fig. 4.5, in other words, we focus on how the wavefunction evolves in time
ψ(t), while the operator  is treated as time independent.
80 4 Time Evolution in Quantum System

Fig. 4.6 Dirac picture is useful for addition of perturbation V̂

Contrary, the Heisenberg picture focus on the time evolution of the physical
observable A(t). As discussed earlier, the observable A is represented as an operator
Â, and the time evolution of this operator Â(t) is the main focus, while the state ψ
is considered time independent. We later show that the time evolution of operator
Â(t) can be prescribed in a simple manner using Heisenberg equation of motion.
As discussed earlier, in classical mechanics, observable such as position x(t) is
calculated from Newton’s equation of motion. In this sense, the Heisenberg picture
resembles more of the classical picture, while Schrödinger picture focuses on the
wavefunction, which has no classical counterpart.
Dirac picture has slightly different benefits compared to Schrödinger and Heisen-
berg picture, however, the approach can be considered as an intermediate picture of
two pictures as shown later in detail. As mentioned above, the Dirac picture becomes
particularly useful when the external perturbation V̂ is applied to an existing quantum
system with Hamiltonian Ĥ0 as shown in Fig. 4.6. The intermediacy of the picture
can be seen from the fact that both state ψI (t) and operator ÂI (t) are time dependent
in this picture.
While describing the differences, it is easy to mix up different pictures unless the
notation is clearly defined. We use following notation for state and operators:

|ψ0  Time-independent initial state

Â0 Time-independent initial operator

ÂS , |ψS  Operator and state in Schrödinger picture

ÂH , |ψH  Operator and state in Heisenberg picture
ÂI , |ψI  Operator and state in Dirac picture

To have a consistent result for all pictures, the expected value of the physical
observable A(t) must be the same. We come back to this value A(t) time to time
4.3 Three Pictures in Quantum Mechanics 81

while introducing how different pictures separate the time evolution of wavefunctions
and operators.

4.3.2 Schrödinger Picture

Up to this point in the book, the dynamics of the quantum system are described with
the initial wave function ψ0 changes with time as ψ(t), and the physical observable
operator  remains constant in time as  = Â0 . Presumably, this is the way many
readers first learn quantum mechanics and this picture is called Schrödinger Picture.
In this picture, the wavefunction evolves in time as

d |ψ(t) i
= − Ĥ |ψ(t) . (4.3.2)
dt 

The quantum state and operators are described as

|ψS (t) ≡ e−i Ĥt/ |ψ0  (4.3.3)

ÂS ≡ Â0 , (4.3.4)

and only the wavefunction evolves in time. The expectation value A evolves in
time as

A(t) = ψ0 | ei Ĥt/ Â0 e−i Ĥt/ |ψ0  (4.3.5)

= ψS (t)| ÂS |ψS (t) , (4.3.6)

along with the time evolution of the wavefunction.

4.3.3 Heisenberg Picture

In the Heisenberg picture, we take a position where the state ψ0 remains constant
while the operators are evolving in time.
In this picture, the time varying operator Â(t) is described as

Â(t) = ei Ĥt/ Â0 e−i Ĥt/ . (4.3.7)

Summarizing the time evolution of state and operator in Heisenberg picture, we have

|ψH  ≡ |ψ0  (4.3.8)

i Ĥt/ −i Ĥt/
ÂH (t) ≡ e Â0 e . (4.3.9)
82 4 Time Evolution in Quantum System

Since the state is stationary in the Heisenberg picture, there is no Schrödinger equa-
H
tion, which describes the time evolution of the state i d|ψ
dt = · · · . Instead, we will
derive the equation of motion for the operator, called the Heisenberg equation of
motion, later.
The expectation value A evolves in time as

A(t) = ψ0 | ei Ĥt/ Â0 e−i Ĥt/ |ψ0  (4.3.10)

= ψH | ÂH (t) |ψH  , (4.3.11)

along with the variation of the operator.

A careful reader may point out that “if the operator changes over time, the Hamil-
tonian itself also changes over time, so the expression containing the Hamiltonian
(4.3.9) should be more complicated.” One might think, the time evolution of the
Hamiltonian is as follows:

ĤH = ei Ĥt/ Ĥe−i Ĥt/ , (4.3.12)

with explicit time dependence. However, using Eq. (3.11.3)

∞
 Ân Â2 Â3
e  ≡ = I +  + + ··· ,
n! 2! 3!
n=0

ei Ĥt/ can be expressed as a polynomial function of Ĥ, which commute with Ĥ,
resulting in

ĤH = ei Ĥt/ Ĥe−i Ĥt/ = Ĥ. (4.3.13)

The Hamiltonian above is actually time-independent and we do not need to worry
about the further complication.
Also, it is important to know how the commutation relation changes in the Heisen-
berg
 picture.
 For example, if the commutation relation in Schrödinger picture is given
as Â, B̂ = Ĉ, how would it change in the Heisenberg picture? Would the operators
commuting in Schrödinger picture also commute in Heisenberg picture? 
Consider a commutation relation in Schrödinger picture, ÂS , B̂S = ĈS , and
define Û = e−i Ĥt/ and Û † = ei Ĥt/ to simplify the notation. The commutation
relation in Heisenberg picture is given as
 
ÂH , B̂H = ÂH B̂H − B̂H ÂH

= Û † ÂS Û Û † B̂S Û − Û † B̂S Û Û † ÂS Û

= Û † ÂS B̂S Û − Û † B̂S ÂS Û
 
= Û † ÂS B̂S − B̂S ÂS Û

= Û † ĈS Û = ĈH
4.3 Three Pictures in Quantum Mechanics 83

 
ÂH , B̂H = ĈH , (4.3.14)

where ÔH = Û † ÔS Û = ei Ĥt/ ÔS e−i Ĥt/ .

As shown above, the form of the commutation relation stays the same as the
Schrödinger picture, with a simple replacement of the operators to the ones in the
Heisenberg picture.

4.3.4 Example: Harmonic Oscillator in Schrödinger and Heisenberg

Picture

Let us try to derive and see the difference between the Schrödinger and Heisenberg
picture using a familiar and one of the most important systems in quantum technology,
the harmonic oscillator. It is cumbersome to write ÂS,H for each operator and here we
use the convention often used in much of the literature; the operators in Schrödinger
picture is denoted as Â, while the operators in the Heisenberg picture is written Â(t),
with the explicit time dependence.
Schrödinger picture: In the Schrödinger picture, the system Hamiltonian, position
x̂, and momentum p̂ operators are given in terms of creation and annihilation operator,
also known as ladder operators, â † and â as,

Ĥ = ω↠â 
x̂ = xzp â † + â
 
p̂ = i pzp â † − â ,

 

where xzp = 2mω and pzp = mω ω
2 . Also, the zero-point energy E 0 = 2 , which
is just an offset energy, is ignored from the Hamiltonian for simplicity. Let us calculate
the canonical commutation relation for x̂ and p̂

x̂, p̂ = x̂ p̂ − p̂ x̂
     
= i xzp pzp â † + â â † − â − â † − â â † + â
  
= i · 2 â â † − â † â = i, (4.3.15)
2

where we used the commutation relation for the ladder operator â, â † = 1.
A general state in this picture is written in terms of number state |n as
84 4 Time Evolution in Quantum System

∞

|ψ(t) = cn (t) |n
n=0
∞
= cn (0)e−iωn t |n , (4.3.16)
n=0

where ωn = E n / = nω and the solution for the time-independent potential dis-

cussed earlier is used in the last line.
The expectation value of position x is calculated as

x(t) ≡ ψ(t)|x̂|ψ(t)
∞

∗
= cm (0)cn (0)e−i(ωn −ωm ) m|xzp â † + â |n
n,m=0
∞ √ √ 
∗
= xzp cm (0)cn (0)e−i(ωn −ωm ) m| n + 1 |n + 1 + n |n − 1
n,m=0
∞ √ √ 
∗
= xzp cm (0)cn (0)e−i(ωn −ωm ) n + 1δm,n+1 + nδm,n−1
n,m=0
∞  
√ ∗ √ ∗
= xzp n + 1cn+1 (0)cn (0)eiωt + ncn−1 (0)cn (0)e−iωt . (4.3.17)
n=0

Previously, we discussed the qualitative understanding of the motion of the har-

monic oscillator state, given in a superposition state using Fig. 4.3. The calculation
above shows the analytical result, confirming the origin of motion as the interference
between different states. Furthermore, the motion is provided only by the interference
of neighboring states, n and n + 1 (or n − 1).
Heisenberg picture: For the Heisenberg picture, let us start with the annihilation
and creation operator (â, â † ). The system Hamiltonian in Schrödinger picture is
Ĥ = ωâ † â and which makes

â(t) = ei Ĥt/ âe−i Ĥt/

= eiωâ ât âe−iωâ ât
† †

= âe−iωt = â(0)e−iωt (4.3.18)

â (t) = â † (0)eiωt ,
†
(4.3.19)

where Baker–Campbell–Hausdorff formula is used in the first derivation. Also, we

use the fact that the operator in Schrödinger picture is the initial state of the operator
in Heisenberg picture as â → â(0), â † → â † (0).
4.3 Three Pictures in Quantum Mechanics 85

From the above results, the Hamiltonian has the same form as the Schrödinger
picture as described previously,

ĤH = ωâ † (t)â(t) = ωâ † eiωt âe−iωt = ωâ † â = ĤS .

Using the results above, the position and momentum operators are derived as

x̂(t) = eiωâ ât xzp â † + â e−iωâ ât

† †

 
= xzp â(0)† eiωt + â(0)e−iωt (4.3.20)
 
p̂(t) = i pzp â(0)† eiωt − â(0)e−iωt . (4.3.21)

From above, we recognize x̂(0) = xzp â(0)† + â(0) and p̂(0) = i pzp â(0)† −
â(0) . Solving these for â and â † plug them back into Eqs. (4.3.20) and (4.3.21), the
time evolution of position and momentum now reads

p̂(0)
x̂(t) = x̂(0)cosωt + sinωt (4.3.22)
mω
p̂(t) = p̂(0)cosωt + mω x̂(0)sinωt. (4.3.23)

As discussed earlier, these operators certainly resemble the classical observable x(t)
and p(t) of the harmonic oscillator.
We can also check the commutation relation for the ladder operators, and position
and momentum operators in the Heisenberg picture as

â(t), â † (t) = 1
x̂(t), p̂(t) = i.
 
Both of them are consistent with our previous observation ÂS , B̂S = ĈS →
 
ÂH , B̂H = ĈH , where both 1 and i are the same in the Heisenberg picture.
The state in the Heisenberg picture, which is intrinsically time-independent, is
given simply as
∞

|ψH = cn (0) |n , (4.3.24)
n=0
from which the expectation value of position x is calculated as before

x(t) ≡ ψH |x̂(t)|ψH 

∞
  
∗
= cm (0)cn (0) m|xzp â † eiωt + âe−iωt |n
n,m=0
∞  
√ ∗ √ ∗
= xzp n + 1cn+1 (0)cn (0)eiωt + ncn−1 (0)cn (0)e−iωt , (4.3.25)
n=0
86 4 Time Evolution in Quantum System

which is, of course, identical to the expectation value calculated in the Schrödinger
picture.

4.3.5 Dirac Picture

Dirac picture has a characteristic that is half-way between the Schrödinger picture and
the Heisenberg picture. In this picture, we consider a time-independent Hamiltonian
Ĥ0 and additional Hamiltonian V̂ in a form

Ĥ = Ĥ0 + V̂ . (4.3.26)

Since the Dirac picture focuses on this extra interaction V̂ , it is also called the
interaction picture. We describe the state and operator in the Dirac picture with
subscript I, referring to the “interaction”.
This picture becomes particularly important for controlling the quantum system.
For example, let us consider controlling the electronic state of a Hydrogen atom
with an external electric field. The electron is influenced by the Hydrogen nucleus
regardless of the applied external field. In this example, Ĥ0 becomes the Hamiltonian
for the Hydrogen nucleus and the effect of the external field is given by V̂ . In other
words, it is convenient because you can write the Hamiltonian that is always of
influence in Ĥ0 , and our operation is solely described in V̂ .
In the Dirac picture, both the wave function and the operator evolves in time,
but from different origin. The wavefunction time evolves from the influence of the
added interaction V̂ , while the operator  evolves from the influence of the stationary
interaction Ĥ0 expressed as

|ψI (t) ≡ e−i V̂ t/ |ψ0  (4.3.27)

ÂI (t) ≡ ei Ĥ0 t/ Â0 e−i Ĥ0 t/ (4.3.28)

The expectation value A evolves in time as

A(t) = ψ0 | ei Ĥt/ Â0 e−i Ĥt/ |ψ0 

= ψ0 | ei V̂ t/ · ei Ĥ0 t/ Â0 e−i Ĥ0 t/ · e−i V̂ t/ |ψ0 
= ψI (t)| ÂI (t) |ψI (t) , (4.3.29)

and the time evolution of the state in Dirac picture is described in terms of the state
in Schrödinger picture as

|ψ(t)I = e−i V̂ t/ |ψ0  = ei Ĥ0 t/ e−i(Ĥ0 +V̂ )t/ |ψ0 
= ei Ĥ0 t/ |ψS . (4.3.30)
4.3 Three Pictures in Quantum Mechanics 87

In the Dirac picture, the operator also evolves in time, so let’s check the time
evolution of Hamiltonian Ĥ = Ĥ0 + V̂ as before.

Ĥ0I = ei Ĥ0 t/ Ĥ0 e−i Ĥ0 t/ = Ĥ0 (4.3.31)

i Ĥ0 t/ −i Ĥ0 t/
V̂I = e V̂ e (4.3.32)

The above shows that the steady Hamiltonian Ĥ0 does not evolve in time, however,
the added Hamiltonian V̂I varies in time from the influence of Ĥ0 .
Now consider the time evolution of the state written in the Dirac picture. Using
|ψS = e−i Ĥt/ |ψ0  and |ψI = ei Ĥ0 t/ |ψS , the state evolution can be written as

d |ψI i d |ψS
= Ĥ0 |ψI + ei Ĥ0 t/
dt  dt
i i Ĥ0 t/ −i Ĥ0 t/ i
= e Ĥ0 e |ψI − ei Ĥ0 t/ Ĥe−i Ĥ0 t/ |ψI
 
i i Ĥ0 t/ −i Ĥ0 t/ i
=− e V̂ e |ψI = − V̂I |ψI , (4.3.33)
 
with final result
d |ψI i
= − V̂I |ψI . (4.3.34)
dt 
As we shall see later, the interaction picture becomes particularly useful in the time-
dependent perturbation theory.
Summary Above is a quick overview of three pictures in quantum mechanics. As
shown above, all pictures describe the same time evolution of the expectation value
of a physical observable A(t), despite the difference in the time evolution of the
state and operators.
There are pros and cons for all pictures. For example, finding the wavefunction for
a potential well, as we all learn in the introductory quantum mechanics, is actually
difficult in the Heisenberg picture. In the next section, we use the Heisenberg picture
to derive the Heisenberg equation of motion, which describes the equation of motion
for an operator and is a counterpart of the Schrödinger equation. In quantum tech-
nology, we often focus on the time evolution of the observables, such as resonator
field mode and external electromagnetic noises that are described by operators. In
such a case, the Heisenberg picture and the use of Heisenberg’s equation of motion
is an excellent alternative.
The Dirac picture is particularly useful when an additional perturbation is provided
to a system with known eigenstates and corresponding eigenenergies. With the Dirac
picture, as shown later, the time-dependent perturbation theory can be described in
a much simpler form than the Schrödinger picture.
88 4 Time Evolution in Quantum System

4.4 Heisenberg Equation of Motion

The time evolution of an arbitrary operator  in the Heisenberg picture is given as

ÂH = ÂH (t) = ei Ĥt/ Â0 e−i Ĥt/ . (4.4.1)

In a form of time derivative, we can expand the above as

d ÂH (t) i i Ĥ −i Ĥt/ ∂ Â0 −i Ĥ0 t/

= Ĥei Ĥt/ Â0 e−i Ĥt/ − ei Ĥt/ Â0 e + ei Ĥ0 t/ e
dt   ∂t
i  i Ĥt/  ∂ Â0 −i Ĥ0 t/
= Ĥe Â0 e−i Ĥt/ − ei Ĥt/ Â0 e−i Ĥt/ Ĥ + ei Ĥ0 t/ e
 ∂t
i  ∂ Â0 −i Ĥ0 t/
= Ĥ, ÂH + ei Ĥ0 t/ e , (4.4.2)
 ∂t
and we get

d ÂH (t) i  ∂ Â(t)
= Ĥ, ÂH (t) + , (4.4.3)
dt  ∂t
H
 
where ∂ Â(t)
∂t = ei Ĥ0 t/ ∂∂tÂ0 e−i Ĥ0 t/ . This equation is called the Heisenberg
H
equation of motion. While Schrödinger equation describes the time evolution of the
wavefunction, in the Heisenberg picture, the system evolution is described focusing
on the evolution of the operators. The last term is included only when Â0 is explicitly
time-dependent, meaning the operator in Schrödinger picture have an explicit time
dependence. Otherwise, the last term can be omitted, simplifying the equation as

d ÂH (t) i 
= Ĥ, ÂH (t) . (4.4.4)
dt 

Let us take a look at the harmonic oscillator previously seen, as an explicit exam-
ple. Here we omit the subscript ‘H’ to avoid clumsiness, but all the operators are
understood to be in the Heisenberg picture. The annihilation and creation operators
follow

d â(t) i
= ωâ † â, â
dt 
= −iωâ(t),

which can be integrated to obtain

â(t) = â(0)e−iωt .
4.5 von Neumann Equation 89

Fig. 4.7 Time evolution of

operators, σ̂z and â in cavity
QED system

The time evolution of the position operator x̂ is given by the Heisenberg equation
of motion as

d x̂(t) i
= ωâ † â, xzp â † + â
dt 
= iωxzp â † â, â † + iωxzp â † â, â
= iωxzp â † − iωxzp â.

Integrating the above result, we obtain the position operator in Heisenberg picture
as
 
x̂(t) = xzp â † eiωt + âe−iωt ,

which is identical to the previously obtained result.

In quantum technology, many signals including the incident electromagnetic
waves to the resonator and the state in the resonator are mainly described as quantum
modes using the ladder operators â and b̂.
By taking the spectroscopy of the resonator QED system as shown in Fig. 4.7 as
an example, the resonator mode and the qubit state are described as operators â and
σ̂z , respectively. Also, the drive field âin from the outside and the reflected light âout
are described by operators as well. In the actual experiment, we are often interested
in the time evolution of the qubit σ̂z (t) by the drive field, the time evolution of the
resonator mode â(t), and the reflected light âout (t). The time evolution of these
operators can be calculated using Heisenberg’s equation of motion, and it is easy
to compare with the measured physical quantities that have a high affinity with the
experiment.

4.5 von Neumann Equation

von Neumann equation is quite similar in appearance to Heisenberg’s equation of

motion and sometimes it is confusing.
von Neumann equation is give as

d ρ̂(t) i  
= − Ĥ, ρ̂(t) . (4.5.1)
dt 
90 4 Time Evolution in Quantum System

Note here that ρ̂(t) ≡ |ψS (t)ψS (t)| is a density operator written in Schrödinger
picture. At first glance, the von Neumann equation seems to be an adaptation of
Heisenberg’s equation of motion to the density operator ρ(t), but the sign is different.
One can reach the von Neumann equation by describing the time evolution of the
density operator in Schrödinger picture as follows,
∂ ρ̂(t) ∂ (|ψS (t)ψS (t)|)
i = i
∂t ∂t
∂ |ψS (t) ∂ ψS (t)|
= i ψS (t)| + i |ψS (t)
∂t ∂t
= Ĥ |ψS (t) ψS (t)| − |ψS (t) ψS (t)| Ĥ
 
= Ĥ, ρ̂(t) . (4.5.2)

Keep in mind that the von Neumann equation is written in Schrödinger picture.
Similar confusion is likely to occur when finding the expectation value of operator
Â. Earlier, we described A as
 
A = Tr ρ̂ Â . (4.5.3)

In this equation, the density operator time evolves in Schrödinger picture, while the
operator  evolves in the Heisenberg picture and care needs to be taken. Form of ρ
and  for different picture is given as

ρ̂S (0) = ρ̂H (4.5.4)

ρ̂S (t) = Û (t)ρ̂S (0)Û † (t) (4.5.5)
ÂH (0) = ÂS (4.5.6)
ÂH (t) = Û † (t) ÂH (0)Û (4.5.7)

and the expectation value  in each picture is

   
A(t)S = Tr ρ̂S (t) ÂS = Tr Û (t)ρ̂S (0)Û † (t) ÂS (4.5.8)
   
A(t)H = Tr ρ̂H ÂH (t) = Tr ρ̂H Û † (t) ÂH (0)Û
 
= Tr ρ̂S (0)Û † (t) ÂS Û
 
= Tr Û (t)ρ̂S (0)Û † (t) ÂS = A(t)S . (4.5.9)

In the last line, the permutation characteristic of the trace

Tr(ABC) = Tr(BC A) = Tr(C AB). (4.5.10)

is used. As described above, the same expectation value A can be calculated for
each picture. However, it is important to be consistent with the picture that you are
using and should not mix the Schrödinger and Heisenberg picture. It might be a good
idea to remember that only one of ρ̂ or  is time dependent in either picture.
4.6 Unitary Transformation to a Rotating Frame 91

4.6 Unitary Transformation to a Rotating Frame

Often it is desirable to analyze the dynamics of a quantum system from a frame

that rotates at a certain frequency. In classical mechanics, say we are to analyze the
trajectory of a baseball, we often ignore that the earth is rotating. We do that by
looking at the baseball from the point of view of the person on the earth, who is also
rotating with the earth. As we know, the dynamics can be much simpler that way and
the same analogy applies to the dynamics in quantum systems.
As shown in Fig. 4.8, we often encounter a situation where a drive with frequency
ω is introduced to a system with Hamiltonian Ĥ and corresponding frequency (or
energy) ωsys . The unitary transformation which moves the system into the frame
rotating at ω can be written in terms of the Hamiltonian H as
 
Û (t) = exp i(Ĥ(ω)/)t . (4.6.1)
The state vector |ψ is now transformed to |φ = Û (t) |ψ. Now the Schrödinger
equation for this new state |φ is transformed as
d|φ dÛ d|ψ
i = i |ψ + iÛ
dt dt dt
˙
= iÛ |ψ + Û Ĥ |ψ
= iÛ˙ Û † |φ + Û ĤÛ † |φ

= (Û ĤÛ † − iÛ Û˙ ) |φ ,

†
(4.6.2)

where we used 0 = d(Û Û † )/dt = Û˙ Û † + Û Û˙ in the derivation.

†

Ĥ is the Hamiltonian before the transformation and the new Hamiltonian is

expressed as

Ĥ = Û ĤÛ † − iÛ Û˙ .

†
(4.6.3)

As in the baseball analysis on the earth, this unitary transformation let us see
the quantum system from the frame rotating at ω, in which we often analyze the

Fig. 4.8 Driven system with

an external drive with
frequency ω
92 4 Time Evolution in Quantum System

time evolution and/or the steady state of the system. Let’s calculate an example of a
harmonic oscillator and a qubit system below.
The first example is a single-mode harmonic oscillator with Hamiltonian Ĥ =
ωc â † â. Let’s use a unitary transformation Û (t) = exp iωâ † ât to switch to a frame
rotating at ω. Û (t) can be Taylor-expanded, but we know that â † â in the exponential
commute with the Hamiltonian Ĥ ∝ â † â, therefore, Û (t) also commute with the
Hamiltonian.
This simplifies the calculation of the unitary transformation (B.5). From
Û ĤÛ † = Û Û † Ĥ = Ĥ, we see

Ĥ = Ĥ − iÛ Û˙ = ωc â † â − iÛ (−iωâ † â)Û †

†

= (ωc − ω)â † â
= − c â
†
â. (4.6.4)

As you can see, the difference between the frequency ωc of the harmonic oscillator
and the frequency ω of the rotating system, c = ω − ωc appears in Hamiltonian
after being on the rotating frame. c is called detuning, frequently used term in many
of the spectroscopic studies of the quantum system.
We can perform a similar procedure for the unitary transformation of spin-1/2 sys-
tem with Hamiltonian Ĥ = ωa σ̂z /2 with Û (t) = exp iω(σ̂z /2)t . The new Hamil-
tonian in a rotating frame with frequency ω is calculated as Ĥ = − a σ̂z /2. We
see the detuning a = ω − ωa again here as well.
As an advanced example, let us consider the Jaynes–Cummings Hamiltonian

ωa
ĤJC = σ̂z + ωc â † â − ig(σ̂+ â − â † σ̂− ). (4.6.5)
2

The unitary operator is now Û (t) = eiωσ̂z t/2+iωâ ât = Û1 (t)Û2 (t), combination of
†

Û1 (t) = exp iωσ̂z t/2 and Û2 (t) = exp iωâ † ât . This means that both the qubit
and photon of the electromagnetic resonator are viewed from the frame rotating at
the drive frequency ω. Since the harmonic oscillator creation/annihilation operator
(â, â † ) and the qubit Pauli operator (σ̂i ) are commutative, the first two terms of
Jaynes–Cummings Hamiltonian are − a σ̂z /2 −  c â † â as in the previous exam-
ple. However, the interaction term (third term) does not commute with the unitary
operator, and it is not that simple.
In order to pursuit, the Baker–Campbell–Hausdorff formula
  1    1    
e− Ŝ Ĥ e Ŝ = Ĥ + Ĥ , Ŝ + Ĥ , Ŝ , Ŝ + Ĥ , Ŝ , Ŝ , Ŝ + · · · ,
2! 3!
(4.6.6)

is quite useful. Additionally, we note a useful commutation relation â † â, â = −â,

â † â, â † = â † , and σ̂z , σ̂± = ±2σ̂± .
4.6 Unitary Transformation to a Rotating Frame 93

Let us now pursuit by first applying U2 (t) to the interaction term

Û2 (t)(σ̂+ â − â † σ̂− )Û2† (t)

(σ̂+ â − â † σ̂− )e−iωâ
† ât † ât
= eiωâ
= (σ̂+ a − â † σ̂− ) + (σ̂+ â − â † σ̂− ), −iωâ † ât
1
+ (σ̂+ â − â † σ̂− ), −iωâ † ât , −iωâ † ât + · · ·
2!
= (σ̂+ â − â † σ̂− ) + (−iωt)(σ̂+ â + â † σ̂− )
(−iωt)2 (−iωt)3
+ (σ̂+ â − â † σ̂− ) + (σ̂+ â + â † σ̂− ) + · · ·
2! 3!
 
(−iωt)2 (−iωt)3
= 1 + (−iωt) + + + · · · σ̂+ â
2! 3!
 
(−iωt)2 (−iωt)3
− 1 − (−iωt) + − + · · · â † σ̂−
2! 3!
= e−iωt σ̂+ â − eiωt â † σ̂− . (4.6.7)

Now, applying U1 (t) to the above Hamiltonian we get

Û1 (t)(e−iωt σ̂+ â − eiωt â † σ̂− )Û1† (t)

= eiωσ̂z t/2 (e−iωt σ̂+ â − eiωt â † σ̂− )e−iωσ̂z t/2
   
= e−iωt â eiωσ̂z t/2 σ̂+ e−iωσ̂z t/2 − eiωt â † eiωσ̂z t/2 σ̂− e−iωσ̂z t/2 . (4.6.8)

Using the following relation:

    
iωσ̂z t/2 −iωσ̂z t/2 ωσ̂z t 1 ωσ̂z t ωσ̂z t
e σ̂± e = σ̂± + σ̂± , −i + σ̂± , −i , −i
2 2! 2 2
   
1 ωσ̂z t ωσ̂z t ωσ̂z t
+ σ̂± , −i , −i , −i + ···
3! 2 2 2
1 1
= σ̂± ∓ (−iωt)σ̂± + (−iωt)2 σ̂± ∓ (−iωt)3 σ̂± + · · ·
2! 3!
= e±iωt σ̂± , (4.6.9)

the Jaynes–Cummings interaction Hamiltonian in the rotating frame is derived as

Û1 (t)Û2 (t)(−ig)(σ̂+ â − â † σ̂− )Û2† (t)Û1† (t)

= Û1 (t)(−ig)(e−iωt σ̂+ â − eiωt â † σ̂− )Û1† (t)
= −ig(σ̂+ â − â † σ̂− ). (4.6.10)
94 4 Time Evolution in Quantum System

After all the calculations, the Jaynes–Cummings Hamiltonian in the rotating frame
is actually the same as the original Hamiltonian. Writing all the terms together, the
Jaynes–Cummings Hamiltonian in the rotating frame can be given as
 a
ĤJC = − σ̂z −  c â
†
â − ig(σ̂+ â − â † σ̂− ). (4.6.11)
2

4.7 Driven Two-Level System

Let us take a moment to see the effect of strong drive in a two-level system, using the
formulation given above. This situation often arises when the drive field is introduced
to control the two-level system, for example, laser spectroscopy of atoms or ions,
Nuclear Magnetic Resonance (NMR), and Electron Spin Resonance (ESR), usually
use a strong external drive field. Here, we omit the derivation of the interaction
Hamiltonian and other details, and focus on the frame rotation, eigenenergies, and
eigenstates of the driven two-level system.
Let us consider a situation where an atomic ground and excited state with corre-
sponding energy E g = −ωa /2 and E e = +ωa /2 is driven by an external electric
field E(t). For a strong drive E = E0 cosωd t, where E 0  E zpf , the ladder opera-
tor can be approximated as a classical number (â † → α, â → α).1 The interaction
Hamiltonian can be given as
 
Ĥint =  σ̂+ αeiωd t + σ̂− αe−iωd t , (4.7.1)

with a rotating wave approximation as discussed later. The Hamiltonian of entire

system can be written as
ωa  
Ĥ = σ̂z +  σ̂+ αeiωd t + σ̂− αe−iωd t
2
 
ωa
αeiωd t
= 2 . (4.7.2)
αe−iωd t − ω2a

Moving into the rotating frame at drive frequency ωd , the system Hamiltonian now
reads
 a 
Ĥ = − σ̂z + σ̂x
2 2
 − a 
= , (4.7.3)
2  a

√
1 Recall the effect of the ladder operator, e.g., â |n
= n |n − 1. For a large photon number (large
√ √
n) â |n  n |n and â → α = n is justified. Another words, with a strong drive annihilating or
creating photons practically changes nothing to the photon number, and eigenvalue is approximated
as a classical number.
4.7 Driven Two-Level System 95

where a = ωd − ωa is drive detuning and Rabi frequency  = 2α is introduced.

As described previously, the system has no time dependence in this rotating frame.
The eigenenergies of this Hamiltonian is easily found as


 a 2
E+ = + 1+ 2
2 a

 a 2
E− = − 1+ 2
,
2 a

with corresponding eigenstates

|ψ+  = +cosθ |g − sinθ |e

|ψ−  = +sinθ |g + cosθ |e ,

where
−
sinθ = 
2 + ( a +  )2
a +
cosθ =  .
 + ( a +  )2
2


Here we introduced generalized Rabi frequency  = a + .
2 2
The eigenenergies are altered by the laser detuning and the Rabi frequency. Also,
the eigenstates with the drive is a superposition of the original states |g and |e.

Without the Drive

Let us first take a look at the case where there is no drive ( = 0). With sinθ = 0 and
cosθ = 1, we recognize |ψ+  = |g, |ψ−  = |e with corresponding eigenenergies
and energy difference E + = + 2 a , E − = − 2 a , and E = E + − E − =  a .
Note here that these eigenenergies are in the rotating frame with respect to the
drive frequency ωd and the actual atomic energies remains E g and E e , since there is
no drive. This is as if you are running the experiment, sweeping the laser frequency,
but your mate is blocking the laser (don’t do that). The eigenenergies with respect to
the laser detuning a is shown in Fig. 4.9a. Without the drive, there is no coupling
between |g and |e, and the energy levels cross each other at detuning a = 0.

With the Drive: Coherent Coupling and Dressed State

Now, with the drive, the states |ψ±  are the superposition of the bare atomic states.
These states are called dressed states. As opposed to the bare atomic states, these
96 4 Time Evolution in Quantum System

Fig. 4.9 Dressed states. a Without the drive. b With the drive

states are dressed by a sea of photons, creating new eigenstates. The energy eigenval-
ues vary with detuning a and Rabi frequency  with energy difference E =  .
An example of the eigenenergies with a fixed Rabi frequency  is shown in Fig. 4.9b.
At zero detuning ( a = 0), the eigenstates have the equal participation from the
ground and excited state as

1
|ψ+  = √ (|g + |e)
2
1
|ψ−  = √ (− |g + |e) .
2

The eigenenergies show the anticrossing (or avoided crossing) with the gap frequency
of  (or energy of ) at zero detuning. As we show later, this drive coherently trans-
fers the population between the ground and the excited state of the atom. At zero
detuning, the population oscillates between the ground and excited state, making
half-and-half participation of two states. It is important to note that the oscillation
frequency of the population is , which is manifested as a gap between the eigen-
states.
In far-detuned region (| a |  ), the eigenstates are nearly that of the bare
atomic states, |ψ±   |g,|e. Also the eigenenergies can be approximated as
 
 a 2  a 2
E+  1+ = +
2 2 a2 2 4 a
 
 a  2  a 2
E−  − 1+ = − − .
2 2 a2 2 4 a

The eigenenergies are also similar to that of the atomic states, but shifted slightly
by 
2
4 a . This shift is called AC Stark shift, or light shift, and originated from the
strong far-detuned drive. The AC Stark shift could be actively used to control the
state energies, or may happen as an artifact of the control. For example, if we have
an additional state |a to the two-level example shown above, creating a three-state
4.7 Driven Two-Level System 97

system. The drive may be on-resonant for |g-|e transition, but very likely far-
detuned for |g-|a or |e-|a transition, creating an AC Stark shift to the energy
level of |a.
Let us wrap up this section by noting the importance of the anticrossing derived
here. We mentioned that the anticrossing results from the coherent transfer of the
population between |g and |e state. When we say “coherent”, we mean that the
quantumness, such as superposition and entanglement, is preserved and the system
follows the Schrödinger equation as we expect. In a real experiment, it is not that sim-
ple. Many disturbances and noises result in unwanted transitions and deterioration
of the quantumness. In quantum technology, different quantum systems are coupled
to enable various quantum state transfers and manipulations. Experimentally, anti-
crossing of the energy levels is one of the first feature to look for, since it manifests
the signature of the preservation of quantumness and clean coupling between states
in the experimental setup.

Problems
These identities may be useful in the following problem set.

eiθ (n̂·σ ) = Iˆcosθ + i n̂ · σ sinθ

eiθ (n̂·σ ) σ e−iθ (n̂·σ ) = σ cos2θ + n̂ × σ sin2θ + n̂ n̂ · σ (1 − cos2θ )

Problem 4-1 Hamiltonian of a two-level system, such as qubit, without any inter-
action can be written as
ωq
H= σ̂z .
2

(i) Calculate time-evolution unitary operator Û (t) in a matrix repre-

sentation.
(ii) Apply this unitary operator to a qubit initial state |ψ(0) =
a0 |0 + a1 |1 to calculate the qubit state |ψ(t) at arbitrary time
t.
(iii) Discuss the motion of the state vector on a Bloch sphere.

Problem 4-2 Consider a “bare” two-level system Hamiltonian,

ωq
H= σ̂z .
2
Calculate time-dependent operators in Heisenberg picture for the
following operators:
(i) σ̂x , σ̂ y , σ̂z .
(ii) σ̂+ = σ̂x + i σ̂ y /2, σ̂− = σ̂x − i σ̂ y /2.
98 4 Time Evolution in Quantum System

Fig. 4.10 Optomechanical

system

(iii) Show that the commutation relation σ̂x (t), σ̂ y (t) = 2i σ̂z (t),
holds for the operators in Heisenberg picture.

Problem 4-3 In cavity optomechanics, a system Hamiltonian is given as

 
H = ωc â † â + b̂† b̂ − g0 â † â b̂ + b̂† , (4.7.4)

where the first and second terms are the harmonic oscillator Hamilto-
nian of the optical cavity mode (â) and mechanical mode (b̂), respec-
tively. The last term is the optomechanical interaction term. In many
situation, the system is driven with an external drive. Experimentally,
a laser beam is introduced to the optical cavity through the partial
mirror of the optical cavity as shown in Fig. 4.10. With this drive,
the system Hamiltonian is given as
   
H = ωc â † â + b̂† b̂ − g0 â † â b̂ + b̂† + iα âeiωd t − â † e−iωd t ,
(4.7.5)
where α and ωd are the drive amplitude (in frequency) and drive laser frequency,
respectively.
Rewrite the system Hamiltonian above in a rotating frame defined by the cavity
field photon number (â † â) rotating at the pump frequency drive frequency ωd so
that the time dependence of the drive can be eliminated (Fig. 4.10).

Problem 4-4 In a rotating frame, a system of harmonic oscillator with Hamil-

tonian H = ωc â † â can be transformed with a unitary operator
†
Û = eiωâ ât . Using the following relation for ladder operators and
operator function (Baker–Campbell–Hausdorff identity)

â † â, â = −â
â † â, â † = â †
  α2   
ˆ ˆ
eα A B̂e−α A = B̂ + α Â, B̂ + Â, Â, B̂ + · · · ,
2!
calculate
(i) Û â Û †
(ii) Û â † Û †
(iii) Û â † â Û †
4.7 Driven Two-Level System 99

Problem 4-5 Consider two-level states. First state is a superposition state

|ψ1  = a0 |0 + a1 |1 , (4.7.6)

and the second state is a mixed state described as

ρˆ2 = |a0 |2 |00| + |a1 |2 |11| . (4.7.7)

(i) Show density matrix for ψ1 , ρ̂1 = |ψ1 ψ1 | and ρ̂2 in matrix rep-
resentation.
(ii) The density operator for an arbitrary state is known to be
expressed in terms of Pauli operators as

1 ˆ 
ρ̂ = I + α σ̂x + β σ̂ y + γ σ̂z , (4.7.8)
2
where α, β, and γ are all real values, regardless of a pure or mixed
state. Write ρ̂1 and ρ̂2 in this form to find appropriate coefficient α,
β, and γ .
(iii) Discuss the difference in a form of density matrix between two
states in the matrix representation.
Problem 4-6 The density operator for an arbitrary two-level system can be expressed
in terms of Pauli operators as,

1 ˆ 
ρ̂ = I + α σ̂x + β σ̂ y + γ σ̂z , (4.7.9)
2
where α, β, and γ are all real values.
For a system Hamiltonian,

ωq 
Ĥ = σ̂z + σ̂x .
2 2
(i) Use von Neumann equation to derive the equations of motion
(differential equations) for α(t), β(t), and γ (t).
(ii) Consider the case where  = 0 and calculate α(t), β(t), and γ (t)
in terms of the initial value α(0), β(0), and γ (0).
Perturbation Theory
5

5.1 Time Independent Perturbation Theory

Perturbation theory is one of the important tools of quantum mechanics used to

quantitatively determine deviation in the quantum system due to external influences.
A cartoon to explain the concept of perturbation theory is shown in Fig. 5.1. Acircle, 
triangle, and square drawn with solid lines can be considered as eigenstates ψn0 of
0
the0 system with a corresponding eigenenergy E n , and here we focus on the state
ψ , the pink circle. When this quantum system is affected by an external influence
0
(perturbation), the circle is pushed a little and moves to a different position from
the original circle (dotted circle). This new  circle |ψ0  is almost
 the original circle,
but a small portion of the triangular ψ10 and square ψ20 components outside the
original  circle
 are now added, causing the state to deviate slightly from the original
state ψ00 . Of course, as the state changes, the energy of the state also changes,
E 00 → E 0 . In perturbation theory, we learn how this new state and energy  (|ψn , E n )
are represented in terms of the original eigenstates and eigenenergies (ψn0 , E n0 ). The
perturbation theory also provides a way to calculate these changes quantitatively.
Experimentally, this “deviation” appears in various forms. It is possible that the
system shifts due to the effects of electric and magnetic field noise that are not noticed,
or Lasers that are used as a probe for the experiment. These unnoticed deviations
can be a source of trouble, but conversely, this deviation can be used as a tool to
cancel various shifts by actively introducing an external field. Here, we introduce
perturbation theory and some of its applications.

Supplementary Information The online version contains supplementary material available at

https://doi.org/10.1007/978-981-19-4641-7_5.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 101
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7_5
102 5 Perturbation Theory

Fig. 5.1 A cartoon depicting

the concept of perturbation
theory

 0  of the system Ĥ before the external

Consider the Hamiltonian 0 influence is added,

its energy eigenstates ψn , and corresponding energies E n0 , satisfying
   
Ĥ0 ψn0 = E n0 ψn0 , (5.1.1)

where n = 1, 2, . . .. Now let us introduce an external influence (perturbation), so

that the system Hamiltonian changes as

Ĥ0 → Ĥ = Ĥ0 + λV̂ , (5.1.2)

where V̂ is the external perturbation and λ is the parameter which characterize

the strength of the perturbation, which can be regarded as λ  1. With this new
Hamiltonian Ĥ, the eigenstate |ψn  and corresponding eigenenergy E n fulfills

Ĥ |ψn  = E n |ψn  . (5.1.3)

The goal here is to describe  new eigenstate |ψn  and the eigenenergy E n in terms
 this
of the original eigenstate ψn0 and energy E n0 . 

From previous discussion, we know |ψn   ψn0 but not quite equal. Using the
parameter λ to expand the system state and energy into a power series of λ
     
|ψn  = ψn0 + λ ψn1 + λ2 ψn2 + · · · (5.1.4)
En = E n0 + λE n1 +λ
2
E n2 + ··· , (5.1.5)

where
 superscript on E i and ψ i describes the ith
 order
 correction, for example,
ψ , E describes the first-order correction, and ψ 2 and E 2 are the second-order
1 1
n n n n
correction.
5.1 Time Independent Perturbation Theory 103
   
We will derive the form of these  corrections for state ψn1 , ψn2 , . . . , and for

energy E n1 , E n2 , . . . , in terms of ψn0 and E n0 .
Substituting Eqs. (5.1.4, 5.1.5) to expand Eq. (5.1.3) as
      
(Ĥ0 + λV̂ ) ψn0 + λ ψn1 + λ2 ψn2 + · · ·
       (5.1.6)
= (E 0 + λE 1 + λ2 E 2 + · · · ) ψ 0 + λ ψ 1 + λ2 ψ 2 + · · · ,
n n n n n n

and now sort them by the order of λ, we get

   
λ0 : Ĥ0 ψn0 = E n0 ψn0 (5.1.7)
       
λ1 : Ĥ0 ψn1 + V̂ ψn0 = E n0 ψn1 + E n1 ψn0 (5.1.8)
       
λ2 : Ĥ0 ψn2 + V̂ ψn1 = E n1 ψn1 + E n2 ψn0 . (5.1.9)

One can see that λ0 term represents the state of the

 original
 system as we expect.
By taking an inner product of Eq. (5.1.8) with ψn0 
 0 0  1  0  0      
ψn  Ĥ ψn + ψn  V̂ ψn = E n0 ψn0 ψn1 + E n1 ψn0 ψn0 . (5.1.10)

Since Ĥ0 is Hermitian, this equation is also

 0  0  1   0 0  1    
ψn  Ĥ ψn = Ĥ ψn ψn = E n0 ψn0 ψn1 , (5.1.11)

which sets the first-order energy correction E n1 as

   
E n1 = ψn0  V̂ ψn0 . (5.1.12)

 1  1
 
 0  ψn of the wavefunction. Expanding ψn
Next, consider the first-order correction
in terms of the original wavefunction ψm , we have
 1  0
ψ = cm ψm (5.1.13)
n
m=n

The case m = n in the equation

  above  is omitted since we are considering the first-
order correction terms for ψn1 and ψn0 is already included in the 0th correction.
 
Now our goal is to find this cm . Similar to the previous case, substituting ψn1 into
 0
Eq. (5.1.8) and taking an inner product with ψk  shows
           0   0  0
L.H.S. ψk0  Ĥ0 ψn1 + ψk0  V̂ ψn0 = cm ψk0  Ĥ0 ψm + ψk  V̂ ψn
m=n
  0   0  0
= cm E m0 ψk0 ψm + ψk  V̂ ψn
m=n
   
= ck E k0 + ψk0  V̂ ψn0
104 5 Perturbation Theory

        0
R.H.S. E n0 ψk0 ψn1 + E n1 ψk0 ψn0 = E n0 cm ψk0 ψm
m=n

= ck E n0 , (5.1.14)

and setting them equal to each other and solving for ck to derive
 0  0
ψk  V̂ ψn
ck = . (5.1.15)
E n0 − E k0

Finally, substituting this into Eq. (5.1.13), we obtain the first-order correction of the
wavefunction
 0   0
 1 ψm  V̂ ψn  0 
ψ = ψ . (5.1.16)
n m
E n0 − E m0
m=n

The same procedure can be taken to calculate the higher-order corrections. We

only show the result of the second-order energy correction

   2
 0  0 
 ψm V̂ ψn 
E n2 = . (5.1.17)
E n0 − E m0
m=n

The higher-order correction for the energy and wavefunction can be calculated,
however, the accuracy of the correction would be diminished. In general, the correc-
tion up to the first or second order is routinely used.
In the above perturbation theory, the case with degeneracy was omitted. Looking
at the first-order energy correction (5.1.12), there is an energy difference E n0 − E m0 for
corresponding states in the denominator, but this term is 0 between the two degenerate
states. It seems that the energy shift will blow up. Many quantum mechanics textbooks
describe in detail that this problem can be avoided by using a basis in which Ĥ0 and
V̂ can be diagonalized at the same time. Please refer to other textbooks for more
details on the perturbation theory with degeneracies.

5.1.1 Zeeman Effect

Let us look at an example of a time-independent perturbation. For electrons with

orbital and spin angular momentum L and S, respectively, the Zeeman effect is
described with a Hamiltonian
e
V̂ = L̂ + 2Ŝ · Bext , (5.1.18)
2m
5.1 Time Independent Perturbation Theory 105

where L̂ and Ŝ are the operators for electron orbital angular momentum and spin
angular momentum, respectively. e, m, and Bext are elementary charge, electron
mass, and applied external magnetic field, respectively.
First of all, the quantum states of atom appears as a result of many kinfs of pertur-
bations. In the introductory quantum mechanics course, we all study the hydrogen
atom where only the central potential from the nucleus is concerned. On top of that,
a real atom has multiple electrons, causing Coulomb interaction between electrons,
spin–orbit interaction, spin–spin interaction of nucleus and electron spin, and so on.
Studying all of that is another subject and a crash course is given in Appendix D. Here
we assume the electronic state is well defined in the basis |L S J m J , where the total
orbital angular momentum quantum number L, the total spin angular momentum
quantum number S, the total electronic angular momentum quantum number J of
operator defined by Ĵ = L̂ + Ŝ, and the second electronic total angular momentum
quantum number m J are used.
In many experiments in atoms and ions, one may encounter the energy state
assignments, such as 3 P1 and 1 D2 . These are labeled conveniently, using the con-
vention
2S+1
LJ (5.1.19)
where L = S, P, D, . . . are used instead of L = 0, 1, 2, . . ..
The Zeeman effect arises from a simple interaction between electrons in atoms
(and in solids) and an external magnetic field. From that aspect, it is quite important
since the effect is quite general and can appear in various physical systems. Elec-
trons with angular momentum and spin generate an effective magnetic field Bint , the
Zeeman effect can be considered as the interaction between the internal magnetic
field (L̂, Ŝ ∝ Bint ) and the external magnetic field Bext , or more simply, the “magnet”
created by the electronic system reacting with the magnet creating the external field.
When Bext  Bint is satisfied, the effect can be considered as a small perturbation.
The first-order energy correction can be derived using Eq. (5.1.12) as

E L1 S J m J = L S J m J | V̂ |L S J m J 
  e  
 
= LSJmJ L̂ + 2Ŝ · Bext  L S J m J
2m 
e   

= L S J m J  L̂ z + 2 Ŝz L S J m J Bz
2m
e
= L z + 2Sz  Bz
2m
where in the last line we defined z-direction along the applied magnetic field
Bext = Bz ẑ.
The expectation value L z + 2Sz  is worked out as
3 S(S + 1) − L(L + 1)
L z + 2Sz  =  + mJ
2 2J (J + 1)

where
3 S(S + 1) − L(L + 1)
gJ = + (5.1.20)
2 2J (J + 1)
is called Landé g-factor.
106 5 Perturbation Theory

Fig. 5.2 Zeeman shift of energy levels

Finally, the first-order energy correction reads

E L1 S J m J = μB g J m J Bz , (5.1.21)

where μB ≡ 2m e
= 5.79 × 10−5 eV/T is a constant called Bohr magneton.
An example of the Zeeman shift is shown in Fig. 5.2. At Bz = 0, magnetic sub-
levels (states with different m J ) are degenerate. The degeneracy is broken with the
applied magnetic field. The energy shift is linear in Bz and the slope depends on the
g J of state. Some state such as 1 S0 has J = 0 (and m J = 0) and is not susceptible
to the applied magnetic field.
In an actual experiment, the Zeeman effect allows one to adjust the energy of the
atoms by the external applied magnetic field. In some experiments using trapped
ions, a magnetic field gradient (the magnetic field with varying strength in space) is
applied to an array of ions. Each ion in the array now feels a different magnetic field,
thus the energies are different for each ion. The effect is used to individually address
the ions, which now have different resonance frequencies due to the magnetic field
gradient.
Another application is for the quantum sensing. Some system intrinsically has
high magnetic field sensitivity (large energy shift from a small field) can be used as
a magnetic field sensor that can indirectly measure the magnetic field by the energy
shift of the states.
The Zeeman effect can sometimes be a double-edged sword. In experiments where
you do not want the energy to change due to the external field, it is necessary to
suppress the effect by introducing a magnetic shield or field compensating coils.
5.2 Treatment of Time-Dependent Perturbation 107

5.2 Treatment of Time-Dependent Perturbation

Fundamentals of quantum technology lies in the capability to perform quantum state

manipulation. To assure that the quantum system complies, the first step is to isolate
the system to retain the quantum coherence in the system. Secondly, we need to
introduce some kind of external signal from the outside to control and operate the
quantum system. It is important to prescribe a method which describes the time
evolution of the quantum system, under the influence of time-varying perturbation
applied as our control signal.
When a time-varying perturbation V̂ (t) is introduced to the system, Hamiltonian
is perturbed as
Ĥ0 → Ĥ = Ĥ0 + V̂ (t), (5.2.1)
from which we derive an equation of motion of the system.
 0Assume
 that the original Hamiltonian Ĥ0 is time independent with eigenstate
ψ and eigenvalue E 0 satisfying
n n
   
Ĥ0 ψn0 = E n0 ψn0 . (5.2.2)

When incorporating time-dependent perturbation, the Dirac picture described pre-

viously is quite convenient. The time evolution of the wavefunction in Schrödinger
picture is written as
d |ψ(t) S
i = Ĥ |ψ(t) S , (5.2.3)
dt
and in Dirac picture, the wavefunction |ψ(t) I can be written in terms of |ψ S as

|ψ(t) I = ei Ĥ
0 t/
|ψ(t) S . (5.2.4)

Now the time evolution of |ψ(t) I is derived as

 
d |ψ(t) I d |ψ i Ĥ 0
= i ei Ĥ t/ ei Ĥ t/ |ψ S
0 S 0
i +
dt dt 

= ei Ĥ
0 t/
Ĥ − Ĥ0 |ψ S

= ei Ĥ V̂ (t)e−i Ĥ ei Ĥ
0 t/ 0 t/ 0 t/
|ψ S ,

d |ψ(t) I
i = V̂I (t) |ψ(t) I (5.2.5)
dt
0i Ĥ t/ V̂ (t)e−i Ĥ t/ is used. Expanding |ψ(t) in terms
0
 0V̂ I (t) = e
In the last line, I
of eigenstate ψn of unperturbed Hamiltonian Ĥ0 as
 
|ψ(t) I = cn (t) ψn0 (5.2.6)
n
108 5 Perturbation Theory

and substituting this into Eq. (5.2.5), we derive

   
i ċn (t) ψn0 = cn (t)λ V̂I (t) ψn0
n n
 0
= cn (t)λ ei Ĥ
0 t/
V̂ (t)e−i Ĥ
0 t/
ψ
n
n
 
cn (t)λ ei Ĥ V̂ (t)e−i E n t/ ψn0 .
0 t/ 0
= (5.2.7)
n
 0
Taking an inner product of the equation above with ψm 

 0  0  0  i Ĥ t/  
i ċn (t) ψm ψ = cn (t) ψm  e 0 V̂ (t)e−i E n0 t/ ψ 0
n n
n n
 0  
= cn (t) ψm  V̂ (t) ψ 0 ei(E m0 −E n0 )t/ (5.2.8)
n
n
 0  0
and using ψm ψ = δmn
n

 0  
iċm (t) = cn (t) ψm  V (t) ψ 0 ei(E m0 −E n0 )t/ . (5.2.9)
n
n

This is the equation of motion for the state amplitude cm (t). Our goal is to solve this
cm (t) to obtain the time-dependent wavefunction

 0
|ψ(t) I = cm (t) ψm . (5.2.10)
m

Time Evolution of Two-Level System

Based on the previous results, let us look at an example of the time evolution of a
quantum system with an added perturbation.
Consider a two-level system, such as the ground state and the first excited state
of an atom. As shown in Fig. 5.3, the population of the ground and excited states

Fig. 5.3 Time evolution of

two-level system
5.2 Treatment of Time-Dependent Perturbation 109

vary in time when an external field oscillating at frequency ωd is applied to this

two-level system. Unperturbed system Hamiltonian Ĥ0 of the two-level system and
the perturbation V̂ (t) is given as
   
E 10 0 ω1 0
Ĥ0 = =  (5.2.11)
0 E 00 0 ω0

  −iωd t

0 2e
V̂ (t) =   iωd t , (5.2.12)
2 e 0

where , called Rabi frequency, can be regarded as an effective amplitude of the

perturbation.
The time evolution of the state amplitude, Eq. (5.2.9), can be written as

     i(ω1 −ω0 )t −iωd t

iċ1 (t) = c0 (t) ψ10  V̂ (t) ψ00 ei(ω1 −ω0 )t = c0 (t) e e
2
     i(ω0 −ω1 )t iωd t
iċ0 (t) = c1 (t) ψ00  V̂ (t) ψ10 ei(ω0 −ω1 )t = c1 (t) e e
2
 
c (t)
which can be conveniently expressed using the matrix representation c(t) = 1
c0 (t)
as,
    i(ω1 −ω0 )t −iωdt
 
ċ1 (t) 0 e e c1 (t)
i =   i(ω0 −ω1 )t iωdt 2
ċ0 (t) 2e e t 0 c0 (t)
 −i(ω −ω )t 
 e d 10 c (t)
0
= i(ωd −ω10 )t c (t) , (5.2.13)
2 e 1

where ω10 = ω1 − ω0 . Eliminating one state amplitude and writing it in terms of

c1 (t), a differential equation

2
c̈1 (t) + i(ωd − ω10 )ċ1 (t) + c1 (t) = 0, (5.2.14)
4
is derived. Solving the differential equation and defining the detuning  = ωd − ω10 ,
the solution is

c1 (t) = Ae−it/2 sin t, (5.2.15)
2
 1/2
where  = 2 + 2 is called generalized Rabi frequency.
The coefficient A is calculated from the initial condition c0 (t = 0) = 1, c1 (t =
0) = 0 as
 1/2
2
A= . (5.2.16)
2 + 2
110 5 Perturbation Theory

Finally, the population of state |0 and |1 are derived as

2 2 2 1 
|c1 (t)|2 = sin t = 1 − cos t (5.2.17)
2 + 2 2 2 + 2 2
|c0 (t)| = 1 − |c1 | .
2 2
(5.2.18)

A cartoon of the two-level system dynamics calculated above is shown in Fig. 5.4.
Let us first assume that the driving field is on resonance, meaning the driving field
frequency matches the frequency difference between the ground and excited state
 = ωd − ω10 = 0 →  = . Initial state at t = 0 is the ground state |0. The time
variation of the population simplifies to

1
|c1 (t)|2 = (1 − cost) (5.2.19)
2
1
|c0 (t)|2 = (1 + cost) . (5.2.20)
2
The external field drives the population from |0 to |1 and oscillates between the
two states. This population oscillation is called Rabi oscillation, which is one of the
most fundamental quantum state manipulation.
On resonance, the occupation probability in the ground and the excited state
oscillates between 0 and 1. Conversely, if ωd is off resonant ( = ωd − ω10 = 0),
only a partial population is going to oscillate. Also, at off-resonant condition, the
oscillation frequency  will increase with the detuning .
As long as the external field continues to perturb the system, the occupation
probability oscillates, but when the external field is turned off, the oscillation stops
and the population at that time is maintained (Fig. 5.4c). If the states |0 and |1
are regarded as a bit of information 0 and 1, the two-level system can be used as a
qubit. The timing of the drive turn-off can be controlled to generate a superposition
state |ψ = c0 |0 + c1 |1 with an arbitrary population difference. Rabi oscillation
is the most fundamental manipulation of a qubit. The detailed derivation of Rabi

Fig. 5.4 Dynamics of two-level state system. a Two-level system with the state energy E 0 and
E 1 . In addition to the unperturbed Hamiltonian Ĥ0 , which defines the two-level system and the
perturbation V̂ , an external field with frequency ω is added. We assume the initial state of the
system is in the ground state |0. b The occupancy probability of the two-level system affected by
the external field Rabi-oscillates between the states |0 and |1. c When the external field is turned
off, the occupation probability stops oscillating and stays at where it ended
5.2 Treatment of Time-Dependent Perturbation 111

oscillation using the Bloch equation and the visualization using the Bloch sphere
will be described later in Sect. 6.4.
Fermi’s Golden Rule
In the previous section, Rabi oscillation was considered in a two-level system, and
found the occupation probability oscillates between states at Rabi frequency .
Often in experiments, it becomes necessary to calculate the transition probability
per unit time from a certain initial state |i to the final state | f , that is, the transition
rate. Fermi’s golden rule is a method for calculating this transition rate using the
time-dependent perturbation theory.
From Eq. (5.2.9), the final state time evolution can be written as
   
i 
cn (t) ψ 0f  V̂ (t) ψn0 ei(E f −E n )t/
0 0
ċ f (t) = −
 n
   
i 
=− cn (t) ψ 0f  V̂ (t) ψn0 eiω f n t ,
 n

where in the last line, ωi j = (E i0 − E 0j )/ is used. By integrating this equation, one
obtains
 t  
i   
c f (t) − c f (0) = − dt cn (t ) ψ 0f  V̂ (t ) ψn0 eiω f n t . (5.2.21)
 n 0

For a weak perturbation, we can assume cn (t)  cn (0). Also, we assume that at
t = 0, only the initial state is populated as ci (0) = 1 (cn (0) = 0 for n = i), then this
equation is simplified as
  
i t
  
c f (t)  − cn (0) dt ψ 0f  V̂ (t) ψn0 eiω f n t (5.2.22)
 n 0
 t  
i   
= − ci (0) dt ψ 0f  V̂ (t) ψi0 eiω f i t (5.2.23)
 0
  
i t   
=− dt ψ 0f  V̂ (t) ψi0 eiω f i t . (5.2.24)
 0
Now consider two scenarios. The first case is when the perturbation V̂ does not
change in time (though we are thinking of time-dependent perturbation theory). It
sounds a bit confusing, but experimentally, for example, we can consider a situation
in which a DC electric field is suddenly applied and then observe how the system
develops.
The probability amplitude c f (t) in the final state is written as

i  0   0  t
c f (t) = − ψ f  V̂ ψi dt eiω f i t (5.2.25)
 0
1  0   0  iω f i t
=− ψ f  V̂ ψi (e − 1), (5.2.26)
ω f i
112 5 Perturbation Theory

and the transition probability of the final state Pi→ f = |c f (t)|2 is derived as

(eiω f i t − 1)(e−iω f i t − 1)  0   0 2

Pi→ f =  ψ f  V̂ ψi  , (5.2.27)
2 ω 2f i

which simplifies to
ωfit 2
4sin    2
2  0  0 
Pi→ f =  ψ f  V̂ ψi  . (5.2.28)
 ωfi
2 2

The denominator ω f i can be very large value, such as in the case of optical transitions,
and the transition probability decreases as the square of ω f i .
For a short time interval or “right at the beginning”, the transition probability is
ωfit 2
4    2
2  0  0 
Pi→ f   ψ f  V̂ ψi 
2 ω 2f i
t 2  0   0 2
=  ψ f  V̂ ψi  ,
2
d Pi→ f
from which the transition rate to the final state  f = dt is calculated as

t  0   0 2
f =  ψ f  V̂ ψi  . (5.2.29)
2
For the second case, we consider a sinusoidal perturbation V̂ (t) = V̂0 e−iωt .
Experimentally, this is quite common, since just about any spectroscopy using
RF/microwave to optical electric field fits into this. Similar to the previous case,
consider the probability amplitude of the final state c f (t), which is written as
    
i t  
c f (t) = − dt ψ 0f  V̂0 ψi0 e−i(ω−ω f i )t (5.2.30)
 0
   
1  
= ψ 0  V̂0 ψi0 (e−i(ω−ω f i )t − 1), (5.2.31)
(ω − ω f i ) f

from which the transition probability of the final state Pi→ f is calculated as

(ω−ω f i )t 2
4sin     2
2  0  0 
Pi→ f =  ψ f  V̂0 ψi  . (5.2.32)
2 (ω − ω f i )2

When the perturbation frequency ω is the same as the frequency difference of the
states ω f i , fulfilling the resonant condition (ω = ω f i ), and for a large t (t → ∞),
one can use the relation
sin2 (αt)
= πδ(α), (5.2.33)
α2 t
5.2 Treatment of Time-Dependent Perturbation 113

which simplifies the transition probability as

    2
   
πt  ψ 0f V̂0 ψi0 
Pi→ f = . (5.2.34)
2
d Pi→ f
Transition rate to the final state  f = dt now takes a simple form

    2
   
π  ψ 0f V̂0 ψi0 
f = , (5.2.35)
2

which is called a Fermi’s golden rule which allows to calculate the transition rate
from the perturbation.
So far, we have been focusing on the transition from the initial state |i to a single
final state | f  as shown in Fig. 5.5a.
As in Fig. 5.5, if there are multiple or continuum of final states, the total transition
probability can be expanded as

Single final state Ptot = Pi→ f

Multiple final states Ptot = Pi→ fn
n

Continuum of final states Ptot = Pi→ f ρ(E f )d E,

where ρ(E) is the density of states. It is probably straightforward in multiple final

states, where the transition probability is a sum of all the different final states. For

Fig. 5.5 Transition probability to various final state. a A single final state, b multiple final states,
and c continuum of final states
114 5 Perturbation Theory

the case of a continuum, the density of states ρ(E) is introduced. The density of state
has a dimension
Number of states at E
ρ(E) = , (5.2.36)
Unit energy
which allows for the incorporation of all the final states within a continuum with a
form ρ(E f )d E.

5.2.1 Time-Dependent Perturbation Expansion

In the case of a two-level system with a sinusoidal drive, simple oscillatory dynamics
in the occupation probability was observed. However, in general, the dynamics of
time-dependent perturbation V̂ (t) is difficult to calculate because there is no analyti-
cal solution for the perturbed Hamiltonian Ĥ = Ĥ0 + V̂ (t) for most cases. Here, we
describe an approximation method that can be calculated by perturbation expansion
of the time dependence of the external field. 
As mentioned previously, the eigenstate ψn0 and the eigenenergy E n0 of the system
before adding the perturbation satisfies
   
Ĥ0 ψn0 = E n0 ψn0 . (5.2.37)

The external perturbation alters the system Hamiltonian as

Ĥ0 → Ĥ = Ĥ0 + λV̂ . (5.2.38)

The time evolution of the wavefunction using the Dirac picture is generally described
as
 
|ψ(t) I = cn (t) ψn0 . (5.2.39)
n
We can expand the state probability amplitude cn (t) in the power series of λ as

cn (t) = cn0 (t) + λcn1 (t) + λ2 cn2 (t) + · · · . (5.2.40)

 0  the goal is to describe this coefficient cn using V̂ (t) and the unperturbed basis
Now, i
ψ .
n
First, the time evolution from the initial state |ψ(t0 ) I ≡ |i at t = 0 in the Dirac
picture can be written using the time-evolution operator Û I (t, t0 ) as

|ψ(t) I = Û I (t, t0 ) |i . (5.2.41)

Substituting the above to the Schrödinger equation in Dirac picture

d |ψ(t) I
i = λV̂I (t) |ψ(t) I , (5.2.42)
dt
5.2 Treatment of Time-Dependent Perturbation 115

we obtain
d Û I (t, t0 )
i |i = λV̂I (t)Û I (t, t0 ) |i . (5.2.43)
dt
This equation is satisfied for an arbitrary initial state |i, and therefore, it can be
rewritten as
d Û I (t, t0 )
i = λV̂I (t)Û I (t, t0 ) , (5.2.44)
dt

which is the equation of motion for the unitary operator. Here, we used V̂I =
ei Ĥ t/ V̂ e−i Ĥ t/ . Û I (t, t0 ) represent the time-evolution operator from the initial
0 0

time t0 with a boundary condition Û (t0 , t0 ) = Iˆ.

Integrating this equation of motion, the left-hand-side of the equation is
  
t d Û I (t , t0 )
i dt = i Û I (t, t0 ) − Û I (t0 , t0 )
t0 dt
 
= i Û I (t, t0 ) − Iˆ , (5.2.45)

and now the form of Û I (t, t0 ) is written as

 t
i
Û I (t, t0 ) = Iˆ − dt λV̂I (t )Û I (t , t0 ). (5.2.46)
 t0

The equation above is a nested equation in Û I (t, t0 ) and it can be expanded as


i t
Û I (t, t0 ) = Iˆ − dt λV̂ I (t )Û I (t , t0 )
 t0
   t
i t i 2 t
= Iˆ − dt λV̂ I (t ) + − dt dt λ2 V̂ I (t )V̂ I (t )Û I (t , t0 )
 t0  t0 t0
   t
ˆ i t i 2 t
= I− dt λV̂ I (t ) + − dt dt λ2 V̂ I (t )V̂ I (t ) + · · ·
 t0  t0 t0
(5.2.47)
∞   tn−1
i n t
= − dt1 . . . dtn λn V̂ I (t1 )V̂ I (t2 ) . . . V̂ I (tn ). (5.2.48)
 t 0 t 0
n=0

In the last line, the term n = 0 in the sum is Iˆ and the time variable is rewritten as
t → t1 , t → t2 , · · · → tn for a simplicity.
Since the form of the time evolution operator Û I (t, t0 ) is obtained, now we con-
sider the time evolution from the initial state |i using Eq. (5.2.41).
 
|ψ(t) I is expanded using the unperturbed eigenstate |n ≡ ψn0 as
116 5 Perturbation Theory

|ψ(t) I = Û I (t, t0 ) |i = |nn| Û I (t, t0 ) |i

n
= n|Û I (t, t0 )|i |n (5.2.49)
n
  t
i
= n| Iˆ|i + λ − dt1 n|V̂ I (t1 )|i
n
 t0
  t1 
i 2 t
+λ2 − dt1 dt2 n|V̂ I (t1 )V̂ I (t2 )|i + · · · |n .
 t0 t0
(5.2.50)

Also, by making correspondence with Eqs. (5.2.39) and (5.2.40)

|ψ(t) I = cn (t) |n

n
 
= cn0 (t) + λcn1 (t) + λ2 cn2 (t) + · · · |n , (5.2.51)
n

each term cni (t) can now be derived as

cn0 (t) = n| Iˆ|i = δni (5.2.52)


i t
cn1 (t) = − dt1 n|V̂I (t1 )|i
 t0

i t
=− dt1 n|V̂ (t1 )|i eiωni t1 (5.2.53)
 t0
 t  t1
1
cn2 (t) = − 2 dt1 dt2 n|V̂I (t1 )V̂I (t2 )|i
 t0 t0
 t  t1
1
=− 2 dt1 dt2 n|V̂I (t1 ) |mm| V̂I (t2 )|i
 m t0 t0
 t  t1
1
=− 2 dt1 dt2 n|V̂ (t1 )|m m|V̂ (t2 )|i eiωnm t1 +iωmi t2 .
 m t0 t0
(5.2.54)

While the derivation, V̂I = ei Ĥ V̂ e−i Ĥ

0 t/ 0 t/
is used to modify n|V̂I (t)|i as

n|V̂I (t)|i = n|ei Ĥ t/ V̂ e−i Ĥ

0 0 t/
|i (5.2.55)
= n|V̂ (t)|i eiωni t , (5.2.56)

where ωni = ωn − ωi .
Time-dependent perturbation theory is often used to calculate the transition prob-
ability to state |n from an initial state |i. When the final state |n of interest is
5.2 Treatment of Time-Dependent Perturbation 117

Fig. 5.6 A particle located

at the bottom of the potential
landscape

different from the initial state, cn0 (t) = 0 and the first-order approximation cn1 (t) is
important. The first-order approximation of the transition probability Pi→n is
  t 2
1 
Pi→n =  dt n|V̂ (t )|i eiωni t  .
 (5.2.57)
i 0

Problems

Problem 5-1 In order to observe quantum phenomena, the experimental system or

a sample of interest is often cooled by a refrigerator or some active
method, such as laser cooling. At a low enough temperature, the
system of interest should be in a ground state, meaning the bottom of
a potential well as shown in Fig. 5.6. The particle position in ground
state is x = x0 at the bottom of the well defined by a potential curve
f (x).
Show that for a small amplitude (or energy), the system can always be
expressed as a harmonic oscillator. (This is not a quantum problem,
but important to recognize why we use the harmonic oscillator as the
building blocks in quantum technology.)
[Hint: Use Taylor expansion.]
Problem 5-2 Consider adding an extra “push” to a harmonic oscillator so the
classical energy of the system reads

p2 mω 2 x 2
E= + + αx.
2m 2
This equation can be rewritten with a displaced equilibrium position
x = 0 → x0 as

p2 mω 2 (x − x0 )2
E= + − E offset .
2m 2
118 5 Perturbation Theory

Now consider the same situation for a quantum harmonic oscillator,

where the system Hamiltonian is described as

Ĥ = Ĥ0 + V̂
= ω â † â + α x̂
= ω â † â + αxzpf (â + â † ),


where xzpf = 2mω is zero-point displacement fluctuation of the
harmonic oscillator.
(i) Solve the classical equilibrium position x0 and energy offset
E offset .
(ii) Use perturbation
 theory
 to calculate the first-order correction for
the wavefunction ψn1 .  
(iii) Use the result in (ii) to calculate the new ground state 0 .
(iv) Find the new ground state energy E 0 by calculating up to second-
order correction.  
(v) Find the expectation value of position x for 0 .
(vi) Compare E 0 and x with x0 and E offset obtained in (i).
Problem 5-3 The Hamiltonian and eigenenergy of a harmonic oscillator is given
as

Ĥ0 = ω0 â † â
E n = nω0

and the energy spacing between states E n = E n − E n−1 = ω0 are

all the same. LC circuit (or LC resonator) is a harmonic oscillator
and have the same Hamiltonian and the energy spacing. In a super-
conducting circuits, there is a nonlinear circuit component called
Josephson junction and its variants, which can introduce a nonlinear
response to the harmonic system.
Suppose, a nonlinear circuit component added to a LC resonator has
a perturbation a form
V̂ = αâ † â â † â.
(i) Write a system Hamiltonian Ĥ = Ĥ0 + V̂ .
(ii) Find the eigenenergy of the system E n = E n0 + E n1 including up
to a first-order energy correction.
(iii) Find the energy separation E n of the system.
(iv) Discuss that this system can actually be used as a qubit.
Problem 5-4 External control is applied to a qubit forming a new Hamiltonian

ωq
Ĥ = Ĥ0 + V̂ = σ̂z + ασ̂x ,
2
5.2 Treatment of Time-Dependent Perturbation 119

Fig. 5.7 Sinusoidal perturbation being slowly turned on

where α  ωq .    
(i) Find the new state 0 and 1 in terms of the original qubit state
|0 and |1, using up to a first-order perturbation.
(ii) Find the energies of the
 new states
 E 0 and E 1 .
(iii) Can the new states 0 and 1 be used as a basis for qubit, other
words, are they orthogonal?
Problem 5-5 Consider a slow turn-on of a sinusoidal perturbation

V̂ (t) = V̂0 e−iωt eγt (5.2.58)

as shown in Fig. 5.7.

(i) Calculate the transition probability Pi→ f .
[Hint: Since it is a “slow” turn-on, take the time integral from
t = −∞].
(ii) Find the transition rate  f .
 Part II
Harmonic Oscillator, Qubit and Coupled
Quantum Systems
Harmonic Oscillator
6

6.1 Harmonic Oscillator and Its Hamiltonian

Lights, sounds, vibrations—such waves we meet in our daily lives are mostly under-
stood as harmonic oscillations supported in various media. Harmonic oscillators are
frequently expressed in the form of Hamiltonian with canonical variables q and p,
as1
p2 1
H= + mω 2 q 2 .
2m 2
Here m is the mass of an object and ω is the angular frequency of the harmonic
oscillation. To switch to quantum mechanics, canonical variables should be replaced
by operators q → q̂ and p → p̂ = (/i)d/dq. Whenever there is no confusion, we
omit hats on the operators. √
By defining
√ annihilation and creation operators by a = mω/2 [q + (i/mω) p]
and a † = mω/2 [q − (i/mω) p], above Hamiltonian can be rewritten as
 
1
H = ω a a +†
2
with a set of eigenstates {|n ; n = 0, 1, 2, · · · } which are called Fock states. Since
the index n in |n represents the number of quanta of the harmonic oscillation, the
Fock states are also dubbed as number states.

Supplementary Information The online version contains supplementary material available at

https://doi.org/10.1007/978-981-19-4641-7_6.

1 Light quantum, photon, is an exception for its massless nature. Nevertheless, we can theoretically

understand it as a harmonic oscillator with slightly different but essentially the same Hamiltonian.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 123
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7_6
124 6 Harmonic Oscillator

Fock states constitute a complete orthonormal basis. Fock state |n has the mean
photon number n and vanishing standard deviation. For these properties and their
physical comprehensibility, Fock states are often used as a fundamental tool to expand
and analyze other quantum states given in this section.

6.2 Electromagnetic Waves and Ladder Operators

6.2.1 Quantization of Electromagnetic Waves

For a quantum-engineering purpose, there is a need to flip the state between the
ground state and excited state, or generate superposition of them by some means.
In general, the ground (|g) and excited (|e) states are non-degenerate, separated in
energy by ωq , and in most systems in reality the quantum state refers to the electronic
and/or spin states. Therefore, the flip operation is done by applying electromagnetic
waves with its frequency around ωq , the transition frequency.
Again for the purpose of quantum engineering, it is often necessary to treat electro-
magnetic fields in a quantum fashion. The quantum theory of light is really insightful
and delightful to learn; however, it is beyond the scope of this book to fully describe
it without messing up the main streamline of the story. So that we shall only scratch
the surface of it and pick several useful topics up. For more detail, please find them
in the literature [8].
We shall focus first on a confined optical mode, or a cavity mode, where one might
imagine an optical resonance mode such as the one in the Fabry–Perót resonator and
any other types of it. To simplify the situation, a single cavity mode is considered
that yields the energy
  
1 1 2
E= ε0 E + B dV
2
(6.2.1)
2 cavity μ0
= 2ε0 Vcavity ωc2 A0 · A∗0 (6.2.2)

where E, B, and A0 denote the electric field, magnetic field, and amplitude of vector
potential,2 ε0 and μ0 the vacuum permittivity and permeability, Vcavity and ωc the
mode volume and the resonant frequency of the cavity mode under concern. By
introducing canonical
 variables q and p together with the polarization vector ε to
rewrite A0 = 1/4ε0 Vcavity ωc2 (ωc q + i p)ε, the energy reads
1 2
E= ( p + ωc2 q 2 ) (6.2.3)
2
which is manifestly a harmonic oscillator. The canonical variables q and p are
replaced by operators q̂ and p̂ to make this Hamiltonian quantum mechanical.3 With

2 Itis assumed that vector potential have the form A0 e−iωc t+ik·r + A∗0 eiωc t−ik·r . Coulomb gauge
and the relations E = −∂A/dt and B = ∇ × A are used as well.
3 All “hats” in the quantum-mechanical operators are not shown throughout this book unless there

is a confusing situation.
6.2 Electromagnetic Waves and Ladder Operators 125

the classical picture, one can imagine continuous trajectory of (q, p) in classical
harmonic oscillator. There q and p are oscillating with the relative phase of π/2
to enclose a circle in the phase space. In quantum-theoretic treatment, the state of
photon is not a point in a phase space but rather a distribution. In analogy with the
classical wave-like understandings of light the variables q and p are called in-phase
and quadrature components ofthe quantum states. As in the usual prescription, we
can define alternatively A0 = /2ε0 Vcavity ωc aε using annihilation operator a to
obtain
 
1
E = ωc a a +†
. (6.2.4)
2

This is a familiar expression of a harmonic oscillator in quantum mechanics and

we mostly omit the zero-point energy ωc /2. This expression is well-understood by
noticing that the operator n̂ = a † a counts the number of photons in the mode. In
such an expression, the electric and magnetic fields are written as

ωc
E=i ε(ae−iωc t+ik·r − a † eiωc t−ik·r ), (6.2.5)
2ε0 Vcavity

ωc
B=i k̂ × ε(ae−iωc t+ik·r − a † eiωc t−ik·r ). (6.2.6)
2ε0 Vcavity c2

Here k̂ is a unit vector parallel to the wavevector. Note that in the rotating frame such
that the time-dependent factors associated with a and a † are hidden, and properly
choosing the phase, these reduce to

E = E zpf (a + a † )ε, (6.2.7)

B = Bzpf (a + a † )k̂ × ε. (6.2.8)

 
where E zpf = ωc /2ε0 Vcavity and Bzpf = ωc /2ε0 Vcavity c2 are vacuum ampli-
tudes or the zero-point fluctuations of the cavity field: the smaller the cavity, the larger
the vacuum amplitudes. The subscript “zpf” stands for the zero-point fluctuation.

6.2.2 Ladder Operators of a Harmonic Oscillator

At this point, let us review the properties of

 the annihilation and creation operators
a and a † . From the commutation relation a, a † = 1, we can immediately see that
 
n̂, a = −a and n̂, a † = a † . Now we want to consider the eigenstate |n of the
126 6 Harmonic Oscillator

number operator n̂ = a † a. As the reader can see, we discriminate the number operator
n̂ and its eigenvalue n in this section. The eigenvalue n is the number of quanta:
n̂ |n = n |n. Then let us try operating above commutation relations on |n:

(n̂a − a n̂) |n = (n̂a − na) |n = −a |n ,

(n̂a † − a † n̂) |n = (n̂a † − na † ) |n = a † |n .

These are rewritten as

n̂(a |n) = (n − 1)(a |n), (6.2.9)

n̂(a |n) = (n + 1)(a |n),
† †
(6.2.10)

which can be interpreted as follows. The quantum states (a |n) and (a † |n) are
eigenstates of the number operator n̂, respectively, with the eigenvalues n − 1 and
n + 1. In other words, the operator a makes the number of quanta decrease by 1
and a † does the number of quanta increase by 1, manifesting themselves as ladder
operators.

6.3 Quantum States of a Harmonic Oscillator

6.3.1 Coherent State

Fock states are the eigenstates of the harmonic oscillator, but in real life it is hard
to encounter them. Real life’s single-mode harmonic oscillations mostly appear as
sinusoidal, monochromatic waves. Then what quantum state is the closest to the
classical, monochromatic wave? Remember that the monochromatic wave in the
classical mechanics can be generated by the forced oscillation. Therefore, we would
like to consider the forced oscillation in a quantum manner.
Driving a harmonic oscillator can be done by injecting energy into the oscil-
lator by a beam-splitter interaction Hd = i(ba † − b† a), where b denotes the
annihilation operator of the driving field. The driving field is usually quite intense
that it is replaced by the classical amplitude β. By mocking up the coefficients
as β = α/τ with τ being the time duration of the drive, the interaction reads
Hd = i (α/τ )a † − (α∗ /τ )a and gives the unitary evolution

Hd † −α∗ a
D(α) = e−i  τ
= eαa .

The unitary operator D(α) is called the displacement operator. To see why it is
“displacement,” we shall see important properties of D(α).
First, suppose that the vacuum |0 is subject to the forced oscillation. The resultant
def
state is |α = D(α) |0 and we expect |α is the closest quantum analogue of the
6.3 Quantum States of a Harmonic Oscillator 127

classical oscillation. Then let us make a act on |α:

† −α∗ a
a |α = aeαa |0
∗ |α|2
= aeαa e−α a e−
†
2 |0
2
∗ − |α|2
= αeαa e−α a e
†
|0
αa † −α∗ a
= αe |0
= α |α .

In the second and fourth lines, the Baker–Campbell–Hausdorff formula e A e B =

e A+B+[A,B]/2 is used. In the third line, the factor aeαa is Taylor-expanded and the
†

repetitive use of commutation relation a, a † = 1 yields eαa a + αeαa . The first
† †

of the two terms acts on |0 to give zero. From above calculation, |α is revealed to
be an eigenstate of the annihilation operator a. √
√ Now, estimating the expectation values of q = 2/mω(a + a † )/2 and p =
2mω(a − a † )/2i is not so difficult. Since a |α = α |α and α| a † = α| α∗ ,

2 α + α∗ 2
q = α| q |α = = Re[α],
mω 2 mω
√ α − α∗ √
 p = α| p |α = 2mω = 2mωIm[α].
2i
In order to calculate the standard deviations of q and p, we can utilize the equalities
α| aa |α = α2 , α| a † a |α = |α|2 and α| aa † |α = |α|2 + 1 to get

Δq ≡ q 2 − q2

2 α2 + α∗2 + 2|α|2 + 1 − α2 − 2|α|2 − α∗2
=
mω 4

= ,
2mω

Δp ≡ p 2 −  p2

−α2 − α∗2 + 2|α|2 + 1 + α2 + α∗2 − 2|α|2
= 2mω
4
mω
= .
2
If we consider the √ p plane, the state |α is represented
phase space, or q√ √ as a distri-
bution
√ located at ( 2/mωRe[α], 2mωIm[α]) with the extent /2mω in q-
and mω/2 in p-axes. Note that ΔqΔ p = /2, implying that the state |α is the
minimum-uncertainty state. Remembering the classical harmonic oscillation, where
128 6 Harmonic Oscillator

Fig. 6.1 Schematic

illustration of the coherent
state in the phase space

the phase-space representation is a point in a rotating frame, the state |α certainly
seems to be the quantum version of the forced oscillation of the harmonic oscillator.
The state |α is thus deduced to be phase-coherent, by which the term coherent state
is given (Fig. 6.1).
We shall rewrite the state |α in terms of the Fock states: |α = ∞
n=0 An |n.
√ From
the equality a |α = α |α,√one can get the recurrence relation√An+1 = α An / n + 1
and deduce An = αn A0 / n!. Therefore, |α = ∞ n=0 (α / n!)A0 |n. The coeffi-
n

cient A0 can be determined by the normalization condition α |α = 1 as A0 =

e−|α| /2 , so that the final form is
2

∞
αn |α|2
|α = √ e− 2 |n . (6.3.1)
n=0
n!

The coefficient squared represents the number distribution of the coherent state:

|α|2n −|α|2
Pn = e .
n!
The standard deviation of the photon number reads

2
Δn ≡ (a † a)2 − a † a = |α|

and together with Pn , the number distribution of the coherent state is revealed to be
a Poissonian one.
We should keep in mind that the set of coherent state {|α ; α ∈ C} is said to be
overcomplete. This stands for the fact that with this basis set one can express any
quantum state composed of a harmonic oscillator; however, any two coherent states
are not rigorously orthogonal to each other. Let us check this by calculating β| α
for α, β ∈ C. To do this math, we can use the following formula:
∗ −α∗ β)/2
D(α)D(β) = e(αβ D(α + β)
6.3 Quantum States of a Harmonic Oscillator 129

which states that the two sequential application of displacement operators with α
and β results in the displacement by α + β with a global phase acquired. This leads
to the following result:

β| α = 0| D † (β)D(α) |0

∗ +β ∗ α)/2
= 0| e(−βα D(α − β) |0
(−βα∗ +β ∗ α)/2
=e 0|α − β
(−βα∗ +β ∗ α)/2
e−|α−β|
2 /2
=e 0| (|0 + |α − β| |1 + · · · )
−(|α|2 +|β|2 −2β ∗ α)/2
=e
= δ(α − β).

Another thing to note here in the end of this section is that the coherent state
analyzed here is a quantum analogue of the classical sinusoidal, forced oscillation,
as described in the beginning. Then why are we dealing with such a stuff? How such
a “classical” state appear in the quantum technology that we are to learn? We have
a simple answer that the electromagnetic waves we have in laboratories and even
in daily lives are the coherent states and principal tool we can poke the quantum
object with is lasers and microwaves in the coherent states. We need to understand
our indispensable tool of electromagnetic waves in terms of the coherent state and
apply the knowledge to properly control the quantum states of photons, atoms, and
artificial quantum systems.

6.3.2 Schrödinger’s Cat State

An interesting states are realized by superposing two coherent states |α and |−α,
which are called Schrödinger’s cat states, or simply the cat states. Here for simplicity
we take the complex number α to be real and consider two cats:
|α + |−α
|cateven  = ,
Aeven
|α − |−α
|catodd  =
Aodd
which are called even cat and odd cat, respectively. Aeven and Aodd normalizes
the states here. Let us express these states in the Fock basis {|n}. By looking at
Eq. (6.3.1), it can be seen that the even (odd) cat state contains only the even (odd)
number Fock states and concrete expressions are given as
∞

2 α2n − α2
|cateven  = √ e 2 |2n ,
Aeven
n=0
(2n)!
∞

2 α2n+1 α2
|catodd  = √ e− 2 |2n + 1 .
Aodd
n=0
(2n + 1)!
130 6 Harmonic Oscillator

These even and odd cat states are apparently orthogonal to each other, manifest-
ing themselves as a simplest example of qubit encoding in the infinite-dimensional
Hilbert space held by a harmonic oscillator.

6.3.3 Squeezed State

Light-matter interaction is so diverse that various nonlinear interactions among pho-

tons can take place. Nonlinear optical phenomena are characterized by their order
m where m + 1 photons are involved in the nonlinear process with strength char-
acterized by nonlinear coefficient χ(m) . The most simple, exemplary process is mth
harmonic generation where m-photon absorption (emission) and one-photon emis-
sion (absorption) occurs at once in a two-level system or certain crystals. Sum- and
difference-frequency generation are classified as χ(2) processes. Frequently appear-
ing nonlinearities are χ(2) nonlinearity, which gives the interaction Hamiltonian with
the products of three operators ig(aaa † − a † a † a), and χ(3) nonlinearity with four
operators. By applying a monochromatic drivings and replace one or two operators
by classical numbers in χ(2) or χ(3) Hamiltonian, one can get so-called squeezing
Hamiltonian
r iφ 2 2
Hs = i (e a − e−iφ a † ).
2τ
and squeezing operator

Hs r iφ a 2 −e−iφ a † 2 )
S(r ) = e−i  τ
= e 2 (e .

for the unitary evolution.

Though its action on an arbitrary state is very tough to calculate, we can deduce
some properties by inspecting the action on the coherent state. Let us work out a
calculation of S † (r )aS(r ) for this purpose. With the use of the Baker–Campbell–
2
Hausdorff formula, S̃ = (r̃a 2 − r̃ ∗ a † ), and r̃ = r eiφ /2, this reads

S † (r )aS(r ) = e− S̃ ae S̃
  1    1    
= a − S̃, a + S̃, S̃, a − S̃, S̃, S̃, a + ···
2! 3!
1 1
= a + 22 |r̃ |2 a + 24 |r̃ |4 a + · · ·
2! 4!
1 1
− 2r̃ ∗ a † − 23r̃ ∗ |r̃ |2 a † − 25r̃ ∗ |r̃ |4 a † − · · ·
3! 5!
∞
 ∗ ∞
|2r̃ |2n r̃ |2r̃ | 2n+1
=a − a†
(2n)! |r̃ | (2n + 1)!
n=0 n=0
† −iφ
= a cosh r − a e sinh r .
6.3 Quantum States of a Harmonic Oscillator 131

Then we are able to calculate the expectation values of q and p with the state
|r , α ≡ S(r ) |α as

q ≡ r , α| q |r , α

= α| S † (r )(a + a † )S(r ) |α
2mω

= (α cosh r − α∗ e−iφ sinh r + α∗ cosh r − αeiφ sinh r )
2mω

= [(α + α∗ ) cosh r − (αeiφ + α∗ e−iφ ) sinh r ]
2mω
and similarly

 p ≡ r , α| p |r , α
mω
= [(α − α∗ ) cosh r + (αeiφ − α∗ e−iφ ) sinh r ].
2

Furthermore, we want to evaluate q 2 ≡ r , α| q 2 |r , α = α| S † (r )q 2 S(r ) |α and

p 2 ≡ r , α| p 2 |r , α = α| S † (r ) p 2 S(r ) |α in order to assess the standard devi-
ations. S † (r )q 2 S(r ) consists of terms like S † (r )aaS(r ) and the maths get rather
straightforward by inserting S(r )S † (r ) = 1 in the middle. For example,
S † (r )aaS(r ) = S † (r )aS(r )S † (r )aS(r ) = (a cosh r − a † e−iφ sinh r )2 . Using such
relations, one finally obtains

Δq = q 2 − q2

= (cosh 2r − cos φ sinh 2r ),
2mω

Δp = p 2 −  p2
mω
= (cosh 2r + cos φ sinh 2r ).
2

Note that ΔqΔ p = (/2) cosh2 2r − cos2 φ sinh2 2r , which takes √ the minimal
value /2 when φ = 0 or r = 0. Looking at Δq for φ = 0, Δq = /2mωe−r . This
value is indeed −r
√ smaller rthan that of coherent state by a factor of e . On the other
hand, Δ p = mω/2e , which is larger than that of the coherent state by a factor
of er , hence the state |r , α is prolate in the phase space, one quadrature even being
smaller than the vacuum fluctuation. In particular, the state |r , 0 = S(r ) |0 ≡ |r 
looks squeezed in q direction, thence is called the squeezed vacuum (Fig. 6.2).
For more characterization, let us express the squeezed vacuum |r  by the Fock
states. For preparation,

S(r )aS † (r ) = a cosh r + a † e−iφ sinh r .

132 6 Harmonic Oscillator

Fig. 6.2 Schematic

illustration of the squeezed
vacuum state in the phase
space

as has been done before. This operates on the squeezed vacuum to yield zero, namely
S(r )aS † (r )S(r ) |0 = S(r )a |0 = 0. Therefore, the squeezed vacuum is an eigen-
state of the operator a cosh r + a † e−iφ sinh r ≡ μa + νa † . Expanding the squeezed
vacuum as |r  = ∞ n=0 C n |n and operating μa + νa on it, we obtain recurrence
†

relations
√ √
μ n + 2Cn+2 + ν nCn = 0.

Then the coefficients

√ to be determined are C0 and C1 , however, (μa + νa † )C1 |1 =
C1 μ |0 + 2νC2 |2 implies that C1 = 0, so that C2n+1 = 0. Thus, the expanded
squeezed vacuum is dictated to be
∞ √

1 n −inφ n (2n)!
|r  = √ (−1) e (tanh r ) |2n
cosh r 2n n!
n=0
√
with C0 = 1/ cosh r is determined by the normalization condition.

6.3.4 Thermal State

Thermal state is not a pure state, but a mixed state with probability of the system
being in |n reads

e−nβω
pn = ∞ −nβω
= e−nβω (1 − e−βω )
n=0 e

as statistical mechanics states about the thermal distribution. Here β = 1/k B T is the
inverse temperature and kB and T are the Boltzmann constant and system tempera-
ture, respectively. With these probabilities, the thermal state is well described by the
density operator
∞

ρth = pn |n n| .
n=0
6.4 Photon Correlations 133

 number n =
As is usually done in the statistical mechanics, the average photon
∞ ∞
n=0 pn n| a a |n = n=0 npn and the standard deviation Δn= n − n are
† 2 2

easily calculated using the relation ∞ n=0 ne

−nβω = ∞
n=0 (−1/ω)(∂/∂β)e
−nβω =
(−1/ω)(∂/∂β)(1 − e −βω ). The results are

e−βω
n = ,
1 − e−βω

Δn = n2 + n.

These results show that for the thermal state, the fluctuation of the photon number
is almost equal to the average photon number.

6.4 Photon Correlations

What is classical and what is quantum? How coherent a state is? Two correlation
functions frequently used in the quantum optics give, to some extent, measures of
the “quantumness” and coherence.

6.4.1 Amplitude Correlation/First-Order Coherence

Quantum analogue of amplitude autocorrelation is called the first-order coherence

and is defined by

a † (t)a(t + τ )
g (1) (τ ) =  .
a † (t)a(t) a † (t + τ )a(t + τ )

For the coherent state, since a(t) = ae−iωt in the Heisenberg picture,

a † a e−iωτ
g (1) (τ ) = = e−iωτ .
a†a

Let us consider another example. Suppose we have two modes a1 and a2 and see
amplitude cross-correlation
 
a1† (t)a2 (t + τ )
(1)
g12 (τ ) =   .
† †
a1 (t)a1 (t) a2 (t + τ )a2 (t + τ )

(1)
For such a situation with both of the two modes being coherent states, g12 (τ ) = e−iωτ
as well. However, for the Fock state |n 1 , n 2 , straightforward calculation shows that
134 6 Harmonic Oscillator

(1)
g12 (τ ) = 0. Recall that the amplitude correlation signals the interference effect, that
is, it tells us that the state has maximal ability of interference when |g (1) | = 1 and the
systems totally lacks interference when g (1) = 0. Thus, the former and the latter cases
are said to be coherent and incoherent, respectively. In other cases, 0 < g (1) < 1, the
state is said to be partially coherent.
Physically, the first-order coherence is nothing but a result of interference of
waves. Therefore, the first-order coherence can be observed if one prepares the
double-slit experiment or its equivalent alternatives.

6.4.2 Intensity Correlation/Second-Order Coherence

Quantum analogue of intensity autocorrelation is called the second-order coherence

and is defined by

a † (t)a † (t + τ )a(t)a(t + τ )
g (2) (τ ) =  .
a † (t)a(t) a † (t + τ )a(t + τ )

For a monochromatic wave, a † (t)a † (t + τ )a(t)a(t + τ ) = a † a † aa = a † (aa † − 1)a =

n 2  − n. Therefore, the second-order coherence reduces to
 2
(2) n 2 − n Δn 1
g (τ ) = =1+ − .
n2 n n
√ (2)
 the coherent state, Δn = n and g (τ ) = 1. For the thermal state, Δn =
For
(2)
n + n so that g (τ ) = 2. In general, classical intensity correlation function
2

always takes values larger than 1, and inversely there is some classical analogue
of the quantum state under consideration whenever g (2) (τ ) ≥ 1. For the Fock state
|n, there is no fluctuation of the photon number and g (2) (τ ) = 1 − 1/n < 1. This
does not have any classical analogue and is definitely of quantum nature. Thus, the
second-order coherence is a measure of the quantumness of a quantum state under
concern.
Intensity correlation is interpreted as a particle-like statistics of a harmonic oscil-
lator. For a photonic mode, it can be measured by preparing two photon-counting
modules and measuring the coincidence.

6.5 Wigner Function

6.5.1 General Remarks

Density matrix contains the complete information about the quantum state, and now
we would like to express it in a phase space. Let us apply Wigner transformation to
6.5 Wigner Function 135

the density matrix:

    
1 ∞
−i pq q   q
W (q, p) = e q +  ρ q − dq .
2π −∞ 2 2

W (q, p) is known as Wigner function. One might immediately see that the Wigner
function is real but not necessarily positive.
Marginal distributions are obtained by integrating with respect to one variable.
For example,
 ∞
W (q) = W (q, p)d p
∞
 ∞  ∞    
1 q   q
= dp dq e−i pq q +  ρ q −
2π −∞ −∞ 2 2
 ∞    
q   q
= dq δq q +  ρ q −
−∞ 2 2
= q| ρ |q .
∞
Here the identity 2πδ(q) = −∞ d pe−i pq is used. W (q) is obviously non-negative.4
For a pure∞state, W (q) is a squared wavefunction in q-representation. How about
W ( p) = −∞ W (q, p)dq? The calculation is a bit tricky but by properly inserting
∞ √
−∞ | p  p| d p = 1 twice and using q | p = e
i pq/ / 2π, one can see

W (q) =  p| ρ | p .

This, as well, is non-negative and equals a squared wavefunction in p-representation

for a pure state. Thus, the marginal distributions are probability distributions, satis-
fying
 ∞  ∞  ∞  ∞
W (q)dq = W ( p)d p = W (q, p)dqd p = 1
−∞ −∞ −∞ −∞

The Wigner function itself can take negative values, therefore it is not a proba-
bility distribution. Instead, it is said to be a quasi-probability distribution, for it is
normalized. But do not overlook the importance of the Wigner function itself, since
the quantum states accompanied with the classical counterpart yield positive Wigner
function. In other words, the negativity of the Wigner function is an evidence of the
non-classicality.

4 Check for ρ = |ci |2 |ψi  ψi |, |ci |2 = 1.

136 6 Harmonic Oscillator

Fig. 6.3 Wigner functions of coherent state |α (left) and superposition |α + |−α (right)

6.5.2 Examples

We pick up a few examples of Wigner function for pure states. For a pure state |ψ,
the Wigner function reduces to
 ∞    
1 q q
W (q, p) = e−i pq ψ q + ρψ ∗ q − dq .
2π −∞ 2 2

Coherent State
First example is a coherent state. First we should derive the q-representation of the
coherent state by evaluating q |α.

∞
αn |α|2
q |α = x| √ e− 2 |n
n=0
n!
∞
 αn − |α|2 † n
= x| e 2 a |0
n!
n=0
|α|2
= x| e− eαa |0
†
2
  
2 mω p̂
|α| α q̂−i mω
= e− 2 x| e 2
|0

|α|2 α2 mω p̂
α 2 q̂ −iα √
= e− 2 e− 4 x| e e 2mω |0

|α|2 α2 α mω √−iα  d
2 q
= e− 2 e− 4 e x |0 e 2mω i dq
  
|α|2 α2 α mω q 
= e− 2 −
e 4 e 2
x −α |0
2mω
6.5 Wigner Function 137

  2
 
|α|2 α2 α mω − mω
2 q−α 2mω
2 q
= Ae− 2 e− 4 e e
  2
− mω 2
2 q−α mω
= Ae

mω 2
In this calculation, identities eγ(d/dq) f (q) = f (x + γ) and x |0 = Ae− 2 q are
used. In the last line, α is assumed to be real to simplify the situation.
∞
Now, let us evaluate the Wigner function. With the knowledge of −∞ e−a(x+b) dx =
2

√
π/a for real a and complex b, one can straightforwardly derive

√ 2   2
2 A − mω q−α 2
p2
e− mω
 mω
Wα (q, p) = √ e
πmω
√
which is the Gaussian
√ √ with its peak located at (α 2/mω, 0) and its
distribution
widths being /2mω in q- and mω/2 in p-axes. See the left panel of Fig. 6.3.
Schrödinger’s Cat State
Let us next consider the Wigner function of a Schrödinger’s cat state |α + eiθ |−α,
a coherent superposition of two coherent states. The normalization factor is omitted
here. The calculation of the Wigner function is a bit lengthier than that of the simple
coherent state, but it basically reduces to Gaussian integrals. The final expression is
 
2 2
− mω 2 − p
W (q, p) = Wα (q, p) + W−α (q, p) + Ãe  q e
mω cos 2 p α+θ .
mω

The Wigner function of the Schrödinger’s cat state is shown in the right panel of
Fig. 6.3. In addition to the Wigner function of each coherent state Wα or W−α , there
is an “interference fringe” in p direction in the middle which originates in the third
term.
Fock State
The wavefunctions of the Fock states are given as
 
1 mω − mωq 2 mω
|ψn (q)|2 = e  Hn q
2n n! π 
 
1 1 p2 1
|ψn ( p)| = n
2
e− mω Hn p
2 n! πmω mω
138 6 Harmonic Oscillator

Fig. 6.4 Wigner functions of Fock states |n

using Hermite polynomials Hn (x) = (−1)n e x /2 (∂ n /∂x n )e x

2 2 /2
. The Wigner func-
tion which generates these as marginals is the following:
 
2   
(−1)n − 1 p
mωq 2 + mω 2 p2
Wn (q, p) = e Ln mωq +
2
.
π  mω

Here L n (x) = nj=0 n C j (−x) j / j! are Laguerre polynomials.

As it was mentioned in the beginning of this section, a quantum state with Wigner
function taking negative values does not have corresponding classical analogue. In
this perspective, the Schrödinger’s cat state and the Fock state (Fig. 6.4) both take
negative values and thus are genuinely quantum mechanical states. On the other
hand, the coherent state and squeezed state does not have intrinsic negativity in the
Wigner function and there are corresponding classical phenomena.

Problems

Problem 6-1 (i) Derive[q, p] = i using p = (/i)d/dq.

(ii) Show a, a † = 1.
Problem 6-2 Show that the creation operator a † does not have any nontrivial eigen-
state, that is, the eigenstate other than the null vector.
6.5 Wigner Function 139

Problem 6-3 We define superoperators (S∗), (∗S) and [S, ] for an operator S such
that for any operator A they act as

(S∗)A = S A, (∗S)A = AS, [S, ]A = [S, A]. (6.5.1)

(i) Check that [S, ] = (S∗) − (∗S) and (S∗)(∗S) = (∗S)(S∗) hold.
(ii) Prove the Baker–Campbell–Hausdorff formula

1 1
e−S Ae S = A − [S, A] + [S, [S, A]] − [S, [S, [S, A]]] + · · · .
2! 3!
(6.5.2)

Problem 6-4 Check that the coherent state has standard deviation |α| of its number.
Problem 6-5 Prove that a marginal distribution W (q) of a Wigner function is non-
negative, here for a state in q-representation ρ= |ci |2 |ψi (q)ψi (q)|,
|ci |2 = 1.
Problem 6-6 Let us consider a Wigner function of a thermal state with average
number of quanta n th .
(i) Discuss qualitatively what it should look like.
(ii) Quantitatively confirm your physical intuition by explicitly cal-
culating the Wigner function of the thermal state starting from the
density-matrix representation of it. The following identities might be
useful:

nn
e−nβω (1 − e−βω ) = (6.5.3)
(1 + n)n+1
−cx
 e 1−c
cn L n (x) = . (6.5.4)
n
1−c

Then discuss the similarities and differences with other quantum

states such as the coherent state.
Problem 6-7 Calculate the normalization constants Aeven and Aodd of the Schrödin
ger’s cat states.
Two-level System and Interaction with
Electromagnetic Waves 7

7.1 Two-level System, Spin, and Bloch Sphere

First of all, what we are interested in is a quantum bit, or a qubit, which is to be

controlled in the quantum engineering activities. A qubit consists of two orthogonal
quantum states |g and |e, that is, g|g = e|e = 1 and g|e = 0. This set of states
 
0
|g = ,
1
 
1
|e = ,
0

or the two-level system is usually constructed by picking-up two orthogonal quantum

states among many others residing in an individual quantum system regarding the
ability of manipulation, excited-state lifetime and coherence, and any other merits
regarding the quantum control of the qubit. Note that labels g and e are often inter-
changed depending on the convention used. Quantum information can be registered
in the complex vector space called Hilbert space, spanned by these two orthonormal
bases set {|g, |e}, or more generally by an assembly of them. Let us, then, write
down the Hamiltonian of the qubit which is the sum of the energies, ωg for the
ground state and ωe for the excited state (Fig. 7.1), associated with the projectors
|gg| and |ee|:

H = ωg |gg| + ωe |ee|. (7.1.1)

Supplementary Information The online version contains supplementary material available at

https://doi.org/10.1007/978-981-19-4641-7_7.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 141
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7_7
142 7 Two-level System and Interaction with Electromagnetic Waves

One can easily verify that if the Hamiltonian is evaluated (sandwiched) by |g, the
value is the ground-state energy, and so for the excited state as well. Here we want
to try rewriting this Hamiltonian by the Pauli operators defined by
 
01
σx = = |eg| + |ge|, (7.1.2)
10
 
0 −i
σy = = −i|eg| + i|ge|, (7.1.3)
i 0
 
1 0
σz = = −|gg| + |ee| (7.1.4)
0 −1

to get an alternative expression of the Hamiltonian as

(ωe − ωg ) (ωe + ωg )
H= σz + I
2 2
ωq
= σz + const. (7.1.5)
2
where the second term in the first line is neglected since it, being proportional to the
identity matrix I , only accounts for the energy shift imposed on the whole system
that only redefines the reference of the energy. Here ωq = ωe − ωg is the transition
frequency, which specifies the frequency of the electromagnetic wave to be applied
in order to flip the qubit, as detailed in later section.
Let us pose here gazing at Pauli operators given in Eqs. (7.1.2)–(7.1.4). First of all,
they are all Hermitian: σξ† = σξ (ξ = x, y, z). This means that the eigenvalues of the
Pauli operators can be interpreted as being some physically observable quantities.
Next point is less obvious but essential. With the commutator [A, B] = AB − B A,
they have relations

[σx , σ y ] = 2iσz , [σ y , σz ] = 2iσx , [σz , σx ] = 2iσ y . (7.1.6)

Fig. 7.1 Energy levels of a

quantum bit
7.2 Interaction Between Two-level System and Electromagnetic Field 143

Fig. 7.2 Illustration of a Bloch sphere

These relations are summarized into following form:

[σi , σ j ] = 2ii jk σk . (7.1.7)
Pauli operators are indeed the generators of the special unitary group SU(2), which
is known to be a good group-theoretical description of a spin-1/2 system. In other
words, a qubit can be identified as a spin-1/2 system! We are glad to see this cor-
respondence, since the expectation value of the Pauli operator σξ is now read as
the spin component in ξ axis and we are allowed to draw a quantum state of the
qubit within a Bloch sphere (see Fig. 7.2), likewise for a single √spin-1/2 system.
√ In
the Bloch sphere,√three sets of orthogonal
√ bases {(|g + |e)/ 2, (|g − |e)/ 2},
{−i(|g + i|e)/ 2, −i(|g + i|e)/ 2} and {|e, |g} are taken, respectively, to be
x, y, and z axes.1 If a qubit is in a pure state |ψ, it can be rewritten as |ψ =
cos (θ/2)|e + eiφ sin (θ/2)|g with parameters θ and φ, or equivalently in the Bloch
sphere as (sin θ cos φ, sin θ sin φ, cos θ ) = (ψ|σx |ψ, ψ|σ y |ψ, ψ|σz |ψ).

7.2 Interaction Between Two-level System and

Electromagnetic Field
The quantum states constructing a qubit are ideally orthogonal to each other, as
mentioned in the previous section. Some mechanism is needed to “mix” these two

1 It is straightforward to check that these bases are eigenvectors of σx , σ y and σz , respectively.

144 7 Two-level System and Interaction with Electromagnetic Waves

states to transform
√ a naturally prepared state, say |g, into a desired state like |e,
(|g + |e)/ 2, and so on. In many cases, |g is the ground state and |e the excited
state of some anharmonic oscillator, and sometimes they are spin down and spin
up. However, one thing seems to be very common: the two states are coupled via
transition dipole moment which is driven by an electromagnetic field resonant on
the transition.2 ,3 The transition dipole moment d is intuitively the classical dipole
moment qr with charge q and some maths will give the correctly quantized expression
of it. However, we dictate the final expression as given, since it is also intuitive. The
dipole operator d is given as [9]

d = μ (|eg| + |ge|) (7.2.1)

= μ(σ+ + σ− ). (7.2.2)

In this expression, the following two operators are newly defined:

 
0 1 σx + iσ y
σ+ = |eg| = = (7.2.3)
0 0 2
 
0 0 σx − iσ y
σ− = |ge| = = . (7.2.4)
1 0 2

These operators act as raising and lowering operators, so that we can control qubit by
“stimulating” the dipole moment that includes these operators by an external field.
Suppose that the electromagnetic field with its polarization parallel to the transi-
tion dipole moment is applied to the two-level system. Then the interaction terms are
generated in the Hamiltonian that is generally described by a polynomial function
of the applied field, say, an electromagnetic field E. To first-order approximation the
interaction term reads d · E and let us neglect other higher-order terms.4 Plugging
the operator expressions of d = μ(σ+ + σ− ) and E = E zpf (a + a † ), we have an
interaction Hamiltonian

Hint = μE zpf (σ+ + σ− )(a + a † ). (7.2.5)

Furthermore, we assume that the frequency of the electric field ωc and the transition
frequency ωq are close together. Products such as σ+ a † are “rapidly oscillating” at

2 There are cases with quadrupole moment taking place, however, the resultant interaction Hamil-
tonian will be of the same form so that we focus on the dipole transitions.
3 There are cases where the auxiliary level |r  mediates the transition between |g and |e. In such

a case the system can be treated as a two-level system as well. After reading through this Chapter,
readers interested in are recommended to refer to Appendix C.
4 This prescription is called the “electric dipole approximation” in the case of electric dipole moment

under concern.
7.2 Interaction Between Two-level System and Electromagnetic Field 145

ωq + ωc and we neglect here. This procedure is called the rotating-wave approxi-

mation. Then the final expression is

Hint = μE zpf (σ+ a + σ− a † )

= g(σ+ a + σ− a † ) (7.2.6)

which is the celebrated Jaynes–Cummings interaction [10] with coupling strength

g = μE zpf /. The physical situation is really simple. The first term represents the
process that a photon is absorbed (a) by the qubit and the qubit is excited (σ+ ) and
the reverse in the second.
Sometimes the anti-Jaynes–Cummings interaction appears depending on the con-
text.5 It is of the form

Hint = g  (σ+ a † + σ− a). (7.2.7)

Keep in mind that whether the Jaynes–Cummings or anti-Jaynes–Cummings inter-

action, or both, survives depends on what system and what “driving” electromag-
netic waves you take into account. By driving the nonlinear system with ωq + ωc
strongly, the nonlinear process may allow simultaneous creation of a photon in the
cavity and the qubit excitation and vice versa. For example, both of them can be
realized by applying the two-tone driving to the trapped atomic ions to implement
Mølmer-Sørensen gate [11], which enables the highest-ever-achieved-fidelity two-
qubit gate [12].

7.2.1 Spontaneous and Stimulated Emission

Here we introduce two kinds of processes, spontaneous emission and stimulated

emission/absorption, as the ones involving electromagnetic waves in close relation to
the two-level system. Spontaneous emission is a process that the excited state decays
into the ground state by emitting one photon. This happens due to the coupling of
the dipole moment of the two-level system to the vacuum field surrounding it. This
can be interpreted as the stimulated emission caused by the vacuum fluctuation, as
introduced in the next paragraph.
Intuitive picture of this process in the classical physics point of view is that with an
atom as an example, the orbiting electron around the nucleus, for instance, generates
the oscillating electromagnetic field and loses its energy. Of course, this is incorrect
interpretation of the dynamics of electron in an atom and the quantum mechanics

5 Here we make a note on the terminology. The (anti-)Jaynes–Cummings interaction usually refers

to those between a qubit and a harmonic oscillator. Similar interactions involving two harmonic
oscillators, whose annihilation operators are a and b, are represented by g(a † b + ab† ) and
g(a † b† + ab), and called beam-splitter and two-mode squeezing interaction, respectively. For
qubit-qubit interactions such as the ones proportional to σx ⊗ σx and σx ⊗ σ y , names like X X and
X Y interactions are frequently used.
146 7 Two-level System and Interaction with Electromagnetic Waves

should take place. In the spontaneous emission process, a single photon is emitted
and the phase of the emitted photon is completely at random. This results in the
dephasing, as we will refer to in the later Section.
The spontaneous emission or spontaneous decay is responsible for the de-
excitation of the qubit with its transition frequency in the optical regime. However,
the de-excitation process itself is actually not limited to the spontaneous decay. For
example, the energy transfer between the qubit and some particle in the environment
results in irreversible loss of excitation energy from the qubit system. The general
name of the de-excitation processes is longitudinal decay or longitudinal relaxation
which is complementary to the dephasing processes or transverse decay/relaxation.
Another important process is stimulated emission and absorption which occur
when a coherent electromagnetic wave is applied to the two-level system. The pro-
cesses under concern are the absorption and emission of photons by the two-level
system, and these two processes are reciprocal to each other and occur at the same
rate proportional to the square root of the intensity of the incident electromagnetic
wave. Therefore, however strongly we apply continuous electromagnetic wave, the
two-level system does not end in a situation that the occupation probability of the
excited state gets unity. The occupation probabilities of the ground and excited states
should be 1/2, or less if the driving field is not sufficiently intense. As it is impor-
tant in the Rabi oscillation and Ramsey interference introduced in later Chapter, the
stimulated emission and absorption are coherent processes in which the phases of
the electromagnetic waves and the qubit state have one-to-one correspondence.

7.3 Two-level Systems and Electromagnetic Waves in Practice

7.3.1 Two-level Systems in Nuclear Magnetic Resonance

Nuclear magnetic resonance (NMR) is frequently utilized in a spectroscopic method

in material sciences, but also in a very familiar method of medical application known
as magnetic resonance imaging (MRI). In NMR, a two-level system is constructed
by two states |↑ and |↓ of a nuclear spin of an atom in a properly chosen molecule.
Energies of these two nuclear-spin states are split by Zeeman effect under applied
magnetic field. Molecules containing nuclear spins usually take a form of aqueous
or other solutions, and nuclear spins of phosphor and hydrogen atoms are frequently
shed light (though radio frequency, as we mention later on). Magnetic field of the
order of 1 T is applied to the solution containing nuclear spins, with which the spin
states are Zeeman-split. Then radio-frequency waves of tens or hundreds of MHz,
equivalent to the Zeeman splitting mentioned just above, are applied to implement
the interaction between a two-level system and electromagnetic waves in the form
of magnetic dipole interaction. Coherence time of nuclear-spin state depends on
molecules and solvents used. In some combinations, the coherence time becomes
too long to be measured, verifying that the nuclear spins are quite nicely isolated
quantum systems even in condensed matter systems.
7.3 Two-level Systems and Electromagnetic Waves in Practice 147

7.3.2 Two-level Systems in Atomic Gases

In systems utilizing neutral and singly ionized atomic gases, their rich internal
structures allow us to construct various two-level systems. Here we focus on
two-level systems operated in the optical regime and in microwave regime, that
appear frequently in the trapped ion quantum systems. Appendix D might help read-
ers unfamiliar with atomic levels such as hyperfine structures and more details will
be found in e.g. Ref. [9].
What we want to introduce first is an optically operated two-level system. For such
a purpose, we summarize the frequently-used quantum states and optical transitions
of a singly-charged calcium ion in Fig. 7.3. A calcium ion, an alkaline-halide atom
with one valence electron removed, possesses one valence electron and so that it has
S state as its electronic ground state. It also has P state as an excited state with a
lifetime of several nanoseconds, and the optical transition from S state to P state
(S-P transition) is dipole-allowed. This ion has D state which is metastable, since
the S-D transition is dipole forbidden. The S-D transition, which is only allowed by
an electric quadrupole transition, has very narrow linewidth of around 1 Hz and thus
D state has very long lifetime of around 1 s. This is one of the benefits of building a
qubit with S state as |g and D state as |e.
For the purpose of introducing a two-level system operated in microwave regime,
we consider an singly-ionized ytterbium atom, especially a specific isotope of
171 Yb+ . Its electronic ground state is S state and its nuclear spin is 1/2. There-

fore, hyperfine structure has two manifolds F = 0 and F = 1, where F denotes the
total angular momentum including electronic spin, electronic orbital, and nuclear
spin degrees of freedom. The hyperfine splitting is about 12.6 GHz in frequency.
Though the F = 0 state has only one spin state with m F = 0 where m F stands for
the component of the angular momentum along the quantization axis, F = 1 man-
ifold consists of three states m F = −1, 0, +1. Since the energies of two m F = 0
states are insensitive to the magnetic field for the absence of the angular momen-
tum component along the quantization axis, |F, m F  = |0, 0 and |F, m F  = |1, 0

Fig. 7.3 Partial energy diagram of calcium ion and ytterbium ion
148 7 Two-level System and Interaction with Electromagnetic Waves

are frequently used as a bases constructing a qubit. By irradiating microwaves of

12.6 GHz frequency or by using optical Raman transition, the qubit state can be
controlled. Since both the |0, 0 and |1, 0 states are electronic ground state, the
lifetime of the qubit excited state is nearly infinite. Therefore, the coherence time of
such a hyperfine qubit is limited by dephasing, mainly due to the fluctuation of the
transition frequency accompanied with the fluctuating magnetic field bias.
Atomic gas or atomic ions as qubits provides us with uniform, long-lived two-level
systems by nature. Cons are the complex experimental systems using frequency-
stabilized lasers and optics for their adjustment, and ultrahigh vacuum setup down
to 10−9 Pa or less. The atomic quantum technologies are fronting these difficulties
of the stability and the scalability.

7.3.3 Two-level Systems in Solid-state Quantum Defects

Atomic defects or impurities in the non-metallic solid materials are called quantum
defects. The optical transition of such a defect center has energy levels within the
bandgap and can be coherently controlled, since it is relatively isolated from the
environment, or bulk phonons and electrons supported by the substrate materials.
However, the isolation is not complete: the optical transitions are in many cases
broadened by phonons, and the spin states are disturbed by surrounding electronic
and/or nuclear spins. The site-controlled preparation of the quantum defects is also
a technical challenge.

7.3.4 Two-level System in an Optically Controlled Semiconductor

Quantum Dot

Semiconductor crystal growth technology has enabled us crystal growth at single-

atom-layer level. This technology has been applied to realize nanoscale material
region with narrower bandgap, indium arsenide for instance, than that of the sub-
strate such as gallium arsenide. If such a nanoscale region confines the electron in
3 dimensions, such a nanostructure is called a quantum dot. Only a single electron
or two can be trapped inside the quantum dot, the optical transition and quantum
manipulation of electronic spin state have gathered attention.
Figure 7.4 depicts the band structure of the quantum dot. An electron in a quan-
tum dot is localized in a region of around 10 nm, so that only a single electron is
trapped in most cases. The excited state of the quantum dot electron just below the
conduction-band edge exhibits the lifetime of a few hundreds of picoseconds to a
few nanoseconds.
An electron confined in the quantum dot energetically close to the valence band
has a spin degree of freedom, so that it can also be used as a qubit. The coherence
time of such a spin qubit is about a few microseconds; therefore, the spin qubit can
be controlled by using laser technique such as stimulated Raman transition.
7.4 Dynamics and Relaxations of Two-level Systems 149

Fig. 7.4 Exciton in a quantum dot (left) and an electron confined in a quantum dot (right). Broken
horizontal line represents the Fermi energy

Although semiconductor quantum dot fits well with the semiconductor-based

nanophotonics, the short coherence time and the dot-to-dot fluctuation of its posi-
tion, lifetime, and emission wavelength hinder the straightforward scaling up of the
quantum dot quantum technology.

7.3.5 Two-level System in Superconducting Circuit

As detailed later in Sect. 9.3, a ultra-low loss, nonlinear LC resonator can be con-
structed by the parallel connection of a capacitor and a Josephson junction using
superconducting wiring. In such a superconducting circuit, anharmonicity of the
resonator results in such situation that the energy difference between the ground
state and the first excited state and that of the first and the second excited states are
quite large to allow us for utilizing the ground and the first excited state as a two-level
system, which is often referred to as a superconducting qubit. The transition energy
of this qubit depends on the circuit design, however, generally it is within a few to
10 GHz. An advantage of the superconducting qubit is that by properly designing a
circuit, the interaction between microwave field and the qubit can be very large that
the coupled qubit-cavity system (circuit QED system) is often in a strong dispersive
regime.

7.4 Dynamics and Relaxations of Two-level Systems

7.4.1 Bloch Equations and Relaxations

In the previous section, we saw that a qubit can be identified as a spin. Primary
dynamics that the spin undergoes is, in the context of quantum engineering, due to the
interaction with applied electromagnetic fields. Or we consider relaxation of it when
it is subject to the interaction with some environment, such as stray electromagnetic
150 7 Two-level System and Interaction with Electromagnetic Waves

fields for highly isolated systems and thermal phonons or surrounding spin degree of
freedom for solid-state systems. In 1946, Felix Bloch constructed a set of equations
describing the spin, or more concretely the magnetization, precession about the
applied magnetic field. This theory was developed principally for the “real” spin
systems such as nuclear magnetic resonance, electron spin resonance for instance,
however, in 1957 Richard P. Feynman realized that the Bloch equation is also a
good description of the dynamics of a fictitious spin-1/2 system, or a qubit. Here
we introduce a density matrix first, rewriting it as a spin-1/2 object, and analyze its
behavior using a fundamental equation named master equation. Then, as a concrete
case, the Bloch equations are derived.
Let us first define a density matrix ρ for a quantum state |ψ = cg |g + ce |e as

ρ = |ψψ| (7.4.1)
= ρgg |gg| + ρge |ge| + ρeg |eg| + ρee |ee| (7.4.2)
 2 ∗ 
|ce | cg ce
= . (7.4.3)
cg ce∗ |cg |2

A density matrix can treat the probabilistic realization of some states i pi |ψi ψi |,
called the mixed state, in contrast to the pure state. How about the dynamics of the
density matrix? To address this, we depart from the state vector in the Schrödinger
picture, where the time evolution of the state reads |ψ(t) = e−i(H/)t |ψ = Ut |ψ
with the system Hamiltonian H. Let us differentiate the density matrix evolving with
ρ(t) = Ut ρUt† :

dρ(t) dUt dU † i i 1
= ρUt† + Ut ρ t = − Hρ(t) + ρ(t)H = [H, ρ(t)] (7.4.4)
dt dt dt   i
which leads us to the von Neumann equation

dρ
i = [H, ρ] . (7.4.5)
dt
Note the difference between the von Neumann equation for a state and the Heisenberg
equation of motion for an operator A: id A/dt = [A, H].
Let us now consider a driven qubit, at a first step without the relaxations. Our
Hamiltonian for the driven qubit is dictated as6

ωq 
H= σz + (σ+ e−iωd t + σ− eiωd t ). (7.4.6)
2 2
Here the parameter = μE d is introduced with the electric field strength E d , which
is in close connection with the coupling strength g of the Jaynes–Cummings interac-
tion. Moving onto the rotating frame (see Appendix B) by the unitary transformation

6 Please refer to the Appendix C for the derivation of the driving term.
7.4 Dynamics and Relaxations of Two-level Systems 151

U HU † − iU U̇ † with U = exp [i(ωd /2)σz t], we get a modified one:

 
H =
q
σz + (σ+ + σ− ), (7.4.7)
2 2
where the detuning is defined by q = ωq − ωd . Substituting this into the von Neu-
mann equation (7.4.5) and evaluating it by sandwiching with |g’s and |e’s, one can
obtain a set of equations

dρee
=i (ρeg − ρge ), (7.4.8)
dt 2
dρgg
= −i (ρeg − ρge ), (7.4.9)
dt 2
dρeg
= −i q ρeg −i (ρgg − ρee ), (7.4.10)
dt 2
dρge
= +i q ρge +i (ρgg − ρee ), (7.4.11)
dt 2
where ρi j = i|ρ| j. Let us define some quantities that look like the components of
the pseudo-spin:

sx = ρge + ρeg , s y = −i(ρge − ρeg ), sz = ρee − ρgg .

Note that there is an intrinsic relation of the density matrix of a pure state ρgg + ρee =
1. The reason why we can regard this as components of the pseudo-spin can be seen
by noting that

ρ = ρgg |gg| + ρge |ge| + ρeg |eg| + ρee |ee| (7.4.12)

1
= (I + sx σx + s y σ y + sz σz ). (7.4.13)
2
Furthermore, these quantities allow us to rewrite above differential equations in the
following neat form, thanks to the transformation onto the rotating frame:

dsx
= − q sy , (7.4.14)
dt
ds y
= + q s x − sz , (7.4.15)
dt
dsz
= sy (7.4.16)
⎛ dt⎞ ⎛ ⎞ ⎛ ⎞
s s
d ⎝ x⎠ ⎝ ⎠ ⎝ x⎠ ds
⇐⇒ sy = 0 × sy ⇐⇒ = B × s. (7.4.17)
dt s sz dt
z q

This is the Bloch equation without relaxations that Feynman found valid for the
two-level system, and it resembles the Bloch equation for real spin systems. The
152 7 Two-level System and Interaction with Electromagnetic Waves

two-level system and the vector B = ( , 0, q ) act like a spin and applied static
magnetic field, just as the Larmor precession.
Master equation (see Appendix E) goes a step further by incorporating
a loss characterized by a dissipator and Lindblad superoperator
L[ ]ρ = 2 ρ † − † ρ − ρ † added in the von Neumann equation to describe
the lossy dynamics of the system. Suppose the dissipator is with its spontaneous
and thermal-bath-induced decay rates and  , respectively, the master equation
has the general form of

dρ i 1 √ 
1 √ 
= − [H, ρ] + L[ + ]ρ + L[ †
]ρ. (7.4.18)
dt  2 2

The effect of thermal bath,  , is mostly negligible in the quantum engineering

experiments, since the whole environment is usually cooled down to sufficiently
low temperature. Therefore, we shall ignore this from now on. We now consider
two sources of relaxations. One is the spontaneous decay or longitudinal relaxation,
which originates in the finite√lifetime of the excited state |e and is represented by
a Lindblad superoperator L[ 1 σ− ]. The other is the dephasing or the transverse
relaxation, whose origins are diverse ranging from fast fluctuations of an environment
to slow variation√of applied electromagnetic fields. This is taken into account by a
superoperator L[ 2 2 σz /2]. Putting these together, we shall investigate the specific
but ubiquitous master equation

dρ i 1 1
= − [H, ρ] + L[ 1 σ− ]ρ + L[ 2 σz ]ρ. (7.4.19)
dt  2 2
As has been done when the Bloch equation was derived, both sides of the equation
are evaluated and reinterpreted in terms of sx , s y , and sz to become Bloch equations
with relaxations:
dsx 1+2 2
= − q sy − sx (7.4.20)
dt 2
ds y 1+2 2
= + q s x − sz − sy (7.4.21)
dt 2
dsz
= s y − 1 (sz + 1) (7.4.22)
dt
⎛ 1 +2 2 ⎞
sx
ds 2
⇐⇒ = B × s − ⎝ 1 +2
2
2
sy ⎠ . (7.4.23)
dt
1 (sz + 1)

As can be seen in above expressions, 1 is principally responsible for the decay

from |e to |g. 2 does not account for the population decay but the Bloch sphere
“shrinks” in sx and s y directions as a result of the loss of phase coherence of a qubit
(see Fig. 7.5).
7.4 Dynamics and Relaxations of Two-level Systems 153

Fig. 7.5 Evolution of the Bloch sphere with spontaneous emission (left) and dephasing (right)

7.4.2 Rabi Oscillation

Given a set of equations that describes the spin dynamics in the presence of driving
field and losses, we are to analyze it in practice. However, the Bloch equation cannot
be solved analytically except for some special situations. We here solve this for
idealized case of no detuning ( q = 0) and no relaxations ( 1 = 2 = 0). Then the
Bloch equations are simplfied as
dsx ds y dsz
= 0, = − sz , = sy
dt dt dt
which yield sx = const. and sz = − cos ( t), s y = sin ( t) with an initial condition
sz (t = 0) = −1, meaning that the qubit is in |g at t = 0. Intuitively in the Bloch
sphere, the qubit state, first in |g, is rotated about the vector ( , 0, 0), the “x” axis,
see left panel of Fig. 7.6. One can easily recognize that determines how rapidly the
qubit is rotated in the Bloch sphere, justifying the notion of the Rabi frequency. Recall
that the Rabi frequency is related to the applied field strength by = μE d /2,
the product of the transition dipole moment and the applied electric field strength.
Therefore, how much the qubit is rotated with an applied pulse, being assumed to be
square-shaped, is under control of the pulse duration Δt and electric field strength
E d . In general, the phase of rotation is determined by the pulse area, the integrated
pulse envelope over relevant time.
We give some remarks on a bit more general case, firstly for a finite detuning
q = 0. As mentioned previously, the Bloch vector, the qubit state represented in the
Bloch sphere, is rotated about the vector ( , 0, − q ), which is inclined by an angle
θ = − arctan ( q / ) with respect to the x axis, see right panel of Fig. 7.6. Therefore,
the population dynamics can be easily deduced from the above consideration: the
154 7 Two-level System and Interaction with Electromagnetic Waves

Fig. 7.6 Rabi oscillation without (left) and with (right) the detuning

Fig. 7.7 Population dynamics in |e during the Rabi oscillation. Left panel: Rabi oscillation for var-
ious detunings δ without spontaneous decay. Right panel: Rabi oscillation for various spontaneous
decay rates γ without detuning. Simulation is done with the help of QuTiP [13]

Bloch vector of the qubit is rotated about ( , 0, q ) and the population, initially in
|g, is not completely transferred to |e unlike the case of q = 0, and the angular
frequency of the rotation is given by the length of the vector ( , 0, q ), namely
2+ q2 which is called the generalized Rabi frequency. The situation is depicted
in the right panel of Fig. 7.6. In the case of finite losses, namely 1 and 2 are not
equal to zero, the Rabi oscillation gradually loses its amplitude and final population
sz (t = ∞) is determined by the balance between the driving intensity and the
decay 1 . Such a situation is illustrated in the left panel of Fig. 7.7. Numerical
evaluation of the system dynamics is useful in such general cases.7

7 For example, a package called Quantum Toolbox In Python (QuTiP) [13] is a good choice if you
are familiar with python (or even if not!).
7.4 Dynamics and Relaxations of Two-level Systems 155

Fig. 7.8 Schematics of the Ramsey interference

7.4.3 Ramsey Interference

We would like to introduce another important quantum-mechanical phenomena8

called Ramsey interference, named after Norman Ramsey. In the Ramsey interfer-
ence, the qubit is transformed into a superposition of ground and excited states and
the interference phenomenon can be observed between these states experiencing
different phase shifts. First we apply a resonant electromagnetic pulse of duration
Δt = π/2 to the qubit so that it acquires the phase of π/2 in the Rabi oscilla-
tion. This pulse is called the π/2-pulse.
√ The Bloch vector, initially in |g, is then
transferred to | − i = (|g − i|e)/ 2 (left panel of Fig. 7.8). Then for some time
duration the qubit is subject to the “free” evolution, that is, we do nothing. At the final
stage, the π/2-pulse is again applied to the qubit that maps the phase information in
x y plane onto zx plane and the projective measurement along z axis is executed. If
the qubit is not subject to any fluctuating environment and/or stray field, the Bloch
vector stays where it is just after the first π/2-pulse and is transferred to |e by the
second π/2-pulse, which gives sz = 1.
If, e.g., the slow variation of the applied field, the phase drift occurs during the
time we leave the qubit between the two π/2-pulses, the phase rotation in the x y
plane is transformed to the one in the zx plane and the final measurement result
of sz oscillates (co-)sinusoidally between 1 and −1 (see middle and right panels of
Fig. 7.9) according to the acquired phase, showing an interference fringe. Or if the
pure dephasing instead of the phase drift is experienced by the qubit, random phase
“diffusion” in x y plane makes sx and s y decrease and finally results in sx = s y = 0
with sufficiently long duration between the two π/2-pulses.

7.4.4 Spin Echo

Even if there are some sources of the phase drifts mentioned above during the
time evolution of the quantum systems, their effects can be canceled by a spin-echo
method, though under a few conditions. The essential idea is simple. The phase drift

8 Originally the Ramsey interference was invented for spectroscopic purpose.

156 7 Two-level System and Interaction with Electromagnetic Waves

Fig. 7.9 Schematics of the spin-echo sequence

of the qubit is intuitively recognized as a rotation around z axis in the Bloch sphere.
Then, if typical timescales of the phase drift are sufficiently long, or in other words
the phase drift is constant over the time duration under concern, we can effectively
“rewind” them by applying a π pulse around x or y axis.
The spin-echo method is particularly useful when, for example, the qubit is subject
to the slowly varying noise during the quantum operations such as π rotation around
x axis. Since the phase drift adds small z rotation during the x-rotation sequence, it
gives rise to an operational error. To avoid this error, spin-echo method tells us that
we can cancel the phase drift by dividing the π pulse into two π/2 pulses temporarily
separated by some duration τ and insert a π pulse around y axis in between, delayed
from the first π/2 pulse by τ/2. To see what is going on in this sequence, see Fig. 7.9.
Let us suppose the qubit is initialized in |g. The first π/2 pulse let the quantum state
in | − i and in the succeeding τ/2 free evolution the Bloch vector is rotated around
z axis by some angle φ. The π pulse in between the two π/2 pulses flips the Bloch
vector around y axis. Here, if the source of the phase drift is constant over the whole
sequence, the Bloch vector again rotates about z axis by exactly the same angle φ
to get back to | − i state. Eventually, the last π/2 pulse let | − i evolve into |e.
In the above discussion, of course, a fact that the phase drift occurs during every
pulse sequence is neglected. Even so, the fine tuning of the pulse durations could
help achieve phase-drift-free qubit operation in actual experiments.
The spin-echo method is originally invented for the spectroscopy of an ensemble
of spins in the presence of the inhomogeneous environment, where the inhomo-
geneous phase evolutions of the spins here and there are all “rewinded” to have
7.4 Dynamics and Relaxations of Two-level Systems 157

echo-like revival of the signals of the spin ensemble. For a single spin, or a qubit,
unwanted phase acquisition is prevented by this pulse sequence, however, the loss of
the phase coherence by the pure dephasing still remains. Therefore, in combination
with the ordinary Ramsey interference, the spin-echo method is frequently used for
the measurement of pure dephasing rate separately from the systematic phase noises.

Problems

Problem 7-1 Find eigenvectors for the Pauli operators σx , σ y and σz .

Problem 7-2 Show that the density matrix of a qubit can be written as

1
ρ= (I + sx σx + s y σ y + sz σz ). (7.4.24)
2
Problem 7-3 Discuss what happens for the Bloch vector (sx , s y , sz ) when ρ is
mixed state. What if ρ = I /2?
Problem 7-4 Discuss the nature of the Bloch equation (7.4.23) for 2 = 0, and
sz (t = 0) = 1 and explain the physical situation and consequence.
Problem 7-5 Consider the steady state of the Bloch equation (7.4.23), namely
ds/dt = 0 to calculate the excited-state population ρee . It will be of
Lorentzian lineshape with respect to the detuning q and check that
power broadening occurs, that is, the linewidth gets thicker as the
Rabi frequency increases.
Electromagnetic Cavities and Cavity
Quantum Electrodynamics 8

Only for the sake of manipulating a single quantum system, one can realize it by
applying intense electromagnetic waves such as lasers and/or microwaves, except for
the photon itself as a quantum system. However, this is not the case when one wants
to convert quantum states back and forth between a fixed quantum system and a
photon, in a deterministic and efficient manner. For such a purpose, electromagnetic
cavities, or electromagnetic resonators, are outstandingly significant components in
quantum technologies for their ability to enhance the interaction between a single
quantum system and a single photon. In this chapter, we will describe some properties
of electromagnetic cavities and general schemes of cavity measurements. To put
it concretely, We will introduce quality factor and lifetime of a cavity, finesse of
a Fabry-Perót cavity, internal loss and external coupling, and input-output theory
with applications to some types of cavities. Succeeding Sections are devoted to
the introduction of cavity quantum electrodynamics where Jaynes–Cummings and
Tavis–Cummings models are introduced and weak, strong and strong dispersive
regimes are analyzed.

8.1 Properties of Cavities

An electromagnetic resonator/cavity, or simply a cavity, is characterized by its res-

onant (angular) frequency ωc and the lifetime τc of a cavity photon. The lifetime
of the cavity photon characterizes the 1/e time scale of the exponential decay of

Supplementary Information The online version contains supplementary material available at

https://doi.org/10.1007/978-981-19-4641-7_8.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 159
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7_8
160 8 Electromagnetic Cavities and Cavity Quantum Electrodynamics

Fig. 8.1 Typical frequency

spectrum of a cavity mode

the cavity photon number. Despite we introduced the lifetime, the cavity is rarely
evaluated by the lifetime itself. We instead often consider a loss rate κc = 2π/τc
of the cavity photon which is proportional to the inverse of the cavity lifetime. In
terms of the electric field of the cavity mode, it oscillates at the frequency ωc /2π
and its amplitude decays exponentially with the rate κc /2. This exponential decay
assumes that the loss rate is independent of the number of photons inside the cavity,
which is justified in almost every situation since the photon–photon interaction, or
the self-Kerr effect, is usually very weak. Frequency spectrum Sc (ω) of such a cavity
mode is, by noting that the exponential function is Fourier transformed to yield a
Lorentzian function, expected to have a form

1
Sc (ω) ∝ . (8.1.1)
(ω − ωc )2+ (κc /2)2

As can be seen in this spectral shape (see Fig. 8.1), the loss rate κc is the full-width-
half-maximum value of this Lorentzian spectrum.
An electromagnetic cavity is a tool for confining electromagnetic waves inside a
definite spatial region. As mentioned above, one of its characteristics is how long
it can trap electromagnetic waves inside it, and frequently used quantities are qual-
ity factor or Q factor Q and finesse F , both being proportional to the lifetime and
inversely proportional to the loss rate of the cavity. Which of these two we should
consider depends on various situations and losses involved in the cavity, however,
basic policy to decide which is somewhat simple, as detailed in the later Section.
To evaluate these quantities, time-domain or frequency-domain measurement is per-
formed where appropriate one is adopted in the actual experiment.
Another quantity to characterize the cavity is the mode volume V , which indicates
that in how small region it can confine electromagnetic waves. As we learned in
Sect. √
6.2, vacuum fluctuation of the electric field of a cavity mode is proportional
to 1/ V , so that it is preferable to make the mode volume as small as possible for
the sake of enhancing electromagnetic waves–matter interaction in the cavity mode.
The mode volume is determined by the shape and size of the cavity.
8.1 Properties of Cavities 161

8.1.1 Q Factor

Q factor is a quantity given by the ratio of the center frequency and the loss rate
(full-width-half-maximum) of the Lorentzian spectrum of a cavity mode
ωc
Q= .
κc

Its physical meaning becomes clear when we rewrite it as Q = τloss /τc , with τc =
2π/ωc and τloss = 2π/κc , respectively, denote the oscillation period of the cavity
electric field and the time constant of the cavity loss. This expression suggests that
the Q factor tells how many times the resonant electromagnetic field can oscillate
until it escapes from the cavity. It is implied that this quantity is a good measure if
the cavity photon is continuously exposed to the loss, which is true in most cases
except for the Fabry-Perót cavity.
Q factor gets infinity for a lossless cavity, however, in reality, various loss mech-
anisms make the Q factor finite. Let multiple loss mechanisms with lossrates
κi = ωc /Q i join the system, which give rise to the total cavity loss rate i κi .
By noting that the loss rate is inversely proportional to the Q factor, therefore, the Q
factor of the cavity becomes
 −1
 1
Q tot = . (8.1.2)
Qi
i

Nonetheless, when one designs experimental systems in practice including internal

loss and external coupling as is done in the input-output theory (see Sect. 8.3), it
might be intuitive to go with the loss rates.

8.1.2 Finesse

The simplest optical cavity is a Fabry-Perót resonator, which consists of a pair of

mirrors that traps light in between (see Fig. 8.2). In this cavity, light wave resonates
when the phase acquisition during a single turnaround in the resonator equals an
integer multiple of 2π , by which the light wave inside the cavity can constructively

Fig. 8.2 Fabry-Perót cavity and its spectra

162 8 Electromagnetic Cavities and Cavity Quantum Electrodynamics

interfere to build up. In other words, the light waves in the cavity otherwise interfere
with various phases modulo 2π to cancel out, which means that the light wave cannot
exist in the cavity. The resonance condition reads
c c
2n L = λN ⇔ ν= = N, (8.1.3)
λ 2n L
by reinterpreting the above condition as the round-trip optical length being equal
to the integer multiple of the optical wavelength. Here N denotes the non-negative
integer. One can immediately see that the cavity mode aligns equidistantly in the
frequency domain, and the frequency interval c/2n L is called the free spectral range
or FSR. Taking into account the loss of the cavity, each spectrum has a Lorentzian
lineshape with its full-width-half-maximum corresponding to the loss rate.
Let us pose here a bit thinking about the loss of the Fabry-Perót cavity. When
the light propagates in the region between the mirrors, the atmosphere between the
mirrors hardly causes loss by light scattering, as long as the wavelength is in the
visible or near-infrared region. The main origin of the loss in the Fabry-Perót cavity
is the cavity mirror, where light unwantedly flies out of the cavity due to various
mechanisms, such as surface roughness and imperfect reflectivity. Therefore, the
loss rate of the Fabry-Perót cavity is proportional to the number of times that light is
reflected by cavity mirrors within 1 s. Given this observation, the longer the cavity,
the larger the Q factor. In such a situation, the Q factor still represents a loss rate,
but not a good parameter characterizing a quality of Fabry-Perót cavity. Finesse F is
used in such a situation, instead of the Q factor. The finesse is represented by using
FSR as
FSR
F= . (8.1.4)
κc /2π

The quantity 1/FSR = 2n L/c stands for how long it takes for light to go around the
cavity and 2π/κc does for the time constant of the loss. Therefore, finesse quantifies
how many times the light can go around the cavity.

8.2 Measurement of a Cavity

Cavities often appear in experimental attempts of quantum technologies and their

measurement take various forms depending on the physical systems. However, every
measurement technique has an essential thing in common, that is, one inputs electro-
magnetic waves to the cavity and inspects the response. This will be clarified more
using input-output theory in the later section, however, let us see basic measurement
schemes here as a preparation.
8.2 Measurement of a Cavity 163

Fig. 8.3 Schematic representation of reflection and transmission measurements

8.2.1 Reflection and Transmission Measurements

Cavity has a structure that traps electromagnetic waves inside, so that the electromag-
netic waves injected from outside are repelled out ideally. However, with a proper
design one can inject a fraction of electromagnetic waves into the cavity. Hence, the
amount of electromagnetic waves reflected back from the cavity is reduced, and the
measurement of this signal reduction is referred to as the reflection measurement. If
we have another outcoupling port incorporated in a cavity, the injected electromag-
netic waves can transmit from the input port to that port. These transmitted waves
are also useful for cavity measurement since it allows for the background-free mea-
surement of the cavity. This latter method is called the transmission measurement.
A schematics in Fig. 8.3 might help the readers to grab a concept.

Spectrum measurement
The most common method of measuring a cavity is to measure its electromagnetic
response in the frequency domain. In concrete, reflection or transmission signal of
the electromagnetic waves is measured with the frequency ω/2π of the incident wave
swept over the resonance, see Fig. 8.3.
This method is valid when the cavity linewidth is broader than that of the incident
electromagnetic wave, which is the case in most experiments. In contrast, if such a
situation is not realized in the experiment, the measured spectrum shows the linewidth
of the incident wave as well and the cavity linewidth is hidden in such a noise.1 Even
in such a case, we can measure the linewidth of the cavity by ring-down method, a
time-domain measurement introduced in the next section.

1 Rather we can measure the linewidth of the electromagnetic wave in this situation
164 8 Electromagnetic Cavities and Cavity Quantum Electrodynamics

Fig. 8.4 Timeline of the input electromagnetic wave (upper panel) and observed signal (lower
panel) in the ring-down experiment

Ringdown measurement
If we continuously inject electromagnetic waves into the cavity resonance, a steady
state is realized such that a certain amount of photons are accumulated in the cavity.
The number of photons stored inside the cavity is determined by the balance between
the photon injection rate and the loss rate and detailed calculation can be done by
using the input-output theory later. Then we suddenly turn off the electromagnetic
wave or tune the frequency of the electromagnetic wave off-resonant. After such an
operation, photon injection no longer exists and the stored photons inside the cavity
leak out. The photon number decreases exponentially as inferred by the Fourier
transformation of the Lorentzian spectrum of the cavity mode. Therefore, the signal
from the cavity also decreases exponentially. This type of measurement is called the
ring-down measurement which is an analogy of ringing down the bell and examining
the decay of the sound of the bell. A schematic illustration of what we observe in
the ring-down measurement is indicated in Fig. 8.4.
In the ring-down measurement, we must quickly modulate or turn of the electro-
magnetic wave into the cavity in time domain. This is difficult for the measurement
of a cavity with thick linewidth, or equivalently a cavity with short lifetime. On the
other hand, ring-down measurement is suitable for a cavity with narrow linewidth,
or long lifetime, manifesting itself as a useful method of the cavity measurement
complimentary to the frequency-domain measurement.

8.2.2 Actual Measurement Systems

Let us take an overview of the measurements for optical cavities. As a typical case,
we pick up Fabry-Perót and ring cavities here. For a Fabry-Perót cavity, a laser beam
is incident on one of the mirrors with the same direction and the same beam profile as
8.2 Measurement of a Cavity 165

Fig. 8.5 Schematics of the actual measurement systems of a Fabry-Perót resonator (left), a ring
resonator (center) and an LC resonator (right)

leaked optical mode out of a cavity. The reflection measurement can be implemented
by collecting the reflected light at the first mirror by properly setting the polarization
of the incident optical beam. If we collect the leakage of cavity-mode light through
the second mirror instead, this measurement becomes the transmission measurement.
The leakage rate of photons of the cavity mode is dependent on the reflectances of
the mirrors. The injection rate of photons into the cavity is additionally dependent
on the spatial overlap between the input mode and leakage mode.
As for the ring cavity (center of Fig. 8.5), it is not easy to have an access to a cavity
mode from free space. In such a case, one can introduce light into the ring cavity by
putting an optical waveguide next to the ring cavity, so that the evanescent wave of
the optical waveguide overlaps that of the cavity mode which results in the coupling
between the modes of the optical waveguide and ring cavity. With this configuration,
one can observe dips in the transmission spectrum of the optical waveguide, which
originates in the resonance of the cavity modes of the ring cavity that extracts light
from the optical waveguide.
Next, we will describe the measurement systems for the cavity constructed by
electrical circuits. Right panel of Fig. 8.5 displays an LC circuit. By adding a capaci-
tor and sending radio-frequency waves or microwaves from one side of that coupling
capacitor, we can get the reflection from the LC resonator and implement the reflec-
tion measurement. Furthermore, if we add other coupling elements and monitor
leakage signals of electromagnetic waves, the transmission measurement is feasi-
ble. The coupling element is not necessarily capacitors, but inductors can also be
used to couple to the LC resonator by utilizing the mutual inductance. The differ-
ence between the use of capacitor and inductor is merely whether the LC resonator
is accessed through its electric or magnetic field. Which one to use is up to the
electromagnetic-field profiles of the resonator and to the actual design parameters
and purposes. Such circumstances are common in various cavities using electrical
components, for example, a coplanar cavity, three-dimensional cavity, and loop-gap
cavity.
166 8 Electromagnetic Cavities and Cavity Quantum Electrodynamics

8.3 Input-Output Theory

8.3.1 Propagating Mode and Input-Output Theory

In this Section, we investigate what signals we should expect in the reflection and
transmission measurements and what information we can deduce about the cavity
from the measurement outcome. In relation to this purpose, we have a problem that
we only know, or have learned through this book, how to treat spatially localized
physical systems, e.g. a qubit that is supposed to be at rest and a cavity confining
electromagnetic waves in some region. However, we want to analyze the situation
that traveling photons in the propagating mode are coupled into and out of the cavity.
We should utilize boundary conditions at which a localized mode and a propagating
mode exchange photons to analyze the system. Below, we overview the basics of the
input-output theory treating such boundary conditions called input–output relations
together with its applications for practical cavity measurements.
Here let annihilation and creation operators of one-dimensional propagating
mode be c(ω) and c(ω)† , respectively, with a commutation relation [c(ω), c† (ω )] =
2π δ(ω − ω ). The operators are dependent on the frequency of the propagating mode,
reflecting the fact that a propagating mode has a continuous spectrum. With this setup,
we shall describe a model of cavity measurement including the coupling of the cavity
mode to the propagating mode [26].
First, Hamiltonian of the whole system including a cavity mode, a propagating
mode and the coupling between them is dictated as
 ∞  ∞
dω √ dω  † 
Htot = ωc a a +
†
ωc† (ω)c(ω) − i κ a c(ω) − c† (ω)a .
−∞ 2π −∞ 2π
(8.3.1)

First term represents the cavity mode and the second one the propagating mode. The
coupling between the cavity mode and the propagating mode is incorporated in the
third term with the rate of photon exchange being represented by the quantity called
external coupling rate κ. We pose here for a moment to clarify the reason why the
external coupling rate is included in the third term by its square root. Consider the
situation where the cavity is pumped and the number of photons are stored inside
it. The quantity κ is the rate that the cavity photon leaks out toward the propagating
mode, which can be in principle evaluated using Fermi’s golden rule. Therefore, the
squared value of the
√ coefficient of the interaction Hamiltonian should coincide with
the rate κ, so that κ appears in the coefficient of the third term.
To return to our subject, let us write down the Heisenberg equation of motion
about the propagating-mode operator c(ω)
 ∞
dc(ω) i i dω  † 
= [Htot , c(ω, t)] = ω c (ω )c(ω ), c(ω)
dt   −∞ 2π
 ∞
i √ dω †
+ −i κ a c(ω ) − c† (ω )a , c(ω)
 −∞ 2π
8.3 Input-Output Theory 167
 ∞  ∞
dω  †  dω √
= iω c (ω )c(ω ), c(ω) + κ a † c(ω ) − c† (ω )a, c(ω)
−∞ 2π −∞ 2π
 ∞  ∞
dω  dω √
=− iω (2π )δ(ω − ω )c(ω ) + κ(2π )δ(ω − ω )a
−∞ 2π −∞ 2π
√
= −iωc(ω) + κa. (8.3.2)

On the other hand, the Heisenberg equation of motion for the cavity mode reads by
denoting the system Hamiltonian by Hs
 ∞
da i √ dω
= [Hs , a] − κ c(ω). (8.3.3)
dt  −∞ 2π

What we do next is to formally solve the Heisenberg equation of motion for the
propagating mode. In doing this, we have two solutions, namely cin referring to a
past time t0 < t and cout referring to a future time t1 > t. These can be written as,
by making use of the equality (dω/2π )e−iω(t−τ ) = δ(t − τ )

√ t
cin (ω, t) = e−iω(t−t0 ) c(ω, t0 ) + κ dτ e−iω(t−τ ) a(τ ), (8.3.4)
t0
 t1
√
cout (ω, t) = e−iω(t−t1 ) c(ω, t1 ) − κ dτ e−iω(t−τ ) a(τ ). (8.3.5)
t

Substituting these into the Heisenberg equation of motion for the cavity mode, we
obtain
da i κ √
= [Hs , a] − a − κcin e−i t , (8.3.6)
dt  2
da i κ √
= [Hs , a] + a − κcout e−i t . (8.3.7)
dt  2
The derivation is aided by the Markov approximation that allows us to neglect m-fold
integrals (m ≥ 2) and the substitutions
 ∞
dω −iω(t−t0 )
cin e−i t = e c(ω, t0 ), (8.3.8)
−∞ 2π
 ∞
dω −iω(t−t1 )
cout ei t = e c(ω, t1 ). (8.3.9)
−∞ 2π

There is one thing to note here about cin and cout . These operators√are now integrated
with respect to frequency. As a result, these have a unit of [ Hz]. An operator
†
n in/out = cin/out cin/out hence possesses a unit of [Hz], implying that n in/out stands
for a photon flux operator.
By combining the above two differential equations, the following input–output
relation can be obtained:
√
cout = cin + κaei t . (8.3.10)
168 8 Electromagnetic Cavities and Cavity Quantum Electrodynamics

The set of the Heisenberg equation of motion and input–output relation allows us
to calculate various experimentally measurable quantities such as absorption√and
emission spectra. The input–output relation is often rewritten as cout = cin + κa
in rotating frames with frequency for both cin and cout . In the following sections,
we will see how to use this method for understanding how cavity measurement works.

8.3.2 Single-Port Measurement

We here consider what we call single-port measurement as a simple and useful

situation. What we call the single-port measurement here refers to the circumstance
that the cavity is intended to couple to only a single propagating mode, e.g., an optical
microring resonator coupled to an optical waveguide, an electrical circuit resonator
with a capacitively or inductively coupling wiring, and so on.
As a Hamiltonian, we only have to take the cavity one H =  c a † a into account
and do not have to explicitly include the propagating mode, for we know how it
affects the Heisenberg equation of motion of a. Whenever it becomes unclear what
the equation should look like with the propagating mode, one would follow the
discussion in the previous Section again to obtain the correct equation. Here we define
the relaxation rate of the cavity photon without the external coupling, or intrinsic
loss, as κin and the external coupling rate between the cavity and the propagating
mode, here for convenience supposed to be the one inside a waveguide of some
electromagnetic field, as κex ; see Fig. 8.6. The Heisenberg equation of motion of a
reads
da κin κex √
= −i ca − a− a − κex ain . (8.3.11)
dt 2 2

Fig. 8.6 Schematics of a single-port measurement

8.3 Input-Output Theory 169

√
and the input-output relation is obtained to be aout = ain + κex a. In a steady state,
da/dt = 0 and this yields
√
κex
a=− κin +κex ain . (8.3.12)
i c + 2

By taking the product of this and its Hermitian conjugate, we obtain a relationship
between the intra-cavity photon number n cav = a † a and input photon flux n in =
†
ain ain as

κex
n cav = n in . (8.3.13)
2
c + [(κin + κex )/2]2

From this expression, we see that the intra-cavity photon number roughly equals
the input photon flux divided by the relaxation rate of the cavity when the driving
frequency is in resonance with the cavity and the intrinsic loss κin is negligible
compared to the external coupling.
Furthermore, by substituting the expression of a into the input–output relation,
we get
 
κex
aout = 1− ain
i c + κin +κ
2
ex

κin −κex
i c +
= 2
κin +κex ain . (8.3.14)
i c + 2

The transmittance T of the waveguide can be calculated as the ratio of output pho-
ton number/flux to the input photon number/flux. Then by using n out /n in =
† †
aout aout /ain ain , the transmittance is revealed to be
 κin −κex 2
2
c + κin κex
T =  κin +κex 2 = 1 −
2
  . (8.3.15)
2 + 2 + κin +κex 2
c 2 c 2

As can be seen immediately from this expression, a Lorentzian dip appears in a

transmission spectrum, manifesting itself as a signal of the cavity mode.
Related to this spectrum, we make a brief note of the three coupling regimes of the
coupled cavity-waveguide system. First, if the external coupling rate κex is smaller
than the intrinsic loss rate κin of the cavity, namely if κex < κin , the system is said to
be in an under coupling regime or one would state that the cavity is under coupled.
The left panel of Fig. 8.7 displays aout /ain in the complex plane or phase space,
using the relationship
 
κex
aout = 1− ain . (8.3.16)
i c + κin +κ
2
ex
170 8 Electromagnetic Cavities and Cavity Quantum Electrodynamics

The quantity aout /ain as a function of the driving frequency forms a circle in the
phase space. In the under coupling regime, the circle does not enclose the origin of
the phase space, hence the spectral dip does not reach zero and the phase variation
is obviously below π . This can be intuitively understood as follows: photons in
the propagating mode transmit more without coupling into the cavity in the under
coupling regime, so that not all photons can be consumed by the cavity.
When κex = κin , that is, the intrinsic loss and the external coupling balances, the
system is said to be in the critical coupling. In such a situation, the resonance circle
in the phase space reaches the origin, resulting in a zero transmittance at the very
resonance of the cavity mode and π -phase variation, see the middle panel of Fig. 8.7.
This means that the photons in the propagating mode are completely consumed in the
cavity mode. In other words, the impedance matching from the propagating mode
to the intrinsic loss channel of the cavity mode is realized. The spectral linewidth of
the cavity mode in the critical coupling becomes κin + κex = 2κin . This situation is
convenient in the experiment because when the spectral dip of the cavity mode shows
Lorentzian lineshape with its linewidth exp and zero transmittance at resonance, one
can immediately see that the intrinsic loss rate of the cavity mode is exp /2. If one has
a single transmission spectrum alone without the phase information, one can uniquely
determine the intrinsic loss rate of the cavity mode only when the measuring system
is in the critical coupling regime.
The third and last regime is the over coupling regime, where the external coupling
rate overwhelms the intrinsic loss rate, namely κex > κin . The resonance circle, in
this coupling regime encircles the origin of the phase space. This leads to a situation
that the transmittance does not reach zero at resonance while the phase variation
always yields 2π . The linewidth of the measured cavity mode is more than two
times thicker than its intrinsic one, solely due to the coupling to the propagating
mode. An over coupled cavity extracts many photons from the waveguide and at the
same time returns many photons back to it. This is the reason why the over coupled
cavity mode shows nonzero transmittance at resonance. The transmission spectra for
various coupling parameters are shown in Fig. 8.8.

Fig.8.7 Transmission signals in the phase space for the cavity mode in under-(left), critical (middle),
and overcoupling regimes
8.3 Input-Output Theory 171

Fig. 8.8 Transmission spectra for various κex for κin = 1

8.3.3 Dual-Port Measurement

The next cavity measurement under concern is what we call dual-port measurement
here, including two-sided Fabry-Perót cavities and LC resonators with two coupling
elements (Fig. 8.9). The input–output relations are considered in this case for two
input–output ports of the electromagnetic waves, as two boundary conditions. At a
port to which the electromagnetic waves are injected, the external coupling rate is
defined as κ1 . The annihilation operators for incident and reflected waves are written
as ain and aout , respectively. At the other port with external coupling rate κ2 , the
transmitted electromagnetic waves are monitored and the annihilation operator of
this field is written as bout . Heisenberg equation of motion for cavity operator a
reads, as in the previous Section

da κ √
= −i ca − a − κ1 ain . (8.3.17)
dt 2
Here, κ = κin + κ1 + κ2 is the total decay rate of the cavity, meaning that in addition
to the intrinsic loss rate of the cavity, photons can escape from the cavity through
the two ports. The input–output relations at the two ports are
√
aout = ain + κ1 a, (8.3.18)
√
bout = κ2 a (8.3.19)

reflecting the incoming and outgoing fields at the two ports, where the one includ-
ing bout does not contain the incident field since there is no input field for this
    dt =0 and calculate
port. We shall consider the steady state to put da
 
the reflectance

† † † †
R = aout aout / ain ain and the transmittance T = bout bout / ain ain . After some
172 8 Electromagnetic Cavities and Cavity Quantum Electrodynamics

Fig. 8.9 Schematics of a dual-port measurement

maths, we obtain

κ1 (κin + κ2 )
R =1− , (8.3.20)
2 + (κ/2)2
c
κ1 κ2
T = . (8.3.21)
2 + (κ/2)2
c

Here the sum of the reflectance and transmittance is not unity: R + T = 1 −

κin κ1 /( 2c + (κ/2)2 ), simply because the intrinsic loss throws photons away in nei-
ther mode aout nor mode bout . One might well note that when κ1 = κin + κ2 , the
reflectance gets zero and the transmittance is maximized at resonance c = 0. This
is the “impedance matching” from the input port to the output port.

8.4 Cavity Quantum Electrodynamics

8.4.1 Jaynes–Cummings Model

Cavity quantum electrodynamics deals with the two-level system(s) interacting with
a cavity to investigate quantum nature of photons inside the cavity or rather to control
the two-level system in between by a cavity photon. Such a physical system is well
described by the Jaynes–Cummings model

ωq
HJC = σz + ωc a † a + g(σ+ a + a † σ− ). (8.4.1)
2
and let us analyze this model here to learn most simple but powerful platform of
engineering a quantum system.
First of all, let us assume a lossless system and see what the eigenenergies of the
Jaynes-Cummings Hamiltonian look like. Here the state vector with a qubit being in
|ξ (ξ = g, e) and n photons is denoted by |ξ, n . It is straightforward to inspect the
matrix elements ξ  , m|HJC |ξ, n and one will find that this Hamiltonian is block-
diagonal. This is because the off-diagonal part, σ+ a + a † σ− , only connects |g, n
8.4 Cavity Quantum Electrodynamics 173

Fig. 8.10 Jaynes-Cummings model and Jaynes-Cummings ladder

(n)
and |e, n − 1 . By explicitly writing down the block HJC ,
 √ 
ωq
(n) − + ωc n g n
HJC = 2 √ ωq (8.4.2)
g n 2 + ωc (n − 1)

where the first and second column/row indicate the elements in bases |g, n and
|e, n − 1 , respectively,
 and n ≥ 1. This matrix has eigenvalues λ± = ωc (n −
1/2) ± (/2) (ωc − ωq )2 + 4g 2 n, which reads, in the case of atom-cavity reso-
nance ωc = ωq = ω0 ,
√
λ± = ω0 ± g n. (8.4.3)

Thus the n-th excited manifold formed √by |g, n and |e, n − 1 has (of course) two
eigenstates energetically split by 2g n. Thus, the coupled system of a harmonic
oscillator and a two-level system described by the Jaynes–Cummings model yields an
anharmonic spectrum dubbed as the Jaynes–Cummings ladder. The energy splitting
2g of the first two excited states is called the vacuum Rabi splitting.
√
As for the eigenstate, the states |ψ± n = (|g, n ± |e, n − 1 )/ 2 represent the

symmetric and anti-symmetric modes, much like the bonding and antibonding
orbitals in molecules. If one somehow excites only the qubit or only injects photons
into the cavity, the quantum state of the system oscillates among these symmetric
and anti-symmetric modes (Fig. 8.10).

8.4.2 Tavis–Cummings Model

Let us consider another situation that N qubits are coupled to a single cavity mode
with Jaynes–Cummings interaction, as schematically shown in Fig. 8.11. For sim-
plicity, we assume here that every qubit has its transition frequency ωq and that
the coupling strengths for qubits, being inhomogeneous in actual experiments, are
supposed to be the same value g. Hamiltonian of the whole system is then written as
174 8 Electromagnetic Cavities and Cavity Quantum Electrodynamics

Fig. 8.11 Tavis–Cummings

model

 ωq 
H= σz(i) + ωc a † a + g(a † σ+(i) + aσ−(i) )
2
i i

with superscript i on the Pauli operators indicate those for i-th qubit. Note here
that by considering N qubits, the Pauli operator such as σz(i) should be rigorously
I ⊗ . . . I ⊗ σz(i) ⊗ I ⊗ . . . I . We should only keep it in mind and usually abbreviate
I ’s. Since the frequencies and coupling strengths for the qubits are identical, we
cannot spectrally distinguish the qubits through the measurement of the cavity’s
response. This situation justifies the use of collective Pauli operators

1  (i)
S+ = √ σ+ ,
N i
1  (i)
S− = √ σ− .
N i

We also assume that qubits do not interact directly with each other, that is, terms like
( j)
σ+(i) σ− (i = j) does not have contributions in the Hamiltonian. We further assume
that only one qubit out of N qubits is excited at most, justifying that the first term
in the Hamiltonian can be rewritten as (ωq /2)S+ S− . The interaction terms can be
 √
rewritten as well using i σ±(i) = N S± and the Hamiltonian now reads

ωq
H= S+ S− + ωc a † a + g̃(a † S+ + aS− ).
2
This looks much √ like a Jaynes-Cummings Hamiltonian with the modified coupling
strength g̃ = g N . This enhancement of the coupling strength is owing to the pres-
ence of multiple identical qubits and is called the collective enhancement. The Hamil-
tonian shown above is called Tavis–Cummings Hamiltonian and the physical system
modeled by this Hamiltonian is called Tavis–Cummings model. The vacuum Rabi
splitting with 2g̃ and related physics are the same as those in the Jaynes–Cummings
model as long as the above assumptions are valid.
8.4 Cavity Quantum Electrodynamics 175

Fig. 8.12 Jaynes–Cummings

model with losses

8.4.3 Weak and Strong Coupling Regimes

Let us analyze Jaynes–Cummings model further here with losses incorporated. We

consider a qubit residing in a cavity as shown in Fig. 8.12, where the longitudinal
decays of the qubit (γ ) and the cavity (κ) are considered. We depart from the Jaynes–
Cummings Hamiltonian in the rotating frame at frequency ω. The transformation
matrix is U = exp iωa † at + i(ω/2)σz t . Do not be afraid of multiple terms in the
exponent since they commute. One can do such transformations one by one. We
have, as a result

 q
HJC = σz +  ca
†
a + g(σ+ a + a † σ− ). (8.4.4)
2

Here q = ωq − ω and c = ωc − ω.2 From this Hamiltonian, we obtain coupled

Heisenberg equations of motion for a and σ− that appear as

da κ
= −i c a − a − igσ− , (8.4.5)
dt 2
dσ− q γ
= −i σ− − σ− + igσz a. (8.4.6)
dt 2 2
The loss terms are introduced on the right-hand side as the second terms. At this
point, we make an approximation that the qubit is mostly in the ground state, that
is, the expectation value of σz is −1, which is valid when the qubit is very weakly
excited. By assuming this nonlinear term +igσz a in the second equation is simplified
as −iga and we here consider the situation that the multiphoton states are negligible.
Above equations of motion lead to the coupled linear differential equations
    κ
  
d a − i c + 2  −ig  a
= . (8.4.7)
dt σ− −ig − i 2q + γ2 σ−

2 Note the phase convention of the interaction term. See also the Problems.
176 8 Electromagnetic Cavities and Cavity Quantum Electrodynamics

Note the resemblance of these equations to the Schrödinger equation when the Her-
mitian conjugate of the two lines of equations are taken and they act on the state
|g, 0 . Therefore, it can be deduced that the eigenvalues of the matrix in the right
hand side are closely related to the energies of the cavity-qubit coupled system. By
an ordinary diagonalization procedure yields eigenvalues

  2
E± 1  κ  γ 1 κ 1 q γ
=− i c+ + i q+ ± i c+ − i + − g2
i 2 2 2 2 2 2 2 2
(8.4.8)

with eigenvectors
  
1 1  κ q γ
|ψ± =√ ic+ − i + (8.4.9)
A± 2 2 2 2
 ⎤ ⎫
    2 ⎬
1 κ 1 γ
− g 2 ⎦ |g, 1 − g|e, 0
q
± i c+ − i +
2 2 2 2 2 ⎭
(8.4.10)
√
where the factor 1/ A± is present for the normalization. As the loss terms are
involved, the eigenenergies E ± are complex-valued in general. Their real and com-
plex parts are regarded as the energies and losses, respectively.
Here we can discuss interesting regimes of this cavity QED system; weak and
strong coupling regimes. These are discriminated by whether the real parts of E ±
exhibit the same value or not. For brevity, we restrict ourselves in the case c = q =
0. When the coupling strength g is smaller than |κ/4 − γ /4|, the weak coupling
regime, the quantity in the square root is always positive and hence Re[E ± ] is equal.
In other words, the qubit and the cavity spectrally coincide with each other except
for the linewidths and there is no frequency shift for either due to the weakness of
the coupling compared to the losses. This situation in turn makes the loss rates of the
qubit and the cavity deviate from the original ones. Let us consider the bad cavity
limit κ/4 γ /4 g and expand the square-root term as follows:
⎡ ⎤
E± 1 κ γ ⎢ 1 κ
1 g2 ⎥ γ
=− + ± − ⎣1 −   ⎦ (8.4.11)
i 2 2 2 2 2 2 2 κ−γ 2
4

κ ∓κ γ ±γ 1 g2
=− − ∓ κ−γ (8.4.12)
4 4 2 4
8.4 Cavity Quantum Electrodynamics 177

Now we write down + = E + /i and − = E − /i separately

   
γ g2 4g 2 γ
+ =− 1 + γ κ−γ 1+ − , (8.4.13)
2 2 2
κγ 2
   
κ g2 κ 4g 2
− =− 1 − κ κ−γ  − 1− 2 . (8.4.14)
2 2 2
2 κ

The first thing we can say is that + and − , respectively, correspond to the qubit
and cavity loss rates, since their absolute values approach their original values as the
coupling strength g vanishes. In the presence of the finite coupling strength, they are
modified from their intrinsic values, e.g., atomic loss increases by F p = 4g 2 /κγ .
This enhancement factor F p is called the Purcell factor, or the atomic cooperativity
in the context of the cavity QED.
Next, we shall see what happens when the coupling strength g is greater than
|κ/4 − γ /4|, the strong coupling regime. In this situation, the quantity in the square
root becomes negative and the real parts of E + and E − are no longer the same. This
is an implication of the fact that the atom and the cavity is hybridized to form two
coupled modes. Again we consider the case c = q = 0 to simplify the maths.
Given the above circumstances, we rewrite the eigenenergies as

 2
κ −γ  κ γ
E ± = ∓ g 2 − −i + (8.4.15)
4 2 2 2

from which it can be seen that the loss rates of the two modes take
 the simple average
 2
of those of them. The energetic shifts are ΔE ± = ∓ 0 = ∓ g − (κ − γ )/4
2
and the energy splitting given by an amount of 2 0 is called the vacuum Rabi
splitting. How linewidths and the energies behave in the strong coupling regime are
plotted in Fig. 8.13.

Fig. 8.13 Linewidths (left, Imaginary part of E ± ), energies (center, Imaginary part of E ± ) and their
dependence on the detuning (right). Parameters are chosen so that the cavity and qubit lienwidths
are 1 and 0.05
178 8 Electromagnetic Cavities and Cavity Quantum Electrodynamics

8.4.4 Dispersive Regime

In the previous Section, the Jaynes–Cummings Hamiltonian is investigated when

the qubit and the cavity are on- or near-resonance. In this Section, however, let us
consider an interesting, alternative regime in which the qubit and the cavity are far-
off resonant, but the coupling constant is moderately large enough so that the qubit
and the cavity couple dispersively with each other.
The consequences of the analysis shown in this subsection are somehow related
to the dressed-state picture and ac Stark shift or light shift therein. We will see that,
as a result of the dispersive coupling, the energy of the cavity is shifted by the qubit
and the amount of the shift depends on whether the qubit is in |g or |e , on one hand.
On the other hand, the energy of the qubit will be shifted as well, and the energy
shift depends on the number of photons, just as in the light shift. Indeed, what we
are going to see here is a fully quantum-mechanical treatment of a light shift.
Let us depart again from the Jaynes–Cummings Hamiltonian

ωq
H= σz + ωc a † a + g(σ+ a + σ− a † ). (8.4.16)
2
 
Riding onto the frame rotating at ωc by U = exp iωc a † at + i(ωc /2)σz t , the
Hamiltonian in the rotating frame gets H = ( /2)σz + g(σ+ a + σ− a † ) where
= ωq − ωc . Then we approximatelydiagonalize the Hamiltonian
 by Schrieffer-
Wolff transformation3 using e S = exp (g/ )(σ+ a − σ− a † ) ,4 which is meaning-
ful only when |g/ |  1. The truncated Hamiltonian reads H = ( /2)σz +
(g 2 / )(a † a + 1/2)σz and back to the original frame,
 
 ωq g 2 1
H = σz + ωc a a +
†
a a+
†
σz . (8.4.17)
2 2

This is often called the dispersive Hamiltonian and according to our intention it is
apparently diagonal, that is, the Hamiltonian consists only of σz and a † a. We shall
interpret this Hamiltonian in two ways: first
 
  g2 g2
H = ωq + σz +  ωc + σz a † a. (8.4.18)
2

There can be seen two types of frequency shifts of qubit and cavity nature. The
frequency shift of the qubit by g 2 / is caused by the coupling of the qubit to the
vacuum fluctuation, which is read as the Lamb shift. The one of the cavity by an
amount (g 2 / )σz is interpreted as the frequency shift depending on the state of the
qubit, which is called the dispersive shift. With this expression, it is apparent that the
cavity peak will split into two, representing the ground and excited states of the qubit.

3 see Appendix F.  
4 For consistency with Appendix F, one might put egS = exp (g/ )(σ+ a − σ− a † ) .
8.4 Cavity Quantum Electrodynamics 179

Fig.8.14 Spectral features (schematic) of Jaynes–Cummings model in the strong dispersive regime.
Left: cavity spectrum. Right: qubit spectrum

This spectral feature is schematically illustrated in Fig. 8.14. The two resonances are
resolved when 2g 2 / > γ . Next, the Hamiltonian can also be written in the form
 
g2 1
H =  ωq + a†a + σz + ωc a † a. (8.4.19)
2

This form tells us that other than the Lamb shift, the qubit frequency is shifted dis-
cretely and exhibits multiple peaks, each of which corresponds to the photon number
(see right panel of Fig. 8.14). Using this property, the photon-number resolving exper-
iments are implemented using a superconducting qubit [16] under the condition that
the peaks are resolved, g 2 / > κ. One can see that if we drive the cavity strongly, we
can make a replacement a → α + a with α being a classical, complex amplitude and
a after the replacement is only responsible for the quantum noise that is negligibly
small compared to the amplitude |α|. Then we define the Rabi frequency gα =
to get the frequency shift of the qubit by the classical driving as 2 / which is the
very thing we saw in the discussion of the dressed-state picture.

8.4.5 Waveguide-Coupled Cavity QED System

We here consider a system where a coupled qubit-cavity system is further coupled

to a propagating mode in a waveguide and analyze what kind of signal one can see
and how the cavity QED system looks like through the transmission measurement
with the waveguide as depicted in Fig. 8.15.

Fig. 8.15 Waveguide-coupled cavity QED system

180 8 Electromagnetic Cavities and Cavity Quantum Electrodynamics

Here, for instance, we suppose that a photonic-crystal cavity is evanescently cou-

pled to the optical waveguide. The Hamiltonian of the system is simply of Jaynes–
Cummings type, that is
ωa
H= σz + a † a − ig(σ+ a − σ− a † ) (8.4.20)
2
In this Section, we write the intrinsic loss rate of the cavity as κin and the external
coupling rate to the waveguide as κex . Total loss rate of the cavity mode is κ =
κin + 2κex . The factor of 2 for the external coupling rate originates in the fact that the
cavity photons couple to both of the left-going and right-going propagating modes.
We assume the incident waveguide mode bin is coming from the left and transmitted
mode and reflected mode are denoted by bout and cout , respectively. In a rotating
frame with angular frequency ω, the drive frequency, for qubit and cavity operators,
the Heisenberg equations of motion are
da κ √
= −i c a − a − gσ− − κex bin , (8.4.21)
dt 2
dσ− a γ
= −i σ− − σ− − gσz a, (8.4.22)
dt 2 2
dσz
= 2g(σ+ a + σ− a † ) (8.4.23)
dt
where a = ωq − ω and c = ωc − ω are detunings of the qubit and the cavity.
Input-output relations are
√
bout = bin + κex a, (8.4.24)
√
cout = κex a (8.4.25)

from which we want to calculate the transmittance and reflectance. However, this
equation is not linear due to the presence of σz in (8.4.22). Furthermore, here we
do not assume that the qubit is only weakly excited and σz = −1. In other words,
we want to consider the situation that the qubit can be strongly driven to make
σz = −1. Keeping this in mind, let us regard coherent electromagnetic waves bin ,
bout and cout as classical quantities with the same terminologies, and also the Pauli
operators σz and σ− of the coherently driven qubit as sz and s, respectively [17]. By
this prescription, we lose the single-photon-level response of the cavity and the qubit
but can know how the coupled system behaves with the classical input and output
electromagnetic waves. After a bit lengthy calculation, one obtains

cout = −r ( c , a )bin , (8.4.26)

bout = [1 − r ( c , a )]bin , (8.4.27)
1
sz = − , (8.4.28)
1+ X

1 1 2ηr0 ( c )
s=− bin . (8.4.29)
η 1 + X 2ηr0 ( c ) + 2(i a + γ /2)
8.4 Cavity Quantum Electrodynamics 181

Here the following parameters are defined:

g2
η= , (8.4.30)
κex
κex
r0 ( c ) = , (8.4.31)
i c + κ/2
2ηr0 ( c ) 1
r ( c, a ) = 1 − r0 ( c ), (8.4.32)
2ηr0 ( c ) + 2(i a + γ /2) 1 + X
|bin |2
X= , (8.4.33)
Ps
  γ 2  κ 2 "
1 2 2+ γ ηκex κ
Ps = 2 a+ c + η2 κex
2 − 2ηκ
ex a c +
8ηκex 2 2 2
(8.4.34)

r ( c , q ) represents the reflectance of the coupled qubit-cavity system and r0 ( c )

the reflectance in the absence of the qubit. Ps is a value of photon flux with which
the qubit is saturated and X denotes a photon flux normalized by Ps .
By using above expressions, we can now examine the behavior of the coupled
qubit-cavity system by evaluating the transmittance |bout |2 /|bin |2 and the reflectance
|cout |2 /|bin |2 . These spectra are plotted in Fig. 8.16 for various values of X in the
case of the frequencies of the qubit and the cavity coincides and κ > γ . Let us look
at the lower panel of Fig. 8.16, where the transmission spectra are displayed. For
small X , that is, when the incident electromagnetic waves are weak enough to only
weakly excite the qubit, the coupled qubit-cavity system exhibits a gross resonance
dip structure of the cavity with a steep peak at resonance as a result of the destructive
interference of the photons from the qubit and the cavity. In other words, the cavity
becomes transparent at resonance as a result of the coupling to the qubit, which is
dubbed as the dipole-induced transparency. Corresponding to this, a reflection dip
appears in the reflection spectrum.
Next, let us take a look at the dependence of the reflection spectrum on the
coupling strength g and the qubit detuning a with c set to be zero (lower panel of
Fig. 8.17). As the coupling strength gets larger, the dip at the center of the cavity peak
becomes larger and the spectrum exhibits two peaks split by 2g (see the upper panel
of Fig. 8.17), implying that the coupled qubit-cavity system is in the strong coupling
regime. The lower panel of Fig. 8.17 plots the spectra when the qubit frequency is
swept over the cavity spectrum. When the qubit-cavity detuning is large, the spectrum
possesses two peaks, i.e., sharp one by the qubit and a thick one by the cavity. As
the qubit peak approaches the cavity one, the lineshape of the qubit peak appears
dispersive, in other words, it exhibits Fano lineshape, at the shoulders of the cavity
peak.
182 8 Electromagnetic Cavities and Cavity Quantum Electrodynamics

Fig. 8.16 Upper panel and its inset, respectively, show the waveguide reflection spectra and expec-
tation values of σz for various normalized input photon flux X . The lower panel shows the waveguide
transmission spectra for various X
8.4 Cavity Quantum Electrodynamics 183

Fig. 8.17 Reflection spectrua for various coupling strengths g (upper panel) and qubit-cavity detun-
ing (lower panel)

Problems

Problem 8-1 Suppose one executes the reflection measurement of a cavity that
has only a single coupling port. Then they obtain a cavity spectrum
like the one shown in Fig. 8.18 by measuring the reflected signal
which is proportional to the number of reflected photons. Given this
spectrum, express the intrinsic loss rate and external coupling rate of
the cavity.
Problem 8-2 Discuss what information is required to determine the external cou-
pling rate uniquely in the single-port measurement.
Problem 8-3 Suppose one executes the reflection and transmission measurements
of a cavity that has two coupling ports. Then they obtain a cavity
184 8 Electromagnetic Cavities and Cavity Quantum Electrodynamics

Fig. 8.18 Cavity spectrum with a single-port measurement

Fig. 8.19 Cavity spectra with a dual-port measurement

spectra like the one shown in Fig. 8.19 by measuring the reflected and
transmitted signals which are proportional to the number of reflected
photons. Here left and right panels of Fig. 8.19 show the reflection
and transmission spectra, respectively. The S0 is the signal level at
which all the input electromagnetic waves are detected. Given these
spectra, express the intrinsic loss rate and external coupling rates of
the cavity.
Problem 8-4 (i) Discuss what assumption is made for the decay process when one
says that the system decays exponentially.
(ii) Show that Fourier transformation of an exponential function is a
Lorentzian function.
Problem 8-5 Interaction terms like g(σ+ a + a † σ− ) sometimes appear in other
context as −ig(σ+ a − a † σ− ). Show that these  can be †regarded
 as
equivalent
 under
 the transformation U = exp i(π/2)a a or U =
exp i(π/2)σz . This phase difference is not significant as long as
the relative phases of each terms in the Hamiltonian do not matter.
Problem 8-6 Derive Eq. (8.4.17) by applying Schrieffer-Wolff transformation (see
Appendix F).
Various Couplings in Quantum
Systems 9

In this chapter, we pick up several quantum systems to see examples of interac-

tions, or couplings, among quantum systems and estimate the values of the coupling
strengths in practice. The basic strategy is to consider classical interaction ener-
gies first, then quantize them to derive interaction Hamiltonians, and finally get the
coupling strengths which include susceptibilities of the particles to an applied field
and/or vacuum fluctuations of the field.

9.1 Interaction Hamiltonians

Before diving into the concrete physical systems, we would like to introduce sev-
eral types of interaction Hamiltonians to which their specific names are given. To
“read” or deduce what is going on with the given interaction Hamiltonian, it is impor-
tant to grasp the involved quantum systems first and then interpret the products of
operators as describing concrete processes that can happen in the coupled system
under concern. Keeping this in mind, let us take an overview of various interactions
Hamiltonians frequently appear in quantum technologies.
Throughout this section, the following notations are used. First, a, a † , b, b† , . . .
represent annihilation and creation operators of harmonic oscillators with eigen-
frequencies ωa , ωb , . . . . Pauli operators σ with proper subscripts x, y, z, + and
− and superscripts (i) are assigned for i-th qubit, while the superscript is not
attached when only a single qubit is under concern. In most examples, g denotes the
coupling strength, however, keep in mind that they are all different and have their

Supplementary Information The online version contains supplementary material available at

https://doi.org/10.1007/978-981-19-4641-7_9.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 185
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7_9
186 9 Various Couplings in Quantum Systems

own expression in terms of susceptibilities and vacuum fluctuations, as introduced

in later Sections. The interaction Hamiltonian is commonly denoted by Hint .

9.1.1 Jaynes–Cummings and Anti-Jaynes–Cummings Interactions

Interaction Hamiltonian of a Jaynes–Cummings interaction involves a qubit and a

harmonic oscillator, exchanging their energy

Hint = g(a † σ− + aσ+ ) (9.1.1)

This type of interaction is ubiquitous for a coupled quantum system of a harmonic

oscillator and a two-level system. The figure below depicts the physical process
realized by the Jaynes–Cummings interaction.

Anti-Jaynes–Cummings interaction is a complementary one to the Jaynes–

Cummings interaction in a sense that “counter-rotating” terms are taken from the
original quantum Rabi interaction g(a + a † )(σ− + σ+ ), from which the Jaynes–
Cummings interaction can be derived through the rotating-wave approximation. It
expresses the simultaneous creation or annihilation of the excitations and the concrete
form reads

Hint = g(a † σ+ + aσ− ). (9.1.2)

and is schematically represented as in the following figure.

In the next section, we introduce the trapped ion system as a first example. In that
system, the phonon mode and the internal state of an ion are coupled by properly
utilizing the energy and momentum of the incident laser. The coupling is either
Jaynes–Cummings or anti-Jaynes–Cummings interaction depending on whether the
frequency of the laser is slightly red-detuned or blue-detuned, respectively. This will
be seen later in Sect. 9.2.2.

9.1.2 Beam-Splitter and Two-Mode-Squeezing Interactions

Two modes of harmonic oscillators can interact with each other and they can exchange
their energy quanta when their eigenfrequencies are sufficiently close together. This
9.1 Interaction Hamiltonians 187

process is described by a beam-splitter interaction

Hint = g(a † b + ab† ). (9.1.3)

which allows an intuitive interpretation as shown below.

The two-mode-squeezing interaction can be introduced in almost the same spirit as

the anti-Jaynes-Cummings interaction, since the beam-splitter interaction is obtained
via an application of a rotating-wave approximation to the original coupling of the
form g(a + a † )(b + b† ). The interaction Hamiltonian

Hint = g(a † b† + ab) (9.1.4)

manifest itself as the one of two-mode-squeezing interaction and graphically it can

be shown as below.

If the two modes involved in the two-mode-squeezing interaction are identical,

the interaction Hamiltonian reads

Hint = g[(a † )2 + a 2 ] (9.1.5)

that is called the squeezing interaction. Since this interaction makes two quanta
created or annihilated simultaneously, the physical situation is interpreted as depicted
in the following figure.

Likewise, the two-mode-squeezing interaction correlates the quantum states of

two harmonic oscillators, and the squeezing interaction correlates a quantum state
of a harmonic oscillator to itself in the phase-space: a Wigner function of it will be
transformed, or squeezed, as in Sect. 6.3.3.
188 9 Various Couplings in Quantum Systems

9.1.3 Interaction Between Qubits

Interaction terms such as gσx(1) σx(2) , gσz(1) σz(2) and gσx(1) σz(2) are respectively
called by their rotation axes as X X , Z Z and X Z interactions. These are literally the
simultaneous rotations of Bloch vectors about the X axes of the two qubits. If the
interaction Hamiltonian is given as some combination of qubit-qubit interaction, say,
(1) (2) (1) (2)
g(σx σx + σ y σ y ), this is never simple, separate rotations of Bloch vectors
about some axes. This becomes clear if we recognize that the time evolution with the
interaction Hamiltonian Hint is described by the unitary operation U (t) = e−i Hint t/ ,
as will be discussed later in detail in Sect. 10.1.2.

9.1.4 Nonlinear Interactions

The last type of interaction is nonlinear interaction, in which the effect of the interac-
tion depends on the number of quanta. Nonlinear interaction is literally diverse, and
we here only pick up two simplest examples of χ (2) and χ (3) nonlinear interactions.

The first example is the χ (2) interaction within a single harmonic oscillator:

Hint = χ (2) [a † a 2 + (a † )2 a]. (9.1.6)

This interaction Hamiltonian represents the three-wave mixing, where two quanta
with energies ω1 and ω2 are absorbed (emitted) and a single quantum with energy
ω1 + ω2 is emitted (absorbed) by some media (see the figure below).

The most frequently appearing χ (2) process takes place in the second-harmonic
generation, often abbreviated as SHG, where for instance an intense 1064 nm-
wavelength laser beam optically pumps the nonlinear crystal and transforms two
such near-infrared photons into a green, 532 nm-wavelength photon. By regarding a
strong pump photons as a classical wave with its complex amplitude α to substitute
α + a into a and apply rotating-wave approximation, we can replace one of the three
operators to get a squeezing Hamiltonian. The optomechanical interaction

Hint = ga † a(b + b† ) (9.1.7)

is also understood as a χ (2) interaction. Strong pump laser allow us to replace a

by α + a again. Depending on the pump detuning, the rotating-wave approximation
then leads to beam-splitter or two-mode-squeezing interaction.
9.2 Atomic Ions 189

Next we would like to introduce the χ (3) interaction in a single harmonic oscillator,
which is also known as a self-Kerr effect. The interaction Hamiltonian is

Hint = χ (3) a † aa † a (9.1.8)

that implies that the four-wave mixing with the absorption of quanta with energies
ω1 and ω2 are absorbed and those with ω3 and ω4 are absorbed. Note that the
energy conservation requires ω1 + ω2 = ω3 + ω4 . This situation is depicted in the
left panel of the figure below.

Its meaning will become clear if the total Hamiltonian Htot = ωa † a +
χ (3) a † aa † a = (ω + χ (3) n̂)n̂ is considered. Its eigenenergy reads (ω + χ (3) n)n
and the level spacing does not become equidistant, as indicated in the right panel
of above figure where χ (3) is assumed to be negative in this case. This means that
the presence of even a single quanta affects the energy of another quanta. In such
a situation, we can regard the quanta interact with themselves to shift their energy
according to the number of quanta present.
As another example, let us think about the interaction Hamiltonian

Hint = χ (3) a † ab† b. (9.1.9)

By combining with the bare energies ωa a † a + ωb b† b, we see that the presence
of one quanta shifts the energy of the quanta of the other harmonic oscillator. This
interaction is called the cross-Kerr effect.

9.2 Atomic Ions

In most cases, atomic or atomic-ion systems allow us to derive analytical expressions

of the coupling strengths that agree well with the experimentally implemented ones.
Let us see the atomic-ion system for example, in which ion interacts with light by its
electric dipole moment, ions’ motional degrees of freedom can be coupled to each
other as well as to light.

9.2.1 Atom-Light Interaction

Let us dictate here again the results of Sect. 7.2 about the light-atom interac-
tion. When we consider the electric dipole interaction d · E, d = μ(σ+ + σ− ) and
190 9 Various Couplings in Quantum Systems

E = E zpf (a + a † ) yield the interaction Hamiltonian

Hint = μE zpf (σ+ + σ− )(a + a † ). (9.2.1)

Furthermore, rotating-wave approximation is applicable to this Hamiltonian when

the frequencies of the optical transition of the ion ωq and the cavity ωc is close to each
other. The interaction Hamiltonian reads then the Jaynes-Cummings Hamiltonian

Hint = μE zpf (σ+ a + σ− a † )

= g(σ+ a + σ− a † ). (9.2.2)

with the coupling strength g = μE zpf /. This coupling strength of the Jaynes-
Cummings interaction is obviously a product of the transition dipole moment and
the vacuum fluctuation of the electric field of the cavity mode.
The transition dipole moment μ and the inverse of the excited-state lifetime,
denoted by , have a relationship

μ2 ωq3 3π 0 c3 
= ⇐⇒ μ= (9.2.3)
3π 0 c3 ωq3

only if the excited-state lifetime is dominated by the radiative recombination and

nonradiative recombination can be neglected. For instance, a 422 nm-wavelength
optical transition of the strontium ion has  = 2π × 23 MHz, which is equivalent
to the transition dipole moment of this transition being 1.39 × 10−29 C·m.
Next, let us consider a Fabry-Perót cavity with its length being L, radii of curvature
R and the beam-waist radius w0 , as depicted in Fig. 9.1. The mirrors are usually
made of multiple layers of dielectric coatings and the effective cavity length should
include the optical penetration depth at the mirrors, however, the penetration depth
is comparable to the wavelength of light, and in most cases, it is far smaller than L.
The optical penetration can thus be neglected and the effective length is here set to
be L. The electric-field amplitude of the fundamental mode of this cavity is written

Fig. 9.1 Schematic of a

cavity QED system with an
atom inside a Fabry-Perót
cavity
9.2 Atomic Ions 191

under the assumption that the cavity length L and the radii of curvature of the mirrors
R are sufficiently large compared to the wavelength λ
 
2π x y2 + z2
E(r) = E 0 sin exp − . (9.2.4)
λ w02

The mode volume of the cavity mode is defined by the integral of the squared electric
field and simple calculation yields
  ∞  ∞
1 L π w02
V = 2
dx dy dz E(r)2 = L. (9.2.5)
E max 0 −∞ −∞ 4

This mode volume is an important quantity since the √ vacuum fluctuation of the
electric field of the cavity mode is defined as E zpf = ωc /2ε0 V , as we mentioned
earlier. Since the beam radius w0 can be written as [14]
  1
Lλ 2R − L 4
w0 = , (9.2.6)
2π L

the mode volume reads

λ
V = L 3 (2R − L). (9.2.7)
8
Assuming that λ = 422 nm, L = 200 µm and R = 1 mm, we have V = 6.3 ×
106 µm3 .
By putting above parameters in g = μE zpf /, we can estimate the coupling
strength of the optical transition of the ion and the cavity mode as g/2π = 0.96 MHz.
Since the natural linewidth, or the inverse of the excited-state lifetime, under concern
is 2π ×23 MHz, the coupling strength is inferior to the natural linewidth and this
cavity QED system is in the weak coupling regime regardless of the decay rate of
the cavity mode. In order to increase the ratio of the coupling strength to the natural
linewidth and the cavity one, we should make some efforts as illustrated below. One
prescription is to use an optical transition with narrower natural linewidth. At first
sight, the narrower linewidth is equivalent to the smaller transition dipole moment
and it seems that the coupling strength gets smaller. However, if the wavelength
of the transition is long enough, it may rather enhance the coupling strength, see
Eq. (9.2.3). Furthermore, the narrower linewidth makes it easy for the system to
get into the strong coupling regime, even if the enhancement is just incremental or
not achieved. Another straightforward prescription is to make a smaller cavity to
make the mode volume smaller and the vacuum fluctuation E zpf larger, to enhance
the coupling strength. The use of mirrors with smaller radii of curvature R is also
beneficial for this purpose.
192 9 Various Couplings in Quantum Systems

Fig. 9.2 Quantized motion of an ion and the sideband transitions

9.2.2 Sideband Transitions: Optomechanics with an Ion

Paul trap is an ion-trapping technique that confines atomic ions in an effective har-
monic potential generated by an oscillating quadrupolar electric field. Ions then
exhibit harmonic oscillations, whose quanta are called phonons. If only a single ion
is trapped, three phonon modes exist, and if there are N ions, 3N phonon modes
are supported by the chunk of the ions connected through Coulomb interaction. In
this section, we consider the situation that only a single ion is trapped, and a phonon
mode only in one direction is taken into account (Fig. 9.2).
By the presence of phonon mode, the ion can oscillate with frequency ωm and
its position fluctuates even if the ion is in its motional ground state by the vacuum
fluctuation of the phonon. Then if the laser is irradiated on the ion, the phase of the
laser field sensed by the ion can vary according to the ion’s position x. We can take
this effect into account by including the ion’s position in the phase factor of the laser
field to rewrite as E(x) = E zpf eikx (a + a † ).1 Here k = 2π/λ is the wavevector of
the laser beam. We now denote an annihilation operator of the phonon mode as b
and the vacuum fluctuation of the phonon as xzpf to rewrite the dipole transition term
μE(x) as

geikx (σ+ + σ− )(a + a † )

= geikxzpf (b+b ) (σ+ + σ− )(a + a † )
†

 g[1 + ikxzpf (b + b† )](σ+ + σ− )(a + a † )

= g(σ+ + σ− )(a + a † )
+ iηg(b + b† )(σ+ + σ− )(a + a † ). (9.2.8)

Here the dimensionless parameter η = kxzpf is defined, which is called Lamb–

Dicke parameter. The first term in the above expression yields Jaynes–Cummings

1 Here for the simplicity the laser field is rewritten as if it is a localized mode, or cavity mode,

however, we can switch to the propagating mode just by replacing localized photon operators a and
a † by corresponding ones.
9.2 Atomic Ions 193

interaction via a rotating-wave approximation as usual, and is called the carrier tran-
sition that does not have something to do with the phonons.
What we are interested in here is the phonon-related transitions contained in
the second term. To simplify the problem, we shall assume that the irradiated laser
light is intense, continuous one, and the photon operators can be replaced by a
coherent amplitude α that is assumed to be real-valued here, multiplied by a time-
evolution factor. In such a situation, if the irradiated laser field has a frequency
ωq − ωm , rotating-wave approximation allows us to eliminate the terms oscillating
with ωq + ωm to yield

iηgα(σ+ b + σ− b† ). (9.2.9)

This is a Jaynes–Cummings interaction between the ion’s optical transition and the
phonon mode. Since the process described by the above interaction can be driven
by the laser with frequency ωq − ωm that is red-detuned from the carrier transition
frequency ωq , this process is called the red-sideband transition.
On the other hand, if the ion is driven with the frequency ωq + ωm , this time we
obtain

iηgα(σ+ b† + σ− b). (9.2.10)

This is an anti-Jaynes–Cummings interaction between the ion’s optical transition and

the phonon mode again. Since the drive frequency ωq + ωm is blue-detuned from the
carrier transition frequency ωq , this process is called the blue-sideband transition.
If the two-tone driving with ωq − ωm and ωq + ωm is executed on the ion, red- and
blue-sideband transitions are driven at the same time.

9.2.3 Phonon–Phonon Interaction

When two monovalence ions are interacting with each other by Coulomb force in
a harmonic potential, phonon–phonon interaction is significant. We here derive the
phonon–phonon interaction Hamiltonian and its coupling strength by considering the
position-dependent electrostatic energy between the two ions. As in Fig. 9.3, ions

Fig. 9.3 Ions’ phonons

interacting with each other in
a harmonic potential
194 9 Various Couplings in Quantum Systems

at Z 1 and Z 2 are separated by a distance Z = Z 2 − Z 1 and the displacements from

their equilibrium positions are denoted by r1 = (x1 , y1 , z 1 ) and r2 = (x2 , y2 , z 2 ).
Then we shall calculate the Coulomb energy E ph−ph between these two ions as

1 e2 1
E ph−ph =
2 4π 0 |r1 − r2 |
e2 1
=
8π 0 (x1 − x2 )2 + (y1 − y2 )2 + (Z + z 1 − z 2 )2
− 21
e2 2(z 1 − z 2 ) (z 1 − z 2 )2 (x1 − x2 )2 (y1 − y2 )2
= 1+ + + +
8π 0 Z Z Z 2 Z 2 Z2
e2 z1 − z2 1 (z 1 − z 2 )2 1 (x1 − x2 )2 1 (y1 − y2 )2
 1− − − − .
8π 0 Z Z 2 Z2 2 Z2 2 Z2
(9.2.11)

In this case, we consider phonons only in z-direction and set x1 = x2 = y1 = y2 = 0.

Then above energy can be simplified as

e2 z1 − z2 1 (z 1 − z 2 )2
E ph−ph = 1− − . (9.2.12)
8π 0 Z Z 2 Z2

Here it is the time for the substitution of the displacements by operators, namely
z i = z izp f (ai† + ai ) for i = 1, 2. ai are the annihilation operators of the phonons. We
focus on the third term alone, since the phonon–phonon interaction resides there but
not in the first two terms. The phonon-phonon interaction Hamiltonian Hph−ph at
this stage reads

e2 (z 1 − z 2 )2
Hph−ph =
16π 0 Z Z2
e2 z 1zp f z 2zp f
= (a1† )2 + a12 + (a2† )2 + a22 + (2a1† a1 + 1) + (2a2† a2 + 1)
16π 0 Z 3
−2(a1† a2† + a1 a2 ) − 2(a1† a2 + a1 a2† ) . (9.2.13)

Among the terms, we can eliminate about a half of them by assuming that two phonon
modes have the same oscillation frequency ωm

e2 z 1zp f z 2zp f
Hph−ph = a1† a1 + a2† a2 − a1† a2 − a1 a2† . (9.2.14)
8π 0 Z 3
The first two terms of above expression just shift the energies of the phonons, there-
fore, we omit them. The final form of the interaction Hamiltonian reads
 
Hph−ph = −gph−ph a1† a2 + a1 a2† (9.2.15)
9.3 Superconducting Circuits 195

which is obviously a beam-splitter-type interaction. This interaction is also phys-

ically intuitive since it expresses the exchange of phonons between the two
phonon modes. The phonon-phonon coupling strength is now defined as gph−ph =
1 z 2 /8π  Z 3 . Therefore, by plugging the expressions of the vacuum fluc-
e2 z zpf 0
zpf
√
tuations of the phonon modes z zpf i = /2mωm , which can be quickly derived by
equating a half of the zero-point energy and the potential energy of the harmonic
oscillation: ωm /2 = mωm 2 (z i )2 /2, we can obtain concrete expression of the cou-
zpf
pling constant as

e2
gph−ph = . (9.2.16)
16π 0 mωm Z 3

This is the case for the circumstance that the two ions have the same mass, however,
one can obtain the coupling strengths for more general situations by replacing the
expressions of the vacuum fluctuation of phonons.
Let us evaluate the coupling strength with realistic parameters at last. As typical
values, we set Z = 10 µm, m = 40m p with m p being proton mass and ωm /2π =
100 kHz, the coupling strength is evaluated to be gph−ph /2π = 220 kHz.

9.3 Superconducting Circuits

In a superconducting material, where Bardeen–Cooper–Schrieffer theory is valid,

electrons form so-called Cooper pairs with the help of attractive interaction in a
momentum space that is mediated by electron–phonon interaction. Cooper pairs
share a macroscopic wavefunction over the entire superconductor to exhibit quantum-
mechanical behavior such as interference effect. The fact that the resistance of the
superconductor is nearly equal to zero means that the quality factor of the LC res-
onator drastically increases, which is otherwise limited by the Joule heating with
parasitic resistance.
In addition to this, Josephson junction, which consists of a nanometer-thick insu-
lating layer sandwiched by superconductors (Fig. 9.4) plays an important role in
the context of quantum technology. Cooper pairs can transmit the insulating film
of the Josephson junction by tunneling effect and exhibit finite conductivity across

Fig. 9.4 Josephson junction

and its circuit diagram
196 9 Various Couplings in Quantum Systems

the insulating film, which is the celebrated Josephson effect. The AC Josephson
effect, in particular, states that the tunneling current I (t) oscillates with respect to
the magnetic flux J piercing the circuit as I (t) = I0 sin (2π J / 0 ) with I0 being
a constant in the unit of current and 0 = h/2e is a flux quantum, where e (< 0) is
the elementary charge. For a back-of-the-envelope derivation of the Josephson effect
can be found in Ref. [15]. In an ordinary inductor with inductance L, the current
through it is proportional to L. In the AC Josephson effect, one can immediately
see that there is no simple proportionality between the current and inductance, and
indeed the Josephson junction is acting as a nonlinear inductor. By replacing the
linear inductor of the LC resonator by such a Josephson junction, one can realize the
anharmonic resonator whose ground state and the first excited state can be utilized
for qubit construction. In this Section, we review the quantum-mechanical treatment
of an LC resonator, an anharmonic resonator described above which is also called a
transmon, and coupling between an LC resonator and a transmon.

9.3.1 Quantum-Mechanical Treatment of an LC Resonator

Let us consider an LC resonator with an inductor L and a capacitor C connected

in parallel,
√ as shown in Fig. 9.5. The resonance frequency is, of course, given by
1/2π LC. The purpose of this chapter is to quantize the LC resonator and to obtain
a Hamiltonian describing it as a harmonic oscillator. The charge accumulated in the
capacitor is here denoted by Q and magnetic flux, or simply “flux” below, in the
resonator by . These quantities are related to voltage V and current I as

Q = CV , = LI. (9.3.1)

The voltage V and current I are further related to the time derivatives of the charge
Q and magnetic flux , respectively as

dQ d
I =− , V = . (9.3.2)
dt dt
By combining these equalities, a differential equation for Q can be obtained as

d2 Q Q
2
=− (9.3.3)
dt LC

Fig. 9.5 LC resonator (left)

and transmon qubit (right)
9.3 Superconducting Circuits 197
√
which actually yields solutions oscillating at the frequency 1/2π LC. In the same
manner, the flux obeys a differential equation

d2
=− . (9.3.4)
dt 2 LC
At this stage, we would like to think about Lagrangian that yields above differential
equations as Euler-Lagrange equations. Such a Lagrangian L is
 2
Q2 L dQ
L= − (9.3.5)
2C 2 dt

and by noting the apparent fact that the canonical conjugate variable of the charge
Q is the flux , Hamiltonian HLC reads

Q2 2
HLC = + (9.3.6)
2C 2L
where one can see that the first term stands for the energy stored in the capacitor,
and the second term does for the energy in the inductor.
Here we succeeded in expressing the LC resonator by the Hamiltonian of a har-
monic oscillator with canonical variables Q and . Then, we require a commutation
relation [Q, ] = i to complete the quantization of the LC resonator. The annihi-
lation and creation operators of a photon in the LC resonator are defined as
 
1 L i C
a= Q+ (9.3.7)
2 C 2 L
 
1 L i C
a =
†
Q− (9.3.8)
2 C 2 L

that fulfills the commutation relation [a, a † ] = 1 and the Hamiltonian is rewritten as
 
1
HLC = ωLC a † a + . (9.3.9)
2

9.3.2 Superconducting Quantum Bit

Next, let us consider the circuit shown in the right panel of Fig. 9.5 in which the
inductor is replaced by a Josephson junction. The Josephson effect tells us that current
I and voltage V across the Josephson junction can be written by the phase difference
θ of the macroscopic wavefunction across the junction as

 dθ
I = I0 sin θ, V = . (9.3.10)
2e dt
198 9 Various Couplings in Quantum Systems

By using the flux quantum 0 = h/2e and defining the flux as J = (θ/2π ) 0 , we
can rewrite above expressions as
 
J d J
I (t) = I0 sin 2π , V = . (9.3.11)
0 dt

Then by noting that the current is given by the time derivative of the charge, we get
 
dQ dV J
=C = −I0 sin 2π (9.3.12)
dt dt 0

and further by using V = d J /dt we obtain

 
d2 J J
C = −I0 sin 2π (9.3.13)
dt 2 0

as a differential equation for J . Lagrangian Lq that yields such an Euler-Lagrange

equation is
 2  
C d J I0 0 J
Lq = + cos 2π . (9.3.14)
2 dt 2π 0

The first term is immediately reinterpreted as the energy of the capacitor Q 2 /2C and
the second term is interpreted as the energy involving Cooper pair tunneling across
the junction. The prefactor E J = I0 0 /2π of the second term is called Josephson
energy. We further define charge energy E C = e2 /2C and number of Cooper pairs
n C = Q/2e to obtain a Hamiltonian Hq of the circuit under concern as

Hq = 4E C n C
2
+ E J cos θ (9.3.15)

where the argument of the second term is written by θ again. If we expand the
cosinusoidal function as cos θ = 1 − θ 2 /2! + θ 4 /4! − · · · and truncate the series to
keep terms up to θ 2 , we see that this Hamiltonian reduces to a one describing mere
harmonic oscillator.
Here we shall make an assumption: we suppose that the Josephson energy is
much larger than the charge energy, namely E J /E C  1. The resonance circuit has
originally become a nonlinear or anharmonic oscillator by the incorporation of the
Josephson junction, a nonlinear inductor. However, above assumption implies that
the nonlinearity is so small that physical quantities of the circuit can be approximated
by the ones of the harmonic oscillator, which turns out to be not so bad as we shall see
below. In such an approximation, the commutation relation [Q, J ] = i is required
to hold as was done in the quantization procedure of the LC resonator. [n C , θ ] = i is
9.3 Superconducting Circuits 199

immediately derived from this commutation relation and we define the annihilation
and creation operators a and a † as follows:
 1
EJ 4 a + a†
nC = √ , (9.3.16)
8E C 2
 1
8E C 4 a − a†
θ= √ . (9.3.17)
EJ i 2

Since we assumed that the nonlinearity caused by the Josephson junction is small, it
is reasonable to put cos θ  1 − θ 2 /2! + θ 4 /4! by truncating the Taylor expansion.
The reason that we keep the term proportional to θ 4 is that this term is the lowest-
order one which brings the nonlinearity into the circuit. Then by plugging n C and θ
into the Hamiltonian we finally get

EC † †
Hq = ( 8E C E J − E C )a † a − a a aa
2
αq † †
= ωq a † a + a a aa (9.3.18)
2
which manifests itself as a harmonic oscillator-like resonator in the first term while a
weakly nonlinear one in the second term. Therefore, the energy levels of this system
are energetically not equidistantly aligned and we can only pick
√ up the lowermost two
levels to construct a qubit. The qubit energy reads ωq = 8E C E J − E C and the
quantity αq = −E C / represents the magnitude of the nonlinearity since the energy
difference between the first excited state and the second one is given by ωq + αq .
As typical parameters, ωq  10 GHz, E J /E C ≥ 50, αq ∼ −200 MHz.
We here make a brief comment on the assumption E J /E C  1 made in the above
discussion. The statement that the Josephson energy E J is much larger than the charge
energy E C means that the nonlinearity of the circuit is very weak, as implied in the
coefficient of θ including a factor (E J /E C )1/4 . One can acquire deeper intuition
by considering the vacuum fluctuation of the charge operator δn C = (E J /8E C )1/4 .
When the E J /E C is small, δn C is also small and a tiny difference of n C , even by
a single quantum, has non-negligible effect on the state and energy of the circuit.
Therefore, the nonlinearity is understood to be large. In contrast to this, δn C is very
large if E J /E C  1 and a small change of n C does not affect the state and energy
of the circuit, meaning that the circuit is almost linear. The superconducting circuit
considered above where E J /E C  1 holds is called a transmon. On the contrary,
if E J /E C  1 do not hold, the nonlinearity gets large and such a circuit is called a
Cooper-pair box.

9.3.3 Coupling Between a Transmon and an LC Resonator

Let us consider here the situation that the transmon and the LC resonator is electrically
connected by a capacitor as depicted in Fig. 9.6. Parameters such as voltages, charges
200 9 Various Couplings in Quantum Systems

Fig. 9.6 Coupled system of

a transmon and an LC
resonator

in capacitors and fluxes are defined as indicated in the figure. A key point is how to
treat the coupling capacitor with capacitance Cc . Here we assume that the coupling
capacitance is negligibly small compared to Cq and Cr , so that the Hamiltonian
can be given by the sum of energies of the transmon, the LC resonator and the
coupling capacitor. The voltage across the coupling capacitor is Vc = Vr − Vq =
Q r /Cr − Q q /Cq and the energy of the coupling capacitor is given by Cc Vc2 /2 using
Vc . Then, the Hamiltonian of the whole system reads

   
Q r2 Cc 2
r
Q q2 Cc
H= 1+ + + 1+ + E J cos θ − Cc Vr Vq
2Cr Cr 2L 2Cq Cq
Q r2 2 Q q2
= + r + + E J cos θ − Cc Vr Vq . (9.3.19)
2Cr 2L 2Cq

Here Cξ = Cξ Cc /(Cξ + Cc ) (ξ = r , c) are defined and the expression of the last

term is intended to be in terms of the voltages. Let us now replace the charges and
fluxes by the annihilation and creation operators of the LC resonator (c† , c) and the
transmon (a † , a), namely by
 
Cr  L
Qr =  (c + c† ), r = (c − c† ), (9.3.20)
L i Cr
 1  1
E J 4 a + a† 8E C 4 a − a †
Qq = √ , θ= √ . (9.3.21)
8E C 2 EJ i 2

By ignoring the constants, we obtain

αq † †
H = ωLC c† c + ωq a † a + a a aa − g(c + c† )(a + a † ) (9.3.22)
2
where coupling strength g is defined as
    1 
Cc  Cr 2e EJ 4 1
g= √ . (9.3.23)
 Cr L Cq 8E C 2
9.4 Optomechanical Interaction 201

This coupling constant is proportional to the coupling capacitance Cc , the vacuum

fluctuation of the voltage of the LC resonator, and the vacuum fluctuation of the
voltage of the transmon. Further application of rotating-wave approximation to this
Hamiltonian yields the Jaynes-Cummings Hamiltonian.
In realistic designs of the superconducting circuits and their couplings, tiny capac-
itor, inductors, and Josephson junctions should be characterized by numerical sim-
ulations and empirical values. This is because the capacitance and inductance held
by wirings other than the capacitor and inductor are not negligible. The original
assumption of Cc Cq , Cr is not the case as well in most situations, hence the
numerical simulations are indispensable at last when one deals with the supercon-
ducting circuits.

9.4 Optomechanical Interaction

9.4.1 Brief Introduction

A gravitational-wave detector consists of a number of mirrors that form an astonish-

ingly large Michelson interferometer extending over kilometers to pickup an unimag-
inably tiny amount of “noise” created in the spacetime by gravitational waves. The
mirrors form extremely high-finesse, coupled optical resonators in order to enhance
the sensitivity of the detector to such noises. Enough disgustingly, the detector should
sense the deviation of the cavity lengths with a precision well below the lattice con-
stants of solids!—and hence, the vibrational motion of the mirrors and even of the
collective motion of atoms on the surface of mirrors, both are described universally
as phonons, may cause fatal noises on the detector that hide the faint signals. To avoid
such unwanted noises, the mirrors are kept in a vacuum, cooled down to a cryogenic
temperature, and suspended by thin wires. Being highly isolated from the environ-
ment, the mirrors embrace the vibrational modes with extremely high-quality factors.
The effect of these vibrational modes on optical signals was studied by Braginsky,
focusing on the quantum-limited measurement of the position of the mirror.
In the series of this kind of study, the cavity optomechanics [30] raised the cry. It
anticipates the interaction between cavity photons and mechanical phonons, where
photons can control phonons and vice versa. Phonons scatter the cavity photons
inelastically to make optical sidebands, which are spectrally discriminated from the
cavity spectrum when the mechanical frequency is much larger than the linewidth of
the cavity. Researchers have searched for high-frequency mechanical modes in the
micromechanical devices to deal with nicely resolved sidebands. Resolved-sideband
regime allows the cooling rate to be maximized to even achieve ground-state cooling
of the mechanical modes where the phonons can be manipulated in the quantum
regime.
Here we introduce the optomechanical interaction and its linearization. These play
essential roles in the rapidly growing field of cavity optomechanics that enchants the
quantum optical systems with novel degrees of freedom.
202 9 Various Couplings in Quantum Systems

Fig. 9.7 Schematics of the system of cavity optomechanics

9.4.2 Optomechanical Interaction

We first dictate the basic Hamiltonian that only includes the energies of cavity photons
(ωc a † a) and mechanical phonons (ωm b† b);

H = ωc (x)a † a + ωm b† b (9.4.1)

in which we assume that the resonant frequency of the cavity ωc (x) is dependent on
the position x of the mechanical oscillator. This assumption applies to the situation
that the cavity length is modulated by the oscillatory motion of some part of the cavity,
e.g., that one of the mirrors which form a Fabry-Perót resonator is ideally fixed and
the other is supported by a spring (see Fig. 9.7). There we can expand the resonant
frequency as ωc (x) = ωc − Gx + · · · with the structure-dependent coefficient G to
get the “perturbed” Hamiltonian

H = (ωc − Gx)a † a + ωm b† b (9.4.2)

= ωc a a + ωm b b − Ga ax
† † †
(9.4.3)
= ωc a a + ωm b b − g0 a a(b + b )
† † † †
(9.4.4)

within the linear regime in the expansion of ωc (x). Here we utilized the fact
√ that the
mechanical displacement is written as x = xzpf (b + b† ) where xzpf = /2mωm
represents the amplitude of the zero-point fluctuation. The effective mass m of the
mechanical oscillator is used in this expression. The coupling strength is thus written
as g0 = xzpf G. This Hamiltonian is widely applicable to the cavity optomechanical
systems including Fabry-Perót type cavities with moving mirrors, membrane-in-the-
middle system, photonic-phononic crystal cavities, and even the electromechanical
systems where the LC resonator involving a trembling capacitor.
Let us look at the form of the optomechanical interaction Hint . It is a coupling
term between photon number in the cavity and the mechanical displacement, hence
its static nature is the radiation pressure, i.e. the mechanical object is pushed by the
momentum kicks exerted by the cavity photons. Since the radiation pressure pushes
back the mechanical object slightly to make the equilibrium position shift, it is also
called the optical spring effect. Its dynamical nature results in the situation that the
9.4 Optomechanical Interaction 203

phonon is either created or annihilated in the virtual absorption and emission of

the cavity photon to result in the Stokes and anti-Stokes scattering. These inelastic
scatterings are frequently called the red- and blue-sideband transitions with the terms
“red” and “blue” refering to the detuning relative to the carrier (phonon-number
conserving) transition. The highlight of the cavity optomechanics is the cooling of
the mechanical modes down close to its motional ground state by imposing the strong
imbalance between red- and blue-sideband transitions.

9.4.3 Linearized Optomechanical Interaction

We would like to see the linearized versions of the optomechanical interaction. We

for brevity ride on a rotating frame of the laser drive frequency ωl . The first term is
transformed to (ωc − ωl )a † a = c a † a while other two remain unchanged. We
consider the circumstance that the cavity mode is driven by the laser with sufficiently
narrow linewidth, so that the cavity operator a can be replaced by α + δa, where
α ∈ C is the amplitude of the coherent state and δa is the noise or fluctuation operator
superposed on the amplitude. By doing this, we have
 
Hint = −g0 |α|2 + α ∗ δa + αδa † + δa † δa (b + b† ) (9.4.5)
√
= −g0 n cav (b + b† ) − g0 n cav (δa + δa † )(b + b† ) (9.4.6)

by neglecting the term in second order of the fluctuation and defining n cav = |α|2 .
√
The coupling strength often include this driving amplitude to be g = g0 n cav . The
first term gives the static shift of the resonant frequency of the cavity, that is, the
radiation pressure and dropped in the following discussion. The second term repre-
sents the linearized optomechanical interaction and takes two forms depending on
the frequency of the driving laser. It might be helpful to explicitly write down the
time-dependent factors as

Hint = −g(δae−ic t + δa † eic t )(be−iωm t + b† eiωm t ) (9.4.7)


−g(δab† + δa † b) when c = ωm ⇒ ω = ωc − ωm
= (9.4.8)
−g(δa † b† + δab) when c = −ωm ⇒ ω = ωc + ωm .

Apparently these are read as the beam-splitter and two-mode-squeezing interactions.

The intuitive understanding is straightforward: when the driving laser is red-detuned
(ωl = ωc − ωm ), the driving photon at ωc − ωm is annihilated, the “noise” at ωc
is created and the phonon is annihilated, see Fig. 9.8a. The “noise” at ωc − 2ωm
is also present, however, it is not nicely supported by the cavity. This asymmetry
of the inelastic scattering make the phonon mode to be cooled, even down to the
motional ground state as detailed in Appendix H. The reverse process takes place
when the driving laser is resonant on the cavity. On the other hand, if the driving
laser is blue-detuned (ωl = ωc + ωm ), the simultaneous creation or annihilation of
the “noise” at ωc and the phonon occurs with entanglement between them [see
204 9 Various Couplings in Quantum Systems

Fig. 9.8 a Beam-splitter and b entangling interactions

Fig. 9.8b]. Due to the presence of the density of states of the cavity, phonon creation
process overwhelms the annihilation one, leading to the motional heating. As the
phonon number increases, the nonlinear nature of the phonon mode (b† b† bb term)
gets relevant and the phonon-phonon interaction has non-negligible effect. Further
phonon pumping results in the observation of phonon lasing.

9.5 Hybrid Quantum Systems and Cooperativity

In the previous section, an interesting “hybrid” quantum system composed of a har-

monic oscillator and an optical cavity is introduced. Such hybrid quantum systems,
where different quantum systems are somehow coupled to each other, provide many
kinds of interesting applications based on the conversion of the quantum state from
one to the other. Then the characteristic quantity should be defined for the hybrid
quantum system representing how efficient the conversion of the quantum state is.
This can be given by the quantity called cooperativity, which we shall introduce in
the following.
First, we shall consider a hybrid system where two harmonic oscillators A and
B are coupled with a beam-splitter interaction as schematically shown in Fig. 9.9.
Frequencies of the harmonic oscillators are denoted as ωa and ωb , and a (a † ) and b

Fig. 9.9 Schematics of a coupled harmonic oscillator

9.5 Hybrid Quantum Systems and Cooperativity 205

(b† ) are annihilation (creation) operators. Intrinsic loss rates γa and γb are present
for the systems A and B, respectively. Propagating-mode operators ain (aout ) and bin
(bout ) represent the input and output waves for the corresponding harmonic oscillator,
with external coupling rate κa for A and κb for B. Further with the coupling strength
g of the beam-splitter interaction, we can write down the full Hamiltonian of this
system as

H = ωa a † a + ωb b† b + g(a † b + ab† ).

What we want to find out with this model system is how efficiently the input waves
from ain can be transmitted to bout by quantifying the ratio of the photon flux
† †
bout bout /ain ain . Therefore, we have no input from bin so that this propagating-
mode operator is neglected hereafter.
Let us then think of driving the system A with frequency ωd . Riding on the
rotating frame with this frequency  for both systems A and B by applying unitary
transformation with U (t) = exp iωd a † at + iωd b† bt , we get

H = a a † a + b b† b + g(a † b + ab† )

where a = ωa − ωd and b = ωb − ωd . Heisenberg equations of motion are then

considered next. For our purpose here, we take into account the input with ain in the
Heisenberg equations of motion using input-output theory:
 
da a √
= − ia + a − igb + κa ain , (9.5.1)
dt 2
 
db b
= − ib + b − iga. (9.5.2)
dt 2

Here the total loss rates of the systems are defined as a = γa + κa for the system
A and b = γb + κb for B. Given these and input-output relations
√
aout = ain + κa a,
√
bout = κb b,

we analyze the behavior of the hybrid system. We consider the situation with
da/dt = 0 and db/dt = 0 here, which corresponds to the steady state of the driven
system. Then above equations result in an easy-going set of linear algebraic equa-
tions. Solving these yields
√
ig κa κb
bout = −    ain
a b
ia + 2 ib + 2 + g2
206 9 Various Couplings in Quantum Systems

Fig. 9.10 Conversion

efficiency as a function of
the cooperativity C

and becomes simpler if we suppose the system is driven resonantly: a = b = 0.

† †
With this assumption, the quantum conversion efficiency η = bout bout /ain ain  is
then written as
4C
η = ηa ηb
(1 + C)2

with newly introduced parameters being given as ηa = κa / a , ηb = κb / b and

4g 2
C=
a b

which is called cooperativity.

Figure 9.10 displays the conversion efficiency as a function of the cooperativity,
when ηa , ηb ∼ 1 which means that the external coupling rates are dominant over
the intrinsic loss rates for both systems A and B. One can immediately see that the
conversion efficiency reaches unity only if C = 1. To see what this situation looks
like, we revisit the Heisenberg equations of motion (9.5.1) and (9.5.2) where b in
Eq. (9.5.1) is now to be rewritten by a. For brevity we put a = b = 0, γa = γb = 0
and db/dt = 0 and get
 
da κa 2g 2 √
= − ia + − a + κa ain (9.5.3)
dt 2 κb

where the loss term of A, originally κa /2, is now modified as κa /2 − 2g 2 /κb . The
unity cooperativity C = 1 in this situation means that the original loss rate of A,
κa /2, is equal to the newly added loss rate due to the coupling, 2g 2 /κb , where a kind
of impedance matching from ain to bout is fulfilled. In real experiments, neither of
γa nor γb equals to zero but finite. Hence, ηa , ηb < 1 always makes the conversion
efficiency less than 1 even if C = 1 is fulfilled. It is important for achieving high
conversion efficiency to get large coupling strength and to tune external coupling
rates at will to make them dominant over the intrinsic loss rates and at the same time
to make C = 1.
9.5 Hybrid Quantum Systems and Cooperativity 207

Problems

Problem 9-1 Red- and blue-sideband transitions of trapped ion and cavity optome-
chanical systems allow us to manipulate mechanical oscillations.

(i) Explain what happens with these transitions in terms of the

ion’s internal state.
(ii) Continuously driving the red-sideband transition leads to the
cooling of ion’s motion. Verify this statement qualitatively.
(iii) Address above two issues in the system of cavity optomechan-
ics.

Problem 9-2 Let us consider a system of coupled Harmonic oscillators A and B

with annihilation operators a and b, respectively. The Hamiltonian
is given as

H = ωa † a + ωb† b + g(a † b + ab† )

where two oscillators are coupled via beam-splitter interaction. Diag-

onalize this Hamiltonian and examine the energy spectrum.
Problem 9-3 Calculate the transition frequencies of the transmon qubit from the
ground state to the first excited state and from the first excited state to
the second excited state. Given α/2π = 100 MHz, how long should
the transmon qubit’s lifetime be for above two transitions spectrally
resolve?
Problem 9-4 Consider a coupled system of three harmonic oscillators as shown
below (Fig. 9.11):
Calculate the quantum conversion efficiency of this system.

Fig. 9.11 A schematics and parameters of three harmonic oscillators coupled in series
 Part III
Quantum Information Processing and
Quantum Technologies
Basics of Quantum Information
Processing 10

10.1 Quantum Gates

As mentioned before, an applied electromagnetic field can rotate the qubit state in
a Bloch sphere and the rotation angle is at our will, by tuning the pulse area of the
applied field. However, this is not everything we should do. In quantum technologies,
one will frequently desire to make any single-qubit state in the Bloch sphere, or more
generally any multi-qubit quantum states which may include heavily entangled states.
For the sake of such an attempt, fortunately, we only have to prepare a few quantum
operations, called the quantum gates, to realize any quantum states. This set of
quantum gates is termed a universal set of quantum gates and is not unique [18]. For
example, single-qubit unitary gates and CNOT gate, which will be described later,
establish a universal set. In this Section, we introduce several frequently used single-
and two-qubit gates.
One thing should be noted before we proceed. In the community of quantum
information processing, the convention about the density matrix is different from
what we have adopted so far. In short, the density matrix reads
 2 ∗ 
|cg | ce cg
ρ= , (10.1.1)
ce cg∗ |ce |2

where the ordering of |g and |e are exchanged. Only in this chapter, we rather
switch to this convention in order for the expression of the quantum gates to be in
common forms seen in other literatures.

Supplementary Information The online version contains supplementary material available at

https://doi.org/10.1007/978-981-19-4641-7_10.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 211
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7_10
212 10 Basics of Quantum Information Processing

10.1.1 Single-Qubit Gates

A quantum state of a single qubit is represented by a point in a Bloch sphere, therefore

any single-qubit gate is a product of a rotation matrix that maps one point on the
Bloch sphere to another, and a global phase shift. The former is given by using a unit
vector n = (n x , n y , n z ) as

θ θ θ
Rn (θ) = e−i 2 (n x σx +n y σ y +n z σz ) = I cos − i(n x σx + n y σ y + n z σz ) sin
2 2
 
cos 2θ − in z sin 2θ −(n y + in x ) sin 2θ
= (10.1.2)
(n y − in x ) sin 2θ cos 2θ + in z sin 2θ

Here, the unit vector n act as an axis about which the rotation by an angle θ takes
place.1 We shall describe special cases where the rotation axes are simply x or z.
Indeed the rotation about y-axis can be decomposed into a product of rotations about
X and Z axes.2

X Gate
First single-qubit quantum gate we consider is the X gate which is defined as
 
0 −i
Rx (π) = = −i X = −i(|eg| + |ge|) (10.1.3)
−i 0

and simply flips the qubit from |g to |e and |e to |g. The X gate often refers to
the operation X = i Rx (π), but in a practical experiment the π-rotation about the x
axis executes the operation −i X . Therefore, it can be easily noticed that the resonant
Rabi oscillation with pulse duration Δt = π/  realizes this operation. This applied
pulse is often called the π-pulse since the acquired phase is π. In the Bloch sphere,
X gate corresponds to the π-rotation around the x axis.

Z Gate
Next we consider the Z gate, which is defined by
 
1 0
Rz (π) = −i = −i Z = −i(|gg| − |ee|). (10.1.4)
0 −1

This operation imposes π-phase shift on |e but leave |g unaffected. This operation
is understood as a π-rotation around z axis, and Z gate often refers to the operation
Z = i Rz (π) as well. Then what procedure is needed to implement the rotation around
z axis? Before answering this question let us remind that the Bloch vector “stops” at
some point in a Bloch sphere only when we are in the rotating frame at the frequency

1n is indeed equivalent to s.
2Y = Rz ( π2 )X Rz (− π2 ).
10.1 Quantum Gates 213

ωq . Namely, the phase of the qubit is at any time evolving in other frames and one
answer to above question comes up with our mind: just wait for time duration π/ωq .
However, for ωq ranging from a few tens of MHz to hundreds of THz, it is tough
to wait precisely for a half-period of the oscillation of the light field. Another way
is to shift the phase convention of the overall experimental sequence by π, which
is relatively easier and quicker. Nonetheless, it is still difficult to achieve ultimately
stable phase of the electromagnetic field especially in the optical domain. In such a
case, active irradiation of the electromagnetic field might be another way. Looking
at the vector causing Rabi oscillation (, 0, −q ), one might drive qubit by an off-
resonant drive with   q for a duration π/q to get a π-phase shift, though this
inevitably leads to errors such as X and Y with tiny but finite possibilities.

Hadamard Gate H, Phase Gate S and π/8 Gate T

Although not being simple rotations, we list three single-qubit gates of frequent use.
The first is the Hadamard gate
 
1 1 1 X+Z
H=√ = √ . (10.1.5)
2 1 −1 2
√ √
Hadamard gate transforms |g → (|g + |e)/ 2 and √ |e → √(|g − |e)/ 2, that
corresponds to π-rotation around the vector n = (1/ 2, 0, 1/ 2). Hadamard gate
satisfies H † = H and transforms X and Z to each other, namely H X H = Z and
H Z H = X.
The other two are related to the Z gate:
  π
10 π
S= = ei 4 R z , (10.1.6)
0i 2
  π
1 0 π
T = iπ = ei 8 R z (10.1.7)
0e4 4

where S 2 = Z , T 2 = S are satisfied. These are two specific cases of the phase gate
which gives a phase factor eiφ to |e and leave |g unchanged. By using the S gate,
the Y gate is decomposed into Y = S X S † .

10.1.2 Two-Qubit Gates

Two-qubit gate is in general a quantum gate acting on a two-qubit system |ψ1  ⊗ |ψ2 
consisting of the first |ψ1  and the second |ψ2  qubits.3 For operations like X 1 ⊗ X 2
by which X i act on i-th qubit, it is simply a simultaneous action of two single-
qubit operations. Another interesting two-qubit operation is a controlled gate where

3 “⊗” is a tensor product.

214 10 Basics of Quantum Information Processing

the “target” qubit (say, the second qubit) undergoes some single-qubit gate if the
“control” qubit (say, the first qubit) is |e. We shall focus on this kind of gates here.

Controlled-NOT (CNOT) Gate

Controlled-NOT gate, or C X gate, is a gate that acts as the X gate on the target qubit
for the control-qubit state |e. The matrix form of this gate is, by using 22 by 22
matrix,
⎛ ⎞
1000
⎜0 1 0 0 ⎟
CNOT = ⎜ ⎟
⎝0 0 0 1⎠ = |gg| ⊗ I + |ee| ⊗ X , (10.1.8)
0010

or more concretely, the operations are the replacements |gg → |gg, |ge → |ge,
|eg → |ee and |ee → |eg.

Controlled-Z Gate
Controlled-Z gate, or C Z gate, is a gate that acts as the Z gate on the target qubit if
the control-qubit state is |e. The matrix form of this gate is
⎛ ⎞
100 0
⎜0 1 0 0 ⎟
CZ = ⎜ ⎟
⎝0 0 1 0 ⎠ = |gg| ⊗ I + |ee| ⊗ Z . (10.1.9)
0 0 0 −1

iSWAP and SWAP Gates

iSWAP and SWAP gates are defined as follows:
⎛ ⎞
1000
⎜0 0 i 0⎟
iSWAP = ⎜ ⎟
⎝0 i 0 0⎠ , (10.1.10)
0001
⎛ ⎞
1000
⎜0 0 1 0⎟
SWAP = ⎜ ⎟
⎝0 1 0 0⎠ . (10.1.11)
0001

It can be easily seen that SWAP gate exchanges the state between the two qubit,
while the iSWAP gate does this with phase factor of i.
10.1 Quantum Gates 215

X X and Z Z Gates
X X and Z Z gates are defined as follows:
⎛ ⎞
cos 2θ 0 0 −i sin 2θ
⎜ 0 cos 2θ −i sin 2θ 0 ⎟
X X (θ) = ⎜ ⎟ = I ⊗ I cos θ − i X ⊗ X sin θ ,
⎝ 0 −i sin 2θ cos 2θ 0 ⎠ 2 2
θ θ
−i sin 2 0 0 cos 2
(10.1.12)
⎛ iθ ⎞
e 2 0 0 0
⎜ −i 2θ ⎟ θ θ
⎜ 0 e 0 0 ⎟
Z Z (θ) = ⎜ θ ⎟ = I ⊗ I cos + i Z ⊗ Z sin . (10.1.13)
⎝ 0 0 e−i 2 0 ⎠ 2 2
θ
0 0 0 ei 2

These are also called Ising coupling gates since they mimic the time evolution of the
system generated by the Ising interaction.

From Hamiltonian to Gate Operation

Every physical system implementing qubits possesses its own ways of interac-
tion between qubits. Then a question arises: how above-listed two-qubit gates
are realized with given interaction Hamiltonian Hint ? To address this question,
let us remind of the fact that the quantum system evolves according to the mas-
ter equation, or Schrödinger equation if we neglect the loss terms. Formally the
Schrödinger equation in the interaction picture4 i(∂/∂t)|ψ = Hint |ψ can be
solved as |ψ(t) = exp [−i(Hint /)t] |ψ(0). Therefore, in the following argument
what we will do is really simple: put Hint in the exponent to clarify the action of
exp [−i(Hint /)t] on target qubit(s).
First, we shall consider “incomplete” but very common, beam-splitter-like qubit-
qubit interaction

g
Hint = g(σ+ σ− + σ− σ+ ) = (σx σx + σ y σ y ). (10.1.14)
2
Here the terms like σ+ σ− and σx σx should be written as σ+ ⊗ σ− and σx ⊗ σx
rigorously, indicating that the first operator acts on the first qubit and the second to
the other qubit. We shall omit the symbol ⊗ whenever there is no risk of confusion.
We will follow the whole maths for this first example. Let us see the explicit form
of (σx σx + σ y σ y )/2:

4 Dubbed as the Tomonaga-Schwinger equation.

216 10 Basics of Quantum Information Processing
⎛ ⎞
⎞ ⎛ ⎞⎛
01 0 −1 00
σx σx + σ y σ y 1 ⎜ O 1 0⎟ ⎜ O 1 0 ⎟ ⎜ O 1 0⎟
⎟+ 1⎜
= ⎜ ⎟=⎜ ⎟ ≡ A.
2 2 ⎝0 1 O ⎠ 2 ⎝ 0 1 O ⎠ ⎝0 1 O ⎠
10 −1 0 00
(10.1.15)

Simple algebra shows that A2 = diag(0, 1, 1, 0) and A3 = A. Now the

exp [−i(Hint /)t] = exp [−ig At] reads

(gt)2 2 (gt)3 3 (gt)4 4

e−ig At = 1 − igt A − A +i A + A + ··· (10.1.16)
2! 3! 4!
(gt)2 2 (gt)3 (gt)4 2
= 1 − igt A − A +i A+ A + ··· (10.1.17)
2! 3! 4!
(gt)2 (gt)4
= diag(1, 0, 0, 1) + 1 − + − ··· diag(0, 1, 1, 0) (10.1.18)
2! 4!

(gt)3 (gt)5
− i (gt) − + − ··· A
3! 5!
(10.1.19)
⎛ ⎞
1 0 0 0
⎜0 cos gt −i sin gt 0⎟
⎜
=⎝ ⎟. (10.1.20)
0 −i sin gt cos gt 0⎠
0 0 0 1

With this, the states |gg and |ee are unaffected but “Rabi oscillation” between |ge
and |eg occurs. By setting t = 3π/2g, this reads the iSWAP gate:
⎛ ⎞
1 0 0 0
⎜0 0 i 0⎟
=⎜ ⎟.
3π
e−i 2 A
⎝0 (10.1.21)
i 0 0⎠
0 0 0 1

Next we consider the two-mode-squeezing-like, counter-rotating interaction

g
Hint = g(σ+ σ+ + σ− σ− ) = (σx σx − σ y σ y ). (10.1.22)
2
Almost the same procedure applies to this and we get
⎛ ⎞
cos gt 0 0 −i sin gt
(σx σx −σ y σ y ) ⎜ 0 1 0 0 ⎟
e−ig 2 t
=⎜
⎝
⎟,
⎠ (10.1.23)
0 0 1 0
−i sin gt 0 0 cos gt

that leads to the oscillation in the subspace {|gg, |ee}. As a special case, this oper-
ation with t = 3π/2g is called the bSWAP gate.
10.1 Quantum Gates 217

Let us consider a simultaneous implementation of above two interactions. This

can be done in practice in quantum systems such as trapped ions and superconducting
qubits. The Hamiltonian reads

Hint = g(σ+ σ− + σ− σ+ ) + g(σ+ σ+ + σ− σ− ) = gσx σx . (10.1.24)

 
This can be immediately exponentiated by using the formula5 exp −iσξ σξ α =
cos αI ⊗ I − i sin ασξ ⊗ σξ
⎛ ⎞
cos gt 0 0 −i sin gt
⎜ 0 cos gt −i sin gt 0 ⎟
e−igσx σx t =⎜
⎝
⎟
⎠ (10.1.25)
0 −i sin gt cos gt 0
−i sin gt 0 0 cos gt

which is identified as an X X gate. Likewise the Z Z gate can also be obtained, which
originates in the state-dependent phase shifts of the Z Z interaction.
The “complete” Ising interaction between qubits is the Heisenberg interaction
dictated as
σ x σ x + σ y σ y + σz σz
Hint = g . (10.1.26)
2

This Hamiltonian gives the following time evolution6 :

⎛ g ⎞
e−i 2 t 0 0 0
⎜ g g g g
e−i 2 t +e3i 2 t e−i 2 t −e3i 2 t
⎟
σx σx +σ y σ y +σz σz ⎜ 0 0 ⎟
e −ig 2 t
=⎜
⎜ g 2 g g 2 g ⎟
⎟ (10.1.27)
⎝ 0 e−i 2 t −e3i 2 t e−i 2 t +e3i 2 t
2 2 0 ⎠
g
0 0 0 e−i 2 t

which gives the SWAP gate when t = π/2g, up to a global phase on two qubits.
⎛ ⎞
1000
σ σ +σ σ +σ σ π
−ig x x y2 y z z 2g π ⎜0 0 1 0⎟
e = e−i 4 ⎜ ⎟
⎝0 1 0 0⎠ . (10.1.28)
0001

The readers might notice that the two-qubit gates such as CNOT and CZ gate—the
controlled gate—do not appear as “natural” quantum gates of frequently used inter-
action Hamiltonians. Usually, these controlled gates are implemented by sequential
application of the above-given two-qubit gates in combination with the single-qubit
gates.

5 You can easily check this!

6 Derivation is not as straight as before. See Appendix G.
218 10 Basics of Quantum Information Processing

10.1.3 Clifford and Non-Clifford Gates

We shall now mention briefly about a more general aspect of the quantum gates. First
we introduce the n-qubit Pauli group7 Pn acting on an n-qubit state by
π
Pn = { ei 2  1 ⊗ · · · n |  = 0, 1, 2, 3; i = Ii , X i , Yi , Z i } (10.1.29)

where the subscript i indicates that the operator acts on i-th qubit. Elements in Pn
are tensor products of n single-qubit gates with overall factors 1, i, −1 or −i. Given
this Pauli group, a Clifford group Cn is defined as a set of unitary operations that
transforms an element in a Pauli group into an element in a Pauli group. If we write
all n-qubit unitaries by Un , this definition takes the form

Cn = { V ∈ Un | V Pn V † = Pn }. (10.1.30)

For example, from H X H = Z , H Z H = X and S X S † = Y , the Hadamard gate

H and the phase gate S are the Clifford gates. There are 6 × 4 = 24 elements in C1
corresponding to how we assign x and z axes on three-dimensional space. How about
the two-qubit gates? For instance, we can see that CNOT(X ⊗ I )CNOT = X ⊗ X
and CNOT(I ⊗ Z )CNOT = Z ⊗ Z . These imply the CNOT gate is also a Clifford
gate, and actually it is [18]! Two-qubit gates such as CNOT and CZ gates are the
Clifford gates, indeed. What is the number of elements in C2 ? Counting up the
number of Clifford gates is not so simple as in the case of C1 . Indeed, there are
11 520 elements in C2 and in general, Cn has superexponentially larger number of
elements as n increases.
Unfortunately, the celebrated Gottesman-Knill theorem states that, starting with
an eigenstate of a Pauli operator, the quantum computation only with the Clifford
gates is simulatable in polynomial time with the probabilistic classical computer.
Furthermore, it is known that Clifford gates alone cannot produce arbitrary quantum
states. Therefore, for the aim of realizing universal quantum computation that might
exceed the classical computer in certain tasks, non-Clifford quantum gates are indis-
pensable.8 Fortunately, a non-Clifford gate is already at our hand—the T gate. T
gate is combined with H gate to form a gate T H T H which has a rotation angle of
θT H T H = 2 arccos [cos2 (π/8)] that is an irrational number times π. This, unlike the
rational number, can cover SU(2) densely and Solovay-Kitaev theorem ensures that
this coverage can be done efficiently, that is, for any given rotation angles one might
reach it with polynomial steps with sufficient precision. Reference [18] provides
more information about above theorems.

7 Group is here a mathemtical concept. Readers unfamiliar to this concept may refer to some

literature on group theory or google it to catch a flavor.

8 With a magic state |ψ  = cos θ |g + sin θ |e as an initial state, one can execute the universal
m 8 8
quantum computation only using Clifford gates.
10.2 Quantum Circuit Model 219

10.2 Quantum Circuit Model

In the previous section, we took an overview of quantum gates and physical imple-
mentation of them using time evolution of interaction Hamiltonians. The quantum
operations are described by operators acting on state vectors; however, it will be
messy when the system scales up, that is, the number of quantum operations and
controlled quantum bits increase. Thence a graphical method of representing a chunk
of quantum operations is desired to facilitate the design of protocols for quantum
technologies. Such a method is called quantum circuit in which quantum gates and
quantum measurements are represented as boxes operating on the quantum state
represented by lines incoming from the left, see Fig. 10.1. A line outgoing on the
right of the quantum operation represents a resultant state, which is a quantum state
for the quantum gate and a classical bit state for the quantum measurement, usually
written as a double line.
Let us coin a few concrete elements of the quantum circuit. A single-qubit gate
is depicted by a box and lines attached to it, with a character indicating the gate
operation such as X , Y , Z , H , S, T , Rx (θ), Rz (θ), and so on. The circuit depicted
in Fig. 10.1a yields the output state X |ψ. Since the state flows from left to right to
sequentially experience the quantum gates, the quantum circuit in Fig. 10.1b results
in the state Z X |ψ.
Next, let us take a look at a two-qubit gate, especially a Controlled-NOT gate,
a CNOT gate, as an example. The CNOT gate deals with two-qubit state as an
input and an output state as well. In this perspective, a way to represent it is as the
one displayed in the left panel of Fig. 10.1c. Another way to represent the CNOT
gate is to regard it as an X gate acting on the target qubit conditionally with the
quantum states of the control qubit. This representation is visualized in the middle of
Fig. 10.1c. The rightmost representation is the most frequently used one, mimicking
the Boolean algebra to imply that the CNOT operation conditionally executes the
flip-flop between |g = |0 and |e = |1.
The last element is the quantum measurement, whose description will be given in
the next section. The quantum-circuit representation of the quantum measurement
is shown in Fig. 10.1. It consists of an icon indicating the measurement and two

Fig. 10.1 Quantum circuits.

a A single-qubit gate. b
Sequential application of two
quantum gates. c A
two-qubit gate
220 10 Basics of Quantum Information Processing

Fig. 10.2 Quantum-circuit

representation of a quantum
measurement

lines attached to it, one on the left side representing a qubit measured and the other
indicating the classical bit information obtained by the measurement. The double line
on the right is often hidden unless it is utilized for controlling other qubits (Fig. 10.2).

10.3 Measurement and Imperfections of Quantum States

10.3.1 Projective and Generalized Measurement

Measurement of a quantum state affects the measured state significantly, in general.

This is the principal reason why we devote one Section to the quantum measurement,
however, it is also because the proper formalism9 of the quantum measurement can
describe the quantum states after the measurement, by which the quantum error
correction has been established [18]. 
Suppose we have a quantum state |ψ = i ci |ϕi  (not necessarily being a qubit
here) that is subject to the measurement. The projective measurement is the one
that extracts the probability pi = |ci |2 of being in a basis state |ϕi  and leave the
measured state in that basis. Let the projection operator be Pi = |ϕi ϕi |. Projective
measurement makes the state |ψ → Pi |ψ = ci |ϕi  without normalization, and with
the normalization we have the state vector and the density matrix

Pi |ψ
|ψ → √ , (10.3.1)
pi
Pi |ψψ|Pi
ρ = |ψψ| → (10.3.2)
pi

after the projective measurement. So far so good for a pure state. For a more general
situaiton, generalized measurement should be introduced. In order to do that, Kraus
operators are introduced first. Aside from the rigorous definition of the Kraus oper-
ators, it is fair to say that the Kraus operators represent how the measured quantum
states end up by the measurement. For example, since the projective measurement
leaves the quantum state as it is measured, Mg = |gg| and Me = |ee| constitute
the Kraus operators. Here it can be seen that the number of Kraus operator equals the
number of bases of the measured system, therefore two Kraus operators Mi (i = g, e)
are assigned to the projective measurement of the qubit. The density operator after a
quantum measurement with the measurement outcome |i becomes Mi ρMi† and by

9 Currently thought so.

10.3 Measurement and Imperfections of Quantum States 221

normalizing it with the probability pi that we find the qubit in the state |i, we have

Mi ρMi†
ρ→ (10.3.3)
pi

Another important concept is positive operator-valued measure (POVM). POVM

are operators that are defined using Kraus operators as E i = Mi† Mi that satisfies

 †
E
i i = I . The probability pi is written by using POVM as pi = Tr M i ρM i =
Tr [E i ρ]. POVM coincides with the projection operators if the measurement does
not include errors. Passing the details to the literature like Ref. [18], we would
like to describe a simple example of measurement with errors. WIth the projection
measurement, one might encounter two types of errors. One is the error that we
missed the signal and the other is the one that we have counted a noise as a signal.
Here we assume these two errors occur with the same probablity . Then the Kraus
operators read
√ √
Mg = 1 − |gg| + |ee|, (10.3.4)
√ √
Me = 1 − |ee| + |gg|. (10.3.5)

We shall make a brief comment on a quantum non-demolition (QND) measure-

ment. QND measurement refers to a projective measurement that preserves the prob-
ability distribution of the measured system. In the quantum engineering community,
one often measures σz of a target system using a probe system interacting with it. If
the interaction is of type Z Z = σz ⊗ σz , it does not affect sz of the target state and
the QND measurement is possible. In the community, QND measurement almost
always refers to the measurement of σz using the Z Z interaction.

10.3.2 State Tomography

As mentioned previously, the density matrix of a single-qubit system is given in

the form ρ = (1/2)(I + sx σx + s y σ y + sz σz ). All we have to do in identifying the
quantum state is to determine the coefficients sx , s y and sz . Using the fact that the
Pauli matrices σξ and thus their product σξ σζ (ξ = ζ) are traceless and σξ2 = I , we
 
can obtain the spin components sξ by using the equality sξ = 2Tr ρσξ . Then, how
this can be done in actual experiments? Firstly we note that in most experiment,
what we extract from the measured signal about the quantum state is the population,
or in other words sz . Other components sx and s y cannot be obtained from this
measurement unless the measurement basis is changed. This can be done rather by
rotating the Bloch vector of a qubit in order for the desired x or y component to
align in z direction. For sx measurement, first the gate R y (π/2)
√ is applied to the
√ so that the transformations |+ = (|g + |e)/ 2 → |g and |− =
quantum state
(|g − |e)/ 2 → |e take place. That is, the x component of the Bloch vector is
transformed into the z component and now the sx is measurable with the measurement
222 10 Basics of Quantum Information Processing
√
√ Rx (π/2) gate makes | + i = (|g + i|e)/ 2 → |g and
of z component. Similarly,
| − i = (|g − i|e)/ 2 → |e, so that the s y measurement becomes feasible by
successive sz measurement. In this manner, the quantity s = (sx , s y , sz ) is determined
through multiple attempts and one acquires the complete information of the quantum
state in terms of the density matrix ρ. This procedure is known as the quantum state
tomography.

10.3.3 Fidelity

Once the quantum state is identified by the state tomography, one needs to evaluate
how close it is to the desired state by using some quantity. A frequently used quantity
for such a purpose is the fidelity. Fidelity does not fulfill the mathematical conditions
of distance between quantum states. However, it yields zero for orthogonal state and
1 for identical states, so that the fidelity is frequently used as an intuitive and easy-
to-calculate characteristics.
For general quantum states represented as density matrices ρ and σ, the fidelity
F is defined by

F = Tr [ρσ] . (10.3.6)
 
It can be immediately seen that for identical states we have F = Tr ρ2 = 1. If the
two states are orthogonal, the two states are pure states and orthogonality leads to
F = 0. The inverse is actually true that one gets F = 0 only if the two quantum
states are orthogonal to each other. Suppose that ρ is unknown quantum state and
σ = |ψψ| is known. The fidelity reads

F = Tr [ρ|ψψ|] = Tr [ψ|ρ|ψ] = ψ|ρ|ψ. (10.3.7)

Further simplifying the above expression by assuming ρ as a pure state ρ = |φφ|,

we get F = | ψ|φ |2 and the fidelity is merely a squared value of the inner product
of the two states.
We shall refer to the fidelity evaluated for two completely mixed states. Since the
completely mixed state for a single qubit is I /2, the fidelity reads F = 1/2 by simply
taking a trace of the matrix I ⊗ I /4. The completely mixed state is a state with no
coherence; however, the fidelity outputs a value of 1/2. If we consider a many-qubit
system, the completely mixed state reads I ⊗n /2n and the fidelity for two such states
becomes 2n (1/2n )(1/2n ) = 1/2n .
One might encounter another definition of the fidelity. Fidelity is defined above
as the trace of product of density matrices; however, here we define it as a trace of
square-root of product of density matrices:
 
√ √
f = Tr ρσ ρ . (10.3.8)
10.3 Measurement and Imperfections of Quantum States 223

Byassuming that ρ is unknown

√ and σ = |ψψ|√is known, the fidelity reads f =
√ √
Tr ρ|ψψ| ρ = ψ|ρ|ψTr [|ψψ|] = ψ|ρ|ψ using the cyclic property
of the trace operation. Further assumption of ρ = |φφ| makes f = | ψ|φ |, from
which we can catch a flavor that f is a square-root version of F.

10.3.4 Estimation of Gate Errors

Quantum gates, such as single-qubit gates, can be implemented by applying a pulse

of electromagnetic wave with a designed pulse area. However, there are many factors
that the desired gate deviates from the ideal one. For example, the fluctuation of pulse
duration or pulse height gives rise to the improper rotation. Or, the drifted phase would
result in the improper rotation axis. Therefore, a method for the evaluation of gate
errors must be developed in order to access how good the gates are. In this section,
we describe such methods.

Noise Superoperators
Quantum processes can be described by applying some operator on a quantum state
|ψ only in the case of pure states under consideration. If one wants more general
prescription applicable to the mixed state, operation on the density matrix should be
dealt with. This is done by writing it as

E (ρ) = Ai ρAi† (10.3.9)
i

where Ai describing the physical process is called the Kraus operator which should
satisfy the conservation of probability. For example, the process described by E (ρ) =
X ρX corresponds to the qubit flip process10 with probability 1. In contrast to this
perfect operation, realistic operation always suffers from noisy processes expressed
as operations on a density matrix. This operation  is expressed as a “superoperator”
since the noise process is an operator on a matrix, an operator as well. If a noise
process flips the qubit with probability p, the noise superoperator x of the bit-flip
error is

x ρ = (1 − p)ρ + p X ρX . (10.3.10)

When, on the other hand, the phase of the qubit is rotated by π by accident, the noise
superoperator z of the phase-flip error reads

z ρ = (1 − p)ρ + p Z ρZ . (10.3.11)

10 See X ρX = X |ψψ|X † , which is just a density matrix of X |ψ.

224 10 Basics of Quantum Information Processing

Another typical noise is the bit-phase-flip error that is

 y ρ = (1 − p)ρ + pY ρY . (10.3.12)

The depolarizing error,

p
d ρ = (1 − p)ρ + (X ρX + Y ρY + Z ρZ ), (10.3.13)
3
where all of the bit-, phase-, and bit-phase-flip noise occur, is not practical in most
experiments. Nevertheless, it is occasionally used because of its symmetry that facil-
itates the theoretical treatment.
The spontaneous decay, often dubbed as the amplitude damping in this context,
requires a bit complicated treatment. It is described by

amp ρ = A0 ρA†0 + A1 ρA†1 (10.3.14)

with Kraus operators

   √ 
1√ 0 0 p
A0 = , A1 = . (10.3.15)
0 1− p 0 0

The first Kraus operator A0 represents the loss of population in |e, and A1 implies
it is due to the transition from |e to |g, both of which are simultaneous processes
with probability p.

Quantum Process Tomography

Quantum process tomography is, like the state tomography, a method for the com-
plete characterization of the quantum process. The goal is to find a way to quantify
how well an experimentally implemented gate resembles the ideal one. First we
apply a quantum gate on a system and write this “unknown” quantum operation

as a result of various noises as E (ρ) = i Ai ρAi† . The unknown Kraus operators
on d-dimensional Hilbert
 space can always be decomposed by orthogonal basis of
operators as Ai = m aim E m , m = 1, 2, . . . , d 2 . Plugging this into E (ρ), we can
rewrite it as

E (ρ) = χmn E m ρE n† (10.3.16)
m,n
 ∗ is called a χ-matrix which thoroughly characterizes the
where χmn = i aim ain
unknown quantum operation in terms of a handy set of orthogonal bases, which are
usually the Pauli matrices together with an identity operator.
Experimentally, first the qubit is prepared in some state ρ j chosen out of d 2
linearly independent set of states {ρ j } and the gate operation is executed on it. Then
the qubit state is evaluated by state tomography to obtain an output. When this is all
10.3 Measurement and Imperfections of Quantum States 225


done, the output state can be written as E (ρ j ) = k c jk ρk . Furthermore, since it can

be said that, by the linear independence of ρ j ’s E m ρk E n† = k βmn jk ρk , we have
  
c jk ρk = χmn βmn jk ρk ⇔ c jk = χmn βmn jk . (10.3.17)
k m,n k m,n

Then inverting βmn jk , we achieve


−1
χmn = βmn jk c jk . (10.3.18)
j,k

On the right-hand side, c jk are experimentally available values through the state

tomography and βmn jk is given by E m ρk E n† = k βmn jk ρk assuming that the ρk is
ideal. Therefore, χ-matrix can be evaluated by repetitive attemps for all ρk ’s and the
unknown quantum operation can now be completely determined. For single- and two-
qubit gates the {E i } are respectively {I , X , Y , Z } and {I I , I X , . . . , Y Z , Z Z } hence
the χ-matrices are 4 × 4 for single-qubit gates and 16 × 16 matrices for two-qubit
ones.
There is one thing to note about quantum process tomography. Quantum com-
putation proceeds firstly with state preparation, secondly the gate operation and in
the end the state is measured, and sometimes the measurement result is fed back
into the system by reflecting it to the gate operation. What errors the quantum pro-
cess tomography senses is the total errors throughout these procedures. In other
words, the errors given rise to in the state preparation and measurement, frequently
abbreviated as SPAM error, are not separately calibrated.

Gate Fidelity
Once the gate is characterized by some means such as quantum process tomography,
a measure is needed to quantify how the implemented gate generating E (ρ) is close
to the ideal gate U which results in U ρU † . One such measure seems to be
 
Fg = Tr U ρU † E (ρ)

which reads
 
  
Fg = Tr U ρU E (ρ) = Tr U ρU
† †
χm E m ρE m
†
(10.3.19)
m
 

= Tr U ρ χmm U †
E m ρE m† UU † (10.3.20)
m
 

= Tr ρ χmm U †
E m ρE m† UU † U (10.3.21)
m
226 10 Basics of Quantum Information Processing
 

= Tr ρ χmm U † E m ρE m† U (10.3.22)
m
= Tr [ρ] (10.3.23)

where we defined the noise superoperator  = m χmm L m ρL †m with L m = U † E m .
If the implemented gates were ideal and E (ρ) = U ρU † , the quantity Fg becomes
 
Fg = Tr ρ2 and is 1 for a pure state. In practice the error comes in and even if the
initial state is a pure state, the errors make the final state ends up with a mixed state,
meaning that the quantity Fg degrades to be less than 1.
One thing we should keep in mind is that the gate fidelity depends on the basis
taken in the trace operation. For example, let the desired operation be a phase-flip
Z gate but with small probability p it appears to be a bit-flip X gate. By taking
the trace with√basis set {|g, |e} √ the gate fidelity is 1 − p/2 < 1, however, with
{(|g + i|e)/ 2, (|g − i|e)/ 2} the gate fidelity gets 1! In order to make things
straight, we should define the gate fidelity not as Fg itself but the minimum value of
it evaluated for any pure state |ψ:
 
Fg = min Tr U |ψψ|U † E (|ψψ|) . (10.3.24)
|ψ

This allows us to avoid underestimation of the gate errors. For example, let us check
with the example that the X gate is contaminated by the Z gate with probability p.
Fg is calculated as

Fg = min Tr [X |ψψ|X E (|ψψ|)] (10.3.25)

|ψ
= min Tr [ψ|X [(1 − p)X |ψψ| + p Z |ψψ|Z ] X |ψ] (10.3.26)
|ψ
= (1 − p) + p min Tr [ψ|Y |ψψ|Y |ψ] (10.3.27)
|ψ
=1− p (10.3.28)

so that the gate fidelity is now evaluated to be 1 − p.

Randomized Benchmarking
In the quantum process tomography, gate errors are not discriminated from the SPAM
errors. Another problem is that as the number of qubit grows, the dimension of χ
matrix rapidly grows and becomes hard to implement full process tomography. To
mitigate these problems, a technique called randomized benchmarking is of frequent
use. Randomized benchmarking has several assumptions and procedures that should
be sufficed, however, the merit of this method is that the gate errors can be separately
extracted from the various errors including SPAM errors.
The procedures for the randomized benchmarking are listed below.

• Initialize the qubit(s) to |ψi 

10.4 Essential Idea of Quantum Error Correction 227

• Apply randomly-chosen l ∈ N Clifford gates

• Apply a gate that inverts the above Clifford gates
• Check if the final state |ψ f  is the same as the initial one

As the number of Clifford gates l increases, the error accumulates and the fidelity
of the quantum states11 F = maxall purification of |ψ f  | ψ f |ψi |2 decreases exponen-
tially. Then by fitting this exponential decrease we can extract the gate fidelity in a
manner discriminated from the SPAM errors!
Randomized benchmarking looks such powerful, however, there is of course a
limitation that the assumption made in above argument seems to be not so relevant
in the realistic experiments. The assumptions to be fulfilled are

• All noise superoperators are equally contained in every Clifford gates.

• Average gate fidelity is calculated as if the noise was the depolarizing error.
• Noises are Markovian, namely the noises are memoryless.

These assumptions, especially the first and the second ones, seem ridiculous from the
experimental viewpoint. For instance, for the first point, certain quantum system is
good at performing some gates but not at others. As for the second issue the modeling
of error solely as the depolarizing noise does not seem to be the case, rather the
spontaneous decay and dephasing are more likely to be the actual ones. The noise
process is not proven to be Markovian regarding the third one, though experimentally
it is fair to assume so. Nonetheless, randomized benchmarking is routinely used in
obtaining the gate fidelities. One might also notice that there is always spontaneous
emission that might let coverage of the state space inhomogeneous. This can be
mitigated by adopting a technique called “twirling”, see literature for the detail [19].

10.4 Essential Idea of Quantum Error Correction

Qubits are always subject to the errors as well as the classical bits are. Realistic
applications of quantum technologies may require numerous, say millions of, state
preparations, quantum gates and measurements applied to the system, so that even if
we have quantum operation of a gate fidelity 0.9999, success probability of a series
of 1 000 gate operations is about 0.9, while 100 000 operations make it on the order
of 10−5 ! Therefore, the errors of the quantum manipulations should be corrected
somehow, as is routinely done in the classical information processing. In the classical
information processing, the error correction can be done by preparing multiple copies
of the bits and adopting majority, assuming the low error rate. In contrast, quantum

11 First, any mixed states can be made pure state by attaching another qubit state. Next, for the pure
 
states |ψi  and |ψ f  the fidelity reads F = | ψ f |ψi |2 . Uhlmann’s theorem states that this fidelity
of the quantum states can be estimated by above equation.
228 10 Basics of Quantum Information Processing

error correction is not that straightforward as it should be implemented under the

following conditions:

(i) Errors are continuous in a Bloch sphere

(ii) No-cloning theorem inhibits the simple redundancy
(iii) Measurement destroys the quantum state

Quantum error correction elegantly avoids violating these conditions. In other words,
one can “digitize” the errors and sense them to implement quantum error correc-
tion, without destroying the quantum state. A drawback is the fact that it requires
redundancy by multiple qubits like the classical majority voting,12 and the error
threshold, over which the quantum error correction does not work, is somehow strin-
gent with current technology—in other words, the fidelity of the quantum operation
must be incredibly high. In this Section, we introduce some essential ideas of quan-
tum error correction to just scratch the surface of it. More details are found in the
literature [18,20].

10.4.1 Stabilizer Formalism

Before proceeding to the error correcting codes, it is instructive to take several para-
graphs about the stabilizer formalism. The stabilizer group S is an Abelian, or com-
mutable, subgroup of n-qubit Pauli group Pn such that
 
S = { Si | − I ∈ / S , Si , S j = 0 for all Si , S j ∈ S } ≤ Pn . (10.4.1)
Here “≤” means that the group on the left-hand side is a subgroup of the one on
the right-hand side. Since S ≤ Pn , S ∈ S is an n-qubit Pauli operator possessing the
eigenvalues ±1. A group is characterized by a generators whose products cover the
entire group. The subset made of generators Sg , or a generator itself, of the stabilizer
group is called the stabilizer generator. By construction any generator cannot be
decomposed into a product of other generators. The stabilizer group contains the
stabilizer generators, S = Sg , and the number of generators required to reproduce
S is known to be at most log |S | where |S | denotes the order, the number of elements,
of S . Here comes one of the merits using stabilizer group to describe the quantum
system, namely we expect economic description of quantum states using stabilizer
formalism rather than writing down the bras and kets of many-qubit system explicitly.
Stabilizer operators all commute with each other by definition, hence have com-
mon eigenstates spanning the eigenspace in common as well. The common eigen-
states are stabilizer states that satisfy Si |ψ = |ψ for any Si ∈ S , with the common
eigenvalue +1. The eigenspace spanned by the stabilizer states is said to be the stabi-
lizer subspace Vs stabilized by S . There is another subspace spanned by eigenstates
with the eigenvalue −1, which are denoted by Vs for convenience.

12 Though it is not majority voting in quantum case.

10.4 Essential Idea of Quantum Error Correction 229

Let us see an example with stabilizer group S = {Z 1 · · · Z n , X 1 · · · X n } with

even n.13 The stabilizer subspace is the “cat state”14

|g · · · g + |e · · · e
|cat = √ .
2

As another example, let us omit X 1 · · · X n from the stabilizer generator and let S =
Z 1 · · · Z n , whose stabilizer subspace is {|g · · · g, |e · · · e}. If one of the qubit is
flipped, the states will be in the subspace spanned by {|eg· · ·g, |ge· · ·g, . . .|gg · · · e,
|ge · · · e, |eg · · · e, · · · |ee · · · g, }. Since these are eigenstates of Z 1 · · · Z n with the
eigenvalue −1, we can see that a single-bit-flip error maps the system from Vs into
Vs , and thus this error can be detected by measuring Z 1 · · · Z n . However, we cannot
tell which qubit is flipped only with this measurement. To correct the error, there are
more tricks to develop the quantum error correction codes.
The quantum error correction code stands for the qubits constructed in the sub-
space of the n-qubit Hilbert space, and error correction codes can be neatly described
by the stabilizer formalism: the logical qubit is encoded in the stabilizer subspace and
the error can be sensed by the measurements of the stabilizer generators, by which
the quantum state can be set back to the original one through the proper feedback
operation.

10.4.2 Three-Qubit Repetition Code

There is one more step to go before the complete15 quantum error correction. We
here see the three-qubit repetition code by which the bit-flip error can be detected
and corrected in a manner capable of keeping the coherence of the qubit.
The stabilizer group for the three-qubit repetition code is S = {Z 1 Z 2 , Z 2 Z 3 }
and the stabilizer subspace reads Vs = {|ggg, |eee}. The logical qubit {|0 L , |1 L }
is encoded as |0 L  = |ggg and |1 L  = |eee, whose logical Pauli operators are
X L = X 1 X 2 X 3 and Z L = Z 1 Z 2 Z 3 . Then, we shall see what happens and what is
inferred from the measurements of the stabilizer generators. First, the measurement
results of Z 1 Z 2 and Z 2 Z 3 with the states in the stabilizer subspace Vs always return
+1 because a single Pauli-Z operator returns +1 for |g and −1 for |e, and the sta-
bilizer generators are the product of two single Pauli-Z operators. This is somewhat
trivial from the definition of Vs .
Next we consider the states after the bit-flip error. If the first qubit is subject
to the bit-flip error, the state a|ggg + b|eee is turned into a|egg + b|gee. Then
the measurement of Z 1 Z 2 returns −1 and that of Z 2 Z 3 returns +1. Readers might
notice that Z i Z j judges whether i-th and j-th qubits are in the same state (+1) or
not (−1), justifying such a measurement as the parity measurement. By these two
parity measurements, we know qubits 1 and 2 are different but 2 and 3 are the same.

13 More details of this example and the stabilizer formalism itself is found in Ref. [20].
14 |ξ  ⊗ |ξ  ⊗ · · · ⊗ |ξ  is denotedby |ξ1 ξ2 · · · ξn .
1 2 n
15 Regardless of the feasibility.
230 10 Basics of Quantum Information Processing

Under the assumption that only one bit-flip error occurs, we can tell from this set
of measurements that the first qubit is flipped and to be corrected again. This set
of parity measurements is thus called the syndrome measurement and even if the
bit-flip error occurs on qubit 2 or 3, the syndrome measurement can specify which
qubit is bit-flipped.
The bit-flipped qubit is then flipped to bring the state back again into the stabilizer
subspace, which constitutes the quantum error correction against the bit-flip error.
Let us see the probability of having the correct state of qubit after the error correction.
The probability of having no errors in all three physical qubits is (1 − p)3 given that
the single-qubit error occurs with probability p. If the error is not corrected, this
contains a term linear in p. If the error is corrected, the single-bit-flip event can be
corrected and the probability of having the correct state is (1 − p)3 + 3(1 − p)2 p =
1 − 3 p 2 + 2 p 3 , where the term linear in p vanishes and when p < 1/2 this gives
larger probability than without the error correction.
Note that the three-qubit repetition code can correct the bit-flip error; however,
the phase-flip and the bit-phase-flip errors cannot be addressed in this logical-qubit
encoding. If one have a stabilizer group S = {X 1 X 2 , X 2 X 3 }, then the code space is
{| + ++, | − −−} and the phase-flip error can be addressed in expense of the ability
to correct the bit-flip error. Another important point is that in actual experiment, we
should prepare an auxiliary qubit that is subject to the syndrome measurements.

10.4.3 Nine-Qubit Shor Code

The three-qubit repetition code is capable of correcting only the bit-flip error. Here
we introduce one of the most popular codes, the nine-qubit Shor code named after
Peter Shor invented it, for it affords the complete correction of single-qubit errors.16
In the nine-qubit Shor code, logical qubit is constructed as follows:
(|ggg + |eee)(|ggg + |eee)(|ggg + |eee)
|0 L  = √ , (10.4.2)
2 2
(|ggg − |eee)(|ggg − |eee)(|ggg − |eee)
|1 L  = √ . (10.4.3)
2 2
These are the stabilizer states of the stabilizer group
S ={Z 1 Z 2 , Z 2 Z 3 , Z 4 Z 5 , Z 5 Z 6 , Z 7 Z 8 , Z 8 Z 9 ,
X 1 X 2 X 3 X 4 X 5 X 6 , X 4 X 5 X 6 X 7 X 8 X 9 }

and the logical Pauli operators are

Z L = X 1 X 2 X 3 X 4 X 5 X 6 X 7 X 8 X 9, (10.4.4)
X L = Z1 Z2 Z3 Z4 Z5 Z6 Z7 Z8 Z9. (10.4.5)

16 And it appeared as a first demonstration that the quantum computer can be, in principle, fault-
tolerant.
10.4 Essential Idea of Quantum Error Correction 231
This Shor code looks like a phase-flip code constructed by a bit-flip code. Such a
code concatenation is a straightforward, powerful way to enhance the fault tolerance
of the logical qubit. Actually, syndrome measurements of Z ⊗2 stabilizer generators,
which contains two Z operators, detect the bit-flip errors in one of the three qubits
in the three blocks, and syndrome measurements of two X ⊗6 stabilizer generators
sense the phase-flip errors.
Naive thought leads us to think that in order to execute the quantum operations
fault-tolerantly, that is, in the way the errors are corrected, the required gate fidelity
seems not so high because no two- or more-qubit error out of nine qubits is sufficient
for the error correction to work. However, in scaling up the quantum operation one
might have millions of gates and errors can be distributed among the physical qubits
through, say, CNOT gates and so on. This error propagation should be avoided by
making the probability of the error propagation square of the error probability itself,
to ultimately make the quantum operations fault-tolerant. Proper analysis shows that
there is a threshold theorem [18] for a specific quantum error correction code that
specifies how much error is admissible to implement the fault-tolerant quantum com-
putations, the error threshold. It is known that with the Shor code the error threshold
is on the order of 10−7 , just an incredibly tiny amount of errors are formidable!

10.4.4 Advanced Quantum Error Correction Codes

The nine-qubit Shor code is a relatively simple code that is not so hard to understand,
however, the error threshold of ∼ 10−7 is so stringent that near-term realization is
desperate given that the current best gate error achieved for the single-qubit operation
of a trapped ion is just ∼ 10−6 and a few-orders-of-magnitude larger for the two-qubit
gates. The error correcting code with a moderate error threshold, or that may demand
less physical resources by utilizing the Fock space of a harmonic oscillator is under
intensive investigation to provide less stringent forms of the fault-tolerant quantum
devices. Two representatives of such are the surface code and the Gottesman-Kitaev-
Preskill code.
The surface code is also called as topological code and toric code, in which a
large number of physical qubits in a two-dimensional array span logical qubits. An
attractive point of the surface code is its high error threshold which amounts to ∼0.01.
It also has relation to the condensed matter physics, opening up a new application of
quantum information theory to new avenue of condensed matter theory. A drawback
is the number of physical qubits required for the implementation of the surface code
is no less than 1 000 that imposes the currently stringent scale-up of the quantum
systems. It is important to estimate how the error rate scales with the number of
connected qubits and how the electrical and/or optical wirings and control units
does, which are under intense investigation.
The second example of the advanced quantum error correction code, the GKP
code [22], gathers attention as a way to implement a logical qubit in a Fock space
of a harmonic oscillator. It utilizes quantum states represented by two-dimensional
arrays of points in the phase space. The spanned logical qubit is also called GKP
qubit. A point in a phase space stands for the situation that +∞-dB squeezing is
 232 10 Basics of Quantum Information Processing

available and not realistic. Therefore, distributions with finite width aligning in a
grid are generated to form approximately orthogonal two GKP qubit states. By now
there are some attempts of GKP qubit generation using an oscillatory motion or
phonon of a trapped ion [23] and a microwave cavity coupled to a superconducting
qubit [24].

10.5 DiVincenzo Criteria

In order to assess how a quantum system is suitable for applications in quantum

technologies, criteria have been coined by David P. DiVincenzo which are listed
below [29]:

(i) Scalable, well-characterized qubit

The system should be almost always settled within a two quantum states among
others. Furthermore, a numerous number of this element of qubit should be assem-
bled together, in the scope of realistic applications.17
(ii) Initialization of qubits to a fiducial state
Preparation of a system in a proper subspace spanning qubits is as important as
the precise operations on it. If the system cannot be initialized properly, one might
start the quantum operations conditionally when they get some signal telling the
completed initialization. However, this tends to be probabilistic and the “firing”
events become rarer and rarer as the system size grows.
(iii) Long coherence times18
In the experiment, one should fight against decoherences exerted by surrounding
environment, including experimental equipments. This process literally lets the
quantum state “decohere” and transforms it gradually into undesired state. Quan-
tum gates must be executed within the duration much shorter than this coherence
time and the lifetime and, roughly speaking, the ratio of the gate pulse duration
to the coherence time gives the estimation of the gate fidelity.
(iv) A universal gate set
Universal gate set is given by arbitrary single-qubit gates, necessarily including
at least one non-Clifford gate, and two-qubit gates. Single-qubit gates require
phase-coherent, individual control of physical qubits, while the two-qubit gates
are usually implemented by some coherent interaction between physical qubits.
(v) A qubit-specific measurement capability
A physical qubit can usually be measured by a projective measurement and pro-

17 Sloppy estimation: if a single logical qubit is spanned by thousands of physical qubits and one
desires to factorize a thousand-digit number by Shor’s algorithm executed fault-tolerantly, then
thousands of logical qubits are required and millions of physical qubits should be integrated into
the system.
18 or “decoherence times”.
 10.5 DiVincenzo Criteria 233

jective measurement combined with single-qubit Pauli gates enables the state
tomography.
(vi) Interconversion between stationary and flying qubits and faithful transmis-
sion of flying qubits
This criterion together with former five criteria constitutes the necessary condi-
tions for the construction of quantum networks [25].

Problems

Problem 10-1 Verify Eq. (10.1.2).

Problem 10-2 Show that Y = S X S † holds.
Problem 10-3 Derive the matrix representation of Y Y and Z X gates.
Problem 10-4 Calculate explicitly the resultant two-qubit state by the quantum
circuit shown in Fig. 10.3.
Problem 10-5 What are the Kraus operators for the luminescence measurement,
where one measures whether a photon is detected or not to deter-
mine the two-level system was in its excited or ground states,
respectively? Is the luminescence measurement projective one?
Problem 10-6 In the quantum process tomography, how many matrix elements
do you have to determine for an n-qubit system?
Problem 10-7 In the randomized benchmarking, we should invert the quantum
operations in the last sequence. How many gates are needed for
such an inversion?
Problem 10-8 Discuss how the random errors and coherent errors can be detected
in the randomized benchmarking. The random error refers to the
random fluctuation of the rotation angle of the quantum gate and
coherent one does to a fixed glitch of the rotation angle.
Problem 10-9 What states are generated by the quantum circuits shown in
Fig. 10.4?
Problem 10-10 Show that the quantum circuit in Fig. 10.5 encodes |ψ into the
logical qubit of the Shor code. Furthermore, find out how to decode
the logical state of the Shor code.
234 10 Basics of Quantum Information Processing

Fig. 10.3 Quantum circuit

with a Hadamard gate and a
CNOT gate

Fig. 10.4 Quantum circuits

with different input states

Fig. 10.5 Quantum circuit

for Shor encoding
Quantum Technologies
11

In this chapter, we will introduce some of the applications that can be realized by
the quantum technologies described so far.

11.1 Quantum Computer

By using quantum systems, it is possible to perform certain kinds of computational

tasks at a speed overwhelming the classical computer. A quantum computer is a
device that implements such tasks. Qubits play the role of fundamental information
carriers in quantum computation. As we have seen, the output of the projective
measurement on a qubit is a discrete number, 0 or 1. Although these values tell us
the output of the quantum computation, the qubit is in superposition states during
calculation and has essentially analog information in the form of complex coefficients
of them. In this sense, quantum computation is very different from the conventional
digital computation.
All physical manipulations have finite errors, therefore, errors cannot be avoided
in the actual quantum computers. These of error are not unique to the quantum world
and the conventional digital computers also suffer from them. However, the errors
become a larger issue, especially for quantum computation due to the analog nature
of qubits. Even if the errors are tiny, they accumulate with each operation, and mak-
ing the final output unreliable. In a large-scale quantum computation, which could
be considered as a huge quantum interferometer, a very small error will propagate
and change the output result. To solve this problem, a quantum computation with
error correcting capability, as current digital computations operate under a finite
error, is being considered. This is called error-tolerant (digital) quantum computa-
tion, achieved by correcting the state of the qubit and the errors in each operation
by quantum error correction. When the various quantum operations are performed
with a error rate less than a certain threshold, the quantum error correction enables
the execution of quantum algorithm without being affected by imperfections in the
quantum operations.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 235
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7_11
236 11 Quantum Technologies

The quantum error correction will definitely be a game changer, however, the
hardware demand is still high and would take some time to perform a large-scale
quantum error correction. Mean while analog quantum computation, in which com-
putation is performed under accumulating errors, has also been studied. The analog
quantum computer such as quantum annealing [33] and NISQ (Noisy Intermediate-
Scale Quantum) [34] computer, are capable of sampling problems and evaluation
and the characteristics on of the ground state of some physical system that are dif-
ficult to calculate with the current state-of-the-art classical super computer. These
analog quantum computers, outputs the computation results without constructing a
huge quantum interferometer.

11.1.1 Grover’s Algorithm

We introduce Grover’s quantum search algorithm as an example of the quantum

algorithm.
Consider the problem of determining integer x0 for a function f as

1 x = x0
f (x) = (11.1.1)
0 x = x0 ,

where x = 0, 1, . . . , 2n − 1 is the integer expressed by n bits. We can find x0 by brute

force, substituting all number from 0 for x until function f returns 1. It requires 2n /2
trials on average, and 2n − 1 times in the worst case.
Here we consider the case of n = 2 for simplicity. The Grover’s algorithm searches
x0 with only a single trial. We define operators Û and Ŵ acting on |x as

−|x x = x0
Û |x = exp(i f (x)π)|x = (11.1.2)
|x x = x0


−|0 x = 0
Ŵ |x = (11.1.3)
|x x = 0.

Note that, only Û depends on the function f .

These two operators and the Hadamard gate act on the initial state |0, 0 in turn,
so that
Ĥ ⊗2 Ŵ Ĥ ⊗2 Û Ĥ ⊗2 |0, 0 = |x0 , (11.1.4)
where Ĥ ⊗2 expresses the simultaneous operations of Hadamard gates on both qubits.
We get integer x0 by measuring the final state |x0 . We need only one Û operator to
find x0 , which is smaller than the number of trials that classical calculations needs
to be done. For general n qubit case and more details, please refer to Reference [18].
11.1 Quantum Computer 237

11.1.2 Phase Estimation Algorithm

Controlled Unitary Operations

Suppose we want to find the eigenvalues of a unitary operator Û .1
The controlled unitary operation with a unitary operator Û can be formed as

Ûc = |0c 0|1̂ + |1c 1|Û . (11.1.5)

Consider this controlled unitary operation is performed on a initial state

1
|ψin  = √ (|0c + |1c ) |u k , (11.1.6)
2

where |u k  is the eigenstate of the operator Û and its eigenvalue is u k . The output
state is
1
Ûc |ψin  = √ (|0c + exp(iu k )|1c ) |u k . (11.1.7)
2
The eigenvalue u k is imparted in the phase of the control qubit. We can determine
these eigenvalues u k by measuring the phase of the control qubit with the quantum
state tomography.

Accuracy
Since each measurement on a single qubit returns only “0” or “1”, it is necessary to
obtain the expectation value of many measurements.
Suppose the each trial are independent, the accuracy of phase estimation, φ is
calculated by a binomial distribution as
√
φ ∝ 1/ N , (11.1.8)

where N is the number of the measurement. The number of measurements needs to

be increased by two digits to improve the accuracy by one digit, and an exponentially
large number of measurements is required for the large number of digits of the phase
to be obtained.

Phase Estimation Algorithm

In a phase estimation algorithm, the number of control qubits used determines the
number of digits of the eigenvalue you would like to calculate. The j-th control qubit
j
control the unitary operator Û(2j) to the target qubits. This leads

j 1   1  
Û(2j) √ |0j + |1j |u k  = √ |0j + exp(2 j iu k )|1j |u k . (11.1.9)
2 2

 
1 For example, from the eigenvalues of the time evolution operator Û
t = exp −it Ĥ / , the energy
eigenvalues are obtained. Here, Ĥ is the Hamiltonian of the system.
238 11 Quantum Technologies

Performing these controlled operators to all control qubits, the final state is
obtained as

20 21 2 N −1 1
Û(0) Û(1) . . . Û(N −1) √ N (|00 + |10 ) . . . (|0 N −1 + |1 N −1 ) |u k  =
2
1  
√ (|00 + exp(iu k )|10 ) . . . |0 N −1 + exp(2n iu k )|1 N −1 |u k , (11.1.10)
2N
where N is the number of control qubits. The eigenvalues are written into the phase of
the control bits, digit by digit. There are two methods for extracting the information
of each digit: quantum state tomography and quantum inverse Fourier transform. As
in the previous section, the quantum state tomography method estimates the state
of the qubit that represents each digit and measures the phase. Unlike the previous
section, the number of measurements required for tomography is constant regardless
of the number of digits, since each digit of the eigenvalue expressed as a binary
number can be obtained from the estimation of each qubit.
Quantum Fourier transform is an unitary operation,

N −1
1 
| j → √ exp(2πi jk/N )|k. (11.1.11)
N k=0

Here, | j express the quantum state of N qubit system such as

| j = | j1 , j2 , . . . j N  ( j = j1 2 N −1 + j2 2 N −2 + . . . j N 20 ). (11.1.12)

With a little algebra, this definition of the quantum Fourier transformation leads
to following expression [18],

1    
| j → √ |00 + exp(2πi j/21 )|10 · · · |0 N −1 + exp(2πi j/2 N )|1 N −1 .
2N
(11.1.13)
Comparing this equation and Eq. (11.1.10), we can see the eigenvalue will be
written in the state of the control qubits through the reverse quantum Fourier trans-
formation.
These methods require the preparation of number of qubits enough to obtain the
precision that we want. The algorithm also require number of measurements which
does not grow exponential with respect to the number of digits (qubits) required.

11.1.3 Shor’s Algorithm

An application of the phase estimation algorithm is Shor’s algorithm for prime fac-
torization. Consider an operator that performs the following operation on an input
bitstring x,
Ûshor |x = |yx mod N . (11.1.14)
11.2 Quantum Key Distribution 239

Table 11.1 Encoding for BB84

0 1
A. linear Horizontal Vertical
B. inclined +45 degree -45 degree

This operator is unitary and the order r (> 0) is the minimum integer, which following
equation will satisfy,
r
Ûshor |x = |y r x mod N  = |x mod N . (11.1.15)

The phase estimation algorithm for the operator Ûshor finds the order r and the
result can be used for the prime factorization. For more details, please refer to Ref-
erence [18].

11.2 Quantum Key Distribution

In the quantum mechanics, successive measurements of non-commutable operators

induce the backaction and make outputs probabilistic. For example, when we mea-
sure the |0 state in Z basis, the output is always 0. On the other hand, we add a
measurement with X basis before that, the outputs become 0 and 1 with a probability
of 50 % for each output. Using this feature of the quantum world, we can share the
secret key for detecting a wiretap. This technique is called quantum key distribution
(QKD) [35]. These shared secret keys are useful for secure cryptography. The first
QKD protocol (BB84) was proposed by Charles Benett and Gilles Brassard in 1984
[7].

11.2.1 BB84

In BB84, we use the polarization of photons is used for the encoding in two different
ways expressed in Table 11.1.
A sender creates a set of random numbers and sends a series of randomly encoded
photons to the receiver using method A or method B for each data. A set of random
numbers and their encoding methods should be recorded one by one by the sender.
The receiver detects the polarization of photons with method A or B, which is ran-
domly chosen. The output of the measurements and their basis should be recorded.
If the method on the sender and the receiver is matched, the output on the receiver
must be the same as the sender’s bit. If not, the output of the receiver is completely
random, because the linear polarization state is the superposition of the two inclined
linear polarization states like

1
|H  = √ (| + 45 + | − 45) . (11.2.1)
2
240 11 Quantum Technologies

To check whether there is a wiretap in the communication channel or not, the

receiver randomly picks up some of the measured results and sends the results and
their methods to the sender over a classical insecure channel. The sender checks the
correlation between the sent data and the receiver’s results only for trials where the
methods were identical between them. If someone peeks at the photon to wiretap a
communication, the state of the photon changes according to the result of someone’s
measurement and breaks the correlation between the data. The sender can know by
the correlation whether the photon has been peeked at or not. If the correlations are
perfect, the rest of the data can be shared with secret keys.

11.3 Quantum Sensing

A method of precisely measuring an object using a quantum system is called quan-

tum sensing [36]. In quantum sensing, quantum technologies such as quantum state
manipulation and quantum measurement are used to construct a sensor. The typical
examples of the currently used quantum sensors are the following:

• Color center in a diamond (NV center) [37,38]

• Superconducting quantum interference device [39–41]
• Atomic gas [42–44]
• Cavity optomechanics [45,46]

In many quantum sensing methods, a shift of the resonant frequency of the quan-
tum systems are measured. For example, from the shift of the resonant frequency of
the spin, we can sense the magnetic field. Ramsey interference is one of the most
widely used methods to precisely measure the frequency change using quantum
coherence.

11.3.1 Ramsey Interference

The frequency shift δω induce Z rotation to a qubit and corresponds to the operation
as

δωt
Ûshift = exp σ̂z , (11.3.1)
2
where t is the interaction time and σ̂z is the Pauli operator of the z component.
The Ramsey interferometer consists of successive π/2 X-rotations and the oper-
ator Ûshift between them. The output state of the interferometer is

|ψ = Rx (π/2)Ûshift Rx (π/2)|0, (11.3.2)

and the probability for the 0 state of the output state is

δωt
p0 = |0|ψ|2 = cos2 . (11.3.3)
2
11.3 Quantum Sensing 241

This leads to frequency detection and senses the external field. Note that, the proba-
bility becomes most sensitive for the frequency shift at the condition of δωt/2 = π/4.
Since the measurement of the quantum state outputs only 0 or 1, the measurement
results have statistical fluctuations. The frequency shift proportional to the probability
of the final state and the fluctuation of the measurement at the condition of δωt/2 =
π/4
Nbin 1
ω ∝  p0 ≥ = √ , (11.3.4)
N 2 N
where N is the number of measurements and Nbin is the statistical fluctuation of
the binomial distribution. This precision limit is called a projection noise limit or
shot noise limit [36].

11.3.2 Quantum Sensing with Entanglement

Entanglement can be used to improve the metrological sensitivity. As described

√ in
the previous section, N independent measurements improve sensitivity by N . This
can be achieved by making N measurements in succession, or it can also be achieved
by increasing the number of qubits to N qubits. When these N qubits are independent,
the
√ sensitivity increases by the same amount as simply making N measurements, or
N . However, when the N qubits are not independent, or other words correlated in
some
√ special way it may be possible to make measurements that exceed the sensitivity
of N . In the following, we will discuss such ultra-sensitive measurements.
Here we consider a following entangle state of n qubit system,

1
|ψ = √ (|111 . . . 1 + |000 . . . 0). (11.3.5)
2

When the n qubits are subject to the perturbation causing the same frequency shifts,
the probability of the |00 . . . 0 state is

⊗n nδωt
p00...0 = 0|Rx (π/2)⊗n Ûshift Rx (π/2)⊗n |0 = cos2 . (11.3.6)
2
The sensitivity for frequency measurement is enhanced by n times;

 p00...0 Nbin 1
ω ∝ ≥ = √ . (11.3.7)
n nN 2n N

This shows that for large n, quantum sensing can be more sensitive than sensors
using n independent qubits. This limitation is called Heisenberg limit [36].
242 11 Quantum Technologies

11.4 Quantum Simulation

In theoretical physics, simulation of the physical systems are performed to understand

and predict physical phenomena. In many cases, the simulations are done by digital
computer, but the complexity of the quantum system increases exponentially with
the number of particles because of the superposition nature of the quantum world.
Even a handful of particles require a supercomputer to simulate [58].
Therefore, in order to simulate complex quantum behaviors, we prepare an ideal
quantum system to model the system. We examine the behavior of the model system
instead of the actual system. This type of calculation method is called quantum
simulation [47,48]. Quantum simulations of several models such as the Hubbard
model and the Heisenberg model of spin have been reported experimentally [59,60]
as well as observations of their quantum phase transitions.
There are not only quantum simulations that actually create the corresponding
quantum system, but also quantum simulations that calculate time evolution and
energy eigenstates on the quantum computer. These algorithms, such as quantum
chemical calculations, have been actively studied in recent years as an application
for NISQ computers.

11.5 Quantum Internet

The Internet has become an indispensable part of our daily lives, either directly or
indirectly. Just as the Internet connects our information devices, we can think about
networks that connect quantum devices. When we think of a network that combines
various quantum technologies such as quantum computers, quantum sensors, and
quantum key delivery, it is important to realize a quantum network that allows the
exchange of quantum states as a communication channel. Such a system consisting of
an entire quantum network is called a quantum Internet [52]. In the quantum internet,
more powerful quantum functions will emerge, such as cloud-based quantum compu-
tation [53], quantum secret computation [61], quantum dense coding communication
[62,63], and increased sensitivity of quantum sensors [54,55]. The quantum internet
will require ultimate quantum technologies such as quantum media conversion from
photons to spins, spins to microwaves, etc., [49–51] and quantum error correction.
These technologies are currently only at the proof-of-principle level and will require
the realization of a test bed and a high-performance interface that can accommodate
the different environments and time scales of individual quantum systems.
Because of the loss of communication channel, signal repeaters are installed even
in the classical communication. The repeaters placed at some distance amplify the
optical signal and compensate for the loss. However, the quantum information cannot
be amplified and relayed because of a property in quantum mechanics called no
cloning theorem, which claims that quantum information cannot be copied without
noise. Instead of installing the amplifier, the entanglement swapping [56,57] and
quantum teleportation [4,5] play an important role for the quantum repeaters. By the
11.5 Quantum Internet 243

quantum teleportation, the quantum state can be transferred to a remote place without
loss with an entangled pair and lossy classical channel. Remote entangled pairs will
be prepared by the entanglement swapping technique. The interested readers might
refer to the Appendix I for detailed explanations.
Position and Momentum
Representations A

In the Dirac notation using bras and kets, a quantum state is represented by a ket vector
| as long as it is a pure state. Indeed, this is a more abstract representation of the
quantum state than the wavefunction like (q),1 since the state vector | is defined
as an object that yields the wavefunction (q) in the position representation and
( p) in the momentum representation. Such position and momentum representations
are of frequent use and let us quickly summarize the basic properties here.
First, let the eigenstate of the position operator q̂ be |q, with its eigenvalue being
q:

q̂ |q = q |q . (A.1)

In the same manner, we define the eigenvector | p and eigenvalue p for the momen-
tum operator p̂ as

p̂ | p = p | p . (A.2)

Then the position and momentum representations of the quantum state |, respec-
tively, read

(q) = q | , ( p) =  p | (A.3)

which result in wavefunctions as functions of position and momentum, respectively.

|q (| p) forms a basis set with continuous variable, so that the completeness relations
are dictated as
 
dq |q q| = 1̂, d p | p  p| = 1̂. (A.4)

1 Position and momentum are, respectively, denoted by q and p in this appendix.

© The Editor(s) (if applicable) and The Author(s), under exclusive license 245
to Springer Nature Singapore Pte Ltd. 2022
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7
246 Appendix A: Position and Momentum Representations

Wavefunctions ( p) and (q) can be derived from each other by the Fourier
transformation

1
q | = √ d pei pq/  p | (A.5)
2π

and on the other hand, by sandwiching d p | p  p| = 1̂ by q| and | we obtain

q | = d pq | p  p | . (A.6)

If we compare this with Eq. (A.5), we see that

1
q | p = √ ei pq/ , (A.7)
2π

namely the position representation of the momentum eigenvector | p results in a plane

wave, or equivalently the state with definite momentum. This is a physically trivial
result, however, above equality is useful in many calculations such as in Sect. 6.5.
Unitary Transformation to a Rotating
Frame B

It is often preferable to analyze the dynamics of the system in the “frame” rotating
at the driving frequency. The unitary transformation that let the frame rotating at
the driving frequency ωd is given by U (t) = exp [i(Hs (ωd )/)t] with the system
Hamiltonian. A state vector |ψ is transformed as |φ = U (t) |ψ and the Schrödinger
equation for |φ reads

d|φ dU d|ψ
i = i |ψ + iU (B.1)
dt dt dt
= iU̇ |ψ + U H |ψ (B.2)
= iU̇U |φ + U HU |φ
† †
(B.3)
= (U HU − iU U̇ ) |φ
† †
(B.4)

where an identity 0 = d(UU † )/dt = U̇U † + U U̇ † is used. H is an original Hamil-

tonian and the above equation indicates that the transformed Hamiltonian is given
by

H = U HU † − iU U̇ † . (B.5)

First example
 is asimple harmonic oscillator with H = ωc a † a transformed by
U (t) = exp iωa at which is defined by the Taylor series of the operator. Since the
†

operator in the exponent commutes with the Hamiltonian, so U (t) does. This greatly
simplifies the transformation of the Hamiltonian: U HU † = UU † H = H and thus

H = H − iU U̇ † = ωc a † a − iU (−iωa † a)U † (B.6)

= (ωc − ω)a a †
(B.7)
= c a a.†
(B.8)

© The Editor(s) (if applicable) and The Author(s), under exclusive license 247
to Springer Nature Singapore Pte Ltd. 2022
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7
248 Appendix B: Unitary Transformation to a Rotating Frame
 
In a very similar way, we can transform H = ωa σz /2 by U (t) = exp iω(σz /2)t
to get the one in the rotating frame H = a σz /2 with the detuning being defined
by a = ωa − ω.
Now let us deal with a second example, the Jaynes–Cummings Hamiltonian

ωa
HJC = σz + ωc a † a − ig(σ+ a − a † σ− ). (B.9)
2
†
We can rotate the frame by the unitary transformation
 U (t)
 = †e 
iωσz t/2+iωa at

= U1 (t)U2 (t), where U1 (t) = exp iωσz t/2 and U2 (t) = exp iωa at . Since a’s
and σ’s commute, the first two terms in the Jaynes–Cummings Hamiltonian are
transformed into a σz /2 + c a † a. However, the transformation of the interac-
tion term is not as straightforward as these two terms. To execute the calculation,
Baker–Campbell–Hausdorff formula
1 1
e−S H e S = H + [H , S] + [[H , S], S] + [[[H , S], S], S] + · · · (B.10)
2! 3!
 
is useful. Another useful tools are the commutation relations a † a, a =
 †   
−a, a a, a † = a † and σz , σ± = ±2σ± . Aided by these, let us transform the
interaction part with U2 (t) first and proceed step by step.

U2 (t)(σ+ a − a † σ− )U2† (t) (B.11)

iωa † at −iωa † at
=e (σ+ a − a † σ− )e (B.12)
 
= (σ+ a − a † σ− ) + (σ+ a − a † σ− ), −iωa † at (B.13)
1   
+ (σ+ a − a † σ− ), −iωa † at , −iωa † at + · · · (B.14)
2!
= (σ+ a − a † σ− ) + (−iωt)(σ+ a + a † σ− ) (B.15)
(−iωt)2 (−iωt)3
+ (σ+ a − a † σ− ) + (σ+ a + a † σ− ) + · · · (B.16)
2! 3!
 
(−iωt)2 (−iωt)3
= 1 + (−iωt) + + + · · · σ+ a (B.17)
2! 3!
 
(−iωt)2 (−iωt)3
− 1 − (−iωt) + − + · · · a † σ− (B.18)
2! 3!
= e−iωt σ+ a − eiωt a † σ− . (B.19)

Now these resultant terms are transformed by U1 (t) as follows, and you will observe
that everything will end up with a really nice (or boring) result!

U1 (t)(e−iωt σ+ a − eiωt a † σ− )U1† (t) (B.20)

−iωt −iωσz t/2
=e (e
iωσz t/2
σ+ a − e a σ− )e
iωt †
(B.21)
 
= e−iωt a eiωσz t/2 σ+ e−iωσz t/2 − eiωt a † eiωσz t/2 σ− e−iωσz t/2 . (B.22)
Appendix B: Unitary Transformation to a Rotating Frame 249

Noting that
    
ωσz t 1 ωσz t ωσz t
eiωσz t/2 σ± e−iωσz t/2 = σ± + σ± , −i + σ± , −i , −i
2 2! 2 2
(B.23)
   
1 ωσz t ωσz t ωσz t
+ σ± , −i , −i , −i + ···
3! 2 2 2
(B.24)
1 1
= σ± ∓ (−iωt)σ± + (−iωt)2 σ± ∓ (−iωt)3 σ± + · · ·
2! 3!
(B.25)
= e±iωt σ± , (B.26)

we have the interaction part in the rotating frame as

U1 (t)U2 (t)(−ig)(σ+ a − a † σ− )U2† (t)U1† (t) (B.27)

−iωt
= U1 (t)(−ig)(e σ+ a − e iωt †
a σ− )U1† (t) (B.28)
= −ig(σ+ a − a σ− )†
(B.29)

which is totally the same form as the one before the transformation. Therefore, the
Jaynes–Cummings Hamiltonian in the rotating frame is finally represented as

 a
HJC = σz + c a † a − ig(σ+ a − a † σ− ). (B.30)
2
Extraction of a Two-Level System from
a Three-Level System C

Let us start from a Hamiltonian of a three-level system {|g , |e , |r } where the two
transitions |g ↔ |r  and |e ↔ |r  are considered (Fig. C.1). The Hamiltonian reads

H = ωg |g g| + ωe |e e| + ωr |r  r | (C.1)

+ g1 (|r  g| a1 + |g r | a1† ) + g2 (|r  e| a2 + |e r | a2† ) (C.2)

and by the coherent driving we justify the replacement a1 → α1 e−i(ωr −ωg +δ1 )t and
a2 → α2 e−i(ωr −ωe +δ2 )t .2 On these rotating frames, that is, |g g| at ωr − ωg + δ1
and |e e| at ωr − ωe + δ2 , we get, by denoting gi αi as i (i = 1, 2),
H = δ1 |g g| + δ2 |e e| + 1 (|r  g| + |g r |) + 2 (|r  e| + |e r |)
(C.3)
where results in Appendix B is used and the overall energy-shift ωr I is ignored.
What we want to do is to adiabatically eliminate the electronic excited state |r 
and make a two-level system by the electronic ground states |g and |e whose
lifetimes are long, usually infinitely, but the transition is dipole-forbidden. To do this,
here the Schrieffer–Wolff transformation is executed with S = −(1 /δ1 )(|r  g| −
|g r |) − (2 /δ2 )(|r  e| − |e r |) (see Appendix F). This transformation partially
diagonalizes the Hamiltonian and we have, up to first order in i /δi ,

21 22
He f f =  δ1 + |g g| +  δ2 + |e e| (C.4)
δ1 δ2
1 2 1 2
+ + (|e g| + |g e|). (C.5)
2δ1 2δ2

2 Rigorously the replacement should be accompanied with the quantum fluctuation, like a1 →
(α1 + δa1 )e−i(ωr −ωg +δ1 )t , however, here we neglect the small δa.

© The Editor(s) (if applicable) and The Author(s), under exclusive license 251
to Springer Nature Singapore Pte Ltd. 2022
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7
252 Appendix C: Extraction of a Two-Level System from a Three-Level System

Fig. C.1 A -type

three-level system

Note that this transformation is valid only when3 1 /δ1 , 2 /δ2 1 and the ac
Stark shift of |r  by −(21 /δ1 + 22 /δ2 ) is present but |r  r |-term itself is omitted
since it does nothing4 to the manifold {|g , |e} in this approximation. Redefining
the parameters as
21 22 1 2 1 2
δ1 = δ1 + , δ2 = δ2 + ,  = + ,
δ1 δ2 2δ1 2δ2
we find our three-level system reduces to the driven two-level system:
He f f = δ1 |g g| + δ2 |e e| +  (|e g| + |g e|). (C.6)

Let us see what is going on in the three-level system by analyzing the eigenstates
of the Hamiltonian in the presence of the third level. Its matrix form reads
⎛ ⎞
δ1 1 0
H =  ⎝1 0 2 ⎠ (C.7)
0 2 δ2
and the eigenvalues λ are the solutions of λ(λ − δ1 )(λ − δ2 ) − 22 (λ − δ1 )−
21 (λ − δ2 ) = 0. Suppose the two-photon detuning is 
zero, namely δ1 = δ2 = δ.
Then the eigenvalues are λ = δ ≡ λ0 and λ = (δ/2) ± (δ/2)2 + 21 + 22 ≡ λ± .
The eigenvector |ψ0  corresponding to λ0 is a superposition of |g and |e only, while
those corresponding to λ± , |ψ± , contains the auxiliary state |r  that might cause
decay into the manifold {|g , |e} accompanied with the emission of a photon. For
such reasons, the former state |ψ0  is often dubbed as the dark state and the latter the
bright state.

3 Usually the detunings are taken to be large compared to the driving strengths.
4 Ifthe loss or decoherence of |r  is non-negligible, it does something, e.g., let the quantum gate
incomplete.
Quantum Theory of Hydrogen Atom
D

Since the astonishing experimental results by Rutherford, an atom is recognized to be

an object consisting of electrons orbiting around an atomic nucleus. Nevertheless, the
puzzle had still remained that the electrons in accelerated motion should have radi-
ated electromagnetic fields and lost their velocities to fall into the nucleus, but they
do not in reality. This puzzle had not been solved until the birth of quantum theory
and quantum mechanics. In this Appendix, let us consider a hydrogen atom, the most
simple atomic species, with the help of Bohr’s model. Back-of-the-envelope calcu-
lations for the fine and hyperfine structures are also given here to make this appendix
a quick introduction to quantum systems accompanied with optical transitions.

D.1 Bohr’s Atom

Bohr’s atomic model considers an electron orbiting around a proton as in the clas-
sical picture, except that the electron behaves as a matter wave. It assumes that the
phase acquisition by the electron as a matter wave with one circulation around the
nucleus equals to an integer-multiple of 2π. Then we reinterpret this argument as the
discretization of the orbital angular momentum of the electron that is expressed by

mvr = n (D.1)

where m is the electron mass, v the velocity, and r the radius of the orbit. n is here a
positive integer. In an inertial system that the electron is at rest, the Coulomb force
and the centrifugal force should equilibrate:

v2 1 e2
m = . (D.2)
r 4π 0 r 2

© The Editor(s) (if applicable) and The Author(s), under exclusive license 253
to Springer Nature Singapore Pte Ltd. 2022
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7
254 Appendix D: Quantum Theory of Hydrogen Atom

By using the above two equations, we can eliminate v and get the radius of electron’s
orbit as

4π 0 2 2
r= n (D.3)
me2
which means that the radius of electron’s orbit has to take discrete values. The energy
E of the hydrogen atom can be calculated as follows:

1 2 1 e2
E= mv − (D.4)
2 4π 0 r
1 e 2
=− (D.5)
4π 0 2r
me4 1
=− . (D.6)
2(4π 0 )2 2 n 2

As expected, the energy of Bohr’s hydrogen atom can only take discrete values

me4 1
En = − . (D.7)
2(4π 0 )  n 2
2 2

The energy difference between E 1 and E ∞ is evaluated to be 13.6 eV that explains the
experimentally obtained ionization energy. Moreover, the above formula reproduces
the empirical formula of the atomic spectra found by Rydberg. These facts made
scientists believe in Bohr’s atomic model.
Let us next summarize the quantum mechanical treatment very briefly. First,
the energy of the hydrogen atom E n calculated above coincides with the one
derived from the non-relativistic quantum mechanics. The quantum states are
labeled by three integers (n, l, m) with n = 1, 2, 3, . . ., l = 0, 1, . . . , n − 1, and
m = −l, −(l − 1), . . . , l are respectively called principal, azimuthal, and magnetic
quantum numbers. One should note that the electronic ground state in Bohr’s model
describes an orbiting electron around the nucleus but it is not the case in quan-
tum mechanics; the ground-state electron is not orbiting but rather have probability
distribution with its peak at the nucleus.
The velocity of the electron in an atom is very large that the relativistic effect also
has non-negligible modifications. Dirac equation, instead of Schrödinger equation,
is to be used for taking into account spins, fine structure, and other effects. Here
after we shall admit the electron and nuclear spin to overview the fine and hyperfine
structures.

D.2 Fine Structure

Suppose again that an electron is orbiting around the nucleus. In the rest frame of
the electron, in turn, the nucleus is orbiting around the electron. Given the electric
Appendix D: Quantum Theory of Hydrogen Atom 255

charge of the nucleus be +e, the orbiting nucleus forms a loop current I = ev/2πr
that generates the magnetic field

μ0 I μ0 ev μ0 eL
B= = 2
= (D.8)
2r 4πr 4πmr 3
at the position of the electron, where L = mvr denotes the orbital angular momen-
tum. The Zeeman effect says that the energy shift of the spin states of the electron
by this magnetic field reads

gμB μ0 e gμ B e2
−µ · B = S· L=− S · L. (D.9)
 4πm r2 3 8πm 2 r 3

Here g 2 denotes the Landé g factor and μB = e/2m is called the Bohr magneton.
Indeed, the above discussion neglects the relativistic effect of Thomas precession and
correct expression is given by replacing g by g − 1. If we look at this phenomena
from the original frame, above energy shift can be interpreted as an interaction energy
between the electron spin and the magnetic field generated by the electron’s orbital
angular momentum. This interaction is called the spin–orbit interaction. In such a
situation, L or S are not good quantum numbers alone, but the sum J = S + L is.
For example, if the electron is in the S state, L = 0 so that the spin–orbit interaction
energy is zero. In contrast, if in the P state, the energy level is split into two depending
on whether the L and S are parallel or anti-parallel. This energy splitting is called
the fine-structure splitting and the energy scale of the splitting is ranging from GHz
to THz, depending on the atomic species.

D.3 Hyperfine Structure

The interaction between the spin and orbital degrees of freedom was discussed in the
previous section, and we mentioned that the total angular momentum J becomes a
good quantum number. In this section, we further estimate the amount of interaction
between the electron’s angular momentum J and nuclear spin I.
Let us consider a simple case that the electron is in the S state, where the electron’s
wavefunction exhibits finite value ψ(0) at the position of nucleus. The nucleus thus
feels magnetic field generated by the electron spin that is proportional to |ψ(0)|2 . The
magnetic field generated by the electron spin is here roughly estimated by substituting
the magnetization M = −μB g|ψ(0)|2 J into the magnetic flux of a uniformly mag-
netized sphere B = (2/3)μ0 M. The nuclear spin μ N I interacts With this magnetic
field to yield the Zeeman shift of

2
−µ N · B = μ N μ0 μB g|ψ(0)|2 I · J ≡ AI · J (D.10)
3
that catches the basic concept of hyperfine interaction. Here μ N denotes the nuclear
magnetic moment. Like in the fine structure, a good quantum number is not J or
256 Appendix D: Quantum Theory of Hydrogen Atom

I alone, but the sum F = J + I = S + L + I acts as a good quantum number. The

energy splitting by the values of I · J is called hyperfine splitting. The amount of the
hyperfine splitting is small compared to the fine splitting, for the nuclear magnetic
moment is about 1/1000 times smaller than that of the electron. The hyperfine splitting
is on the order of GHz in most atomic species.5

5 The above discussion takes only the magnetic dipole interaction into account, however, it is known

that the quadrupole interaction also have a sizable effect.

Master Equation
E

E.1 Density Matrix

In quantum mechanics, the physical system is described by operators and state vec-
tors, which respectively represent the physical quantities and the probabilities that
the system is in some state. Suppose that a Hamiltonian of our interest is given. How
do we analyze the system, or in other words, what quantity do we keep track of?
One way is to look at the operators that obey the Heisenberg equations of motion,
which is useful for directly inspecting the evolution of the experimentally observable
physical quantities. Another way is to focus on the state to see that with how much
probability the system evolves into a certain state. For the latter purpose, density
matrix nicely addresses the time evolution of the system including the transition of
one state to another to allow us to analyze the rate equation, optical Bloch equation,
Raman transition and electromagnetically induced transparency, and so forth.
When state vectors {|ψk } of the system are given, the density matrix is constructed
by

ρ= wk |ψk  ψk | (E.1)
k

where the coefficients satisfy k wk = 1. If the density matrix can be written by
a single state vector |ψ, i.e., ρ = |ψ ψ|, the system is said to be in a pure state,
otherwise in a mixed state. The pure and mixed states can be discriminated by the
quantity Tr[ρ2 ] that exhibits a value of 1 for the pure state and less than 1 for the
density matrix using some basis set {|i}. By inserting
mixed state. Let us express the
the completeness condition i |i i|, we get
 
ρ= |i i| ρ | j  j| = |i ρi j  j| . (E.2)
i, j i, j

© The Editor(s) (if applicable) and The Author(s), under exclusive license 257
to Springer Nature Singapore Pte Ltd. 2022
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7
258 Appendix E: Master Equation

The ρi j in the above expression dictates the matrix element of the density matrix
represented in the current basis. The expectation value of the operator A is written
by the density matrix as
 
wk ψk | A |ψk  = wk ψk | A |i i|ψk  (E.3)
k k i

= wk i|ψk  ψk | A |i (E.4)
k i

= i| ρA |i (E.5)
i
= Tr[ρA]. (E.6)

It is instructive to give an example of the density matrix

 witht a two-level system
where a state is represented by |ψ = cg |g + ce |e = ce cg . The density matrix
constituted by this state vector is

ce  ∗ ∗  |c |2 ce cg∗
ρ = |ψ ψ| = ce cg = ∗e . (E.7)
cg ce cg |cg |2

One can immediately see that the diagonal elements of the density matrix express the
populations in the basis states. The off-diagonal elements are called the coherences.
At first sight these are mysterious, however, the off-diagonal elements connect the
different basis states, hence one can guess that they have something to do with the
interference or transition among the states. Actually, the coherence terms play an
essential role in the optical Bloch equations as we describe later.
How about the dynamics of the density matrix? To address this, we depart from
the state vector in the Schrödinger picture, where the time evolution of the state
reads |ψk (t) = e−i(H /)t |ψk  = Ut |ψk  with the system Hamiltonian H . Let us
differentiate the density matrix evolving with ρ(t) = Ut ρUt† :

dρ(t) dUt dU † i i 1
= ρUt† + Ut ρ t = − H ρ(t) + ρ(t)H = [H , ρ(t)] (E.8)
dt dt dt   i
which leads us to the von Neumann equation

dρ
i = [H , ρ] . (E.9)
dt
Note the difference between the von Neumann equation for a state and the Heisenberg
equation of motion for an operator: id A/dt = [A, H ].
Appendix E: Master Equation 259

Basic properties of trace operation

Here for the later use we list the elementary algebraic properties of trace operation.
Most of the following properties can be derived by straightforwardly calculating the
trace of the matrices involved. First, the trace operation is linear:

Tr[A + B] = Tr[A] + Tr[B], (E.10)

Tr[c A] = cTr[A]. (E.11)

The trace is invariant under cyclic permutations of the matrices:

Tr[AB] = Tr[B A], (E.12)

Tr[ABC] = Tr[C AB] = Tr[BC A], (E.13)
··· . (E.14)

The trace of the tensor product of matrices is the product of the traces of the matrices:

Tr[A ⊗ B] = Tr[A]Tr[B]. (E.15)

E.2 System-Bath Interaction

Let us now take a look at how to deal with the system–bath interaction using the
density matrix. The starting point is the same Hamiltonian as the input-output theory
(see Sect. 8.3). The total Hamiltonian Htot = Hs + Hb + Hi is the sum of the ones
of the system Hs , bath Hb and interaction between them Hi that have concrete forms
of

Hs = ωc a † a, (E.16)
 ∞
dω
Hb = ωc† (ω)c(ω), (E.17)
−∞ 2π
 ∞
dω  
Hi = −i f (ω)a † c(ω) − f ∗ (ω)c† (ω)a . (E.18)
−∞ 2π

Note that the system-bath

√ coupling here is now in the general form f (ω), which is
to be replaced by κ in most of the situations. We shall define
 ∞ dω
R− = f (ω)c(ω), (E.19)
−∞ 2π
 ∞
dω ∗
R+ = f (ω)c† (ω) (E.20)
−∞ 2π
260 Appendix E: Master Equation

to simplify the interaction Hamiltonian as

 
Hi = −i a † R − − a R + . (E.21)

Let us shortly make a note on the time evolution of the state and operator in the
interaction representation. We list the equations governing the time evolution of these
quantities without proof:

d|ψ(t)
i = Hi |ψ(t) , (E.22)
dt
dρ(t)
i = [Hi , ρ(t)] , (E.23)
dt
d A(t)
i = [A(t), Hs + Hb ] . (E.24)
dt
The first, second, and third ones are respectively called the Tomonaga–Schwinger,
Liouville–von Neumann, and Heisenberg equations. It can be immediately observed
that in the interaction picture, the time evolution of the states is governed by the
interaction Hamiltonian Hi , while that of the operators by the “unperturbed” Hamil-
tonian Hs + Hb . This formalism was first invented in the context of developing the
perturbation expansion of the quantum electrodynamics.
In the following, we switch to the interaction picture from the Schrödinger picture.
Transformation of the density matrix and the Hamiltonian reads
Hs +Hb Hs +Hb
ρ̃(t) = ei  t
ρ(t)e−i  t
, (E.25)
Hs +Hb Hs +Hb
H̃i = ei t
Hi e−i t  (E.26)

= −i a † eiωc t R̃ − − ae−iωc t R̃ + (E.27)

where we have defined

 ∞
Hs +Hb Hs +Hb dω
R̃ − = ei  t
R − e−i  t
= f (ω)c(ω)e−iωt , (E.28)
−∞ 2π
 ∞
Hs +Hb Hs +Hb dω ∗
R̃ + = ei  t
R + e−i  t
= f (ω)c† (ω)eiωt . (E.29)
−∞ 2π

The Liouville–von Neumann equation in the interaction picture becomes

dρ̃(t)  
i = H̃i (t), ρ̃(t) . (E.30)
dt
In what follows, we will work on the interaction picture so that the tilde is suppressed.
In order to dive deeper into the dynamics of the density matrix in the presence
of the system–bath interaction, we rewrite the Liouville–von Neumann equation by
using infinitesimal time interval Δt as
Appendix E: Master Equation 261

 t+Δt
1  
Δρ(t) = ρ(t + Δt) − ρ(t) = dt  Hi (t  ), ρ(t  ) (E.31)
i t
 t+Δt
1  
= dt  Hi (t  ), ρ(t) (E.32)
i t
  
1 2 t+Δt  t    
+ dt dt Hi (t  ), Hi (t  ), ρ(t)
i t t
(E.33)

where we truncate the infinite series up to the second order.

Since we are here interested in the dynamics only of the system part, we define
the reduced density matrices of the system σ(t) = Tr b [ρ(t)] and complementarily
the one of the bath σb (t) = Tr s [ρ(t)] by tracing out the bath- and system-degrees of
freedom, respectively. At this point we make a crucial assumption that the density
matrix of the whole system can be written as the tensor product of the two reduced
matrices:

ρ(t) = σ(t) ⊗ σb (t) (E.34)

which implies the absence of the temporal correlation between the system and bath.
We further assume that the bath is always stationary, i.e. σb (t) = σb (0) = σb , hence
the σb commutes with the bath Hamiltonian Hb . Now we are to trace out Δρ(t) with
respect to the bath degree of freedom to get the dynamics of the system.

1 t+Δt  
Tr b [Δρ(t)] = dt  Tr b Hi (t  ), ρ(t) (E.35)
i t
2  t+Δt  t  
1
+ dt  dt  Tr b Hi (t  ), Hi (t  ), ρ(t) .
i t t
(E.36)

Here the first term on the right-hand side vanishes, since

   
Tr b Hi (t  ), ρ(t) = Tr b Hi (t  ), σ(t) ⊗ σb (t) (E.37)
      
= −i a † eiωc t Tr b R − (t  )σb − ae−iωc t Tr b R + (t  )σb σ(t) (E.38)
      
+ iσ(t) a † eiωc t Tr b R − (t  )σb − ae−iωc t Tr b R + (t  )σb (E.39)
       
= −i a † eiωc t R − (t  ) b − ae−iωc t R + (t  ) b σ(t) (E.40)
       
+ iσ(t) a † eiωc t R − (t  ) b − ae−iωc t R + (t  ) b (E.41)
= 0, (E.42)
 
where we used the fact that c(ω)b = c† (ω) b = 0. ·b stands for the expectation
value evaluated over the bath degree of freedom, which we shall omit the subscript
below.
262 Appendix E: Master Equation

Then the remaining second term matters. The commutator inside the trace is
expanded to give messy appearance; however, they are traced out to give eight sur-
viving terms
 
Tr b Hi (t  ), Hi (t  ), ρ(t)
= (E.43)
(−i)2
 −  +   †   
− R (t )R (t ) a (t )a(t  )σ(t) − R + (t  )R − (t  ) a(t  )a † (t  )σ(t) (E.44)
   
+ R − (t  )R + (t  ) a(t  )σ(t)a † (t  ) + R + (t  )R − (t  ) a † (t  )σ(t)a(t  ) (E.45)
   
+ R − (t  )R + (t  ) a(t  )σ(t)a † (t  ) + R + (t  )R − (t  ) a † (t  )σ(t)a(t  ) (E.46)
   
− R − (t  )R + (t  ) σ(t)a † (t  )a(t  ) − R + (t  )R − (t  ) σ(t)a(t  )a † (t  ). (E.47)

In this expression, the exponential factor for the cavity operator is absorbed by the
−iωc t
 = ae
 +  and−a (t)
 = a e . The two-time correlation
operator itself, † † iωc t
 − namely
 +
a(t)

functions R (t )R (t ) and R (t )R (t ) should be evaluated to eliminate the
bath operator from above expression. These read

 ∞   
  dω ∞ dω    
R − (t  )R + (t  ) = f (ω) f ∗ (ω  ) c(ω)c† (ω  ) e−iωt eiω t (E.48)
−∞ 2π −∞ 2π
 ∞ 
dω ∞ dω    
= f (ω) f ∗ (ω  ) [n(ω) + 1] 2πδ(ω − ω  )e−iωt eiω t
−∞ 2π −∞ 2π
(E.49)
 ∞
dω  
= | f (ω)|2 [n(ω) + 1] e−iω(t −t ) , (E.50)
−∞ 2π
 ∞
 +  −   dω  
R (t )R (t ) = · · · = | f (ω)|2 n(ω) e−iω(t −t ) . (E.51)
−∞ 2π

In integrating with respect to time t  and t  , we transform the domain of integration

by putting τ = t  − t  as
 t+Δt  t  t+Δt  t  −t  Δt  t+Δt
  
dt dt = dt dτ = dτ dt  (E.52)
t t t 0 0 t+τ
   
and further assume that the quantities R − (t  )R + (t  − τ ) and R + (t  − τ )R − (t  )
have non-zero contribution only around τ Δt, which allows for the extension of
the domain of integration:
 Δt  t+Δt  ∞  t+Δt  ∞  t+Δt
dτ dt  = dτ dt  dτ dt  . (E.53)
0 t+τ 0 t+τ 0 t

Note that since the limit Δt → +0 is taken afterwards, this assumption is justified
by the Markov approximation—the dynamics of the bath mode is memoryless. By
Appendix E: Master Equation 263

Fig. E.1 Transforming the

domain of integration

executing these transformations, we arrive at the expressions

  t    ∞
t+Δt     
dt  dt  R − (t  )R + (t  ) eiωc (t −t ) = Δt dτ R − (τ )R + (0) eiωc τ
t t 0
(E.54)
 ∞  ∞
dω
= Δt | f (ω)|2 [n(ω) + 1] e−i(ω−ωc )τ e− τ ,
dτ (E.55)
0 −∞ 2π
 t+Δt  t    ∞
    
dt  dt  R + (t  )R − (t  ) eiωc (t −t ) = Δt dτ R + (0)R − (τ ) eiωc τ
t t 0
(E.56)
 ∞  ∞ dω
= Δt dτ | f (ω)|2 n(ω) e−i(ω−ωc )τ e− τ . (E.57)
0 −∞ 2π

Here the convergence factor e− τ has been introduced in above equations in order
not to make these integrals diverge at ω = ωc . These integrals are further calculated
to yield

 ∞  ∞
dω
Δt dτ | f (ω)|2 [n(ω) + 1] e−i(ω−ωc )τ e− τ (E.58)
0 −∞ 2π
 ∞   ∞ 
dω
=i | f (ω)|2 [n(ω) + 1] −i dτ e−i(ω−ωc )τ e− τ Δt (E.59)
−∞ 2π 0
 ∞
dω 2 Δt
=i | f (ω)| [n(ω) + 1] (E.60)
−∞ 2π (ωc − ω) + i
 ∞
→+0 dω | f (ω)|2 [n(ω) + 1]
−→ iP Δt (E.61)
−∞ 2π ωc − ω
 ∞
1
+ dω| f (ω)|2 [n(ω) + 1] δ(ωc − ω)Δt (E.62)
2 −∞
 + 
= i( +  )Δt + Δt (E.63)
2
264 Appendix E: Master Equation
 ∞  ∞
dω
Δt dτ | f (ω)|2 n(ω) e−i(ω−ωc )τ e− τ (E.64)
0 −∞ 2π
 ∞   ∞ 
dω
=i | f (ω)|2 n(ω) −i dτ e−i(ω−ωc )τ e− τ Δt (E.65)
−∞ 2π 0
 ∞
dω Δt
=i | f (ω)|2 n(ω) (E.66)
−∞ 2π (ω c − ω) + i
  
→+0 ∞ dω | f (ω)|2 n(ω) 1 ∞
−→ iP + dω| f (ω)|2 n(ω) δ(ωc − ω) Δt
−∞ 2π ωc − ω 2 −∞
(E.67)
 

= i + Δt (E.68)
2

where we used the Dirac identity lim →+0 1/ [(ωc − ω) + i ] = P /(ωc − ω) −

iπδ(ωc − ω). The parameters ,  are defined by
 ∞ dω | f (ω)|2
=P , (E.69)
−∞ 2π ωc − ω
 ∞
dω | f (ω)|2 n(ω)
 = P (E.70)
−∞ 2π ωc − ω

both representing the radiative shift. On the other hand,  and   stand for
 ∞
= dω| f (ω)|2 δ(ωc − ω) = | f (ωc )|2 , (E.71)
−∞
 ∞
 = dω| f (ω)|2 n(ω) δ(ωc − ω) = n(ωc )  (E.72)
−∞

which respectively represents the spontaneous and induced decay rates (Fig. E.1).
By using results obtained so far, the quantity dσ(t)/dt = limΔt→+0 Tr b [Δρ(t)] /
Δt reads

∂σ(t) i    +   
= − a † a, σ(t) + 2aσ(t)a † − a † aσ(t) − σ(t)a † a (E.73)
∂t  2
  † 
+ 2a σ(t)a − aa † σ(t) − σ(t)aa †
2
(E.74)

which looks in the Schrödinger picture like

dσ(t) i    +   
= − (ωc + )a † a, σ(t) + 2aσ(t)a † − a † aσ(t) − σ(t)a † a
dt  2
(E.75)

  † 
+ 2a σ(t)a − aa † σ(t) − σ(t)aa †
2
(E.76)
Appendix E: Master Equation 265

This is the celebrated master equation. The master equation is somewhat simplified
by the Lindblad superoperator

L [A] σ(t) = 2 Aσ(t)A† − A† Aσ(t) − σ(t)A† A (E.77)

that lead to
dσ(t) i   1 √  1 √ 
=− (ωc +)a † a, σ(t) + L  +   a σ(t) + L   a † σ(t).
dt  2 2
(E.78)

The first term describes the coherent dynamics of the system itself. Very roughly
speaking, second and third terms have meanings of the decay into and excitation
from the bath mode, respectively.

E.3 Rate Equation

The master equation

dσ(t) i    +   
= − (ωc + )a † a, σ(t) + 2aσ(t)a † − a † aσ(t) − σ(t)a † a
dt  2
(E.79)

  † 
+ 2a σ(t)a − aa † σ(t) − σ(t)aa †
2
(E.80)

describes the dynamical behavior of the system when sandwiched by various basis
states. In this section, we evaluate the master equation with the Fock-state basis {|N }
to derive the rate equation. By sandwiching the master equation by |N , we get
d P(N )  
=  +   [(N + 1)P(N + 1) − N P(N )] +   [N P(N − 1) − (N + 1)P(N )]
dt
(E.81)
= ↓ [(N + 1)P(N + 1) − N P(N )] + ↑ [N P(N − 1) − (N + 1)P(N )] (E.82)

and hence the rate equation

dN  d 
= N P(N ) (E.83)
dt dt
N
  
= N (N + 1)P(N + 1)↓ + N P(N − 1)↑ − (N + 1)P(N )↑ − N P(N )↓ (E.84)
N
 
= ↑ (N + 1)P(N ) − ↓ N P(N ) (E.85)
N N
= ↑ N + 1 − ↓ N  (E.86)
= − N  +   (E.87)
= − N  +  n(ωc ) . (E.88)
266 Appendix E: Master Equation

 ∞
made use of the identities  ∞
Note that we  N =0 N P(N ) =
2
N =0 (N + 1)
2
∞ ∞
P(N + 1) and N =0 N (N + 1)P(N + 1) = N =0 (N − 1)N P(N ) in the third
line.

E.4 Optical Bloch Equation

Next, let us consider the two-level system and sandwich the master equation for it
with the basis states |g and |e. The master equation for the atomic density matrix
ρ(t) is dictated by

dρ(t) i  +   
= − [Hs , ρ(t)] + 2σ− ρ(t)σ+ − σ+ σ− ρ(t) − ρ(t)σ+ σ−
dt  2
(E.89)
  
+ 2σ+ ρ(t)σ− − σ− σ+ ρ(t) − ρ(t)σ− σ+ .
2
(E.90)

The system Hamiltonian is set here to be Hs = (q /2)σz + (/2)(σ+ + σ− ), a

driven atom, for simplicity (see Appendix C). Evaluating the both sides of the master
equation by multiplying |g’s and |e’s, one can obtain a set of equations

dρee 
= −( +   )ρee +   ρgg + i (ρeg − ρge ), (E.91)
dt 2
dρgg 
= ( +   )ρee −   ρgg − i
(ρeg − ρge ), (E.92)
dt 2
dρeg q  + 2  
= −i − ρeg − i (ρgg − ρee ), (E.93)
dt 2 2 2
dρge q  + 2  
= i − ρge + i (ρgg − ρee ) (E.94)
dt 2 2 2

where ρi j = i| ρ | j. We shall make a short note that the atomic decay rate  and
the atomic dipole moment μ are related to each other by
!
μ2 ωa3 3π 0 c3 
= ⇐⇒ μ= . (E.95)
3π 0 c3 ωa3

We are now close to the final form of the equation. We further consider the
situation that the transition energy of the two-level atom is in the optical domain.
With this assumption, almost no photon is found in the thermal bath, since the
optical photon has the energy of ωc ∼ k B × 10 000 K,6 that is far larger than the
bath temperature ∼300 K and make n(ωc ) = 1/(eωc /k B T − 1) 0. Therefore, the

6 Temperature of the surface of the sun is about 6000 K.

Appendix E: Master Equation 267

“bath-stimulated” decay   = n(ωc )  can be safely ignored to make the above set
of equations

dρee 
= −ρee + i (ρeg − ρge ), (E.96)
dt 2
dρgg 
= ρee − i(ρeg − ρge ), (E.97)
dt 2
dρeg  
= −iq − ρeg − i (ρgg − ρee ), (E.98)
dt 2 2
dρge  
= iq − ρge + i (ρgg − ρee ) (E.99)
dt 2 2

which are called the optical Bloch equations. By putting w = ρgg − ρee , the optical
Bloch equations yield

dw
= −w − i(ρeg − ρ∗eg ) + , (E.100)
dt
dρeg  
= −iq − ρeg − i w (E.101)
dt 2 2

with the use of the relations ρgg + ρee = 1 and therefore 2ρee = ρee + (1 −
ρgg ) = −w + . Inspecting the steady-state solution and combining these two, w
and ρee are found to be

1
w= , (E.102)
1+s
1 s (/2)2
ρee = = (E.103)
21+s ˜
(q )2 + (/2)2

with parameters defined by

s0
s= , (E.104)
1 + (2q / )2
22
s0 = , (E.105)
"2
˜ =  1 + s0 . (E.106)

There are two points to be noted: one is that as the driving strength becomes larger,
the population in the ground and excited states approaches 1/2 as a result of the
balance between the stimulated absorption and emission including spontaneous and
stimulated ones. The other point is that the effective linewidth of the transition is
˜ which gets broader as the driving field increases. This phenomenon is
given by ,
called the power broadening.
Schrieffer–Wolff Transformation
F

F.1 General Prescription

In a system of nicely isolated multiple particles, the energy of the system is given by
the sum of the energies of particles involved. For example, when an electromagnetic
field and an atom are not interacting with each other, the energy of the whole system
is written by the sum of the energies of the photon and the atom. However, we always
consider a system of interacting particles which generally possesses eigenenergies
different from the ones in the non-interacting case. If the interaction between particles
is weak enough, perturbation theory allows for obtaining approximated energies and
eigenstates. In such an attempt, Schrieffer–Wolff transformation describes how to
arrive at the diagonalized effective Hamiltonian of the system of weakly interacting
particles.
Here we present the general prescription of how we can derive perturbatively
diagonalized Hamiltonian from the original Hamiltonian H = H0 + λV consisting
of the non-interacting part H0 and the interaction part λV where the parameter λ
is introduced so that the order of the perturbation becomes apparent. Without loss
of generality, H0 is set to be diagonal with the eigenbasis |m of itself, and λV to
be off-diagonal, that is, m| λV |m = 0 for all eigenstates |m. Let us consider the
unitary transformation of this Hamiltonian by eλS , which reads
λ2 λ3
eλS H e−λS = H0 + λV + λ [S, H0 ] + λ2 [S, V ] + [S, [S, H0 ]] + [S, [S, V ]] + · · · . (F.1)
2! 2!

Note that since e S is unitary, S should be anti-Hermitian S † = −S. In order for the
first-order terms in λ vanishes, one should have

[H0 , S] = V (F.2)

© The Editor(s) (if applicable) and The Author(s), under exclusive license 269
to Springer Nature Singapore Pte Ltd. 2022
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7
270 Appendix F: Schrieffer–Wolff Transformation

and the Hamiltonian is perturbatively diagonalized up to the first order as

λ2
He f f = eλS H e−λS = H0 + [S, V ] + O(λ3 ). (F.3)
2!
How to find an appropriate S for such a diagonalization greatly relies on the “educated
guess”, therefore it should be helpful to list some examples below to get the flavor.

F.2 Examples
F.2.1 Driven Spin

First example is a driven spin expressed by a Hamiltonian

H = ωsz + g(s+ + s− ) (F.4)

   
where the spin operators sz , s+ and s− satisfy sz , s± = ±s± and s+ , s− = 2sz . We
are working with the unit where  =  1. The specific
 transformation for the Schrieffer–
Wolff transformation is U = exp α (s+ − s− ) , with which we can exactly diago-
nalize H by correct choice of the parameter α. To find this “correct” value, we shall
transform the Hamiltonian in practice.
   
U HU † = ωsz + g(s+ + s− ) + α (s+ − s− ), ωsz + α (s+ − s− ), g(s+ + s− )
1   
+ α (s+ − s− ), α (s+ − s− ), ωsz
2!
1   
+ α (s+ − s− ), α (s+ − s− ), g(s+ + s− ) + · · ·
2!
  1  
= ωsz + g(s+ + s− ) − α ω(s+ + s− ) − 4gsz − α2 4ωsz + 4g(s+ + s− )
2!
1  
+ α3 4ω(s+ + s− ) − 16gsz + · · ·
3!
 
(2α)2 (2α)3
= sz ω 1 − + · · · + 2g 2α − + ···
2! 3!
 
(2α)2 ω (2α)3
+ (s+ + s− ) g 1 − + ··· − 2α − + ···
2! 2 3!
 ω 
= [ω cos 2α + 2g sin 2α] sz + g cos 2α − sin 2α (s+ + s− )
2

Now remember the motivation of this calculation—to eliminate the off-diagonal

terms. We can accomplish this to arbitrary order in the present case, by setting

2g
tan 2α = (F.5)
ω
Appendix F: Schrieffer–Wolff Transformation 271

so that
2g
1 ω
cos 2α =  , sin 2α =  . (F.6)
4g 2 4g 2
1+ ω2
1+ ω2

This leads to the effective Hamiltonian transformed into a diagonal form

!
4g 2
He f f = ωsz 1 + (F.7)
ω2

This form of the effective Hamiltonian says that the energy difference between the
ground and excited state of the spin system is blue-shifted as the driving strength
increases, which can be regarded as the resonant version of the ac Stark shift.

F.2.2 Generalized “Spin” Subject to the “Driving Field”

Let us consider more general situation that more generalized “spin” is driven;

H = X z + g(X + + X − ) (F.8)
   
where operators are defined to support X z , X ± = ±X ± and X + , X − = P(X z ).
P(X z ) in the second commutation relation is defined as a polynomial function of
X z . Here for the second term to be treated as the perturbation, g/
 1 should hold.
Given these, the unitary transformation is guessed to be U = exp (g/)(X + − X − )
and this leads straightforwardly to the effective Hamiltonian of the form

g2
He f f = X z + P(X z ). (F.9)

which is clearly a diagonal one.
Derivation of the SWAP Gate from the
Heisenberg Hamiltonian G

The starting point is the Heisenberg interaction

σ x σ x + σ y σ y + σz σz
Hint = g . (G.1)
2
Before we exponentiate this, let us calculate D = σx σx + σ y σ y + σz σz and its poly-
nomials D n . First the concrete form of D reads

D = σ x σ x + σ y σ y + σz σz (G.2)
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
01 0 −1 1 0 0 0
⎜ O ⎟ ⎜ O ⎟ ⎜0
1 0 1 0 −1 0 0⎟
=⎜⎝01
⎟+⎜
⎠ ⎝ 0 1
⎟+⎜
⎠ ⎝0
⎟ (G.3)
0 −1 0 ⎠
O O
10 −1 0 0 0 0 1
⎛ ⎞
1 0 0 0
⎜ 0 −1 2 0 ⎟
=⎜⎝ 0 2 −1 0 ⎠
⎟ (G.4)
0 0 0 1

Since D is block-diagonal, we have only to diagonalize the two-by-two matrix in

the middle. Then we have
n  n
−1 2 1 1 1 1 0 1 1 1
= √ √ (G.5)
2 −1 2 1 −1 0 −3 2 1 −1
1 1 1 1 0 1 1 1
=√ √ (G.6)
2 1 −1 0 (−3)n 2 1 −1
1+(−3)n 1−(−3)n
= 2 2
1−(−3)n 1+(−3)n . (G.7)
2 2

© The Editor(s) (if applicable) and The Author(s), under exclusive license 273
to Springer Nature Singapore Pte Ltd. 2022
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7
274 Appendix G: Derivation of the SWAP Gate from the Heisenberg Hamiltonian

∞
Next we calculate the exponential of e−i Dα . It is expanded as n=0 (−i Dα) /n!
n

and
∞
 (−i)n αn
e−i Dα = Dn (G.8)
n!
n=0
⎡ ⎛ ⎞⎤
0 0 0 0
∞
(−i)n αn ⎢ ⎢diag(1, 0, 0, 1) + ⎜
⎜0 1+(−3)n 1−(−3)n 0⎟ ⎥
⎟⎥
= ⎣
2
⎝0 1−(−3)n 1+(−3)n
2 (G.9)
n=0
n! 2 2 0 ⎦
⎠
0 0 0 0
⎛ ⎞
0 0 0 0
−iα −iα
⎜0 e +e 3iα e −e 3iα
0⎟
= diag(e−iα , 0, 0, e−iα ) + ⎜⎝0 e −e
−iα
2
3iα e −iα
2
+e3iα ⎠
⎟ (G.10)
2 2 0
0 0 0 0
⎛ −iα ⎞
e 0 0 0
⎜ e−iα +e3iα e−iα −e3iα ⎟
⎜ 0 0 ⎟
=⎜ −iα
2
−iα
2 ⎟ (G.11)
⎝ 0 e −e e +e
3iα 3iα
2 2 0 ⎠
0 0 0 e−iα

Thus, finally we have

⎛ g ⎞
e−i 2 t 0 0 0
⎜ g g g g
e−i 2 t +e3i 2 t e−i 2 t −e3i 2 t
⎟
σx σx +σ y σ y +σz σz ⎜ 0 0 ⎟
e −ig 2 t
=⎜
⎜ g 2 g g 2 g ⎟.
⎟ (G.12)
⎝ 0 e−i 2 t −e3i 2 t e−i 2 t +e3i 2 t
2 2 0 ⎠
g
0 0 0 e−i 2 t
Cavity Cooling of a Mechanical Mode
H

H.1 Quantum Noise Spectrum and Rate Equation

One fascinating feature of the cavity optomechanical system is the ability to cool
down, i.e., reduce the number of phonons in the mechanical mode under concern.
This is done by employing the beam-splitter-type interaction described above, since
the phonon-annihilating anti-Stokes scattering overwhelms the reverse process with
the help of the cavity. This is often called the “laser cooling” or “sideband cooling”
of mechanical mode in close similarity with those that have been familiar techniques
to cool down the motions of trapped atoms or ions.
To illustrate how it works and how cold the modes can be laser-cooled, we fol-
low the quantum noise approach. The noise spectrum with a quantum-mechanical
quantity is defined as
 ∞
Scc (ω) = dτ eiωτ c(τ )c(0) (H.1)
−∞

mimicking the one with a classical quantity. The angled braces stand for the expec-
tation value. With c(t) being defined by the difference between operator itself and
its expectation value, Scc is actually the noise spectrum. We are here interested in the
optical noise generated by the presence of mechanical motion through the optome-
chanical interaction, and how these noises can be utilized in the cavity cooling.
Our starting point is the photon noise in the interaction part
  
Hint = g0 a † a − a † a (b + b† ) = g0 ν(t)(b + b† ) (H.2)

of the Hamiltonian H = H0 + Hint , where the overall phase of the interaction is

changed so that the minus sign vanishes. Since the optomechanical interaction acts as
a perturbation, things are going to be handy in the interaction picture. Let the phonon

© The Editor(s) (if applicable) and The Author(s), under exclusive license 275
to Springer Nature Singapore Pte Ltd. 2022
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7
276 Appendix H: Cavity Cooling of a Mechanical Mode

number state in the Schrödinger picture be |N  that satisfies b† b|N  = N |N . This is

represented in the interaction picture as |N , t = U I (t)|N  with the unitary operator
satisfying the equation

∂U I (t)
i = Hint
I
(t)U I (t) (H.3)
∂t
where
H0 H0
Hint
I
(t) = ei  t Hint e−i  t (H.4)
−iωm t
= g0 ν(t)(be +b e † iωm t
). (H.5)

A formal solution of this differential equation is approximated as


i t
U I (t) = 1 − dτ Hint
I
(τ )U I (τ ) (H.6)
 0
 t    
i i t  I 
=1− dτ Hint (τ ) 1 −
I
dτ Hint (τ ) 1 + · · · (H.7)
 0  0
 t
i
1− dτ Hint
I
(τ ). (H.8)
 0

We are to evaluate the probability amplitude α N →N ±1 of the event that after a

interaction time t, |N , t evolves to the phonon number state |N ± 1.

α N →N +1 = N + 1 |N , t (H.9)
 t
= N + 1 |N  − ig0 dτ ν(τ ) N + 1| be−iωm τ + b† eiωm τ |N 
0
(H.10)
 t  
= −ig0 dτ ν(τ ) N + 1| b |N  e−iωm τ + N + 1| b† |N  eiωm τ
0
(H.11)

√ t
= −ig N + 1 dτ ν(τ )eiωm τ , (H.12)
0
α N →N −1 = · · · (H.13)
√  t
= −ig N dτ ν(τ )e−iωm τ . (H.14)
0

We then obtain probabilities of finding the system in |N ± 1 after the time evolution
by taking square of α’s and its expectation value:
 t  t
    
PN →N +1 = α∗N →N +1 α N →N +1 = g02 (N + 1) dτ dτ  ν(τ )ν(τ  ) e−iωm (τ −τ )
0 0
(H.15)
Appendix H: Cavity Cooling of a Mechanical Mode 277
 t
= g02 (N + 1) dτ Sνν (−ωm ) (H.16)
0
= g02 (N + 1)t Sνν (−ωm ), (H.17)
 ∗ 
PN →N −1 = α N →N −1 α N →N −1 = · · · (H.18)
= g02 N t Sνν (ωm ). (H.19)

by using the Wiener–Khinchin theorem. Given these formulae, the transition rates
of above two processes are

d PN →N +1
 N →N +1 = = g02 (N + 1)Sνν (−ωm ) = (N + 1)↑ , (H.20)
dt
d PN →N −1
 N →N −1 = = g02 N Sνν (ωm ) = N ↓ . (H.21)
dt
At this point we notice that the quantum noise spectrum is closely related to the
transition rate and thus plays an essential role in the optomechanical interaction.
Here the transition rates

↑ = g02 Sνν (−ωm ) (H.22)

↓ = g02 Sνν (ωm ) (H.23)

are the ones of increasing and decreasing phonon number by 1.

Before deriving the concrete form of Sνν (ω), let us step a bit more into the
dynamics of the phonon number. What we want to do is to see the temporal variation
of the average number of phonons N  = N N P(N ). The rate of change of the
average number of phonon reads

dN  d 
= N P(N ) (H.24)
dt dt
N
  
= N (N + 1)P(N + 1)↓ + N P(N − 1)↑ − (N + 1)P(N )↑ − N P(N )↓
N
(H.25)
 
= ↑ (N + 1)P(N ) − ↓ N P(N ) (H.26)
N N
= ↑ N + 1 − ↓ N  (H.27)
= −(↓ − ↑ ) N  + ↑ (H.28)
= −opt N  + ↑ . (H.29)
 ∞
Note that we  made use of the identities  ∞ N =0 N P(N ) =
2
N =0 (N + 1)
2
∞ ∞
P(N + 1) and N =0 N (N + 1)P(N + 1) = N =0 (N − 1)N P(N ) in the third
line. This is the rate equation for mechanical phonon without any intrinsic loss.
opt = ↓ − ↑ denotes the optical damping, which is essential for the cavity cool-
ing. If we incorporate the intrinsic loss of the mechanical mode m and the effect
278 Appendix H: Cavity Cooling of a Mechanical Mode

of thermal phonons with the average phonon number being n th , the rate equation is
modified as
dN 
= ↑ N + 1 − ↓ N  + N + 1 ↑th − N  ↓th (H.30)
dt
= −(opt + m ) N  + ↑ + n th m (H.31)

where use has been made of ↑th = n th m and ↓th = (n th + 1)m . The final number
of phonons reached after the cavity cooling is thus estimated to be

↑ + n th m
N t→∞ = . (H.32)
opt + m

H.2 Limits on the Cavity Cooling

In order to know how cold you can get the mechanical mode with the cavity cooling,
remaining thing is to evaluate the quantum noise spectrum
 ∞
Sνν (ω) = dτ eiωτ ν(τ )ν(0) (H.33)
−∞
 
with ν(t) = a † a − a † a . Plugging this into ν(τ )ν(0), we have
     
ν(τ )ν(0) = a † (τ )a(τ ) − a † (τ )a(τ ) a † (0)a(0) − a † (0)a(0) (H.34)
 †  † 
= a (τ )a(τ )a (0)a(0) − a (τ )a(τ ) a (0)a(0)
† †
(H.35)
 †  †   †  † 
− a (0)a(0) a (τ )a(τ ) + a (τ )a(τ ) a (0)a(0)
(H.36)
 †    2  2
= a (τ )a(τ )a † (0)a(0) − 2 a † (0)a(0) + a † (0)a(0) (H.37)
 † 
= a (τ )a(τ )a (0)a(0) − n cav .
† 2
(H.38)

Suppose that the photon operator a can be written in a form a(t) = [α + d(t)] e−iωl t
where α denotes the amplitude of the coherent drive and d(t) the fluctuation. By
 † a(t) |ψ(t)
definition, when it acts on a state, = αe−iω

l t |ψ(t) and d(t) |ψ(t) =

ψ(t)| d (t) = 0. In calculating a (τ )a(τ )a (0)a(0) , the only surviving terms are
† †

the ones having d on the leftmost and d † on the rightmost and the one without any
operators. Thus the above expression turns out to be
 
ν(τ )ν(0) = a † (τ )a(τ )a † (0)a(0) − n 2cav (H.39)
 
= α∗ αα∗ α + α∗ α d(τ )d † (0) − n 2cav (H.40)
 
= n cav d(τ )d † (0) (H.41)
Appendix H: Cavity Cooling of a Mechanical Mode 279
 
There appears the term d(τ )d † (0) which seems to be difficult to approach. However,
“quantum regression theorem” [31] allows us to write down the equation for this
quantity by the same form as for d(t), that is,
 
d d(τ )d † (0)   κ 
= −ic d(τ )d † (0) − d(τ )d † (0) for t ≥ 0, (H.42)
dt 2
 
therefore d(τ )d † (0) = exp [[−ic t − (κ/2)t]], where c = ωc − ωl and κ the
loss rate of the cavity. Then the quantum noise spectrum can be explicitly calcu-
lated:
 ∞  κ 
Sνν (ω) = dτ eiωτ exp −ic t − t (H.43)
−∞ 2
κn cav
= . (H.44)
(c − ω)2 + (κ/2)2

In the rate equation, noise spectra appear in the forms Sνν (±ωm ). These are the
“noises” (or in another viewpoint, the “signals”) exposed to the optical mode by the
mechanical mode, each of which has Lorentzian peak at ±ωm .
We immediately get the full expression of the transition rates ↑ and opt using
this noise spectrum, leading to

↑ = g02 Sνν (−ωm ) (H.45)

κ
= g02 n cav (H.46)
(c + ωm )2 + (κ/2)2
opt = g02 [Sνν (ωm ) − Sνν (−ωm )] (H.47)
 
κ κ
= g02
− n cav (H.48)
(c − ωm )2 + (κ/2)2 (c + ωm )2 + (κ/2)2

Now we revisit the lowest number of phonons achievable with the cavity cooling

↑ + n th m
N t→∞ = . (H.49)
opt + m

We assume here that the frequency of the mechanical mode is much larger than the
linewidth of the cavity mode, so that the noise spectrum originated in the mechanical
mode is well separated from the signal of the cavity in the frequency domain. This
regime is called the resolved sideband regime. When the cooling rate opt is much
smaller than m , it holds trivially that N t→∞ = n th , where the system remains in
the initial thermal state. When the spectrum of the mechanical mode is sufficiently
narrow and opt  m holds, we have

↑ (c − ωm )2 + (κ/2)2
N t→∞ = = (H.50)
opt 4ωm c
280 Appendix H: Cavity Cooling of a Mechanical Mode

This is minimized when c = ωm and gives a minimum available number of phonons

κ2 /16ωm 2 , which is well below 1 since κ ωm in the resolved sideband regime.
Another thing we should note is that the cooling rate opt is proportional to the num-
ber of photons in the cavity, therefore the cooling rate can relatively easily become
larger than the decay rate of the high-Q mechanical mode. Given these estimations,
the cavity optomechanical system has been a cutting-edge of the modern quantum
physics as the one being capable of seeing the quantum behavior of macroscopic (
atomic scale) objects.
Entangled States and Quantum
Teleportation I

This appendix introduces the quantum entanglement and its celebrated applications.

I.1 Separable and Entangled States

Suppose there are two quantum systems labeled 1 and 2. A composite quantum
def
system can be like |ψ1  ⊗ |ψ2  = |ψ1 ψ2  that is friendly to our intuition. This kind of
composite quantum systems, which is decomposed into the product of the constituent
individual quantum systems, are said to be separable states. In general, the separable
state is written in terms of the density matrices of the subsystems ρi1 and ρi2 as

ρ = i pi ρi1 ⊗ ρi2 with probabilities pi .
On the other hand, there are states that cannot + ±  be decomposed into √such a
product. The most famous ones are Bell states +12 = (|e1 g2  ± |g1 e2 )/ 2 and
+ ± √
+ = (|e1 e2  ± |g1 g2 )/ 2. Bell states cannot be written as a separable states
12
in any basis transformation as one can immediately check by reductio ad absur- √
dum. Other non-separable states are, e.g. a Schrödinger’s cat (|g |α + |e |−α)/ 2
with coherent states |±α of a harmonic √ oscillator, Greenberger–Horne–Zeilinger
(GHZ) state (|g · · · g + |e
√ · · · e)/ 2 of a many-qubit system, N00N (“Noon”)
state (|N  |0 + |0 |N )/ 2 of two harmonic oscillators. These family of non-
separable states are named entangled states.

I.2 Measures of Entanglement

One might expect there is some measure of the degree or at least the existence of
the entanglement. Actually there are several attempts to develop such a quantity,
however, there is no satisfactory one that can be used to judge whether a quantum
state is entangled or not. However, for a two-qubit system, there are good measures
telling the existence and degree of entanglement.

© The Editor(s) (if applicable) and The Author(s), under exclusive license 281
to Springer Nature Singapore Pte Ltd. 2022
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7
282 Appendix I: Entangled States and Quantum Teleportation

I.2.1 Entanglement Entropy

We can see below what information can be extracted when we focus on one of
the subsystems in the composite system, represented by a reduced density matrix.
Moreover, we will try to calculate the von Neumann entropy of the reduced density
matrix, a quantity called an entanglement entropy.
First, the reduced density matrix ρ̃1 or ρ̃2 is a density matrix of the com-
posite system ρ12 traced out for one of the two constituent quantum systems:
ρ̃1 = Tr 2 ρ12 or ρ̃2√= Tr 1 ρ12 . Let us consider two
+ cases,
 |0  =
the separable state √
|g1  (|g2  + |e2 )/ 2 and one of the Bell state ++ = (|g1 , g2  + |e1 , e2 )/ 2. On
one hand, reduced density matrices of |0  is easily obtained as ρ̃1 = |g1  g1 | and
ρ̃2 = (|g2  + |e2 )(g2 | + e2 |)/2, the density matrices
√ that can be derived from+ the
original individual states |g1  and (|g2  + |e2 )/ 2. On the other hand, for ++
the reduced density matrices reads ρ̃1 = I /2 and ρ̃2 = I /2 as well, meaning that by
taking a trace over one Hilbert space, the quantum state in the other Hilbert space is
completely ruined, down to the completely mixed state.
Next, let us define one of the measures of the entanglement. The quantum von
Neumann entropy is defined in analogy to the Shannon entropy as −tr{ρ log ρ}. For a
pure state |ψ, quantum von Neumann entropy can be calculated by using eigenvalues
λk of the density matrix as −tr{ρ log ρ} = − k λk log λk = 0, since the eigenvalues
are either 0 or 1. If the density matrix is a completely mixed state ρ = I /2, the von
Neumann entropy reads log 2. In a similar manner, the entanglement entropy S1 or
S2 is defined by the von Neumann entropy of the reduced density matrix ρ1 or ρ2 ,
respectively:

S1 = −Tr[ρ1 log ρ1 ], S2 = −Tr[ρ2 log ρ2 ]. (I.1)

+ 
For the quantum state |0 , S1 = S2 = 0. In contrast, the Bell state ++ yields the
entanglement entropy of log 2. One might want to think that non-zero entanglement
entropy signals the entanglement, however, the state ρ12 = (|g, g + |g, e + |e, g +
|e, e)/4, the completely mixed state, also gives ρ1 = I /2 and ρ2 = I /2 to yield S1 =
S2 = log 2. The entanglement entropy is a good measure only when we consider pure
states.

I.2.2 Negativity

Instead of considering the reduced density matrix, we shall take a look at the density
matrix ρ of the composite quantum system comprised of qubit A and qubit B. ρ is of
∗
course Hermitian, that is, ρ† = ρT = ρ. Since its diagonal components represent
the probability, its eigenvalues should be non-negative.
Appendix I: Entangled States and Quantum Teleportation 283

Let us then think about what happens with the partial transposition of ρ, which is
denoted by ρT A . We explicitly write the density matrix in general as

ρ= pξ A η A ξ B η B |ξ A  η A | ⊗ |ξ B  η B | (I.2)
ξ A ,η A ,ξ B ,η B

with |ξ A  , |η A  ∈ {|g A  , |e A } and |ξ B  , |η B  ∈ {|g B  , |e B }. According to this

expression, the partial transposition reads

ρT A = pξ A η A ξ B η B |η A  ξ A | ⊗ |ξ B  η B | . (I.3)
ξ A ,η A ,ξ B ,η B

This is Hermitian again, meaning that the eigenvalues of ρT A is still real. However,
the eigenvalues are not necessarily non-negative. An exceptional situation is that ρ
is separable. Peres–Horodecki criterion [32] states that if ρ is separable, then all the
eigenvalues of ρT A are non-negative. Therefore, an existence of negative eigenvalues
in ρT A heralds the existence of quantum entanglement in ρ. On top of that, the
converse is true with the two-qubit system: the separability of ρ and non-negativity
of the eigenvalues of ρT A are equivalent in two-qubit systems.
Here we saw an essential idea of negativity, an alternative entanglement measure.
Negativity N (ρ) is defined as a sum of absolute values of the negative eigenvalues
of ρT A :

N (ρ) = |λn | (I.4)
λn <0

where λn are eigenvalues of ρT A . Since ρT A is just a partial transposition of

ρ, the trace should be preserved, i.e. Tr[ρ] = Tr[ρT A ] = 1. Thence it holds that
   def
1 = n λn = λn >0 λn − λn <0 |λn |. We further define a trace norm ||ρT A ||1 =
 †
Tr[ ρT A ρT A ]. Since this definition sums up the absolute values of the eigenval-
 
ues of ρT A , ||ρT A ||1 = n |λn | = 2 λn <0 |λn | + 1. Here an alternative expression
of the negativity is achieved:

||ρT A ||1 − 1
N (ρ) = (I.5)
2
The Peres–Horodecki criterion for the negativity is valid for mixed state, providing
a more general measure than the entanglement entropy that is only useful for pure
states. In order for the entanglement measure to be additive when the two systems are
def
synthesized, One might use logarithmic negativity L(ρ) = log [2N (ρ) + 1]. With
this quantity, we have a nice formula L(ρ1 ⊗ ρ2 ) = L(ρ1 ) + L(ρ2 ).
284 Appendix I: Entangled States and Quantum Teleportation
+ 
As an example, let us take a glimpse of the Bell state ++ . The density matrix
ρ+ reads
⎛ ⎞
1001
1 ⎜0 0 0 0 ⎟
ρ + = ⎜ ⎟ (I.6)
2 ⎝0 0 0 0 ⎠
1001

and its partial transposition does

⎛ ⎞
1 0 0 0
TA 1⎜
⎜0 0 1 0⎟⎟.
ρ = ⎝ (I.7)
+
2 0 1 0 0⎠
0 0 0 1

The eigenvalues of this partially transposed density matrix are three 1/2 and one
−1/2. Provided these, the negativity is 1/2 and the logarithmic negativity is log 2.
A more interesting example is Werner state
+  + I⊗I
ρW = p ++ + + + (1 − p) (I.8)
⎛ ⎞ ⎛4 ⎞
1001 10 0 0
p⎜ 0 0 0 0 ⎟ 1 − p ⎜0 1 0 0⎟
= ⎜ ⎟+ ⎜ ⎟ (I.9)
2 ⎝0 0 0 0⎠ 4 ⎝0 0 1 0⎠
1001 00 0 1
⎛ 1+ p p
⎞
0 0
⎜ 4 1− p 2
⎟
⎜ 0 0 0 ⎟
=⎜ 4 ⎟. (I.10)
⎝ 0 0 1−4 p 0 ⎠
p
2 0 0 1+4 p

Whether this state is entangled or not depends on the value of p and we are to clarify
the threshold. The partial transposition of this appears as
⎛ 1+ p ⎞
4 0 0 0
⎜ 1− p p ⎟
⎜ 0 0 ⎟
ρTWA = ⎜ 4
p 1− p
2 ⎟, (I.11)
⎝ 0 2 4 0 ⎠
0 0 0 1+4 p

which yields eigenvalues three (1 + p)/4 and one (1 − 3 p)/4. The latter takes
a negative value when 1/3 < p ≤ 1, so that according to the Peres–Horodecki
criterion, the Werner state is entangled only if 1/3 < p ≤ 1 and separable when
0 ≤ p ≤ 1/3. Here the term “entangled” stands for another sense: the Werner state
with 1/3 < p ≤ 1 cannot be generated through local operation and classical com-
munication (LOCC, see the last section of this appendix) starting from the separable
state, whereas the one with 0 ≤ p ≤ 1/3 can be.
Appendix I: Entangled States and Quantum Teleportation 285

I.3 Quantum Teleportation

In this section, we introduce the concept of quantum teleportation as an applica-

tion of the entangled state, as well as an interesting phenomena exhibited by the
quantum +world.
 All the thing wanted be done is transmitting an unknown state
|ψ = cg +gψ + ce |eψ from a transmitter Alice, to a receiver Bob. The subscript
ψ is the reminder of the quantum states belonging to |ψ. Alice should keep |ψ
unknown in contrast to the classical communication because once Alice knows the
quantum state by measurement, the measured state has no longer be the same state as
|ψ. One might keep a copy of the transmitted information in the classical commu-
nication, however, no-cloning theorem inhibits that. Therefore, sending an unknown
state requires special trick that fully utilizes the bizarre
+ −  nature of entangled state.√
First thing to do is to share an entangled pair, say +ab =(|ea  |gb  + |ga  |eb )/ 2,
between Alice and Bob. Here the subscripts a and b indicates the qubit systems
+ −  and Bob. Combining together with |ψ, a whole quantum state reads
of Alice
|ψ +ab . Let us rewrite this state
+ in terms+ of Bell states involving Alice’s qubit
+ ± √
+ −
and the qubit system ψ. Using +ψa (|ψ ab ) = (±cg |gb  − ce |eb )/2 2 and
+  + − √
+ ±
+ψa (|ψ +ab ) = (ce |gb  ∓ cg |eb )/2 2, we have
+  + 
+ −  cg +gψ + ce +eψ |ea  |gb  + |ga  |eb 
|ψ +ab = √ √ (I.12)
2 2
+  + 
+ + + −
+ψa +ψa
= (cg |gb  − ce |eb ) − (cg |gb  + ce |eb ) (I.13)
2+  2+ 
+ + + −
+ψa +ψa
+ (ce |gb  − cg |eb ) + (ce |gb  + cg |eb ) (I.14)
2 2
Alice then operates an X gate, X a , on their qubit and successively the CNOT+gate with

+ ±
Alice’s qubit being the control qubit and ψ the target one. Then we have +ψa →
+  +  √ +  + 
+ ± + +
CNOT · X a +ψa = +gψ (± |ga  + |ea )/ 2 and +ψa → CNOT · X a +ψa =
± ±
+  √
+eψ (|ga  ± |ea )/ 2 for the subsystem ψa. Finally Alice operates the Hadamard
gate Ha for their qubit to obtain
+  + 
+ + + +
+ψa (cg |gb  − ce |eb ) → Ha · CNOT · X a +ψa (cg |gb  − ce |eb ) (I.15)
+ 
= +gψ |ga  (cg |gb  − ce |eb ), (I.16)
+  + 
+ − + −
+ψa (cg |gb  + ce |eb ) → Ha · CNOT · X a +ψa (cg |gb  + ce |eb ) (I.17)
+ 
= +gψ |ea  (cg |gb  + ce |eb ), (I.18)
+  + 
+ + +
+ψa (ce |gb  − cg |eb ) → Ha · CNOT · X a ++ ψa (ce |gb  − cg |eb ) (I.19)
+ 
= +eψ |ga  (ce |gb  − cg |eb ), (I.20)
286 Appendix I: Entangled States and Quantum Teleportation

Fig. I.1 Schematics of quantum teleportation

+  + 
+ − +
+ψa (ce |gb  + cg |eb ) → Ha · CNOT · X a +− ψa (ce |gb  + cg |eb ) (I.21)
+ 
= − +eψ |ea  (ce |gb  + cg |eb ), (I.22)
+  + 
With this knowledge, Alice measures the system ψa in the bases +gψ |ga , +gψ |ea ,
+  + 
+eψ |ga  and +eψ |ea , which is termed Bell measurement. Alice should tell Bob the
measurement result+ in order to finish their work. If Alice’s measurement says the
system +ψa is in +gψ |ga , then Bob should operate Z gate +Z b on their qubit. When
it says +eψ |ga , Bob does Z b gate and then X b gate. For +
+ eψ |ea , X b gate is fine,
and nothing is to be done for the measurement result +gψ |ea . All the way here,
Bob has a quantum state |ψ = cg |gb  + ce |eb , the very state that Alice wanted to
transmit. At the time when Alice made measurement, the quantum state |ψ had been
destroyed, so that the quantum state is not copied (Fig. I.1).

I.4 Entanglement Swapping

As a close relative of the quantum teleportation, we here introduce entanglement

swapping which is of frequent use in the quantum communication.+Theproblem is as
follows. Alice and Bob have pairs of known entangled states, say ++A = (|g1 g2  +
√ + + √
+
|e1 e2 )/ 2 for Alice and  B = (|g3 g4  + |e3 e4 )/ 2 for Bob. Then Alice and
Bob wants to somehow generate an entangled state shared between them. This can
be accomplished by the Bell measurement of the two-qubit system consisting of one
Alice’s qubit and one Bob’s, and successively applying local quantum operations.
Let us briefly see how this works.
Appendix I: Entangled States and Quantum Teleportation 287

The total system | including Alice’s and Bob’s entangled states is written as
+  + +
| = ++A ⊗ B
+ (I.23)
|g1 g2 g3 g4  + |g1 g2 e3 e4  + |e1 e2 g3 g4  + |e1 e2 e3 e4 
= . (I.24)
2
By defining the Bell states of the
+ ±subsystem
 + involving
 + ±  the+first and the fourth (the
second and the third) qubits as +14 and +±
14 ( + and +± ) as in the previous
23 23
section, we can rewrite | as
+ + + +  + − + −  + ++ + + −+ −
+ + + + + + + + + + +
14 23 14 23 14 23 14 23
| = . (I.25)
2
Then let us make a Bell measurement for the second and the third qubits, with which
the subsystem involving + + the first and the fourth qubits can be turned into the desired
entangled state, say + . If the Bell measurement says the measured system is in the
+  14 + +
state ++ 23 , the remaining system
+ −is projected onto the state + and there is nothing
14
we have to do. If in the state + , Z gate, a Z gate on the first qubit, transforms
+ − +  23 1 + 
+ into ++ and we get the desired state. With the measured state being + + ,
14 14 + + + +  23
X 1 gate, an + X gate on the first qubit, allows us to get X 1 +14 = +14+. Finally, when
− −
the+ state 23 is obtained, we apply Z 1 X 1 on the remaining system +14
+ , resulting

in ++ 14 . In this manner, by passing ones of Alice’s and Bob’s entangled pairs and
measuring those qubits in the Bell basis, Alice and Bob acquire an entangled pair
between them after the proper local operation (Fig. I.2).

I.5 Local Operation and Classical Communication (LOCC)

In quantum teleportation and entanglement swapping, procedures are summarized

as follows: (i) measurement of qubit(s) of one party, represented by a Kraus oper-
ator M A (ii) classical communication to tell the measurement result and (iii) quan-
tum operation U B on qubit of another party. These are termed as local opera-
tion and classical communication (LOCC). If we naively neglect the communi-
cation time, the density matrix evolution after above LOCC can be written as
 † †
r U B (r )M A (r )ρM A (r )U B (r ).
We do not intend to introduce every theoretical detail but list some important
properties of LOCC.

• LOCC cannot increase the entanglement measures with unit probability.

• LOCC cannot entangle the separable state.
• LOCC can probabilistically increase the entanglement measure, which is dubbed
as the entanglement distillation.
288 Appendix I: Entangled States and Quantum Teleportation

Fig. I.2 Schematics of entanglement swapping

Quantum No-Cloning Theorem
J

An important operation that is allowed in classical information and not possible in

quantum information is copying the state. For example, if a value x is given and
would like to y = x 2 + x, then one can first copy x and add 1 to the original x and
multiply by the copied x to perform,

y = x 2 + x = x(x + 1). (J.1)

In quantum information, however, Quantum No-Cloning Theorem prohibits the

copying of a quantum state and therefore the calculation cannot be performed in
the same way. This first seems to be a huge disadvantage for the quantum compu-
tation, but there are number of tricks known so that the quantum computation can
actually perform numerous calculations. So please do not worry. Here we show a
simple proof of the no-cloning theorem.
Suppose there is an arbitrary quantum state |ψ = a |0 + b |1 and copy it to
another quantum state |i, so that |i → |ψ, using a unitary transformation Û .
Another words, there exist a two-qubit gate,

Û |ψ |i = |ψ |ψ . (J.2)

Similarly, using the same operation Û , different quantum state |φ can also be copied,
thus
Û |φ |i = |φ |φ . (J.3)
Using above results,

φ| i| U † U |ψ |i = φ| ψ i| i

= φ| ψ , (J.4)

© The Editor(s) (if applicable) and The Author(s), under exclusive license 289
to Springer Nature Singapore Pte Ltd. 2022
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7
290 Appendix J: Quantum No-Cloning Theorem

is calculated and likewise,

φ| i| U † U |ψ |i = φ| φ|ψ |ψ

= φ| ψ2 . (J.5)

These results demand φ| ψ = φ| ψ2 , however, it can be only fulfilled with either
φ| ψ = 0 or φ| ψ = 1. The later is a trivial case, |φ = |ψ, and be ignored. Then,
the first case tells us that two states can be duplicated only when they are orthogonal,
and not for an arbitrary quantum state.
Let us take a look at an actual example. Suppose, operation Û is capable of
duplicating both |0 and |1 as,

Û |0 |i = |0 |0 (J.6)

Û |1 |i = |1 |1 . (J.7)

Now, let’s try to copy |ψ = a |0 + b |1. In quantum information, a measurement
will destroy the quantum state, so we need some method of copying an arbitrary
quantum state without knowing the state itself.
Performing the same operation to |ψ,

Û |ψ |i = Û (a |0 + b |1) |i = a Û |0 |i + bÛ |1 |i
= a |0 |0 + b |1 |1 , (J.8)

is obtained. It may seem like the copying worked, but the state to be copied is
|ψ = a |0 + b |1, so if the actual copying worked, we should have obtained

|ψ |ψ = (a |0 + b |1)(a |0 + b |1)

= a 2 |00 + b2 |11 + ab(|01 + |10). (J.9)

So we were not duplicating the state. It is possible to find an operation Û , which

copies a known state and another state which is orthogonal to the state, however, the
same operation cannot be used to duplicate an arbitrary quantum state.
References

1. Shor, P.W.: Algorithms for quantum computation: discrete logarithms and factoring. In: Pro-
ceedings 35th Annual Symposium on Foundations of Computer Science, pp. 124–134. IEEE
Computer Society Press (1994)
2. Shor, P.W.: Scheme for reducing decoherence in quantum computer memory. Phys. Rev. A
52, R2493 (1995)
3. Einstein, A., Podolsky, B., Rosen, N.: Phys. Rev. 47, 777 (1935)
4. Bennett, C.H., Brassard, G., Crépeau, C., Jozsa, R., Peres, A., Wootters, W.K.: Phys. Rev.
Lett. 70, 1895 (1993)
5. Bouwmeester, D., Pan, J.-W., Mattle, K., Eibl, M., Weinfurter, H., Zeilinger, A.: Nature 390,
575 (1997)
6. Boschi, D., Branca, S., De Martini, F., Hardy, L., Popescu, S.: Phys. Rev. Lett. 80, 1121 (1998)
7. Bennett, C.H., Brassard, G.: Proceedings of IEEE International Conference on Computers
Systems and Signal Processing, p. 175 (1984)
8. Loudon, R.: The Quantum Theory of Light, 3rd edn. Oxford University Press (2000)
9. DeMill, D.: Atomic Physics, 3rd edn. Oxford University Press (2000)
10. Jaynes, R., Cummings, A.: J. Phys. B: Mol. Opt. Phys. 38(9), S551 (2005)
11. Molmer, K., Sorensen, A.: Multiparticle entanglement of hot trapped ions. Phys. Rev. Lett.
82, 1835 (1999)
12. Gaebler, J.P., Tan, T.R., Lin, Y., Wan, Y., Bowler, R., Keith, A.C., Glancy, S., Coakley, K.,
Knill, E., Leibfried, D., Wineland, D.J.: High-fidelity universal gate set for 9 Be+ ion qubits.
Phys. Rev. Lett. 117, 060505 (2016)
13. Johansson, J.R., Nation, P.D., Nori, F.: QuTiP 2: a Python framework for the dynamics of open
quantum systems. Comp. Phys. Comm. 184, 1234–1240 (2013)
14. Yariv, A.: Quantum Electronics. Wiley, New York (1989)
15. Feynmann, R.P., Leighton, R.B., Sands, M.: The Feynman Lectures on Physics, vol. 3, Chapter
21 (Basic Books); https://www.feynmanlectures.caltech.edu/III_21.html
16. Schuster, D.I., Houck, A.A., Schreier, J.A., Wallraff, A., Gambetta, J.M., Blais, A., Frunzio,
L., Johnson, B., Devoret, M.H., Girvin, S.M., Schoelkopf, R.J.: Resolving photon number
states in a superconducting circuit. Nature 445, S15 (2007)
17. Auffeves-Garnier, A., Simon, C., Gerard, J.-M., Poizat, J.-P.: Giant optical nonlinearity induced
by a single two-level system interacting with a cavity in the Purcell regime. Phys. Rev. A 75,
053823 (2007)

© The Editor(s) (if applicable) and The Author(s), under exclusive license 291
to Springer Nature Singapore Pte Ltd. 2022
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7
292 References

18. Nielsen, M., Chuang, I.: Quantum Computation and Quantum Information, 10th anniversary
ed. Cambridge University Press (2010)
19. Emerson, J., Alicki, R., Zyczkowski, K.: Scalable noise estimation with random unitary oper-
ators. J. Opt. B: Quantum Semiclass. Opt. 7, S347 (2005)
20. Fujii, K.: Quantum Computation with Topological Codes. Springer (2015)
21. Fukui, K., Tomita, A., Okamoto, A., Fujii, K.: High-threshold fault-tolerant quantum compu-
tation with analog quantum error correction. Phys. Rev. X 8, 021054 (2018)
22. Gottesman, D., Kitaev, A., Preskill, J.: Encoding a qubit in an oscillator. Phys. Rev. A 64,
012310 (2001)
23. Fluhmann, C., Nguyen, T.L., Marinelli, M., Negnevitsky, V., Mehta, K., Home, J.P.: Encoding
a qubit in a trapped-ion mechanical oscillator. Nature 566, 513 (2019)
24. Campagne-Ibarcq, P., Eickbusch, A., Touzard, S., Zalys-Geller, E., Frattini, N.E., Sivak, V.V.,
Reinhold, P., Puri, S., Shankar, S., Schoelkopf, R.J., Frunzio, L., Mirrahimi, M., Devoret,
M.H.: Quantum error correction of a qubit encoded in grid states of an oscillator. Nature 584,
368 (2020)
25. Schuster, D.I., Houck, A.A., Schreier, J.A., Wallraff, A., Gambetta, J.M., Blais, A., Frunzio,
L., Johnson, B., Devoret, M.H., Girvin, S.M., Schoelkopf, R.J.: Resolving photon number
states in a superconducting circuit. Nature 445, S15 (2007)
26. http://www.qc.rcast.u-tokyo.ac.jp/lecture_en.html
27. Igi, K., Kawai, H.: Quantum Mechanics II. Kodansha (1994)
28. Auffeves-Garnier, A., Simon, C., Gerard, J.-M., Poizat, J.-P.: Phys. Rev. A 75, 053823 (2007)
29. DiVincenzo, D.P.: Fortschr. Phys. 48, 771 (2000)
30. Aspelmeyer, M., Kippenberg, T.J., Marquardt, F.: Rev. Mod. Phys. 86, 1391 (2014)
31. Clerk, A.A., Devoret, M.H., Girvin, S.M., Marquardt, F., Shoelkopf, R.J.: Rev. Mod. Phys.
82, 1155 (2010)
32. Peres, A.: Separability criterion for density matrices. Phys. Rev. Lett. 77, 1413 (1996);
Horodecki, M., Horodecki, P., Horodecki, R.: Separability of mixed states: necessary and
sufficient conditions. Phys. Lett. A 223, 1–8 (1996)
33. Kadowaki, T., Nishimori, H.: Quantum annealing in the transverse Ising model. Phys. Rev. E
58, 5355 (1998)
34. Preskill, J.: Quantum computing in the NISQ era and beyond. Quantum 2, 79 (2018)
35. Scarani, V., Bechmann-Pasquinucci, H., Cerf, N.J., Dušek, M., Lütkenhaus, N., Peev, M.: The
security of practical quantum key distribution. Rev. Mod. Phys. 81, 1301 (2009)
36. Degen, C.L., Reinhard, F., Cappellaro, P.: Quantum sensing. Rev. Mod. Phys. 89, 035002
(2017)
37. Rondin, L., Tetienne, J.-P., Hingant, T., Roch, J.-F., Maletinsky, P., Jacques, V.: Magnetometry
with nitrogen-vacancy defects in diamond. Rep. Prog. Phys. 77, 056503 (2014)
38. Schirhagl, R., Chang, K., Loretz, M., Degen, C.L.: Nitrogen-vacancy centers in diamond:
nanoscale sensors for physics and biology. Annu. Rev. Phys. Chem. 65, 83 (2014)
39. Jaklevic, R.C., Lambe, J., Mercereau, J.E., Silver, A.H.: Macroscopic quantum interference
in superconductors. Phy. Rev. 140(5A), A1628 (1965)
40. Fagaly, R.L.: Superconducting quantum interference device instruments and applications. Rev.
Sci. Instrum. 77, 101101 (2006)
41. Hatridge, M., Vijay, R., Slichter, D.H., Clarke, J., Siddiqi, I.: Dispersive magnetometry with
a quantum limited SQUID parametric amplifier. Phys. Rev. B 83, 134501 (2011)
42. Kominis, K., Kornack, T.W., Allred, J.C., Romalis, M.V.: A subfemtotesla multichannel atomic
magnetometer. Nature 422, 596 (2003)
43. Shah, V., Knappe, S., Schwindt, P.D.D., Kitching, J.: Subpicotesla atomic magnetometry with
a microfabricated vapour cell. Nat. Photonics 1(11), 649 (2007)
44. Jensen, K., Budvytyte, R., Thomas, R.A., Wang, T., Fuchs, A.M., Balabas, M.V., Vasilakis,
G., Mosgaard, L.D., Stærkind, H.C., Mueller, J.H., Heimburg, T., Olesen, S.-P., Polzik, E.S.:
Non-invasive detection of animal nerve impulses with anatomic magnetometer operating near
quantum limited sensitivity. Sci. Rep. 6, 29638 (2016)
References 293

45. Aspelmeyer, M., Kippenberg, T.J., Marquardt, F.: Cavity optomechanics. Rev. Mod. Phys. 86,
1391 (2014)
46. Rugar, D., Budakian, R., Mamin, H.J., Chui, B.W.: Single spin detection by magnetic resonance
force microscopy. Nature 430, 329 (2004)
47. Trabesinger, A.: Quantum simulation. Nat. Phys. 8, 263 (2012)
48. Georgescu, I.M., Ashhab, S., Nori, F.: Quantum simulation. Rev. Mod. Phys. 86, 153 (2014)
49. Awschalom, D., et al.: Phys. Rev. X 2, 017002 (2021)
50. Safavi-Neaini, A.H., van Thourhout, D., Baets, R., van Laer, R.: Optica 6, 213 (2019)
51. Lambert, N.J., Rueda, A., Sedlmeir, F., Schwefel, H.G.L.: Adv. Quantum Technol. 3, 1900077
(2020)
52. Kimble, H.J.: The quantum internet. Nature 453, 1023 (2008)
53. Buhrman, H., Röhrig, H.: Distributed quantum computing. In: Mathematical Foundations of
Computer Science, Lecture Notes 2747 (2003)
54. Gottesman, D., Jennewein, T., Croke, S.: Longer-baseline telescopes using quantum repeaters.
Phys. Rev. Lett. 109, 070503 (2012)
55. Kómár, P., Kessler, E.M., Bishof, M., Jiang, L., Sørensen, A.S., Ye, J., Lukin, M.D.: A quantum
network of clocks. Nat. Phys. 10, 582–587 (2014)
56. Bennett, C., Wiesner, S.: Communication via one- and two-particle operators on Einstein-
Podolsky-Rosen states. Phys. Rev. Lett. 69, 2881 (1992)
57. Pan, J.W., Bouwmeester, D., Weinfurter, H., Zeilinger, A.: Experimental entanglement swap-
ping: entangling photons that never interacted. Phys. Rev. Lett. 80, 3891 (1998)
58. Feynman, R.: Simulating physics with computers. In: International Journal of Theoretical
Physics (1981)
59. Greiner, M., Mandel, O., Esslinger, T., Hänsch, T.W., Bloch, I.: Quantum phase transition
from a superfluid to a Mott insulator in a gas of ultracold atoms. Nature 415, 39 (2002)
60. Kim, K., Chang, M.-S., Korenblit, S., Islam, R., Edwards, E.E., Freericks, J.K., Lin, G.-D.,
Duan, L.-M., Monroe, C.: Quantum simulation of frustrated Ising spins with trapped ions.
Nature 465, 590 (2010)
61. Broadbent, A., Fitzsimons, J., Kashefi, E.: Universal blind quantum computation. In: 2009
50th Annual IEEE Symposium on Foundations of Computer Science. IEEE (2009)
62. Mattle, K., Weinfurter, H., Kwiat, P.G., Zeilinger, A.: Dense coding in experimental quantum
communication. Phys. Rev. Lett. 76, 4656 (1996)
63. Galindo, A., Martín-Delgado, M.A.: Information and computation: classical and quantum
aspects. Rev. Mod. Phys. 74, 347 (2002)
Index

A Cavity optomechanics, 201, 275

Ac Stark shift, 96 Cavity Quantum Electrodynamics (QED),
Amplitude correlation, 133 cavity, 172
Amplitude damping, 224 Charge energy, 198
Anharmonic oscillator, 144, 196, 198 Clifford group, 218
Annihilation operator, 123 Code concatenation, 231
Anti-crossing, 96 Coherence, 258
Anti-Jaynes-Cummings interaction, 145, 186 Coherent, 134
Atomic gas, 147 Coherent state, 126
Atomic ion, 189 Commutation relation, 66
Avoided crossing, 96 Completely mixed state, 63
Composite system, 36
Controlled gate, 213
B
Controlled-NOT (CNOT) gate, 211, 214, 219
Baker-Campbell-Hausdorff formula, 127, 130,
Controlled-Z gate, 214
139, 248
Cooperativity, 177, 204, 206
BB84 protocol, 239
Cooper-pair box, 199
Beam-splitter interaction, 187, 203
Coulomb interaction, 192
Bell measurement, 286
Coupling capacitor, 200
Bell state, 47, 281 Coupling strength, 145, 177, 185
Bit-flip error, 223 Creation operator, 123
Bit-phase-flip error, 224 Cross-Kerr effect, 189
Bloch equation, 151, 266
Bloch sphere, 43, 143
D
Blue-sideband transition, 193
Dark state, 252
Bohr magneton, 255
Density matrix, 63, 257
Bohr’s model, 253
Density of states, 114
Boundary condition, 166
Density operator, 61, 63
Bra vector, 31 Dephasing, 152
Bright state, 252 Depolarizing error, 224
Dirac notation, 30
C Dirac picture, 78, 86
Carrier transition, 193 Direct product, 36
Cavity, 159 Dispersive Hamiltonian, 178
Cavity cooling, 275 Displacement operator, 126
Cavity mode, 124 Dissipator, 152
© The Editor(s) (if applicable) and The Author(s), under exclusive license 295
to Springer Nature Singapore Pte Ltd. 2022
A. Osada et al., Introduction to Quantum Technologies, Lecture Notes in Physics 1004,
https://doi.org/10.1007/978-981-19-4641-7
296 Index

DiVincenzo criteria, 232 Ising coupling gate, 215

Dual-port measurement, 171
J
E Jaynes-Cummings Hamiltonian, 172, 248
Electric dipole interaction, 189 Jaynes-Cummings interaction, 145, 186, 190
Entangled state, 47, 281 Jaynes-Cummings ladder, 173
Entanglement entropy, 282 Jaynes-Cummings model, 172
Entanglement swapping, 243, 286 Josephson effect, 196, 197
Error threshold, 228, 231 Josephson energy, 198
Even cat state, 129 Josephson junction, 149, 195
External coupling, 168
K
F Ket vector, 31
Fabry-Perót cavity, 190 Kraus operator, 220, 223
Fano lineshape, 181
Fault-tolerant, 231 L
Fermi’s golden rule, 111, 166 Ladder operator, 125
Fidelity, 222 Lamb-Dicke parameter, 192
Finesse, 160, 161 Landé g factor, 255
Fine structure, 254 Laser cooling, 275
First-order coherence, 133 Light shift, 96
Flux quantum, 196 Lindblad superoperator, 152
Fock state, 123 Local Operation and Classical Communication
Four-wave mixing, 189 (LOCC), 284, 287
Logarithmic negativity, 283
G Logical qubit, 229
Gate fidelity, 225 Longitudinal relaxation, 146, 152
Generalized Rabi frequency, 95, 154
Global phase, 42 M
Gottesman-Kitaev-Preskill (GKP) code, 231 Magic state, 218
Gottesman-Knill theorem, 218 Magnetic flux, 196
Greenberger-Horne-Zeilinger (GHZ) state, 49, Marginal distributions, 135
281 Master equation, 150, 152, 257
Ground-state cooling, 201 Measurement with errors, 221
Grover’s algorithm, 236 Minimum-uncertainty state, 127
Mixed state, 62
H Mode volume, 160, 191
Hadamard gate, 213 Momentum representation, 245
Harmonic oscillator, 123
Heisenberg equation of motion, 88, 150 N
Heisenberg picture, 78, 81 Negativity, 282
Heisenberg uncertainty principle, 67 No-cloning theorem, 228
Hermitian operator, 54 Noise superoperator, 223
Hybrid quantum system, 204 Noisy Intermediate-Scale Quantum (NISQ)
Hyperfine structure, 147, 255 device, 236, 242
Non-classicality, 135
I Non-Clifford gate, 218
Incoherent, 134 Nonlinear interaction, 188
Input-output relation, 167 N00N (“Noon”) state, 281
Input-output theory, 166 Nuclear magnetic resonance, 146
Intensity correlation, 134 Number distribution, 128
Intrinsic loss, 168 Number operator, 125, 126
Ion trap, 189, 192 Number state, 123
Index 297

O Quantum Non-Demolition (QND) measure-

Odd cat state, 129 ment, 221
Operator function, 53 Quantum number, 254
Optical Bloch equation, 266 Quantum Phase Estimation (QPE), 237
Optical damping, 277 Quantum process tomography, 224
Optical spring effect, 202 Quantum regression theorem, 279
Optomechanical interaction, 201 Quantum repeater, 242
Quantum sensing, 240
P Quantum simulation, 242
Parity measurement, 229 Quantum state tomography, 221
Partially coherent, 134 Quantum teleportation, 243, 285
Partial trace, 65 Qubit, 141
Partial transposition, 283
Pauli group, 218 R
Pauli operator, 59, 142 Rabi frequency, 95, 153
Paul trap, 192 Rabi oscillation, 110, 153
Peres-Horodecki criterion, 283 Radiation pressure, 202
Perturbation theory, 101 Raising and lowering operators, 144
Phase estimation algorithm, 237 Ramsey interference, 155
Phase-flip error, 223 Randomized benchmarking, 226
Phonon, 192, 201 Rate equation, 265, 277
Phonon-phonon coupling, 195 Red-sideband transition, 193
Photon flux, 167 Reflection measurement, 163
Poissonian distribution, 128 Relaxation, 149
Population, 258 Resolved-sideband regime, 201, 279
Position representation, 245 Resonator, 159
Positive Operator-Valued Measure (POVM), Rotating frame, 91, 247
221 Rotating-wave approximation, 145
Power broadening, 157, 267
p-representation, 245 S
Projection operator, 56, 220 Schrieffer-Wolff transformation, 178, 251, 269
Projective measurement, 220 Schrödinger equation, 26, 73
Propagating mode, 166 Schrödinger picture, 78, 81
Purcell factor, 177 Schrödinger’s cat state, 129, 137, 281
Pure state, 62 Second-order coherence, 134
Self-Kerr effect, 189
Q Separable state, 281
q-representation, 245 S gate, 213
Quality factor, Q factor, 160, 161 Shor code, 230
Quantum bit, 141 Shor’s algorithm, 238
Quantum circuit, 219 Sideband cooling, 275
Quantum computation, 235 Sideband transition, 192
Quantum computer, 235 Single-port measurement, 168
Quantum conversion efficiency, 206 Single-qubit gate, 219
Quantum defects, 148 Solovey-Kitaev theorem, 218
Quantum dot, 148 SPAM error, 225, 226
Quantum error correction, 227 Spin echo, 155
Quantum error-correction code, 229 Spontaneous emission, 145
Quantum Fourier transform, 238 Spontnaneous decay, 152
Quantum gates, 211 Squeezed state, 130
Quantum internet, 242 Squeezed vacuum, 131
Quantum Key Distribution (QKD), 239 Squeezing Hamiltonian, 130
Quantum measurement, 219, 220 Squeezing interaction, 187
Quantum noise, 275, 278 Squeezing operator, 130
298 Index

Stabilizer generator, 228 Two-mode-squeezing interactions, 187, 203

Stabilizer group, 228 Two-qubit gate, 213, 219
Stabilizer subspace, 228
Stimulated emission, 145 U
Strong coupling regime, 177 Unitary operator, 58
Strong dispersive regime, 149, 179 Universal set of quantum gate, 211
Superconducting quantum bit, 197
Superconducting qubit, 149 V
Surface code, 231 Vacuum amplitude, 125
SWAP gate, 214 Vacuum fluctuation, 185
Syndrome measurement, 230 Vacuum Rabi splitting, 173
System-bath interaction, 259 Von Neumann entropy, 282
Von Neumann equation, 89, 150, 258
T
Tavis-Cummings Hamiltonian, 174 W
Tavis-Cummings model, 173 Wavefunction, 26
T gate, 213, 218 Waveguide-coupled cavity QED, 179
Thermal distribution, 132 Weak coupling regime, 176
Thermal state, 63, 132 Werner state, 284
Three level system, 251 Wigner function, 135
Three-qubit repetition code, 229 W state, 49
Three-wave mixing, 188
Threshold theorem, 231 X
Time-dependent perturbation theory, 107 X gate, 212
Time-independent Schrödinger equation, 74 X X gate, 215
Trace operation, 259 χ-matrix, 224
Transition dipole moment, 190 χ(2) interaction, 188
Transition frequency, 142 χ(3) interaction, 188
Transmission measurement, 163 χ(2) nonlinearity, 130
Transmon, 196, 199
Transverse relaxation, 146, 152 Z
Trapped ion, 147 Zeeman effect, 104
Traveling photon, 166 Zero-point fluctuation, 125
Twirling, 227 Z gate, 212
Two-level system, 141 Z Z gate, 215