Teoria Grupos Quantum Mechanics

MasahitoHayashi
Group
Representation
for Quantum
Theory
Group Representation for Quantum Theory
Masahito Hayashi
Group Representation
for Quantum Theory
123
Masahito Hayashi
Graduate School of Mathematics
Nagoya University
Nagoya
Japan
ISBN 978-3-319-44904-3
DOI 10.1007/978-3-319-44906-7
ISBN 978-3-319-44906-7
(eBook)
Library of Congress Control Number: 2016950886

Springer International Publishing Switzerland 2017
Springer received copyright for English version only.
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microlms 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 specic 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, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made.
Printed on acid-free paper
This Springer imprint is published by Springer Nature
The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
This book is the English edition of the Japanese book Group Representations for
Quantum Theory, which was originally published by Kyoritsu shuppan, Tokyo,
Japan in January 2014. The original Japanese book covers several topics in representation theory that is related to quantum theory. As is well known, group
representation theory is a very strong tool for quantum theory, in particular, angular
momentum, hydrogen-type Hamiltonian, spinorbit interaction, quark model,
quantum optics, and quantum information processing including quantum error
correction. Therefore, many departments of physics have lecture courses for
mathematics for physics, in particular, they have a graduate lecture course for
representation theory for physics. This book conducts lecture courses on mathematics for physicists. When the contents of this book are too much for the lecture
course, this book can be used by skipping several detailed parts.
To describe a big picture of application of representation theory to quantum
theory, the book needs to contain the following six topics, permutation group, SU(2)
and SUd, Heisenberg representation, squeezing operation, Discrete Heisenberg
representation, and the relation with Fourier transform from a unied viewpoint by
including projective representation. Unfortunately, although there are so many good
mathematical books for a part of six topics, no book contains all of these topics
because they are too segmentalized. Further, some of them are written in an abstract
way in mathematical style and, often, the materials are too segmented. At least, the
notation is not familiar to people working on quantum theory. Others are good
elementary books, but do not deal with topics related to quantum theory. In particular, such elementary books do not cover projective representation, which is more
important in quantum theory. On the other hand, there are several books for
physicists. However, these books are too simple and lack the detailed discussion.
Hence, they are not useful for advanced study even in physics.
To resolve this issue, the author published the Japanese version of this book. It
starts with the basic mathematics for quantum theory. Then, it proceeds to the
foundation of group representation theory by including nite group. During this
discussion, this book deals with mathematical formulation of boson and fermion,
which are fundamental concepts in quantum theory. After this preparation, it
v
vi
Preface
discusses representation theory of Lie group and Lie algebra. Based on these
mathematics, this book addresses bosonic system and its discretization, which are
more related to quantum optics and quantum information, respectively.
However, the original Japanese version has less application to quantum physics.
The author has newly added Chap. 5, the later half of Chap. 1, crucial contents of
Sect. 4.4, and the discussion for Wigner functions in Chap. 7 as follows. Originally,
Chap. 1 discusses only the minimum preparation of mathematics of quantum theory. For smooth connection from basic mathematics to representation theory, the
author has added a section for Hamiltonian, which explains the role of Hamiltonian
and the relation to representation theory. Also, the author has added a section for
symmetry, which describes the details of roles of representation theory in quantum
theory. Additionally, the author has added a section for the unbounded case, which
is applied only to the innite dimensional case and can be skipped when the reader
is not interested in the difculty of the innite-dimensional case. Originally, Sect. 4.
4 briefly explained boson and fermion based on the representation theory for Lie
group and Lie algebra. In this English version, the author has added several helpful
examples including spin, which clarify the need for representation theory for
understanding boson and fermion. Unfortunately, many existing books explain
boson and fermion without group representation. The author believes that this
section is helpful for the study of boson and fermion on nite-dimensional system.
Chapter 5 is devoted to applications of the representation theory to physical
systems. This additional chapter starts with the spectral decomposition of the
Hamiltonian on 3-dimensional physical system with a rotational invariant potential
by using the group SO3; R. As a generalization of a part of this topic, we proceed
to the spectral decomposition of the Laplacian on the general dimensional sphere,
which is essentially the same as the Hamiltonian. Hydrogen-type atom is a special
case of 3-dimensional physical system with a rotational invariant potential. The
Hamiltonian of this case has a hidden symmetry of SO4; R. Indeed, this
Hamiltonian has a large degeneracy, which cannot be explained by the visible
symmetry of SO3; R. The hidden symmetry of SO4; R explains such a large
degeneracy. Then, it proceeds to the spin-orbit interaction, in which, the irreducible
decomposition of group representation on the composite system (Clebsch Gordan
coefcient) plays an important role. This kind of system frequently appears in spin
magnetics. Finally, this chapter addresses quark model, which is the key concept of
fundamental particle theory. The discussion on the quark model is composed of the
analysis on nite-dimensional system, which can be regarded as a good exercise of
fermion and group representation theory of a nite-dimensional system. To discuss
quark model precisely, we need to discuss the notion of flavor, spin, and color,
simultaneously. Although we need many preparations to discuss the whole topic
thoroughly and consistently, this book covers all the essential preparation. We also
note that several books on the quark model discuss the subject inconsistently due to
a lack of basic preparation at the beginning of the books. In particular, this book
explicitly writes down the wave functions of all of Baryon and Meson with SU3
symmetry while they are not given in many basic books for particle physics.
This description will be helpful for students of particle physics. In the nal section
Preface
vii
of Chap. 5, we discuss uncertainty relation for wake packets on various spaces, i.e.,
not only on the set of real numbers R but also on the one-dimensional and
three-dimensional spheres S1 and S3 .
Additionally, the author has newly added recent progresses for design theory as
discretization of Lie group and homogeneous space in Sect. 4.5 and recent progresses for mutually unbiased bases (MUB) and symmetric informationally complete (SIC) vectors in Sect. 8.4. Another unique feature of this book is the
clarication on the relation between group representation and Fourier analysis in
Sects. 2.8 and 3.8. Fourier analysis is another important mathematical structure of
quantum physics. In this revision, based on this relation, in Chap. 7, the author has
added the description for Winger function, which is a key concept for the duality
between the position and the momentum in foundation of quantum theory. Then, we
discuss the uncertainty relation for Winger function. Since the uncertainty relations
on the one-dimensional and three-dimensional spheres S1 and S3 and on Winger
function were obtained recently, these topics have not been discussed in other books.
So, the reader can understand the duality in the phase space more deeply.
Further, for better understanding, the author has added many gures, tables, and
exercises to help the reader to understand the materials better so that this book
contains 54 gures, 23 tables, and 111 exercises with solutions. This book is
organized as follows. First of all, when the section or the example is too advanced,
the symbol * is indicated in the title. Since such parts will be used only in the later
parts with the symbol *, the reader can still understand the contents well even if the
reader omit them in the rst reading. The author recommends a beginner of representation theory to omit these parts in the rst time. Since the symbols of representations are too complicated, this book summarizes the symbols as several
tables (Tables 1-8). The reader can refer the tables to recall the symbols.
We now describe the whole structure of this English edition. As the author
explained before, this book treats projective representation as well as conventional
representation. Chapter 1 starts the mathematical preparation for quantum theory,
and explains the back ground of group representation theory. Chapter 2 introduces
group representation theory with fundamental concepts. Chapter 3 deals with
general theories that do not depend on the types of Lie groups and Lie algebras.
Then, it introduces the Fourier transform for Lie groups. Chapter 4 treats representations of special Lie groups and special algebras, SU2, SU1; 1, and SUd.
Based on them, we discuss boson and fermion as indistinguishable particles in
Sect. 4.4. We also discuss the coherent states, which are not discussed in the
conventional textbooks. Using these preparations, Chap. 5 proceeds the
above-mentioned applications. As advanced topics, Chap. 6 deals with representations of Lie group and Lie algebra with general form based on a root system
including representations of non-compact Lie group. Since these topics are more
advanced, the proofs of many theorems are omitted. The reader may need to spend
a long time to understand some of these proofs. This chapter summarizes such
advanced results very reasonably so that a reader can grasp the contents intuitively
based on analogies with simple cases, which is another advantage of this book.
viii
Preface
Chapter 7 deals with Heisenberg representation, which gives the bosonic system
and plays an important role in quantum optics. Using the bosonic system, this
chapter explains second quantization, which is a key concept of quantum eld
theory. In the end of Chap. 7, we discuss multi-mode squeezing, which requires
knowledge for representations of Lie group and Lie algebra based on a root system,
which are explained in Chaps. 3, 4, and 6. Chapter 8 deals with discrete Heisenberg
representation as a discrete version of Heisenberg representations addressed in
Chap. 7. This representation is useful in quantum information, especially, quantum
error correcting codes, and designs of quantum circuits. Since any quantum information process is constructed based on a combination of quantum circuits, designs
of quantum circuits are crucial for quantum information. Chapter 8 is organized so
that the contents can be understood with the contents of Chaps. 1 and 2.
Finally, the author emphasizes the difference from existing books for representation theory as follows. There are so many books for representation theory. At
least, there exist several good books containing a part of the contents of this book,
in particular, the major parts of Chaps. 3, 4, and 6. However, no book contains the
whole contents of this book. Further, many existing mathematical books do not
adopt the notation familiar to physicist. Mathematical books containing the detail of
projective representation are often too advanced. Typically, such books for representation theory do not explain the relation between the representation theory and
the foundation of quantum theory, e.g., boson, fermion, second quantization,
Winger function, and quantum circuits. Indeed, since representation theory requires
too complicated notations, students have trouble to interpolate notations across
several books by themselves. To resolve this problem, they need a single book that
incorporates representation theory, which brings them a big picture of quantum
theory. Therefore, the author believes that this book is helpful for students for
representation theory for quantum theory.
The author is grateful when the readers would be interested in representation
theory and quantum theory via this book. Finally, the author expresses the
acknowledgments to all persons who cooperate to this English version. Especially,
the author would like to thank Prof. Hideyuki Ishi of Nagoya University, Prof.
Kwek Leong Chuan of Nanyang Technological University and National University
of Singapore, Prof. Serge Richard of Nagoya University, Dr. Huangjun Zhu of
University of Cologne, Prof. Soichi Okada of Nagoya University, Prof. Hiroyuki
Kanno of Nagoya University, Prof. Toru Uzawa of Nagoya University, and CafDavid, which is mathematics salon in Graduate School of Mathematics, Nagoya
University for providing helpful comments for this English version. The author
would also like to thank Dr. Claus E. Ascheron of Springer Science+Business
Media for his encouragement and patience during the preparation of the manuscript.
Nagoya, Japan
Masahito Hayashi
Preface to the Japanese Version
Group representational symmetry is one of most fundamental concepts in quantum

theory, and has been applied to various areas in physics, e.g., particle physics,
nuclear physics, condensed matter physics, and statistical physics. It also plays an
important role in quantum information, which enables us fruitful information
processing by using quantum phenomena. Especially, since various types of representations of various groups appear in various quantum systems, group representation theory reveals so many aspects of quantum theory. Unfortunately, besides
various topics underlie group representation theory, such relations are not sufciently recognized. In fact, such a recognition often leads us not only to deeper
understanding of the topic but also to generalization of the topic. However, since
the methods of operator algebra and partial differential equation have been
emphasized in the area of mathematical foundation of quantum theory, that of
group representation has not taken a sufciently prominent position in this area.
Hence, few books cover various methods in representation theory in a unied
viewpoint. On the other hand, representation theory has been signicantly developed in various directions as a part of mathematics. However, such fruitful
developments are not accessible for students and researchers of quantum theory due
to the problems explained later.
As useful knowledge of representation theory for quantum theory is divided into
various subtopics in representation theory, they are so segmentalized that no
mathematical book provides a simple collection of such knowledges from a unied
viewpoint. If we deal with representation theory from mathematics, we usually
specify an individual topic of representation theory. As its own characteristic features, representation theory inevitably employs very complicated symbols. When
we address a topic across distinct books, we need to spend much effort to adjust the
different notations among these books. Especially, representations of real Lie
groups play an important role in quantum theory, and they are classied via representations of real Lie algebras. However, since many mathematical introductory
books are written based on representations of complex Lie algebras, only a few
mathematical introductory books emphasize real Lie algebras. In fact, although
representations of real Lie algebras are obtained from representations of complex
ix
Lie algebras via a suitable conversion, few introductory books describe this conversion carefully. In particular, many mathematicians avoid to handle representations of real Lie algebras with the above conversion because it requires more
complicated notations while it is not mathematically interesting. Since there are so
many elegant mathematical introductory books for individual subtopics of representation theory in the above way, we need a book to connect these individual
books while few mathematical introductory books carefully describe such connection parts.
On the other hand, many books in physics care such a point; however, they omit
their descriptions so much that the reader cannot understand the contents precisely.
Further, projective representations play an central role in quantum theory, but many
mathematical books do not cover them sufciently. Especially, major aspects of
projective representation can be treated by trivial extension, but projective representation has several blind points that requires special treatment different from
conventional representation. Indeed, such subtle points are linked to variety of
quantum phenomena. As another problem, many mathematical books are written in
a too generalized form, and do not explain examples related to quantum theory.
Also, their descriptions are far from the description of quantum theory. There are so
many factors that inhibit researchers of quantum theory from accessing useful
results of representation theory in this way.
The contents of this book are composed of mathematical knowledge for representation theory that are well-known for many mathematicians, and reorganized by
using notations of quantum theory so that they can be easily applied to various
topics in quantum system from the viewpoint of quantum theory. Since this book
deals with various topics of representation theory essential for quantum system, the
whole structure of representation theory will be claried from quantum theory. This
book emphasizes the similarities among several related topics so that their analogies
enable the readers to easily understand some complicated topics based on related
simpler topics. Hence, the reader will grasp the key points of these advanced topics
of representation theory easily. That is, this book will work as a basic infrastructure
to understand quantum theory from representation theory. Since the author studies
quantum information mainly, the contents are related to applications to quantum
information. However, as quantum information is related to foundation of quantum
theory, they will be useful for understanding quantum theory beyond quantum
information. Therefore, once the readers complete to read this book, they will be
able to understand quantum theory much more deeply based on the representation
theory. Also, they can proceed to read another book A Group Theoretic Approach
to Quantum Information.
This book is organized as follows. First of all, when the section or the example is
too advanced, the symbol * is indicated in the title. Since such parts will be used
only in the latter parts with the symbol *, the reader can understand the contents
except for such parts even if the reader omits them. The author recommends a
beginner of representation theory to omit these parts in the rst time. Since the
symbols of representations are too complicated, this book summarizes the symbols
xi
as several tables (Tables 18). The reader can refer the tables to remember the
symbols.
Chapter 1 introduces basic concepts of quantum theory, measurement, state,
composite system, and entanglement. It also prepares mathematical notations for
quantum systems. Although these notations are specied to quantum systems, they
are helpful for group representation. Hence, this book consistently deals with
representation theory based on these notations. Although the second quantization is
an important topic in quantum eld theory, we explain it in Chapter 6 (new Chapter
7) because it needs the Bosonic system.
Chapter 2 discusses representations for group in a general framework including
projective representations, which are important in quantum theory. Since projective
representation is closely related to extension of group, this chapter focuses on this
relation. To discuss representation theory including projective representations, we
need to handle the factor system, i.e., the set of phase factors, which requires
complicated notations. Since the discussions with projective representations are
complicated and do not seem essential for mathematics, many mathematically
standard textbooks omit them. However, since such discussions are essential for
quantum theory, this chapter handles projective representations by using factor
systems and we keep this style during the whole book, which is a distinct point
from other related books. This chapter proceeds to the details of representations of
nite groups so that it introduces the Fourier transform for nite groups, which
connects analysis and algebra. As a typical example, we analyze representations of
a permutation group by using Young diagrams.
Chapters 3, 4, and 5 address representation theory for Lie groups and Lie
algebras. Especially, Chapter 3 deals with general theories that do not depend on the
types of Lie groups and Lie algebras. Chapter 3 treats projective representations of
Lie groups and Lie algebras by combining the contents of Chapter 2 although few
traditional introductory textbooks touch them. Then, it introduces the Fourier
transform for Lie groups including the case of projective representations. It also
prepares several concepts for Chapter 6 (new Chapter 7). Also, Chapter 3 introduces complex Lie groups and complex Lie algebras.
Chapter 4 treats representations of special Lie groups and special algebras. Since
representations of a Lie algebra can be classied with maximum weight, those of a
Lie group can be easily treated through those of the corresponding Lie algebra.
Also, Lie algebras provide several concepts important for quantum theory. Hence,
Chapter 4 is organized so that it constructs representation of a Lie group via that
of the corresponding Lie algebra. Chapter 4 starts with representations of Lie
algebras su2 and su1; 1. Because the Lie algebra su2 is compact and the Lie
algebra su1; 1 is not compact, they require different treatment caused in this
difference. As they have unexpected common features, we handle both in a unied
way. Then, we proceed to representation of the Lie algebra sur by using Young
diagrams. Especially, the representation of the Lie algebra sur on the tensor
product space is closely related to that of the permutation group on the same tensor
xii
product space. The relation is called Schur duality. We also consider how a nite
subgroup can take a role of the Lie group when its representation is given. Such a
problem is called design, and is discussed in this chapter.
Chapter 5 (new Chapter 6) deals with representations of a Lie group via those
of the corresponding Lie algebra. This chapter discusses a noncompact Lie group as
well as a compact Lie group. To discuss both, we focus on the relation between a
(real) Lie algebra and a complex Lie algebra. Since representations of a noncompact
group are very complicated, this chapter treats only a part of its representations that
are related to quantum theory. Since such a special representation has analogies
with representations of a compact Lie group, they can be more easily understood
than the general case. As this chapter is composed of very advanced topics and such
sections are labeled with *, The author recommends the reader to omit this chapter
in the rst time. Indeed, this book is organized so that the reader can understand the
large part even though such advanced parts are omitted.
Chapter 6 (new Chapter 7) deals with Heisenberg representation, which gives the
Bosonic system. Since Heisenberg representation is projective representation, the
general theory for projective representation given in Chapter 3 plays an essential
role. It also treats the representation of su1; 1 that describes the squeezing
operations. We discuss the multi-mode Bosonic system as well as the one-mode
Bosonic system. Such a quantum system is called the continuous system, and
attracts attention because it can be easily implemented in a particular sense. Further,
using the Bosonic system, we explain the second quantization and the physical
meaning of the tensor product state. Only sections with symbol * require knowledge given in the latter part of Chapter 5.
Chapter 7 (new Chapter 8) deals with discrete Heisenberg representation as a
discrete version of the representations addressed in Chapter 6 (new Chapter 7).
Although discrete Heisenberg representation is usually constructed from nite
elds, there is another type of discrete Heisenberg representation based on the
commutative algebra Zd . So, this book covers both types of discrete Heisenberg
representations. Further, we address a discrete version of the representation corresponding to the squeezing operation given in Chapter 6 (new Chapter 7). This
representation is closely related to quantum circuits and quantum error correction.
While Chapter 7 (new Chapter 8) is related to Chapter 6 (new Chapter 7), this
chapter can be read independently of Chapter 6 (new Chapter 7). Since it addresses
representations of nite groups, it can be read after Chapter 2. The nal section
discusses mutually unbiased bases (MUB) and symmetric informationally complete
(SIC) vectors, which have several applications.
In the above-mentioned way, this book summarizes knowledges of representation theory that is useful for quantum theory. Then, based on these, this book also
explains the foundation of quantum theory, e.g., second quantization and quantum
circuits. The author is grateful when the readers would be interested in representation theory and quantum theory via this book.
xiii
Finally, the author expresses the acknowledgments to all persons who cooperate
to this book. Especially, the author would like to thank Prof. Hideyuki Ishi of
Nagoya University, Dr. Wataru Kumagai of Kanagawa university, Prof. Akito Hora
of Hokkaido University, and Dr. Tsuyoshi Miezaki of Yamagata University for
providing helpful comments. The author would also like to thank Mr. Hideo
Kotobuki and Ms. Kei Akagi of Kyoritsu Shuppan for their supports.
Nagoya, Japan
December 2013
Masahito Hayashi
Contents
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
1
6
6
7
10
11
11
12
13
14
15
2 Group Representation Theory . . . . . . . . . . . . . . . . . . . . . .

2.1 Group and Homogeneous Space . . . . . . . . . . . . . . . . .
2.1.1 Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.2 Homogeneous Space . . . . . . . . . . . . . . . . . . .
2.2 Extension of Group . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1 General Case . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.2 Central Extension of a Commutative Group .
2.2.3 Examples for Central Extensions . . . . . . . . . .
2.3 Representation and Projective Representation . . . . . . .
2.3.1 Denitions of Representation and Projective
Representation . . . . . . . . . . . . . . . . . . . . . . . .
2.3.2 Schurs Lemma . . . . . . . . . . . . . . . . . . . . . . .
2.3.3 Representations of Direct Product Group . . .
2.4 Projective Representation and Extension of Group . . .
2.4.1 Factor System of Projective Representation . .
2.4.2 Irreducibility and Projective Representation . .
2.4.3 Extension by U1 . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
21
21
21
24
28
28
33
34
37
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
37
41
44
45
45
47
48
1 Mathematical Foundation for Quantum System . . . .

1.1 System, State, and Measurement. . . . . . . . . . . . .
1.2 Composite System . . . . . . . . . . . . . . . . . . . . . . .
1.2.1 Tensor Product System . . . . . . . . . . . . .
1.2.2 Entangled State . . . . . . . . . . . . . . . . . . .
1.3 Many-Body System . . . . . . . . . . . . . . . . . . . . . .
1.4 Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4.1 Dynamics and Hamiltonian . . . . . . . . . .
1.4.2 Simultaneous Diagonalization . . . . . . . .
1.4.3 Relation to Representation . . . . . . . . . . .
1.5 Relation to Symmetry . . . . . . . . . . . . . . . . . . . . .
1.6 Remark for Unbounded Case* . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
xv
xvi
Contents
2.5
2.6
2.7
2.8
2.9
Semi Direct Product and Its Representation . . . . . . . . . . . . . . . . .

2.5.1 From H o K to K and H . . . . . . . . . . . . . . . . . . . . . . . . .
2.5.2 From K and H to H o K . . . . . . . . . . . . . . . . . . . . . . . .
Real Representation and Complex Conjugate Representation . . . .
2.6.1 Real Linear Space and Its Complexication . . . . . . . . . .
2.6.2 Real Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.6.3 Complex Conjugate Representation . . . . . . . . . . . . . . . . .
Representation on Composite System . . . . . . . . . . . . . . . . . . . . . .
Fourier Transform for Finite Group . . . . . . . . . . . . . . . . . . . . . . .
2.8.1 Discrete Fourier Transform . . . . . . . . . . . . . . . . . . . . . . .
2.8.2 Character and Orthogonality . . . . . . . . . . . . . . . . . . . . . .
2.8.3 Fourier Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Representation of Permutation Group and Young Diagram . . . . .
2.9.1 Young Diagram and Young Tableau . . . . . . . . . . . . . . . .
2.9.2 Permutation Group and Young Diagram . . . . . . . . . . . . .
2.9.3 Plancherel Measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
49
49
50
52
52
52
53
55
58
58
59
63
64
64
66
67
3 Foundation of Representation Theory of Lie Group

and Lie Algebra. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.1 Lie Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.1.1 Basic Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.1.2 Symmetry in Analytical Mechanics . . . . . . . . . . . . . . . . . 72
3.1.3 Complex Lie Group. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.1.4 Other Examples of Real Lie Groups . . . . . . . . . . . . . . . . 74
3.2 Lie Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.3 Relation Between Lie Group and Lie Algebra I . . . . . . . . . . . . . . 78
3.3.1 Innitesimal Transformation and Lie Algebra . . . . . . . . . 78
3.3.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.3.3 Central Extension of Real Lie Algebra . . . . . . . . . . . . . . 83
3.4 Representation of Lie Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.4.1 Representation of Real Lie Algebra . . . . . . . . . . . . . . . . . 84
3.4.2 Real Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3.4.3 Representation of Complex Lie Algebra . . . . . . . . . . . . . 87
3.4.4 Adjoint Representation . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.4.5 Projective Representation. . . . . . . . . . . . . . . . . . . . . . . . . 88
3.4.6 Semi Direct Product Lie Algebra and Representation . . . 89
3.5 Killing Form and Compactness. . . . . . . . . . . . . . . . . . . . . . . . . . . 90
3.5.1 Killing Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
3.5.2 Compactness of Real Lie Algebra g . . . . . . . . . . . . . . . . 91
3.5.3 Casimir Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.6 Relation Between Lie Group and Lie Algebra II . . . . . . . . . . . . . 95
3.6.1 Universal Covering Group . . . . . . . . . . . . . . . . . . . . . . . . 95
3.6.2 Relation to Representation . . . . . . . . . . . . . . . . . . . . . . . . 98
3.6.3 Projective Representation. . . . . . . . . . . . . . . . . . . . . . . . . 100
3.6.4 Representation for Complex Lie Groups . . . . . . . . . . . . . 101
Contents
3.7
3.8
xvii
Invariant Measures on Group and Homogeneous Space . . . .

Fourier Transform on Lie Group . . . . . . . . . . . . . . . . . . . . .
3.8.1 Commutative Case . . . . . . . . . . . . . . . . . . . . . . . . .
3.8.2 Non-commutative Case . . . . . . . . . . . . . . . . . . . . . .
4 Representations of Typical Lie Groups and Typical

Lie Algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1 SL2; C and Its Subgroup . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.1 Gauss Decomposition of SL2; C . . . . . . . . . . . . . .
4.1.2 Structure of SU2. . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.3 Structure of SU1; 1 . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Irreducible SkewHermitian Representation of su2
and su1; 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.1 Construction of Irreducible SkewHermitian
Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.2 Unitary Representation and Real Representation
of Lie Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.3 Casimir Operator . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.4 Tensor Product . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.5 Coherent Vector . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3 Irreducible Unitary Representations of SUr and Ur . . . .
4.3.1 Irreducible Unitary Representation of SUr . . . . . .
4.3.2 Irreducible Unitary Representation of Ur . . . . . . .
4.3.3 Dimension of Irreducible Representation Space . . . .
4.3.4 Tensor Product . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.5 Irreducible Representation of SLr; C, GLr; C . .
4.4 Many-Particle System: Composite System of n Quantum
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.4.1 Distinguishable Particles: Schur Duality . . . . . . . . .
4.4.2 Indistinguishable Particles: Boson and Fermion . . . .
4.5 Discretization of Lie Group and Homogeneous Space . . . . .
4.5.1 Design on Group . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.5.2 Design on Homogeneous Space . . . . . . . . . . . . . . .
4.5.3 Approximating Design. . . . . . . . . . . . . . . . . . . . . . .
5 Application to Physical Systems . . . . . . . . . . . . .
5.1 3-Dimensional Physical System . . . . . . . . .
5.1.1 Orbital Angular Momentum . . . . .
5.1.2 Hamiltonian with Central Potential
5.2 Laplacian on Sphere . . . . . . . . . . . . . . . . . .
5.2.1 General Case . . . . . . . . . . . . . . . . .
5.2.2 Three-Dimensional Sphere . . . . . . .
5.3 Hydrogen-Type Hamiltonian and SO4; R
5.3.1 Puzzle for Degeneracy . . . . . . . . . .
5.3.2 Symmetry Based on SO4; R . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
101
104
104
107
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
113
113
114
114
116
....
119
....
119
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
122
123
124
127
129
129
131
133
134
136
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
137
137
139
143
143
147
148
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
151
151
151
156
158
158
162
164
164
166
xviii
5.4
5.5
5.6
Contents
5.3.3 Fourier Transform of Coulomb Potential* . . . . . . . .

Spin-Orbit Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Quark Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5.1 Quark and Anti-quark . . . . . . . . . . . . . . . . . . . . . . .
5.5.2 Meson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5.3 Baryon. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5.4 Charm Quark . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5.5 Six Quarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Uncertainty Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.6.1 Uncertainty Relation on R . . . . . . . . . . . . . . . . . . . .
5.6.2 Periodic Function Space and Mathieu Equation* . . .
5.6.3 Uncertainty Relation on S1 * . . . . . . . . . . . . . . . . . .
5.6.4 Uncertainty Relation on S3 * . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
6 Representation of General Lie Groups and General

Lie Algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.1 Root System and Its Classication* . . . . . . . . . . . . . . . . . . . . . . .
6.2 Structure of Compact Lie Algebra* . . . . . . . . . . . . . . . . . . . . . . .
6.3 Structure of Complex Semi Simple Lie Algebra* . . . . . . . . . . . . .
6.3.1 Complex Semi Simple Lie Algebra . . . . . . . . . . . . . . . . .
6.3.2 Relation to Compact Semi Simple Lie Algebra . . . . . . . .
6.4 Skew-Hermitian Representation of Compact Semi Simple Lie
Algebra* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.5 Subalgebra and Root Subsystem* . . . . . . . . . . . . . . . . . . . . . . . . .
6.6 Structure of spr* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.7 Complex Structure and Coherent State* . . . . . . . . . . . . . . . . . . . .
6.8 Structure of Non-compact Semi Simple Lie Algebra* . . . . . . . . .
6.9 Skew-Hermitian Representation of Non-compact Simple Lie
Algebra* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.10 General Non-compact Lie Algebra and Its Skew-Hermitian
Representation* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7 Bosonic System and Quantum Optics . . . . . . . . . . . . . . . . . . . . .
7.1 One-Mode Bosonic System . . . . . . . . . . . . . . . . . . . . . . . . .
7.1.1 One-Variable System . . . . . . . . . . . . . . . . . . . . . . . .
7.1.2 One-Mode Bosonic System and One-Variable
System. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.1.3 Extended Heisenberg Representation . . . . . . . . . . . .
7.2 Multi-mode Bosonic System . . . . . . . . . . . . . . . . . . . . . . . . .
7.2.1 Multi-variable System . . . . . . . . . . . . . . . . . . . . . . .
7.2.2 Relation Between Multi-mode Bosonic System
and Multi-variable System . . . . . . . . . . . . . . . . . . . .
173
175
177
177
179
181
183
185
185
186
187
188
193
201
201
205
208
208
210
211
213
219
221
223
225
227
....
....
....
231
231
231
.
.
.
.
.
.
.
.
233
237
239
239
....
240
.
.
.
.
.
.
.
.
Contents
7.3
xix
....
....
242
242
....
245
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
246
248
249
253
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
255
255
257
259
8 Discretization of Bosonic System . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

8.1 Discrete Heisenberg Representation . . . . . . . . . . . . . . . . . . . . . . .
8.1.1 Discrete Heisenberg Representation on Zd . . . . . . . . . . .
8.1.2 Discrete Heisenberg Representation on Fq . . . . . . . . . . . .
8.1.3 Multi-mode Discrete Heisenberg Representation . . . . . . .
8.1.4 Irreducible Representation of Subgroup . . . . . . . . . . . . . .
8.2 Discrete Symplectic Group and Clifford Group . . . . . . . . . . . . . .
8.2.1 Denitions of Discrete Symplectic Group and Clifford
Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2.2 Generator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2.3 Structure of Self-orthogonal Subgroup . . . . . . . . . . . . . .
8.3 Metaplectic Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.3.1 Existence of Projective Unitary Representation . . . . . . . .
8.3.2 Concrete Construction of Representation . . . . . . . . . . . . .
8.3.3 Properties of Irreducible Decomposition . . . . . . . . . . . . .
8.4 MUB and SIC Vectors* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
263
263
263
266
268
271
273
7.4
7.5
7.6
7.7
7.8
Transformation by Unitary Group Ur . . . . . . . . . . . . . . . .

7.3.1 Correlation by Beam Splitter . . . . . . . . . . . . . . . . . .
7.3.2 Transformation Corresponding to the Irreducible
Decomposition. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.3.3 Representation of Semi Direct Product Lie Group
H2r; Ro Ur and Its Coherent State* . . . . . . . . .
Second Quantization and Boson . . . . . . . . . . . . . . . . . . . . . .
Squeezed State on One-Mode Bosonic System . . . . . . . . . . .
Squeezed State on Two-Mode Bosonic System* . . . . . . . . .
Wigner Function and Fourier Transform of Heisenberg
Representation* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.7.1 Fourier Transform of Heisenberg Representation . . .
7.7.2 Uncertainty Relation with Wigner Function . . . . . .
Squeezed State on Multi-mode Bosonic System* . . . . . . . . .
273
274
277
278
278
279
281
285
Appendix A: Solutions of Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
Curriculam Vitae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333
About the Author
Masahito Hayashi was born in Japan in 1971. He received his B.S. degree from
the Faculty of Sciences in Kyoto University, Japan, in 1994 and his M.S. and Ph.D.
degrees in Mathematics from Kyoto University, Japan, in 1996 and 1999, respectively.
He worked in Kyoto University as a Research Fellow of the Japan Society of the
Promotion of Science (JSPS) from 1998 to 2000, worked in the Laboratory for
Mathematical Neuroscience, Brain Science Institute, RIKEN from 2000 to 2003,
worked in ERATO Quantum Computation and Information Project, Japan Science
and Technology Agency (JST) as the Research Head from 2000 to 2006. He also
worked in the Superrobust Computation Project Information Science and
Technology Strategic Core (21st Century COE by MEXT) Graduate School of
Information Science and Technology, The University of Tokyo as Adjunct
Associate Professor from 2004 to 2007. He worked in the Graduate School of
Information Sciences, Tohoku University as Associate Professor from 2007 to
2012. In 2012, he joined the Graduate School of Mathematics, Nagoya University
as Professor. He also worked in the Centre for Quantum Technologies, National
University of Singapore as Visiting Research Associate Professor from 2009 till
now. In 2011, he received the Information Theory Society Paper Award (2011) for
Information-Spectrum Approach to Second-Order Coding Rate in Channel Coding.
In 2016, he received the Japan Academy Medal from the Japan Academy and the
JSPS Prize from Japan Society for the Promotion of Science.
He is a member of the Editorial Board of the International Journal of Quantum
Information and International Journal on Advances in Security. His research
interests include classical and quantum information theory, information-theoretic
security, and classical and quantum statistical inference.
xxi
Symbols
Table 1 Symbols for groups (see Tables 3.1, 3.2, and 3.3 for other groups)
Symbol
Name (Denition)
Page
HK
HoK
HK : H
Ker f
G; G
H2r; Zd ,
H2r; Fq
CG M
CG
NG M
U1
Uk
Direct product group of groups H and K

Semi direct product of groups H and K
Group of extension of group H by group K
Kernel of homomorphism f
Commutator subgroup of G
Discrete Heisenberg group of degree r
28
28
34
27
27
35, 36
Centralizer of subgroup M of group G

Center of G (CG G)
Normalizer of M
Unit circle
26
26
26
23
23
Sn
GLH
UH
SUH
G0
1 G
~
G
^
G
G!
Sp2r; Fq ,
Sp2r; Zd
Subgroup fei2j=k jj 2 Zg of U1
Permutation group of degree n
General linear group on H
Unitary group on H
Special unitary group on HUH\ SLH)
Connected component of G including identity element e
Fundamental group
Extension of commutative group C by G (universal covering
group when C is 1 G)
Double covering group of G
23
37
37
37
70
95
46, 47, 98
97
Extension of group G by U1 dened by 2 CG
49
Discrete symplectic group
273
(continued)
xxiii
xxiv
Symbols
Table 1 (continued)
Symbol
Name (Denition)
Page
G=H
25
Fk
q
Quotient space, quotient group (residue group) when H is

normal subgroup
Z=kZ
Finite eld of order prime p (Z=pZ)
Algebraic extension over Fp of degree n (q pn )
k-dimensional subspace spanned by e1 ; . . .; ek over Fq
Fq or Zd
268
Zk
Fp
Fq
26
36
36
277
Table 2 Symbols for Lie algebras (see Tables 3.4, 3.5, and 3.6 for other Lie algebras)
Symbol
Name (Denition)
Page
g1 g2
g1 o g2
gC
cg
h
k
glH
uH
oV
Direct sum of g1 and g2

Semi direct product of g1 and g2
Complexication of g
Center of g
Cartan subalgebra
Maximal compact subalgebra
Set of linear maps from H to H
Set of skew-Hermitian matrices on H
Set of real alternative matrices on V
77
90
77
78
205
223
37
85
86
Table 3 Sets
Symbol
^
G
Name (Denition)
Page
Set of irreducible unitary representations of G
41
^f
G
ig;g0
^
Gfe
gg;g0
^
Set of nite-dimensional representations in G
Set of irreducible projective unitary representations of factor
0
system feig;g gg;g0 2 HG : U1
48
^ f feig;g0 g 0
G
g;g
Set of nite-dimensional projective representations in

ig;g0
^
Gfe
gg;g0
48
^g
^gf
Yr
Yn
Y rn
YU;r n
YS n
Set of irreducible skew-Hermitian presentations of Lie algebra g

Set of nite-dimensional representation in ^
g
Set of Young diagram of depth not greater than r
Set of Young diagram of size n
Y r \Y n
Set of Semistandard Young tableau corresponding to n
Set of Standard Young tableau corresponding to n
Root system
Root system of Lie algebra g
99
99
65
65
65
65
65
130, 201
209
Fundamental system
Unit circle
130, 202
117
41
Symbols
xxv
Table 4 Symbols for Representation

Symbol
Name (Denition)
Group (Lie
algebra)
Representation
space
Page
f1 f2
f1 f 2
Direct sum representation

Tensor product
representation
Gg
Gg
H1 H2
H1 H2
39, 86
39, 86
G1 G2
Gg
HoK
H1 H2
g
44
87
50
2
f1 f
ad
fH ofK
Adjoint representation
Semi direct product
representation
~ T
fH of
HoK
HoK
fH ofK f0K
fH of
~ T f0T
50
51
HoK
51
Complex conjugate
representation of f
f
~f
53
~
G
SU2 su2
116
Heisenberg representation
R2 , H2; R
Ur
L2 R
233
L2 R
238, 242
r-mode Heisenberg
representation
R , H2r; R
L R
241
SU1; 1 su1; 1
L2 R
250
SU1; 1 su1; 1
L R
254
U
Wr
S
2
S
Sr
~Z
W
46, 98
2r
Sp2r; R
L R
259
Z2d
Cd
264
Z2d
Cd
264
H2; Zd
265
F2q
Cd
Cq
F2q
Cq
267
2r
F2r
q , Zd
Cd , Cq
2r
F2r
q , Zd
Cd , Cq
WZ;H
H2r; Zd
Cd
WF;H
H2r; Fq
SrX g
VrX
Sp2r; X
C ,C
X2r o Sp2r; X
Cd , Cq
WZ
Pre-discrete-Heisenberg
representation
Discrete Heisenberg
representation
WZ;H
~F
W
WF
WrX
~r
W
X
representation
Discrete Heisenberg
representation
Discrete Heisenberg
representation
representation
Metaplectic representation
267
268
268
270
Cq
270
dr
qr
278
282
xxvi
Symbols
Table 5 Vector
Symbol
Name (Denition)
jXiiA;B
Vector on HA HB
j; ji
j; i
j; j; j0 i
j : i
j : i
1

:
CONS of U G weight vector for SU2

Maximum weight vector
j; ji jj0 i
Coherent vector
Coherent vector
Coherent vector
41, 121
212, 225, 228
41
127
222, 227, 228
261
jni
jêl i
jêZ li,jêF li
Number vector
Dual computational base
Dual computational base
233
58
265, 268
Page
Table 6 Symbols for vector space

Symbol
Name (Denition)
Page
CG
Group algebra
Set of square integrable functions of G
59
61, 104, 107
L2 G
^
L2 G
61, 104, 107
L G; inv
Set of square integrable functions of conjugated class
63
L2 G; inv; R
^
L2 GE
Set of real-valued functions in L2 G; inv
63
U G; U
^
Irreducible representation space of 2 G
Complexication of V with complex inner product
Complexication of V with Hermitian inner product
V C

V C
64
41
52
52
Table 7 Symbols for matrices and operators

Symbol
Page
J
Jd
E0;1
K ;1
K;1
M
P
Q
F1x
54
54
114
114
114
114, 275
114, 275
275
114
114
F1y
F1z
x
F1
114
116
(continued)
Symbols
xxvii
Table 7 (continued)
Symbol
Page
y
F1
z
F1
E0;1
K ;1
K;1
W
Q : P
XZ
ZZ
XF s
ZF t
DFT
x
Fj;l
116
116
117
117
117
118
186
278
278
267
267
58
130
y
Fj;l
130
z
Fj;l
130
Ej
Kj;l;
Kj;l;
Fjz;u
130
130
130
131
a
Q
P
N
Uk;j
19, 186, 234

19, 232
19, 232
234
243
Table 8 Other symbols

Symbol
Name (Denition)
Complex conjugate matrix of X
XT
Xy
Transposed matrix of X
Transposed complex conjugate matrix of X
Order (number of elements) of G

Residue class with representative g
Index of H with respect to G
Conjugate class of a in G
23
25
25
26
Product set of sets H and K

Factor system
Dimension or formal dimension of U G
Multiplicity
28
31
41, 102
56
jGj
g
jG : Hj
aG
HK
fTk; `k; k 0 g
d
00
C;
0 G
Page
(continued)
xxviii
Symbols
Table 8 (continued)
Symbol
Name (Denition)
Page
suppf
Character of representation f
Irreducible character
Support of character f
Factor system
41
41
43
46
Element of HG : U1 corresponding to projective

representation f
Label of complex conjugate representation of f
46
Fourier transform
Inverse Fourier transform
63, 108
63, 108
Killing form
90
Casimir operator
Normalized invariant measure over group G
Normalized invariant measure over homogeneous space
Plancherel measure
Plancherel measure
94
102
103
67
110
feig;g gg;g0
Hf
^
2 G
F
F 1
X; Yg
Cf
G
n r
GE
^
FW
!Z
Z
!F
F
m1 ; . . .; mr1
m1 ; . . .; mr1
in
Dynkin index
Young index
53
118
255
264
264
267
267
130, 202
132
132
137, 212
Chapter 1
Mathematical Foundation for Quantum

System
Abstract This book introduces group representation theory in terms of quantum

theory. For this purpose, this chapter introduces basic concepts of quantum theory,
measurement, state, composite system, many-body system, and entanglement. It also
prepares mathematical notations for quantum systems. Although these notations are
specified to quantum systems, they are helpful for group representation. Hence, this
book consistently deals with representation theory based on these notations.
1.1 System, State, and Measurement

In the framework of quantum theory for a microscopic system, the object of interest
is called the Quantum system or the System, and is mathematically described as a
complex vector space H with a Hermitian inner product, which is finite-dimensional
or infinite-dimensional. Such a complex vector space H is called a Hilbert space
even though it is finite-dimensional.1 Since H is a vector space with a Hermitian
inner product, we can choose a completely orthonormal system (CONS) {ei }. Each
normalized base ei represents a state in the quantum system that is distinguished from
each other. An arbitrary state of the system is given as a normalized vector x H.
Once a CONS has been fixed as a standard
basis, the vector x describing an arbitrary
state is written as a linear combination i xi ei . In quantum theory, there are two
methods to describe the element x H. One is the description by a ket vector |x,
and the other is that by a bra vector x|. Although these descriptions are defined to
satisfy the linearity with respect to real coefficients, the multiplication of a complex
coefficient a C is defined as
|ax = a|x, ax| = ax|.
(1.1)
In particular, for a base ei of the standard basis, |ei and ei | are simplified to |i and
i|, respectively. On the other hand, we define the Hermitian inner product x|y for
x, y H to satisfy the condition ax|by = abx|y
for a, b C. Then, the inner

1 It is required to satisfy the completeness under the given inner product in the infinite-dimensional
case. see Sect. 1.6.

M. Hayashi, Group Representation for Quantum Theory,
DOI 10.1007/978-3-319-44906-7_1
1 Mathematical Foundation for Quantum System
product x|y between x and y can be regarded as the multiplication x| |y of the
bra vector and the ket vector.
In addition to a state, a physical quantity is also an important concept. When
the system obeys the classical mechanics, the value of the physical quantity is also
determined according to the state of the system, hence, the physical quantity is given
as a function whose input is the state of the system. However, the physical quantity
does not necessarily take a fixed value, and is defined as a Hermitian linear map on
the Hilbert space.2 In the following, for the simplicity, we employ the terminologies
in the finite-dimensional case.
The most important physical quantity is the Hamiltonian, which is written as H .
This is because it is the physical quantity describing the time evolution. Since H is
a Hermitian matrix, eit H is a unitary, where the real number t describes the time.
Then, the map |x eit H |x describes the evolution of the state of the system, i.e.,
the state change during time period t.
As mentioned above, a physical quantity A does not take a fixed value. What is its
exact meaning? To answer this question, we need to consider the measurement of the
physical quantity A. This is because the value of the physical quantity is determined
via the measurement. In quantum theory, any physical quantity A is written as a
Hermitian matrix or a self-adjoint operator on H, and the Hermitian matrix or the
self-adjoint operator is also written as A with this context. When a Hermitian matrix
A has no multiplicity for eigenvalues, it is called non-degenerate. Otherwise, it
is called degenerate. To consider the measurement, we focus on a non-degenerate
Hermitian matrix A. Then, the eigenvalue decomposition of A is given as
A=
ai |i i |,
(1.2)
where ai is an eigenvalue of A and i is an eigenvector corresponding to the eigenvalue ai .

Once the physical quantity A is measured, the possible outcomes are limited to
the eigenvalues ai of the Hermitian matrix A. The realized outcome is determined
according to the state of the system. The outcome is determined only probabilistically
except for several special cases. When the state is a normalized vector x H, the
probability to obtain the outcome ai is given to be
x|i i |x = |x|i |2 .
(1.3)
The
above values gives the probability distribution because the relation
set of the
2
|x|
|
=
1 can be shown by the fact that {i }i is a CONS. In particular, when
i
i
the state is known to be a base of the CONS {i }i , we can identify which base is the
state |x via the above measurement. In general, once a CONS {i }i is given, it is
2 It
is defined as a self-adjoint operator in the infinite-dimensional case. The precise discussion is

given in Sect. 1.6.
3
state
Fig. 1.1 Measurement
measurement
{ E i} i
outcome
ai
possible to realize the measurement whose outcome corresponding to the base i is

obtained with the probability (1.3) in the above way.
However, a Hermitian matrix might be a degenerate, i.e., might have multiplicity
for eigenvalues, in general. The above fact can be generalized to the degenerate case
by considering the spectral decomposition
A=
ai E i
(1.4)
instead of the eigenvalue decomposition (1.2). Then, the eigenvalues of A are {ai },
and the eigenspace corresponding to ai is the range of the projection E i . When the
physical quantity A is measured, the probability to obtain the outcome ai is given
to be x|E i |x = Tr E i |xx|. Hence, in general, given an orthogonal decomposition
i Hi of the Hilbert space H, a measurement can be defined by the set {E i }i of the
projections E i on the Hilbert space Hi . Such a measurement is called a projectionvalued measurement (PVM). As Fig. 1.1, the outcome is given as ai .
Here, we should remark that the above probability is determined by the onedimensional projection |xx| rather than the vector |x. Hence, even if the vector
|x is changed to |ei x by multiplying the phase factor ei , the corresponding onedimensional projection |xx| does not change.
In general, it is difficult to prepare the state of the system to be a specific state |x
exactly due to the difficulty of control of the quantum system. The mixture of two
states |x and |y with a ratio (1 p) : p is rather natural than the exact preparation
of the state |x. In this case, when the above measurement is applied, the probability
to obtain the outcome ai is given to be
(1 p) Tr(E i |xx|) + p Tr(E i |yy|) = Tr E i ((1 p)|xx| + p|yy|). (1.5)
How can we describe such a mixture of states? Since we cannot recognize the system
by means other than measurements in the quantum system, it is better to identify a
mixture of state with another mixture when they cannot be distinguished via any
measurement. Thus, it is reasonable to define a mixed state by using the right hand
side (RHS) of (1.5). That is, we describe the mixed state as a Hermitian matrix
(1 p)|xx| + p|yy|. In general, the state corresponding to the mixture of several
states |i with the probability pi is expressed as
:=
pi |i i |.
(1.6)
The above is a Hermitian matrix, in which, the trace is 1 and the eigenvalues
are not less than zero. Thanks to the above discussion, we can correctly describe the
probability to obtain each outcome of each physical quantity once we know the above
matrix . Conversely, when a Hermitian matrix satisfies the two conditions, (1) the
trace is 1 and (2) the eigenvalues are not less than zero, it can be described by the RHS
of (1.6) via the eigenvalue decomposition of . Here, the set of the states {|i } in the
RHS of (1.6) does not necessarily form a CONS. In the following, a Hermitian matrix
is called a density matrix when it satisfies the above two conditions. In contrast,
a state described by a normalized vector in H is called a wave function. When a
density matrix is written as a wave function, it is called a pure state. Otherwise, it
is called a mixed state.
Here, let us consider the measurement of the physical quantity A. Then, when the
density matrix of , the expectation of the outcome is given as
ai Tr E i = Tr A.
(1.7)
Hence, the density matrix can be regarded as a non-commutative extension of the

probability distribution. On the other hand, the physical quantity A can be regarded
as a non-commutative extension of a random variable. In probability theory, the
expectation is the sum of the product of the possible values of the random variable
and their probability. Since, in quantum theory, the expectation is the trace of the
multiplication of a density matrix and a physical quantity, the above correspondence
can be thought to be a natural non-commutative extension. Also, the variance is
given as
2 A :=
ai2 Tr |i i | (Tr A)2
= Tr A2 (Tr A)2 = Tr(A (Tr A)I )2 .
(1.8)
When two physical quantities A and B are commutative with each other, a simultaneous decomposition of A and B, i.e., the simultaneous measurement of A and
B, is given as follows. Let ai , Hi , and E i be an eigenvalue of A, the eigenspace
of A associated to ai , and the projection to Hi , respectively. Since B(A ai ) =
(A ai )B, any vector |x in Hi satisfies (A ai )B|x = 0. So, B|x belongs to
Hi , i.e., E i B|x = B|x.
For any element |y, we have E i B E i |y = B E i |y, i.e.,
E i B E i = B E i . Since i E i = I , we have
B=
Ei B Ei .
(1.9)
Thus, we can make

the spectral decomposition {E j,i } j of E i B E i on Hi as E i B E i =

b
E
.
Since
j j,i j,i
j,i E j,i = I , the decomposition {E j,i } j,i forms a PVM. As (1.9)
guarantees that j,i b j,i E j,i = B, the decomposition {E j,i } j,i gives the simultaneous
measurement of A and B.
In this case, the covariance of two outcomes is given as

A B
1
:= Tr[(B (Tr B))(A (Tr A)) + (A (Tr A))(B (Tr B))]. (1.10)
2
Now, we consider the case when A and B are non-commutative. Although A B
can be defined and can be regarded as a kind of correlation between the two physical
quantities A and B, it cannot be thought to be a covariance because the simultaneous
measurement of A and B cannot be defined.
In the above formulation, a measurement is given from a CONS. This formulation
can be generalized as follows. Given a set {|i }i of non-normalized vectors in H,
we assume that

|i i | = I,
(1.11)
i
where I is the unit matrix. Then, the set {|i i |}i gives a decomposition of the
unit matrix I , which gives a measurement as well [58]. Such a decomposition is
called a positive operator-valued measure (POVM). When the above measurement is
applied to the system whose state is given as the density matrix , the measurement
outcome i is obtained with the probability i ||i . Such a measurement is possible
within the framework of PVMs by extending the system [46, 54, 58]. Such a type of
measurement can be generalized to the continuous case. That is, when normalized
vectors {| } are parameterized by the set and the measure on the parameter
space satisfies the condition

| |(d ) = I,
(1.12)
the decomposition of the unit matrix is called a POVM and gives a measurement in
the above-mentioned sense.
In summary, the key points of quantum theory are summarized as the following
basic concepts, a physical quantity (especially the Hamiltonian), its spectral decomposition, a wave function as an element of H, and a density matrix.
3
Exercise1.1 Define
M= {M

i }i=1 and thedensity
matrix as M1 =

1 thei POVM
1 i
0
0
1
1
1
2 4 , M =
2
4 , M =
, and = 2
. Calculate the distri2
3
i 1
11
0 43
4i 18
4 8
bution PM .

0 i
01
Exercise 1.2 Give the spectral decomposition of A =
and B =
.
i 0
10
Exercise 1.3 Calculate the expectation when the measurement

isthe spectral decom10
position of A given in Exercise 1.2 and the state is = 21
01
Exercise 1.4 Calculate the variance A in Exercise 1.3.

Exercise 1.5 Calculate the value A B when A, B and are given in Exercises
1.2 and 1.3, respectively.
1.2 Composite System

1.2.1 Tensor Product System
When we have two quantum systems H A and H B and we treat them as one quantum
system, we need a description for the whole system as Fig. 1.2. For example, when
the system H A describes the internal freedom of a particle and the system H B does
its position, we need the quantum system that describing the whole freedom of the
particle, whose typical example is given in Sect. 5.4. Such a quantum system is
called the composite system of H A and H B . The composite system is given by the
k
tensor product space H A H B . When H A and H B have their CONSs {|vi }i=1
l
and {|u j } j=1 , respectively, the tensor product space H A H B is given as the linear
space whose CONS is {|vi , u j }1ik,1 jl . One might consider that the composite
system is given as the product space H A H B . However, the composite system is
as the tensor product space H A H B . To understand the reason, we consider the
typical classical case. Remember that the product space H A H B has the CONS
k
{|u j }lj=1 .
{|vi }i=1
In the classical case, the state is given as a probability distribution over the set of
events. That is, to identify the system, we need to identify the set of events. The basis
of the quantum system corresponds to the set of events. In the classical case, an event
of the composite system is given as a pair of events of the respective systems. That
is, the set of events of the composite system is the product set of the sets of events of
the respective systems. Considering the relation between the basis of the quantum
system and the events of the classical system, we find that the basis of the composite
k
{|u j }lj=1 . Thus, the composite
system is {|vi , u j }1ik,1 jl and is not {|vi }i=1
system is the tensor product system H A H B . When we need to identify the basis of
the composite system with a single number, we number it as |ei+k( j1) = |vi , u j .
Next, let us consider the case when the respective states of the systems H A and
H
B
l to be the density matrices and . When =
k are independently prepared
i,i =1 ai,i |vi vi | and =
j, j =1 b j, j |u j u j |, the state of the composite sysk
l
tem is given as the tensor product state := i,i
=1
j, j =1 ai,i b j, j |vi , u j
Fig. 1.2 Composite system

of two quantum systems
vi , u j |. In the following, for simplicity, we consider the case when these density
k

matrices are diagonal, i.e., = i=1
ai |vi vi | and = lj=1 b j |u j u j |. Since
k l
= i=1 j=1 ai b j |vi , u j vi , u j |, the event (vi , u j ) occurs with the probability ai b j . Hence, the above independent situation does not contradict the independence in the classical case. Therefore, when the respective states are independently
prepared in the quantum system, the state of the composite system is given as the
tensor product state.
k l
Here, we should remark that the tensor product |v |u = i=1
j=1 ci f j
k
l
|vi , u j of two wave functions |v = i=1 ci |vi and |u = j=1 f j |u j is independent of the choice of the bases of the respective systems. To check this fact,
k
and
it is sufficient to show that the tensor product based on other CONSs {|vi }i=1
l
{|u j } j=1 of the respective systems H A and H B equals the tensor product |v |u =
k l
i=1
j=1 ci f j |vi , u j based on the original bases. When the wave functions |v and

|u are given as |v = ks=1 cs |vs and |u = lt=1 ft |u t , the tensor product based

we employ the unitary
on the basis {|vs }ks=1 and {|u t }lt=1 is s,t cs ft |vs , u t . Now,
matrices U and V defined as |vs = i Vs,i |vi and |u t = t Ut, j |u j . Then, we

have cs = s cs Vs,i and f t = t ft Ut, j , which implies that
cs ft |vs , u t =
s,t

s,t,i, j
cs ft Vs,i Ut, j |vi , u j =
ci f j |vi , u j .
i, j
Thus, given a matrix X on H A and a matrix Y on H B , their tensor product is defined

as a linear map on the composite system H A H B as follows.
X Y (|v |u) := (X |v) (Y |u).
(1.13)
Since the tensor product of |v and |u is defined independently of the choice of the
bases, the definition of the linear map X Y does not
depend on the choice of the
k
bases. Hence, the tensor product state of = i,i
=1 ai,i |vi vi | and =
l
l
k

b
|u
u
|
is
defined
as
a
b
j
j
j, j =1 j, j
i,i =1
j, j =1 i,i j, j |vi , u j vi , u j |, which
does not depend on the choice of the bases. States of the composite system H A H B
cannot be restricted to tensor product states or their convex combinations

j p j j j , which are called separable states, where p j is a probability and j
and j are density matrices on H A and H B , respectively.
1.2.2 Entangled State

We need to notice that all states on the composite
system are not necessarily separable, i.e., do not necessarily havethe form j p j j j . For example, the pure
state corresponding to the vector j p j |v j , u j cannot be described as a convex

combination of tensor product states unless p j is a deterministic distribution. For
Fig. 1.3 Entangled state

between two quantum
systems
1 ( u ,v + u ,v ) x y
1 1
2 2
2
example, the state 12 (|u 1 , v1 + |u 2 , v2 ) cannot be written with any convex combination (Fig. 1.3). It is not a separable state. Such a state is called an entangled
state, which has been studied as a resource of quantum information processing. In
particular, such a property is called entanglement.
When the state of the composite system H A H B is given as the density matrix
=
k
l

ci, j,i , j |vi , u j vi , u j |,
i,i =1 j, j =1
there exists a density matrix Tr H B such that

Tr(Tr H B )X = Tr (X IH B )
for any matrix X on the system H A . Thus, it is natural to consider that the state of
the smaller system H A is given as the density matrix Tr H B . The state Tr H B is
called the partial trace of , and is calculated as
Tr H B =
k
l

i,i =1
ci, j,i , j |vi vi |.
j=1
When the density matrix is diagonal under the basis {|vi , u j }i, j , the partial trace
coincides with the marginal distribution of the distribution composed of the diagonal
elements. The partial trace can be defined when is not necessarily a density matrix.
When it is a density matrix, the partial trace is also called the reduced density matrix
of . In the following, when there is no possibility for confusion, Tr H B is simplified
to Tr B .
Now, we consider the linear space of Hermitian matrices on H A as the normed
linear space B(H A ) with respect to the matrix norm. The partial trace Tr H B can be
regarded as a linear map from B(H A H B ) to B(H A ). The norm of the linear map
is min{dim H B , dim H A } because the relation
Tr H B X
min{dim H B , dim H A }
X
holds for a Hermitian matrix X on H A H B .
(1.14)
In this book, if there is no possibility for confusion, for a matrix X on H A and

a matrix Y on H B , the matrices X IH B and IH A Y on H A H B are simplified
to X and Y . Hence, X Y is written as X Y . Once standard bases of the systems
H A and H B are fixed, we use the following simplified notations. Since two quantum
systems are addressed, the ket vectors of their standard bases are distinguished as
| j A , | j B . Hence, the basis of the composite system H A H B is |k A | j B , which
is simplified to |k, j A,B .
In the following, given a matrix X = (xk, j ), we denote the complex conjugate
matrix (The entries are given as the complex conjugate of the original entries.)
under the standard basis by X , and the transposed matrix under the standard basis
by X T . Then, we denote the complex conjugate transposed matrix by X . Although
the matrices X and X T depend on the choice of the standard basis, the matrix X
depends only on the Hermitian matrix and does not depend on
the choice of the
standard basis. Using the matrix X = (xk, j ), we denote the vector k, j xk, j |k, j A,B
on the composite system H A H B by |X A,B . Then, we have
Y Z |X A,B = |Y X Z T A,B .
(1.15)
The inner product of two vectors |X A,B and |Y A,B is given as

A, BX |Y A,B = Tr X Y.
(1.16)
Hence, the vector |X A,B is a wave function, i.e., its norm is 1 if and only if
Tr X X = 1.
(1.17)
We also have the following formula for the partial trace;

Tr B |X A,B
A,B X |
= X X , Tr A |X A,B
A,B X |
= XT X.
(1.18)

1 1 0
1 1 0
1 0 1
Further, the states 2
, 2
, 2
, and
01
0 1
10
A,B
A,B
A,B

0 1
1
are called Bell states. Since they form a CONS on C2 C2
2
1 0
A,B
(see Exercise 1.7), they give a PVM, i.e., a measurement on the composite system
C2 C2 . The measurement is called Bell measurement. To realize this measurement, the measurement device needs to handle the quantum correlation between two
systems (Fig. 1.4).
Fig. 1.4 Bell measurement
10
Exercise 1.6 Assume that X is a 2 2 matrix. Show that the state |X A,B A,B X |
is entangled if and only if det X = 0.

1 1 0
1 0 1
1 1 0
, 2
, 2
, and
Exercise 1.7 Show that 2
01
0 1
10
A,B
A,B
A,B

0 1
1
forms a CONS on C2 C2 .
2
1 0
A,B
1.3 Many-Body System

Consider the case when there are n particles and their quantum systems are given as
Hi (i = 1, . . . , n) (Fig. 1.5). When they can be distinguished from each other, the
composite system of these n particles is given as (((H1 H2 ) ) Hn ). Since the
tensor product space does not depend on the order of the tensor products, the above
tensor product space is the same as (H1 ( (Hn1 Hn ))). Hence, we denote it
by H1 H2 Hn . In particular, when each system Hi is isometric to H, the
tensor product space is simplified to Hn .
However, when these particles are the same particles and cannot be distinguished
from each other, we cannot use the above notation. Such a difficult case will be
addressed as bosonic system or fermionic system in the latter chapter (Sects. 4.4.2
and 7.4). Even if these n particles are the same particles, they might be distinguished
from each other. For example, when the system Hi describes the internal freedom of
the particle and the position of each particle can be distinguished with probability 1,
the composite system is given as H1 H2 Hn .
When the state of each system Hi is independently prepared to be the density
matrix i , the state of the composite system is given as (((1 2 ) ) n ). Since
the tensor product of the matrices does not depend on the order of the tensor product,
the density matrix given by the tensor product is written as 1 2 n . In
particular, when i = , i.e., the state is independently prepared in n quantum
systems, the density matrix of the composite system is simplified to n , and is
called an n-fold tensor product state of . This notation can be applied to the case
when and i are not restricted to density matrices.
Given a matrix A on the tensor product system H1 H2 Hn , we denote
the partial trace of A with respect to the system Hi by Tr Hi A. Conversely, we denote
the partial trace with respect to the all systems except for the system Hi by Tr H i A.
In particular, when A is a density matrix , Tr H i is simplified to Hi or i . On the
other hand, given a matrix A on the system Hi , the matrix I i1 A I ni on
Fig. 1.5 Composite system

of n quantum systems
1.3 Many-Body System
11
the composite
n system H1(n) H2 Hn is simplified to Ai . We also abbreviate
Ai to A .
the matrix i=1
As mentioned above, to consider the quantum system, we need to discuss a physical quantity (especially Hamiltonian), its spectral decomposition, a wave function, and a density matrix. However, even when the respective systems Hi are
2-dimensional, the composite system has the dimension 2n and requires very complicated treatment. So, we have serious troubles to handle physical quantities and
density matrices. However, when they have invariance with respect to physical transformation, the number of parameters to describe them can be reduced so that their
mathematical treatment is simplified. The purpose of this book is to provide the
systematic method to treat physical quantities and density matrices by reducing the
freedom under the suitable physical invariance.
Exercise 1.8 Give the spectral decomposition of n when = 13 E 0 + 23 E 1 on C2 .
1.4 Hamiltonian
1.4.1 Dynamics and Hamiltonian
In the isolated quantum system, time evolution of the state of the system is given by
the time span t and the Hamiltonian H as
UU ,
(1.19)
where the unitary matrix U is defined as U := eit H (Fig. 1.6). However, when we
cannot estimate the time t, we need to take the average with respect to T as
1
T
eit H eit H dt.
(1.20)

In particular, taking the limit T , we have
the state j E j E j when the Hamiltonian H has the spectral decomposition j h j E j (see Exercise 1.9). Then, only
eigenvectors are stable. Hence, it is an important topic to calculate eigenvectors of
the Hamiltonian H . That is, we need to the following equation;
H |x = E|x
with eigenvalue E, which is called Schrdinger equation.
Fig. 1.6 Time evaluation
(1.21)
12
Hamiltonian has another role in quantum theory. When the system is correlated to
a huge system called an environment system or a heat bath during a long time, the state
of the system approaches to the state e H / Tr e H called the thermal equilibrium
state or thermal state, where the parameter is called the inverse temperature. On
the other hand, under the limit that the inverse temperature goes to infinity, the
state converges to an eigenvector of H associated with the minimum eigenvalue.
Especially, when the eigenspace is the one-dimensional, the state is called the ground
state. Such a state can be realized by keeping the system at a lower temperature.
Therefore, it is a central issue to derive the ground state for a given Hamiltonian. In
general, the number of particles in the given system is not necessarily one. When
the particle is the type of fermion, more than one particles cannot share the same
state. That is, when one particle takes the ground state, another particle takes a state
orthogonal to the ground state. Hence, to clarify all of states of these particles, we
need to calculate several eigenvectors of the Hamiltonian H associated to lower
eigenvalues.
However, in general, it is often difficult to describe the exact form of the Hamiltonian. In some cases, the main term H0 of the Hamiltonian is given and the fluctuation H is given as a small constant times of another Hermitian matrix H1 .
In particular, when H0 and H1 are commutative with each other, we have common
eigenvectors as a basis. Then, we can take a eigenvector system of H independently
of the coefficient .
T

Exercise 1.9 Show that lim T T1 0 eit H eit H dt = j E j E j when H = j
h j E j.
1.4.2 Simultaneous Diagonalization

Now, we consider the diagonalization of a given Hamiltonian H on a given composite system H1 H2 . Usually, this task needs our complicated operation for the
composite system H1 H2 . However, when a non-degenerate Hermitian matrix A
on H2 is commutative with H as a Hermitian matrix on H1 H2 , this task can be
decomposed to several
operations on the subsystems H1 and H2 as follows. First, we
diagonalize A as i ai |xi xi |. When |xi xi | is identified with |xi xi | IH2 , this
diagonalization can be regarded as the diagonalization of a matrix on the composite
system H1 H2 . Then, we have a Hermitian
matrix Hi := Tr H2 |xi xi |H |xi xi |
on H1 . Now, we diagonalize Hi as Hi = j h j,i E j,i . Since
|xi xi |H |xi xi | =
h j,i (E j,i |xi xi |),
(1.22)

we obtain the spectral decomposition of H as H = j,i h j,i E j,i |xi xi |. from the
discussion before (1.10). This method is employed in Sect. 5.1.
1.4 Hamiltonian
13
Notice that this method does not work when the Hermitian matrix A is degenerate
because (1.22) does not hold in general. To resolve this problem, we consider a
set of Hermitian matrices {Ak }k on H2 such that the simultaneous measurement
of the Hermitian matrices {Ak }k is composed of rank-one projections, i.e., has the
form {|xi xi |}i . In this case, we can show that |xi xi |H |xi xi | = H |xi xi |, which
implies that
H=
|xi xi |H |xi xi |.
(1.23)
Then, we can define the Hermitian matrix Hi on H1 in the same way. Using the
spectral decomposition of Hi , we obtain that of H in the above way.
Exercise 1.10 Give the simultaneous diagonalization of A I and I B, where
A and B are given in Exercise 1.2.
1.4.3 Relation to Representation

Due to the above discussion, it is important to derive the spectral decompositions of
the Hamiltonian H and the related Hermitian matrices. When the system H is too
large and/or is too complicated, it is very difficult to find their spectral decompositions. That is, it is almost impossible to derive them without reducing the problem
to an easier problem.
When a set of Hermitian matrices {A j } j satisfies the following condition, the
condition almost determines the structure of these Hermitian matrices. There exists
a set of real numbers ck, j,l such that
i[A j , Ak ] =
ck, j,l Al , x
(1.24)
where [X, Y ] := X Y Y X . Indeed, the condition vanishes a larger part of the freedom of these Hermitian matrices. Then, the difficulty of the spectral decomposition is
transformed to an algebraic problem. When we use terminologies roughly, this commutation condition (1.24) is called a representation of Lie algebra, and brings very
powerful methods. This method is related to the fact that the set of skew-Hermitian
matrices is closed with respect to the commutation relation.
Next, we focus on the set of unitary matrices. The set of unitary matrices is closed
with respect to the matrix product. Now, we assume that a subset S of the set of unitary
matrices are closed with respect to the matrix product. When the Hamiltonian H is
invariant for S, i.e., U HU = H for any element U S, the algebraic structure of
S for the matrix product determines a larger part of the structure of H . At least,
the structure vanishes a larger part of freedom of the Hilbert space H. Hence, the
problem of the analysis of the Hamiltonian can be reduced to much more simple
14
problem. When we use terminologies roughly, the condition for the subset S is called
a representation of a group. A representation of a group is a powerful tool as well as
a representation of Lie algebra. To discuss a representation of a group, we need more
rigorous formulation. So, in the next chapter, we start our study of a representation
of a group with the formal definition of a group. Since a representation of Lie algebra
requires much more preparation, we can start its discussion in the middle in Chap. 3.
1.5 Relation to Symmetry

As above-mentioned, symmetry plays an important role in the derivation of
Schrdinger equation. In fact, the three dimensional space has natural symmetry
with respect to the three-dimensional rotation and displacement. The symmetry of
three-dimensional rotation is crucial to discuss quantum system. These symmetries
are related to the freedom of the physical space and appear even in classical mechanics.
However, quantum mechanics has a different type of symmetry as Fig. 1.7. Since
quantum mechanical object is composed of microscopic particles, we cannot distinguish them when they are the same kind of particles, like, electrons. This problem
appears when we have more than one particles while this problem does not occur
for the single particle case. Indeed, in quantum theory, it is the most important thing
to identify the Hilbert space H to describe the quantum system because we cannot
discuss anything without identifying the Hilbert space H. Hence, we need a proper
Hilbert space H to describe the possible physical freedom reflecting the indistinguishability. Since the indistinguishability is related to a special kind of symmetry,
this description is closely related to the symmetry, which never appear in the classical mechanics. The symmetry is rooted in group representation theory and requires
many preparations. Indeed, at least, we have two kinds of indistinguishable particles,
boson and fermion. We need different type of symmetry dependently of the type of
indistinguishable particles. Hence, we discuss both types of indistinguishable particles in Sect. 4.4.2 after many discussions for group representation theory. Then,
we apply this description of the Hilbert space H to several important real physical
systems in Chaps. 5 and 7.
Fig. 1.7 Classical and

quantum symmetries
classical
symmetry
quantum
symmetry
rotation symmetry
rotation symmetry
translation symmetry
translation symmetry
internal (spin) symmetry
permutation symmetry
(due to
indistinguishability)
1.5 Relation to Symmetry
15
As another relation between quantum system and symmetry, we have discrete

symmetry of available unitaries. When we artificially control the quantum system,
we need to consider the set of possible unitaries, which usually forms a finite set
because it is quite hard to prepare the devices corresponding to infinitely many
unitaries. In this case, it is important to discuss what subset of unitaries generates
the given subset of unitaries. To discuss this issue, we need algebraic structure of
the set based on the matrix product. In this context, representation theory plays a
crucial role for the design of quantum operation. To protect quantum state from the
noises, we often the quantum noise based on this algebraic structure. Hence, this
algebraic structure brings us an essential utility for quantum technologies. Since this
kind of symmetry has analogy with boson, we discuss it in Chap. 8 after Chap. 7 that
discusses the detail of boson.
1.6 Remark for Unbounded Case*

Finally, we remark several mathematical issues for precise description. If a reader is
not interested in such mathematical issues, he/she can skip this part. When the system
H is infinite-dimensional, we need more careful discussions. Firstly, the system H
needs to satisfy the completeness under the given inner product, i.e., the system
H needs to contain the limit of Cauchy sequences in H. When it does not satisfy
the condition, we need to attach the completion for a given vector space H with
Hermitian inner product, which is explained below. First, consider the set of Cauchy
Second, we introduce the equivalence of the set as follows.
sequences {xn } in H.
Given two Cauchy sequences {xn } and {yn }, we define equivalence {xn } {yn } when
the sequence {xn yn } goes to zero. Finally, we take the quotient space composed
of sequences with respect to this equivalence. So, the resultant vector space satisfies
the completeness.
Even though the system H satisfies the completeness, we need to care about the
x
for an operator H on H. When the operator
operator norm H := supxH H
x
norm H takes finite value, the operator is called bounded. Otherwise, it is called
unbounded and we need another mathematical delicate problem. In the infinitedimensional case, there is a possibility that an operator H on H is unbounded. In such
a case, the operator H is called unbounded and requires a more delicate treatment.
When we omit this careful treatment, the discussion contains several inconsistencies.3
When the operator norm of an operator H is infinite, we need to restrict the domain
of H to a subspace D(H )
of H so that H x < for x D(H ). Then, we define
the graph norm x H := x2 + H x2 . When the domain D(H ) is closed in the
sense of the graph norm, H is called a closed operator. Now, we define the adjoint
operator H of H on the domain
D(H ) := {x H|y H such that y|z = x|H z for z D(H )}
3 Even
if a reader ignores such inconsistencies, so serious problem will not be caused when the
Hilbert space L 2 (Rd ) is discussed. So, it is possible to skip this section.
16
as
H x|z = x|H z
(1.25)
for z D(H ). When D(H ) D(H ) and H x = H x for x D(H ), H is called

symmetric. When D(H ) = D(H ) and H x = H x for x D(H ), H is called selfadjoint.
In general, it is quite difficult to directly define a self-adjoint operator H on the
proper domain D(H ) because the domain cannot be described explicitly in many
cases. Hence, we often define a self-adjoint operator indirectly as follows. Firstly,
we usually define an operator H on a dense subspace D( H ) of H so that H is
symmetric. Then, we take the completion of the subspace D( H ) in the sense of the
graph norm of H . The completion (the resultant Hilbert space of the completion)
is different from H because we consider the graph norm of H not the norm of the
original Hilbert space H. Then, we denote the completion by D(H ) and the operator
H on D(H ) by H . Since the operator H is automatically a closed operator, it is
called the closed extension of H . When H is symmetric, the closed extension H
is also symmetric. However, in general, the closed extension H is not necessarily
self-adjoint even though H is symmetric. Indeed, when an operator H has a closed
extension H that is self-adjoint, it is called essentially self-adjoint.
For a self-adjoint operator H , we have the following lemma.
Lemma 1.1 For a self-adjoint operator H , there exists a PVM E on R such that

y|H x =
sy|E(ds)x.
(1.26)
Conversely, we have the following lemma.

Lemma 1.2 Given a PVM E taking the outcome in R, we define the operator X E
as

X E x :=
s E(ds)x
(1.27)
R
on

D(X E ) := x H s 2 x|E(ds)x < .

(1.28)
Then, X E is a self-adjoint operator.

Due to this lemma, for any self-adjoint operator X and any real-valued function
f , we can define the self-adjoint operator f (X ) as R f (s)E X (ds), where E X is the
PVM defined for the self-adjoint operator X in Lemma 1.1.
17
Proof Assume that, for an element x H, there exists an element y H such that
x|X E z = y|z
(1.29)

for any element z D(X E ). Hence, it is sufficient to show that R s 2 x|E(ds)x <
. Firstly, we define the bounded subset S R := {s R||s| < R}. Then, we have

y|z = x|X E z = lim x
R

s E(ds)z = lim
R
SR

s E(ds)x z
(1.30)
SR

for any element z D(X E ). Since D(X E ) is a dense subset of H, SR s E(ds)x
2

converges to y as R . Thus, R s x|E(ds)x = SR s E(ds)x| SR s E(ds)x <
, which is the desired statement.

When a Hamiltonian is given as a self-adjoint
operator H , the unitary matrix U
describing the time evolution is given as U := R eits E(ds).
Next, we consider the case when the Hilbert space H is given as the set of square
summable sequences with an index n = 0, 1, 2, . . . , . An infinite-dimensional
matrix {h i, j }i, j0 is called a band-diagonal matrix with width l when h i, j = 0 for
|i j| > l.
Lemma 1.3 For two band-diagonal matrices {h i, j }i, j0 and {gi, j }i, j0 with width
l, the sum {h i, j + gi, j }i, j0 is a band-diagonal matrix with width l, and the product

{ f i, j }i, j0 is a band-diagonal matrix with width 2l, where f i, j := i+l
j =il h i, j g j , j .
The above lemma shows that the set of band-diagonal matrices are closed with
respect to the sum and the matrix multiplication. A band-diagonal matrix {h i, j }i, j0
is called Hermitian when h i, j = h j,i .
Lemma 1.4 ([68, Theorem 2.3]). For a band-diagonal matrix {h i, j }i, j0 , we define
an operator H on
l

2

<
D(H ) := {ak }
H
h
a

nn+k
n+k
k=0
n=0
(1.31)
k=l
as
(H a)n :=
l
h nn+k an+k
(1.32)
k=l
for a D(H ). When {h i, j }i, j0 is Hermitian, the operator H is self-adjoint.

This lemma shows that any band-diagonal matrix can be regarded as a self-adjoint
operator.
Proof Since a D(H ) satisfies H a = H a, it is sufficient to show that for an element x H \ D(H ), no element y H satisfies that
18
x|H z = y|z
(1.33)
for z D(H ). When y Hsatisfies the above condition,

with z =
the conditions
l
2
implies
that
y
=
h
x
.
Since
|y
|
<
,
x
D(H ).
{m,n }
m
n=0
k=l m,m+k m+k
n=0 n
This, we obtain the desired statement.

Lemma 1.5 Let {h i, j }i, j0 be a Hermitian band-diagonal matrix. When a dense
subspace D( H ) of H contains the sequence {m,n }
n=0 for any non-negative integer
m and is included in D(H ), the operator H defined by the Hermitian band-diagonal

matrix {h i, j }i, j0 on D( H ) is symmetric. Also, the closed extension of H is H .
Proof The first part is trivial because it is included in D(H ). Since the domain
of the closed extension of H is included in D(H ), it is enough to show that any
element of D(H ) is contained in the domain of the closed extension of H . Any finite
in the domain of the closed extension of H . When a
sequence {an }m
n=0 is contained
l
2
sequence {an }n=0 satisfies n=0 | k=l h n,n+k an+k |2 < and
n=0 |an | < ,
the sequence {an }n=0 belongs to the domain of the closed extension of H , which
implies the desired statement.

We often consider the case when the Hilbert space H is given as L 2 (R), which
is defined as follows. Firstly, we consider the set S (R) of rapidly decreasing functions as

d

S (R) := C (R) sup |x f (x)|2 < , , Z+ ,
dx
xR
(1.34)
where C (R) is the set of smooth functions. Then,

the space L 2 (R) as
we define
2
2
2
the completion of the L norm defined as := R |(x)| d x. The multiplication
operator (x) x(x) and the differential operator (x) ddx (x) are defined in
the following way. So, we can define the symmetric operators Q and P on S (R) by
(Q)(x)
:= x(x), (P)(x)
:= i
d
(x).
dx
The set S (R) contains the n-th Hermitian function n (x) :=
(1.35)
2
x
1
d n x 2
,
e 2 dxn e
n!2n
1
+ i P)
which is identified with |n. We define the operators a and a as a := 2 (Q
i P),
and a := 12 (Q
respectively. Since a|n
= n|n 1 and a |n =
n + 1|n + 1, the operators a and a have band-diagonal forms. So, the opera-
tors Q and P have Hermitian band-diagonal forms. Due to Lemmas 1.4 and 1.5,
the closed extensions Q and P of Q and P are self-adjoint. Now, we consider a
two-variable ordered polynomial f . That is, when the degree is 2, f (a, b) is written
P)
is defined on
as linear sum of a, b, a 2 , ab, ba, and b2 . Then, an operator f ( Q,
S (R) as a linear sum of multiplications of P and Q, and it has a band-diagonal

form. In this case, the ordered polynomial f equals the opposite ordered polynomial
19
of f if and only if f (a, b) is written as linear sum of a, b, a 2 , ab + ba, and b2 . When

P)
is symmetric and has a
this condition is additionally assumed, the operator f ( Q,
Hermitian band-diagonal form. Due to Lemma 1.5, the closed extension f (Q, P) of
P)
is self-adjoint. So, the operators a and a are defined as the closed operators
f ( Q,
in this way.
Further, we can define the set S (Rd ) of d-dimensional space rapidly decreasing functions in the same way. Taking the completion, we can define the ddimensional L 2 space L 2 (Rd ). Then, we can define self-adjoint operators Q j , P j
and f (Q1 , . . . , Qd , P1 , . . . , Pd ) in the above way when f is 2d-variable ordered
polynomial and it equals the opposite ordered polynomial of f .
Now, we consider
the case with the Coulomb potential. Firstly, we define the
self-adjoint operator 3j=1 Q2j on a dense subspace of L 2 (R3 ). Then, according
tothe method explained after Lemma 1.2, we define the self-adjoint operator
1
( 3j=1 Q2j ) 2 , which is called the Coulomb potential. This operator is also constructed by a closed extension of an operator defined in the domain S (R3 ).
Lemma 1.6 Let f be a smooth function densely defined on Rd . Then, we can define
the self-adjoint operator f (Q1 , . . . , Qd ) according to the method explained after
d ) on S (Rd ). Then, the closed
1, . . . , Q
Lemma 1.2. We also define the operator f (Q
extension of f (Q1 , . . . , Qd ) is f (Q1 , . . . , Qd ).

Proof Since the domain of f (Q1 , . . . , Qd ) includes S (Rd ), the domain of f (Q1 ,
1, . . . , Q
d ). So, it is
. . . , Qd ) includes the domain of the closed extension of f (Q
sufficient to show that any element g of the domain of f (Q1 , . . . , Qd ) is expressed
as the limit of sequence in S (Rd ) in the sense of the graph norm.
Given R > 0, we choose a function g R S (Rd ) such that g R (x) = 1 for | f (x)| <
R and g R (x) = 0 for | f (x)| > 2R. We also choose a sequence of functions gn
S (Rd ) such that g gn 0 as n . Now, we choose an integer n R such that
g R (g gn R ) < R12 .
Then, we have gm gn m S (Rd ). Now, we show that gm gn m converges to g in the
sense of graph norm as follows. We have
gm gn m g gm gn m gm g + gm g g 0.
(1.36)
and
f (Q1 , . . . , Qd )(gm gn m g)
f (Q1 , . . . , Qd )(gm gn m gm g) + f (Q1 , . . . , Qd )(gm g g)

2m(gm gn m gm g) +
|g(s)|2 ds
2m
1
+
m2

|s|>m
|s|>n m
|g(s)|2 ds 0.
(1.37)
20
Although the self-adjoint operator f (Q1 , . . . , Qd , P1 , . . . , Pd ) is defined only on

a dense subspace of the Hilbert space H, to avoid the complexity of description,
we simply say that the self-adjoint operator f (Q1 , . . . , Qd , P1 , . . . , Pd ) is defined
on the Hilbert space H in this book when the self-adjoint operator is defined in the
above way.
Exercise 1.11 Show that the functions x f (x) and
also belong to S (Rd ).
d
(x)
dx
for a function f in S (Rd )
Exercise 1.12 Consider two functions f and g in S (Rd ). Show that the product
f (x)g(x) also belongs to S (Rd ).
Chapter 2
Group Representation Theory
Abstract This chapter discusses representations for group in a general framework

including projective representations, which are important in quantum theory. Since a
projective representation is closely related to extension of group, this chapter focuses
on this relation. To discuss representation theory including projective representations,
we need to handle the factor system, i.e., the set of phase factors, which requires complicated notations. The discussions with projective representations are complicated
and are often omitted in introductory book for mathematics. However, they are more
essential for quantum theory. So, this chapter handles projective representations by
using factor systems and we keep this style during the whole book. In the latter part
of this chapter, we proceed to the details of representations of finite groups so that
it introduces the Fourier transform for finite groups, which connects analysis and
algebra. As a typical example, we analyze representations of a permutation group by
using Young diagrams.
2.1 Group and Homogeneous Space

2.1.1 Group
It is natural that physical rule does not depend on the coordinate or the origin. The
change of the origin is generated by the replacement, and that of the coordinate is by
the rotation or the reflection. Hence, in physics, we often consider various transformations, e.g., rotation, reflection, and replacement. Important physical quantities and
density matrices often satisfy the invariance or other proper properties with respect
to these transformations. To describe such useful properties, we need to employ the
mathematical theory for such a transformation.
In general, when a transformation g1 and another transformation g2 are given, we
can apply the transformation g2 after the transformation g1 , which can be regarded as
another transformation. That is, there exists a transformation combining the transformation g1 and the transformation g2 . Further, when the transformation g1 is invertible,
we have the inverse transformation of g1 . Hence, we impose the following mathematical properties to the set G of invertible transformations. For any two elements
DOI 10.1007/978-3-319-44906-7_2
21
22
2 Group Representation Theory
Fig. 2.1 Product
g2
g1
g1 g2
Fig. 2.2 Associative law
g1
g2
g1 g2
g3
g2 g3
g1 g2 g3 = ( g1 g2 ) g3 = g1 ( g2 g3 )
Fig. 2.3 Identity element 1
g
eg = g
e
Fig. 2.4 Identity element 2
e
g
ge = g
g1 , g2 G, we have the product g1 g2 G, as Fig. 2.1, that satisfies the following

conditions.
Associative law: Any three elements g1 , g2 , g3 G satisfy the condition (g1 g2 )g3
= g1 (g2 g3 ) as Fig. 2.2.
Identity element: There uniquely exists an element e of G satisfying the following
condition. Any element g G satisfies eg = ge = g as Figs. 2.3 and 2.4. (The
element e is called the identity element, and corresponds to the situation that we
apply no transformation.)
Inverse element: For any element g G, there uniquely exists an element g 1
G such that g 1 g = gg 1 = e as Fig. 2.5. (Such an element is called the inverse
element of g.)
For example, we can easily check that the set of rotations and the set of replacements satisfy these conditions. In mathematics, when the set G satisfies the above
conditions, it is called a Group, which has been studied very extensively. In the
following, to address a group precisely, we prepare several mathematical concepts
required in the latter discussions. When a subset H of G forms a group with respect

Fig. 2.5 Inverse element
23
g
g 1
g 1 g = gg 1 = e
to the same product of G it is called a subgroup of G. When the number of elements

of the group G is finite, it is called a finite group and the number of elements is
called the order of the group G and is denoted by |G|. As a typical example of a
finite group, we consider the permutation group Sn of degree n, which is defined
as the set of permutations among n letters. The order of Sn is n!. When an element of
Sn is written as the exchange between two letters a and b, it is written as (a, b) and
is called a transposition. It is known that any permutation is given as products of
transpositions. When an element of Sn is written as the product of an even number of
transposition, it is called an even permutation. The set of even permutations among
n letters is called the alternating group of degree n and is written as An . We say that
a subset G of the group G generates G when all of elements of G can be written as
products of elements of G and their inverses. In this case, the subset G is called a
generator of G. A group G is called finitely generated when there exists a subset
G with finite elements of G that generates G.
For an element g G, the minimum natural integer n satisfying g n = e is called
the order of g. The order of an element g G is a divisor of the order of G.
The group G is called a commutative group when it satisfies the following
commutative law.
Commutative law:
g2 g1 .
Any two elements g1 , g2 G satisfy the condition g1 g2 =
For example, when the Hamiltonian H is given as |11|, the time evolution is given
by the unitary eit H = |00| + eit |11|. By considering the diagonal element of the
basis |1, the time evolution is identified with the unit circle U(1) := {z C| |z| =
1} = {eit | t R}. Since eit1 eit2 = ei(t1 +t2 ) , the unit circle is a commutative group
with respect to the multiplication. The subset Uk := {ei2 j/k | j Z} of U(1) is also
a subgroup. Since any element of Uk can be written as the n-th power of ei2/k
with a natural number n less than k, the subgroup Uk is called a cyclic group and
ei2/k is a generator of Uk . As another commutative group, we have Klein fourgroup, which is composed of four elements e, i, j, k with products i j = ji = k,
jk = k j = i, ki = ik = j, i 2 = j 2 = k 2 = e. This group is denoted by Z2 Z2 .
Exercise 2.1 List up all of elements of the permutation group S3 . Calculate the
orders of all elements.
Exercise 2.2 Show that the subset {e, (1, 2)(3, 4), (1, 3)(2, 4), (1, 4)(2, 3)} of S4 is
a subgroup.
24
T ( g)
Fig. 2.6 Action of group
T ( g )
2.1.2 Homogeneous Space

Now, we assume that the target of interest is written by the set and that the
transformation on the set is given as a group G. That is, for a given element
g G, we have a transformation T (g) on the set as Fig. 2.6, which satisfies the
following condition.
Action of group: The relation T (g1 ) T (g2 ) = T (g1 g2 ) holds and the transformation T (e) is the identity map.
Here, we can show that T (g) is a bijection. In the following, T is called an action of
the group G on the set . Such a situation often occurs among natural phenomena.
A subset of is called an invariant subset with respect to the action T on the set
when the following condition holds.
Invariance:
The relation T (g) holds for any elements g G and .
Therefore, when has an invariant non-trivial proper subset, there exist two elements
and of such that and cannot be mapped to each other via the action T (g)
with any element g G. Conversely, when has no invariant non-trivial proper
subset, the action T is called transitive and the image of an element coincides
with the whole set , i.e., the following transitive law holds.
Transitive law: For any two elements , , there exists an element g G
such that T (g) = .
When there exists a transitive action T for the set , the set is called a homogeneous space. In general, even though the action T is transitive, the set is not
identical to the set G.
In the following, we describe the structure of the homogeneous space based
on the action of G via constructing a homogeneous space essentially identical to .
For this purpose, we need to define the equivalence as homogeneous spaces. Two
actions T1 , T2 on two sets 1 , 2 are called equivalent with each other when there
exists a map T3 from 1 to 2 that satisfies the condition (2.1) as Fig. 2.7.
T3 T1 (g) = T2 (g) T3 .
(2.1)
25
Fig. 2.7 Equivalence

between two actions
T1 ( g )
T1 ( g )1
T3
T2 ( g )
T2 ( g ) 2
Fig. 2.8 Residue class and

quotient space
g'
gH
G
g ' gH
In particular, when these actions are transitive as well as equivalent, we write 1

=
2 .
From the condition of action of group, we can show that the subset {g
G|T (g)0 = 0 } of G is a subgroup for a given 0 (Exercise 2.6). Since the
subgroup does not change the element 0 , it is called a stabilizer of 0 . In fact, the
homogeneous space can be reconstructed from the group G and the stabilizer H .
For an element g G, the subset [g] := g H is called a residue class with respect
to the subgroup H . (In general, for a subset B G, the subsets g B := {gh|h B}
and Bg := {hg|h B} are defined.) Then, the set of all of residue classes is written as G/H = {g H |g G} and is called the quotient space. When the number of
elements of G/H is finite, the number is written as |G : H | and is called the index
of H with respect to G. The map g g H is called the projection with respect to
the residue G/H . Now, for an element g G, we can naturally define the action
g H (g g)H on the quotient space G/H as Fig. 2.8. When the subgroup H is the
stabilizer of the element 0 with respect to the action T on the set , the homogeneous space is equivalent to the quotient space G/H as homogeneous spaces.
Although the stabilizer H depends on the choice of the element 0 , the quotient space
G/H is equivalent to the quotient space based on another choice of the element 0 .
Next, for an element g G, we define the action g gg g 1 on the group G.
When a subgroup H is invariant with respect to this action, the subgroup H is
called an invariant subgroup or a normal subgroup. This condition is equivalent
to the condition that g H = H g for any element g G as Fig. 2.9. For example, for a
26
g
H
gH
Hg
Fig. 2.9 Residue class and quotient space
g2
g1
g1 g2
f ( g2 )
f ( g1 )
f ( g1 g2 )
Fig. 2.10 Homomorphism
subgroup M G, the subset C G (M) := {h G|gh = hg, g M} is a subgroup

of G and is called a centralizer of M. In particular, C G (G) is a commutative normal
subgroup of G. So, we simplify it to C(G) and call it the center of G. Since the center
C(G) is a commutative group, any subgroup of C(G) is also a normal subgroup of G.
For a subset M G, the subset NG (M) := {g G|g M = Mg} is a subgroup of G,
and is called the normalizer of M. Especially, when M is a subgroup of G, NG (M)
is a normal subgroup of M. For an element a G, the set a G := {gag 1 |g G}
is called the conjugacy class of G containing a. For example, the conjugacy class
containing a C(G) is {a}.
For a normal subgroup H , the quotient space G/H forms a group with respect
to the multiplication g H g H := (gg )H and is called the residue group or the
quotient group for H . The residue group for the center C(G) is called the index
group. A map f from a group G 1 to a group G 2 is called a homomorphism
when the condition f (g1 ) f (g2 ) = f (g1 g2 ) holds for any two elements g1 , g2 G 1
as Fig. 2.10. In addition, when f is a bijection, f is called an isomorphism. When
there is an isomorphism from a group G 1 to a group G 2 , the two groups G 1 and
G 2 are called isomorphic to each other and we write G 1
= G 2 . For example, the
set Z of integers is a normal subgroup of the set R of real numbers with respect to
addition, and the quotient group R/Z is isomorphic to U(1). Similarly, the quotient
group Zk := Z/kZ is isomorphic to the cyclic group Uk . Hence, the group Zk is also
called the cyclic group. In particular, when G 1 = G 2 , an isomorphic map f from

Fig. 2.11 Kernel of
homomorphism f
27
Kerf
G1
g2
g1
f ( g1 )
f ( g2 )
G2
G 1 to G 2 is called an automorphism of G 1 . An action T of a group G 1 on a group

G 2 is called an automorphic action of the group G 1 on the group G 2 when T (g) is
an automorphism of G 2 for any g G 1 . For a homomorphism f from a group G 1 to
a group G 2 , the kernel of f is defined as f 1 (e) and is written as Ker f (Fig. 2.11).
It can be shown to be a normal subgroup of G 1 . Then, we have G 1 / Ker f
= G2,
which is called the homomorphism theorem.
When a group G do not have any subgroup except for {e} and G, the group G is
called simple. When a group G has a normal subgroup, we can analyze the structure
of the group G via the analysis on the structure of the normal subgroup. That is, the
analysis of the group can be reduced into that of simple groups. For example, for a
group G, the subgroup [G, G] of G generated by the set {g 1 h 1 gh|g, h G} is a
normal subgroup of G and is called the commutator subgroup of G.
Lemma 2.1 The following conditions are equivalent for a subgroup H of a group G.
(1) The subgroup H is a normal subgroup and G/H is a commutative group.
(2) [G, G] H .
Proof (1)(2): Assume that the subgroup H satisfies the condition (1). Since
[a][b] = [b][a] holds for a, b G, [a]1 [b]1 [a][b] is the identity element. This
fact implies that a 1 b1 ab H . Hence, we obtain the condition (2).
(2)(1): Assume that the subgroup H satisfies the condition (2). Since the relation
d 1 cd = cc1 d 1 cd c[G, G] H holds for c H and d G, the subgroup H is
a normal subgroup. Further, since any elements a, b G satisfies ab = baa 1 b1 ab

and a 1 b1 ab [G, G] H , we obtain [a][b] = [b][a].
Example 2.1 The subgroup Z2 Z2 given in Exercise 2.2 is a normal subgroup of
S4 (see Exercises 2.3 and 2.4). Then, we have the quotient group S4 /(Z2 Z2 ),
which is isometric to S3 (see Exercise 2.5).
28
Exercise 2.3 Show that the subgroup given in Exercise 2.2 is isometric to Klein
four-group Z2 Z2 .
Exercise 2.4 Show that the subgroup given in Exercise 2.2 is a normal subgroup of
S4 .
Exercise 2.5 Show that the quotient group S4 /(Z2 Z2 ) is isometric to S3 .
Exercise 2.6 Show that a subset {g G|T (g)0 = 0 } of a group G is a subgroup
of G.
Exercise 2.7 Show that G/H is a homogeneous space equivalent to when is
a homogeneous space for G and H is the stabilizer of an element 0 .
Exercise 2.8 Show that [G, G] is a normal subgroup of a group G.
2.2 Extension of Group

2.2.1 General Case
When two different sets of transformations are given, we often consider the pair
of two transformations. When these two transformations act independently, we can
easily handle these transformations. However, these two transformations effect each
other in general. In such a case, it is very difficult to describe the set of the pair
of two transformations. Group extension is a concept to handle such a complicated
situation.
Given two different groups H and K , the product set H K is a group under
the product (h, k)(h , k ) := (hh , kk ), which is called the direct product (group)
of H and K , and is denoted by H K . (In this book, the direct product as a set is
denoted by H K and the direct product as a group is denoted by H K .) The direct
product corresponds to the case when the transformations given by the groups H and
K act independently. For example, Klein four-group is isometric to the product group
Z2 Z2 . So, it is written as Z2 Z2 .
Next, we consider the case when the transformation by the group K effects the
action of the group H . To handle such a situation, we focus on an automorphic action
T of the group K on the group H . Then, the following product is defined for the
direct product set H K by reflecting the structure of the automorphic action T .
(h, k)(h , k ) := (hT (k)(h ), kk ).
(2.2)
This product is different from the product in the direct product group H K . The
group based on this product is called a semi direct product of H and K and is
denoted by H K . When the elements h H k K are identified with (h, e) and
(e, k), respectively, we have
29
H = Kerf
Fig. 2.12 Regular

hexahedron group
( h ', k ') G
( h, k )
h'
h
hk = (h, k), khk 1 = T (k)(h).
k'
(2.3)
Further, when H and K are identified with the subsets {(h, e)}hH and {(e, k)}kK ,
respectively, H is a normal subgroup of H K and K is a subgroup H K .
Example 2.2 Consider the case when H = Z2 Z2 = {e, i, j, k} and K = S3 .
Define the action T of S3 on Z2 Z2 as the permutation among three elements
{i, j, k}. Then, the semi direct product (Z2 Z2 ) S3 is isometric to S4 because
the map h khk 1 for h Z2 Z2 and k S3 behaves as the permutation on
the three elements {(1, 2)(3, 4), (1, 3)(2, 4), (1, 4)(2, 3)} with the correspondence
(1, 2)(3, 4) 3, (1, 3)(2, 4) 2, (1, 4)(2, 3) 1 (see Exercise 2.9). On the other
hand, the group of rotations preserving the regular hexahedron is called the regular
hexahedron group. The group (Z2 Z2 ) S3 is isometric to the regular hexahedron
group as follows (Fig. 2.12). The elements of i, j, k correspond to the rotations
of x, y, z axises. The elements of S3 correspond to the rotations permuting x, y, z
axises. Using this correspondence, we have the above isometric relation, because the
rotation except for the rotations of x, y, z axises permutes these three rotations.
The direct product and the semi direct product are methods to give a group G
that contains H as a normal subgroup and satisfies the condition G/H
= K . Even
if a group structure for the direct product set H K is restricted to satisfy these
conditions, it is not limited to the direct product and the semi direct product. In
general, such a group G is called an extension of the group H by the group K .
Further, an extension of a group is not necessarily given as the direct product or the
semi direct product. In fact, studying the extensions of a group, we can understand
what transformation is possible for a given set G of transformations caused by the
groups H and K under the condition G/H
= K . When the quotient G/H is given
by a homomorphism f , this situation is illustrated as Fig. 2.13. Especially, when
the group H is the center C(G) of an extension G of the group H as Fig. 2.14, the
30

-axis
Fig. 2.13 Relation among

G, H , and K
Z-axis
X-axis
Fig. 2.14 Central extension
T (k )
H = C (G )
extension G plays an important role and is called a central extension of the group
H by the group K .
Let G be an extension of the group H . When we choose a representative gk G
for an element k K , the residue class is given as gk H and the group G is written as
gk1 H gk2 H gkn H . Now, for an element k K , we define the automorphism
T (k) of the group H as
T (k)(h) := gk hgk1 , k K , h H.
(2.4)
Then, for two elements k, k K , we define the element (k, k ) H by

gk gk = (k, k )gkk .
(2.5)
Then, for two elements h, h H , we have

(hgk )(h gk ) = hgk h gk1 gk gk = hT (k)(h )(k, k )gkk .
(2.6)
Thus, since G is a group, the combination of (2.4) and (2.5) yields the relations
T (k)(T (k )(h)) = (k, k )T (kk )(h)(k, k )1
(k, k )(kk , k ) = T (k)((k , k ))(k, k k ).
(2.7)
(2.8)
31
When the above two conditions are satisfied, {T (k), (k, k )} is called a factor system of K . The factor system depends on the choice of the representative gk . Due to
the above discussion, the factor system is determined by the extension of the group.
Conversely, we have the following lemma [9, Theorem 8.1].
Lemma 2.2 When a factor system {T (k), (k, k )} satisfies the conditions (2.7) and
(2.8), the element e H := (e, k)1 H does not depend on k K . Then, we can
define the following product for the set H K by replacing hgk by (h, k);
(h, k)(h , k ) := (hT (k)(h )(k, k ), kk ).
(2.9)
The product satisfies the condition for the group, in which, the identify element is
e H,K :=(e H , e) and the inverse element of (h, k) is (e H (k 1 , k)1 T (k 1 )(h)1 , k 1 ).
This group contains the subset {(h, e)}hH as a normal subgroup. Hence, the subset
{(h k , k)}kK satisfies the conditions (2.4) and (2.5) for gk when the element h k H
is chosen as an arbitrary element of H .
Hence, the factor system {T (k), (k, k )} satisfies the conditions (2.7) and (2.8) if
and only if there exists an element gk G satisfying (2.4) and (2.5) for an element
k K . In particular, when the identity element e is chosen as the representative ge ,
T (e) is the identity map and the relation (k, e) = (e, k) = e holds. So, the identity
element is given as (e, e). This condition is satisfied when h e is chosen to be e in
Lemma 2.2. Such a factor system is called a normalized factor system1
Further, when (k, k ) = e, the product (2.9) is the product of the semi direct
product group. So, the extension of the group is a generalization of the semi direct
product group.
Proof of Lemma 2.2 The condition (2.8) with k = e implies that
(k, e) = T (k)((e, k )),
which yields that T (k)1 ((k, e)) = (e, k ). That is,
T (k)1 ((k, e)1 ) = (T (k)1 ((k, e)))1 = (e, k )1 .
(2.10)
Since k and k are chosen independently, both sides are independent of k and k .
Hence, we have
(h, k)(e H , e) = (h, k)(T (k)1 ((k, e)1 ), e)
=(hT (k)(T (k)1 ((k, e)1 ))(k, e), ke)
=(h(k, e)1 (k, e), k) = (h, k).
The condition (2.8) with k = e implies that
1 The
book [99] calls a normalized factor system a factor system.
(2.11)
32
T (e)(T (k )(h)) = (e, k )T (k )(h)(e, k )1 .

Since any element h can be written as T (k )(h), we have
(e, k )1 T (e)(h )(e, k ) = h .
(2.12)
Hence,
(e H , e)(h, k) = ((e, k)1 , e)(h, k) = ((e, k)1 T (e)(h)(e, k), e) = (h, k).
(2.13)
Thus, (e H , e) satisfies the condition for the identity element.
Next, we show that the inverse element of (h, k) is (e H (k 1 , k)1 T (k 1 )
(h)1 , k 1 ).
(e H (k 1 , k)1 T (k 1 )(h)1 , k 1 )(h, k)
=(e H (k 1 , k)1 T (k 1 )(h)1 T (k 1 )(h)(k 1 , k), k 1 k) = (e H , e).
(2.14)
The condition (2.8) with k = k 1 and k = k implies that

T (k)((k 1 , k)) = (k, k 1 )(e, k)(k, e)1 .
(2.15)
Hence, we have
T (k)(e H (k 1 , k)1 T (k 1 )(h)1 )
=T (k)(e H )T (k)((k 1 , k))1 T (k)(T (k 1 )(h 1 )))
(a)
=(k, e)1 ((k, k 1 )(e, k)(k, e)1 )1 (k, k 1 )T (e)(h 1 )(k, k 1 )1

=(e, k)1 T (e)(h 1 )(k, k 1 )1
=(e, k)1 T (e)(h 1 )(e, k)(e, k)1 (k, k 1 )1
(b) 1
=h e H (k, k 1 )1 ,
(2.16)
where (a) and (b) follow from (2.15) and (2.12), respectively. Thus,
e H (k 1 , k)1 T (k 1 )(h)1 = T (k)1 (h 1 e H (k, k 1 )1 ).
(2.17)
Therefore,
(h, k)(e H (k 1 , k)1 T (k 1 )(h)1 , k 1 )
=(h, k)(T (k)1 (h 1 e H (k, k 1 )1 ), k 1 )
=(hh 1 e H (k, k 1 )1 (k, k 1 ), kk 1 ) = (e H , e).
(2.18)
33
Combining (2.14) and (2.18), we find that the inverse element of (h, k) is (e H (k 1 ,

k)1 T (k 1 )(h)1 , k 1 ). For the associative rule, see Exercise 2.10.
Exercise 2.9 Show that the map h khk 1 given in Example 2.2 gives the permutation on the three elements {(1, 2)(3, 4), (1, 3)(2, 4), (1, 4)(2, 3)} with the correspondence (1, 2)(3, 4) 3, (1, 3)(2, 4) 2, (1, 4)(2, 3) 1.
Exercise 2.10 Show the associative rule for the product given in Lemma 2.2.
2.2.2 Central Extension of a Commutative Group

Since a central extension of a commutative group plays an important role in the latter
section, we discuss the structure of the set of central extensions. When the group
H is a commutative group, the extension of the group H by the group K based on
the factor system {T (k), (k, k )} is a central extension if and only if T (k) is the
identity map for any element k K . In the following, T (k) is fixed to be the identity
map because we discuss only the central extension. Hence, when the group H is a
commutative group, a factor system describing a central extension can be obtained
from the set {(k, k )} H satisfying the condition:
(k, k k )(k , k ) = (k, k )(kk , k )
(2.19)
because (2.8) is simplified to (2.19). This condition is equivalent to the existence

of the element gk G satisfying (2.5) for an element k K . Then, the product can
be simplified as (h, k)(h , k ) = (hh (k, k ), kk ). This book addresses a central
extension mainly for the case when H is a commutative group because such a case is
closely related to a projective representation. So, the subset {(k, k )} H satisfying
(2.19) is called a factor system. In particular, when (k, e) = (e, k) = e, it is a
normalized factor system, however, we do not assume this condition for a factor
system.
Now, we discuss the set Z (K : H ) of factor systems {(k, k )}k,k . This set
has the structure of group with the product { (k, k )}k,k {(k, k )}k,k :={ (k, k )
(k, k )}k,k . When {e} expresses the factor system satisfying the condition (k, k ) =
e for any two elements k, k K , {e}k,k is the identity element of Z (K : H ). The
central extension corresponding to {e}k,k is the direct product group of H K .
Next, we focus on the subgroup B(K : H ) of Z (K : H ) defined by the set of factor

systems {h k h k h 1
kk }k,k by using {h k }k . Two factor systems {(k, k )} and { (k, k )}
are called equivalent to each other when there exists a subset {h k }kK H such that

(k, k ) = h k h k h 1
kk (k, k ). This equivalence condition holds if and only if the two

factor systems {(k, k )} and { (k, k )} belong to the same element of the quotient
group H (K : H ) := Z (K : H )/B(K : H ). So, the central extensions corresponding
to {(k, k )} and { (k, k )} are isometric to each other if and only if they are equivalent to each other. Thus, there is a one-to-one correspondence between an element
34
of the quotient group H (K : H ) and a central extension of the commutative group

H by the group K . In particular, when the factor system is given by an element of B,
the constructed central extension is isomorphic to the direct product group H K .
Hence, H (K : H ) is called the group of extension of the group H by the group
K .2 Using this relation H (K : H ) = Z (K : H )/B(K : H ), we can characterize the
order of an element of the group of extension H (K : H ) as follows [9, Theorem 8.5].
Theorem 2.1 The order of an element of the group of extension H (K : H ) is a
divisor of the order of H when H is a finite group. Also, the order is a divisor of the
order of K when K is a finite group. That is, the order is a common divisor of the
orders of H and K when H and K are finite groups. In particular, when the orders of
H and K are coprime, a central extension of H by K is limited to the direct product
H K.
Proof Assume that H is a finite group. Since (k, k )|H | = e for k, k K ,
n
{h k h k h 1
kk }k,k is the identity element of Z . Hence, the order of an element of
H (K : H ) is a divisor of the order of H .

Assume that K is a finite group. Using bk := k K (k, k ), we have
(k, k )|K | bkk = (k, k )|K |
(a)

k K
(kk , k ) =
k K
((k , k )(k, k k )) = (
((k, k )(kk , k ))
k K
(k , k ))(
k K
(k, k k )) = bk bk ,
k K
1
where (a) follows from (2.19). So, we have (k, k )|K | = bk bk bkk
B. Hence, the
order of an element of the group of extension H (K : H ) is a divisor of the order
of K .
Exercise 2.11 Show the relation (2.3) with respect to the semi direct product H K .
Exercise 2.12 Show that H is a normal subgroup of the semi direct product H K .
Exercise 2.13 Show that the factor system {T (k), (k, k )} gives a central extension
if and only if T (k) is the identify map for an element k K when H is a commutative
group.
2.2.3 Examples for Central Extensions

In the following, we list examples for central extensions.
when the automorphism T (k) is not necessarily the identity, when T (k)T (k ) = T (kk ), we
can define the group of extension by replacing the condition (2.19) by the condition (2.8). In this
case, the extension corresponding to {e}k,k is the direct semi product group of H K , not the
direct product group of H K . Also, we can show Theorem 2.1 in the same way.
2 Even
35
Example 2.3 (Cyclic group Z pn ) We focus on the cyclic group Z pn for a prime
number p. For m < n, the subgroup {kp m |kp m Znp } of Z pn is a normal subgroup
isomorphic to Z pnm . Then, Z pn /Z pnm is isomorphic to Z pm . Hence, Z pn is a central
extension of Z pnm by Z pm . For two elements k1 , k2 Z pm , the factor system is
given as

(k1 , k2 ) =
0 when k1 + k2 p m
1 when k1 + k2 < p m .
(2.20)
Thus, for the group Z pn , we can construct a sequence of normal subgroups {0}
Z p1 Z p2 Z pn .
Example 2.4 (Discrete Heisenberg group on Zd with degree 1) Assume that d is an

odd number. The set Z2d Ud forms a group with the product (s, t, dn ) (s , t , dn ) :=

(s + s , t + t , dn+n +st ts ), where d is the d-th root of 1. This group is called the
discrete Heisenberg group on Zd with degree 1 and is written as H(2, Zd ). The
subset {(0, 0, dn )} is a normal subgroup isomorphic to Ud , hence it is also written as
Ud . Also the quotient group H(2, Zd )/ Ud is isomorphic to Z2d . When we choose the
representative (s, t, 1) H(2, Zd ) of the element of H(2, Zd )/ Ud corresponding to

an element (s, t) Z2d , we have (s, t, 1) (s , t , 1) = (s + s , t + t , dst ts ). That
is, H(2, Zd ) is a central extension of Ud by Z2d , and the factor system is given as

((s, t), (s , t )) = dst ts . The center of H(2, Zd ) is Ud .
Assume that d is an even number. The set Z2d U2d forms a group with the prodn+n +st ts
n
n
uct (s, t, 2d
) (s , t , 2d
) := (s + s , t + t , 2d
). The group is called the
discrete Heisenberg group on Zd with degree 1 and is written as H(2, Zd ). Similar
n
)} is denoted by U2d , and we have
to the odd case, the normal subgroup {(0, 0, 2d
2
Zd = H(2, Zd )/ U2d . Then, H(2, Zd ) is a central extension of U2d by Z2d , and the
factor system can be calculated similarly. The center of H(2r, Zd ) is U2d .
Example 2.5 (Discrete Heisenberg group on Zd with multi-degree) Assume that d
n
is an odd number. The subset Z2r
d Ud forms a group with the product (s, t, d )

n+n
+(s,t
)
(t,s
)
r
Z
Z
), where (s, t)Z := j=1 s j t j . The
(s , t , dn ) := (s + s , t + t , d
group is called the discrete Heisenberg group on Zd with degree r and is
written as H(2r, Zd ). The normal subgroup {(0, 0, dn )} is written as Ud . The quotient group H(2r, Zd )/ Ud is isomorphic to Z2r
d . When we choose the representative (s, t, 1) H(2r, Zd ) of the element of H(2r, Zd )/ Ud corresponding to an
(s,t )Z (t,s )Z

).
element (s, t) Z2r
d , we have (s, t, 1) (s , t , 1) = (s + s , t + t , d
That is, H(2r, Zd ) is a central extension of Ud by Z2r
d . The factor system is given as

((s, t), (s , t )) = d(s,t )Z (t,s )Z . The center of H(2r, Zd ) is Ud .
Assume that d is an even number. The set Z2r
d U2d forms a group with the
n+n +(s,t )Z (t,s )Z
n
n

). The group is
product (s, t, 2d ) (s , t , 2d ) := (s + s , t + t , 2d
called the discrete Heisenberg group on Zd with degree r and is written as H(2r, Zd ).
n
)} is denoted by U2d , and we
Similar to the odd case, the normal subgroup {(0, 0, 2d
2r
have Zd = H(2r, Zd )/ U2d . Then, H(2r, Zd ) is a central extension of U2d by Z2r
d .
The factor system can be similarly calculated. The center of H(2r, Zd ) is U2d .
36
Example 2.6 (Discrete Heisenberg group on Fq with degree 1) Let q be a power p n

of a prime p, and Fq be the n-th algebraic extension of the finite filed F p := Z/ pZ.
Assume that p is an odd prime. The set Fq2 U p forms a group with the product

+tr (st ts )
(s, t, np ) (s , t , np ) := (s + s , t + t , n+n
), where tr z is defined as folp
lows. Consider the linear space Fmp isomorphic to Fq . Let Mz be the matrix on the
finite filed F p corresponding to the linear map x zx defined by the multiplication
under the finite filed Fq . Then, we denote Tr Mz by tr z. The group is called the
discrete Heisenberg group on Fq with degree 1, and is written as H(2, Fq ). The subset {(0, 0, np )} is a normal subgroup isomorphic to U p , and is written as U p . The
quotient group H(2, Fq )/ U p is isomorphic to Fq2 . When we choose the representative (s, t, 1) H(2, Fq ) of the element of H(2, Fq )/ U p corresponding to an ele

ment (s, t) Fq2 , we have (s, t, 1) (s , t , 1) = (s + s , t + t , trp (st ts ) ). That is,
H(2, Fq ) is a central extension of U p by Fq2 . The factor system is ((s, t), (s , t )) =

trp (st ts ) . The center of H(2, Fq ) is U p .
Assume that p = 2. The set Fq2 U4 forms a group with the inner product
(s, t, 4n ) (s , t , 4n ) := (s + s , t + t , 4n+n +tr (st ts ) ). The group is called the

discrete Heisenberg group on Fq with degree 1, and is written as H(2, Fq ). Similar
to the odd prime power case, the normal subgroup {(0, 0, 4n )} is denoted by U4 , and
we have Fq2
= H(2, Fq )/ U4 . Then, H(2, Fq ) is a central extension of U4 by Fq2 . The
factor system can be calculated in the same way. Also, H(2, Fq ) is the center of U4 .
Example 2.7 (Discrete Heisenberg group on Fq with multi-degree) Assume that

p is an odd prime. The set Fq2r U p forms a group with the product (s, t, np )

+(s,t )F (t,s )F
(s , t , np ) := (s + s , t + t , n+n
), where (s, t)F := rj=1 tr (s j t j ).
p
The group is called the discrete Heisenberg group on Fq with degree r , and is
written as H(2r, Fq ). Then, the normal subgroup {(0, 0, np )} is denoted by U p , and
the quotient group H(2r, F p )/ U p is isomorphic to F2r
p . When we choose the representative (s, t, 1) H(2r, Fq ) of the element of H(2r, Fq )/ U p corresponding to

(s,t )F (t,s )F
),
an element (s, t) F2r
p , we have (s, t, 1) (s , t , 1) = (s + s , t + t , p

)F (t,s )F
which implies that ((s, t), (s , t )) = (s,t
.
That
is,
(2r,
F
)
is
the
central
H
q
p

extension of U p by Fq2r . The factor system is ((s, t), (s , t )) = trp (st ts ) . The center of H(2r, Fq ) is U p .
Assume that the prime p is 2. The set Fq2r U4 forms a group with the product
(s, t, 4n ) (s , t , 4n ) := (s + s , t + t , 4n+n +(s,t )F (t,s )F ). The group is called the

discrete Heisenberg group on Fq with degree r , and is written as H(2r, Fq ). Similar
to the odd prime power case, the normal subgroup {(0, 0, 4n )} is denoted by U4 , and
we have Fq2r
= H(2r, Fq )/ U4 . Then, H(2r, Fq ) is a central extension of U4 by Fq2r .
The factor system can be similarly calculated. The center of H(2r, Fq ) is U4 .
2.3 Representation and Projective Representation
37

2.3.1 Definitions of Representation and Projective
Representation
Next, we consider what transform can be realized of a given quantum system H based
on a given group G. For this purpose, we need several additional notations. Let gl(H)
be the set of linear maps on a finite-dimensional Hilbert space H, and GL(H) be
the set of invertible elements of gl(H). The set GL(H) has a group structure when
the product is defined as the composition of maps. So, it is called the general linear
group on H. In particular, the subset of elements of GL(H) with determinant 1 is
called the special linear group on H and is written as SL(H). An element of GL(H)
is called a unitary map when it preserves the inner product. The set of unitary maps
is written as U(H), and its subgroup U(H) SL(H) is as SU(H). The set U(H)
(SU(H)) has a group structure when the product is given as the composition of
maps. So, it is called the (special) unitary group on H. A homomorphism f from a
group G to a group GL(H) is called a representation of G on H, and H is called
its representation space. In particular, a representation f on H is called a unitary
representation when the image of f is included in U(H). A unitary representation
plays an important role in quantum theory as follows. A unitary representation f of
group G gives a transformation f(g)f(g) for density matrices by an element
g G as Fig. 2.15. That is, a unitary representation f determines an action of a
group G on the set of density matrices. However, a physical transformation by a
group G appearing in quantum theory cannot be necessarily limited to a unitary
representation f.
To cover all of physical transformations by a group G, we need to consider a
larger class of actions of a group G. A map f from a group G to the group GL(H)
is called a projective representation when it satisfies the following condition. For
any elements g, g G, there exists a real number (g, g ) R such that
f(g)f(g ) = ei(g,g ) f(gg ).
(2.21)
In particular, it is called a projective unitary representation when the image of f is

included in U(H). One might consider that the phase factor ei(g,g ) breaks the physical
symmetry. However, the phase factor does not appear in the action on the set of the
density matrices. Hence, even when f is not a representation but a projective unitary
Fig. 2.15 Transformation

for density matrix
( g)
( g ) ( g )
S (H )
38
representation, we have the relation f(g)f(g )(f(g)f(g )) = f(gg )f(gg ) . That is,

the phase factor ei(g,g ) does not appear in the action with respect to the density
matrix . On the other hand, when a representation or a projective representation is
not unitary, it does not transform a density matrix to another density matrix. Thus, in
quantum system, actions of a group are written by projective unitary representations,
which contain unitary representations.
Since possible unitary representation and possible projective unitary representations are limited to a small class due to the property of the group of interest, the
transformation with respect to the given quantum system derives various restrictions for the states of the quantum system. Sometimes, these limitations simplify
our analysis. This book explains this kind of restrictions for the states based on the
properties of the group. Although only a unitary representation and a projective unitary representation give physical transformations of quantum systems, a non-unitary
representation or a non-unitary projective representation often plays an important
role for the analysis of a unitary representation or a projective unitary representation.
Due to the above reason, this book covers a representation and a projective representation that are not necessarily unitary. The role of representations is not limited to
analysis of physical transformations. A density matrix or a physical quantity is often
determined by the properties of a representation. For example, given a (projective)
unitary representation f of G, an important physical quantity is often given as the
Hermitian matrix i log f(g) with use of an element g G or the Hermitian matrix
i limt0 f(gt t)I with use of a 1-dimensional subgroup {gt }tR of G. As explained in
Sect. 4.4.1, given a non-unitary (projective) representation f of GL(n, C), a density
matrix is often given as f(g) with use of g G. In quantum theory, we are usually
interested in the matrix forms of a physical quantity and a density matrix. In the above
case, their matrix forms are often given via the properties of the group of interest. In
this way, a non-unitary (projective) representation f often plays an important role in
quantum theory.
On the other hand, since unitary operators can be defined even on an infinitedimensional Hilbert space H, we can define the set U(H) even for an infinitedimensional Hilbert space H. Hence, we can define a unitary representation and
a projective unitary representation for an infinite-dimensional Hilbert space H in
the same way as the finite-dimensional case. However, a non-unitary representation
is not useful in the infinite-dimensional case, even though it can be defined in the
infinite-dimensional case. Hence, we discuss a non-unitary representation and a nonunitary projective representation only in the finite-dimensional case. That is, when H
is finite-dimensional, a representation and a projective representation are not limited
to unitary. However, when H is infinite-dimensional, they are limited to unitary.
In the following, we summarize mathematically important concepts of representation and projective representation. A representation f is called faithful when it is
an injective homomorphism. In this case, any element of G except for the identity
element e changes the system. Given two representations f1 and f2 on H1 and H2 ,
the two representations f1 and f2 are isomorphic to each other when there exists an
invertible linear map A from H1 to H2 such that
Af1 (g) = f2 (g)A, g G.
39
(2.22)
Then, A is called an isomorophism (with respect to representation of the group

G). As mentioned in Sect. 2.8, any representation of a finite group G is isomorphic
to a unitary representation of G. Similarly, two unitary representations are called
(unitarily) isomorphic to each other when there exists a unitary operator A satisfying (2.22). The concept of isomorphic can be extended to the case of projective
representations. Two projective representations f1 and f2 are called isomorphic to
each other when there exists a linear map A satisfying (2.22). Two projective unitary
representations f1 and f2 are called (unitarily) isomorphic to each other when there
exists a unitary operator A satisfying (2.22).
Given two representations f1 and f2 of group G on H1 and H2 , the representation
f1 f2 defined below on the direct sum space H1 H2 is called the direct sum
representation of f1 and f2 , and the representation f1 f2 defined below on the
tensor product space H1 H2 is called the tensor product representation of f1
and f2 .
f1 f2 (g) := f1 (g) f2 (g), f1 f2 (g) := f1 (g) f2 (g).
(2.23)
When f1 and f2 are projective representations, the tensor product projective representation f1 f2 can be defined in the same way. However, as explained in the end of
Sect. 2.4, the direct sum projective representation f1 f2 can be defined only under
the restricted case.
When a representation f on the direct sum space H1 H2 is given as a direct
sum representation f1 f2 , we can easily discuss its property. In general, to discuss
the matrix form of the representation f(g), we need to calculate the matrix entries.
However, when the representation f is a direct sum representation, there does not
exist a non-zero matrix entry that expresses the transformation from the space H1 to
the space H2 . Similarly, there does not exist a non-zero matrix entry of the opposite
direction.
For example, when the dimensions of the spaces H1 and H2 are d1 and d2 , respectively, we need to consider (d1 + d2 )2 entries. However, since 2d1 d2 entries are 0 as
explained in (2.24), it is enough to consider the remaining d12 + d22 entries.

f(g) =

f1 (g) 0
.
0 f2 (g)
(2.24)
Further, since the remaining d12 + d22 entries can be divided into d12 entries and d22
entries, the treatment is much easier than the case when the representation f is not a
direct sum representation. The difficulty of analysis of a representation on a higher
dimensional space depends on whether it is a direct sum representation.
A (projective) representation f of a group G on H is called reducible when there
exists a subspace K of H satisfying the following invariance condition except for
{0} nor H.
40
Invariance:
f(g)u K, u K, g G.
In this case, a subspace K is called an invariant subspace of H. We can define f(g)|K

as a restriction of f(g) to the invariant subspace K. The map f|K : g f(g)|K is a
representation of G on K, and is called the subrepresentation of f on K. Conversely,
the representation f is called irreducible when there does not exist a non-trivial subspace K satisfying the above invariance condition. The representation f is called
decomposable when there exist two non-trivial subspaces K1 and K2 satisfying the
above invariance condition and the relation H = K1 K2 , i.e., the representation f
is given as the direct sum representation f1 f2 of the subrepresentations f1 and f2 of
f on the respective subspaces. In particular, in the case of (projective) unitary representation, the above given invariant subspaces K1 and K2 are orthogonal. Conversely,
the representation f is called indecomposable when it is not decomposable.
Although any irreducible (projective) representation (unitary representation) f is
indecomposable, we have the following lemma as a converse statement.
Lemma 2.3 An indecomposable (projective) unitary representation f is irreducible.
That is, a reducible (projective) unitary representation f is decomposable, and is
given as the direct sum representation of the subrepresentations on the invariant
subspace.
Proof We show its contraposition. Assume that a (projective) unitary representation
f of a group G on H is reducible. Then, there exists a non-trivial subspace K of H
that is invariant for the action of G. As shown later, the orthogonal complement K
of K is also an invariant subspace for the action of G. Hence, we can conclude that
a (projective) unitary representation f is decomposable.
Since any two vectors v1 K and v2 K satisfy f(g)(v1 + v2 )2 = v1 +
1
v2 2 , we have f(g)v1 , f(g)v2 = 0. Since f(g)f(g 1 ) = ei(g,g ) I (f(g)f(g 1 ) = I
in the case of projective representation), the image of K with respect to the linear
map f(g) is K. Hence, f(g)v2 is contained the orthogonal complement K of K. Thus,

K is an invariant subspace for the action of G.
However, there exists a non-unitary representation that is indecomposable and
is not irreducible. For example, the following representation of the group R with
addition on the space C2 is indecomposable and is not irreducible.

f(t) :=

1t
, t R.
01
(2.25)
A (projective) representation f on H is called completely reducible when it

is given as a direct sum representation of (projective) irreducible representations.
Hence, when H is finite-dimensional, Lemma 2.3 guarantees that a (projective) unitary representation f is completely reducible. When the representation is completely
reducible, all of matrix entries of the representation can be calculated from all of
the matrix entries of irreducible subrepresentations of this representation. Hence,
to discuss (projective) unitary representation on the finite-dimensional space, it is
enough to discuss irreducible subrepresentations.
41
Let G be the set of indexes describing the irreducible unitary representations of

the group G. In this notation, we identify representations isomorphic to each other.
we denote the corresponding representation space by U (G), and the corFor G,
responding representation by f . Especially, we denote the set of finite-dimensional
representations by G f . The dimension of the representation space U (G) is written
. If there is no possibility for confusion,

as d , and its CONS is written as {|; j}dj=1
U (G) is simplified to U .
Hence, when a unitary representation f is completely reducible, the representation
space H has the following isomorphic representation:
H
=
U (G) Cn ,
(2.26)
where n is the multiplicity of the irreducible representation space U (G) in the space
H. Then, f is written as G n f , and the RHS of (2.26) is called the irreducible
decomposition of the representation f. Under the above irreducible decomposition,
is the CONS of the space Cn corresponding to the multiplicity of U (G).

{| j}nj=1
Then, we abbreviate |; j | j to |; j; j .
For a finite-dimensional representation f, the function f (g) := Tr f(g) is called a
character. When g and g are contained in the same conjugate class, their characters
are the same function. That is, the relation
f (g ) = f (g 1 g g), g, g G
(2.27)
holds for any representation f. Especially, the character of an irreducible unitary

representation f is called an irreducible character, and is written as .
An irreducible decomposition gives the direct sum decomposition i Hi of the
representation space. By denoting the projection to the subspace Hi by E i , {E i }
forms a projection-valued measurement. Assume that a density matrix is given as
the matrix f(g) by using an element g G and a completely reducible representation
f of G. When the projection-valued measurement {E i } is applied to the system with
the state , the probability to obtain the outcome i is Tr f(g)E i . The probability
Tr f(g)E i is nothing less than the character fi (g) of the irreducible representation fi
on Hi . A character has an important physical meaning in this way.
2.3.2 Schurs Lemma

We have Schurs lemma for an irreducible representation as follows.
Lemma 2.4 Assume that f is a (projective) unitary representation of a group G on
finite-dimensional space H. Then, the following two conditions are equivalent.
(1) The (projective) unitary representation f is irreducible.
42
(2) When a matrix A on H satisfies

f(g)A = Af(g)
(2.28)
for any element g G, the relation A = cI holds with a constant c C.

Hence, Schurs lemma (Lemma 2.4) guarantees that the matrix f(g) is a constant
for any element g of the center C(G).
An important physical quantity A often satisfies the condition (2.28). In this case,
we say that A is invariant with respect to the representation f. When the representation
f is irreducible, Lemma 2.4 guarantees that the physical quantity A is determined to
be a fixed value.
Proof (1)(2): Since A is a matrix with complex entries, it has an eigenvalue c.
Hence, the matrix A cI has a non-trivial Kernel. The Kernel is an invariant subspace for the representation f. The irreducibility implies that the Kernel is H.
(2)(1): We show the contraposition of (2)(1). Assume that (1) does not hold.
Then, f is reducible. So, Lemma 2.3 guarantees that f is decomposable. That is, there
exist two orthogonal invariant subspaces K1 and K2 of H. Hence, H has a direct
sum decomposition H = K1 K2 . Thus, when we define the matrix A to be the pair
of the constant 1 on K1 and the constant 0 on K2 , the relation (2.28) holds, but the
relation A = cI does not hold. So, the condition (2) does not hold.

Lemma 2.4 can be extended to an irreducible (projective) unitary representation
on an infinite-dimensional space H. The proof of the part (2)(1) is still valid even
in the infinite-dimensional case. The other part (1)(2) can be shown as follows.
Proof Firstly, we show the case when A is a self-adjoint operator. We have
f(g)Af(g) = A for any element g G. When the spectral decomposition of A
is given as E(d), the integral f(g)E(d)f(g) is a spectral decomposition
of the operator f(g)Af(g) because f(g) is a unitary operator. Hence, the relation
E([a, b]) = f(g)E([a, b])f(g) holds for any closed interval [a, b] R. Thus, the
range of the operator E([a, b]) is an invariant subspace. The irreducibility guarantees
that the above range coincides with H or {0}. Hence, we find that A is a constant
operator on H. When A is a unitary or a constant times of a unitary, we can show
that A is a constant operator on H by considering the spectral decomposition with
the same way.
Even when A is not necessarily self-adjoint, A A is self-adjoint. So, the relation
1
f(g )A f(g 1 ) = A holds for any element g G. Hence, we can show the relation
f(g)A A f(g) = A A for any element g G by replacing g 1 by g. Since A A is
self-adjoint, A A is a constant operator cI on H. If A has a non-trivial kernel, since
the kernel is an invariant subspace, the irreducibility of f guarantees that the kernel
coincides with the whole space H. Thus, we consider the case when the kernel of A
is {0}. In this case, the image of A is H because A A = cI . So, A is a constant times
of a unitary operator. Therefore, the analysis of this case can be reduced to the case
of unitary operators. Similarly, we can show the same fact when the kernel of A
is {0}.
43
We define the support supp(f ) := {g G|f (g) = 0} of the character f . Then,

we obtain the following lemma as an application of Schurs lemma.
Lemma 2.5 Assume that the index group G/C(G) of a group G is a commutative
group. Further, we assume that f is an irreducible character and the relation f(g) =
f(e) holds for g = e C(G). Then, we have supp(f ) = C(G).
Proof Since Lemma 2.1 guarantees that the commutator group of G is contained
in C(G), we have g gg 1 g 1 C(G) for g G \ C(G) and g G. Since g is not
contained in C(G), there exists an element g G such that z := g gg 1 g 1 =
e. Schurs lemma guarantees that f(z) is a constant . The assumption implies
that = 1. Hence, f (g) = f (g gg 1 ) = f (zg) = f (g), which implies that

f (g) = 0.
Lemma 2.5 can be generalized as follows.
Lemma 2.6 Assume that the index group G/C(G) of a group G is a commutative
group. Also, we assume that f is an irreducible character of normal subgroup N of
G. Additionally, we suppose that an element g C(G) N satisfies f(g) = I if and
only if it is the identity element. Then, we have supp(f ) = C(N ).
Proof Since Lemma 2.1 guarantees that the commutator group of G is contained
in C(G), the relation g gg 1 g 1 C(G) N holds for g N \ C(N ) and g G.
Since g is not contained in C(G), there exists an element g G such that z :=
g gg 1 g 1 = e. Because z N , Schurs lemma guarantees that f(z) is a constant
= 1. So, similar to Lemma 2.5, we obtain the desired argument.

Schurs lemma can be generalized to the completely reducible case as follows.
Lemma 2.7 Assume that a (projective) representation f of a group G on the space
H is irreducibly decomposed as (2.26). Then, the following two conditions for a
matrix A on H are equivalent to each other.
(1) The matrix A can be written in the following way by using a matrix A on Cn :
A=
I A ,
(2.29)
where I is the identity matrix on U .

(2) The relation
f(g)A = Af(g)
(2.30)
holds for any element g G.

When there is no pair of isomorphic subspaces in the irreducible decomposition
(2.26), the meaning of this lemma is clear. In this case,
when A satisfies the commutative condition (2), A can be written to be the form a E by using the projection
44
E to the irreducible space U (G). That is, A has a fixed value on each irreducible
space. However, when there is a pair of isomorphic subspaces in the irreducible
decomposition (2.26), we cannot derive the same conclusion.
Proof Since the part (1)(2) is trivial, we show only the part (2)(1) as follows.
We firstly show (1) for the case when H consists of only one kind of irreducible
n
n
representation,
i.e., H = U C . When the vector u v U C satisfies
A(u v) = i u i vi , u i is a constant times of u because of the condition (2)
and Lemma 2.4. Hence, A(u v) can be written as u v by using v Cn . That
is, A satisfies (1).
Next, we show the general case. In this case, we have H = G U Cn , and
denote the projection to U Cn by P . The relation P A P f(g) = f(g)P A P
holds for = and g G. However, since U is not isomorphic to U , we have
P A P = 0. Hence, the matrix A can be written as G A by using the matrix A
on U Cn . Finally, applying the above discussion to each matrix A , we obtain
the desired argument.
2.3.3 Representations of Direct Product Group

Consider two given representations f1 and f2 of two groups G 1 and G 2 on H1 and H2 .
When the group G 1 does not cause any change on the space H2 and the group G 2 does
not cause any change on the space H1 , it is natural to consider the action on the tensor
product space H1 H2 (the composite system) as the following representation of
the direct product group G 1 G 2 . That is, we define the (unitary) representation
2 of the direct product group G 1 G 2 on the tensor product space H1 H2 as
f1 f
follows.
2 )(g1 , g2 ) := f1 (g1 ) f2 (g2 ), (g1 , g2 ) G 1 G 2 .
(f1 f
(2.31)
Then, we have the following formula for characters

f1 f
2 (g1 , g2 ) = f1 (g1 )f2 (g2 ).
(2.32)
2 is also reducible. That is, when f1 f

2 is irreducible,
When f1 is reducible, f1 f
f1 is irreducible. As the converse statement, we have the following lemma.
Lemma 2.8 When the representations f1 and f2 are irreducible, the (unitary) repre 2 is also irreducible. Conversely, any irreducible (unitary) representasentation f1 f
2 by using irreducible
tion of the direct product group G 1 G 2 can be written as f1 f
(unitary) representations f1 and f2 of the groups G 1 and G 2 . The same fact holds for
projective (unitary) representations.
Now, given a representation f of G on H, we consider the following representation
on the set gl(H) of matrices on H of G G. For any element (g1 , g2 ) G G, we
45
define a map D(f)(g

1 , g2 ) on X gl(H) as
T
D(f)(g
1 , g2 )(X ) := f(g1 )X f(g2 ) ,
(2.33)
which satisfies the condition for a representation of G G. Thanks to Lemma 2.8
and its above discussion, the representation D(f)

is irreducible if and only if f is
irreducible. In the irreducible case, we can symbolically express this relation as
follows.
gl(U )
= U U .
(2.34)
Proof Any irreducible subspace of the group subgroup {e} G 2 can be written
as {v1 v2 |v2 H2 } by choosing a suitable element v1 H1 . Hence, an irreducible subspace of H1 H2 includes the subset {v1 v2 |v2 H2 } unless the irreducible subspace is {0}. Similarly, an irreducible subspace of the subgroup G 1 {e}
is written as {v1 v2 |v1 H1 } by choosing a suitable element v2 H2 . Hence,
when an irreducible subspace of H1 H2 contains v1 v2 , it includes the subset
{v1 v2 |v1 H1 }. Since an irreducible subspace of H1 H2 includes the subset
{v1 v2 |v2 H2 } unless it is {0}, it is coincides with the whole space H1 H2 .
Let f be an irreducible (unitary) representation of the direct product group G 1 G 2
on the space H. We define the (unitary) representation f1 of the group G 1 on the space
H by g1 f(g1 , e). Let H1 be an irreducible subspace of f1 that is different from {0}.
We denote the representation on H1 by f1 . Then, the irreducibility guarantees that the
subset {f(e, g2 )H1 } generates the space H. Hence, we obtain the irreducible decomposition H
= U (G 1 ) Cn . Since f(e, g2 ) : H1 f(e, g2 )H1 is an isomorophic
map for (unitary) representation of G 1 , the map f(e, g2 ) maps v1 Cn to v1 Cn
for any element v1 U (G 1 ). Define the (unitary) subrepresentation f2 of G 2 on
2 . Since f2 is not an irreducible (unitary)
v1 Cn as g2 f(e, g2 ), we have f
= f f
representation of G 2 , f also is not an irreducible (unitary) representation of G 1 G 2 .
2 , we obtain the
By contradiction, we find that f2 is also irreducible. Since f is f1 f
desired argument.

Exercise 2.14 Show that any irreducible representation of G has a one-dimensional
representation space when G is a commutative group.
2.4 Projective Representation and Extension of Group

2.4.1 Factor System of Projective Representation
As explained in the previous section, any transformation on a quantum system can be
written by using a projective unitary representation. However, since the phase factor
appears in a projective representation, its treatment is more complicated than that of
46
Fig. 2.16 Relation between
central extension and
projective representation
projective (unitary)
representation
(unitary)
representation
(unitary) matrices
+
central extension
factor system
a representation. In fact, a projective representation of a group G can be reduced to

a representation of a central extension of the group G as Fig. 2.16. The purpose of
this section is to investigate this reduction.
A projective (unitary) representation f of group G can be reduced to a (unitary)
representation of a central extension of the commutative group U(1) by the group
G as follows. Remember that a central extension of G is a larger group than G.

As given in (2.21), we define the complex number ei(g,g ) := f(g)f(g )f(gg )1 =
1

i(g,g )
}g,g satisfies the condition (2.19) for a
f(gg ) f(g)f(g ). In fact, we find that {e
factor system because f(gg g )1 f(g)f(g g )(f(g g )1 f(g )f(g )) = f(gg g )1 f(gg )

(f(gg )1 f(g)f(g ))f(g ). Then, we call {ei(g,g ) }g,g the factor system of the projective (unitary) representation f. Let G be an extension of a commutative group U(1) by

G whose factor system is {ei(g,g ) }g,g . Based on a projective (unitary) representation f
of G, we can define the (unitary) representation of G by f(ei , g) := ei f(g). When the
factor system is {1}g,g , the extension G is the direct product group G U(1). Then,
the projective (unitary) representation f is a (unitary) representation. When f and f
are projective unitary representations and f (g) = f(g)ei(g) , both actions for a density
matrix satisfy that f(g)f(g) = f (g)f (g) . So, these actions are the same action.
Due to this reason, two projective (unitary) representations f and f are called equivalent to each other when there exist a matrix (unitary operator) A and a phase factor

ei(g) such that f (g) = ei(g) Af(g)A1 . Then, the factor system {ei (g,g ) }g,g of the

projective (unitary) representation f is given by ei (g,g ) = ei(g) ei(g) ei(gg ) ei(g,g ) .

That is, {ei (g,g ) }g,g is a factor system equivalent to {ei(g,g ) }g,g . Then, the (unitary)
representation f of G and the (unitary) representation f of the group G satisfy the relation f(g ei , g) := ei f (g). Hence, an element of the group of extension H (G : U(1))
(the set of extensions of U(1) by G) uniquely corresponds to a equivalent class of projective (unitary) representations. The element of the group of extension H (G : U(1))
is written as H (f). In particular, when a projective (unitary) representation f is equivalent to a (unitary) representation, i.e., the factor system H (f) is the identity element of
the group H (G : U(1)), the projective (unitary) representation f is called reducible
to a (unitary) representation. Conversely, a projective (unitary) representation f is
called a proper projective (unitary) representation when it is not reducible to a

(unitary) representation. When the set {ei(g,g ) }g,g is included in a subgroup C of the
i(g,g )
}g,g can be regarded as a factor system corresponding to
group U(1), the set {e

an extension of C by G. Hence, the set {ei(g,g ) }g,g can be regarded as an element of
the group of extension H (G : C). When the identity element e of G satisfies f(e) = I ,
47
the factor system H (f) given by the projective representation f is a normalized factor
system.
In the following, we study the properties of a projective (unitary) representation
f of a group G via the order of the element H (f) H (G : U(1)).
Lemma 2.9 Let f be a projective (unitary) representation of a group G on an kdimensional space H. Then, the order of H (f) H (G : U(1)) is a divisor of k.
Proof Firstly, we define another projective (unitary) representation f (g) := f(g)
(det f(g))1/k from the given projective (unitary) representation f. The projective
(unitary) representation f satisfies the condition det f (g) = 1 for any element g G.

Then, (2.21) implies that f (g2 )f (g1 )f (g2 g1 )1 = ei (g2 ,g1 ) . Since the determinant of
ik (g2 ,g1 )
= 1. Therefore, the order of H (f) H (G : U(1))
the LHS is 1, we have e
is a divisor of k.

Combining the above lemma and Theorem 2.1, we have the following corollary.
Corollary 2.1 Let f be a projective (unitary) representation of a finite group G on
H. When the order G is coprime to the dimension k of H, H (f) H (G : U(1)) is
the identity element, i.e., f is a (unitary) representation.
2.4.2 Irreducibility and Projective Representation

In the previous subsection, we have seen that any projective representation of a group
G can be described by a representation of a central extension G of a commutative
finite group C by the group G. In the following, we investigate how an irreducible
projective (unitary) representation of G can be characterized by an irreducible (uni
tary) representation f of a central extension G.
In this case, Schurs lemma guarantees that f(c) is a constant for any element
c C. Letting m be the order of C, we have f(c)m = 1. Then, for any element
g G = G/C,
we define
f(g) := f(g)
(2.35)
by choosing an element g G among the equivalent class g. Then, since f(c) is a

constant, f is a projective (unitary) representation of G. Given two elements g,
g G

Since the condition
among the equivalent class g, f(g ) is a constant times of f(g).
(2.28) for f is equivalent to the condition (2.28) for f, f is an irreducible projective
(unitary) representation. From an irreducible (unitary) representation of a central
extension G by G, we can construct an irreducible projective (unitary) representation
of group G. Hence, the factor system of a projective representation f is an element
of group of extension H (G : C). Conversely, when f is an irreducible projective
(unitary) representation and a (unitary) representation f of G satisfies (2.35), the
(unitary) representation f is irreducible. In general, if f is irreducible and f is defined
48
by (2.35), the factor system of f depends on the representation f. Using this fact, we
obtain the following corollary of Lemma 2.9.
Corollary 2.2 Let f be a projective (unitary) representation on H. Assume that the
condition (2.35) holds for an irreducible (unitary) representation f of an extension
G of a commutative group C by G. When the order of the commutative group C is
coprime to the dimension k of H, H (f) H (G : C) is the identity element, i.e., f is
an irreducible (unitary) representation.
Next, let us consider the case when a (unitary) representation f of a group G is not
necessarily irreducible. In this case, since Schurs lemma does not hold, f(c) is not
necessarily a constant. Hence, there does not necessarily exist a projective (unitary)
representation f satisfying (2.35). In general, the matrix entries of f(c) depend only
on the irreducible component.
i(g,g )
}g,g ]) be the set of labels of irreducible projective unitary repreLet G([{e

sentations whose factor system is equivalent to [{ei(g,g ) }g,g ] H (G : U(1)), and

G f ([{ei(g,g ) }g,g ]) be the set of labels of finite-dimensional projective unitary representations. When f is an irreducible unitary representation, since f(c) is a constant for
the center C(G)
does not change any density matrix. Hence, the group
c C(G),
Thereof transformations for density matrices is not G but rather G = G/C(

G).
the group
fore, even when f is not irreducible and f(c) is a constant for c C(G),
expresses the essentially physical transforms. Then, there exists a

G = G/C(
G)
projective unitary representation f of G satisfying (2.35) and f describes the action
for density matrices.
When we choose irreducible projective unitary representations f1 , . . . , fl whose

i(g,g )
factor systems are {ei(g,g ) }g,g for 1 , . . . , l G([{e

}g,g ]), their direct sum
representation f1 fl is a projective unitary representation. For example,
when C is the cyclic group U2 with order 2, H (G : U2 ) has only two elements. Now,
When
let G be central extension of U2 by G and f1 , . . . , fk be representations of G.
the projective representations f1 , . . . , fk of G are proper projective representations,
their factor systems are equivalent to each other. Hence, we can add phase factors
to the projective representations so that their suitable factor system is the same.
After this replacement, the direct sum representation f1 fk is a projective
representation.
2.4.3 Extension by U(1)

Next, we consider an extension of the group G by U(1), which is different from an
extension of U(1) by the group G because the roles of G and U(1) are opposite to
the previous discussion. Since this kind of extensions play a certain role in the latter
discussion, we discuss them here. We choose an element of C(G) and denote its
order by m. Using the element , we define an extension of G by U(1) as follows.
That is, for two elements (ei1 , g1 ), (ei2 , g2 ) U(1) G, their product is defined as

(ei1 , g1 ) (ei2 , g2 ) :=
(ei(1 +2 ) , g1 g2 )
(ei(1 +2 2) , g1 g2 )
49
if1 + 2 < 2/m

if1 + 2 2/m.
Then, the center of the defined group is C(G) U(1). The group defined above is
written as G . Let f be an irreducible representation of G on H. Schurs lemma
guarantees that there exists an integer l(f) such that f() = ei2l(f)/m I for a generator of C(G). Note that a generator of C(G) is an element of C(G), but it is
not necessarily unique. A representation of an extension of G of G is given as
f(ei1 , g) = f(g)ei1 (nm+l(f)) by using an integer n Z.
2.5 Semi Direct Product and Its Representation

2.5.1 From H K to K and H
Let us study the relation between a (projective) (unitary) representation of a semi
direct product H K and (projective) (unitary) representations of the groups K
and H . We discuss how to construct a (projective) (unitary) representation of the
groups K and H from that of the semi direct product H K . Consider a (unitary)
representation f on H of the semi direct product H K based on an action T of a
group K on a group H . Define (unitary) representations f H and f K of the subgroups
H and K by restricting the domain of the homomorphism f to the subgroups H and
K , respectively. Then, the relation (2.3) guarantees that
f(h, k) = f H (h)f K (k)
f H (T (k)(h)) = f K (k)f H (h)f K (k
(2.36)
1
).
(2.37)
The map f H,k : H h f H (khk 1 ) defined for any element k K is a (unitary)

representation of H .
In the same way, given a projective (unitary) representation f of the semi direct
product H K , we define the projective (unitary) representations f H and f K of the
k)
subgroups H and K . Then, due to (2.3), there exist real numbers (h, k) and (h,
such that
ei(h,k) f(h, k) = f H (h)f K (k)
e
i (h,k)
f H (T (k)(h)) = f K (k)f H (h)f K (k
(2.38)
1
).
(2.39)
Then, for any k K , the map f H,k : H h f H (khk 1 ) is a projective (unitary)

representation of H . In this way, we can construct a (projective) (unitary) representation of H and K from a (projective) (unitary) representation of H K .
50
2.5.2 From K and H to H K *

Next, we investigate the opposite direction, i.e., how to construct a (projective) (unitary) representation of H K from (projective) (unitary) representations of H and
K . In particular, we discuss the case when these (projective) (unitary) representations
are irreducible.
Construction 1: Let f H be a (unitary) representation of H on H. Assume that
a (unitary) representation f H,k : H h f H (khk 1 ) is isomorphic to f H for any
element k K . Then, there exists a (unitary) matrix f K (k) such that f H,k (h) =
f K (k)f H (h)f K (k)1 . The map k f K (k) is a projective (unitary) representation of
K , and is called the associated representation of K with respect to f H . Here, we
assume that the associated representation f K of K with respect to f H is a representation of K . Hence, defining f(h, k) := f H (h)f K (k), we find that f satisfies the condition
(2.37) and forms a (unitary) representation of H K . The representation is called
the semi direct product (unitary) representation of f H and f K , and is written as
f H f K . In particular, when f H is irreducible, f H f K is also irreducible.
Construction 2: Let f H be a (unitary) representation of H on H. Define the subset
T consisting of elements t K such that the (unitary) representation f H,t is isomorphic to f H . The subset T is a subgroup of K and is called the inertia group of the
representation f H . Here, we assume that the associated representation fT of T with
respect to f H is a representation of T . Now, we focus on the representation space
H[k] of the representation f H,k (h) dependently of the element [k] of the homogeneous
space K /T . The representation f H,k (h) does not depend on the choice of the representative k. Choosing the representatives k1 , . . . , kn , . . . of all of elements of K /T ,
we define a linear map U[kn ] from H to H[kn ] as f H,kn (h) = U[kn ] f H (h)U[k1n ] , where the
representative of the equivalent class containing the identity element e should be cho1
sen to be e. We define the unitary matrix f K (k ) := [k]K /T U[k k] fT (t (k , [k]))U[k]

on the representation space [k]K /T H[k] for an element k K , where t (k , [k]) is
given as k21 k k1 by using the above-defined representatives k1 and k2 of the equivalent classes [k] and [k k]. Then, f K forms a (unitary) representation of the group
K . Also, the map h f H (h) := [k]K /T f H,k (h) forms a (unitary) representation
of H . Hence, we can define the (unitary) representation f H f K of the semi direct
product H K . Since the the representation f H f K depends only on f H and fT , it
T . When f H is irreducible, f H f
T is also irreducible. When the
is denoted by f H f
subgroup T consists only of the identity element, the representation fT is limited to
the trivial representation, and is written as f H 1.

Construction 3: Assume that a (unitary) representation f H,k : H h
f H (khk 1 ) is isomorphic to f H for any element k K . Consider the case when the
associated representation f K of K with respect to a representation f H of H on H
defined in Construction 1 is not necessarily a representation. Let fK (k) be the projective (unitary) representation of K on H whose factor system is the complex conjugate
of the factor system of f K (k). Then, f K fK forms a (unitary) representation of K on
H H . Hence, we can define the (unitary) representation (f H I ) (f K fK ) of
H K . Since the (unitary) representation (f H I ) (f K fK ) depends only on f H ,
2.5 Semi Direct Product and Its Representation
51
f K , and fK , it is written as f H f K fK . In particular, when f H and fK are irreducible,
f H f K fK is irreducible, too.
Construction 4: We define the subgroup T of K as the same way as Construction 2. Consider the case when the associated representation fT of T with respect to a
representation f H of H on H is not necessarily a representation. Let fT be the projective (unitary) representation of T whose factor system is the complex conjugate of the
factor system the projective representation fT (t). Then, fT fT is a (unitary) repre T fT )
sentation of T . Hence, we can define the (unitary) representation (f H I )(f

of H K on [k]K /T H[k] H . Since it depends only on f H , fT , and fT , it is writ T fT . In particular, when f H and fT are irreducible, f H f
T fT is also
ten as f H f
irreducible.
We obtain the following generalization of Lemma 2.8.
Theorem 2.2 Any irreducible (unitary) representation f of H K is written as
T fT by using an irreducible representation f H of H , the associated represenf H f
tation fT of T with respect to f H , and an irreducible projective (unitary) representation
fT of T , where T is the inertia group of f H and the factor system of projective (unitary)
representation fT is the complex conjugate of the factor system of fT .
Proof We regarded the representation f as the representation of H , and focus on
its irreducible decomposition. We choose one irreducible representation f H among
irreducible representations appearing in the irreducible decomposition, and denote
its representation space by H. Let m be the multiplicity of f H in the irreducible
decomposition and T be the inertia group of the representation f H . Also, we denote
the associated representation of T with respect to f H by fT . Here, we consider the
tensor product space H Cm . Due to the definition of the inertia group, for |u |v
and t T , there exists a vector |u such that f(t)|u |v = |fT (t)u |u . Thus,
letting fT (t) be the matrix mapping |u to |u , we can show that fT is an irreducible
projective (unitary) representation of T .
In the following, we show this fact. We notice that fT fT a (unitary) representation of T . We choose an irreducible projective (unitary) subrepresentation fT of fT
with respect to T . Since fT fT is a (unitary) representation, the factor system of fT is
the complex conjugate of that of fT . Recalling the action on the homogeneous space
K /T given in Construction 2 (or Construction 4), we define the irreducible (unitary)
T fT of H K , which is a (unitary) subrepresentation of f with
representation f H f
T fT . Now, we
respect to T . Due to the irreducibility of f, f coincides with f H f

denote the representation space of fT by H , and we consider the irreducible decom T fT with respect to the representation
position of the representation space f H f
f H of the subgroup H . Due to the discussions in Construction 4, any representation
subspace isomorphic to f H is included in H H . Hence, we find that H coincides
T fT .

with Cm , which implies that fT equals fT . Thus, f equals f H f
52
2.6 Real Representation and Complex Conjugate

Representation
2.6.1 Real Linear Space and Its Complexification
Since quantum theory mainly discusses complex Hilbert spaces, we rarely treat real
vector spaces. However, to discuss several physical-meaningful representations, we
often need to handle real vector spaces. In this section, we treat representations on
real vector spaces and the relation between real vector spaces and complex vector
spaces. For this purpose, we need to define several kinds of complexification of a
real vector space.
For a real linear space V with inner product, we define the complex-multiplication
on the real vector space V V as follows.
(x + i y)(a, b) := (xa yb, ay + xb), (a, b) V V, x + i y C.
We let V C be the above-defined complex linear space with the following inner
product.
((a , b ), (a, b)) := (a a b b) + i(a b + b a), (a , b ), (a, b) V V.
be the above-defined complex linear space with the following inner
Also we let V C
product.
(a , b )|(a, b) := (a a + b b) + i(a b b a), (a , b ), (a, b) V V.
and
Then, we have z v |zv = z zv |v and (z v , zv) = z z(v , v) for v, v V C
z, z C.
2.6.2 Real Representation

Up to the previous section, we have discussed only representations whose representation space is a complex space. For a given real linear space V , we denote the set
of invertible linear maps on V by GL(V ). A homomorphism f from a group G to
GL(V ) is called a real representation of G on V . In this case, V is called the representation space of the real representation f . In general, a representation means
a representation whose representation space is a complex vector space. When we
need to distinguish it from a real representation, we call it a complex representation.
In the case of a real representation, similar to the case of complex representation, we can define invariance, irreducibility, reducibility, completely reducibility,
decomposability, irreducible decomposition, character, irreducible character, and
2.6 Real Representation and Complex Conjugate Representation
53
multiplicity. When a real representation f preserves the inner product, it is called

orthogonal representation. Similarly, we can define a projective representation.
However, Schurs lemma does not hold for a real representation. That is, there
exists an irreducible real representation f of a group G on a real linear space V
that satisfies the following condition. There exists a real matrix A on V that is not
a constant but satisfies the relation f(g)A = Af(g) for any element g G. For an
example, we define the representation
f ofthe group G = U(1) on the real linear

cos
sin
space V = R2 by f(ei ) :=
. Then, the matrix A := f(ei0 ) satisfies
sin cos
the above condition.
Hence, the irreducibility with respect to a real representation is not useful. Thus,
in this book, even though f is a real representation on a real linear space V , the irreducibility of f means the irreducibility of the complex representation on the complex
linear space V C.
2.6.3 Complex Conjugate Representation

Given a representation f of a group G on a finite-dimensional complex Hilbert space
H, the representation f defined by
f(g) := f(g)
(2.40)
is called the complex conjugate representation of f. (In general, we cannot define

the complex conjugate operator for an operator A on an infinite-dimensional complex Hilbert space H. Hence, we discuss complex conjugate representations only
of finite-dimensional representations.) The definition of a complex conjugate representation depends on the choice of the standard basis. However, even when we
define the complex conjugate representation based on another basis, it is isomorphic to the complex conjugate representation based on the original basis. When the
representation f is isomorphic to f, the representation f is called a self-conjugate
presentation.3 Then, the character is a real-valued function. Further, given a label
the complex conjugate representation of the irreducible representation f is
G,
That is, f is the
an irreducible representation, whose label is written as G.
complex conjugate representation of f .
Now, we modify the discussion in (2.33) by using complex conjugate representation. Given a projective representation f of G on H, we consider the following representation on the set gl(H) of matrices on H of G G. For any element
(g1 , g2 ) G G, we define a map D(f)(g1 , g2 ) on X gl(H) as
3 Such a representation is usually called a self-adjoint representation. However, since an adjoint

representation will be used in the different meaning in this book, we express such a representation
self-conjugate.
54
D(f)(g1 , g2 )(X ) := f(g1 )X f(g2 ) .
(2.41)
Then, D satisfies the condition for a representation of G G even when D is proper

projective representation. Thanks to Lemma 2.8 and its above discussion, the representation D(f) is irreducible if and only if f is irreducible. When f is a unitary projective representation, D(f) is also a unitary representation under the inner product
X, Y := Tr X Y . In the irreducible case, we can symbolically express this relation
as follows.
gl(U )
= U U .
(2.42)
A real linear map A from a complex Hilbert space H1 to a complex Hilbert space
H2 is called conjugate linear when the relation A(c|v) = c A(|v) holds for any
complex number c C and any vector |v H1 . Two (unitary) representations f1 and
f2 on H1 and H2 are called conjugate isomorphic when there exists a conjugate linear
map A such that Af1 (g) = f2 (g)A for any element g G. Then, the conjugate linear
map A is called a conjugate isomorophism for (unitary) representation of G. When
two (unitary) representations f1 and f2 of G are conjugate isomorphic, f2 is isomorphic
to the conjugate representation f1 of f1 . In fact, an irreducible real representation f on
C. The complex representation
V can be regarded as complex representation on V
is a self-adjoint representation because its conjugate representation has the same
form.
Conversely, given a complex representation f on the complex Hilbert space H, f
can be regarded as a real representation and is called a real representation when H is
regarded as a real vector space. When there exists a real vector subspace V H such
that H = VC and the real vector subspace V is invariant for the real representation
f, we can define a real representation fR on the real linear space V as a restriction of
the real representation f. In this case, the complex representation f is called underlying real form. However, a self-adjoint complex representation is not necessarily
underlying real form.
When f1 is a projective (unitary) representation and its factor system is given

as E := [{ei(g,g ) }g,g ] H (G : U (1)), its conjugate representation f1 has the factor

system E 1 = [{ei(g,g ) }g,g ] H (G : U (1)). We can define the concept of conjugate isomorophic for the projective case. When a projective (unitary) representation
has a conjugate isomorphic projective (unitary) representation, its factor system E
satisfies E 1 = E, i.e., it has order 2 as an element of the group H (G : U (1)).
On a 2d-dimensional
real representation space V , using the two-dimensional

0 1
matrix J =
, we define the matrix
1 0
Jd := Id J.
(2.43)
2.6 Real Representation and Complex Conjugate Representation
55
Then, considering the complex-valued vector (x1 + i xd+1 , . . . , xd + i x2d ) for x V ,

we can regard V as a d-dimensional complex linear space V. When a real representation f forms a complex representation on a complex linear space V, it is called
complexifiable. Here, we should remark that V is different from V C. A real
representation f is complexifiable if and only if
f(g)Jd = Jd f(g).
(2.44)
Exercise 2.15 Show that the character of a self-adjoint representation takes real
numbers.
Exercise 2.16 Show that (2.44) is holds if and only if a real representation f is
complexifiable.
2.7 Representation on Composite System

In Sect. 2.3, we have mentioned that, the treatment of a representation can be simplified by its irreducible decomposition because the treatment can be divided into
each irreducible representation. However, in the physical context, the representation
of our interest is usually given in a form with a tensor product representation.
As the simplest example, we consider the case when the first quantum system
is given as the representation space U (G) of the irreducible representation f of a
group G and the group G does not act on the second quantum system. In this case,
we are interested in the behavior of a vector |X U (G) Cd under the action G.
In the following, we focus on the case when the dimension of the second quantum
system is larger than the dimension d of U (G).
Now, we remember that a vector in the tensor product space U (G) Cd can be
written as |X by using a d d matrix X . We employ a polar decomposition of
a matrix X . That
is, we choose suitable CONSs {|ei } and {| f i } of U (G) and Cd
such that |X = dj=1

a j |e j | f j . Let K be the subspace of Cd generated by { f j }dj=1
.
Then, the vector f (g)|X is contained in U (G) K for any element g G. Hence,
to discuss the orbit of |X by the action of G, we do not need to discuss the whole
space U (G) Cd and only need to treat the subspace U (G) K. Since the subspace is isomorphic to U (G) Cd , it is sufficient to treat at most the dimension d .
In general, when a representation of a group G is completely reducible, as given
in (2.26), the representation space has the irreducible decomposition G U (G)
Cn . However, to address the orbit of a specific vector by the action of G, the
above discussion guarantees that it is enough to discuss the representation subspace
G U (G) Cd . As will be explained in Sect. 2.8, restricting our analysis in the
subspace enables us a deeper analysis.
Next, we consider the case when two quantum systems are given as two representation spaces U (G) and U (G) of two irreducible representations f and f
of a group G. Then, we have the tensor product representation f f on the ten-
56
sor product space U (G) U (G) corresponding to the composite system. When
these two representations f and f are finite-dimensional unitary representations,
the tensor product representation is completely reducible due to Lemma 2.3. When
the tensor product representation f f is completely reducible, we define the mul
tiplicity coefficient C,
(G) that describes the multiplicity of U (G) in the space
U (G) U (G) as follows
U (G) U (G)
=
U (G) CC, (G) .
(2.45)
G
When G is a commutative group, its irreducible representation is one-dimensional

Then, the map
(Exercise 2.14). That is, the relation f : G U(1) holds for G.
We denote the label of the repreg f (g)f (g) is a representation for , G.
The set G forms a commutative group with this product.
sentation by + G.
In particular, the multiplicity coefficient satisfies the following formula.

C,
(G)

=
1 when = +
0 otherwise.
(2.46)
As a special case, we consider the case when the second representation space
U (G) is U (G).
In this case, the relation f = f holds under a suitable choice of the standard
basis {| j A }, {| j B }. By using the standard basis, any vector on the composite system
can be written as |X with a square matrix X . Then, we have the following lemma.
Lemma 2.10 The following conditions are equivalent to each other for any vector
in the composite system |X .
(1) |X is a constant times of |I .
(2) The relation
f(g) f(g)|X = |X , g G
(2.47)
holds for any element g G.

Proof (1)(2): This part can be shown from the relation f(g) f(g)|I =
|f(g)f(g)T = |I .
(2)(1): Assume the condition (2). Since the relation X = f (g)X f (g)T =
f (g)X f (g)1 holds, Schurs lemma restricts the matrix X to a constant times of
the identity matrix. Hence, we obtain the condition (1).

We also have the following lemma.
Lemma 2.11 The following conditions are equivalent to each other for a representation f of a group G on a space H.
(1) The representation space H contains the irreducible representation space U
with multiplicity 1.
2.7 Representation on Composite System
57
(2) There uniquely exists a wave function, (a normalized vector) |X in the composite
system H U such that
f(g) f (g)|X = |X , g G.
(2.48)
Proof (2)(1): Assume the condition (2). The irreducibility of U guarantees that
Tr H |X X | is a constant times of the identity matrix. Hence, Tr U |X X | is a constant times of the projection PK to the range K of Tr U |X X |. Since Tr U |X X |
is invariant with respect to the representation f, the space K is also invariant with
respect to the representation f. Hence, we have PK X = f|K (g)PK X (f (g))T , which
implies that f|K (g) = PK X ((f (g))T )1 (PK X )1 = PK X f (g)(PK X )1 . This relation shows that the representation f|K (g) is isomorphic to the representation f (g).
Thus, H contains the irreducible representation space U at least with the multiplicity
1. If the multiplicity is strictly larger than 1, there exist more than two wave functions
satisfying the invariance (2.48), which contradicts the condition (2). Therefore, we
can conclude that H contains the irreducible representation space U only with the
multiplicity 1, which yields the condition (1).
(1)(2): Assume the condition (1). When X is a constant times of the identity
matrix, |X satisfies (2.48). So, it is sufficient to show the uniqueness of |X to
satisfy (2.48). Assume that |X satisfies (2.48). Since Tr U |X X | is invariant
with respect to the representation f, the space K is also invariant with respect to the
representation f. Due to the condition (1), the space K is U . That is, the support
of |X is limited to U U . In this case, the condition (1) is equivalent to the
condition f (g)X f (g) = X . Schurs lemma guarantees that X is a constant times
of the identity matrix. Hence, we obtain the condition (2).

Applying Lemma 2.11 to a wave function on the composite system consisting of
three quantum systems, we obtain the following lemma.
Lemma 2.12 The following four conditions are equivalent to each other for three
irreducible unitary representations 1 , 2 , 3 G of a group G.
(1) The tensor product space U1 U3 contains the space U2 as an irreducible
component (with multiplicity 1).
(2) There (uniquely) exists a wave function as |x invariant for the action of G in
the composite system U1 U3 U2 as
f1 (g) f3 (g) f2 (g)|x = |x, g G.
(2.49)

Proof Lemma 2.11 yields (1)(2), (3)(2), and (4)(2).
58
Thus, when U1 U3 contains the space U2 as an irreducible component (with

multiplicity 1), U1 U2 and U1 U2 contain the spaces U3 and U3 , respectively.
The following lemma discusses how these subspaces are contained.
Lemma 2.13 Assume that U1 U3 contains the space U2 as an irreducible component with multiplicity 1. We have the following items.
(1) The range of the following map from U3 to U1 U2 is the subspace U3 of
U1 U2 , and the map gives an isomorophism for representation of G.
|u (I1 Q 2 )|u |I1 , |u U3 ,
(2.50)
where Q 2 is the projection to U2 .

(2) The range of the map |u( U3 ) u|I1 from U3 to U1 U2 is the subspace U3 of U1 U2 , and the map gives an isomorophism for representation
of G.
Proof (1): When |u = 0, we have Tr U |uu| |I1 I1 | = |uu| I1 , which
1
implies that (I1 Q 2 )|u |I1 = 0. Hence, the linear map given in (2.50) is a
homomorphism for representation of G. The irreducibility guarantees that it is an
isomorophism. So, we obtain Item (1).
Item (2) can be shown by the same way.

Here, we should be careful when f and f are projective (unitary) representations.
When the factor systems of f and f are E and E H (G : U (1)), respectively, the
factor system of the projective (unitary) representation f f is E E H (G :
U (1)).
2.8 Fourier Transform for Finite Group

2.8.1 Discrete Fourier Transform
Firstly, we focus on the group Zd as a typical example of commutative group. Let H
be the d-dimensional Hilbert space
spanned by the standard basis {|0, . . . , |d 1}.
The unitary matrix DFT := 1q l, jFq ei2l j/d |l j| on H is called the discrete
Fourier transform and plays an important role in quantum information processing.
Regarding the basis | j as elements of cyclic group Zd , we define the representation f of the cyclic group Zd by f(k)| j := | j + k, ( j, k Zd ). Then, we
have the irreducible decomposition of H as H = H0 Hd1 , where Hl is the
one-dimensional subspace generated by the dual base
1
2ilk/d
e
|k = DFT |l,
|el :=
d kZd
(2.51)
59
and the action of fl on Hl is given by fl (k)|el := e2ilk/d |el . Since Zd is a commutative group, its irreducible representation is limited a one-dimensional representation,
which is one of f0 , . . ., and fd1 . Hence, the set Zd of labels corresponding
d1 to the
irreducible representations of Zd is {0, . . . , d 1}. Since DFT|v = l=0
el |v|l,
the discreteFourier transform can be regarded as a conversion operation from a vec
tor |v := d1
j=0 v j | j in H to a function l el |v on Ud , where the vector |v is
regarded a function from Zd to C. In the following, we investigate how to extend
this characterization to the case of non-commutative groups. In fact, a finite group
plays an important role for the description of the asymmetry in a quantum system
as well as a continuous group (Lie group), which is discussed in the next chapter.
This section addresses a generalization of discrete Fourier transform for finite groups
from a viewpoint of representation theory.
2.8.2 Character and Orthogonality

In this subsection, we discuss a finite group G and its representation, which has
several good properties. For any representation f of G on H, there exists an inner
product of H such that f(g) is a unitary matrix [98, Remark in Sect. 1.3]. Such an
inner product is given by
1
f(g)v|f(g)v
|G|
gG
(2.52)

1
1

because the relation
gG |G| f(g)v|f(g)v =
gG |G| f(g)f(g )v|f(g)f(g )v
holds for any element g G and two vectors v, v . Hence, using this inner product,
without loss of generality, we can assume that a representation f of a finite group G
is unitary. This fact holds even for a projective representation because the relation

f(g)f(g )v|f(g)f(g )v = ei(g,g ) f(gg )v|ei(g,g ) f(gg )v = f(gg )v|f(gg )v holds

for g, g G and two vectors v, v . Hence, Lemma 2.3 guarantees that any representation f of a finite group G on the Hilbert space H is completely reducible. Therefore,
due to the discussions in Sect. 2.7, it is enough to consider the case when the representation space is the direct sum product of U Cd . This subsection and the next
subsection address various methods to describe the representation space U Cd ,
which bring mathematically easier analyses.
The character of the representation on

a representation space U Cm is m , where m = 0 means that H contains no irreducible representation isomorphic to f . We employ this relation in the
following discussion.
To study the structure of representations of a finite group G, we consider the
complex linear space C[G] whose CONS is composed of elements g of the finite
group G. When we define the following product in the complex linear space C[G],
it is called the group algebra of G.
60
ag g,

bg g
C[G] C[G]
ag bg gg C[G].
(2.53)
g
This product satisfies the associative law and the bilinearity, which means the linearity for both inputs. The identity element e of G works the identify element with
respect to the product of the group algebra C[G]. Such a linear space with an associative product and the bilinearity is called a linear algebra. For example, the set gl(U )
of linear maps on U forms a linear algebra when the product is defined by the composition of maps. As defined in (2.41), the space gl(U
) also has the
representation
:=
D(f ) of G G. Hence, their direct sum L 2 (G)

G gl(U ) =
G U U
2
also has the representation of G G as well as that of G. Also, the space L (G)
forms a linear algebra. Here, a linear map from a linear algebra V to another linear
algebra W is called a homomorphism for linear algebra when it preserves the struc is isomorphic to U Cd with respect to
ture of the product. The space L 2 (G)
is isomorphic
the application of G from the left hand side. That is, the space L 2 (G)
d
to representation space U C .
as follows.
Now, we define the homomorphism f from C[G] to L 2 (G)
f

ag g
ag f (g).
(2.54)
Since the homomorphism is injective as shown later, the dimensions of C[G] is less
which yields
than that of L 2 (G),
|G|
d2 .
(2.55)
=1
Because the opposite inequality can be shown from Lemma 2.14, we obtain
|G| =
d2 ,
(2.56)
=1
Hence, the character coincides with

which
implies that C[G] is isomorphic to L 2 (G).

d
.
When
we
focus
on
the
representation
corresponding to the action of G

from the left hand side on C[G], the character of the representation takes the value 0
for a non-identity element g because {g }g G forms a basis of C[G] and the relation
gg = g holds for non-identity element g. Hence, we obtain
d (g) = 0.
(2.57)
61

Proof of the injectivity of f. Assume
that f( g ag g) = 0. The left-hand-side appli
cation of anyelement of G to g ag g is the application of the zero matrix. Hence,
the relation g ag f(g) = 0 holds for any unitary representation f. Since C[G] can
be regarded as a representation
space with respect to the left-hand-side application,

a
g
to
C[G]
is written by the zero matrix. That is, the relathe application
of
g g

tion g ag g = ( g ag g)e = 0 holds for the identity element e of C[G]. Hence, we
obtain the injectivity of f.

Considering the correspondence between an element g ag g of C[G] and the
function g ag from G to C, we find that C[G] is isomorphic to the set L 2 (G) of
functions from G to C.

1
h 1 (g)h 2 (g),
We define the inner product in L 2 (G) as h 1 , h 2 := gG |G|
2
(h 1 , h 2 L (G)). Choosing a CONS |, j of U , we define R, j , j L 2 (G) as
R, j , j (g) := , j |f(g)|, j. Given a unitary representation f of G and a matrix A
on its representation space, the matrix
f(g)1 Af(g)
(2.58)
gG
is commutative with f(g ) for any g G. Using this fact, we can show the Plancherel
formula as follows.
Lemma 2.14 The following orthogonality holds.
1
1
R, j, j (g)R,
, j, j j , j .
j,
j (g) =
|G|
d
gG
(2.59)
Proof We choose the inner product of the space U such that the space U is
Then, we denote the unitary
orthogonal to the space U for different labels and .
representation of G on the space U by f. Hence, we have
1
j|f
j
(g)|;
; j |f (g)1 |, j;
R, j, j (g)R,
j,
j (g) =
|G|
|G|
gG
gG
1
f(g)1 |; j;
j .
j|
f(g) |;
= ; j |
|G|
gG

1
1
f(g) is commutative with f(g), Schurs
j|
Since the matrix

|,
j
;
f(g)
gG |G|
lemma guarantees that the matrix is written as c I , where I is the identity on
the LHS of (2.59) is zero because the vector |; j is
the space U . When = ,
orthogonal to the vector |; j.
62
we have
When = ,
1

f(g)1 |; j;
j|
f(g) = ; j|; j I ,
|G|
d
gG
because the left hand side is a constant times of the identity matrix and the above
observation yields that the traces of both sides of this equation are the same. Hence,
we obtain (2.59).

Since R, j, j L 2 (G), this lemma shows that dim L 2 (G) d2 . Because
2
L (G) is isomorphic to C[G], we obtain the inequality opposite to (2.55). Hence,
R, j, j forms a CONS of L 2 (G).
By considering the trace of the matrix (R, j, j ) j, j , the orthogonality discussed in
Lemma 2.14 yields the following orthogonality for characters.
1
(g) (g 1 ) =
(g) (g)
|G|
|G|
gG
gG
1
1 if =
(g) (g) =
=
0 if = .
|G|
(2.60)
gG
Similarly, we can characterize the projection to the irreducible subspace U as follows.
d
f (g) (g 1 ) =
f (g) (g)
|G|
|G|
gG
gG
d
I if =
f (g) (g) =
=
0 if = .
|G|
(2.61)
gG
When the representation space H of the unitary representation f is given by

G U Cm , the orthogonality (2.60) yields
|f(g)|2
gG
|G|
m.
(2.62)
Hence, we obtain the following lemma.

Lemma 2.15 The following conditions are equivalent to each other for a representation f
(1) f is an irreducible representation.

|f(g)|2
(2)
gG |G| = 1.
63
The orthogonality (2.60) derives the coefficient C,

given in (2.45) as follows.
C,
=
1
(g) (g) (g) =
(g) (g) (g).
|G|
|G|
gG
gG
(2.63)
Since (2.63) implies that C,

and C , equal the RHS of (2.63), we have
C,
= C , = C , .
(2.64)
The relations (2.56) and (2.57) yield another orthogonality:
d
Tr(f (g)) f (g ) =
Tr f (g 1 g )
|G|
|G|
d
1
if
g
= g
(g 1 g ) =
=
0 if g = g .
|G|
(2.65)
Let L 2 (G, inv) be the set of functions f in L 2 (G) whose value f (g) depends
only on the conjugate class of g. That is, L 2 (G, inv) is the set of functions in L 2 (G)
satisfying (2.27). Especially, let L 2 (G, inv, R) be the set of functions satisfying
(2.27). Then, we have the following lemma.
Lemma 2.16 The set { } of irreducible characters forms a CONS of L 2 (G, inv).
Especially, the set { + } forms a CONS of L 2 (G, inv, R).
2.8.3 Fourier Transform

k
Next, we define the inner product A, B L 2 (G)

:=
=1 Tr A B for any two ele
=
ments A = (A1 , . . . , Ak ) and B = (B1 , . . . , Bk ) of L 2 (G)
G gl(U ). Then,
1 , . . . , k ) L 2 (G)
for a function
we define the Fourier transform
F[]
=
(

1
f (g) (g). The relation (2.65) yields the
L 2 (G) by (F[]) := d gG |G|
2
relation L 2 (G) = F L 2 (G)
for L (G). By defining the Inverse Fourier

as F 1 [A](g) := d Tr f (g)A , the relatransform F 1 [A] for A L 2 (G)
tion (2.65) implies that
F 1 F[] = .
(2.66)
That is, the relation between F and F 1 is summarized as Fig. 2.17.

Hence, the behavior of an vector of any representation space of G can be characterized as an element of L 2 (G) via the Fourier transform. This property plays an
important role in the estimation of the unknown group action.
64
F 1

F and F 1
F
The above discussion can be extended to the case of projective representations.

Using the factor system E := [{ei(g,g ) }g,g ] H (G : U (1)), we modify the product
of the group algebra defined in (2.53) as follows.

ag g,

bg g
C[G] C[G]
ag bg ei(g,g ) gg C[G]. (2.67)
g
Using the set G[E]

of labels of irreducible projective representation of G, we define
the linear space L 2 (G[E])

:= G[E]
gl(U ). Similar to (2.54), we define the homo
2
morphism f from C[G] to L (G[E]) and can show the inequality (2.55). Since we can
show the commutativity between f(g) and the matrix defined in (2.58), we can show
Plancherel formula, i.e., Lemma 2.14. Hence, (2.56) holds. Similarly, Plancherel
formulas yields the relations (2.60)(2.62), (2.65), and Lemma 2.15. Therefore, we
can define the Fourier transform and the Inverse Fourier transform. Since the relation
(2.65) holds, we have (2.66). That is, the contents in this section can be extended to
the case of projective unitary representations with the above modification.
2.9 Representation of Permutation Group

and Young Diagram
2.9.1 Young Diagram and Young Tableau
In this section, we focus on the set of permutation among the integers 1, . . . , n. This
set forms a group when the product is defined by the composition of two permutations.
So, the set is called the permutation group of degree n and is written as Sn . The
order of Sn is n!. Since the permutation group describes operations for n particles
with the same kind, it is a very natural object in quantum theory.
To discuss representations of the permutation group Sn of degree n, we introduce
Young diagram and Young tableau as follows. The Young diagram with depth r is
defined to be a collection of r rows like
. (This picture describes the case
with r = 3.) The i-th row has n i boxes. A Young diagram is described by a vector
n = (n 1 , . . . , n r ) that consists of a monotone decreasing sequence of r non-negative
2.9 Representation of Permutation Group and Young Diagram
65

integers. The integer |n| := rj=1 n j is called its size. We let Y r be the set of Young
diagram whose depth is not greater than r and Yn be the set of Young diagram whose
size is not greater than n. Then, we define the set Ynr := Y r Yn . When |n| boxes
of a Young diagram n have |n| different integers from 1 to |n| like 2 3 6 4 , it is
7 1 8
5 9
called a Young tableau and is written as T . In particular, it is called a Standard
Young tableau when the integer is monotone decreasing from left to right and from
up to down like 9 8 6 1 . The set of Standard Young tableau corresponding to n is
7 4 3
5 2
denoted by Y S (n).
On the other hand, when each box of a Young diagram n has an integer such
that the number is monotone decreasing from left to right and it is strictly monotone
decreasing from up to down like 3 3 3 3 and 3 3 2 2 , it is called a Semistan2 2 2
2 2 1
1 1
1 1
dard Young tableau. Here, the same number can be used in these boxes. Usually, a
Semistandard Young tableau is not a Young tableau. Let YU,r (n) be the set of Semistandard Young tableaus given by a Young diagram n whose numbers are restricted
to the integers 1, . . . , r .
It is relatively easy to evaluate the cardinality |Ynr | of Young diagrams with depth r
the number
and size n. When we remove the condition that n i is monotone

1decreasing,
,
which
is
not greater
of sequences (n 1 , . . . , n r ) of non-negative integers is n+r
r 1
than (n + 1)r 1 . Since we impose the condition that n i is monotone decreasing, the
relation

n +r 1
(2.68)
(n + 1)r 1
|Ynr |
r 1
holds. Now, we consider a correspondence between a sequence (n 1 , . . . , n r ) of nonnegative integers and a sequence (n 1 , . . . , n r ) of non-negative integers with the
monotonically decreasing condition. Applying a suitable permutation, the former
can be converted to the latter. One of the latter sequence corresponds to r ! former
sequences at most. Hence, we have
|Ynr |

n +r 1
/r !.
r 1
(2.69)
Let us consider the cardinality |Yn | of Young diagrams with size n by removing
the condition for the depth. The cardinality |Yn | is called the Partition function, and
has been studied for a long time. When the size n is large, we have the following
asymptotic formula (HardyRamanujanRademacher formula) [43, 86], [19, p. 169],
[85, p. 925].
66
|Yn |
=

2n
1
.
exp
3
4n 3
(2.70)
2.9.2 Permutation Group and Young Diagram

A permutation Sn is called a horizontal permutation of a Young tableau T
when it permutates the numbers only in the horizontal direction in the given Young
tableau T . Similarly, a permutation Sn is called a vertical permutation of the
Young tableau T when it permutates the numbers only in the vertical direction in the
given Young tableau T . The set of horizontal permutations and the set of vertical permutations are subgroups of Sn , which are called the horizontal permutation group
and the vertical permutation group and are written as H (T ) and V (T ), respectively.
Given a
Young tableau T with size n, we define the Young horizontal
symmetrizer

cTH := H (T ) and the Young vertical symmetrizer cTV := V (T ) sgn() as
an element of C[Sn ], where sgn() is defined as follows. Also, define the Young
symmetrizer cT := cTH cTV . A permutation is called a transposition when it permuates only two integers. A transposition exchanging j and j is written as ( j, j ). Any
permutation Sn can be written as products of transpositions. We set sgn() = 1
when the number of the above transpositions is an even number, and set sgn() = 1
when the number is an odd number. We can show that this definition does not depend
on the choice of the product form of transpositions. It is also known that the subspace
C[Sn ]cT := {xcT |x C[Sn ]} is an irreducible representation space for the action of
the group Sn from the left hand side. The representation of Sn on the space depends
only on the Young diagram n = (n 1 , . . . , n r ) and does not depend on the choice
of the Young tableau T . Hence, we let Un (Sn ) be the representation space that is
isometric to this representation.
Now, we define a vector := (r 1, r 2, , 0). We also define the action
of the element Sr (not Sn ) for an r -dimensional vector v := (v1 , , vr )
as (v) := (v(1) , , v(r ) ). Additionally, we define n! := n 1 ! n r ! for an r dimensional vector n that consists of non-negative integers. The dimension of Un (Sn )
equals the number of Standard Young tableaus corresponding to the Young diagram
n and is known to be given as [37, Corollary 9.1.5]
d Sn ,n := dim Un (Sn ) =
Sr
sgn()
n!
[n + ()]!

n!
(n i n j + j i),
(n + )! i, j{1,...,r }:i< j
(2.71)
(2.72)
where the sum in (2.71) means the sum of Sr satisfying that n + () 0.

Let nT be the Young diagram that is the transpose of the Young diagram n by
exchanging the horizontal and vertical directions. Then, when the Standard Young
2.9 Representation of Permutation Group and Young Diagram
67
tableau corresponds to n, the transposed Standard Young tableau does to nT . So, we

have
dim Un (Sn ) = dim UnT (Sn ).
(2.73)
Exercise 2.17 Assume that r = 2 and that k is an integer between n and n/2 so that
n k is an even number. Show the following relation using (2.72);
d
nk
Sn ,( n+k
2 , 2 )
k+1
=
n+1

n+1
.
n+k
+1
2
(2.74)
2.9.3 Plancherel Measure

Now, using the formula (2.56), we define the Plancherel measure n (r ) :=

d S2n ,n
nYn :n 1 =r n! , where we used the fact that the order of the permutation group
d2 T

d2
Sn is n!. The relation (2.73) yields that n (r ) = nYn :n 1 =r Snn!,n = nYnr Sn!n ,n .
That is, n (r ) expresses the probability that the depth of the Young diagram is not
d2
greater than r when the the Young diagram n is generated with the probability Sn!n ,n .
Let R be the random variable subject to the Plancherel measure n . Then, it is
known that the random variable Rn converges to 2 in probability under the limit
n . More precisely, the random variable R 2 n behaves with the order n 1/6
2
as mentioned below. Letting u be the solution of the Painlev II equation: dd xu2 (x) =
2u(x)3 + xu(x), we define the TracyWidom distribution function [102]: FT W (t) :=
exp( t (x t)u(x)2 d x). Then, the asymptotic behavior of Plancherel measure n

can be characterized as [12]

r 2 n
lim n r
t
= FT W (t).
n
n 1/6
(2.75)
Chapter 3
Foundation of Representation Theory

of Lie Group and Lie Algebra
Abstract This chapter deals with general theories that do not depend on the types
of Lie groups and Lie algebras. As generalizations, it addresses projective representations of Lie groups and Lie algebras by combining the contents of Chap. 2. Then,
it introduces the Fourier transform for Lie groups including the case of projective
representations. It also prepares several concepts for Chap. 6. Also, this chapter introduces complex Lie groups and complex Lie algebras, which are helpful for real Lie
groups and real Lie algebras.
3.1 Lie Group

3.1.1 Basic Examples
As an example of a parametrized group, we consider the set of rotations in the
2-dimensional space:

SO(2, R) :=

cos sin
0 < 2 .
sin cos
(3.1)
The group can be written as the set of the 2-dimensional orthogonal matrices whose
determinant is 1. An element of the group can be characterized by the continuousvalued parameter . A group is called a Lie group when an element of the group
is characterized by continuous-valued parameters = (1 , . . . , l ). The group of
orthogonal matrices on d-dimensional real linear space is called the orthogonal group
and is written as O(d, R). Any matrix g in the group O(d, R) has determinant 1 or
1. Especially, the set of elements in O(d, R) with determinant 1 is a subgroup. It
is written as SO(d, R) and is called the special linear group. As an example of an
important transformation in quantum system, we can list the set of rotations on the
3-dimensional space, which is written as SO(3, R).
Since the determinant is a continuous function of a matrix, the subset of matrices
with determinant 1 is completely separated from the subset of matrices with determinant 1 in O(d, R). Conversely, an element of SO(d, R) has determinant 1, and
DOI 10.1007/978-3-319-44906-7_3
69
70
3 Foundation of Representation Theory of Lie Group and Lie Algebra
all of its elements are continuously connected to each other. Such a Lie group is
called a connected Lie group. In general, a Lie group is not necessarily connected
and consists of several connected components. For example, the group O(d, R) has
the connected component of matrices with determinant 1 and the connected component of matrices with determinant 1. Especially, for a Lie group G, we denote the
connected component containing the identity element by G 0 . For example, we have
O(d, R)0 = SO(d, R) (see Exercises 3.3). Then, the following lemma is known.
Lemma 3.1 The connected component G 0 of a Lie group G is a Lie group. Hence,
we have gg G 0 for any two elements g, g G 0 .
A coordinate transformation in the d-dimensional real linear space is given as
an invertible real square matrix of size d. The set of such matrices forms a group,
so called the general linear group GL(d, R). The set of invertible linear maps on a
complex Hilbert space H is written as GL(H). The set of elements of determinant 1
in GL(d, R) forms a group so called the special linear group SL(d, R). The groups
O(d, R) and SO(d, R) are bounded closed sets, i.e., they are bounded closed subsets
of set of all d d matrices, which is a d 2 -dimensional Euclidean space. Such a group
is called a compact Lie group. Since GL(d, R) or SL(d, R) is not bounded, either is
not compact. They are called non-compact groups. In fact, any compact Lie group is
isomorphic to a subgroup of GL(H) [66, Corollary 4.22]. A larger part of properties
of a Lie group is determined by the local structure around the identity element even
though it is not compact. It is also known as Ados Theorem that the local structure
of any Lie group is isomorphic to a subgroup of GL(H) [104, Theorem 3.17.8]. This
book discusses any Lie group as a subgroup of GL(H).
Physical transformation cannot be limited to linear transformations on the space
time, e.g., a rotation. For example, a displacement of the particle on the 3-dimensional
space is a physical transformation. The group of these displacements is written as
the 3-dimensional real linear space R3 . A pair of a rotation in the 3-dimensional
system and a displacement in the 3-dimensional system is also a transformation in
the 3-dimensional space. More generally, we focus on a pair of an element of the ddimensional special linear group SO(d, R) and an element of Rd . In this combination,
the transformations by SO(d, R) and Rd are not independent of each other, but effect
each other. When a pair of a displacement a Rd and a rotation g SO(d, R) is
applied to a particle with position x Rd , the position is transformed as follows.
x (g , a )x := g (x) + a .
(3.2)
When we apply another pair of another displacement a Rd and another rotation

g SO(d, R) after the above transformation, the position is transformed as follows.
g x + a (g, a)(g x + a ) = g(g (x) + a ) + a = gg (x) + ga + a.
(3.3)
When the product of two elements (g, a) and (g , a ) as (g, a)(g , a ) := (gg , ga +
a), the direct product set Rd SO(d, R) forms a semi direct product Rd SO(d, R)
and is called the Euclidean group. The transformation (3.2) is a natural transformation
3.1 Lie Group
71
by the semi direct product Rd SO(d, R). Similarly, the semi direct product Rd
GL(d, R) is called the affine group and is written as Aff(n, R).
When our transformation rotates the time and the position in the 3-dimensional
space, the transformation can be regarded as a rotation in the 4-dimensional space
time. The special relativity provides such a rotation as a Lorentz transformation as
follows. The set of Lorentz transformations is given as the following Lie group and
is called the Lorentz group.

O(1, 3) := g GL(4, R)|g T I1,3 g = I1,3 ,
1 0 0 0
0 1 0 0
I1,3 :=
0 0 1 0 ,
0 0 0 1
(3.4)
(3.5)
where Ip,q is the diagonal matrix of size p + q satisfying that the initial p diagonal
entries are 1 and the remaining q diagonal entries are 1. The Lorentz group is
divided into four connected components, and each component contains one of the
following matrices.
1
0
0
0
0
1
0
0
0
0
1
0
0
0
,
0
1
1
0
0
0
0
1
0
0
0
0
1
0
0
0
,
0
1
1
0
0
0
0
1
0
0
0
0
1
0
0
0
,
0
1
1
0
0
0
0
1
0
0
0
0
1
0
0
0
.
0
1
(3.6)
The most important connected component is the component containing the identity
element, which is written as O(1, 3)0 = SO(1, 3)0 . Since it is complicated to handle
the rotation between the time and the position in the 3-dimensional space, we firstly
discuss the spacetime with the 2-dimensional space and the one-dimensional time.
In this case, the Lorentz group is written as

O(1, 2) := g GL(3, R)|g T I1,2 g = I1,2 .
(3.7)
Similarly, the most important connected component is the component containing the
identity element, i.e., O(1, 2)0 = SO(1, 2)0 .
Although SO(3) and SO(1, 2)0 are physically important, it is complicated to
handle them because they are written as square matrices of size 3. If we treat the
same transformation with square matrices of size 2, our calculation becomes simpler. For this purpose, we denote the set of invertible complex square matrices of
size d by GL(d, C) and the subset consists of complex square matrices with determinant 1 by SL(d, C). We also denote the set of complex unitary matrices of size
d by U(d) and define SU(d) := U(d) SL(d, C). We also define U(p, q) := {g
GL(p + q, C)|gIp,q g = Ip,q }, and SU(p, q) := U(p, q) SL(d, C). In fact, SO(3)
and SO(1, 2)0 are deeply related to SU(2) and SU(1, 1), respectively, e.g., the isomorphic relations
72
SO(3)
= SU(2)/Z2 , SO(1, 2)0
= SU(1, 1)/Z2
(3.8)
hold, as shown in Sect. 4.1.2 and Exercise 4.1. Since the group Z2 is contained in
SU(2) and SU(1, 1) as the normal subgroup {I}, SU(2) and SU(1, 1) are central
extensions of SO(3) and SO(1, 2)0 , respectively. Hence, the discussion in the initial
part of Sect. 2.4.2 gives a one-to-one correspondence between projective representations of SO(3) (SO(1, 2)0 ) and SU(2) (SU(1, 1)), we can discuss SU(2) and SU(1, 1)
instead of SO(3) and SO(1, 2)0 , respectively. Similarly, as shown in the beginning
of Sect. 5.2.2, the following relation holds between SL(2, C) and SO(1, 3)0 .
SO(1, 3)0
= SL(2, C)/Z2 .
(3.9)
Exercise 3.1 Show that SO(2, R) is connected.

Exercise 3.2 Show that given an element g SO(d, R) there exists another element
g SO(d, R) such that g gg 1 is block-diagonal, with each block either a 2 2
rotation matrix or 1.
Exercise 3.3 Show that SO(d, R) is connected.
3.1.2 Symmetry in Analytical Mechanics

Next, we discuss the symmetry in analytical mechanics. When we consider a dynamical system of r real parameters, analytical mechanics discusses these r real parameters and their conjugate parameters. That is, we need to discuss 2r parameters as
(q, p) := (q1 , . . . , qr , p1 , . . . , pr ). In such a system, only the following elements g
in GL(2r, R) are allowed as possible transformations, so called symplectic transformations.
gJr g T = Jr ,
(3.10)
where Jn is the matrix defined in (2.43) in the previous chapter. The condition can
be rewritten as follows by using the symplectic inner product
(q, p), (q , p )
:= q p q p .
(q, p), (q , p )
= g(q, p), g(q , p )
.
(3.11)
The set of elements g in GL(2r, R) satisfying the condition is called the symplectic
group and is written as Sp(2r, R). This condition is illustrated in Fig. 3.1.
Let us consider the reason of the condition from the viewpoint of quantum theory.
For simplicity, we consider the case of r = 1. The group composed of displacements
of the position in the 2-dimensional space R2 can be regarded as itself R2 . To discuss
the transformation in quantum system by the group R2 , we need to a projective unitary
3.1 Lie Group
73
Fig. 3.1 Condition for

symplectic group
( p, q ),( p ', q ') = g ( p, q ), g( p ', q ')
g ( p, q )
( p, q )
g( p ', q ')
( p ', q ')
representation of R2 and we cannot restrict it to unitary representations of R2 . To

address its projective unitary representation, we need to consider central extensions
of U(1) by R2 . As shown in Sect. 3.3.3, for any central extension of U(1) by R2 ,
there exists a real number such that
(q, p, ei ) (q , p , ei ) = (q + q , p + p , ei(+ + (q,p),(q ,p )

/2) )
(3.12)
for (q, p, ei ), (q , p , ei ) R2 U(1). That is, the factor system ei (q,p),(q ,p )
/2 of
any projective representation of R2 can be characterized by the above central extension of U(1) by R2 . To avoid the variety of central extensions dependently of the
parameter , we treat the central extension of R by R2 . That is, we consider the
central extension of R by R2 as the direct product set R2 R with the product
(q, p, ) (q , p , ) := (q + q , p + p , + + (q, p), (q , p )
/2)
(3.13)
for (q, p, ), (q , p , ) R2 R. The group is called the Heisenberg group of degree
1 and is written as H(2, R). Any projective representation of R2 is given from a representation of the Heisenberg group H(2, R) of degree 1. For example, the projective

representation f with the factor system ei (q,p),(q ,p )
/2 is given by the representation
i
(q, p, ) f(q, p)e of the Heisenberg group H(2, R).
Now, we define the action of GL(2, R) on the Heisenberg group H(2, R)
as g(q, p, ) := (g(q, p), ) for g GL(2, R). When an element g GL(2, R)
satisfies
g(q, p, ) g(q , p , ) = g((q, p, ) (q , p , ))
(3.14)
if and only if the element g belongs to Sp(2, R). That is, to reserve the composition
under the projective representation f of R2 , i.e., to satisfy f(g(q, p))f(g(q , p )) =
f(g(q + q , p + p )), we need to restrict the linear transformation on R2 to elements
of Sp(2, R). For any extension of U(1) by R2r , there exists a square matrix of size
2r such that

/2) )
(3.15)
for (q, p, ei ), (q , p , ei ) R2r U(1). When the matrix is invertible, by applying
a suitable coordinate transformation, the product is simplified as
74


/2) ).
(3.16)
Similarly, for simplicity, we consider central extension of R by R2r , which is the

direct product set R2r R with the product
(q, p, ) (q , p , ) := (q + q , p + p , + + (q, p), (q , p )
/2)
(3.17)
for (q, p, ), (q , p , ) R2 R. The group is called the Heisenberg group of degree
r and is written as H(2r, R). Hence, to reserve the composition under the projective
representation f of R2r , i.e., to satisfy f(g(q, p))f(g(q , p )) = f(g(q + q , p + p )),
we need to restrict the linear transformation on R2r to elements of Sp(2r, R). In
this way, the property required by a projective representation restricts our linear
transformations to symplectic transformations. Thus, to consider groups of the above
linear transformation on the space R2r , it is sufficient to treat the semi direct product
H(2r, R) Sp(2r, R). Chapter 8 addresses unitary representations of this group.
3.1.3 Complex Lie Group

Although we have discussed Lie groups with real number parameters up to the previous subsection. Now, we consider Lie groups with complex parameter. Additionally,
we assume that the product is given as a holomorphic function for the complex parameter. Such a Lie group is called a complex Lie group. On the other hand, when a
Lie group is not necessarily a complex Lie group and it is needed to be distinguished
from a complex Lie group, it is called a real Lie group. In fact, a complex Lie group
has simpler structure than a real Lie group because the holomorphic condition is so
strong.
The most of physical meaningful transformations are not complex Lie groups
and but real Lie groups. Hence, we do not need to treat complex Lie groups from
the physical viewpoint. However, real Lie groups have too complicated structure to
directly treat without considering complex Lie groups. So, we consider complex Lie
groups, whose typical examples are summarized in the following Table 3.1.
Many real Lie groups have their corresponding complex Lie groups as their own
complexifications, respectively. However, each complex Lie group has more than
one corresponding real Lie groups so that a complicated treatment is required.
3.1.4 Other Examples of Real Lie Groups

Although we have listed several real Lie groups up to now, there are so many remaining real Lie groups, which are listed below. Compact real Lie groups corresponding
to complex Lie groups in Table 3.1 are given in Table 3.2. (The last 2 examples in
Table 3.2 do not have their corresponding compact real Lie groups in Table 3.1.) We
3.1 Lie Group
75
Table 3.1 Complex Lie group

Name of group
Complex general linear group
Complex special linear group
Complex orthogonal group
Complex special orthogonal group
Complex symplectic group
Table 3.2 Compact Lie group

Name of group
Notation
Definition
GL(n, C)
SL(n, C)
O(n, C)
SO(n, C)
Sp(2n, C)
{n n complex regular matrix }

{g GL(n, C)| det g = 1}
{g GL(n, C)|g T g = I}
O(n, C) SL(n, C)
{g GL(2n, C)|g T Jn g = Jn }
Definition
Dimension
Fundamental
group
{g GL(n, C)|g g = I}
U(n) SL(n, C)
O(n, C) GL(n, R)
n2
n2 1
n(n 1)/2
SO(n, R)
SO(n, C) GL(n, R)
n(n 1)/2
Sp(n)
D(n)
Sd(n)
n(2n + 1)
U(2n) Sp(2n, C)
diagonal matrices in U(n) n
diagonal matrices in
n1
SU(n)
1
1
Nonconnected
Z2 (n 3) Z
(n = 2)
1
Unitary group
U(n)
Special unitary group SU(n)
Real orthogonal group O(n, R)
Real special linear
group
Hyper-unitary group
Notation
Zn2
Zn1
2
denote the set of real regular matrices of size n by GL(n, R). The definition of the
fundamental group is given in Sect. 3.6.
Non-compact real Lie groups are listed in Table 3.3. Since a complex Lie group
usually has more than one corresponding non-compact real Lie groups, the table of
non-compact real Lie groups (Table 3.3) is very complicated. In the lower dimensional case, several groups in Table 3.3 are the same or isomorphic to each other
although all of them are not isomorphic to each other in the higher dimensional case.
SU(1, 1)
= SL(2, R) = Sp(2, R)
(3.18)
3.2 Lie Algebra

The previous section addressed real Lie groups and complex Lie groups. Although
complex Lie groups can be more easily treated than real Lie groups, complex Lie
groups still have difficulty in their own treatment. The structures of real Lie groups
and complex Lie groups are almost determined by their local structures around their
76
Table 3.3 Non-compact real Lie group

Name of group
Notation
Real general linear
group
GL(n, R)
Non-definite unitary U(p, q)

group
U (2m)
Real special linear
group
Real non-definite
unitary group
Non-definite
orthogonal group
elements of group
Dimension
{n n real regular matrices} n2

g GL(p + q, C)

g
g Ip,q g = Ip,q

g GL(2m, C)

g
gJm = Jm g
(p + q)2
(2m)2
SL(n, R)
GL(n, R) SL(n, C)
n2 1
SU(p, q)
SL(p + q, C) U(p, q)
(p + q)2 1
SU (2m)
SL(n, R) U (2m)
(2m)2 1
O(p, q)
GL(p + q, R) U(p, q)
(p+q)(p+q1)
2
SL(p + q, R) U(p, q)
(p+q)(p+q1)
2
SO(2m, C) U (2m)
GL(2n, R) Sp(2n, C)
m(2m 1)
n(2n + 1)
U(2p, 2q) Sp(2p +

2q, C)
n(2n + 1)
Special non-definite SO(p, q)

orthogonal group
SO (2m)
Real symplectic
Sp(2n, R)
group
Non-definite
Sp(p, q)
symplectic group
Heisenberg group
H(2n, R)
Affine group
Aff(n, R)
2n + 1
n2 + n
identity element, respectively. Further, their local structures around their identity
element are much simpler than their global structures. Hence, it is better to discuss
only their local structures before discussing the whole structures of the groups. For
this purpose, we introduce the concept of Lie algebra, which reflects the local
structure of each Lie group. This section gives only the formal definition of Lie
algebra, and the latter section explains the relation with Lie group. A real (complex)
linear space V is called a real (complex) Lie algebra when there is a map [, ] :
V V V (called a commutator) that satisfies the following conditions.1
1 In physics, we identify elements of a real Lie algebra and a Hermitian matrix as a physical quantity.
However, in our definition, the

real Lie algebra of SU(d) consists of skew-Hermitian matrices. That
is, in physics, the products of 1 and the elements of Lie algebra defined here are regarded as
elements of a Lie algebra.
3.2 Lie Algebra
77
Bilinearlity: For any element X V , the maps Y [X, Y ] and Y [Y , X] are

linear.
Skew-symmetry (skew-symmetricity): [X, Y ] = [Y , X].
Jacobi law: [X, [Y , Z]] + [Y , [Z, X]] + [Z, [X, Y ]] = 0.
For example, the set gl(V ) of linear maps on a real (complex) linear space V form
a real (complex) Lie algebra with the following commutator
[X, Y ] := XY YX.
(3.19)
Similarly, the set u(H) of skew-Hermitian matrices on a Hilbert space H forms a real
Lie algebra with the commutator (3.19). When a Lie algebra g is given as a subset
of matrices with the commutator (3.19), two elements X, Y g satisfy [X, Y ] = 0 if
and only if they are commutative with each other as matrices. Hence, even when a
Lie algebra g is not given as a subset of matrices, two elements X, Y g are called
commutative when [X, Y ] = 0. We need to notice that a complex Lie algebra is a
real Lie algebra but a real Lie algebra is not a complex Lie algebra in general.
A real (complex) linear map f from a real (complex) Lie algebra g1 to a real (complex) Lie algebra g2 is called a homomorphism for real (complex) Lie algebras
when [f (X), f (Y )] = f ([X, Y ]), (X, Y g1 ). Especially, when f is injective, f is
called an isomorphism. When there exists an isomorphism from a real (complex)
Lie algebra g1 to a real (complex) Lie algebra g2 , g1 and g2 are called isomorphic
to each other as Lie algebras, whose relation is written as g1
= g2 . In this book, we
simply denote a real Lie algebra by a Lie algebra, and a homomorphism for real Lie
algebras by a homomorphism for Lie algebras. In general, even though a map from
a complex Lie algebra to another complex Lie algebra is a homomorphism for real
Lie algebras, it is not necessarily a homomorphism for complex Lie algebras.
For two real (complex) Lie algebras g1 and g2 , their direct sum g1 g2 forms a
real (complex) Lie algebra with the commutator [X1 X2 , Y1 Y2 ] := [X1 , Y1 ]
[X2 , Y2 ]. Since a real Lie algebra g forms a real linear space, we define the complexification as the linear space by replacing the real coefficients by the complex
coefficients. Then, the commutator can be extended to the complexification so that it
forms a complex Lie algebra, which is called the complexification of g and is denoted
by gC . In general, even though two real Lie algebras g1 and g2 satisfy g1,C
= g2,C ,
the relation g1
= g2 does not necessarily hold, whose counterexample will be given
later.
Here, we define several fundamental concepts for real (complex) Lie algebras.
For this purpose, for real (complex) linear subspaces a and b of a real (complex)
Lie algebra g, we introduce the notations [a, b] := {[A, B]|A a, B b}, g(0) :=
g, g(n) := [g, g(n1) ], and D(0) g := g, D(n) g := [D(n1) g, D(n1) g]. Then, a is
called a real (complex) Lie subalgebra when [a, a] a. In particular, it is called an
ideal when [g, a] a. The quotient space g / a forms a real (complex) Lie algebra
and is called a quotient real (complex) Lie algebra. A real (complex) Lie algebra
g is called simple when its ideal is restricted to {0} or itself. As mentioned later,
the simplicity means the impossibility of a non-trivial decomposition. For any linear
78
subspace k of g, a linear subspace cg (k) := {X g |[X, Y ] = 0, Y k} forms a Lie

subalgebra of g and is called the centralizer of k with respect to g. When k is the Lie
algebra g, cg (g) forms an ideal of g, which is called the center of g and is simplified
to c(g).
On the other hand, when g(1) = 0, all of elements of g are commutative with each
other and g is called commutative. When there exists an integer k such that g(k) = 0,
g is called nilpotent. Moreover, when there exists an integer k such that D(k) g = 0,
g is called solvable. In the nilpotent case, g(1) is an ideal. Hence, we find that the
concepts nilpotent and simple are opposite to each other. When a solvable ideal
of g is limited to {0}, g is called semi simple. Even though g is not semi simple, we
have the following characterization. Let {gi }i be the set of all solvable ideals of g.
Then, i gi is also a solvable ideal of g, which is called the radical and is denoted
by rad g. Then, the quotient real (complex) Lie algebra g /rad g is semi simple.
Given a semi simple real (complex) Lie algebra g, we can choose simple real
(complex) Lie subalgebras g1 , . . . , gk such that g = ki=1 gi . Hence, the analysis
of a semi simple real (complex) Lie algebra can be reduced to that of simple real
(complex) Lie algebras.
Finally, given two real (complex) Lie subalgebras g3 and g4 of a real (complex) Lie
algebra g, the real (complex) linear subspace g3 + g4 := {X3 + X4 |X3 g3 , X4 g4 }
forms a real (complex) Lie subalgebra of g. In particular, when g3 g4 = {0}, g3 + g4
is isomorphic to g3 g4 .
3.3 Relation Between Lie Group and Lie Algebra I

3.3.1 Infinitesimal Transformation and Lie Algebra
As mentioned in Sect. 3.2, an element of real (complex) Lie algebra corresponds to an
infinitesimal transformation in a real (complex) Lie group around the identity element
I. The following discussion concretely gives this relation. Remember that a real
(complex) Lie group G is defined as a subgroup of GL(n, C) = GL(Cn ). We focus on
a one-dimensional differentiable subset {g(t) G|t R} satisfying g(0) = I, which
is called a path of G. Then, the real (complex) Lie algebra g(G) corresponding to G
as Fig. 3.2. As another definition of
is defined as the set of the limits X = limt0 g(t)I
t
the real (complex) Lie algebra g(G) corresponding to G, we often adopt
definition
the
Xn
g(G) := {X gl(Cn )|etX G, t R}, where eX := exp(X) :=
.
n=0 n!

Lie algebra and Lie group
e
e tY
X
e tX
Y
79
Since any two elements X, Y g(G) satisfy

t2
etY etX = et(X+Y )+ 2 [Y ,X]+O(t
(3.20)
as the limit t 0, we can show that

etY etX etY etX = et
[Y ,X]+O(t 3 )
(3.21)
Using this fact, we have

etY etX etX etY
t0
t2
etY etX etY etX I
etY XetY X
,
= lim
= lim
2
t0
t0
t
t
[Y , X] = lim
(3.22)
which naturally derives the commutator of Lie algebra from the product of Lie group.
Hence, the set of matrices X given as the above limits forms a real (complex) Lie
algebra with the commutator (3.19).
Conversely, when a real (complex) Lie algebra g is a real (complex) Lie subalgebra
of gl(H) with the commutator (3.19), the set
G(g) := {eX1 eXm |X1 , . . . , Xm g}
forms a real (complex) Lie group, whose product is the conventional matrix multiplication. When G is compact or nilpotent, we can simplify it as G(g) = {exp(X)|X
g}. This definition depends on the choice of the Hilbert space H. That is, even though
g1 gl(H1 ) is isomorphic to g2 gl(H2 ) as a Lie algebra, the Lie group G(g1 ) is
not necessarily isomorphic to the Lie group G(g2 ). For example, as mentioned later,
su(2) and so(3, R) are isomorphic to each other, SU(2) and SO(3, R) are not isomorphic to each other.
When the Lie group is simple as a group, the corresponding Lie algebra is simple.
However, the converse is not true in general. For example, although su(2) is a simple
real Lie algebra, SU(2) is not a simple group because it has the normal subgroup U2 .
When two Lie groups G and H correspond to Lie algebras g and h, respectively, and
H is a subgroup G, then h is a Lie subalgebra of g. Additionally, when H is a normal
subgroup of G, h is an ideal of g.
Then, the following theorem holds.
Theorem 3.1 Given a homomorphism f from a Lie group G 1 to a Lie group G 2 , the
map
f (etX ) I
, X g(G 1 ).
t0
t
f (X) := lim
is a homomorphism from the Lie algebra g(G 1 ) to the Lie algebra g(G 2 ).
80
However, we cannot necessarily make a homomorphism for Lie groups from a homomorphism for Lie algebras.
Proof The relations (3.21) and (3.22) imply that
f (et
[X,Y ]
)I
t2
f (e e e e ) I
f (etY )f (etX )f (etY )f (etX ) I
= lim
=
lim
t0
t0
t2
t2
tf (Y ) tf (X) tf (Y ) tf (X)
e
e
e
e
I
= lim
= [f (X), f (Y )]
2
t0
t
f ([X, Y ]) = lim
t0
tY tX tY tX
for X, Y g. Hence, we obtain the desired argument.
Although our aim is the characterization of real Lie groups, the process of this
aim is summarized as follows.
Characterization of
Characterization of
Characterization of
=
=
complex Lie algebras
real Lie algebras
real Lie groups
Exercise 3.4 Show that g(G 1 ) g(G 2 ) = g(G 1 G 2 ) based on the above definition of direct sum for Lie algebras.
Exercise 3.5 Show (3.20) and (3.21).
3.3.2 Examples
Table 3.4 summarizes the complex Lie algebras corresponding to the complex Lie
groups given in Table 3.1 in Sect. 3.1, where the notations of the Lie algebras are
given as the Fraktur of the notations of the corresponding Lie groups.
Table 3.5 lists the real Lie algebras corresponding to the compact Lie group given
in Table 3.2, where gl(n, R) is gl(Rn ). In the latter section, we define the compactness
for a real Lie algebra when a real Lie algebra is defined as a linear space with a
commutator. All of examples in Table 3.5 are compact under this definition.
Table 3.4 Complex Lie algebra

Notation of Lie algebra
Definition
gl(n, C)
gl(Cn )
sl(n, C)
{X gl(n, C)| Tr X = 0}
{X gl(n, C)|X T = X}
{X gl(2n, C)|X T Jn + Jn X = 0}
o(n, C)
sp(2n, C)

Table 3.5 Compact real Lie algebra
Notation of Lie algebra
Definition
u(n)
su(n)
o(n, R)
so(n, R)
sp(n)
d(n)
sd(n)
gl(n, C)|X
{X
= X}
u(n) sl(n, C)
gl(n, R) o(n, C)
= o(n, R)
u(2n) sp(2n, C)
Diagonal matrices in u(n)
Diagonal matrices in su(n)
81
Dimension
n2
n2 1
n(n 1)/2
n(n 1)/2
n(2n + 1)
n
n1
Table 3.6 Non-compact real Lie algebra

Symbol of Lie algebra
Definition
gl(n, R)
u(p, q)
u (2m)
sl(n, R)
su(p, q)
su (2m)
o(p, q)
so(p, q)
so (2m)
sp(2n, R)
sp(p, q)
h(2n, R)
aff(n, R)
gl(Rn )
gl(p + q, C)|X Ip,q
{X
+ Ip,q X = 0}
{X gl(2m, C)|XJm = Jm X}
sl(n, C) gl(n, R)
sl(p + q, C) u(p, q)
sl(n, C) u (2m)
gl(p + q, R) u(p, q)
sl(p + q, R) u(p, q)
so(2m, C) u (2m)
gl(2n, R) sp(2n, C)
u(2p, 2q) sp(2p + 2q, C)
Dimension
n2
(p + q)2
(2m)2
n2 1
(p + q)2 1
(2m)2 1
(p + q)(p + q 1)/2
(p + q)(p + q 1)/2
m(2m 1)
n(2n + 1)
n(2n + 1)
2n + 1
n2 + n
Table 3.6 lists real Lie algebras corresponding to non-compact Lie groups given
in Table 3.3. Similar to the compactness, in the latter section, we define the noncompactness for a real Lie algebra when a real Lie algebra is defined as a linear
space with a commutator. All of examples in Table 3.6 are non-compact under this
definition.
In the above table, for z R, x Rn , z R2n , A gl(n, R), the elements
Xh (z, z) and Xa (A, x) are defined as
82
0 x T z
Xh (z, z) := 0 0n y
0 0T 0

Ax
Xa (A, x) :=
,
00
(3.23)
(3.24)
where z is defined as z := (x1 , y1 , . . . , xd , yd )T for x, y Rn . Note that the real Lie

algebra o(n, R) is isomorphic to the real Lie algebra so(n).
The following relation holds between a real Lie algebra and a complex Lie algebra
when p + q = n.
gl(n, C)
= u (2m)C
= gl(n, R)C
= u(p, q)C , gl(2m, C)
= u(n)C
sl(n, C)
= su(n)C
= su (2m)C
= sl(n, R)C
= su(p, q)C , sl (2m, C)
o(n, R)C = so(n, R)C
o(n, C) = so(n, C) =
= o(p, q)C = so(p, q)C
o(2m, C) = so(2m, C)
= o (2m)C = so (2m)C
sp(2n, C)
= sp(n)C
= sp(2n, R)C
= sp(p, q)C .
(3.25)
Although the real Lie algebras su(n)C , sl(n, R)C , and su(p, q)C are not isomorphic
to each other as real Lie algebras, their complexifications are isomorphic to each
other, as explained in the above table. Since su(n)C is a compact real Lie algebra and
sl(n, R)C and su(p, q)C are non-compact real Lie algebras, sl(n, R)C and su(p, q)C
are not isomorphic to su(n)C as real Lie algebras. Even though the complexifications
of given Lie algebras are isomorphic to each other, the original real Lie algebras are
not necessarily isomorphic to each other. This fact indicates that a real Lie algebra
has a more complicated structure than a complex Lie algebra.
The following isomorphic relation for real Lie algebras holds in the lower dimensional case.
sp(2, R) = sl(2, R)
= su(1, 1)
su(2),
so(4, R)
so(3, R)
= su(2) su(2)
=
su(4)
sp(4, R), so(6, R) =
so(5, R) =
sl(2, C) = so(1, 3).
(3.26)
(3.27)
(3.28)
In particular, the second isomorphic relation in (3.27) is discussed precisely in

Sect. 5.2.2. Hence, the following isomorphic relation for complex Lie algebras holds
from (3.25).
sp(2, C)
= sl(2, C)
= sl(2, C), so(3, C)
so(4, C) = sl(2, C) sl(2, C), so(5, C)

= sp(4, C)
sl(2, C).
so(6, C) =
(3.29)
83
As mentioned later, it is known that complex simple Lie algebras are completely
classified. That is, they are limited to classical types sl(n, C), so(n, C), sp(n, C) and
exceptional types E6 , E7 , E8 , F4 , G 2 . In these examples, as explained in (3.29), only
so(4, C) is not simple.
Similarly, as mentioned later, since any compact simple Lie algebra uniquely
corresponds to a complex simple Lie algebra, compact simple Lie algebras are completely classified. Although the classification of non-compact simple Lie algebras is
not easy, they has been completely classified [66, Theorem 6.105, Sect. 10, Chap.
VI].
3.3.3 Central Extension of Real Lie Algebra

We have discussed the central extension by a real Lie group in Sect. 3.1.2. Similarly,
we can consider the central extension by a real Lie algebra.
For a real Lie algebra g, we consider the direct sum space R + g as a linear space.
Then, we denote the generator of the subspace R by Z. So, any element of R + g can
be written as aZ + X by using a R and X g. We define the commutator for two
elements aZ + X, a Z + X R + g as
[aZ + X, a Z + X ] = c(X, X )Z + [X, X ],
(3.30)
where c is a bilinear form satisfying

c(X, X ) = c(X , X).
(3.31)
That is, using a bilinear form c on g satisfying (3.31), we can define the above
commutator on the direct sum space R + g. The constructed Lie algebra gc is called
the central extension of R by g. Unless c is 0, the center c(gc ) of gc is R.
Next, we will show that the central extension by a Lie algebra is the Lie algebra
corresponding to the central extension by a Lie group. Consider the central extension
of U(1) by Lie group G. Let Z be the generator of the normal subgroup U(1) of the
central extension. Since Z is commutative with any elements of the central extension,
the relations
1
lim (et(aZ+X) (a Z + X )et(aZ+X) (a Z + X ))
t0 t
1
= lim (etX X etX X ) = c(X, X )Z + [X, X ]
t0 t
(3.32)
hold for X, X g(G). The property of Lie algebra guarantees the bilinearity of c
and the condition (3.31).
For example, the bilinear form c characterizing the central extension by the commutative Lie algebra R2 is given as ((p, q), (p , q ))
by using a symplectic inner
84
product and a real number . In the central extension, the generators Q := (1, 0) and
P := (0, 1) of the commutative Lie algebra R2 satisfy the following relations
[Q, P] = Z, [Q, Z] = [P, Z] = 0.
(3.33)
Especially, the above Lie algebra with = 1 is called the Heisenberg algebra with
degree 1 and is written as h(2, R). This fact shows that (3.12) gives the product
of the central extension of U(1) by the commutative Lie group R2 . An element of
Heisenberg group H(2, R) is written as gh (z, ) := epQqP eZ by using real numbers
z := (p, q) and .
Similarly, any central extension by the commutative Lie algebra R2r is given
by a bilinear form c, which is given as ((p, q), (p, q))
with a matrix and the
symplectic inner product ( , )
. In particular, when is the identity matrix, the central
extension is called the Heisenberg algebra of degree r and is written as h(2r, R). Let
ej be the element of the commutative Lie algebra R2 that has the j-th entry 1. The
Lie algebra h(2r, R) is generated by Q := ej , Pj := er+j , and the generator Z of the
center. The relations
[Pj , Qj ] = j,j Z, [Pj , Z] = [Qj , Z] = [Pj , Pj ] = [Qj , Qj ] = 0
hold. Hence, (3.15) gives the product of the central extension of U(1) by the
2r
of Heisenberg group H(2, R) is writcommutative Lie group
rR . Any element
ten as gh (z, ):= exp( j=1 pj Qj qj Pj )eZ by using a real vector z:=(q1 , . . . , qr ,
p1 , . . . , pr )T and . Since the Lie algebra h(2r, R) is nilpotent, it requires a different
treatment from semi simple Lie algebras.
3.4 Representation of Lie Algebra

3.4.1 Representation of Real Lie Algebra
The structure of a Lie group and its representations can be easily understood via the
structure of a Lie algebra and its representations. Hence, we focus on a representation
of a Lie algebra.
Since this subsection and the next subsection address only representations of a real
Lie algebra g, we abbreviate real Lie algebra to Lie algebra. Given a representation
f of the Lie group G corresponding to a Lie algebra g, we define the element of Lie
algebra gl(H) by
f(etX ) I
t0
t
f(X) := lim
(3.34)
85
for an element X g. Theorem 3.1 guarantees that f is a homomorphism from the

Lie algebra g to the Lie algebra gl(H) on the complex Hilbert space H. Generally,
a homomorphism f from the Lie algebra g to the Lie algebra gl(H) on the complex
Hilbert space H is called a representation of the Lie algebra g on H. As mentioned after Theorem 3.1, a representation of the Lie group G uniquely determines a
representation of the Lie algebra g, however, the converse does not hold in general.
A representation f of a Lie algebra g is called faithful when it is injective. It is
called a skew-Hermitian representation when all elements of the image of f are
skew-Hermitian matrices, i.e., the image of f is included in the Lie algebra u(H).
Let f be a unitary representation of a Lie group G. As mentioned in Sect. 3.6, the
corresponding representation of the Lie algebra g is a skew-Hermitian representation.
Hence, the skew-Hermitian representation of a Lie algebra is directly related to the
quantum theory.
The final aim is the characterization of unitary representations of real Lie groups.
The procedure of this book is the following.
Characterization of
skew-Hermitian representation of
real Lie algebra
Characterization of
unitary representation of
real Lie group
There exists another reason why a skew-Hermitian representation of a real Lie algebra
is important for quantum theory. In quantum theory, an important physical quantity
is often written as if(X) by using an element X of a real Lie algebra g and a skewHermitian representation f of the real Lie algebra g. This is a reason for the importance
of skew-Hermitian representations of real Lie algebras.
Given a representation f of a Lie algebra g on H, a linear subspace K of H is
called an invariant subspace when it satisfies the invariance condition given below;
Invariance:
f(X)u K, u K, X g.
The representation f of a Lie algebra g is called irreducible when its invariant subspace is limited to {0} or H. On the other hand, f is called reducible when there
exists an invariant subspace different from {0} and H. Then, employing the above
defined invariance, for a representation of a Lie algebra g, we can define the concepts of decomposable, indecomposable, completely reducible, complex conjugate representation and self-adjoint representation in the same way as the case
of a representation of a group.
When H is an infinite-dimensional Hilbert space, f is defined to be a skewHermitian representation if if(X) is a self-adjoint operator for any X g. Then,
we have the Schurs lemma as follows.
Lemma 3.2 Let f be an irreducible representation of a Lie algebra g on a finitedimensional Hilbert space H. When the relation
Af(X) = f(X)A, X g
holds for a matrix A on H, A is a constant matrix.
(3.35)
86
Proof Let c be an eigenvalue of A. This lemma can be shown by considering

the Kernel of A cI in the same way as Schurs lemma for a representation of a

group.
In fact, Lemma 3.2 can be extended to the case of an irreducible skew-Hermitian
representation on an infinite-dimensional space H. The extension can be shown by the
following steps. Firstly, we show the equivalence between A exp(f(X)) = exp(f(X))A
and (3.35). Secondly, we focus on the spectral decompositions of A and AA in the
same way as Lemma 2.4.
Given two representations f1 and f2 of a Lie algebra g on two different spaces H1
and H2 , we can define the direct sum representation f1 f2 of the Lie algebra g
on the direct sum space H1 H2 and the tensor product representation f1 f2 of
the Lie algebra g on the tensor product space H1 H2 as follows.
f1 f2 (X) := f1 (X) f2 (X), f1 f2 (X) := f1 (X) I + I f2 (X).
On the other hand, given two representations f1 and f2 of the Lie algebras g1 and g2
2 of the Lie algebra
on H1 and H2 , respectively, we can define the representation f1 f
g1 g2 on the tensor product space H1 H2 as follows.
2 )(X1 X2 ) := f1 (X1 ) I + I f2 (X2 ), X1 X2 g1 g2 .
(f1 f
Then, we can show the following lemma similar to Lemma 2.8.
Lemma 3.3 Given two irreducible representations f1 and f2 of the Lie algebras g1
2 is also irreducible. Conversely, any irreducible repand g2 , the representation f1 f
2 by using two irreducible
resentation of the Lie algebra g1 g2 can be written as f1 f
representations f1 and f2 of g1 and g2 .
Therefore, when a Lie algebra g is decomposed as a direct sum g1 gl ,
f
l by using
an irreducible representation of the Lie algebra g is given as f1
g
g
irreducible representations f1 , . . . , fl of 1 , . . . , l .
3.4.2 Real Representation

On the other hand, a homomorphism from a Lie algebra g to the Lie algebra gl(V )
on a real linear space V is called a real representation of Lie algebra g on V . A
representation with a complex Hilbert space is called a complex representation when
it is needed to distinguish it from a real representation. A real representation of a
Lie algebra g corresponds to a real representation of a real Lie group G. A real
representation f of a Lie algebra g is called an alternative representation when the
image of f is included in the Lie algebra o(V ), which is the set of alternative matrices
on V . It corresponds to an orthogonal representation of a Lie group G. For a real
representation, we can define the concepts of irreducible, faithful, decomposable,
87
indecomposable, and completely reducible in a way similar to those for a complex

representation. For a finite-dimensional complex representation of a Lie algebra, we
define complex conjugate representation in the same way as (2.40). Also, we define
a representation underlying real form and a complexifiable representation of a Lie
algebra in the same way as the case of a group. Schurs lemma does not hold for a
real representation of a Lie algebra g due to the same reason as the case of a real
representation of a group. In order to employ Schurs lemma, the irreducibility of a
real representation f on a real linear space V means the irreducibility of the complex
representation on the complex representation space V C.
3.4.3 Representation of Complex Lie Algebra

Next, we focus on a complex Lie algebra g, a homomorphism for complex Lie
algebras from g to the Lie algebra gl(H) on a complex linear space H is called a
representation for complex Lie algebras of g on H. For simplicity, we simplify it
to a C-representation of g on H. On the other hand, a representation of a complex Lie
algebra g on H means a homomorphism for real Lie algebras from g to gl(H). So, it
is not necessarily a C-representation. For a C-representation, we define the concept
of irreducible, faithful, decomposable, indecomposable, completely reducible.
Schurs lemma holds in the same way as the case of a real Lie algebra.
3.4.4 Adjoint Representation

As an example of a representation, given a real (complex) Lie algebra g, we define
the real representation ad of a real linear space g (the representation ad of a complex
Lie algebra g for complex Lie algebras) on the real (complex) linear space g by
ad(X)(Y ) := [X, Y ], which is called the adjoint representation ad of g. Hence, an
ideal can be defined as an invariant subspace with respect to the adjoint representation.
Since
ad(X)ad(Y )(Z) = [X, [Y , Z]] = [Y , [X, Z]] + [[X, Y ], Z]
=(ad(Y )ad(X) + ad([X, Y ]))(Z),
we have
ad([X, Y ]) = [ad(X), ad(Y )].
(3.36)
In particular, if and only if a real (complex) Lie algebra g is simple, the adjoint
representation is irreducible. For example, when g is a subspace of u(H) and the
inner product of g is defined as X, Y
:= Tr X Y , the adjoint representation is skew-
88
Hermitian. Since elements of Lie algebras gl(r), sl(r), u(r), and su(r) are written
as matrices on Cr , we can define natural representations of these Lie algebras on
Cr . These representations on Cr are called the fundamental representation and
are irreducible. Then, we can naturally define their representations on the n-fold
symmetric tensor product space of Cr , which are irreducible representations and are
called n-fold symmetric tensor representations. Similarly, their representations on
the n-fold alternative tensor product space of Cr are irreducible representations and
are called the n-fold alternative tensor representations.
Next, we define the adjoint representation ad of the Lie group G corresponding
to the Lie algebra g on the vector space g by ad(g) := exp(ad(tX)) for an arbitrary
element g = exp(tX) G. Although this definition seems to depend on the choice
of t and X, the representation is independent of this choice, as shown in Sect. 3.6.
When g is given as a linear subspace of gl(H), the relation ad(g)Y = gY g 1 holds
for Y g.
3.4.5 Projective Representation

We can consider a projective representation for a real Lie algebra. A linear map f is
called projective representation of the Lie algebra g when the relation
[f(X), f(Y )] = f([X, Y ]) + c(X, Y )iI.
(3.37)
holds, where c(X, Y ) is a real-valued bilinear map satisfying (3.31). In particular,

the linear map f is called a projective skew-Hermitian representation of the Lie
algebra g when all elements of the image of f are skew-Hermitian matrices. (skewself-adjoint operators in the infinite-dimensional case)
Then, we have the following lemma.
Lemma 3.4 The following conditions are equivalent to each other for a projective
representation f of g with a real-valued bilinear map c(X, Y ).
(1) The real number c(X, Y ) depends only on [X, Y ] g.
(2) There exists a representation f such that
f (X) = f(X) + c (X)iI,
(3.38)
where c is a real-valued linear function on g.

Proof The direction (2) (1) can be easily shown by a simple calculation. The
direction (1) (2) can be shown as follows. Condition (1) guarantees the existence
of a real-valued linear function c on g such that c(X, Y ) = c ([X, Y ]) for X, Y g.
When we define the projective representation f of g by (3.38), f is a representation

of g.
89
Therefore, when the condition of Lemma 3.4 holds, the projective representation f
of g can be replaced by the representation f of g. Hence, to avoid the above situation,
we call a projective representation f a proper projective representation when it
does not satisfy the above condition.
Lemma 3.5 A projective representation of a Lie algebra g on a finite-dimensional
space is not a proper projective representation.
Proof We consider a projective representation f of the Lie algebra g on the finitedimensional space H. Then, the map
f (X) := f(X)
1
(Tr f(X))I
dim H
(3.39)
is a projective representation of g. We give a real-valued bilinear function c according

to (3.37) for the projective representation f . The relation
[f (X), f (Y )] = f ([X, Y ]) + c(X, Y )iI
(3.40)
holds. Since the traces of the left hand side and f ([X, Y ]) are 0, c is also 0. So, f is

a representation. That is, f is not a proper projective representation.
Since a proper projective skew-Hermitian representation is limited to a projective representation on an infinite-dimensional space, it is sufficient to treat skewHermitian representations for a finite-dimensional quantum system.
3.4.6 Semi Direct Product Lie Algebra and Representation

Given an automorphic action T of a Lie group G 2 on a Lie group G 1 , the semi direct
product G 1 G 2 can be defined according to (2.2). When g1 and g2 are the Lie
algebras corresponding to G 1 and G 2 , respectively, we define the action T of G 2 on
g1 as
T (g)(etX1 ) T (g)(I)
, g G 2 , X1 g1 .
t0
t
T (g)(X1 ) := lim
Since g1 is a linear space, T is a real representation of G 2 on g1 . Then, the map

T (etX2 )(X1 ) X1
, X2 g2 , X1 g1 .
t0
t
T (X2 )(X1 ) := lim
is a real representation of Lie algebra g2 on g1 .

Given a real representation T of the Lie algebra g2 on g1 , we define the bilinear
map for the direct product set g1 g2 of g1 and g2 .
90
[(X1 , X2 ), (Y1 , Y2 )] := (T (X2 )Y1 T (Y2 )X1 + [X1 , Y1 ], [X2 , Y2 ]).
(3.41)
The above bilinear map satisfies the conditions for a Lie algebra, and the Lie algebra
is called the semi direct product Lie algebra g1 g2 of Lie algebras g1 and g2 .
In particular, when T is 0, the semi direct product Lie algebra g1 g2 is the direct
sum Lie algebra g1 g2 . For example, the Lie algebra aff(n, R) is given as the semi
direct product Lie algebra Rn gl(n, R) of the Lie algebras Rn and gl(n, R).
When representations f1 and f2 of Lie algebras g1 and g2 on H satisfy
[f2 (X2 ), f1 (X1 )] = f1 (T (X2 )X1 ),
(3.42)
we can define the semi direct product representation f1 f2 of the semi direct product
Lie algebra g1 g2 on H as
f1 f2 ((X1 , X2 )) := f1 (X1 ) + f2 (X2 ).
(3.43)
Finally, we remark that the semi direct product representation of the semi direct
product Lie algebra can be derived from the semi direct product representation of
the semi direct product Lie group.
Exercise 3.6 Given a real representation T of a Lie algebra g2 on another Lie algebra
g1 , we define the bilinear map on the direct product set g1 g2 by (3.41). Show that
the bilinear map satisfies the conditions for a Lie algebra.
Exercise 3.7 Assume that the condition (3.42) holds. Show that the map f1 f2
defined in (3.43) is a representation of the semi direct product Lie algebra g1 g2
on H.
3.5 Killing Form and Compactness

3.5.1 Killing Form
We define a bilinear form on a real (complex) Lie algebra g that is invariant for Lie
algebra. That is, we define the Killing form (X, Y )g := Tr ad(X)ad(Y ) for arbitrary
X, Y g. Then, we have
Tr ad([X, Y ])ad(Z) = Tr[ad(X), ad(Y )]ad(Z)
= Tr ad(X)[ad(Y ), ad(Z)] = Tr ad(X)ad([Y , Z]).
The Killing form satisfies the following associative law.
Associative law:
91
([X, Y ], Z)g = (X, [Y , Z])g .
This fact implies the equation

(ad(Y )X, Z)g = (X, ad(Y )Z)g .
(3.44)
Applying this property to the Lie group, we find that the action ad(g) of the Lie
group preserves the Killing form. That is, the equation (ad(g)X, ad(g)Z)g = (X, Z)g
holds.
The following lemmas hold with respect to the Killing form.
Lemma 3.6 The Killing form is non-degenerate for a real (complex) Lie algebra g
if and only if g is semi simple.
Let A be the dual map of a linear map A on a semi simple Lie algebra g with
respect to Killing form. Then, (3.44) implies that
ad(Y ) = ad(Y ).
(3.45)
That is, ad(Y ) is an alternative map with respect to the Killing form.
Lemma 3.7 When a real (complex) Lie algebra g is simple, any inner product satisfying the associative law is a constant times of the Killing form.
Lemma 3.8 When a real (complex) Lie algebra g is nilpotent, its Killing form satisfies (X, Y )g = 0. In this case, any inner product satisfying the associative law is
degenerate. That is, there exists a non-zero element X g such that (X, X)g = 0.
When a complex Lie algebra is semi simple, its Killing form is non-degenerate.
When a real Lie algebra is semi simple, we have another information for its Killing
form in addition to the non-degeneracy. Even though an inner product of a real
linear space is non-degenerate, there are many possibilities for the signs of the inner
product. That is, the matrix corresponding to the inner product might have positive
and negative eigenvalues. When all of eigenvalues are positive, the inner product is
called positive definite. When all of eigenvalues are negative, it is called negative
definite. When the matrix has both positive and negative eigenvalues, it is called
non-definite.
3.5.2 Compactness of Real Lie Algebra g

In this subsection, we address the compactness of a real Lie algebra g, and abbreviate
real Lie algebra and real Lie group to Lie algebra and Lie group, respectively. Firstly,
we discuss the compactness of an element X of the Lie algebra g. When {exp(tX)}t is a
compact subgroup of the corresponding Lie group G, the one-dimensional Lie group
{ad(exp(tX))}t = {exp(ad(tX))}t is compact. There is no one-to-one correspondence
92
between a Lie algebra g and a Lie group G. When a group {exp(ad(tX))}t is included
in a commutative compact subgroup of ad(G), X is called compact.
Lemma 3.9 The following conditions are equivalent to each other for an element
X of a Lie algebra g.
(1) X is compact.
(2) ad(X) is diagonalizable, and its eigenvalues are pure imaginary numbers.
(3) By choosing a suitable basis of g, ad(X) can be diagonalized as
0 c1
c1 0
..
0 cl
ad(X)
=
c
l 0
..
(3.46)
0
Proof (2) (1): Trivial.
(2) (3): Trivial.
(1) (2): Let ad(X1 ), . . . , ad(Xl ) be the generator of the Lie algebra corresponding to the commutative compact subgroup including {exp(ad(tX))}t . Due
to the compactness of the subgroup, there exist real numbers t1 , . . . , tl such that
exp(tj ad(Xj )) is the identity matrix. Thus, ad(X1 ), . . . , ad(Xl ) are diagonalizable
and their eigenvalues are pure imaginary numbers. Since ad(X) is written as a linear
sum of ad(X1 ), . . . , ad(Xl ), which are commutative with each other, Condition (2)

holds.
Hence, when X g is compact, the relation (X, X)g < 0 or ad(X) = 0 holds.
Then, we have the following relation between compactness and the Killing form for
a semi simple Lie algebra g.
Theorem 3.2 The following conditions are equivalent to each other for a semi simple Lie algebra g.
(1) The Killing form is negative definite.
(2) The Killing form is not non-definite.
(3) Any element of g is compact.
Proof (1) (2): Trivial.
(2) (3): Since any element X g satisfies (3.45) under the non-degenerate inner
product that is not non-definite, Condition (3) holds.
(3) (1): Since any eigenvalue of ad(X) is a pure imaginary number and its
Killing form is non-degenerate, Tr ad(X) is a negative number. So, Condition (1)

holds.
93
Then, we obtain the following characterization for the compactness of an element

of the Lie algebra g.
Theorem 3.3 The following three conditions are equivalent to each other for a Lie
algebra g.
(1) Any element of g is compact.
(2) There exists a compact Lie group G corresponding to the Lie algebra g.
(3) There exists a faithful skew-Hermitian representation f of g on a finitedimensional Hilbert space H.
Proof (1) (2): Since the Lie group {exp(ad(X))}g is compact, Condition (2)
holds.
(2) (1): This relation follows from the definition of compactness of X.
(1) (3): The adjoint representation satisfies Condition (3).
(3) (1): Condition (1) guarantees that f(g) is a Lie subalgebra of u(H). For any
element X g, the matrix ad(f(X)) on u(H) is a diagonalizable and has pure imaginary eigenvalues. The subspace f(g) is an invariant subspace with respect to the action
ad(f(X)) = f(ad(X)). When the action ad(f(X)) = f(ad(X)) is limited to f(g), the
matrix f(ad(X)) on f(g) is diagonalizable and all of its eigenvalues are pure imaginary. This fact guarantees that ad(X) is diagonalizable and all of its eigenvalues are

pure imaginary because f is faithful. Thus, Condition (1) holds.
Now, a Lie algebra g is called compact when the condition of Theorem 3.3 holds.
All of Lie algebras of Table 3.5 are compact under this definition, and all of Lie
algebras of Table 3.6 are non-compact. Then, the Killing form of the quotient Lie
algebra g / c(g) with respect to the center c(g) of g is negative definite. That is, g / c(g)
is semi simple. Thus, g / c(g) = [g / c(g), g / c(g)], which implies that
g = [g, g] + c(g).
(3.47)
In other words, the compact Lie algebra g is decomposed to the direct sum of the
commutative part c(g) and the semi simple part [g, g].
Due to Theorem 3.3, there exists a faithful skew-Hermitian representation of the
Lie algebra g on a finite-dimensional Hilbert space H if and only if g is compact.
Hence, to realize a faithful skew-Hermitian representation of a non-compact Lie
algebra g, we need to consider a representation on an infinite-dimensional Hilbert
space H. However, we need some additional cares to discuss a representation on an
infinite-dimensional Hilbert space. When we focus on a unitary representation of a
Lie group, the appearing operator on H is bounded so that no serious problem occurs.
However, when we focus on a skew-Hermitian representation of a Lie algebra, the
appearing operator on H is unbounded so that we need to care about the domain of
the operator. To consider this problem, we need to employ self-adjoint operators [88]
because this concept covers the difficulty of the domain. That is, a skew-Hermitian
representation f of a Lie algebra g on an infinite-dimensional Hilbert space H is
defined as a homomorphism for Lie algebras satisfying that if(X) is a self-adjoint
94
operator for any element X g. In particular, when the Killing form of a semi simple
Lie algebra g is negative definite, the representation space of its irreducible skewHermitian representation is limited to finite-dimensional. When the Killing form of
a semi simple Lie algebra g is non-definite, the representation space of its irreducible
skew-Hermitian representation is limited to infinite-dimensional.
In general, even though the physical system is given as a representation space of a
representation f of a non-compact group, it seems that the Hamiltonian of our interest
is usually given as the operator if(X) with a compact element X of the Lie algebra g.
In this book, even for a non-compact Lie algebra, we treat its representation based
on its compact elements due to the above physical reason.
Exercise 3.8 Show that (ad(g)X, ad(g)Z)g = (X, Z)g .
3.5.3 Casimir Operator

Next, we define an invariant operator called the Casimir operator. We choose
a basis {e1 , . . . , ed } of a semi simple real (complex) Lie algebra g and its dual
basis {e1 , . . . , ed } with respect to its Killing form. That is, we choose the dual basis
{e1 , . . . , ed } such that (ek , ej )g = k,j . Then, identifying the dual space g of g with

g via the Killing form, we define the Casimir operator C := dj=1 ej ej as an
on g. When A is a linear map
element of g g . Then, C is the identity operator

preserving the Killing form, we have C = dj=1 (Aej ) (Aej ). Since the action of
the Lie group G on g preserves the Killing form, C is invariant for the action of
the Lie
group G. For
a representation f on the Hilbert space H, we define the map
f : t Xt Xt t f(Xt )f(Xt ) from g g to the set of linear maps on H. Then,
space H as the image of C with
defining the Casimir operator Cf on the Hilbert
respect to the map f , we have the relation Cf = dj=1 f(ej )f(ej ). Cf is commutative
with the action of the Lie group G(g) corresponding to the Lie algebra g. That is,
the relations
f(g)Cf f(g)1 =
d

f(g)f(ej )f(g)1 f(g)f(ej )f(g)1
j=1
d

f(ad(g)ej )f(ad(g)ej )
j=1
d
d

ej ej = Cf
=f (ad(g)ej ) (ad(g)ej ) = f
j=1
j=1
holds for an element g G(g). Considering the derivative of g = exp(tX) with

respect to t, we have the following relation for any element X g
95
f(X)Cf = Cf f(X).
(3.48)
As mentioned in the previous section, many physical quantities are given as

if(X) by using a representation f of a Lie algebra. Some of other physical quantities are written as the Casimir operator Cf . When the Casimir operator Cf takes
different values dependently of irreducible representations, we can realize the measurement corresponding to the irreducible decomposition by measuring the physical
quantity Cf .
3.6 Relation Between Lie Group and Lie Algebra II

3.6.1 Universal Covering Group
This section discusses the relation between a representation of a Lie group and that of
a Lie algebra. From a representation (unitary representation) f of the real (complex)
Lie group G on H, we can define a representation (skew-Hermitian representation)
of the corresponding real (complex) Lie algebra g on H.
f(X) := limt0 f(exp(tX))I
t
However, the converse is not true in general. That is, we cannot necessarily define
a representation (unitary representation) of real (complex) Lie group G from a representation (skew-Hermitian representation) of a real (complex) Lie algebra g. The
aim of this section is clarifying this relation.
For simplicity, we assume that the group is connected. To address this problem,
we consider a differentiable curve c(t) (a path) in G. Two paths c(t) and c (t) with
the same initial and final points are called equivalent to each other and their relation
is written as g g when they are continuously converted to each other under the
constraint that the initial and final points are fixed.
When the final point of c(t) is the initial point of c (t), we define the product
between c(t) and c (t) as

(c c )(t) :=
c (2t)
0 t 21
c(2t 1)) 21 t 1.
(3.49)
For a path c(t), we also define the inverse path c1 (t) := c(1 t). Then, the set
of paths with the initial and final points g0 forms a group whose identity element
is a path g(t) = g0 , t [0, 1]. Especially, the set of paths that are equivalent to
the identity element forms a normal subgroup. Hence, we can define the quotient
group with respect to the normal subgroup. The quotient group has the product
[g] [g ] := [g g ], is called the fundamental group of G, and is written as 1 (G).
When G is connected, the fundamental group does not depend on the choice of g0 .
Since G is a group, we define another product for any two paths c(t) and c (t) as
follows.
96
(cc)(t) := c(t)c(t).
(3.50)
These two products have the following relation.

(c c )(c c )(t) = (c c )(t)(c c )(t) =
0 t 21
c(2t)c (2t)

c (2t 1)c (2t 1) 21 t 1
= (cc ) (cc )(t).

Hence, we have [(c c )(c c )] = [(cc ) (cc )]. Thus, denoting the path that constantly maps the element e by e, we have
[cc ] = [(c e)(e c )] = [(ce) (ec )] = [c c ]
[cc ] = [(e c)(c e)] = [(ec ) (ce)] = [c c].
Since [c c ] = [c c], we have the following lemma.
Lemma 3.10 The fundamental group 1 (G) of a real (complex) Lie group G is a
commutative group.
When the fundamental group consists only of the identity element, G is called
simply connected. Then, when the Lie group G is simply connected, the representation f of the Lie group G can be uniquely defined from a representation f of the
corresponding Lie algebra, as shown in the following way.
Since the neighborhood of the identity element e of the Lie group G can be
identified with the corresponding Lie algebra g, the representation of the element
et1 X1 can be given as f(et1 X1 ) = et1 f(X1 ) . In general, given a path c(t) in G, we can
define the representation of the Lie group G along the path. That is, we can define
f(c)(t) when the initial point of c is e. Hence, we can define f(g) for an element g G
as follows. First, we choose a path c(t) such that the initial point is e and the final
point is g. Then, we define f(g) to be f(c)(1).
Since this definition depends on the choice of the path c(t), we need to show
that it does not depend on the choice. Now, we consider another path c (t) such that
the initial point is e and the final point is g. Then, the initial and final points of the
path (c c 1 )(t) are g. Then, the initial and final points of the path (f(c) f(c)1 )(t)
in GL(H) are f(c)(1) and f(c)(1). Since G is simply connected, this path can be
continuously converted to the path that always maps to g. Since the image of f is also
simply connected, the path (f(c) f(c)1 )(t) in GL(H) can be continuously converted
to a path that maps to a fixed point in GL(H), where the initial and final points are
fixed during this continuous conversion. Hence, f(c)(1) is f(c)(1). Next, we show
that the map f is a representation of the Lie group G. For two elements g1 , g2 G,
we choose three paths c1 (t), c2 (t), and c3 (t) such that their initial points are e and
their final points are g1 , g2 , and g1 g2 , respectively. Then, it is sufficient to show that
f(c1 )(1)f(c2 )(1) = c3 (1). Since f(c1 )(t)f(c2 )(t) is the image along the path c1 c2 (t),
it is f(c1 c2 )(t). Since the path c1 c2 (t) has the initial point e and the final point g1 g2 ,
we can show the relation f(c1 c2 )(1) = f(c3 )(1) in the same way as the above discus-
97
[c ]
Fig. 3.3 Universal covering

group
G
[cc ]
[e ]
[c '] [cc ]
[c ]
e
[c ]
G
[cc ]
[ c'] [ cc ]
sion. Therefore, we find that the above defined map f is a representation of the Lie
group G.
On the other hand, when the connected Lie group G is not simply connected, there
is a possibility that no representation of the connected Lie group G corresponds
to the representation f of the Lie algebra g. However, even in such a case, there
exists a projective representation of the connected Lie group G corresponding to the
representation f of the Lie algebra g, as explained later. That is, there exists a quantum
mechanical symmetry of the connected Lie group G on H if and only if there exists
a skew-Hermitian representation of the Lie algebra g on H. To see this fact, we
make a central extension of the fundamental group 1 (G) by the real (complex) Lie
group G as follows. We consider the group of paths in G of the initial point e with the
product defined in (3.50). The identity element of this group is the path that constantly
maps to e. The existence of the inverse element can be easily checked. The quotient
group of this group with respect to the equivalent relation is called the universal
covering group of G and is written as G (see Fig. 3.3). Now, we define the map from
the universal covering group G to G by : c(t) c(1). We have 1 (e) = 1 (G).
Thus, the universal covering group G is a central extension of a fundamental group
(G) by G. For example, SU(2) is the universal covering group of the Lie group
SO(3, R) that expresses the rotations in the 3-dimensional space. In this case, since
the center of SU(2) is {e} isomorphic to Z2 and SU(2)/Z2
= SO(3, R), and SU(2)
is a central extension of SO(3, R).
When a central extension of the fundamental group 1 (G) by G is connected,
Further, we focus on a subgroup H of
it must be the universal covering group G.
the fundamental group 1 (G). Then, when a central extension of a quotient group
Let [(g1 , g2 )] be the residue

1 (G)/H by G is connected, it also must be G/H.
is the
class containing (g1 , g2 ) with respect to H. Then, the quotient group G/H
central extension of the factor system {[(g1 , g2 )]}. In particular, when 1 (G)/H is
2
satisfies the relation G/Z
Z2 , the quotient group G := G/H
= G, and is a central
extension of Z2 by G. Then, G is called the double covering group of G.
98
For example, we focus on the 2-dimensional Lorentz group that describes the
relativistic symmetry between the 2-dimensional space and time. The fundamental
group of its connected component SO(2, 1, R)0 is Z. Here, we choose the subgroup
2Z consists of even integers as subgroup H of Z. Then, we have Z/2Z = Z2 . Especially, we have SO(2,1, R)0 /2Z
= SU(1, 1), which implies that SU(1, 1) is a central
extension of Z2 by SO(2, 1, R)0 .
of the
In general, when we choose a discrete subgroup H of the center C(G)
universal covering group G of the Lie algebra g, the Lie algebra corresponding to
the Lie group G/H

is the Lie algebra g. Thus, the relation
= ad1 (I)
C(G)
(3.51)
The relation of (3.51) is trivial from

holds for an adjoint representation ad of G.
the definition. The relation of (3.51) follows from the fact that any element of G
can be written as exp(X1 ) . . . exp(Xl ) with a suitable choice of X1 , . . . , Xl g.
Hence, we can define the adjoint representation ad for the Lie group G/H.
In
is a discrete subgroup of
particular, when g is semi simple, since the center C(G)
the Lie algebra corresponding to the Lie group G/C(
is the Lie algebra g, and

G,
G)
is trivial.2 When
hence, (3.51) guarantees that the center of the Lie group G/C(
G)
the representation is irreducible, the action on the set of density matrices is the Lie
because any element in C(G)

does not change any density matrix.
group G/C(
G)
is a simple group (as a group). That is, the

When g is simple, the Lie group G/C(
G)
does not change any density matrix. Hence, in these cases,
action of the center C(G)
not the applications of SU(2) and SU(1, 1) but those of SO(3, R) and SO(1, 2, R)0
are directly reflected to the action on the set of density matrices.
3.6.2 Relation to Representation

Since an infinitesimal change in the universal covering group G equals an infinitesimal change in the Lie group G, the Lie algebra corresponding to G equals the Lie
algebra g corresponding to the Lie group G. Then, we can define the representation
f of the universal covering group G from the representation f of the Lie algebra g as
follows. We choose a path c(t) corresponding to an element of the universal covering
group G, and define f([c]) to be f(c)(1) by considering the image f(c)(t) along the
path c(t). The independence of the choice of c in this definition can be shown in the
same way as the simply connected case. In particular, when a representation f of g
is an irreducible representation, the representation f of the universal covering group
G is also an irreducible representation. Summarizing the above discussion, we have
the following theorem.
might be a continuous group. In particular, when g is
g is not semi simple, the center C(G)
is a continuous group, and hence, the Lie algebra of the
a nilpotent Lie algebra, the center C(G)
is different from the Lie algebra g.

Lie group G/C(
G)
2 When
99
Theorem 3.4 Given a connected Lie group G, there is a one-to-one correspondence

between a (unitary) representation f of the universal covering group G and a (skewHermitian) representation f of the Lie algebra g, and the relation
tX
f(etX ) = exp(tf(X)), f(X) = lim f(e ) I
t0
t
holds.
Since the universal covering group G is a central extension by G, the discussion
in Sect. 2.4 guarantees that an irreducible representation f of G uniquely yields a
projective irreducible representation of G. The following theorem holds for this
relation.
Theorem 3.5 Let G be a connected Lie group. Assume that its fundamental group
1 (G) is a finite group. Let k be the maximum value of the orders of elements of
1 (G). Then, given a finite-dimensional projective representation f, there exists a

factor system {ei (g,g ) }g,g taking values in Uk that is equivalent to the factor system

{ei(g,g ) }g,g of the projective representation f.
Hence, to discuss a factor system of a finite-dimensional projective representation,
it is enough to treat the group H(G : Zk ). So, given a projective representation f with a
factor system E H(G : Zk ), there exists a representation f of the universal covering
1 (G), there exists a representative
group G such that for an element g G = G/
g G of g satisfying that f(g) = f(g).

This fact can be shown from the one-to-one
correspondence given in Theorem 3.4.
To express this fact from another viewpoint, we introduce notations as follows. Let
g be the set of labels of irreducible skew-Hermitian representations of the Lie algebra
g, and g f be the subset of labels corresponding to finite-dimensional representations.
Applying this fact to irreducible projective unitary representations, we have

G f (E) = G f = g f .
(3.52)
EH(G:Zk )
When the dimension k of the representation space of the irreducible representation f

of the Lie algebra g is coprime to the maximum order k of elements of 1 (G), due
to Corollary 2.2, any projective representation f of G is a representation.
Proof Let {ei(g,g ) }g,g be the factor system of a given d-dimensional projective
representation f of G. We define a projective representation f equivalent to f by

f(g) 1/d
f(g) := f(g)( | det
)
SL(H). Then, the factor system {ei (g,g ) }g,g of f takes
det f(g)|
values in Ud . Let G be the central extension of the group Ud by the Lie group G
based on the factor system, and G 0 be the connected component of G containing e.
There exists a subgroup Ul of Ud such that Ul is included in the center of G 0 and

G 0 / Ul = G. Due to this fact and Lemma 3.1, the factor system {ei (g,g ) }g,g takes
values in Ul .
100
Since the Lie algebra corresponding to G 0 is the Lie algebra g corresponding to

G, the universal covering group of G 0 is the universal covering group G of G, which
1 (G 0 ) = G 0 . Since G/
1 (G) = G, Ul is a subgroup of 1 (G). This
implies that G/
fact indicates that l is a divisor of the maximum order k of elements of 1 (G). Hence,

the factor system {ei (g,g ) }g,g takes values in Uk .
The following corollary holds.
Corollary 3.1 The compact Lie group G and the maximum order k of elements of

1 (G) satisfies that EH(G:Z ) G(E)

= G = g .
k
Proof Theorem 3.7 in the next section guarantees that any irreducible projective
unitary representation of a compact Lie group G has a finite-dimensional representation space. Then, the discussion after Theorem 3.5 yields the argument of this

corollary.
3.6.3 Projective Representation

As mentioned above, when a Lie group G is a universal covering group, there is
one-to-one correspondence between a (skew-Hermitian) representation of the corresponding Lie algebra and a (unitary) representation of the Lie group G. Even
though the Lie group G is not a universal covering group, as mentioned before
Theorem 3.5, we can make a projective (unitary) representation of the Lie group G
from a (skew-Hermitian) representation of the corresponding Lie algebra. Applying
the discussion in Sect. 3.6.2 to a projective (unitary) representation, we can make a
projective (skew-Hermitian) representation of the corresponding Lie algebra from a
projective (unitary) representation of the Lie group G. Conversely, when G is a universal covering group, we can make a projective (unitary) representation of the Lie
group G from a projective (skew-Hermitian) representation of the corresponding Lie
algebra. Further, due to the discussion in Sect. 2.4, from a projective (unitary) representation of a universal covering group, we can make a projective (skew-Hermitian)
representation of another Lie group that corresponds to the same Lie algebra but is
not a universal covering group. Hence, even when the Lie group is not necessarily
a universal covering group, we can make a projective (unitary) representation of the
Lie group from a projective (skew-Hermitian) representation of the corresponding
Lie algebra.
However, Lemma 3.5 shows that there does not exist a proper projective representation on a finite-dimensional representation space. Due to Theorem 3.7 in the
next section, any irreducible projective unitary representation of a compact Lie group
has a finite-dimensional representation space. Hence, whenever the Lie algebra is
compact, it is sufficient to treat only a (skew-Hermitian) representation of the Lie
algebra. On the other hand, a non-compact Lie algebra has a representation whose representation space is infinite-dimensional. Hence, we need to treat a skew-Hermitian
representation of a non-compact Lie algebra.
101
3.6.4 Representation for Complex Lie Groups

We can define the universal covering group G for a complex Lie algebra g. For
this definition, we give the definition of a representation for complex Lie groups. A
homomorphism for groups from a complex Lie group G 1 to a complex Lie group
G 2 is called a homomorphism for complex Lie groups when the homomorphism
is a holomorphic function for both coordinates. In particular, a homomorphism for
complex Lie groups from a complex Lie group G to the complex Lie group GL(H)
is called a representation for complex Lie groups of a complex Lie group G. For
simplicity, it is simplified to a C-representation of a complex Lie group G.
Then, we have the same relation between a C-representation of a complex Lie
group G and a C-representation of the corresponding complex Lie algebra g as that
between a representation of a real Lie group and a representation of real Lie algebra.
We should remark that a representation (for real Lie groups) of a complex Lie group
G is not necessarily a C-representation of a complex Lie group G. Since a unitary
group is not a complex Lie group, a unitary representation of a complex Lie group
is not a C-representation.
3.7 Invariant Measures on Group and Homogeneous Space

This section addresses an integral (a measure) (dg) on a group G. The measure is
defined by determining the values (B) with every Borel set B of G. The measure
is called left-invariant when the relation (gB) = (B) holds for an element g G
and a Borel set B of G. Similarly, it is called right-invariant when (Bg) = (B).
Both left-invariant and right-invariant measures are uniquely determined except for
a constant factor [66, Sect. 2, Chap. VIII]. When they are the same, it is called the
invariant measure (Haar measure) of G. When the group G consists of finite
elements, the left-invariant and right-invariant measures are the same, and given as
(B) = #B. Then, the following theorem is known [66, Sect. 2, Chap. VIII].
Theorem 3.6 Assume that a Lie group G consists of finite number of connected
components. The left-invariant and right-invariant measures of the Lie group G are
the same if and only if the corresponding Lie algebra g satisfies that
Tr ad(X) = 0, X g .
(3.53)
The group G is called unimodular when it satisfies the above condition. For
example, when the corresponding Lie algebra g is simple, any element X g
can be written as [Y , Z] with two elements Y , Z g. Since (3.36) implies that
ad(X) = [ad(Y ), ad(Z)], we have (3.53), which guarantees the existence of the
invariant measure of G. Since a semi simple Lie algebra can be written as a direct
sum of simple Lie algebras, it satisfies (3.53) and thus, the Lie group G has the
invariant measure. Further, since any element of the Lie algebra corresponding to
102
a compact Lie group is compact, it satisfies (3.53). So, any compact Lie group
has the invariant measure. On the other hand, the affine group Aff(n) satisfies the
relation ad((X, y))(X , y ) = (ad(X)X , Xy X y) for (X, y), (X , y ) Aff(n)
=
gl(n) Rn , which implies Tr ad((X, y)) = Tr X. So, the affine group Aff(n) does
not satisfy (3.53).
A Lie group G (that might be a discrete group) is compact if and only if the total
measure (G) of the invariant measure of G is finite. Hence, a compact Lie group
(that might be a finite group) has the normalized invariant measure G , whose total
measure is normalized to 1. Thus, when G is a compact Lie group, an irreducible
unitary representation f of G on H satisfies

f(g ) f(g)|v
v|f(g) G (dg)f(g )
G

= f(g g)|v
v|f(g g) G (dg) =
f(g)|v
v|f(g) G (dg)
G
for v H and g G. Thus, Lemma 2.4

guarantees that f(g)|v
v|f(g) G (dg) is a
constant times of I. Also, the relation G u|f(g)|v
v|f(g) |u
G (dg) > 0 holds for

u H. Letting d1f be the above integral, we have df G f(g)|v
v|f(g) G (dg) = I.
Comparing the traces of both sides, we find that H is a finite-dimensional space and
its dimension is df . Summarizing this discussion, we have the following theorem.
Theorem 3.7 When G is a compact Lie group, an irreducible unitary representation

of G is a finite-dimensional representation. Hence, letting d be the dimension of the
we obtain
representation space of f for G,

f (g)|v
v|f (g) G (dg) = I.
(3.54)
The same fact can be shown for an irreducible projective unitary representation in the
same way. Notice that the integral (3.54) is not only a mathematical formula but also
a POVM, which gives a measurement in the physical system because any decomposition of the identity matrix (the identity operator corresponds to a measurement, as
explained in Sect. 1.1.
Next, we consider the case when G is non-compact. In this case, while G has the
invariant measure G , the total measure G is not necessarily finite. Also, an irreducible representation space is not necessarily finite-dimensional. In the following,
we assume that there exist a vector v0 with norm 1 and a real number d such that

d1 =
| v0 |f (g)|v0
|2 G (dg) < .
(3.55)
G
Then, we have the following theorem.

Theorem 3.8 Assume that G is a Lie group satisfying (3.53). G is not necessarily
compact. For a representation f , we choose a real number d satisfying (3.55). The
3.7 Invariant Measures on Group and Homogeneous Space
103
relation (3.54) holds for any vector v of norm 1. The value d does not depend on
the choice of v0 .
In the following, the real number d satisfying (3.55) is called the formal dimension
of the representation f with respect to the measure G .

Proof Schurs lemma guarantees that d G f (g)|v0
v0 |f (g) G (dg) is a constant times of I. By considering the trace of the product with the projection
|v0
v0 |, the relation (3.55) guarantees that the operator is I. Hence, the relation
1 = d G | v0 |f (g)|v
|2 G (dg) holdsfor any vector v of norm 1. Similarly, Schurs
lemma guarantees that the operator d G f (g)|v
v|f (g) G (dg) is also a constant

times of I. Using the same discussion, we find that this operator is also I.
A subgroup K of G is called a maximal compact subgroup of G when there is no
compact subgroup K of G such that K K . When a Lie group G has a non-trivial
compact subgroup, it is known that there exist a maximal compact subgroup K of
G and a subgroup P such that G = KP and any element g G uniquely has two
elements k K and p P such that g = kp. Such a decomposition of G is uniquely
determined for a given Lie group G, and is called the Cartan decomposition of
G [67, p.3]. Hence, the invariant measure G of G can be written as G (dg) =
K (dk)P (dp) by using the invariant measures K and P of K and P. When K
is the normalized invariant measure of K, the measure G (dg) = K (dk)P (dp) is
called the normalized invariant measure of G. When a Lie group G has a non-trivial
compact subgroup, we denote the invariant measure of G by G .
Next, we focus on the homogeneous space := G/H of Lie group G with a
stabilizer H. A measure on is called the invariant measure of when the
relation (B) = (gB) holds for any Borel subset B of . The invariant measure is uniquely determined except for constant factor [66, Theorem 8.36, Sect. 3,
Chap. VIII].
When G has the invariant measure , we define a measure on by (B) :=
(H1 (B)), and find that it is the invariant measure of , where H is a projection
from G to G/H. When the state |v
v| is invariant with respect to the action of H,
the action onto the state |v
v| can be described by an element of G/H. Hence, we
can choose the representative g() of the residue class for an element G/H so
that the relation (3.54) can be rewritten as

f (g())|v
v|f (g()) (d) = I.
(3.56)
d
Especially, when G has a non-trivial compact subgroup, we denote the normalized

invariant measure of by as we do that of G by G . Then, d in (3.56) is the
formal dimension (the dimension when G is compact).
On the other hand, when a non-compact group G does not have a non-trivial
compact subgroup, we cannot define the normalized invariant measure in this method.
For example, when G is the group R with respect to addition, which is a non-compact
b
group, we choose the invariant measure R ([a, b]) = a 12 dx as the normalized
invariant measure of R.
104
3.8 Fourier Transform on Lie Group

3.8.1 Commutative Case
In the above discussion, we have seen that irreducible representations of a Lie group
G can be derived from a representation of the corresponding Lie algebra. However,
a part of properties of the irreducible representations can be characterized without
through the corresponding Lie algebra. For the characterization, Fourier transform
plays a key role via the invariant measure of the Lie group G. The purpose of this
section to characterize an irreducible representation of a Lie group G by using Fourier
transform. For this purpose, we need to generalize Fourier transform to a general Lie
group.
In this subsection, we discuss the commutative case. Let us consider the unit
circle U(1) as the simplest example of Lie group. Since U(1) is a commutative
group, the representation space of an irreducible representation f of U(1) is onedimensional and f has the form f() = n , ( U(1)) with use of an integer k.
of its irreducible representation can be identified with the set Z of
Hence, the set U(1)
2
integers. Then, the set L 2 (U(1)) := {| 0 |(ei )|2 d < } of square-integrable

2
:= {{an }nZ |
functions on U(1) is isomorphic to the set L 2 (U(1))
n= |an | < }
= Z as linear spaces. That is, we define the
of square-integrable functions on U(1)
Fourier transform FU(1) and the inverse Fourier transform FU1(1) as the linear map
and the linear map F 1 from L 2 (U(1))
to L 2 (U(1))
FU(1) from L 2 (U(1)) to L 2 (U(1))
U(1)
2
1
FU(1) [](n) :=
(ei )ein d
2 0

an n .
FU1(1) [{an }]() :=
(3.57)
(3.58)
nZ
Then, the maps FU(1) and FU1(1) are unitary maps, and the maps FU(1) FU1(1) and
FU1(1) FU(1) are the identity map. (Exercises 3.9 and 3.10)
Next, we proceed to the case of G = R, in which, any irreducible representation
of R is one-dimensional because R is a commutative group. Then, the irreducible
is given as R x f (x) := eix .
unitary representation corresponding to R
Then, the Fourier transform and the inverse Fourier transform are as

1
(x)eix dx,
(3.59)
F[]() :=
2 R

1
(x)eix d = F[](x).
(3.60)
F 1 [](x) :=
2 R
105
These transformations are isomorphic transformations preserving the L 2 norm, and

the maps F F 1 and F 1 F are the identity map on L 2 (R) (Exercise 3.16). Here,
we list useful properties as
F[P]() = F[]()
(3.61)
FP = QF.
(3.62)
That is, we have
Applying F 1 twice, we have PF 1 = F 1 Q. Thus,

F[Q]() = i
d
F[]().
d
(3.63)
As an integral version, we have

F[eipQ ]() = F[]( p)
iqP
(x) = (x q).
(3.64)
(3.65)
When we define the convolution 1 2 for two functions 1 , 2 S(R) as

1
1 2 (x) :=
2
1 (y)2 (x y)dy,
(3.66)
we have
F[1 2 ] = F[1 ] F[2 ].
(3.67)
In the d-dimensional space Rd , the Fourier transform and the inverse Fourier
transform are as

1
(x)eix dx,
(3.68)
Fd []() :=
d
d
2
(2) R
Fd1 []() := Fd []().
(3.69)
Similarly, these transformations are isomorphic transformations preserving the L 2
norm, and the maps Fd Fd1 and Fd1 Fd are the identity map on L 2 (Rd ) (Exercise
3.16). The d-dimensional Fourier transform Fd has similar properties as
Fd [Pj ]() = j Fd []()
Fd []().
Fd [Qj ]() = i
j
(3.70)
(3.71)
106
Similarly, for two functions 1 , 2 S(Rd ), the convolution 1 2 is defined as

1 2 (x) :=
1
(2)
d
2
1 (y)2 (x y)dy,
(3.72)
we have
Fd [1 2 ] = Fd [1 ] Fd [2 ].
(3.73)
Remark 3.1 The precise derivations of the above relations require slightly difficult
mathematics. That is, we need to discuss the set S (R) of rapidly decreasing functions, which was introduced in Sect. 1.6. These derivations are given in Exercises in
this subsection. In these derivations, we firstly derive required relations on S (R).
Then, since S (R) is a dense subspace of L 2 (R), we extend these properties to the
whole of L 2 (R). Since Exercises 3.113.22 are too difficult, a reader who is not
interested in mathematical details can skip them.
Exercise 3.9 Show that the maps FU(1) FU1(1) and FU1(1) FU(1) are the identity
maps.
Exercise 3.10 Show that the maps FU(1) and FU1(1) are unitary maps.
df
Exercise 3.11 Consider a function f in S (Rd ). Show that iFd [ dx
]()=j Fd [f ]
j
(), i.e., Fd [Pj f ]() = j Fd [f ]().
Exercise 3.12 Show the equation Fd1 [] = Fd [], where is the complex conjugate function of .
Exercise 3.13 Show the equation Fd []() = Fd1 [].
Exercise 3.14 Consider a function in S (Rd ). Show that Fd [] also belongs to
S (Rd ).
Exercise 3.15 Consider a function 1 , 2 in S (Rd ). Show that
1 |Fd [2 ]
= Fd1 [1 ]|2
.
Exercise 3.16 Define the function (x) :=
1
1
(2) 4
x2
e 2 . Show that the maps Fd
Fd1 and Fd1 Fd are the identity map on L 2 (Rd ) by the following steps below.
a: Show that F[1/ ] = .

2
1 14
b: Show that ( 2
) (x y) (y)dy = 12 F[](y)eixy ey /2 dy for
S (R) by using Exercise 3.15.
c: Show the desired statement with d = 1 on S (R).
d: Show the desired statement with d = 1.
107
e: Show the desired statement with the general case.

Exercise 3.17 Show that Fd and Fd1 are unitary maps on L 2 (Rd ) by using Exercises
3.15 and 3.16
Exercise 3.18 Show (3.61) and (3.70) by using Exercise 3.11
Exercise 3.19 Show (3.64).
Exercise 3.20 Show (3.65) by using (3.64)
Exercise 3.21 Show the convolution 1 2 belongs to S (Rd ) for two functions
1 , 2 S (Rd ).
Exercise 3.22 Show the following relation for two functions 1 , 2 S (Rd ).
Fd [1 2 ]() =
1
d
(2) 2
Rd
Fd [1 ]( )Fd [2 ]( )d .
(3.74)
3.8.2 Non-commutative Case

The aim of this section is generalizing this discussion to compact Lie groups. Firstly,
any (projective) representation of a compact group G can be converted to a (projective) unitary representation in the same
way as the case of a finite group via the
inner product (2.52) with replacement of g by the Haar measure G . Due to the
same reason as the finite group case, to investigate the orbit of a vector in a representation
of a compact Lie group, it is sufficient to treat the representation
space

d
U
C
,
which
is
isomorphic
to
the
direct
sum
space
gl(U
)
with
G
G
respect to the left
side action by the Lie group G. We
define
the
inner
product
in
the

Tr
A
B
for
two
elements
direct sum space G gl(U ) as A, B
gl(U ) := kG
A = (A1 , . . .) and B = (B1 , . . .) in the direct

sum space d G gl(U ), which is the
inner product of the representation space G U C . We denote the complete
Then, we study how to treat the
space with respect to the inner product by L 2 (G).
under the action of the Lie group G. As defined in (2.41), the space
space L 2 (G)
gl(U ) also has the unitary representation D(f ) of G G, that is, the space gl(U )
also has the unitary representais written as U U . Hence, the space L 2 (G)
tion
of
G
G
as
well
as
that
of
G
and
can
be
regarded as the complete space of

U
.
G
For this purpose, we make a similar discussion for representations of a compact
Lie group G as that in Sect. 2.8. Since it is a compact Lie group G is usually not a finite
group, we have several technical difficulties to make a parallel discussion as that in
Sect. 2.8. So, we omit the detail of the derivation in the following discussion. Firstly,
we focus on the set L 2 (G) := {| G |(g)|2 G (dg) < } of square-integrable functions on the normalized invariant measure G of G.
108
Theorem 3.7 guarantees that any irreducible unitary representation of a compact

Lie group G has a finite-dimensional representation space. The number of its irreducible unitary representations is countable at most. We denote these irreducible
unitary representations by f , and their representation spaces by U . Letting |, m
be a CONS of U , we define the function R(, m, m ) := , m |f (g)|, m

. Then,
the orthogonal relation

G
R,m,m (g)R,
m
(g)G (dg) =
,m
1
m,m m ,m
d ,
(3.75)
holds.
Similar to (2.60), the relation (3.75) implies the orthogonal relation of characters

(g) (g)G (dg) =
(g) (g)G (dg) = ,
(3.76)
and the relation

(g)f (g)G (dg) =
G
,
I .
d
(3.77)
Especially, when the irreducible representations f and f are still irreducible under
the representation of a subgroup G of G, an orthogonal relation similar to (3.76)
holds even though the integral is replaced by the integral with the invariant measure
of G . Hence, similar to (2.64), (3.76) implies the relation
C , = C,
= C, = C , .
(3.78)
Here, the first equation follows from the Definition 2.45 of the multiplicity

coefficient C,
.
and Inverse Fourier transWe define Fourier transform F : L 2 (G) L 2 (G)
L 2 (G) as
form F 1 : L 2 (G)

f (g) (g)G (dg)
(3.79)
(F[]) := d
G

F 1 [A](g) :=
d Tr f (g)A .
(3.80)
Then, the relation (3.77) implies that FF 1 [A] = A. As shown later, the kernel of
F is {0}. Hence, the domain L 2 (G) of F is isomorphic to the image of F 1 , which
implies that F 1 F[] = . Thus, the set of functions { 1d R,m,m },m,m forms a
1
CONS of L 2 (G). Here, we should remark that
the
function
F
[A] L 2 (G) contains

with respect to the
all information for the entangled state A = G |A
L 2 (G)
representation.
109
1
Proof of Ker F = {0}. Section 2.8 employs the relation G (g) = |G|
in the proof of
1
the relation F F[] = . However, in the case of Lie group, we cannot employ
the same method because the denominator is infinity. Assume that Ker F = {0}.
Then, there exists
a continuous function such that F[] = 0. Any representation
f satisfies that G (g)f (g) G (dg) = 0.
Now, we focus on the left hand side action of G to L 2 (G), which is given
as (g ) (g 1 g ) for an element g of G. We denote the representation of G
by L. Since the representation L is given as a direct sum
representation of irreducible representations, the above relation
implies
that
G (g)L(g)G (dg) = 0.

Since L(g)(g ) = (g 1 g ), we have G (g)(g 1 g )G (dg) = 0. We choose
2
satisfying that (g 1 ) = (g) and substitute e into g . Then, we have

L (G)
2
G |(g)| G (dg) = 0, which implies that (g) = 0. So, we obtain Ker F = {0}.
Proof of (3.75). We choose an inner product of the space H such that the spaces
Letting
U and U are orthogonal to each other for distinct representations and .
f be the representation of G on U , we have

G
R,m,m (g)R,
m
(g)G (dg)
,m
m|f
m
, m |f (g)|, m
,
(g)|,

G (dg)

= , m |

.
f(g)|, m
, m|
f(g)G (dg) |,
The choice of the inner product guarantees that this value is 0 when = .
On the other hand, when = , Schurs lemma and the equality for the traces on

m|
I , where I is
f(g) G (dg) = ,m|,m
both sides guarantee that G f(g)|, m

,
d

the identity matrix on U . Thus, we obtain (3.75).
The above discussion can be extended to the case with projective representations
with factor system E as follows. Using the set G[E]

of irreducible projective rep
resentations corresponding to the factor system E := [{ei(g,g ) }g,g ] H(G : U(1)),

we define the linear space L 2 (G[E])

:= G[E]
gl(U ). Defining Fourier transform
and Inverse Fourier transform based on (3.79) and (3.80), we can show that they
in the same way. Hence, we obtain

are isomorphisms between L 2 (G) and L 2 (G[E])
(3.75)(3.77) by replacing the set G by the set G[E].

In this case, the Fourier transform is defined for a function L 2 (G) as well as in the representation case. To clarify
the dependence of the factor system F, we denote the Fourier transform by FE .
Next, we consider the case when G is non-compact. A unimodular group G is
called Type I when the following condition holds with the following definition [32];
f(G) := {A|f(g)A = Af(g), g G}
f(G) := {A|BA = AB, B f(G) }.
(3.81)
(3.82)
110
When any element of the set f(G) f(G) is a constant times of I, the representation f is limited to an irreducible representation with multiplicity. In this case, since
in general,
the representation space U might be infinite-dimensional for G[E],
we define the linear space gl(U )
as the set of HilbertSchmidt operators on U ,
:= G[E]
gl(U ). When the cardinality of G[E]
is
and define the space L 2 (G[E])
countable or finite and the formal dimension d of U is finite for any G[E], we
can define the Fourier transform and the Inverse Fourier transform based on (3.79)
Similarly, the
and (3.80). These are isomorphisms between L 2 (G) and L 2 (G[E]).
relations (3.75)(3.77) hold. Even though G[E] does not satisfy these conditions,
when the formal dimension d is finite for an element of a countable or finite
subset S of
G[E], the inverse Fourier transform can be defined on the linear space
2
L (S) := S gl(U ).
Even when G[E]

has cardinality of the continuum, when Lie group G is Type I,
the Fourier transform can be defined as follows [34, 32]. We define

f (g) (g)G (dg)
(3.83)
(F[]) :=
G
for G[E].
Then, it is known in [34, 32] that there uniquely exists a measure G[E]
(called the Plancherel measure) on G[E] such that F is a unitary under the norm
defined by

F[] :=
2
G[E]
Tr(F[]) (F[]) G[E]

(d).
(3.84)
Then, the inverse Fourier transform is given as follows.

F 1 [A](g) :=
G[E]
Tr f (g)A G[E]
(d).
(3.85)
Finally, we see how to recover the Fourier transform (3.59) and the inverse Fourier
transform (3.60) in the case of G = R. Since any irreducible representation of R is
one-dimensional due to the communitativity of R, the irreducible unitary representa of
is given as R x f (x) := eix . That is, the set R
tion corresponding to R
2
irreducible unitary representations of R is R. Hence, L (R) can be identified with the
square-integrable space L 2 (R). Choosing the invariant measure G as dx2 , as mentioned in the end of Sect. 3.7, the Plancherel measure is given as d2 . Therefore, the

Fourier transform is given as F[]() = 12 R (x)eix dx, and the inverse Fourier

transform is as F 1 [](x) = 12 R (x)eix d. The above general discussion also
shows that these transformations are isomorphic transformations preserving the L 2
norm, and the maps F F 1 and F 1 F are the identity map.
111
Finally, we consider the physical meaning of the Inverse Fourier transform

F 1 .

For this purpose, we prepare the operator T on the representation space G U
Cd as

T :=
d |I

d I | .
G

When the state is a pure entangled state |A
and we apply the measurement
corresponding to the POVM {f(g)T f(g) }gG , the measurement outcome g obeys the
distribution |F 1 [A](g)|2 G (dg). For example, for G = R, we consider the representation p eipx . The above measurement is the measurement of the momentum
P. This measurement relates the optimal measurement of the group action, which is
discussed in [50, Sect. 4.3]. In the projective representation case with the group R2 ,
the Inverse Fourier transform F 1 is closely related to Wigner function. That is, the
discussion of this section gives the bridge between group representation and Fourier
analysis.
Chapter 4
Representations of Typical Lie Groups

and Typical Lie Algebras
Abstract This chapter deals with representations of special Lie groups and special
Lie algebras. Since a representation of a Lie algebra can be classified with the highest
weight, those of a Lie group can be easily treated through those of the corresponding Lie algebra. Also, Lie algebras provide several concepts important for quantum
theory. Hence, this chapter is organized so that it constructs a representation of a Lie
group via that of the corresponding Lie algebra. This chapter starts with representations of Lie algebras su(2) and su(1, 1). Since the Lie algebra su(2) is compact
and the Lie algebra su(1, 1) is not compact, they require different treatment caused
in this difference. As they have unexpected common features, we handle both in a
unified way. Then, we proceed to representation of the Lie algebra su(r ) by using
Young diagrams. Especially, the representation of the Lie algebra su(r ) on the tensor
product space is closely related to that of the permutation group on the same tensor
product space. The relation is called Schur duality. We also consider what a finite
subgroup of a Lie group can replace the Lie group when its representation is given.
Such a problem is called design, and is discussed in this chapter.
4.1 SL(2, C) and Its Subgroup

The groups SO(3, R) and SO(1, 2, R)0 play a very important role in physics. The
previous chapter clarified that their Lie algebras are isomorphic to su(2) and su(1, 1),
respectively. Important physical quantities are expressed by the skewHermitian
representations of their Lie algebras. In the following, we investigate the structure of
the group SL(2, C) firstly. Then, based on this investigation, we study the structures
of the groups SU(2) and SU(1, 1), which are required to discuss the skewHermitian
representations of su(2) and su(1, 1), respectively.

DOI 10.1007/978-3-319-44906-7_4
113
114
4 Representations of Typical Lie Groups
4.1.1 Gauss Decomposition of SL(2, C)

Since the complex Lie group SL(2, C) has the simplest structure among noncommutative complex Lie groups, this section studies typical properties of SL(2, C),
its subgroup, and its complex Lie algebra, which are the foundation of the latter discussion. We choose the following three elements in the complex Lie algebra sl(2, C)
of SL(2, C) as its basis.
E 0,1 :=
1
2

1 0
00
01
, K +,1 :=
, K ,1 :=
,
0 1
10
00
(4.1)
where the lower subscript 1 indicates the Lie group SU(2), i.e., this basis is
deeply related the real Lie group SU(2) because these matrices produce a Gauss
decomposition of SU(2) as (4.5). This basis satisfies the commutation relation
[E 0,1 , K ,1 ] = K ,1 , [K +,1 , K ,1 ] = 2E 0,1 , which plays an important role in the
representation theory of a Lie algebra. Any element of sl(2, C) can be written as a
linear combination of this basis. The same fact can be shown for almost all the
elements of SL(2, C). That is, any matrix g SL(2, C) can be written as
g = exp( K ,1 ) exp( E 0,1 ) exp( K +,1 )

/2

/2
0
+ e/2 e/2
1 0
e
1
e
=
=
e/2
e/2
0 e/2
1
01
(4.2)

b a 1
with a suitable choice of x, y, z C unless it is written as
. In other words,
a 0
1

0
1
by letting M :=
, P :=
, g is written as g = P M J P J 1 unless
0
01

b a 1
g is written as
.
a 0

b a 1
On the other hand, when g is written as
, we have g = Ma J Pba . Howa 0
ever, the set of such elements has measure 0 under the proper measure of SL(2, C).
That is, except for a subset with measure 0, an element of SL(2, C) has the decomposition (4.2), which is called the Gauss decomposition of g and plays an important
role for studying the structures of SL(2, C) and its subgroup.
4.1.2 Structure of SU(2)

The basis
F1x
1
:=
2
0 i
i 0

,
y
F1
1
:=
2
0 1
1 0
1
=
J,
2
F1z
1
:=
2
i 0
0 i
115
is very useful for discussing the real Lie algebra su(2) of the subgroup SU(2)
y
y
of SL(2, C), and satisfies the commutation relation [F1z , F1x ] = F1 , [F1z , F1 ] =
y
z
x
x
F1 , [F1 , F1 ] = F1 . We replace the super subscripts x, y, z by 1, 2, 3 and define
j,k,l as
j,k,l
1 when ( j, k, l) = (1, 2, 3), (2, 3, 1), (3, 1, 2)

= 1 when ( j, k, l) = (3, 2, 1), (2, 1, 3), (1, 3, 2)
0 otherwise.
(4.3)
Then, the commutation relation can be rewritten as

j
[F1 , F1k ] =
3

j,k,l F1l .
(4.4)
l=1
Combining these basis, we can recover the above basis of sl(2, C) as E 0,1 = i F1z ,
y
K ,1 = i F1x F1 . The matrices E 0,1 and K ,1 are not elements of su(2), however,
they are useful for discussing representations of su(2) later. We focus on the subgroup
H = {exp(t F1z )} of SU(2) isomorphic to U(1). By using two complex numbers
2
2
1, any element
and satisfying
of SU(2) can be written
the condition || + ||
=it/2
eit/2

e
by applying the element
as
, which is converted to
eit/2 eit/2

z
exp(t F1 ) of H from the right hand side. Hence, an element of the homogeneous space
SU(2)/H with the stabilizer H can be written as the ratio and . Hence, SU(2)/H
which is the addition of to C, where corresponds to
is Riemann sphere C,
= 0 (see Fig. 4.1). When = 0, an element of SU(2)/H is given by / C. So,
we choose the representative of the element SU(2)/H

= 0 such
with the condition
cos
ei sin
by using
that R. The representative can be written as
ei sin cos
R, R. Using := ei tan and := log(1 + ||2 ), we obtain its Gauss
Fig. 4.1 Riemann sphere
| |= 1
=0
116
decomposition given in (4.2) as follows.

cos
ei sin
i
e sin cos

= exp( K ,1 ) exp( E 0,1 ) exp( K +,1 ).
Letting := ei , we also have the relations

cos
ei sin
i
e sin cos
ei
= exp
i
e 0
0

= exp( K ,1 K +,1 ), (4.5)
which play an important role in the latter sections. Since the matrix is uniquely
determined by = ei tan and = ei , the above matrix is denoted by g,1 and
g,1 .
The set of traceless 2 2 Hermitian matrices is closed with respect to the
action of SU(2). For an element g
SU(2),
we define themap on the set of
z
x yi
z
x yi
traceless matrices as
g
g 1 . Since (x 2 +
x + yi z
x + yi z

z
x yi
z
x yi
2
2
y + z ) = det
g 1 , So, any element
= det g
x + yi z
x + yi z
g SU(2) gives a 3 3 orthogonal matrix (g) based on the coordinate (x, y, z)
R3 , which is a representation of SU(2) on R3 . Thus, the relations
0 0 0
0 0 1
0 1 0
y
(F1x ) = 0 0 1 , (F1 ) = 0 0 0 , (F1z ) = 1 0 0
0 1 0
10 0
0 0 0
hold. Hence, the Lie algebras su(2) and so(3, R) are isomorphic to each other. However, the homomorphism from SU(2) to SO(3, R) is not an isomorphism because
the matrices I, I SU(2) are mapped to the unit matrix on R3 . The fundamental
group of the Lie group SU(2) is trivial, i.e., {e} and that of the Lie group SO(3, R) is
Z2 . The central extension of Z2 by SO(3, R) is isomorphic to SU(2). So, we obtain
the first equation in (3.8).
4.1.3 Structure of SU(1, 1)

Next, let us consider the Lie algebra su(1, 1) of the Lie subgroup SU(1, 1) of
SL(2, C). For this purpose, we choose the following basis
x
F1
1
:=
2
0 i
i 0

,
y
F1
1
:=
2

0 1
,
1 0
z
F1
1
:=
2

i 0
,
0 i
where the lower subscript 1 expresses the Lie group SU(1, 1). They satisfy the
y
y
y
z
z
x
x
x
commutation relations [F1
, F1
] = F1 , [F1
, F1 ] = F1
, and [F1
, F1 ] =
117
z
F1
. Combining these basis, we can recover the basis of sl(2, C) as follows
E 0,1 :=
z
i F1
1
=
2
y
x
+ F1
K +,1 := i F1
y
x
F1
K ,1 := i F1

1 0
,
0 1

0 0
=
,
1 0

01
=
.
00
They satisfy the commutation relation [E 0,1 , K ,1 ]= K ,1 , [K +,1 ,

K ,1 ]= 2E 0,1 . The matrices E 0,1 and K ,1 are not elements of su(1, 1),
but are useful for discussing representations of su(1, 1) later. The group H defined
above is a subgroup of SU(1, 1), which is isomorphic to U(1). By using two com2
2
= 1, any element
of SU(1, 1) can be
plex numbers

and
satisfying that ||
||it/2
eit/2
e
by applying the element
written as
, which is converted to
eit/2 eit/2
exp(t F1z ) of H from the right hand side. Hence, an element of the homogeneous space
SU(1, 1)/H with the stabilizer H is given as the ratio between and . Hence, the
homogeneous space SU(1, 1)/H is the unit disk D := { C||| < 1} (Fig. 4.2),
and its elements are given as / C. So, we choose the representative of the element SU(1,
R.Then, by using R, R, the representative is
1)/H such that
cosh ei sinh
. Letting := ei tan and := log(1 ||2 ),
written as i
e sinh cosh
and using (4.2), we have its Gauss decomposition based on K ,1 , E 0,1 as follows
cosh ei sinh
i
e sinh cosh

= exp( K ,1 ) exp( E 0,1 ) exp( K +,1 ).
(4.6)
Hence, letting := ei , we obtain
Fig. 4.2 Unit diskUnit disk
=0
| |< 1
118
cosh ei sinh
i
e sinh cosh

= exp
ei
i
e 0
0

= exp( K ,1 K +,1 ), (4.7)
which will play an important role later. Since the matrix is uniquely determined by
= ei tan and = ei , the above matrix is denoted by g,1 and g,1 .
As mentioned in (3.18), the Lie group SU(1, 1) is isomorphic to the Lie groups
SL(2, R) and Sp(2, R). In particular, the Lie groups Sp(2, R) and SL(2, R) are the
subgroup of SL(2, C). Now, we study their relation with SU(1, 1). By using two
complex numbers = a ib and = c + id satisfying that ||2 ||2 = 1, an
element g of SU(1, 1) can be described as

g=

=
a ib c + id
c id a + ib

.
Then, the map : SU(1, 1) SL(2, R) is defined as

a ib c + id
(g) := W
c id a + ib

1
1 1
W :=
.
2 i i

W

=
a+c b+d
b + d a c

,
(4.8)
(4.9)
Here, we consider the action of g to the pair of complex numbers z = x + i y and

z = x i y (defined by two real numbers x and y) as
x + iy
x iy

gz
gz

:= g
x + iy
x iy

.
(4.10)
When we focus on the real numbers x and y, by using defined the above, the
transformation (4.10) can be written as

x
x
(g)
.
y
y
Then, the matrix (g) is contained in SL(2, R). Since the map g (g) preserves
the product of matrices and is one-to-one, SU(1, 1) and SL(2, R) = Sp(2, R) are
isomorphic to each other. Their fundamental group is Z. We denote the universal
1). The central extension of Z2 by SU(1, 1)
covering group of SU(1, 1) by SU(1,
1). Then, taking
is the double covering group of SU(1, 1) and is denoted by SU(1,
the derivative in the group isomorphic
relation given
the map
the
above, we obtain

y
z+x
zi y + xi
. Since the
=
from su(1, 1) to sl(2, R) as
z + x y
y xi zi
map preserves the product of Lie algebras, the map gives the isomorphic relation between su(1, 1) and sl(2, R), which is described in (3.26). Hence, the basis
119

01
1 0
y
x
, (F1 ) = 21
, and
of sl(2, R) = sp(2, R) is given as (F1
) = 21
10
0 1

0 1
z
(F1
= 1
) = 21
J.
2
1 0
Exercise 4.1 Show the relation SO(1, 2)0
= SU(1, 1)/Z2 by steps below.

z
x yi
|(x, y, z) R3 is invariant under
a: Show that the set M1,1 :=
x + yi
z
the map X g X g for any matrix g SU(1, 1).
b: Show that det X = det g X g for X M1,1 and any matrix g SU(1, 1).
c: Show the relation SO(1, 2)0
= SU(1, 1)/Z2 .
4.2 Irreducible SkewHermitian Representation

of su(2) and su(1, 1)
4.2.1 Construction of Irreducible SkewHermitian
Representation
This section treats irreducible unitary representations of the Lie groups SU(2) and
SU(1, 1), which have the simplest structure among non-commutative Lie groups.
For this purpose, we deal with irreducible skewHermitian representations of the
corresponding Lie algebras su(2) and su(1, 1). The complexifications of these real
Lie algebras are sl(2, C). However, the structures of their irreducible skewHermitian
representations are completely different to each other.1
In the following, we deal with irreducible skewHermitian representations of
su(2) and su(1, 1) based on a unified approach, which will clarify their definitive
difference. For this purpose, we choose the parameter s taking the value 1 or 1,
where 1 indicates the Lie algebra su(2) and 1 does the Lie algebra su(1, 1). Then,
we obtain the commutation relation
[Fsz , Fsx ] = Fsy , [Fsz , Fsy ] = Fsx , [Fsx , Fsy ] = s Fsz
(4.11)
with respect to the basis. We also denote the images of the generators with respect
to the skewHermitian representation f by
Fas := f(Fsa ), a = x, y, z, s = 1, 1.
Then, we obtain
E0,s := iFsz , K,s := iFsx Fsy , s = 1, 1.
1 For
the details of skewHermitian representations of su(1, 1), see the reference [59].
(4.12)
120
Since the Hamiltonian is often given by E0,s in a quantum system, it is important to

study the Hermitian matrix (the self-adjoint operator) E0,s .
By extending the representation f to the representation of the complex Lie algebra
sl(2, C), the matrices/ the operators E0,s and K,s can be formally described by using
E 0,s , K ,s given in Sect. 4.1 as follows
E0,s := f(E 0,s ), K,s := f(K ,s ).
These operators satisfy the following relations
[E0,s , K,s ] = K,s , [K+,s , K,s ] = 2sE0,s
E0,s
K+,s
= E0,s ,
(4.13)
= K,s .
Since we can conversely recover Fsx , Fs , and Fsz from E0,s , K,s as
Fsz = iE0,s , Fsx =
i
1
(K,s + K+,s ), Fsy = (K,s K+,s ),
2
2
it is sufficient to discuss the actions of E0,s and K,s . Further, the two relations
n
n
n
n1
] = nK,s
, [K+,s , K,s
] = K,s
ns(2E0,s (n 1)I )
[E0,s , K,s
(4.14)
hold (see Exercise 4.3).

In the following, we assume that there exists an eigenvector v0 corresponding to
the eigenvalue of the Hermitian matrix /the self-adjoint operator E0,s such that
the eigenvector v0 is contained in the Kernel of K+,s and
v0
= 1. It is known that
this assumption holds for an irreducible skewHermitian representation of su(2),
but does not necessarily hold for an irreducible skewHermitian representation of
su(1, 1). In fact, there exists a skewHermitian representation of su(1, 1) that does
not satisfy this assumption.
When K,1 has a non-trivial Kernel instead of K+,1 , the following discussion
is valid by exchanging the roles of K+,1 and K,1 . However, it is known that
there exist irreducible skewHermitian representations of su(1, 1) such that K+,1
nor K,1 does not have a non-trivial Kernel. A part of such representations are
called principal series representations and another part of them are called complementary series representations [59]. Since this book does not deal with such an
irreducible skewHermitian representation, we assume the above assumption. The
n
v0 satisfies
relation (4.14) guarantees that the vector vn := K,s
n
n
n
n
E0,s + [E0,s , K,s
])v0 = (K,s
nK,s
)v0 = ( n)vn .
E0,s vn = (K,s
Hence, the vector vn is the eigenvector of the Hermitian matrix /the self-adjoint
operator E0,s corresponding to distinct eigenvalues. So, the vector vn is orthogonal
to the vector vm with n = m. Since
4.2 Irreducible SkewHermitian Representation of su(2) and su(1, 1)
K+,s K,s vn1 = ns(2 (n 1))vn1 ,
121
(4.15)
(see Exercise 4.2) we have
K,s |vn1 = ns(2 (n 1))vn1 |vn1 .

vn1 |K,s
(4.16)
The property of the inner product of the representation space of the skewHermitian
representation guarantees the relation ns(2 (n 1)) 0 for any integer n 0.
So, the condition with n = 1 implies that s 0.
y
When = 0, we have K,s v0 = 0. Thus, we have Fs = Fsx = 0, which implies that
the skewHermitian representation is the trivial one-dimensional skewHermitian
representation. Hence, we discuss only the case with s > 0. Assume that s = 1.
We have > 0. When 2 is an integer, we have K,s v2 = 0, which implies that the
skewHermitian representation is a 2 + 1-dimensional irreducible skewHermitian
representation. When 2 is not an integer, the vector vn does not vanish for any
integer n 0. So, there exists a positive integer n such that (2 (n 1)s) is a
negative value, which yields the contradiction. Thus, we find that 2 is an integer. Now, we assume that s = 1. So, we have < 0, which implies the inequality
ns(2 (n 1)) > 0 for any integer n 1. Hence, the space spanned by {vn |n 0}
is an irreducible representation space.
The eigenvalue of E0,s corresponding to the vector in the kernel Ker K,s
uniquely determines the skewHermitian representation. So, in the following, we
regard the eigenvalue as the element of the set G labeling the irreducible
skewHermitian representation so that the representation space is written as U .
In particular, we denote it by U (g) when we need to clarify the Lie algebra g of our
interest.
K,s |vn1 = ns(2 (n 1))vn1 |vn1 ,

Further, since vn |vn = vn1 |K,s
we have vn |vn = n!s(2 (n 1))s(2 (n 2)) s(2). Hence, the normalized eigenvector |; n :=
v1n
vn of E0,s corresponding to the eigenvalue
1
n satisfies K,s |; m + 1 = K,s
vn1
v
= ns(2 (n 1))
v1n
vn =
n1
s( m)( + m + 1)|; m when m + 1 = (n 1). The eigenvalue n

of E0,s is called a weight, and the normalized eigenvector |; n is called the
weight vector. So, we obtain
E0,s |; m = m|; m

K+,s |; m = s( m)( + m + 1)|; m + 1

K,s |; m = s( m + 1)( + m)|; m 1.
(4.17)
(4.18)
(4.19)
Since the real number is the minimum value among weights in the representation
space U , is called the highest weight. Hence, we denote the skewHermitian
representation with the highest weight by f .
122
In the quantum theory, the Hamiltonian is often given by E0,s . In this case, the
value m is the energy of the eigenstate |; m. In particular, the highest weight state
has the minimum energy and is called the ground state.
Remark 4.1 In the case of U(1, 1), the dimension of H is infinite. Since the operators
y
y
Fsx , Fs , and Fsz are bounded, we need careful treatment for them. Since iFsx , iFs , and
z
iFs are Hermitian band diagonal with respect to the base {|; m}, Lemmas 1.4 and
1.5 guarantee that they are essentially self-adjoint.
Remark 4.2 The skewHermitian representation of su(2) on U can be extended to
an irreducible C-representation of the complex Lie algebra sl(2, C). However, the
same discussion is not true for the skewHermitian representation of su(1, 1) on U
due to the infinite-dimensionality of the representation space. This is because there is
no method to define a non-skewHermitian representation on an infinite-dimensional
space.
Exercise 4.2 Show (4.15) by using (4.13).
Exercise 4.3 Show the equations in (4.14) by using (4.13).
4.2.2 Unitary Representation and Real Representation

of Lie Group
We assume that > 0. Based on a skewHermitian representation of su(2) on U ,
we can define a unitary representation of simply connected Lie group SU(2) by
considering the map exp. Hence, when > 0 is an integer, the dimension of the
representation space U is an odd number and coprime to the order of the fundamental
group of SO(3, R), which is 2. Since SU(2) is a central extension of the fundamental
group of SO(3, R) by SO(3, R), Lemma 2.2 guarantees that the unitary representation
of SO(3, R) on the representation space U is an irreducible unitary representation.
On the other hand, when is a half-integer, the skewHermitian representation
of su(2) cannot be extended to a unitary representation of SO(3, R). However, as
mentioned in the end of Sect. 2.4, it can be extended to a unitary representation of
SU(2), which is a central extension of U2 by SO(3, R). So, there exists a common
factor system for the projective representations whose highest weight is a half-integer
. Hence, when 1 , . . ., and k are odd numbers, the direct sum representation
f1 . . . fk can be extended to a projective representation of SO(3, R).
We assume that < 0. Based on a skewHermitian representation of su(1, 1)
1)
on U , we can define a unitary representation of universal covering group SU(1,
by considering the map exp. The skewHermitian representation of su(1, 1) can
be extended to a unitary representation of SU(1, 1) if and only if 2 is an integer.
Otherwise, it can be extended only to a projective unitary representation of SU(1, 1).
When 4 is an odd number, it cannot be extended to a unitary representation of
1).
SU(1, 1), but can be to a unitary representation of a double covering group SU(1,
123
Due to the same discussion, when c1 , . . ., and ck are odd numbers, the direct sum
representation fc1 /4 . . . fck /4 can be extended to a projective representation of
1).
SU(1,
Here, we consider the complex conjugate representation of an irreducible representation f . We can make the complex conjugate representation by replacing
E0,s , K,s by E0,s , K,s in (4.17)(4.19). Hence, when s = 1, the complex conjugate representation is isomorphic to f via the isomorphism |; m |; m,
i.e., it is a self-adjoint representation. When is an integer, it is underlying real
form. However, when is a half-integer, it is not underlying real form. On the other
hand, when s = 1, in the complex conjugate representation, the operator K,1 has
a non-trivial kernel instead of K+,1 . That is, while the representation space of the
representation f is generated by the highest weight vector, the representation space
of the complex conjugate representation is generated by the minimum weight vector.
Hence, f is not a self-adjoint representation.
4.2.3 Casimir Operator

Next, we calculate the Casimir operator of these representations. For this purpose,
y
we calculate the Killing form. Let x = 1, y = 2, z = 3. Based on the basis Fsx , Fs ,
y
z
x
z
Fs , we make the matrix expression of ad(Fs ), ad(Fs ), and ad(Fs ) as follows.
00 0
0 01
0 1 0
ad(Fsx ) = 0 0 1 , ad(Fsy ) = 0 0 0 , ad(Fsz ) = 1 0 0 .
0s 0
s 0 0
0 0 0
Then, the matrix expression of the Killing form is given as follows.
[(Fsa , Fsb )g ]1a,b3
2s 0 0
= 0 2s 0 .
0
0 2
Then, the Casimir operator C of f can be calculated as

1
C = (s(Fsx )2 + s(Fsy )2 + (Fsz )2 )
2
1 2
s
= (E0,s + (K+,s K,s + K,s K+,s )).
2
2
(4.20)
Hence, since C v0 = 21 (2 + )v0 , we have C = 21 ( + 1). Considering the

y
matrix-valued vector F1 := (F1x , F1 , F1z ), we have
F1 F1 = ( + 1).
(4.21)
124
When s = 1( su(2)), since there is a one-to-one correspondence between and

+ 1) under the condition > 0, the representation is uniquely determined by
the Casimir operator. However, when s = 1(su(1, 1)), there is no one-to-one correspondence between and 21 ( + 1) under the condition 1 < < 0 and = 21 .
This fact is so crucial that the negative Casimir operator cannot uniquely determine
the representation. Hence, to determine the representation, we need a more careful
discussion. In other words, in the case of su(2), we can realize the measurement
corresponding to the irreducible decomposition by measuring the physical quantity
corresponding the Casimir operator Cf . However, in the case of su(1, 1), even though
we measure the physical quantity corresponding the Casimir operator Cf , we cannot
necessarily realize the measurement corresponding to the irreducible decomposition.
That is, this method cannot distinguish two distinct irreducible components realizing
the same value as the Casimir operator.
1
(
2
4.2.4 Tensor Product

Given two quantum systems, their composite system is the tensor product space.
Hence, when two quantum systems have transformations by the same group SU(2)
or SU(1, 1), the tensor product space also has transformations by the same group. To
investigate what transformation is caused by the group in the tensor product space,
we need the irreducible decomposition of the representation on the tensor product
space. For this purpose, we assume that the representation of the generator Fas on
U1 U2 is given by Fas;1 + Fas;2 , where Fas;1 := Fas I and Fas;2 := I Fas . We
denote the tensor product |1 ; m 1 |2 ; m 2 by |1 , 2 ; m 1 , m 2 .
In the case of su(2), we have
U1 (su(2)) U2 (su(2)) =
U (su(2)),
(4.22)
|1 2 |1 +2
where takes values in {|1 2 |, |1 2 | + 1, . . . , 1 + 2 }. (4.22) can be

checked as follows. In the space U1 U2 (su(2)), the eigenspace of E0,1 with the
eigenvalue 1 + 2 is the one-dimensional space spanned by |1 , 2 , 1 , 2 , which
is the generator of U1 +2 (su(2)).
The eigenspace of E0,1 with the eigenvalue 1 + 2 1 is the two-dimensional
space spanned by |1 , 2 ; 1 1, 2 and |1 , 2 ; 1 , 2 2. Then, the vector in
the eigenspace orthogonal to K,1 |1 , 2 ; 1 , 2 is the generators of the different
representation space U1 +2 1 (su(2)). The eigenspace of E0,1 with the eigenvalue
1 + 2 2 is three-dimensional. From the eigenspace, we choose the generator of
the representation space U1 +2 2 (su(2)). We repeat this process up to the eigenvalue
|1 2 |. Then, when the eigenvalue is larger than |1 2 |, the dimension of the
eigenspace does not increase. This discussion shows (4.22).
In the case of su(1, 1), we can apply the same discussion with the order decreasing
the eigenvalues. Then, the dimension of the eigenspace increases up to infinity. Thus,
125
we obtain
U1 U2 (su(1, 1)) =
U1 +2 n (su(1, 1)).
n=0
Notice that the linear space of right hand side of (4.22) is generated by the basis
{|; m}|1 2 |1 +2 ,|m| . Then, an element |1 , 2 ; m 1 , m 2 of U1 (su(1, 1))
U2 (su(1, 1)) is expressed by using the above basis as
|1 , 2 ; m 1 , m 2 =
|; m; m|1 , 2 ; m 1 , m 2 .
,m
Since the coefficient ; m|1 , 2 ; m 1 , m 2 has the ambiguity for the phase factor, we fix the phase factors of the basis such that ; |1 , 2 ; 1 , 1 > 0.
When our algebra is restricted to su(2), the coefficient ; m|1 , 2 ; m 1 , m 2 is called
Clebsh-Gordan coefficient. The coefficient is always a real number, and the following relation is known [78, Chap. 13 and Appendix C].
; m|1 , 2 ; m 1 , m 2 = (1)1 +2 ; m|2 , 1 ; m 2 , m 1

2 + 1
1 ++m 2
= (1)
1 ; m 1 |, 2 ; m, m 2
21 + 1

2 + 1
2 ; m 2 |1 , ; m 1 , m
= (1)2 m 1
22 + 1
= (1)1 +2 ; m|1 , 2 ; m 1 , m 2 .
(4.23)
To characterize the coefficient, we focus on the relation

Fs Fs + Fs;1 Fs;1 + Fs;2 Fs;2 = 2Fs;1 Fs;2
= 2E0,1;1 E0,1;2 + K+,1;1 K,1;2 + K,1;1 K+,1;2 ,
where K,1;x and E0,1;x are defined as the same way as Fa1;x . Thus, multiplying |; m
and ; m| from both sides, we have
( + 1) 1 (1 + 1) 2 (2 + 1)

1 , 2 ; m 1 , m 2 |(2E0,1;1 E0,1;2 + K+,1;1 K,1;2 + K,1;1 K+,1;2 )
=
m 2 ,m 2
|1 , 2 ; m 1 , m 2 , m|1 , 2 ; m 1 , m 2 , m|1 , 2 ; m 1 , m 2

m 1 m 2 , m|1 , 2 ; m 1 , m 2 2
=
m2
126

(1 m 1 )(1 + m 1 + 1)(2 m 2 + 1)(2 + m 2 )
m2
, m|1 , 2 ; m 1 , m 2 , m|1 , 2 ; m 1 + 1, m 2 1

(1 m 1 + 1)(1 + m 1 )(2 m 2 )(2 + m 2 + 1)
+
m2
, m|1 , 2 ; m 1 , m 2 , m|1 , 2 ; m 1 1, m 2 + 1,
(4.24)
where m 1 and m 1 are given as m m 2 and m m 2 , respectively.

For example, it can be calculated as follows.
Example 4.1 We consider the case when 2 = 21 . We put := , m|1 , 21 , m
1 1
, and := , m|1 , 21 , m + 21 , 21 . Equation (4.24) implies that
2 2
( + 1) 1 (1 + 1)
3
4

1 2
1
1 2
= (m ) (m + ) + 2 (1 + )2 m 2 .
2
2
2
Employing the new number t :=

m
+1
t ( ) + 2 1
2
(4.25)
and the relation 2 + 2 = 1, we have

t 2
when = 1 +
1 when = 1
1
2
1
.
2
(4.26)
So, when = 1 + 21 , we have

=
1+t
=
2
1 + m + 21
, =
21 + 1
1t
=
2
1 m + 21
.
21 + 1
(4.27)
When = 1 21 , we have

=
1t
=
2

1 m + 21
1 + m + 21
1+t
, =
=
.
21 + 1
2
21 + 1
(4.28)
Example 4.2 When m = , the following relation is known [78, Appendix C].
; |1 , 2 ; m 1 , m 2 = ; |2 , 1 ; m 2 , m 1

(2 + 1)!(1 + 2 )!(1 + m 1 )!(2 + m 2 )!
= (1)1 m 1
(1 + 2 + + 1)!( + 1 2 )!( + 2 1 )!

1
,
(1 m 1 )!(2 m 2 )!
(4.29)
127
where = m 1 + m 2 . Hence, using (4.23), we can show the following relation when
m 1 = 1
; m|1 , 2 ; 1 , m 2 = (1)1 +2 ; m|1 , 2 ; 1 , m 2

(2 + 1)(21 )!( + 2 1 )!( + m)!(2 m 2 )!
,
=
( + 2 + 1 + 1)!(1 + 2 )!(1 + 2 )!( m)!(2 + m 2 )!
(4.30)
where 1 = m + m 2 .
Example 4.3 When = 1 + 2 , the following relation is known [78, Appendix C].
; m|1 , 2 ; 1 , m 2

(21 )!(22 )!( + m)!( m)!
.
=
(2)!(1 + m 1 )!(1 m 1 )!(2 + m 2 )!(2 m 2 )!
Exercise 4.4 Show (4.26) by using the relations (4.25) and 2 + 2 = 1.
4.2.5 Coherent Vector

Next, when a representation of the group SU(2) or SU(1, 1) is given on a quantum
system, we investigate the orbit of the ground state by the action of the group.
Based on discussion in Sect. 4.1, we define the coherent vector based on an element
or the unit disk D with the
of the homogeneous space, the Riemann sphere C
stabilizer H = {eiE0,s } [52].2 The coherent vector is the orbit of the ground state.
For a given , we choose C, R according to (4.5) and (4.7) in Sect. 4.1.
Then, exp( K +,s K ,s )) = g,s = g,s is an element of the Lie group SU(2) or
SU(1, 1). We obtain
,s )
U,s := f (g,s ) = f (exp( K +,s K ,s )) = exp(K+,s K
= exp(K,s ) exp(E0,s ) exp(K+,s ),
(4.31)
where U,1 is exp( 2 (K+,s K,s )). Using U,s and the highest weight vector |; ,
we define the coherent vector | : := U,s |; .
Since exp() = (1 + s||2 ) , the coherent vector | : can be written as
| : = (1 + s||2 ) exp(K,s )|;
2 When the group is SU(1, 1), this complex parametrization is different from that by Perelomov
[82]. Lisiecki [74, Sect. 4] defined the same coherent vector with the same parametrization as a
holomorphic function via the one-to-one correspondence between a square integrable function and
a holomorphic function.
128
= (1 + s||2 )

(K,s )n
n=0
n!
|; ,
where is 2 for s = 1 and is for s = 1.

is
The normalized invariant measure C on the compact homogeneous space C,
d xd y
x+i
y
written as C (d) = 2(1+||2 )2 , where = 2 . When > 0, the dimension of U

is 2 + 1, since the vector |; ; | is invariant with respect to the action of the
stabilizer H1 , the relation (3.56) yields (see Exercise 4.5)

(2 + 1)
| : : |
d xd y
= I.
2(1 + ||2 )2
(4.32)
On the other hand, the invariant measure D on the other homogeneous space D
d xd y
1
is also written as D (d) = 2(1||
2 )2 . When < 2 , since the vector |; ; |
is invariant with respect to the action of the stabilizer H1 , the relation (3.56) yields

| : : |
(2 + 1)
D
d xd y
= I,
2(1 ||2 )2
(4.33)
where the constant factor (2 + 1) will be given in Exercise 4.6. Notice that the
above integral diverges when 21 < 0. When 21 > , the formal dimension of
U is (2 + 1).
Hence, when > 0 or < 21 , the invariant measure on s is unifiedly written
d xd y
as s (d) = 2(1+s||
2 )2 and the relations (4.32) and (4.33) are unified to

s(2 + 1)
| : : |s (d) = I.
(4.34)
As long as s(2 + 1) is positive, this equation shows that the coherent states give
a measurement (POVM). Since the POVM has a good symmetry, it is useful for
information processing with a symmetry.
Exercise 4.5 Show (4.32) by using only the invariance of the LHS of (4.32) without
use of (3.56).
Exercise 4.6 Show (4.33) by using only the invariance of the LHS of (4.33).
y when the measurement correspondExercise 4.7 Calculate the outcome of = x+i
2
ing to the POVM given in (4.32) is performed to the system whose state is |; .
y when the measurement correspondExercise 4.8 Calculate the outcome of = x+i
2
ing to the POVM given in (4.33) is performed to the system whose state is |; .
4.3 Irreducible Unitary Representations of SU(r ) and U(r )
129
4.3 Irreducible Unitary Representations of SU(r) and U(r)

4.3.1 Irreducible Unitary Representation of SU(r)
The unitary group U(r ) and the special unitary group SU(r ) with a general dimension
r often appear in a quantum system. To deal with an irreducible unitary representation
of SU(r ), this section addresses a skewHermitian representation of su(r ). Since the
Lie group SU(r ) is simply connected, any skewHermitian representation of the
real Lie algebra su(r ) can be directly converted to a unitary representation of the Lie
).
group SU(r ). Hence, when it is irreducible, it can be regarded as an element of SU(r
An irreducible C-representation of the complex Lie algebra sl(r, C) can be obtained
by extending a representation of the real Lie algebra su(r ) to complex coefficients.
Then, since the complex Lie group SL(r, C) is simply connected, a C-representation
of SL(r, C) is also obtained from the C-representation discussed here.
To treat a representation of the Lie algebra su(r ), we denote the Lie subalgebra
consisting of the diagonal elements of su(r ) by sd(r ), d is the Fraktur of d. The Lie
subalgebra sd(r ) is the maximal Cartan subalgebra, which will be defined later. As
y
extensions of F1x , F1 , F1z defined in Sect. 4.2, we introduce
i
(|l j| + | jl|), 1 j < l r
2
1
:= (| jl| |l j|), 1 j < l r
2
:= i(| j j| |ll|), j = 1, . . . , r 1.
x
F j,l
:=
y
F j,l
z
F j,l
(4.35)
(4.36)
(4.37)
1
x
forms a basis of sd(r ). Adding {F j,l
, F j,l }1 j<lr to the basis of
The set {F j,z j+1 }rj=1
sd(r ), we obtain a basis of su(r ). Given an irreducible skewHermitian representation
y
x
) f(F j,l ) in the
f of su(r ) on H, we define E j := if(F j,z j+1 ) and K j,l; := if(F j,l
same way as that of su(2). Defining E j := | j j| + | j + 1 j + 1|, K j,l;+ :=|l j|,
and K j,l; :=| jl|, we can define E j :=f(E j ) and K j,l; :=f(K j,l; ) formally.
Notice that sd(r ) is the set of i times of traceless diagonal Hermitian matrices.
Given 1 j < l r , we define the element j,l of the dual space sd(r ) of sd(r )
as follows
y
j,l (X ) := i( j|X | j l|X |l).
(4.38)
Then, we have
[E j , E j ] = 0
[E j , K j,l; ] =
j,l (F jz , j +1 )K j,l;
[K j,l;+ , K j,l; ] = E j + + El1 .
(4.39)
(4.40)
130
Since the representation is skewHermitian, we find that (K j,l;+ ) = K j,l; and E j

are Hermitian. Hence, the relation (4.39) guarantees that the representation space H
can be decomposed to the direct sum of the simultaneous eigenspace with respect to
sd(r ). Hence, the simultaneous eigenvalues of elements of sd(r ) can be described
by an element m of the dual space sd(r ) of sd(r ). The simultaneous eigenspace
corresponding to m is denoted by Umw , i.e.,
Umw := {v H|f(X )v = im(X )v, X sd(r )}.
Then, an element m sd(r ) is called a weight when Umw = {0}. Using the basis
z
, . . . , Frz1,r of sd(r ), we can make a vector representation of m as [m 1 , . . . , m r 1 ]
F1,2
z
), . . . , m(Frz1,r )], which is called the Dynkin index. Then, the set
:=[m(F1,2
:= { j,l , j,l }1 j<lr is called a root system, and its element is called a
root. We abbreviate the root j, j+1 to j . The set := { j }1 jr 1 is called
the fundamental system, and its element is called simple root. An element of
+ := { j,l }1 j<lr ( := { j,l }1 j<lr ) are called a positive root (a negative
root). Then, any element of + = { j,l }1 j<lr is given as a sum of simple roots
with positive integer coefficients.
The following array of the Dynkin indexes of all of simple roots 1 , . . . , r 1 is
called Cartan matrix
2 1 0 0 0
1
1 2 1 0 0
2
3 = 0 1 2 1 0
(4.41)
.
. . . .
.
...
.. .. .. .. . . . ..
r 1
0 0 0 0 1 2
w
Due to the relation (4.40), for v Umw , K j,l; v is an element of Um
unless
j,l
K j,l; v is 0. For an arbitrary irreducible skewHermitian representation space H,
there uniquely exists a weight such that
(1) H is a one-dimensional space.

(2) An element v0 of H satisfies K j,l;+ v = 0.
Such
a weight is calledthe highest weight. This representation space is generated

n j,l
by
, and is uniquely determined by the highest weight .
1 j<lr K j,l; v0
{n j,l }
There exists a skewHermitian representation whose highest weight is if and only

if all entries of Dynkin index of are non-negative integer. This equivalence relation
holds for both representations of the Lie group SU(r ) and the Lie algebra su(r ).
As mentioned before, since SU(r ) is a universal covering group, an irreducible
skewHermitian representation f of su(r ) uniquely gives an irreducible unitary
representation of SU(r ). Hence, we regard a highest weight as an element of
), i.e., a label indicating an irreducible representation, and denote the
the set SU(r
representation space by U (SU(r )).
131
An element of su(r ) is an r r matrix, which directly gives a skewHermitian

representation on the representation space Cr . This representation is called the fundamental representation. The highest weight of this representation is [1, 0, . . . , 0].
Given a representation = [1 , . . . , r 1 ], its complex conjugate representation
is = [r 1 , . . . , 1 ]. For example, the representation with the highest weight
[1, . . . , 1] is a self-adjoint representation. The adjoint representation is a self-adjoint
representation, and has the highest weight = [1, 0, . . . , 0, 1]. The n-fold symmetric tensor representation (the n-fold alternative tensor representation) has the
highest weight [n, 0, . . . , 0, 0] ([0, 0, . . . , 0, n]). When we employ Dynkin index, we
have no possibility to confuse another group. So, we simplify U[1 ,...,r 1 ] (SU(r )) to
U[1 ,...,r 1 ] (r ). When the space Cr is written as H, we may denote it by U[1 ,...,r 1 ] (H).
When the quantum system has a representation of the group SU(d), the Hamiltonian is often given as a sum of Ei with non-negative coefficients. In this case, the
ground state, i.e., the minimum energy state, is given as the highest weight vector
|; . Hence, we are often interested in the orbit of the ground state with respect to
the representation. The subgroup H of SU(r ) that stabilizes the state |; ; |
plays an important role in the analysis of the orbit. In this case, the orbit of the ground
state is characterized not by SU(r ) but by the homogeneous space := SU(r )/H .
That is, given an element , the coherent state |, is given with its representative g as
|, := f (g )|; .
(4.42)
Further, the normalized invariant measure on the homogeneous space satisfies

d
|, , | (d) = I,
(4.43)
which can be shown by Schurs lemma (Lemma 2.4) because the left hand side is
commutative with f (g) for an arbitrary element g of SU(r ). Hence, we can construct
a measurement with symmetric property by using of coherent states. Similar to the
case with SU(2), the LHS of (4.43) gives a POVM, which describes the statistical
behavior of the measurement outcome.
4.3.2 Irreducible Unitary Representation of U(r)

Next, we deal with an irreducible representation of the Lie group U(r ). Similar to the
case of su(r ), we focus on the Lie subalgebra composed of diagonal elements of the
Lie algebra u(r ). The Lie subalgebra is given by adding the one-dimensional space
spanned by i I to sd(r ), and is written as d(r ). Then, the basis of d(r ) is given by
F jz,u := i| j j|,
j = 1, . . . , r.
(4.44)
132
Since all of elements of d(r ) are commutative with each other, we can define
the simultaneous eigenspace Umw := {v H|f(X )v = im(X )v, X d(r )}. An element m d(r ) is called the weight of the representation of u(r ) when Umw = {0}.
Using the basis F1z,u , . . . , Frz,u of d(r ), we can express the weight m by the vector form (m 1 , . . . , m r 1 ) := (m(F1z,u ), . . . , m(Frz,u )), which is called the Young
index. The entries of Young index are expressed by ( ) and those of Dynkin
index are by [ ]. We can check that the entries of Young index z,u
of the weight are
integers by considering the representation of the subgroup {eF j } isomorphic to
U(1). Now, we focus on the dual map in : d(r ) sd(r ) of the immbeding
in : sd(r ) d(r ). Then, we find that the vector m = (m 1 , . . . , m r ) d(r ) satisfies
in (m) = [m 1 m 2 , . . . , m r 1 m r ] sd(r ) .
Since a representation of U(r ) is also that of the subgroup SU(r ), similar to a
representation of su(r ), an irreducible unitary representation of the Lie group u(r )
is given by an element d(r ) such that in () sd(r ) satisfies Conditions (1)
and (2) given above, which is called the highest weight of U(r ).
Since the Lie algebra u(r ) is isomorphic to su(r ) u(1), the highest weight
d(r ) generates an irreducible skewHermitian representation of the Lie algebra
u(r ) if and only if each entry j j+1 of Dynkin index of in () is a non-negative
integer. Then, the highest weight d(r ) generates an irreducible unitary representation of U(r ) if and only if each entry j of Young index is an integer as well as
the above condition holds.
) labelIn the following, we regard the highest weight as an element of the set U(r
ing an irreducible representation, and denote its representation space by U (U(r )).
Especially, when all of j are non-negative integer, the highest weight of an
irreducible unitary representation of U(r ) is given as a Young diagram of depth r .
Each basis of the representation space U (U(r )) is given as a Semistandard Young
tableau of Young diagram . The eigenvalue of f (F jz,u ) corresponding to the respective base is given as the number of j contained in the corresponding Semistandard
Young tableau.
Next, using the contents of Sect. 2.4, we consider the relation between irreducible
unitary representations of U(r ) and SU(r ). The centralizer C(SU(r )) of the Lie group
SU(r ) is Ur . Defining := ei2/r I C(SU(r )) and using the notation in Sect. 2.4,
we find that the Lie group U(r ) is SU(r ) . That is, an element of the Lie group U(r )
can be written as (g, ei ) by using g SU(r ) and [0, 2/r ). Given an irreducible
unitary representation f of the Lie group U(r ) whose highest weight is d(r ) ,
we choose an integer l such that f () = ei2l/k I . Then, the representation f can be
written by using an integer n Z as
f (g, ei ) = fin () (g)ei(nr +l) .
Then, we have nr + l =
r
j=1
j.
(4.45)
133
4.3.3 Dimension of Irreducible Representation Space

The dimension of U (SU(r )) is the number |YU,r (n)| of Semistandard Young tableau,
which is given as
dim Un (U(r )) = |Tu,r (n)| =

1 j<lr
l j + nl n j
,
l j
(4.46)
where n j is Young index and the second formula is known as Weyls dimension
formula [37, Theorem 7.1.7]. Applying this formula to the representation of SU(r ),
we have

l j + l1
t= j t
,
(4.47)
dim U (SU(r )) =
l j
1 j<lr
where t is Dynkin index. We should remark that the eigenspace Hm corresponding
to a non-highest weight m is not one-dimensional. For example, for r = 3, we have
dim U (SU(3)) =
(1 + 1)(1 + 2 + 2)(2 + 1)
.
2
When 1 = 1 and 2 = 1, the dimension is 8 while the number of possible weights

is 7. This typical example appears in Baryon octet in Sect. 5.5.3. In particular, the
vectors 0 and 0 given in Sect. 5.5.3 have the same weight, but are orthogonal to
each other.
For r = 4, we have
dim U (SU(4))
(1 + 1)(1 + 2 + 2)(1 + 2 + 3 + 3)(2 + 1)(2 + 3 + 2)(3 + 1)
.
=
2!3!
Even for a general number r , when all entries of Dynkin index j except for the first
and second entries are zero, we can calculate the dimension as follows.
dim U[1 ,2 ,0,...,0] (SU(r )) =
(1 + 2 + r 1)!(1 + r 2)!(1 + 1)
.
(r 1)!(r 2)!(1 + 2 + 1)!2 !
(4.48)
When we employ Young index n j , this relation can be rewritten as

dim U(n 1 ,n 2 ,0,...,0,0) (U(r )) =
(n 1 + r 1)!(n 2 + r 2)!(n 1 n 2 + 1)
.
(r 1)!(r 2)!(n 1 + 1)!n 2 !
Exercise 4.9 Show the following relation
(4.49)
134
dim U[n,0,...,0] (U(r )) = dim U(n,0,...,0) (U(r )) =
(n + r 1)!
.
(r 1)!n!
(4.50)

dim U[1,1,0,...,0] (U(r )) =
(r + 1)r (r 1)
.
3
(4.51)
4.3.4 Tensor Product

Next, we consider the irreducible decomposition of the tensor product representation
of two irreducible skewHermitian representations f and f of U(r ). For simplicity,
we assume that all of entries of and are non-negative integers.
n
The multiplicity coefficient C,
(U(r )) given in (2.45) satisfies the relation
U (U(r )) U (U(r )) =
CC, (U(r )) Un (U(r )),

n
(4.52)
nYnr
and is called the LittlewoodRichardson coefficient, where n := || + | |. The

coefficient can be calculated from the relation among three Young diagrams , ,
and n as follows. Given n and , we define the skew-diagram n\ as a diagram given
as the Young diagram n with substraction of the Young diagram . For example, the
skew-diagram n\ is defined as the diagram
substraction of
that is
with
. Then, we insert the letters 1, . . . , r in the skew-diagram n\
n
according the following rule. The LittlewoodRichardson coefficient C,
(U(r )) is
given as the number of possible types of the insertions of the letters 1, . . . , r . This
fact is called the LittlewoodRichardson rule [75, p.68].
(1) The letter j is written in j boxes among the skew-diagram n\.

(2) The letter is strictly increasing from up to down in each column of the skewdiagram n\.
(3) The letter is increasing from left to right in each low of the skew-diagram n\.
(4) We make the order for the boxes in the skew-diagram n\ as follows. The first
to n 1 1 -th boxes are given in the first low from left to right. (This order is
opposite to Condition (3).) Next, we make the order of the boxes in the second
low in the same way. We repeat this numbering method up to the |n| ||-th
box. Then, for an arbitrary integer 0 m |n| ||, the number of boxes with
j is larger than that with j + 1 among the first to m-th boxes.
135
Example 4.4 Assume that = (1, 0, . . . , 0). The LittlewoodRichardson coeffin

r
satisfies the following two conditions. It is 0
cient C,
(U(r )) is 1 when n Y
otherwise.
(1) Only one entry n i of n satisfies that n i = i + 1.
(2) The remaining entries n j of n satisfy that n j = j .
Example 4.5 Assume that is (1, . . . , 1, 0). The LittlewoodRichardson coeffin
r
satisfies the following two conditions. It is 0
cient C,
(U(r )) is 1 when n Y
otherwise.
(1) Only one entry n i of n satisfies that n i = i .
(2) The remaining entries n j of n satisfy that n j = j + 1.
Examples 4.4 and 4.5 yield that
U(1,0,...,0) U(0,...,0,1) = U(0,...,0) U(2,1,...,1,0) .
(4.53)
When we employ Dynkin index, we have

U[1,0,...,0] (U(r )) U[0,...,0,1] (U(r )) = U[1,0,...,0,1] (U(r )) U[0,...,0] (U(r )).
(4.54)
Example 4.6 Assume that =(m, 0, . . . , 0) and =(n, 0, . . . , 0) (m n). The

n
LittlewoodRichardson coefficient C,
(U(r )) is 1 when n=(n + m k, k, 0, . . . ,
0) and k=0, 1, . . ., or n. Otherwise, it is 0. That is,
U(m,0,...,0) (U(r )) U(n,0,...,0) (U(r )) =
n

(U(m+nk,k,0,...,0) (U(r )).
(4.55)
k=0
Example 4.7 Assume that is the adjoint representation (2, 1, . . . , 1, 0). We have
the following three cases.
(1) When all entries n i of n satisfies that n i = i + 1, the LittlewoodRichardson
n
coefficient C,
(U(r )) is the number of i satisfying the condition n i > n i+1 .
(2) Assume that one entry n i of n satisfies that n i = i + 2, another entry n j of n
satisfies that n j = j , and the remaining entries n l of n satisfy that n l = l + 1.
n
Then, the LittlewoodRichardson coefficient C,
(U(r )) is 1.
n
(3) Otherwise, the LittlewoodRichardson coefficient C,
(U(r )) is 0.
Example 4.8 Assume that = (m, 0, . . . , 0), =(n, 0, . . . , 0), =(1, 0, . . . , 0),
and m n. In the following, we omit U(r ).
(1) When r = 2 and m = n, we have
U(n,0) U(n,0) U(1,0)
= U(2n+1,0)
n

k=1
(U(2n+1k,k) C2 ).
(4.56)
136
(2) When r = 2 and m > n, we have

U(m,0) U(n,0) U(1,0)
= U(m+n+1,0) U(m,n+1)
n

(U(m+n+1k,k) C2 ).
k=1
(3) When r 3 and m = n, we have

U(n,0,...,0) U(n,0,...,0) U(1,0,...,0)
= U(2n+1,0,...,0)
n

(U(2n+1k,k,0,...,0) C )
2
k=1
n

U(2nk,k,1,...,0) .
(4.57)
k=1
(4) When r 3 and m > n, we have

U(m,0,...,0) U(n,0,...,0) U(1,0,...,0)
= U(m+n+1,0,...,0) U(m,n+1,...,0)
n

(U(m+n+1k,k,0,...,0) C2 )
k=1
n

U(m+nk,k,1,0,...,0) .
k=1

dim U[1,0,...,0,1] (U(r )) = r 2 1.
(4.58)
4.3.5 Irreducible Representation of SL(r, C), GL(r, C)

An irreducible unitary representation of SU(r ) can be extended to an irreducible Crepresentation of SL(r, C), and an irreducible unitary representation of U(r ) can be
extended to an irreducible C-representation of GL(r, C). In this case, a non-unitary
element of GL(r, C) does not directly give a physical transformation, however, the
following discussion plays an important role for analysis of a density matrix. For
simplicity, we consider the case with r = 2. So, Young index n can be written as
, nk
) with an integer k between n and n/2 so that n k is an even number.
( n+k
2
2
nk (g) of the repreAssume that g has the eigenvalues a and b. The character n+k
2 , 2
sentation fn is given as
nk (g) =
n+k
2 , 2
k/2

m=k/2
a 2 +m b 2 m =
n
nk
2
n+k
2
a 2 +1 b
1 ab
n+k
nk
2
137
In particular, when a = 1 p and b = p with p (1/2, 1),

nk (g) =
n+k
2 , 2
n+k
2 +1
(1 p)
nk
2
nk
p 2 (1 p)
(2 p 1)
n+k
2 +1
(4.59)
Next, we consider the case with GL(r, C). When an element g GL(r, C) is diagonalizable, and its eigenvalues a1 , . . . , ar are distinct with each other, by using
:= (r 1, r 2, . . . , 0), the character of the representation fn is given as the
Weyls dimension formula [37, Theorem 7.1.1].

n (g) =
Sr

n +
sgn() rl=1 al (l) (l)

.
j<l (a j al )
(4.60)
The right hand side is a function of the eigenvalues a1 , . . . , ar , and is called the
Schur function n (a1 , . . . , ar ).
4.4 Many-Particle System: Composite System

of n Quantum Systems
4.4.1 Distinguishable Particles: Schur Duality
Given an element g of the Lie group SU(r ) or U(r ), we consider its n-fold tensor
product g n := g g , which gives a unitary representation of SU(r ) on the

! "
n
composite system (Cr )n . This unitary representation is called the n-fold tensor
product representation of SU(r ), and describes the situation that n quantum systems
(n particles) with the representation space Cr are independently prepared and the
same transformation is applied on all of them. Further, this description assumes that
these systems or these particles are distinguishable. When such a transformation is
applied on n quantum systems, this representation is helpful. However, even when
such a transformation is not given, this representation is still helpful for the composite
system (Cr )n due to the following symmetry.
In the previous section, given n Y r 1 , we gave the irreducible unitary representation Un (U(r )) of U(r ) and the irreducible unitary representation Uin (n) (U(r )) of
SU(r ). On the other hand, we consider the following unitary representation fn of the
permutation group Sn on (Cr )n .
fn (g)(v1 vn ) = vg(1) vg(n) , g Sn .
(4.61)
138
Hence, combining these two unitary representations, we obtain the unitary representation of the direct product group SU(r ) Sn or U(r ) Sn on (Cr )n , which is
called Schur duality.
of the weight of E0,+1 .
When r = 2, we rearrange 2n basis of (C2 )n dependently
# $
In this case, the number of basis with weight m is mn . When we additionally assume
that n = 4, we can classify the basis as the following figure, where m is the weight
and is the highest weight. In this figure, the action of su(2) moves the basis in the
vertical direction. K+,+1 moves the basis upward, and K,+1 does them downward.
On the other hand, the permutation group group S4 moves the basis in the horizontal
direction. Hence, the vertical arrows indicate irreducible representation spaces of
su(2), and the horizontal arrows do irreducible representation spaces of S4 . That is,
irreducible representation spaces can be classified by (Fig. 4.3).
Now, we concretely construct each representation space as follows. Let | be the
vector of weight m = 1/2 in C2 , and | be the vector of weight m = 1/2. Then, the
leftmost representation space of su(2) (i.e., the representation space with the highest
weight = 2) is a symmetric tensor product space, the vector of weight m = 1 in
this space is | , , , + | , , , + | , , , + | , , , . Hence, the
vector of weight m = 1 in the representation space with the highest weight = 1
is the vector of weight m = 1 orthogonal to the above vector. The vector of weight
m = 0 in the representation space with the highest weight = 1 is given by applying
K,+1 to this vector. The representation space with the highest weight = 0 contains
only the vector of weight m = 0. This vector is the vector of weight m = 0 orthogonal
to the vector of weight m = 0 in the representation space with the highest weight
= 2 and the vector of weight m = 0 in the representation space with the highest
weight = 1. For a general integer n, we can make the basis of each representation
in the same way.
For a general integer r , we can make the following irreducible decomposition
of (Cr )n with respect to the unitary representation of the direct product group
SU(r ) Sn .
Fig. 4.3 Irreducible

decomposition of (C2 )4
=2
m =2
m =1
m=0
m = 1
m = 2
=1
=0
Representation of
Representation of
S4
(2)
4.4 Many-Particle System: Composite System of n Quantum Systems
(Cr )n =
139
Uin (n) (SU(r )) Un (Sn )
(4.62)
Un (SU(r ) Sn ).
(4.63)
nYnr
nYnr
Then, given n Ynr , the space Un (SU(r ) Sn ) is given as the irreducible representation space Uin (n) (SU(r )) Un (Sn ) of the direct product group SU(r ) Sn . In the
irreducible space, the permutation group Sn acts only on Un (Sn ), and its action on
Uin (n) (SU(r )) is given as the identity matrix. Similarly, the Lie group SU(r ) acts only
on the space Uin (n) (SU(r )), and its action on the space Un (Sn ) is given as the identity
matrix. Hence, in the irreducible space, the multiplicity of the unitary representation
Un (Sn ) of Sn is the dimension of Uin (n) (SU(r )), and the multiplicity of the unitary
representation Uin (n) (SU(r )) of SU(r ) is the dimension of Un (Sn ). These dimensions are given in (4.47) and (2.72). Especially, when n = (n, 0, . . . , 0), Un (Sn ) is a
one-dimensional space, and the subspace Uin (n) (SU(r )) Un (Sn )
= Uin (n) (SU(r ))
of (Cr )n is the n-fold symmetric tensor product space.
The tensor product representation can be naturally extended to a C-representation
of GL(r, C). In this case, the relation (4.62) also gives the irreducible decomposition
with respect to the direct product group GL(r, C) Sn of (Cr )n (Table 4.1).
When we denote the projection to irreducible component corresponding to n by
Pn , the irreducible decomposition (4.63) gives the measurement {Pn }. When the
densities of all systems are given as the same density matrix and these systems
are independent of each other, the density matrix of the composite system is given
as n . When does not have eigenvalue 0, it can be regarded as an element of
GL(r, C). Hence, when the above measurement is applied, the probability to obtain
the outcome n is
Tr Pn n = Tr fn () Tr Un (Sn ).
(4.64)
Since Tr fn () is the character of , its calculation formula is available in (4.60). On the

other hand, the calculation formula of Tr Un (Sn ) is given in (2.72). Several formulas
given in group representation theory are useful for the calculations of physically
important values in this way.
4.4.2 Indistinguishable Particles: Boson and Fermion

In Sect. 1.1, we have explained that the composite system is given as Hn when
every particle is distinguishable from each other and its representation space is H.
However, even when the quantum system of the internal degree is H, when the
particles are in the same place as another particle, they cannot be distinguished from
each other. In this case, the composite system is not Hn . Instead of this description,
we have other two types of descriptions as follows. One is the case when more than
140
Table 4.1 Young diagram and representation for r n

Name
Young
Young index Dynkin index
diagram
of U (su(r ))
of U (su(r ))
Fundamental
representation
(1, 0, . . . , 0)
2-fold
symmetric
tensor product
representation
(1, 1, 0, . . . , 0) [0, 1, 0, . . . , 0]
"
n-fold
symmetric
tensor product
representation
n-fold
alternative
tensor product
representation
Adjoint
representation
(only when
n = r 1, no
name in other
cases.)
[1, 0, . . . , 0]
..
..
.
(n, 0, . . . , 0)
! "
(1, . . . , 1,
0, . . . , 0)
n1
! "
n 1 (2, 1, . . . , 1,
0, . . . , 0)
[n, 0, . . . , 0]
n1
! "
[0, . . . , 0,
1, 0, . . . , 0]
Dimension of
U (su(r ))
Dimension of
U (Sn )
r (r 1)
2
#n+r 1$
r 1
#r $
n
n2
! "
[1, 0, . . . , 0,
1, 0, . . . , 0]
n2 1
n 1
one particles can simultaneously take the same state. The other is the case when
more than one particles cannot. The choice depends on the type of the particle. The
particle choosing the former is called a boson, and the particle choosing the latter
is called a fermion. A typical example of a boson is a photon and examples of a
fermion particle are an electron and a quark.
When there are n boson particles in the quantum system H, the representation
system of the composite system is the n-fold symmetric tensor product representation
of H. As Fig. 4.4, the space is the subspace of vectors invariant for the permutation
in the n-fold tensor product space Hn , which has the representation of SU(r ) Sn .
That is, the quantum system is given as the irreducible space U[n,0,...,0] (r ) when
H = Cr .
As a typical example, we consider the case when H = Cr C2 . the twodimensional system C2 is called spin and is spanned by {| , | }.Hence, t
Its physical meaning is explained in Sect. 5.4. Then, the symmetric subspace
141
Fig. 4.4 boson and fermion
U[3,0,...,0] (2r ) is irreducible for SU(2r ), but is not necessarily irreducible for the
subgroup SU(r ) SU(2) SU(2r ) as follows. To discuss this issue, we define the
normalized symmetrized vector of |k1 s1 , k2 s2 , k3 s3 with respect to the permutation
by S(|k1 s1 , k2 s2 , k3 s3 ).
Theorem 4.1 The space U[3,0,...,0] (2r ) is irreducibly decomposed for the representation of the subgroup SU(r ) SU(2).
U[3,0,...,0] (2r )
= U[3,0,...,0] (r ) U[3] (2) U[1,1,0,...,0] (r ) U[1] (2).
(4.65)
The space U[3,0,...,0] (r ) U[3] (2) can be generated by the element |1, 1, 1, , ,
with the action of SU(r )
) U[1,1,0,...,0] (r ) U[1] (2) can
)SU(2). The other space
be generated by the element
action of SU(r ) SU(2).
2
S(|2
3
, 1 , 1 )
1
S(|2
3
, 1 , 1 ) with the
Proof Any element of the subspace U[3,0,...,0] (r ) U[3] (2) is permutation-symmetric.

So, the subspace is included in U[3,0,...,0] (2r ). Next, we focus on the subspace spanned
by S(|2 , 1 , 1 )
) The subspace U[3,0,...,0] (r ) U[3] (2)
) and S(|2 , 1 , 1 ).
, 1 , 1 ) + 23 S(|2 , 1 , 1 ). Under the action

)
)
of SU(r ) SU(2), the orthogonal vector 23 S(|2 , 1 , 1 ) 13 S(|2 , 1
, 1 ) generates the space U[1,1,0,...,0] (r ) U[1] (2). Hence, we obtain the relation
in (4.65). Exercises 4.9 and 4.10 guarantee that the dimension of LHS of (4.65)
+1)
+1)r
and that the dimension of RHS of (4.65) is 4 (r +2)(r
+2
is 2r (r +1)(2r
3
6
(r +1)r (r 1)
2r (r +1)(2r +1)
=
.
Hence,
we
obtain
the
equality
of
(4.65).

3
3
contains the vector
1
S(|2
3
On the other hand, when there are n fermionic particles in the quantum system H, the representation system of the composite system is the n-fold alternative tensor product representation of H. That is, the quantum system is given as
U[0, . . . , 0,1,0,...,0] (r ). Only the sign is changed under the action of permutation in this

! "
n1
space. So, a density matrix in the space is invariant for the action of permutation.
Therefore, to discuss bosons or fermions, we need to discuss the special unitary
group SU(r ) in the above way.
As a typical example, we consider the case when H = Cr Cn , which is spanned
by {|k, l} with k = 1, . . . , r and l = 1, . . . , n. Then, the n-fold anti-symmetric subspace U[0, . . . , 0,1,0,...,0] (nr ) is irreducible for SU(nr ), but is not necessarily irreducible

! "
n1
142
for the subgroup SU(r ) SU(n) SU(nr ) as follows. Let | be the vector spanning the one-dimensional space U[0,...,0] (n) of (Cn )n . Then, || can be regarded
as a projection in Hn .
Theorem 4.2 The n-fold anti-symmetric space U[0, . . . , 0,1,0,...,0] (H) satisfies

! "
n1
||U[0, . . . , 0,1,0,...,0] (H) = U[n,0,...,0] (r ) U[0, . . . , 0] (n),

! "
n1
! "
(4.66)
n1
which is an irreducible subspace of the subgroup SU(r ) SU(n).

This theorem shows that the subspace of the anti-symmetric space can be regarded
as the symmetric space of the smaller system. This statement plays an important role
in Sect. 5.5.
Proof Any vector in ||U[0, . . . , 0,1,0,...,0] (H) has a from || with an element of

! "
n1
r n
(C ) . Since any permutation g satisfies f(g)| = sgn(g)|, the other vector |

is invariant for any permutation. Hence, | belongs to the h-fold symmetric space

U[n,0,...,0] (r ).
To discuss another typical example, we define the spaces U[1,0,...,0] (r ), U[0,...,0,1] (r ),
dk=1 ,
rq=1 , {|k}dk=1 , and {|k}
U[1,0,...,0] (d), and U[0,...,0,1] (d) spanned by {|q}rq=1 , {|q}
respectively. Then, we focus on the 2r d-dimensional space H := U[1,0,...,0] (r )
U[1,0,...,0] (d) U[0,...,0,1] (r ) U[0,...,0,1] (d). Hence, we can naturally define the representation
of the subgroup SU(r ) SU(d) SU(2r d) on H. The vector | :=
d ) 1
|k,
k) spans the one-dimensional space U[0,...,0] (d) of U[1,0,...,0]
(|k, k
k=1
2d
(d) U[0,...,0,1] (d). Hence, | | can be regarded as a projection in H2 .

Theorem 4.3 The two-fold alternative tensor product space U[0,1,0,...,0] (H) satisfies
| |U[0,1,0,...,0] (H)
= U[1,0,...,0] (r ) U[0,...,0,1] (r ) U[0,...,0] (d),
(4.67)
which is closed with respect to the subgroup SU(r ) SU(d) SU(2r d).
Proof Any vector in H2 can be written as a liner combination of the following
q , k |q, k, q , k , |q,
q , k , and |q,
q , k . The
vectors; |q, k, q , k , |q,
k,
k,
k,

q , k vanishes by the application of the projection
k,
vectors |q, k, q , k and |q,
| |. We also have
= | ||q,
q , k
k
k,
| ||q , k , q,
= k,k (|q , q
+ |q,
q )
d

k=1
1
|k,
k).
(|k, k
2d
143
Hence, the space | )

|U[0,1,0,...,0] (H) is given as linear combination of the vectors
d

|k,
k). We can make one-to-one correspon + |q,
q ) k=1 1 (|k, k
(|q , q
2d
dence the basis and the basis of U[1,0,...,0] (r ) U[0,...,0,1] (r ). So, we obtain (4.67).

Exercise 4.12 Show
U[0,0,...,3] (2r )
= U[0,0,...,3] (r ) U[3] (2) U[0,...,0,1,1] (r ) U[1] (2).
(4.68)
4.5 Discretization of Lie Group and Homogeneous Space

4.5.1 Design on Group
Section 3.7 deals with the invariant measure G on the Lie group G. Although many
quantum information processings employ the invariant measure G , it is not easy to
realize such quantum information processings because the support of G is so large.
Hence, to resolve this problem, instead of the measure G , we often employ another
measure whose support is much smaller than G. To discuss how another measure
works as an alternative of the measure G under a quantum information processing,
we need to clarify the required property of for the quantum information processing.
If we need to realize the completely same performance in any situation, we cannot
replace the invariant measure G by another measure. The invariant measure G can
be used in various purposes, however, it is used for a specific purpose in practice.
Hence, for the specific purpose, it is sufficient to employ a measure that satisfies a part
of properties of the invariant measure G . Thus, to decide whether another measure
realizes the same performance under the specific purpose as the invariant measure
G , we need to clarify the required property in terms of the group representation.
However, in the current situation, we have not discussed its concrete application,
we give only a mathematical definition of the required property for the measure
while the usefulness of the definition is discussed in [20, 27] and [50, Chap. 4]. Since
the invariant measure does not necessarily exist for a non-compact Lie group G, we
discuss a Lie group G whose invariant measure exists.
Now, we prepare the following theorem.
Theorem 4.4 The following conditions are equivalent to each other for a measure
on G and a representation f of G on H, where we assume that any irreducible
component of H has the formal dimension when the representation space H is infinitedimensional.
*
*
(1) G f(g) f(g) G (dg) = G f(g) f(g) (dg).
144
*
*
(2) Any matrix X on H satisfies G f(g)X f(g) G (dg)= G f(g)X f(g) (dg). (When
the representation space H is infinite-dimensional, any trace class operator X
satisfies the above equation).
*
*
(3) Any Hermitian matrix X satisfies G f(g)X f(g) G (dg) = G f(g)X f(g) (dg).
(When the representation space H is infinite-dimensional, any trace class selfadjoint operator X satisfies the above equation).
Proof (2)(3): Any matrix X can be written as X 1 + X 2 i by using two Hermitian
matrices X 1 and X 2 . Hence, Conditions (2) and (3) are equivalent to each other.
(1)(2): Assume that v1 , v2 , v3 , and v4 are arbitrary vectors in U . Considering the
matrix |v1 v2 | |v3 v4 |, we have

f(g) f(g) G (dg)|v1 v2 | |v3 v4 |
Tr
G

= v2 |f(g)|v1 v4 |f(g) |v3 G (dg)
G

f(g)|v1 v4 |f(g) G (dg)|v3 .
(4.69)
=v2 | Tr
G
Hence,

v2 | f(g)|v1 v4 |f(g) G (dg)|v3 = v2 | f(g)|v1 v4 |f(g) (dg)|v3 . (4.70)
G
Since v3 and v2 are arbitrary, we have

f(g)|v1 v4 |f(g) G (dg) =
f(g)|v1 v4 |f(g) (dg).
(4.71)
Since an arbitrary matrix X can be written as a linear combination of matrices with

the form |v1 v4 |, we obtain Condition (2).
(2)(1): Assume that v1 , v2 , v3 , and v4 are arbitrary vector in U . To show (1),
it is sufficient to show that the traces of the products with the matrix |v1 v2 |
|v3 v4 | take the same value in both side in the equation in (1). Condition (2) yields
(4.71). Hence, we have (4.70). Thus, using (4.69), we find that the traces of the
products with the matrix |v1 v2 | |v3 v4 | take the same value in both side in the
equation in (1).

When the measure on G satisfies the condition of this theorem under the representation f of G, the measure on G is called an f-design. Using Condition (1), we
obtain the following lemma.
Lemma 4.1 For a measure on the group G and an element g of G, we define
the measure g on the group G as g (B) := (g B). The following conditions are
equivalent to each other.
(1) is an f-design.
145
(2) g is an f-design.
The discretization of the invariant measure G can be realized by choosing a
measure whose support is a finite subset G of G. So, we focus on a finite subset
G of G. In this case, we define the measure G as the |G1 | -times of the counting
measure on G . Then, when the measure G is an f-design G is called an f-design.
We abbreviate an f -design to a -design. When a subset G G is a -design for
the subset G is called an S-design.
any element of a subset S G,
Lemma 4.2 When a probability measure is an S design, the relation

G
(g) (g)(dg) = ,
,
holds for , S. Then, when S and { G|C
= 0} S, the relation

(g) (g) (g)(dg) = C,

holds.
Lemma 4.3 Assume that G is a compact Lie group and f is its finite-dimensional
representation. A finite subgroup G of G is an f-design if and only if the irreducible
decomposition of the restriction to G of the representation f is the irreducible decomposition of the representation f of G
Proof Let K be an arbitrary irreducible non-trivial subspace of the restriction to G
of the representation f. For a vector u K, we have

G
f (g)|uu|f (g) G (dg) =
u
2
IK ,
dK
(4.72)
where dK is the dimension of K. Let K be the irreducible space with respect to the
representation of G including u, and dK be its dimension. Then, we have

f (g)|uu|f (g) G (dg) =
G
u
2
I K .
dK
(4.73)
Hence, when G is an f-design, K and K are the same space. So, the irreducible
decomposition of the restriction to G of the representation f is the irreducible decomposition of the representation f of G.
Conversely, we assume that any irreducible component of f is still irreducible for
the restriction to G . Since any Hermitian matrix is written with a linear combination
of matrices |uu|, it is sufficient to show that any vector u H and any matrix |uu|
satisfy
146

f (g)|uu|f (g) G (dg) =
G
f (g)|uu|f (g) (dg).
(4.74)
Now,we decompose the vector u into vectors of irreducible representation spaces

as u = j u j . Let K j be the irreducible space containing u j and d j be its dimension.
Then,

f (g)|uu|f (g) G (dg) =
G

u j
2
j
dj
IK j .
Since the right hand side depends only on the irreducible decomposition, we obtain
(4.74).

Notice that the map X f(g)X f(g) is a completely positive map on H for a
representation f on a d-dimensional space H. Since the dimension of the convex set
of completely positives map on H has the dimension d 4 d 2 , applying the following
Caratheodorys theorem (Theorem 4.5) to this convex set, we find that there exists
a probability distribution p such that its support has cardinality d 4 d 2 + 1 and p
is an f-design. A similar conclusion has been be shown by using a different method
[53].
Theorem 4.5 (Caratheodorys Theorem) [92, Sect. 17] Assume that X is an mdimensional convex set and the convex set X is given as the convex closure of the
set Ext(X ) of its extremal points. For an element x X , there exist m + 1 extremal
xm+1 Ext(X ) and a probability distribution p on {x1 , . . . , xm+1 }
points x1 , . . . ,
m+1
p(i)xi .
such that x = i=1
Generalizing the proof of Lemma 4.3, we can show the following corollary.
Corollary 4.1 Assume that G is a compact Lie group and f is its finite-dimensional
representation. We also assume that f is a map from a finite group G to a Lie group
G and f f is a projective representation. Then, a subset f(G ) of G is an f-design if
and only if the irreducible decomposition of the projective representation f f of G
is the irreducible decomposition of the representation f of G.
Next, we concentrate the case when G is SU(d). A measure on G or a subgroup
G of G is called a (t1 , t2 )-design when it is an ft1 ,t2 -design for the representation
ft1 ,t2 (g) := g t1 (g t2 ) . Then, a measure on SU(d) is a (t1 , t2 )-design if and only
if any matrices X 1 , . . . , X 2t1 +22 on Cd satisfy

Tr X 1 g X 2 g X t1 g X t1 +1 g X t1 +t2 g X t1 +t2 +1 g
G
X 2t1 +t2 g X 2t1 +t2 +1 g X 2t1 +2t2 gG (dg)
Tr X 1 g X 2 g X t1 g X t1 +1 g X t1 +t2 g X t1 +t2 +1 g
G
X 2t1 +t2 g X 2t1 +t2 +1 g X 2t1 +2t2 g(dg).
147
This fact can be shown in the same way as Theorem 4.4 by replacing X i by a linear
sum of matrices |u i vi |. Hence, a measure on SU(d) is a (t1 , t2 )-design if and only
if it is a (t1 + t2 , 0)-design. Since it is enough to discuss a (t, 0)-design due to this
relation, we simplify it to a t-design.3 When we need to emphasize the design for set
of unitaries, we call it a unitary t-design. The properties were discussed in details
in [21, 24, 39, 45, 94, 96]. Recently, it was shown that Clifford group is a 3-design
when the dimension is a power of 2 (including 2) [106, 117]. (For the definition of
Clifford group, see Sect. 8.2). Deep connection with discrete Wigner functions was
revealed in [117, 250].
By using Dynkin index, the representation f1,1 can be decomposed into the
direct sum representation of the d 2 1-dimensional representation with the highest weight [1, 0, . . . , 0, 1] and the one-dimensional representation with the highest
weight [0, . . . , 0]. On the other hand, the representation f2,0 can be decomposed into
the direct sum representation of the d(d + 1)/2-dimensional representation with the
highest weight [2, 0, . . . , 0] and the d(d 1)/2-dimensional representation with the
highest weight [1, 1, 0, . . . , 0, 0]. Hence, a measure on SU(d) is a [1, 0, . . . , 0, 1]design if and only if it is a [2, 0, . . . , 0]-design and is a [1, 1, 0, . . . , 0]-design. That
is, for a subgroup G of SU(d) or U(d), the restriction to G of the representation
f[1,0,...,0,1] is still irreducible if and only if both restrictions to G of the representations
f[2,0,...,0] and f[1,1,0,...,0] are still irreducible.
4.5.2 Design on Homogeneous Space

The previous subsection is the discretization of the group G. We can consider the
discretization of the homogeneous space in a similar way as follows. In this case,
we employ the invariant measure of instead of a measure of . A measure
of is called a -design for G when the irreducible representation f of the
Lie group G satisfies

d
| : : |(d) = I .
(4.75)
As mentioned in Sect. 4.3.1, | : : | gives a POVM. In the above definition,

the existence of the formal dimension is assumed when the representation f is an
infinite-dimensional representation. Especially, a finite subset of is called a
-design when its counting measure is a -design.
For example, when a given finite subgroup G of the Lie group G is a -design,
the orbit {g0 |g G } of 0 with respect to the action of G is a -design as
a subset of the homogeneous space . Let be the set of coherent vectors in the
fundamental representation of SU(d), i.e., the set of pure states on H = Cd
3 Originally,
the name of t-design was given in [23].
148
When a measure on satisfies

t1
t2
|uu| |uu| (d) =
|uu|t1 |uu|t2 (d),
(4.76)
the measure is called a (t1 , t2 )-design. Additionally, a finite subset of is called

a (t1 , t2 )-design when its counting measure is a -design.
When G is a compact group, the dimension d of the representation space of the
representation f is finite. Then, (6.40) guarantees that the matrix I is contained in
the convex set generated by {| : : |} , whose dimension is d2 1 at most.
Hence, applying Theorem 4.5 to the pair of the convex set and the matrix I , we find
that there exists a probability distribution p on such that the cardinality of the
support is d2 and it is a -design.
Then, letting be the operation to take transpose only in the space (Cd )t2
at (4.76), we have (|uu|t1 |uu|t2 ) = |uu|t1 +t2 . Since the condition for
(t1 , t2 )-design and that for (t1 + t2 , 0)-design can be converted to each other via the
map , this fact yields the following theorem.
Theorem 4.6 A measure on is a (t1 , t2 )-design if and only if it is a (t1 + t2 , 0)design.
Moreover, the following theorem is known [16, 57, 73, 72, 95].
Theorem 4.7 When a measure on is a (t, 0)-design, the relation | supp()|
#d+t/2 1$2
holds. When the equality holds, the relation (|u) = 1/| supp()| holds
t/2
for any vector |u supp(), i.e., is a distribution on supp() with equal weight.
Since it is enough to discuss a (t, 0)-design due to this relation, we simplify it to
a t-design. Sections 8.3 and 8.4 give several (2, 0)-designs. A more efficient (2, 0)design was given in [21] when the dimension is a power of 2. In this case, since
Clifford group is a unitary 3-design [106, 117], the set of stabilizer states forms a
3-design as a set of pure states [69, 117]. (For the definition of stabilizer state, see [50,
Sect. 3.3]). Also, t-design is related to several topics. For example, the relation with
spherical codes is discussed in [15, 26]. The connection with SIC vectors defined in
Sect. 8.4 is discussed in [7, 90, 89, 95]. Since t-design guarantees the symmetry, it
can be applied to quantum state estimation [53, 95, 121, 114, 116], quantum state
discrimination [2, 76], and detecting maximally entangled state [47]. Further, using
t-design, we can construct phase retrieval with the phase lift algorithm [40]. Also,
we can discuss the designs in projective spaces [56].
4.5.3 Approximating Design

In general, it is not easy to find an f-design for a representation f. Even though we could
find an f-design, its structure might be too complicated for a practical use. Hence,
149
as an alternative idea, we often seek to approximately realize it. As measures of the

quality of the approximation, we introduce two criteria, one is based on the matrix
(operator) norm, and the other is based on the trace norm. Given a representation f
of G, a measure on G is called a trace norm approximate f-design when

+
+
+
+
f(g) f(g) (dg)+ .
+ f(g) f(g) G (dg)
G
Tr
Similarly, it is called a matrix (operator) norm approximate f-design when

+
+
+
+
f(g) f(g) (dg)+ .

+ f(g) f(g) G (dg)
G
A measure on is called a trace norm approximate -design when the following

relation (4.77) holds, and it is called a matrix (operator) norm approximate -design
when the following relation (4.78) holds.
+
+
+
+
| : : |(d) I + ,
+d
Tr

+
+
+
+
| : : |(d) I + .
+d
(4.77)
(4.78)
Chapter 5
Application to Physical Systems
Abstract We apply representation theory to several real quantum systems. A real

microscopic system has the symmetry of three-dimensional rotation as well as the
classical macroscopic system. We firstly discuss this rotation symmetry on the function space L 2 (R3 ) by using the algebra su(2). Since this symmetry is related to the
three-dimensional rotation, it is called orbital angular momentum. Then, we discuss
the Schrdinger equation with central potential. As a typical example, we address
the case of hydrogen-type atom and explain the degeneracy of Hamiltonian in this
case by using SO(4, R). We also extend the discussion of the rotation symmetry to
the d-dimensional case. Then, we introduce the spin system and discuss an interaction between the spin system and the orbital angular momentum system. Also,
we discuss quarks that constitutes elements of atoms based on several groups. Since
a quark is a fermion and our discussion is limited to the cases with two or three
particles, the discussion for quarks in this chapter is much simpler than that of the
photonic system because we need to discuss unlimited number of bosons in the photonic system. Finally, we discuss uncertainty relation for wake packets on various
spaces, i.e., not only on the set of real numbers R but also on the one-dimensional
and three-dimensional spheres S 1 and S 3 .
5.1 3-Dimensional Physical System

5.1.1 Orbital Angular Momentum
This chapter discusses spectral decompositions of physically important operators.
As a typical example of the group representation, we focus on the physical system
L 2 (R3 ) for a particle. For this analysis, we focus on two self-adjoint operators Qk
and P k on L 2 (R3 ) defined as1
k when = 1. For the precise

this book, Pk expresses the operator i x k , which equals P
definition of the operators in this chapter, see Sect. 1.6.
1 In

DOI 10.1007/978-3-319-44906-7_5
151
152
5 Application to Physical Systems
Qk f (x) := xk f (x), P k f (x) := i

f (x)
xk
(5.1)
for x R3 , where is the Planck constant. Here, Qk is called the position operator of
the k-th coordinate, and Pk is called the momentum operator of the k-th coordinate.
These operators satisfy the commutation relation as
[Ql , P k ] = il,k I.
(5.2)
Then, we define the orbital angular momentum operator L t on L 2 (R3 ) as

L t :=
t,k,l Qk P l
(5.3)
k,l
for x R3 . Due to the commutation relation (5.2), the self-adjoint operator L k

satisfies the commutation relation

t,k,l L l .
(5.4)
[L t , L k ] = i
l
Hence, i
L satisfies the commutation relation (4.11) for su(2), which is equivalent
k
to so(3). This is natural because the operator L k generates the rotation with k-axis
in L 2 (R3 ).
We also define the operators L and the total orbital angular momentum L 2 as
L := L 1 i L 2 ,
L 2 :=
3
L 2k .
(5.5)
k=1
In this chapter, we denote the vector composed of operators by the bold capital
letter as L, and the sum of the squares of components by the bold capital letter
with superscript 2 as L 2 . We also denote the sum of product of components of two
vectors composed of operators by the inner product of the two bold capital letter as

k Qk .
P Q := 3k=1 P
So, we have
L + = L ,
1
L 2 = L 23 + (L + L + L L + ).
2
(5.6)
Now, we introduce the spherical coordinate (r, 1 , 2 ) as (Fig. 5.1)

x1 = r sin 1 cos 2 , x2 = r sin 1 sin 2 , x3 = r cos 1 ,
(5.7)
where r [0, ), 1 [0, 2), and 2 [0, ). That is, (1 , 2 ) gives a coordinate
on the two-dimensional sphere S 2 .
153
Fig. 5.1 Spherical

coordinate
1 r
2
We define a linear map I from L 2 (R+ ) L 2 (S 2 ) = L 2 (R+ S 2 ) to L 2 (R3 ) as

follows, where R+ is the set of positive real numbers. For L 2 (R+ S 2 ), we
define
I()(x) :=
1
4r
(r, 1 , 2 )
with the condition (5.7). Since we have

|I()(x)|2 d 3 x
R3

1
2

=
4r 2
(r, 1 , 2 ) dr 2 (d1 , d2 )
4r
0
S2

2

=
(r, 1 , 2 ) dr 2 (d1 , d2 )
0
(5.8)
(5.9)
S2
with the normalized invariant measure 2 on S 2 , the map I is an isometry. So, we

can identity the system L 2 (R3 ) with the composite system L 2 (R+ ) L 2 (S 2 ) via the
isometry I defined in (5.8).
Then, we have

cos 1 cos 2
(5.10)
+
L 1 = i sin 2
1
sin 1
2

cos 1 sin 2
(5.11)
L 2 = i cos 2
1
sin 1
2
L 3 = i
.
(5.12)
2
154
Hence, we have

1
1
2
sin 1
+
L =
sin 1 1
1
sin2 1 22

cos 1
L = ei2
.
+i
1
sin 1 2
2
(5.13)
(5.14)
That is, the orbital angular momentum operator L j can be given as an operator on
L 2 (S 2 ).
L to satisfy the commutation relation
Now, we choose the generator F1k := i
k
(4.11) because of (5.4). Hence, we can employ the discussion in Sect. 4.2. In this case,
we have F 1 F 1 = 12 L 2 and F1 = 1 L . Also, F13 and F 1 F 1 are commutative. The
operator F 1 F 1 is a constant times of the Casimir operator, as explained in (4.20).
So, it takes only the value ( + 1) with a non-negative half integer , as explained
in (4.21). This constant determines the type of the irreducible representation as the
maximum weight. Under the irreducible representation with the maximum weight
, the operator i F13 takes only the value , 1, . . . , . So, m is a (half-) integer
when is a (half-) integer.
Now, we denote the eigenvector with eigenvalues ( + 1) and m of F 1 F 1 and
i F13 by the function Ym L 2 (R3 ). Due to the eigen equations for F 1 F 1 and
i F13 , we have
m
Y = mYm
2

1
1
2
sin 1
+
Y m = ( + 1)2 Ym .
2
sin 1 1
1
sin2 1 22
i
(5.15)
(5.16)
Since these partial differential equations depend on 1 and 2 , Ym can be regarded
as a function of 1 and 2 . Now, we define another function m := Ym eim2 . Then,
(5.15) and (5.16) are converted to
sin 1 1

sin 1

+
m2
sin2 1
m
=0
2

+ ( + 1) m = 0.
(5.17)
(5.18)
That is, m is a function only of 1 .

Firstly, we discuss these equations only when m = . Since the relation (4.18)
with m = shows that the operator L + annihilates the function Y , the relation
(5.14) implies

e
i2
cos 1
Y = 0,
+i
1
sin 1 2
(5.19)
155
which is equivalent to

cos 1
= 0,

1
sin 1
(5.20)
1
i.e., we have 1 log = cos
. Since log sin 1 is a solution of log , the general
sin 1
solution of is written as c sin 1 with a constant c . That is, the eigenfunction Y
is c ei2 sin 1 . However, since 1 takes values in [0, 2), the eigenfunction has to
be smooth at the neighborhood of 1 = , i.e., sin 1 = 0. To satisfy this requirement,
the maximum weight needs to be an integer. In this way, the maximum weight is
restricted to be an integer. Due to this restriction, our representation of su(2) can be
extended to the unitary representation of SO(3, R). It is natural because the function
space L 2 (S 2 ) has the symmetry of SO(3, R) naturally.
Now, we introduce another variable := cos 1 . Then, the eigenfunction
Pm () := m (1 ) satisfies Legendre equation;

m2
d2 m
d m
P m = 0.
(1 ) 2 P 2 P 1
d
d
1 2
2
(5.21)

(2+1)(+m)!
1
d m
The solution is given as Pm () = (1)
(1 2 ) , which is
4(m)! (1 2 ) d m
2
called Legendre polynomial. Therefore, the eigenfunction Ym is given as
Ym (1 , 2 ) = Pm (cos 1 )eim2 .
(5.22)
The function Ym is called the spherical harmonic function. Indeed, while Pm () has a
singularity at = 1, the singularity is resolved with the parameter 1 . So, Ym (1 , 2 )
has no singularity even at 1 = . When we treat the function Ym with an element
of L 2 (S 2 ), we denote it by |, m. It is known that {|, m},m forms a CONS of
L 2 (S 2 ), which implies that {|, m, m|},m is a PVM, i.e., gives a simultaneous
measurement of L 2 and L 3 .
The number is called the azimuthal quantum number and the number m is
called the magnetic quantum number. The azimuthal quantum number does not
depend on the choice of z-axis, but the magnetic quantum number depends on the
choice of z-axis. That is, the choice of the wave function Ym depends on the choice
of z-axis. The azimuthal quantum number corresponds to the subshell of the orbit
of electron as Table 5.1.
Finally, we give the concrete forms of Ym (1 , 2 ) with = 0, 1, 2 as follows.
156
Table 5.1 Structures of subshell

Azimuthal quantum
Name of subshell
number ()
Shape
s (sharp)
Sphere
p (principal)
Dumbbel
d (diffuse)
Double dumbbel
f (fundamental)
Complex
1
Y00 (1 , 2 ) =
4

3
cos 1
Y10 (1 , 2 ) =
4

3
Y11 (1 , 2 ) =
sin 1 ei2
8

5
0
(3 cos2 1 1)
Y2 (1 , 2 ) =
16

15
1
Y2 (1 , 2 ) =
sin 1 cos 1 ei2
8

15
2
Y2 (1 , 2 ) =
sin2 1 e2i2 .
32
Picture
(5.23)
(5.24)
(5.25)
(5.26)
(5.27)
(5.28)
5.1.2 Hamiltonian with Central Potential

Next, we consider a particle in the three-dimensional
system, which has the potential

3
2
V (r ) as a function of the radius r =
j=1 x j . When there is only a electric
charge only at the charge, the potential satisfies this condition. In this case, the total
1
P 2 as
Hamiltonian is given as the sum of the potential and the kinetic energy 2
H :=
1 2
P + V (r ),
2
(5.29)
157
where is the constant describing the reduced mass of the system consisting of the
M
with the mass of the electron m e
electron and the nucleus, which is given by mme e+M
and the mass of the atom M.
Now, we discuss the eigenequation of the Hamiltonian H as

1 2
P + V (r ) f (x) = E f (x).
2
(5.30)
For this purpose, we consider the relation among three operators L 2 , L 3 , and the
Hamiltonian H . From the definition of L 2 and L j , we have
L 2 = Q 2 P 2 ( Q P)2 + i Q P.
(5.31)

2
2
f (x) + 2r
L f (x) = r P f (x) + r
f (x) .
r 2
r
(5.32)

1
2 2
2
1 2
P f (x) =
f
(x)
+
f
(x)
+
L 2 f (x).
2
2 r 2
r r
2r 2
(5.33)
So, we have
2
Thus,
Hence, the eigenequation (5.30) can be simplified as

2
( + 1)2
2
r
+
+
V
(r
)
f = E f.
2r 2 r
r
2r 2
(5.34)
However, to discuss the Hamiltonian on L 2 (R+ S 2 ) = L 2 (R+ ) L 2 (S 2 ), we need

the unitary transformation I 1 , i.e., we focus on f(r, 1 , 2 ) := 2r f (r, 1 , 2 ).
Then, (5.34) is converted to
H f =

2
12 ( + 1)
+
+
V
(r
)
f = E f.
2 r 2
2r 2
(5.35)
1
P 2 and V (r ) are commutative with the operators
Hence, we find that the operators 2
L 2 and L 3 , i.e., the Hamiltonian H is commutative with the operators L 2 and L 3
defined on L 2 (S 2 ).
Now, we apply the discussion in Sect. 1.4.2. The simultaneous measurement of
L 2 and L 3 on L 2 (S 2 ) is given as {|, m, m|}l,m . Then, we have
H,m =
2
1
+
( + 1) + V (r )
2 r 2
2r 2
(5.36)
158
as an operator on L 2 (R+ ). Hence, we obtain the following Schrodinger equation

dependently of ;
2
E, +
2 r 2

( + 1)2
+ V (r ) E, = E E, .
2r 2
(5.37)
That is, when the function E, in L 2 (R+ ) is a solution of the above equation,
(r )
is a solution of the original (5.30).
f E,,m (r, 1 , 2 ) := Ym (1 , 2 ) E,2r
The eigen equation (5.37) can be regarded as a Schrdinger equation in the one2
2
dimensional system L 2 (R+ ) with the potential term ( (+1)
+V (r )). That is, (+1)
2r 2
2r 2
can be regarded as the centrifugal force.
In some case, the main term H0 of the Hamiltonian is given in (5.29), and the
fluctuation H is given as a small constant times of L 3 . For example, when there
is an interaction with the external magnetic field. Then, { f E,,m }E, , m gives the
eigenvectors of the Hamiltonian, as discussed in the end of Sect. 1.4. In this case, we
need to choose the z-axis dependently of the direction of the external magnetic field.
5.2 Laplacian on Sphere

5.2.1 General Case
Now, we generalize the above discussion to general d-dimensional case. For this
analysis, we focus on two self-adjoint operators Qk and P k on the function space
L 2 (Rd ) on the d-dimensional space Rd , which is defined as (5.1). Then, the total
kinetic energy is given as
d
d
1 2
2
,
Pj =
2m j=1
2m k=1 k
(5.38)
where k := x k . To discuss the total kinetic energy, we introduce the Laplacian

Rd := dk=1 k2 on the function space L 2 (Rd ) on the d-dimensional space.
As another important self-adjoint operator, we introduce the angular momentum
operator
k = i(Qk j Q j k )
L k, j := Qk P j Q j P
(5.39)
for k = j. For mathematical convenience, we employ another operator Fk,l :=

Qk j Q j k . Now, we investigate the relation between the operators Fk,l and the
Laplacian Rd . Since
Q2k
Qk k
2
Fk,l
=
k>l
159
Q2k l2
k =l
Qk Ql k l (d 1)
k =l
Q2k k2 +
k =l
Qk k +

l2 =
Q2k k2 +
Q2k l2 ,
Qk k
(5.40)
Qk Ql k l
(5.41)
k =l
(5.42)
we have
L 2k,l +
k>l
Qk k
Q2k l2 +
k =l
Q2k k2 (d 2)
Qk k

=
Q2k
l2 (d 2)
Qk k .
k
(5.43)
Now, in the d-dimensional space Rd , we introduce the spherical coordinate r, w

with r > 0 and w S d1 as xk = r wk , where S d1 is the d 1-dimensional sphere.
More precisely, we employ spherical coordinate r and = (1 , . . . , d1 ) as
r=
xk2
(5.44)
xk
(5.45)
= sin 1 sin k1 cos k
r
xd
= sin 1 sin d .
(5.46)
wd =
r

for k = 1, . . . , d 1. Then, we have r

= k wk k , i.e., k xk k = r
. To analyze
r
2
. The
the Laplacian Rd , we introduce the spherical Laplacian S d1 := k>l Fk,l
relation (5.43) is simplified to
wk =
= r 2 Rd (d 2)r ,
S d1 + r
r
r
(5.47)
which implies that

R d =
1
2
d 1
S d1 + 2 +
2
r
r
r r
(5.48)
because (r r
) = r 2 r
2 + r r . The relation (5.48) is the generalization of (5.33).
Now, we show that the operator Fk,l and the spherical Laplacian S d1 can be
regarded as operators on the function space L 2 (S d1 ). First, we notice that
2
160

d

xk
r =
xk2 , k = cos1
d
k=1
k
=k
(5.49)
xk2
for k = 1, . . . , d 1. Since
wk
r
=
xk
r

d
2
k
k
=k+1 wk
=

xk
r dk
=k wk2
wk w j
k
=

x j
d
d
2
2
r
k
=k wk
k
=k+1 wk
(5.50)
(5.51)
(5.52)
for k = 1, . . . , d 1, j > k.
d
cos1 f (x) = 1 2 dd xf . Hence, we have
dx
1 f (x)
k
r
=
+
x j
x j k
x j r
k=1
d1
j1
r

d
k=1
wk w j

d
d
2
2 k
k
=k wk
k
=k+1 wk
wk2

wj
+

r r
r dk
= j wk2
j
k
= j+1
(5.53)
for j = 1, . . . , d 1, and
k
r
=
+
xd
x
x
d
k
d r
k=1
d1
d1

k=1
Thus, for j > l, we have
d
k
=k
wk wd
wd
.
+

r r
d
2
k
wk2
k
=k+1 wk
(5.54)
F j,l =
F j,d =
l1
161
w j wl wk

d
2
2 k
k= j
wk
k
=k+1 wk

d
d
2
2
w
k
= j+1 wk

k =l+1 k
+ w j d
w
d
l
2
2
l
j
k
=l wk
k
= j wk
d2
d
k
=k
wk w j wd

d
d
2
2 k
k= j
k
=k wk
k
=k+1 wk

d
2
k
= j+1 wk

wd1
+ wj 2
w

d
d
2
j
wd1 + wd2 d1
k
= j wk
(5.55)
(5.56)
for k > l > d 1. Therefore, the operatorF j,l can be written as an operator on
2
is also written as an operator
L 2 (S d1 ). The spherical Laplacian S d1 = k>l Fk,l
2 d1
on L (S ).
In the following, we consider the function space L 2 (S d1 ), and discuss the eigen
equation with eigenvalue E of the spherical Laplacian S d1 as
S d1 f () = E f ().
(5.57)
We define the set Pn (d) of polynomials with degree n as a subspace of L 2 (Rd ).

Dividing all elements of Pn (d) by r n , the set Pn (d) can be regarded as a subspace
L 2 (S d1 ). The Laplacian Rd can be regarded as a linear map from Pn (d) to Pn1 (d).
For any element f of the kernel Ker Rd in Pn (d), (5.48) implies that
2
d 1
1
d1 f +
f
f +
S
2
2
r
r
r r
1
= 2 ( S d1 f + n(n 1) f + n(d 1) f ),
r
0=
(5.58)
which yields that

S d1 f = n(n + d 2) f.
(5.59)
The kernel of the Laplacian Rd on the subspace Pn (d) is the eigenspace of the
spherical Laplacian S d1 with the eigenvalue n(n + d 2). In the following,

, the
we denote the subspace by K n (d). Since the dimension of Pn (d) is n+d1
n
dimension of K n (d) is calculated as
162

n+d 1
n+d 3
n
n2
(2n + d 2)((n + d 3) (n + 1))
.
=
(d 2)!
dim K n (d) =
(5.60)
Indeed, the spherical Laplacian S d1 is closely related to the Lie algebra so(d)
as follows. Define the element X k,l of so(d) as |kl| |lk|. Since we have commutation relations
[X k,l , X k
,l
] = X k,l
l,k
X l,l
k,k
X k,k
j, j
+ X l,k
k,l
[L k,l , L k
,l
] = L k,l
l,k
L l,l
k,k
L k,k
l,l
+ L l,k
k,l
,
(5.61)
(5.62)
L k,l can be regarded as the representation of X k,l on L 2 (S d1 ). Since the killing

form is given as (X k,l , X k,l )so(d) = Tr ad(X k,l )ad(X k
,l
) = (d 1)k,k
l,l
for
1
S d1 .
k > l, k
> l
, the Casimir operator C is given as d1
Exercise 5.3 Show (5.50)(5.52), (5.55), and (5.56).
Exercise 5.6 Show that the function (x2 + i x1 )n belongs to K n (d).
Exercise 5.7 Show that the function
n2
k=0 (1)
n2k 2k
k n
x
x2
2k 1
belongs to K n (d).
5.2.2 Three-Dimensional Sphere

Since we have discussed the case with the two-dimensional sphere, we discuss
the detail with the three-dimensional sphere, i.e., d = 4. The case d = 4 has
the symmetry by the 4-dimensional special orthogonal group SO(4, R). The Lie
see this relation,
we express
group SO(4, R) is isometric to SU(2) SU(2)/Z2 . To
a
+
bi
ci
+
d
4
4
the real vector space R as the matrix space R :=
, whose
ci d a bi
element is a constant times of an element of SU(2). For an
element (g1 , g2 )
a + bi ci + d
SU(2) SU(2), we give an element of SO(4, R) that maps

ci d a bi

a + bi ci + d
g21 in the sense of coordinate (a, b, c, d) R4 . That is, this corg1
ci d a bi
respondence gives a homomorphism from SU(2) SU(2) to SO(4, R). The kernel is
{(I, I ), (I, I )}. Hence, we find that SO(4, R) is isometric to SU(2) SU(2)/Z2 ,
which implies that so(4)
= su(2) su(2).
163
The set of elements of R 4 with a 2 + b2 + c2 + d 2 = 1 is SU(2). Hence, we can

regard L 2 (S 3 ) as L 2 (U(2)). Due to Fourier transform on the group SU(2) defined in
Sect. 3.8, we find that
= SU(2)
L 2 (SU(2))
= L 2 (SU(2))
U U .
(5.63)
is the irreducible component of L 2 (SU(2))

Hence, the space U U each SU(2)
with respect to the representation of so(4). The dimension of U U is (2 + 1)2 .
Due to (5.60), dim K n (4) = (n + 1)2 . Since the space K n (4) is closed with respect
to the representation of so(4), we have U n2 U n2 = K n (4).
Now, we consider the subgroup SO(3, R) SO(4, R) in the sense of the subset
{(g, g)|g SU(2)} of SU(2)SU(2). This subset does not change a and b2 +c2 +d 2 .
That is, this action of SO(3, R) can be regarded as rotation for the three parameters
b, c, d. In this subrepresentation of so(3) = su(2), due to (4.22), the space U n2 U n2
is irreducibly decomposed to
U n2 U n2 = Un Un1 U0 .
(5.64)
Now, we consider an integral operator T on L 2 (S 3 ) as

T ( f )(w) =
S3
f (v)
3 (dv),
|v w|2
(5.65)
where 3 is the invariant measure on S 3 whose full measure is 2 2 , i.e., the measure
of the volume of 3-dimensional unit sphere in R4 . Since the operator T is invariant
with respect to the action of SO(4, R), the operator T is a constant on the irreducible
component K n (4). Hence, we have the following theorem.
Theorem 5.1 The operator has the following spectral decomposition;

2 2
Pn ,
T =
n+1
n=0
(5.66)
where Pn is the projection to K n (4).

Proof Now, we choose a function f (v) = (v 2 + iv 1 )n in K n (4). When w =
(1, 0, 0, 0), we have |v w|2 = (2 sin 21 )2 = 2(1 cos 1 ) as Fig. 5.2.
Hence, when n 1,
(sin 1 cos 2 + i cos 1 )n

sin2 1 sin 2
2(1
cos
)
1
0
0
0
2
2 n
sin 1
(a) 2i
(b) 2 i
=
d1
sin(n1 ) =
,
(5.67)
n+1 0
2(1 cos 1 )
n+1
T ( f )(w) =
d1
n
d2
d3
164
Fig. 5.2 Difference between

v and w
2sin 2
1
2
v
where the steps (a) and (b) follow from Exercises 5.8 and 5.9. Since f (w) = i n , the
2 2
. Also, the case when n = 0 can be shown as (Exercise 5.10)
eigenvalue is n+1
T ( f )(w) =
0
d1
d2
0
d3
0
sin2 1 sin 2
= 2 2 .
2(1 cos 1 )
(5.68)
Exercise 5.8 Show the step (a) in (5.67).

2 sin n
Exercise 5.9 Show the equation 0 sin1cos
d = 2 for any integer n 1 by
following steps below.

2
2
a: Show that 0 sin sin nd = 0 cos cos nd = 0 for n 2.
2 sin2
b: Show the equation 0 1cos d = 2.
2 2 sin 2
d = 2.
c: Show the equation 0 sin1cos
2 sin sin(n+1)
2 sin n
d: Show the equation 0
d = 0 sin1cos
d for n 2.
1cos
5.3 Hydrogen-Type Hamiltonian and SO(4, R)

5.3.1 Puzzle for Degeneracy
Next, we discuss the orbital Hamiltonian of an electron around hydrogen-type atom.
Z e2
,
In this case, the Hamiltonian H is given in (5.29) with the potential V (r ) = 4
0r
where 0 is the permittivity of the vacuum, Z is the atomic number (charge of the
nucleus in unit e; number of protons in the nucleus), e is the elementary charge (charge
of an electron), is the reduced mass of the system consisting of the electron and
the nucleus. Then, it is known [77] that Schrdinger equation is given as2
2 The Hamiltonian 1 P 2
2
i.e., the domain of P 2 .
Z e2
40 r
is defined as a self-adjoint operator on the second Sobolov space,
165
Table 5.2 Summary of quantum numbers

Eigen equation
Range of quantum number
2
H nm = Z2nE2h nm
L 2 nm = 2 ( + 1)
nm
L z nm = m nm
Name of quantum number
n = 1, 2, . . . ,
Principal quantum number
= 0, 1, . . . , n 1
m = , 1 , . . . ,
Azimuthal quantum number

Magnetic quantum number

1 2
Z e2
P
= E.
2
40 r
(5.69)
In this case, the solution n, of the equation (5.37) is given as

n, (r ) =
2Z
na
3/2
(n 1)!
2n[(n + )!]
1/2
e
Zr/na
2Zr
na
l
L 2+1
n1
2Zr
na
(5.70)
0
with a = 4
and a new quantum number n = 1, 2, . . . , , which is called the
e2
principal quantum number. Here, the azimuthal quantum number takes values
2
2
n 1, n 2, . . . , 0. The eigenvalue is Z2nE2 h with E h := ( 4e 0 )2 . Hence, the total
wave function on L 2 (R+ S 2 ) is given as
2
n,,m (r, 1 , 2 ) := n, (r ) Y,m (1 , 2 ).
(5.71)
That is, the wave function on L 2 (R3 ) is given as

1
n,,m (x) :=
n, (r ) Y,m (1 , 2 )
4r
(5.72)
with (5.7). Further, the LHS of (5.37) also has continuous spectrum on (0, ) [100,
(10.6) p. 222]. These spectrums correspond to states that are not bounded by the
atom. Such a solution is called a free electron. Therefore, the quantum numbers
of states bounded by the atom are summarized as Tables 5.2 and 5.3. That is, the
eigenvalue of the Hamiltonian depends only on the principal quantum number n,
which decides the shell of orbit, which is illustrated in Fig. 5.3. So, this equation has
a large degeneracy n 2 of the Hamiltonian. Since the above discussion explain the
symmetry with respect to 2-dimensional sphere, it can explain the degeneracy 2 + 1
in each subshell, cannot explain the large degeneracy n 2 in each shell. In the next
subsection, we solve this puzzle by using the larger symmetry based on SO(4, R).
166
Table 5.3 Relation among principal, azimuthal, and magnetic quantum numbers
Shell n
Azimuthal
Subshell
Magnetic
Number
Number of
quantum
quantum
orbitals of
orbitals of
number
numbers
subshell
shell
K-shell
(n = 1)
L-shell
(n = 2)
M-shell
(n = 3)
N-shell
(n = 3)
1s
2s
1
0
2p
3s
1, 0, 1
0
3
1
1
2
0
3p
3d
4s
1, 0, 1
3
2, 1, 0, 1, 2 5
0
1
1
2
3
4p
4d
4f
1, 0, 1
3
2, 1, 0, 1, 2 5
3, 2, 1, 0, 1, 7
2, 3
16
Fig. 5.3 Shells of hydrogen

type atom
N
M
L
K
Ze
n=1
n=2
n=3
n=4
5.3.2 Symmetry Based on SO(4, R)

When the eigenvalue E is negative, the eigenspace L E associated with the eigenvalue
E has symmetry based on the group SO(4, R) behind [30]. When the eigenvalue
E is positive, the eigenspace L E associated with the eigenvalue E is irreducible
representation (principal series) of the group SO(3, 1, R) [83]. We discuss only the
hidden symmetry of SO(4, R) according to [13], [22, Chap. 9] because the symmetry
by the group SO(3, 1) is too complicated. To see the symmetry, we apply Fourier
167
p) := F3 []( p) given in (3.68) to a function in L 2 (R2 ). Then, due to

transform (
the discussion in the next subsection, the eigenequation (5.69) with eigenvalue E is
converted to
2 p 2
Z e2
( p) 3
2
8 0
Using h :=
,
2

R3
p
)
(
p).
dp
= E (
| p p
|2
(5.73)
we have

p
)
p2
(
Z E h h
p).
dp
= Eh (
( p)
2
2
2
2
R3 | p p |
(5.74)
Dependently of the eigenvalue E, we introduce the coordinate of S 3 based on the

momentum coordinate R3 by the inverse of the stereographic projection from S 3
to R3 as
w( p) j :=
2 p0 p j
,
p 2 + p02
w( p)4 :=
p 2 p02
p 2 + p02
j = 1, 2, 3
(5.75)
(5.76)
where p 2 = 3j=1 p 2j and p0 is set to be 2Eh . That is, using the vector n =
(0, 0, 0, 1), w( p) is given as
w( p) =
p 2 p02
2 p0
n+ 2
p.
p 2 + p02
p + p02
(5.77)
p0 wi
1 w4
(5.78)
Then, pi (w) is given as

p(w)i :=
for i = 1, 2, 3. That is, by using the vector n = (0, 0, 0, 1)T , the relation between w
and p is illustrated in Fig. 5.4 and is given as
w=
2 p0
p 2 p02
n+ 2
p.
p 2 + p02
p + p02
(5.79)
Now, we have the following theorem.

Theorem 5.2 When the points w and w
in S 3 correspond to the points p and p
in
R3 in the above way, we have
168
Fig. 5.4 Stereographic

projection from S 3 to R3
n
S
w
3
p p0
| p p
|2 =
( p 2 + p02 )( p
2 + p02 )
|w w
|2 .
(2 p0 )2
(5.80)
Proof It is sufficient to show the

special case p0 = 1 because the special case with
replacing p and p
by pp0 and pp0 yields the general case. Our proof consists of five
steps.
Step 1: We show that
| p n| |w n| = 2
(5.81)
when p is outside the unit sphere S 3 . We choose the point s as the south pole point
as Fig. 5.5. Since the angles nws and nop are 2 , the triangle snw is homothetic
to the triangle pno. So, we have
|s n|
|w n|
=
.
|o n|
| p n|
(5.82)
|w n| | p n| = |o n| |s n| = 2.
(5.83)
Thus,
Step 2: We show that (5.81) when p is on the unit sphere S 3 . Since p = w = a, we

have (5.83).
Step 3: We show that (5.81) when p is inside of the unit sphere S 3 . We can show
(5.83) in the same way as Step 1 (See Fig. 5.6).
Step 4: We show that the triangle nww
is similar to the triangle np
p. We can
show (5.81)
for p
and w
. Thus, | p n| |w n| = | p
n| |w
n|, i.e.,
| pn|
p n|
= ||wn|
. Also, nww
= nww
(common). Therefore, the triangle nww
|w
n|
is similar to the triangle np
p.
Step 5: We show (5.80). The similarity relationship between the triangles nww
and np
p guarantees that
169
Fig. 5.5 Cross section with

onwp when p is inside of the
unit sphere S 3
w
o
s
Fig. 5.6 Cross section with
onwp when p is outside of
the unit sphere S 3
s
| p
n| (a)
| p
n| | p n|
= |w w
|
|w n|
2

2
2
p +1 p +1
(b)
,
=|w w
|
2
| p p
| =|w w
|
where (a) follows

from (5.81) and (b) follows from the relations | p n| =
and | p n| = p
2 + 1. Taking the squares, we obtain (5.80).
(5.84)

p2 + 1
Now, we derive the description 3 (dp) of the volume integral 3 (dw) in the
3-dimensional unit sphere S 3 in terms of the coordinate p = ( p1 , p2 , p3 ), whose full
measure is 2 2 . The metric is given as
g( p)k,l :=
4

w( p) j w( p) j
.
pk
pl
j=1
As shown later, we have the following lemma.

Lemma 5.1 The determinant of g( p)k,l is ( p22+p0p2 )6 .
0
Hence, the normalized invariant volume integral is
(5.85)
170
3 (dp) =

det g( p)d p =
3
2 p0
2
p + p02
3
d 3 p.
(5.86)
Now, we define the map FE from L 2 (R3 ) to L 2 (S 3 ) as

FE ( f )(w) :=
p(w)2 + p02
2 p0
2

f p(w) .
(5.87)
Then, the square of the norm of FE ( f ) is given as

FE ( f )(w) S 3 :=

=
R3
R3

=
R3
|FE ( f )(w( p))|2 3 (dp)
p 2 + p02
2 p0
4

| f ( p)|2
2 p0
p 2 + p02
3
d3 p
p 2 + p02
| f ( p)|2 d 3 p.
2 p0
(5.88)
That is, the map FE is not necessarily a unitary map, in general.

However, we have the following lemma.
Lemma 5.2 When the map FE is restricted to the subspace L E , the map FE is an
isometry.
1 3
Proof We define the operator A := 2

k=1 (Pk Qk + Qk Pk ). Then, using the commutation relation (5.2), we have the commutation relation
i[H, A] =
1 2
P + H.
2
(5.89)
For any normalized wave function f L E , we have f, H A f = E f, A f =

f, AH f , which implies that
3

1 2
0 = i f, [H, A] f = f, (
Pj + H) f
2 j=1
1 2
= f, (
P )f + E =
2 j=1 j

R3
2 p 2
| f ( p)|2 d 3 p + E.
2
171
Thus,

R3
p 2 | f( p)|2 d 3 p = 2Eh = p02 .
(5.90)
Combining (5.88) and (5.90), we have

FE ( f )(w) S 3 =
p02 + p02
= 1,
2 p0
which implies that the map FE preserves the norm for any element in L E .
(5.91)
Lemma 5.3 Any element of FE (L E ) satisfies

S3
(w
)
p0
3 (dw
) =
(w).
|w w
|2
Z E h h
(5.92)
Proof The equation (5.74) is written as
p2
2
Eh
FE1 ()( p)
E h h
2 2

R3
FE1 ()( p
)
dp .
| p p
|2
(5.93)
The LHS of (5.93) is calculated to
p2

p 2 + p02 1
Eh FE1 ()( p) =
FE ()( p)
2
2
2 p02
p 2 + p02
2 p0 2
(w(
p))
=
(w( p)).
=
2
p 2 + p02
p 2 + p02
(5.94)
Let w and w
be w( p) and w( p
). Then, Theorem 5.2 guarantees that
| p p
|2 =
( p 2 + p02 )( p
2 + p02 )
|w w
|2 .
(2 p0 )2
(5.95)
The LHS of (5.93) is calculated to

E h h
FE1 ()( p
) 3
d p
2
2
2
R3 | p p |
2 p0 2

()(w
)
p(w
)2 + p02 3
Z E h h
p(w
)2 + p02
=
(dw
)
( p2 + p02 )( p(w
)2 + p02 )
2 2
2
p
|2
0
R3
|w(
p)
w
(2 p0 )2

2 p0 E h h
()(w
)
= 2 2
3 (dw
).
(5.96)
2
2 ( p + p0 ) R3 |w( p) w
|2
Z
Combining (5.93), (5.94), and (5.96), we have
172

2 p02
()(w
)
2 p0 E h h
(w)
=
3 (dw
),
2
2
p 2 + p0
2 2 ( p 2 + p0 ) R3 |w w
|2
(5.97)
which implies the desired equation.
2
2
p0
Z E h h
E h h
. So,
n
Now, we recall Theorem 5.1. From Lemma 5.3, we see that the eigenvalue
is restricted to
2 2
n
with a positive integer n. That is, p0 is

p2
2 2 Z E h h
2 2 n
the energy E is 20h = Z2nE2 h .

In summary, we obtain the following theorem.
2
Theorem 5.3 The negative eigenvalue E is restricted to Z2nE2 h with a positive integer n. Then, the map FE |L E is a unitary map from L E to Kn1 (4).
Now, we consider the representation of SO(3, R) on L 2 (R3 ). When we consider the L 2 (R3 ) with the momentum coordinate, the irreducible decomposition of
2
SO(3, R) is the same as that with the position coordinate. When E = Z2nE2 h , due to
(5.64), the eigenspace FE |L E (L E ) is irreducibly decomposed to
Un1 Un2 U0 ,
(5.98)
which clarifies the degeneracy of the Hamiltonian for the azimuthal quantum
number .
Proof of Lemma 5.1 As shown later, the diagonal element g( p)k,k is ( p22+p0p2 )2 and
0
the off diagonal element g( p)k,l is 0 with k = l. Hence, the determinant of g( p)k,l
is ( p22+p0p2 )6 .
0
To obtain the diagonal element g( p)k,k , we calculate
w j
pk
as
4 p0 pk p j
( p2 + p02)2
when k = j, j = 1, 2, 3
w j
4 p0 ( p 2 p2j + p02 )
when k = j, j = 1, 2, 3
=
( p2 + p02 )2
pk
2
4
p
p
0 k
when j = 4.
( p2 + p2 )2
(5.99)
So, we have
g( p)k,k =

4

w j 2
j=1
pk

=
2 p0
p 2 + p02
2
.
(5.100)
Similarly, we have
g( p)k,l =
4

w j w j
= 0.
pk pl
j=1
(5.101)
173
Exercise 5.11 Show the commutation relation (5.89).

5.3.3 Fourier Transform of Coulomb Potential*

This subsection discusses Fourier transform of Coulomb potential. That is, we show
the following lemma.

1
Lemma 5.4 Any element D(( 3j=1 Q2j ) 2 ) satisfies
F3

3
j=1
Q2j
21

F3 [](
)
1
() =
d .
2 2 R3 |
|2
(5.102)
Proof For this proof, we recall that the Coulomb potential is defined as the closed
extension of the operator defined in S (R3 ). in Sect. 1.6. Hence, it is sufficient to
show (5.102) when belongs to S (R3 ). Now, we define the convolution
for
two functions and
as

1
(x
)
( x
)d x
.
(5.103)
(x) :=
3
(2) 2 R3
1
2
and 4 (x) := |x|
We define two functions 3 (x) := |x|
2 . Indeed, these functions 3
and 4 are not square integrable. However, when a function 5 is a rapidly decreasing
function, we can define 5 |3 and 5 |4 as

5 (x) j (x)d 3 x
(5.104)
5 | j :=
R3
for j = 3, 4. Since S (R3 ) is a dense subset, it is enough to show that

1 |F3 (3 2 ) = 1 |4 F3 (2 )
(5.105)
for any 1 S (R3 ).

Thus, we have
1 |F3 (3 2 ) = F31 (1 )|3 2 = F31 (1 )2 |3
(a)
=F31 F3 (F31 (1 )2 )|3 = F31 (1 F3 (2 ))|3

(b)
=F31 (1 F31 (2 ) )|3 ,
(5.106)
174
and
(c)
1 |4 F3 (2 ) = 1 F31 (2 ) |4 .
(5.107)
where (a), (b), and (c) follow from Exercises 3.22, 3.12, and 5.13, respectively.
Since the set S (R3 ) is closed for Fourier transform, product and convolution (see
Exercises 1.11, 1.12, 3.14, and 3.21), it is enough to show that
F31 (5 )|3 = 5 |4
(5.108)
for any 5 S (R3 ). Now, we choose the function (x) := e|x| . For any 6
S (R3 ), we have 6 6 |3 0 as 0. Since
lim F3 ( 3 ) = 4 ,
(5.109)
(see Exercise 5.14), we have

F31 (5 )|3 = lim F31 (5 ) |3 = lim F31 (5 )| 3
0
= lim 5 |F3 ( 3 ) = 5 | lim F3 ( 3 ) = 5 |4 .

0
(5.110)
The proof is completed.

Exercise 5.14 Show (5.109) by following steps below.
a: Choose the spherical coordinate x = r sin cos , y = r sin sin , and z =

r cos . Show that

i(x p1 +yp2 +zp3 )
|x| e
r 4

e
d xd ydz = e
sin(| p|r )dr.
(5.111)
| p| 0
x 2 + y2 + z2
R3
b: Show that
4
0 | p|
lim
c: Show (5.109).
er sin(| p|r )dr =
4
.
| p|2
(5.112)
5.4 Spin-Orbit Interaction
175

In the previous section, we have discussed the freedom of the orbit of an electron.
In addition to this freedom, an electron has another internal freedom, spin, which is
represented by a two-dimensional space C2 and is spanned by | and | (Fig. 5.7).
In the spin system C2 , we consider three Hermitian matrices

S1 :=
2

0 i
1 0
01
, S2 :=
, S3 :=
.
10
2 i 0
2 0 1
(5.113)
Then, the total system of an electron is given as L 2 (R3 ) C2 = L 2 (R+ ) L 2 (S 2 )

C2 . When an electron has a homogeneous interaction between the orbital angular
momentum and the spin, we discuss the following type of Hamiltonian
H :=
1 2
P + V1 (r ) + V2 (r )L S,
2
(5.114)
where the additional term V2 (r )L S expresses spin-orbit interaction. Many kinds

of Hamiltonian can be written as special cases of the above Hamiltonian, e.g., Larmor interaction (V2 (r ) = m22 cB2 r d Vdr1 (r ) ) and Thomas interaction [101]. Since matrices
e
{ i Sk }3k=1 satisfy the commutation relation (4.4), they give a skew-adjoint representation of su(2). When we define the total angular momentum Jk := Sk + L k , the
matrices { i Jk }3j=1 give a skew-adjoint representation of su(2) on the composite
system. Since
J =
2
3

k=1
3
Jk2 = L 2 + 2L S + S2 = L 2 + 2L S + ,
2
(5.115)
we have
LS=
3
1 2
( J L2) .
2
4
(5.116)
By using (5.33), the Hamiltonian is simplified to
Fig. 5.7 Spin and orbital

angular momentum
orbital momentum
spin
176
2
2
2
3
+ V1 (r ) V2 (r )
+
2 r 2
r r
4
V2 (r ) 2
V2 (r ) 2
1
L .
J +
+
2
2r 2
2
H =
(5.117)
as an operator on L 2 (R3 ) C2 . However, we need to treat the Hamiltonian on the

space L 2 (R+ ) L 2 (S 2 ) C2 . Then, we have
H =
3
V2 (r ) 2
1
V2 (r ) 2
2 2
L .
J +
+ V1 (r ) V2 (r ) +
2
2
2 r
4
2
2r
2
(5.118)
as an operator on L 2 (R+ ) L 2 (S 2 ) C2 .
The operators H , J 2 , L 2 , and J3 are commutative with each other and J 2 , L 2 ,
and J3 are operators on L 2 (S 2 ) C2 . Based on the discussion in Sect. 4.2.4, in
particular, Example 4.1, we focus on the integer identifying the irreducible representation for the orbital angular momentum L and the integer j identifying the
irreducible representation for the total angular momentum J. When the irreducible
representation for the orbital angular momentum is identified with , the vector | j; m
by |; j; m. That is, the basis {|; j; m}=1,2,..., j= 21 ,m= j,..., j forms a CONS of
the space L 2 (S 2 ) C2 . In fact, |; 21 ; m is given as a linear combination of
|, 21 ; m + 21 , 21 and |, 21 ; m 21 , 21 , whose coefficients (Clebsch-Gordan coefficients) are given in Example 4.1. Then, we simultaneously diagonalize J 2 , L 2 , and
J3 as

j ( j + 1)|; j; m; j; m|
(5.119)
J2 =
, j,m
L =
2
( + 1)|; j; m; j; m|
(5.120)
m|; j; m; j; m|,
(5.121)
, j,m
J3 =
, j,m
which gives the simultaneous measurement of J 2 , L 2 , and J3 .

Now, we apply the discussion in Sect. 1.4.2. From (5.118), we have
2 2
3
+ V1 (r ) V2 (r )
2
2 r
4
1
V2 (r ) 2
V2 (r ) 2
( + 1).
j ( j + 1) +
+
2
2r 2
2
H, j,m =
(5.122)
as an operator on L 2 (R+ ). Hence, we obtain the following Schrdinger equation

dependently of and j;
177
H, j,m E,, j = E E,, j .
(5.123)
That is, when the function E,, j in L 2 (R+ ) is a normalized solution of the above
equation, the wave function E,, j |; j; m on L 2 (R+ ) L 2 (S 2 ) C2 is an eigenvector of the Hamiltonian H given in (5.114) associated to the eigenvalue E.
5.5 Quark Model

5.5.1 Quark and Anti-quark
In particle physics, the atom is not the most fundamental particle. An atom is composed of protons and neutrons and electrons. While the electrons are elementary
particles, the protons and neutrons are composed of more fundamental constituents
called quarks. These quarks constitute the sub-structure of a proton or a neutron.
More generally, composite particles that are made up from these quarks are called
hadrons. Typical hadrons are made up of two or three quarks. Hadrons with two
quarks held together are called mesons, whereas those with three quarks are termed
baryons. Hence, we introduce quarks as substructures of a proton and a neutron.
Firstly, we introduce up quark u and down quark d, and explain that a proton is
composed of two up quarks and a down quark, and a neutron is composed of an up
quark and two down quarks. In this explanation, an up quark u has an electric charge
of + 23 e and a down quark d has an electric charge of 23 e so that a proton has an
electric charge of e and a neutron has an electric charge of 0 as Fig. 5.8. Besides
the proton and the neutron, other exotic hadrons exists. To explain the strangeness
of these particles, a strange quark is introduced. A strange quark s has an electric
charge of + 23 e. To discuss the strangeness, we introduce a new quantum number
strangeness S so that only a strange quark has strangeness 1. For simplicity, we
introduce the quantum number, electric charge Q, whose e times is the electric charge.
Since we have three quarks, up u, down d, and strange s, the quantum system of one
quark is C3 and has the fundamental representation of SU(3). The space is spanned
by |u, |d, |s, and the Dynkin index is [1, 0] so that the system is U[1,0] (3). This
freedom is called flavor. To clarify the group action SU(3) for flavor, we denote the
group by SUf (3). The representation space U[n 1 ,n 2 ] (3) is written as U[n 1 ,n 2 ] (3)f .
Fig. 5.8 Proton and neutron
proton
neutron
178
Now, the quantum numbers, electric charge Q and strangeness S are mathematically defined as the following Hermitian matrices;
1
1
2
|uu| |dd| + |ss|
3
3
3
S := |ss|.
Q :=
(5.124)
(5.125)
We also employ quantum numbers, isospin I3 and hypercharge Y , which are defined
as
1
|uu|
2
1
Y := |uu|
3
I3 :=
1
|dd|
2
1
2
|dd| + |ss|.
3
3
(5.126)
(5.127)
So, Q = I3 + 21 Y . The skew-adjoint representation f[1,0] of su(3) is given as

i
1
y
(|ql q j | + |q j ql |), f[1,0] (F j,l ) := (|q j ql | |ql q j |),
2
2
z
z
) := i(|q1 q1 | |q2 q2 |), f[1,0] (F2,3
) := i(|q2 q2 | |q3 q3 |)
f[1,0] (F1,2
x
f[1,0] (F j,l
) :=
for 1 j < l 3, where q1 = u, q2 = d, and q3 = s. In particle physics, to discuss

the Lie algebra sd(3) of diagonal matrices on C3 , we usually use
F3 :=
1
1
|11| |22|,
2
2
1
1
1
F8 := |11| + |22| |22|
2 3
2 3
3
z
z
and F2,3
. So, we have
instead of F1,2
f[1,0] (F3 ) = I3 , f[1,0] (F8 ) =
3
Y,
2
(5.128)
Also, these quarks have an internal freedom of spin 21 , i.e., the internal freedom for
rotation whose quantum system is C2 with the fundamental representation of SU(2),
whose Dynkin index is [1] so that the system is U[1] (2). Due to relativistic quantum
mechanics, when a particle has spin 21 , it is needed to be a fermion. So, a quark is
a fermion. To express this freedom, we add subscript s . Further, quark has another
internal freedom for color, whose system is C3 and is spanned by blue |b, green
|g, and red |r . To express this freedom, we add subscript c . This space is the
fundamental representation of SUc (3). Hence, under the representation of the group
SUf (3) SUs (2) SUc (3), the total one-quark system is written as U[1,0] (3)f
U[1] (2)s U[1,0] (3)c .
Further, each quark has its anti-quark, i.e., anti-up quark u,
anti-down quark
and anti-strange quark s . The anti-quark has the electric charge and strangeness,
d,
i.e., Q and S that are opposite to those of the original quark, and has a spin 21
5.5 Quark Model
179
as well as the original quark. Since their charge and strangeness are opposite, the
representation of SUf (3) in this system is the complex conjugate representation of
the the fundamental representation. That is, the Dynkin index is [0, 1] so that the
system is U[0,1] (3)f . The skew-adjoint representation f[0,1] (3)f of su(3) is given as
i
1
y
x
f[0,1] (F j,l
) := (|ql q j | + |q j ql |), f[0,1] (F j,l ) := (|q j ql | |ql q j |),
2
2
z
z
) := i(|q1 q1 | |q2 q2 |), f[0,1] (F2,3
) := i(|q2 q2 | |q3 q3 |)
f[0,1] (F1,2
and q3 = s . The quantum numbers I3
q2 = d,
for 1 j < l 3, where q1 = u,
and Y are characterized as
3
Y.
(5.129)
f[0,1] (F3 ) = I3 , f[0,1] (F8 ) =
2
Since the complex conjugate representation of the fundamental representation of
SUs (2) is the fundamental representation of SUs (2), the anti-quark has the fundamental representation of SUs (2) so that the system is U[1] (2)s because the complex
conjugate of U[1] (2)s is also U[1] (2)s . Further, anti-quark has another internal free anti-green
dom for color, whose system is C3 and is spanned by anti-blue |b,
|g,
and anti-red |r . This space is the complex conjugate of the fundamental representation space U[0,1] (3)c with respect to SU(3)c . Hence, under the representation of the group SUf (3) SUs (2), the total one-anti-quark system is written as
U[0,1] (3)f U[1] (2)s U[0,1] (3)c .
5.5.2 Meson
Hadrons are classified to two types, mesons and baryons. A meson is composed of
a quark and an anti-quark. For example, the meson is composed of d and u as
Fig. 5.9. First, we omit the freedom of color for simplicity. Then, since a quark is
different from an anti-quark, under the representation of the group SUf (3) SUs (2),
the one-meson system is given as (U[1,0] (3)f U[1] (2)s ) (U[0,1] (3)f U[1] (2)s ).
Fig. 5.9 Meson
180
Due to (4.54), the irreducible decomposition is

(U[1,0] (3)f U[1] (2)s ) (U[0,1] (3)f U[1] (2)s )
= (U[1,0] (3)f U[0,1] (3)f ) U[1] (2)2
s
= (U[0,0] (3)f U[1,1] (3)f ) (U[0] (2)s U[2] (2)s )
= U[0,0] (3)f U[0] (2)s U[0,0] (3)f U[2] (2)s
U[1,1] (3)f U[0] (2)s U[1,1] (3)f U[2] (2)s .
(5.130)
As U[1,1] (3)f is the adjoint representation, the representation space can be identified
with the Lie algebra (3). The dimensions are calculated as dim U[0,0] (3)f = 1 and
dim U[1,1] (3)f = 8. The total dimension is 66 = 36 = 11+13+81+83.
Since dim U[0,0] (3)f = 1 and dim U[1,1] (3)f = 8, the systems U[0,0] (3)f and
U[1,1] (3)f are called meson singlet and meson octet, respectively. A meson particle is an orthogonal base under this irreducible decomposition. When a meson
particle belongs to a system with U[0] (2)s it has spin 0 and is called a pseudoscalar
meson, and when a meson particle belongs to a system with U[2] (2)s it has spin 1
and is called a vectorial meson. Mesons are listed as Table 5.4.
Therefore, the values of the quantum numbers I3 and Y are summarized as
Figs. 5.10 and 5.11.
Now, we consider the freedom of color. As a physical law, the state of color in a
+ |g g
|gg
+ |r r
|bb
|rr ), which is
meson is restricted to |
:= 16 (|bb
called color confinement. Since a quark is a fermion, the total system is restricted
to the two-fold anti-symmetric space. Even though we take account into the color
freedom, Theorem 4.3 with r = 3 and d = 3 and the color confinement guarantee
that the system of meson is reduced to (U[1,0] (3)f U[1] (2)s ) (U[0,1] (3)f U[1] (2)s ).
Table 5.4 Lits of mesons: the wave function |spin is chosen to be 1 (| + | ) for a
2
pseudoscalar meson, and is chosen to be | for a vector meson. However, for a vector meson, the
wave function |spin can be chosen to be any element of U[2] (2)s
Pseudoscalar meson
Vectorial meson
Wave function
K+
K+
K0
K
0
K
K0
K
0
K
|u s |spin
|d s |spin
|s u|spin
|s d|spin
|u d|spin
|d u|spin
1 (|u u
|d d)|spin
2
1
2|s s )|spin
(|u u
+ |d d
6
+ |s s )|spin
1 (|u u
+ |d d
3
5.5 Quark Model

Fig. 5.10 Pseudoscalar
meson
181
Y
1
0
1
Fig. 5.11 Vectorial meson
Y
1
0
1
K+
K0
K
1 12 0
K0
1
1
2
0
K *0
K *
1
I3
K *+
K *0
1
2
1
2
I3
5.5.3 Baryon
Another type of hadrons, a baryon is composed of three quarks, and an anti-baryon
is composed of three anti-quarks. To discuss baryon, we need to discuss the freedom
of color, first. Due to a physical law, the state of color of a baryon is restricted to
| := 16 (|bgr + |gr b + |r bg |gbr |r gb |br g), which is called color
confinement. That is, a baryon is composed of three different colors as Fig. 5.8.
We apply Theorem 4.2 with r = 6 and n = 3 to the system H := U[1,0] (3)f
U[1] (2)s U[1,0] (3)c . Then, our system of possible state is the three-fold symmetric
tensor product system U[1,0] (3)f U[1] (2)s , which is a 6-dimensional space. That is,
due to the color confinement, a quark behaves as a boson only with the freedoms of
flavor and spin while it is truly a fermion when including the freedom of color.
Under the group SU(6), we denote the total one-quark system by U[1,0,0,0,0] (6),
Then, the one-baryon system is given as U[3,0,0,0,0] (6), which is an irreducible space
of the representation of SU(6). However, it is not irreducible for the subgroup
SUf (3) SUs (2) SU(6). Due to Theorem 4.1, the system U[3,0,0,0,0] (6) is irreducibly decomposed as
U[3,0,0,0,0] (6) = U[3,0] (3)f U[3] (2)s U[1,1] (3)f U[1] (2)s
(5.131)
in the sense of the representation of SUf (3) SUs (2). The dimensions are calculated as dim U[3,0] (3)f = 10 and dim U[1,1] (3)f = 8. Due to Exercise 4.9, the total
= 56 = 10 4 + 8 2.
dimension is 876
32
The space U[3,0] (3)f U[3] (2)s is called baryon decuplet, and is generated
by the vector |uuu with the action of SUf (3) SUs (2). the other space
182
Table 5.5 Lits of baryon

octet: It is possible to
exchange and . This
exchange corresponds to the
freedom in U[1] (2)s
Name
Wave function
p
n
0
+
0
|duu|
+ cyclic terms
|udd|
+ cyclic terms
(5.132)
|suu|
+ cyclic terms
(5.133)
|sdd|
+ cyclic terms
|uss|
+ cyclic terms
|dss|
+ cyclic terms
Table 5.6 Lits of baryon

decuplet: |spin is any
element of U[3] (2)s
Name
Wave function
++
+
0
+
0

0

|uuu|spin
|uud|spin + cyclic terms
|udd|spin + cyclic terms
|ddd|spin
|uus|spin + cyclic terms
|uds|spin + cyclic terms
|dds|spin + cyclic terms
|uss|spin + cyclic terms
|dss|spin + cyclic terms
|sss|spin
U[1,1] (3)f U[1] (2)s is called baryon octet, and, with the action of SUf (3)
SUs (2),
is generated by the symmetrization of the vector |duu|
, where |
:= 23 |

16 (| + | ). This vector is the proton | p. Baryons are summarized
as Tables 5.5 and 5.6. However, 0 and 0 are not so simple. They have the same
weight, but they are given as orthogonal vectors in the following way.
1
|0 = (|uds |dus)(| | )
2 3
+ (|sud |sdu)(| | )

+ (|dsu |usd)(| | ) ,
1
| 0 =
(|uds |dus)(| + | 2| )
6
(5.132)
5.5 Quark Model
183
+ (|sud |sdu)(| + | 2| )

+ (|dsu |usd)(| + | 2| ) .
(5.133)
Hence, these two vectors give an example for an eigenspace Hm of a non-highest

weight m that is not one-dimensional, which is discussed in Sect. 4.3.3.
Therefore, the values of the quantum numbers I3 and Y are summarized as
Figs. 5.12 and 5.13.
Similarly, the anti-one-baryon system is given as U[0,0,0,0,3] (6). Due to Exercise 4.12, the system U[0,0,0,0,3] (6) is irreducibly decomposed as
U[0,0,0,0,3] = U[0,3] U[3] U[1,1] U[1] .
(5.134)
Exercise 5.15 Show that |n and |0 are orthogonal to each other even though they
have the same spin.
Exercise 5.16 Show that |0 , | 0 , and | 0 are orthogonal to each other even
though they have the same spin.
5.5.4 Charm Quark

The existence of a fourth quark had been discussed by several authors [18, 35], and
is named to be a charm quark c. The first charmed particle (a particle containing
a charm quark) was discovered in 1974 [11, 10]. A charm quark has the quantum
Fig. 5.12 Baryon octet
Fig. 5.13 Baryon decuplet
1 12 0
0
1
2
0
1
++
*+
*0
*0
I3
3
2
1
2
1
2
3
2
I3
184
numbers Q = 23 , I3 = 0, S = 0, and C = 1, where C is the new quantum number

and is taken to be 0 in other quarks. So, charm quark has the quantum number
Y = 23 . When we include a charm quark, the quantum system of one quark is C4
and has the fundamental representation of SU(4). The flavor space is spanned by
|u, |d, |s, |c, and the Dynkin index is [1, 0, 0] so that the system is U[1,0,0] . When
we include the freedoms of spin and color, the total one-quark system is written as
U[1,0,0] (4)f U[1] (2)s U[1,0] (3)c . under the representation of the group SUf (4)
|s , |c,
|d,
SUs (2) SUc (3). The flavor space of one-anti quark is spanned by |u,
and the total one-anti-quark system is written as U[0,0,1] (4)f U[1] (2)s U[0,1] (3)c .
A meson is composed of a quark and an anti-quark. Due to the color confinement
and Theorem 4.3, under the representation of the group SUf (4) SUs (2), the onemeson system is essentially given as (U[1,0,0] (4)f U[1] (2)s )(U[0,0,1] (4)f U[1] (2)s ).
Due to (4.54), the irreducible decomposition is
(U[1,0,0] (4)f U[1] (2)s ) (U[0,0,1] (4)f U[1] (2)s )
= (U[1,0,0] (4)f U[0,0,1] (4)f ) U[1] (2)2
s
= (U[0,0,0] (4)f U[1,0,1] (4)f ) (U[0] (2)s U[2] (2)s )
= U[0,0,0] (4)f U[0] (2)s U[0,0,0] (4)f U[2] (2)s
U[1,0,1] (4)f U[0] (2)s U[1,0,1] (4)f U[2] (2)s .
(5.135)
As U[1,0,1] (4)f is the adjoint representation, the representation space can be identified
with the Lie algebra (4). The dimensions are calculated as dim U[0,0,0] (4)f = 1 and
dim U[1,0,1] (4)f = 15. The total dimension is 8 8 = 64 = 1 1 + 1 3 + 15
1 + 15 3.
Since a baryon is composed of three quarks, the color confinement and Theorem 4.2 guarantee that our analysis is reduced to the analysis on the three-fold symmetric space of the 8-dimensional system U[1,0,0] (4)f U[1] (2)s . Hence, we need to
consider the larger special unitary group SU(8) Then, we denote the total one-quark
system by U[1,0,...,0] (8). Then, the one-baryon system is given as the irreducible space
U[3,0,...,0] (8) for SU(8). Due to Theorem 4.1, the system U[3,0,...,0] (8) is irreducibly
decomposed as
U[3,0,...,0] (8) = U[3,0,0] (4)f U[3] (2)s U[1,1,0] (4)f U[1] (2)s
(5.136)
in the sense of SUf (4) SUs (2). Since the dimensions are calculated as
= 20 and dim U[1,1,0] (4)f = 5432
= 20 (Exercises 4.9
dim U[3,0,0] (4)f = 543
32
32
1098
and 4.10). The total dimension is 32 = 120 = 20 4 + 20 2.
5.5 Quark Model
185
5.5.5 Six Quarks

In the current standard model, there are 6 quarks, i.e., we have top quark t and bottom
quark b in addition to up quark u, down quark d, charm quark c, and strange quark
s so that the flavor system is 6-dimensional. So, the total one-quark system and the
total one-anti-quark system are written as U[1,0,0,0,0] (6)f U[1] (2)s U[1,0,0] (3)c and
U[0,0,0,0,1] (6)f U[1] (2)s U[0,0,1] (3)c , respectively, under the representation of the
group SUf (6) SUs (2) SUc (3).
Since a meson is composed of a quark and an anti-quark, due to the color
confinement and Theorem 4.3, under the representation of the group SUf (6)
SU(2)s , the one-meson system is essentially given as (U[1,0,0,0,0] (6)f U[1] (2)s )
(U[0,0,0,0,1] (6)f U[1] (2)s ). Due to (4.54), the irreducible decomposition is
(U[1,0,0,0,0] (6)f U[1] (2)s ) (U[0,0,0,0,1] (6)f U[1] (2)s )
= (U[1,0,0,0,0] (6)f U[0,0,0,0,1] (6)f ) U[1] (2)2
s
= (U[0,0,0,0,0] (6)f U[1,0,0,0,1] (6)f ) (U[0] (2)s U[2] (2)s )
= U[0,0,0,0,0] (6)f U[0] (2)s U[0,0,0,0,0] (6)f U[2] (2)s
U[1,0,0,0,1] (6)f U[0] (2)s U[1,0,0,0,1] (6)f U[2] (2)s .
(5.137)
The dimensions are calculated as dim U[0,0,0,0,0] (6)f = 1 and dim U[1,0,0,0,1] (6)f =
35. The total dimension is 12 12 = 144 = 1 1 + 1 3 + 35 1 + 35 3.
Since a baryon is composed of three quarks, due to the color confinement and
Theorem 4.2, we need to consider the three-fold symmetric space of U[1,0,0,0,0] (6)f
U[1] (2)s , whose dimension is 12. That is, the one-baryon system is given as the irreducible space U[3,0,...,0] (12) for SU(12). Due to Theorem 4.1, the system U[3,0,...,0] (12)
is irreducibly decomposed as
U[3,0,...,0] (12) = U[3,0,0,0,0] (6)f U[3] (2)s U[1,1,0,0,0] (6)f U[1] (2)s
(5.138)
in the sense of SUf (6) SUs (2). Since the dimensions are calculated as
= 56 and dim U[1,1,0,0,0] (6)f = 765
= 70 (Exerdim U[3,0,0,0,0] (6)f = 876
32
3
141312
cises 4.9 and 4.10). The total dimension is 32 = 364 = 56 4 + 70 2. That
is, we have 364 baryons and 364 anti-baryons.
5.6 Uncertainty Relations

In this section, we discuss the uncertainty relation for various types of wave functions. Firstly, we discuss the uncertainty relation for wave functions on the real line
R. Then, to address wave functions on the one-dimensional and three-dimensional
spheres S 1 and S 3 , we prepare several knowledges for periodic function space and
Mathieu function. After this preparation, we derive the uncertainty relation for wave
186
functions on the one-dimensional and three-dimensional spheres S 1 and S 3 . The

uncertainty relation on the one-dimensional sphere S 1 is directly derived from Mathieu function. However, the uncertainty relation on the three-dimensional sphere S 3 is
more complicated because Mathieu function is defined only for the one-dimensional
system. To fill the gap, we employ the representation theory of SU(2), which is
mainly discussed in Lemma 5.8.
5.6.1 Uncertainty Relation on R

In the quantum system L 2 (R), it is impossible to determine simultaneously the values
of the position Q and the operator P, which equals the momentum when = 1. To
clarify this issue, we discuss the trade-off relation between P and Q defined
in (1.8). For this purpose, we introduce two operators a := 12 (Q + iP) and a :=
1 (Q
2
iP), which implies the relations Q =
1 (a
2
+ a ) and P =
(a
2
a ).
Now, we focus on Hermitian inner product (X, Y ) := 21 Tr(X Y + Y X ) defined

for two operators X and Y . We apply Schwarz inequality for the inner product to the
case with X = Q (Tr Q)I and Y = i(P (Tr P)I ).
Then, we obtain the KennardRobertson type uncertainty relation [63, 91]
1
1

P Q i Tr[P (Tr P)I, Q (Tr Q)I ] = .
2
2
(5.139)
Since the inequality x + y 2x y holds for x, y 0, we have

2 P + 2 Q 2 P Q 1.
The equality condition will be discussed in Sect. 7.1.2.
(5.140)

Q Q P
More precisely, the determinant of the matrix [Q : P] :=
Q P P
defined by the notations given in (1.10) satisfies

det [Q : P]
1
.
2
(5.141)
To show (5.141), we consider the Hermitian matrix

1
0
Tr[Q0 , P0 ]
Tr Q20 Tr Q0 P0
[Q
:
P]
+
=
0
Tr P0 Q0 Tr P20
2 Tr[P0 , Q0 ]

i
0 1
= [Q : P] +
,
2 1 0

J :=
187
where Q0 := Q Tr Q and P0 := P Tr P. Since J 0 (Exercise 5.17), we

have det [Q : P] 41 = det J 0.
Exercise 5.17 Show the matrix inequality J 0.
5.6.2 Periodic Function Space and Mathieu Equation*

To discuss the uncertainty relation to the wave functions on the one-dimensional and
three-dimensional and spheres S 1 and S 3 , we prepare several mathematical properties
on the space of periodic function. For a positive real number L, we introduce the
notations as follows.

L
dx
< ,
(5.142)
| f (x)|2
L 2p ((L , L]) := f | f (x + 2L) = f (x),
2L
L
where the subscript p expresses the periodic. As a generalization, we define the space
of anti-periodic functions

L a2 ((L , L]) :=

f | f (x + 2L) = f (x),
L
L
| f (x)|2

dx
< ,
2L
(5.143)
which is a subspace of L 2p ((2L , 2L]). Further, we denote the spaces of even functions and odd functions in L 2p ((L , L]) and L a2 ((L , L]) by L 2p,even ((L , L]),
2
2
L 2p,odd ((L , L]), L a,even
((L , L]), L a,odd
((L , L]), respectively. For any f, g
2
L p ((L , L]), we define the inner product as

f |g :=
L
L
f (x)g(x)
dx
.
2L
(5.144)
The subspaces L 2p,even ((L , L]) and L 2p,odd ((L , L]) (the subspaces
2
2
((L , L]) and L a,odd
((L , L])) are orthogonal to each other. Also, the two
L a,even
subspaces L 2p ((L , L]) and L a2 ((L , L]) are orthogonal to each other. Therefore,
the space L 2p ((2L , 2L]) can be written as L 2p,even ((L , L]) L 2p,odd ((L , L])
2
2
((L , L]) L a,odd
((L , L]).
L a,even
Now, we consider Mathieu equation:
d2
() + (a 2q cos(2))() = 0.
d2
(5.145)
A function satisfies the above equation if and only if the function is the eigenfunction of the differential operator P2 + 2q cos(2Q). The operator X (q) := P2 +
2q cos(2Q) preserves the subspaces L 2p,even (( 2 , 2 ]), L 2p,odd (( 2 , 2 ]),
188
2
2
L a,even
(( 2 , 2 ]), and L a,odd
(( 2 , 2 ]). Then, we denote the minimum eigenvalues in
2
2
L 2p,even (( 2 , 2 ]), L 2p,odd (( 2 , 2 ]), L a,even
(( 2 , 2 ]), and L a,odd
(( 2 , 2 ]) by a0 (q),
b2 (q), a1 (q), and b1 (q), respectively [109, Sect. 28.2]. We call their eigenfunctions
Mathieu functions ce0 (, q), se2 (, q), ce1 (, q), and se1 (, q). The eigenvalues
a0 (q), b2 (q), a1 (q), and b1 (q) satisfy the conditions [109, Sect. 28.2(v)]
a0 (q) = a0 (q), b2 (q) = b2 (q), a1 (q) = b1 (q).
(5.146)
When q < 0, the ordering relation a0 (q) < a1 (q) < b1 (q) < b2 (q) holds [109,
Sect. 28.2(v)].
According to the reference [109, Sect. 28.2(v)], let a0 (q) be the minimum value
of a having the solution in L 2 ((/2, /2]) of the above differential equation, and
b2 (q) be the minimum a having the odd solution in L 2 ((/2, /2]) of the above
differential equation. The solution with a0 (q) is Mathieu function ce0 (, q) and the
solution with b2 (q) is Mathieu function se2 (, q) [109, Sect. 28.2(vi)]. Further, for a
large q, the functions a0 and b2 have the following asymptotic expansions for a large
h as
1
a0 (h 2 )
= 2h 2 + 2h
4
5
2
2
b2 (h ) = 2h + 6h
4
1
3
8 2
5
2 h
2 h
9
45
8 2.
5
2 h
2 h
(5.147)
(5.148)
Further expansion is available in [109, Sect. 28.8].

For a small q, the functions a0 (q) and b2 (q) have the following asymptotic expansions as
1
7 4
a0 (q)
q
= q2 +
2
128
1
5
q 4.
b2 (q)
= 4 q2 +
12
13824
(5.149)
(5.150)
Further expansion is available in [109, Sect. 28.6].
5.6.3 Uncertainty Relation on S1 *

Now, we extend the uncertainty relation to the wave functions on the one-dimensional
sphere S 1 . A wave function on the one-dimensional sphere S 1 can be identified with
a periodic function of the period 2 on R. In this space, we can consider the pair of
operators (cos Q, sin Q). For a function L 2p ((, ]), instead of the uncertainty
2 Q, we focus on the uncertainty
189
2 (cos Q, sin Q) := 2 cos Q + 2 sin Q

= | cos2 Q| + | sin2 Q| | cos Q|2 | sin Q|2
= 1 | cos Q|2 | sin Q|2 .
Then, as the uncertainty relation, we consider the trade-off relation between
2 (cos Q, sin Q) and 2 P. For this purpose, we consider the minimum
U(1) (s) :=
=
min
L 2 ((,]):=1
min
|(I cos(Q)) + sP2 |
L 2 ((/2,/2]):=1
|(I cos(2Q)) +
sP2
|.
4
So, U(1) (s) can be characterized as the minimum value of U(1) having the solution
in L 2 ((/2, /2]) of the following differential equation.
s d2
() + (U(1) 1 + cos(2))() = 0,
4 d2
(5.151)
d2
() +
d2

4(U(1) 1) 4
+ cos(2) () = 0,
s
s
(5.152)
In order to find the minimum U(1) (s), we employ Mathieu equation (5.145), whose
detail is summarized in Sect. 5.6.2. Hence, using the function a0 given in Sect. 5.6.2,
we have U(1) (s) =
sa0 ( 2s )
4
+1 =
sa0 ( 2s )
4
+ 1. Now, we prepare the following lemmas.
Lemma 5.5 Given two self-adjoint operators X and Y on a Hilbert space H, we

consider the functions
(E) :=
(s) :=
inf
{|X ||Y E}
(5.153)
inf
|X + s|Y .
(5.154)
H:=1
H:=1
Assume that X 0 Then, (E) is convex and monotone increasing, and (s) is
concave and monotone increasing. When (E) > lim E (E) and E > 0, there
exists s E such that
(s E ) = E, where
is the derivative of . Then, we have
(E) = (s E ) s E E = max (s) s E.
s>0
(5.155)
Further, s E is positive and monotone decreasing with respect to E.

Lemma 5.5 will be shown in the end of this subsection. Applying Lemma 5.5 to
the case when X = I cos(Q) and Y = P2 , we obtain the following lemma.
190
Lemma 5.6 ([49]) We have the following relation:

U(1) (E) :=
{|I cos(Q)|||P2 | E}
min
L 2p ((,]):=1
= max
s>0
sa0 ( 2s )
+ 1 s E.
4
(5.156)
(5.157)
The minimum (5.156) is attained if and only if () = ce0 ( 2 , s2E ), where s E is

argmaxs>0
sa0 ( 2s )
4
+ 1 s E and the function ce0 is given in Sect. 5.6.2.
Using Lemma 5.6, we derive the trade-off relation between both uncertainties
2 (cos Q, sin Q) and 2 P as follows.
Theorem 5.4 ([49]) The minimum of the uncertainty 2 (cos Q, sin Q) under the
constraint for 2 P is calculated as
min
L 2p ([,]):=1
{2 (cos Q, sin Q)|2 P E}
sa0 ( 2s )
sa0 ( 2s ) 2
s E = max 1 s E
= 1 max
.
s>0
s>0
4
4
(5.158)
The minimum in (5.158) is realized by if and only if is given as a shift of the

Mathieu function ce0 ( 2 , s2E ).
Proof Due to the symmetry, without loss of generality, we can assume that
| sin Q| = 0 and |P| = 0. Hence, we obtain
min
{2 (cos Q, sin Q)|2 P E}
min
{1 | cos Q|2 ||P2 | E}
L 2p ([,]):=1
L 2p ([,]):=1
=1
max
L 2p ([,]):=1
2
{| cos Q|||P2 | E} .
Thanks to Lemma 5.6, we have

1
max
L 2p ([,]):=1
{| cos Q|||P2 | E}
2
2
sa0 ( 2s )
sa0 ( 2s )
+ 1 s E = 1 max
sE .
= 1 1 max
s>0
s>0
4
4
Since maxs>0
sa0 ( 2s )
4
+ 1 s E 1, s E
sa0 ( 2s )
4
0. Hence,
191
2
sa0 ( 2s )
sa0 ( 2s ) 2
1 max
s E = 1 min s E
s>0
s>0
4
4
2 2
sa0 ( s )
sa0 ( 2s ) 2
= 1 min s E
= max 1 s E
,
s>0
s>0
4
4
which implies (5.158). Then, the minimum uncertainty in (5.158) is realized by if

and only if is given as a shift of the Mathieu function ce0 ( 2 , s2E ).
sa ( 2 )
Now, for the asymptotic expansion of 1 (maxs>0 04 s s E)2 , we discuss the

behavior of U(1) (E) when E approaches infinity. By using the expansion (5.147)
for a0 , as s 0, U(1) (s) can be expanded to
U(1) (s)
=

s 2 2s + 2 2s 41
4

+1=
s
s
.
2 16
As is shown in Lemma 5.5, s E is decreasing as a function of E. Hence, when E is

1
large, solving the equation U
(1) (s E ) = E, we approximately obtain s E
= 8(E+1/16)
2.
Hence,
U(1) (E) = U(1) (s E ) s E E
=
sE
sE
sE E =
2
16

1
sE
sE E +
2
16
1
1
1
.
=
8(E + 1/16)
8E
128E 2
(5.159)
So, we have

1
sa0 ( 2s )
max
sE
s>0
4
2
1
1
= 2U(1) (E) U(1) (E)2

. (5.160)
=
4E
128E 2
Next, we consider the case when E is small. When s is large, by using the expansion (5.149) for a0 , U(1) (s) can be expanded to
U(1) (s)
=
s( 21 ( 2s )2 +
4
7 2 4
( ) )
128 s
+1=1
1
7
+
.
2s
32s 3
When E is small, solving the equation U

(1) (s E ) = E, we approximately obtain

sE
= 1 + 21
= 1 (1 + 21 E). Hence,
2E
128
2E
32
192
1
7
U(1) (E) = U(1) (s E ) s E E
sE E +
=1
2s E
32s E3

21
21
7 3
E
E
7 2 3
E 1+
E +
E2.
2E = 1 2E +
=1 1
32
32
32
16
2
2
(5.161)
Thus,

1
sa0 ( 2s )
sE
max
s>0
4
2
2
7
= 1 1 U(1) (E) = 1 2E + E 2 . (5.162)
4
Proof of Lemma 5.5 To show the convexity of , we prepare the following lemma.
Lemma 5.7 Le V be a two-dimensional system, and A and B be two Hermitian
matrices on V. For any density matrix S(V), there exists a normalized vector
V such that
Tr A = |A|, Tr B = |B|.
(5.163)
Proof of Lemma 5.7 We discuss only the case when A and B are not constant.
Otherwise, this statement is trivial. In this case, it is enough to show the case when
the eigenvalues of A and B are 1 and 1. In this case, we can choose a suitable basis
such that A and B can be written as sums of 1 and 3 , j are Pauli matrices defined
in (5.175) in the next subsection. When is 21 (I + a1 + b2 + c3 ), we choose the
2 + c3 ) such that a 2 + b 2 + c2 = 1. Then, we
pure state || := 21 (I + a1 + b
have (5.163).

Lemma 5.7 guarantees that
(E) =
inf
H:==1
{|X |||Y | E}
= inf {Tr X | Tr Y E}
(5.164)
Since (E) is decreasing and the limit lim E (E) 0 exists, when (E) >
lim E (E), we have
(E) =
inf
H:==1
{|X |||Y | = E}.
(5.165)
Now, for simplicity, we show the convexity for only for the case when the minimum
exists in (5.153). For two real numbers E 1 and E 2 , and a real number p [0, 1], we
assume that (E i ) > lim E (E) for i = 1, 2. Then, we choose 1 , 2 H such
that
(E i ) = i |X |i , i |Y |i = E i
(5.166)
193
So, we have
Tr( p|1 1 | + (1 p)|2 2 |)Y = pE 1 + (1 p)E 2
Tr( p|1 1 | + (1 p)|2 2 |)X = p(E 1 ) + (1 p)(E 2 ),
(5.167)
(5.168)
which implies
inf {Tr X | Tr Y E} p(E 1 ) + (1 p)(E 2 ).
(5.169)
So, is convex. Thus,

monotone increasing. We choose s E such that s E =
(E).
Now, we fix E 0 := sup E {E|(E) > lim E (E)}. So, the minimization in
(5.154) is restricted to the case when |Y | E 0 . Thus, for s > 0, we have
(s) = inf (E) + s E = inf (E) + s E,
EE 0
(5.170)
where the second equation follows from the relation (E) = (E 0 ) for E > E 0 .
Thus, is also concave. Generally, as the duality of Legendre transformation [92,
Theorem 26.6], any convex function f satisfies
f (x) = sup(inf
f (x
) + x
y) yx.
(5.171)
(E) = sup (s) s E.
(5.172)
So, we obtain
s>0
Due to the concavity of (s), there exists s E such that

(s E ) = E. So, we have
(E) = (s E ) s E E = max (s) s E.
s>0
(5.173)
Since
is monotone decreasing, s E is positive and monotone decreasing with respect
to E.
5.6.4 Uncertainty Relation on S3 *

Next, we consider the relation between the above results and the uncertainty relation
on wave functions on the three-dimensional sphere L 2 (S 3 ). For this issue, we focus
on the fact that the group SU(2) is isomorphic to the 3-dimensional sphere S 3 by
(g), x2 (g), x3 (g)), where x j (g) ( j = 0, 1, 2, 3)
the correspondence
g (x0 (g), x1
x0 + i x1 x2 + i x3
for g SU(2). That is, we have L 2 (SU(2)) =
is given as g =
x2 i x3 x0 i x1
194
L 2 (S 3 ). Then, we define the operator Q j as the multiplication of x j (g). For the set
of operators Q := (Q0 , Q1 , Q2 , Q3 ), we focus on the uncertainty
2 Q
:=
3
2 Q j
j=0
3
3

2
=
|Q j |
|Q j | 2
j=0
j=0
3

=1
|Q j | 2
i=0
for L 2 (SU(2)). Next, we define the momentum operator P j on L 2 (SU(2)) as

j
d (eit 2 g)
|t=0 ,
P j (g) := i
dt
(5.174)
where Pauli matrices j is defined as

1 :=

01
0 i
1 0
, 2 :=
, 3 :=
.
10
i 0
0 1
(5.175)
For the set of momentum operators P := (P1 , P2 , P3 ), we also consider the

uncertainty
2 P
:=
3
2 P j
j=1

3

j 2
j=1
3

j=1
2
j
F1

k k
k=0
2 2

+ 1 I k2 ,
where I k2 is the projection to the irreducible space with the highest weight 2k . Then,
as the uncertainty relation, we consider the trade-off between 2 Q and 2 P.
Theorem 5.5 ([49]) The minimum of the uncertainty 2 Q under the constraint for
2 P is calculated as

min
L 2 (SU(2)): =1
{2 Q|2 P E} = 1
1 sb2 ( 8s ) 2
min s E +
.
s>0
4
16
(5.176)
The minimum in (5.158) is realized by if and only if there exists g SU(2) such
that (g g,1 ,2 ) =
se2 ( 4 , s8 )
sin
, where s E is given in Lemma 5.9.
Further, the RHS of (5.176) has the asymptotic expansion

and 1 43 E + 85 E 2 as E 0.
9
16E
253
9 E 2 as E
Proof of Theorem 5.5 To prove Theorem 5.5, we prepare several steps. For a given
normalized function L 2 (SU(2)), we can choose an element g0 SU(2) such
195
that g0 |Q j | g0 = 0 for j = 1, 2, 3, where g0 (g) := (g0 g). Since

2 P = 2 g P, without loss of generality, we can assume that |Q j | = 0 for
j = 1, 2, 3. In this case, 2 Q = 1 |Q0 | 2 . Thus, we obtain
min
L 2 (SU(2)): =1
{2 Q|2 P E}
{2 Q|2 P E, |Q j | = 0, j = 1, 2, 3}
2

2 P E,
1 |Q0 |
=
min
|Q j | = 0, j = 1, 2, 3
L 2 (SU(2)): =1

2

2 P E,
|Q0 |
.
=1
max
|Q j | = 0, j = 1, 2, 3
L 2 (SU(2)): =1
(5.177)
min
L 2 (SU(2)): =1
Now, we consider another maximum

{ |Q0 | |2 P E}.
max
L 2 (SU(2)): =1
(5.178)
This maximum can be attained when |Q j | = 0 for j = 1, 2, 3 due to the

following reason. If not, we can choose g0 such that g0 |Q0 | g0 > |Q0 |
and g0 |Q j | g0 = 0 for j = 1, 2, 3. Thus, we have
max
{ |Q0 | |2 P E, |Q j | = 0, j = 1, 2, 3}
max
{ |Q0 | |2 P E}.
L 2 (SU(2)): =1
L 2 (SU(2)): =1
(5.179)
To discuss this maximum, we employ a parametrization of SU(2) as

g,1 ,2 :=
x0 + i x1 x2 + i x3
x2 i x3 x0 i x1

,
(5.180)
where x0 = cos 2 , x1 = sin 2 cos 1 , x2 = sin 2 sin 1 cos 2 , x3 = sin 2 sin 1 sin 2
with the range (2, 2], 1 ( 2 , 2 ], 2 ( 2 , 2 ].
Then, we employ the following lemma.
Lemma 5.8 We consider the Fourier transform of a normalized wave function
L 2 (SU(2)) as F[ ] =
k=0 k2 k2 , where k2 0 and 2k = 1. Then,
|Q0 | =
SU(2)
1
1 (g)| (g)|2 SU(2) (dg)
2 2
2
1
d
| k2 k
| =
cos |()|2 ,
2
2 k,k
=0
2
2
2
(5.181)
196

k+1
where () is defined as the odd function
k=0 k2 sin 2 .

k+1
When (g,1 ,2 ) sin 2 is written as
k=0 k2 sin 2 with k2 0, the equality
1
I k2 .
holds. This condition is equivalent to the condition that 2k is | k+1
This lemma will be shown in the end of this subsection. When is written as above
way,
2 P

k
k
k=0

1
+ 1 2k = |P2 | .
2
2 2
4
(5.182)
Since the function with the above form is an odd periodic function with period
4, we have
max
L 2 (SU(2)): =1
{ |Q0 | |2 P E}

Q
1
.
| cos ||P2 | E +
2
4
L 2p,odd ([2,2]):=1
max
(5.183)
To calculate the RHS of (5.183), we consider the minimum

Q

(I cos ) + sP2
2
2
L p,odd ((2,2]):=1

sP2

=
min
(I cos(2Q)) +
.
16
L 2p,odd ((/2,/2]):=1
SU(2) (s) :=
min
So, SU(2) (s) can be characterized as the minimum value of SU(2) having the solution
in L 2p,odd ((/2, /2]) of the following differential equation.
s d2
() + (SU(2) 1 + cos(2))() = 0,
16 d2
(5.184)

16(
16
d2
SU(2) 1)
+
cos(2)
() = 0.
()
+
d2
s
s
(5.185)
In order to find the minimum SU(2) (s), we employ Mathieu equation (5.145), whose
detail is summarized in Sect. 5.6.2. Hence, using the function b2 given in Sect. 5.6.2,
sb ( 8 )
sb ( 8 )
2 s
we have SU(2) (s) = 216 s + 1 = 16
+ 1. So, applying Lemma 5.5 to the case
2
when X = I cos(Q) and Y = P , we obtain the following lemma.
197
Lemma 5.9

Q
1

I cos |P2 | E +
2
4
L 2p,odd ((2,2]):=1
8

sb2 s
1
.
= max
+1s E +
s>0
16
4
max
(5.186)
(5.187)
The minimum (5.186) is attained and |Q j | = 0 for j = 1, 2, 3 if and only if

se2 ( ,
4
sE
(,1 ,2 ) = sin
, where s E is argmaxs>0
2
function se2 is given in Sect. 5.6.2.
sb2 ( 8s )
16
+ 1 s(E + 41 ) and Mathieu
This lemma shows that

Q
1
| cos( )|||P2 | E +
2
4
L 2p,odd ((2,2]):=1
min

1 sb2 ( 8s )
1.
= min s E +
s>0
4
16
(5.188)
The equality condition is the same as Lemma 5.9. Combining (5.177), (5.179),
(5.183) and (5.188), we obtain (5.176).
Next, we give the asymptotic expansion. By using the expansion (5.148) for b2 ,
as s 0, SU(2) (s) can be expanded to
SU(2) (s)
=

s(2 8s + 6 8s 45 )
16
+1=
3
2
5s
s
.
2 64
As is shown in Lemma 5.5, s E is decreasing as a function of E. Hence, when E
(s E ) = E + 14 , we approximately obtain s E
is large, solving the equation SU(2)
=
3
)2 . Hence,
2 ( 8E+21/8

1 3
1
5s E
sE E +
SU(2) (E) = SU(2) (s E ) s E E +
= sE
4
64
4
2 2

2 21
( 8 + 8E)
21
3
3
9
+E
2
= sE sE
=
64
2(8E
+
21/8)
8E
+
21/8
8
2 2
3
9
9
73
9
9
=
11 2 . (5.189)
=
=
2(8E + 21/8) 4(8E + 21/8)
4(8E + 21/8)
32E
2 E
198
Thus, we have
9
1 sb2 ( 8s ) 2
7 33 2
1 min s E +
11 2
=1 1
s>0
4
16
32E
2 E
3
9
5
(5.190)
9 2,
=
16E
2 E
which is the asymptotic expansion for the RHS of (5.176) when E approaches to
infinity.
Next, we consider the case when E is small. By using the expansion (5.150) for
b2 , when s is large, SU(2) (s) can be expanded to
SU(2) (s)
=
s(4
1 8 2
( )
12 s
+
16
(s)
When E is small, since SU(2)
=
5
( 8 )4 )
13824 s
+1=
s
1
5
+1
+
.
4
3s
54s 3
+ 3s12 18s5 4 solving the equation SU(2)

(s E ) =

1
1
5
1
5
E + 4 , we approximately obtain s E = 3E + 6 = 3E (1 + 4 E). Hence,

1
4

1
SU(2) (E) = SU(2) (s E ) s E E +
4

1
1
1 5 18s E2
sE
5
E
+
sE E
+
1
=
1
+
+
s
=
E
4
3s E
4
3s E
54s E3
18s E2

1

+ 56
5 18 3E
5
1
1

1
E
E
1
+
= 1 +
3E
4
18 3E + 56
1
5
1+ 4E
3
3E

1 + 5E
5
E
E
3

=1
E
1
+
3
4
1 + 5E
3 1 + 54 E
2

5
5E
5E
5
E
E
1+ E
=1 1 E +
4
3
2
3
4
3

25
2
5
E
E
5
3
E
1+ E =1
=1 1
E + E 2 (5.191)
12
3
4
3
3
6 3
That is,
2
5
3
E E2.
1 SU(2) (E)
=
3
6 3
(5.192)
Thus, we obtain the asymptotic expansion for the RHS of (5.176) with E 0 as
8
4
1 (1 SU(2) (E))2
= 1 E + E 2.
3
5
(5.193)
199
Now, our Proof of Theorem 5.5 is completed.

Proof of Lemma 5.8 Since x0 (g) = 21 21 (g), we have
|Q0 | =

SU(2) (dg)
SU(2)

=
SU(2) (dg)
SU(2)
SU(2) (dg)
SU(2)
,
G[L]

k,k
k,k
| (g)|2
2
21 (g)

Tr f k2 (g) k2 d k2 k2

2
21 (g)
2
k2 k
Tr f (g) f
(g)(k + 1) k2 (k
+ 1) k

k2 k
(k + 1)(k
+ 1)
2
Tr
G[L]

=
21 (g)
SU(2) (dg)
21 (g)
f (g) f
(g)
k + 1 k2 k
+ 1 k
2

k2 k
(k + 1)(k + 1) Tr k,k
k + 1 k2 k
+ 1 k
SU(2)
2
(5.194)
hold, where we use the matrix

k,k
1
:=
21 (g)f (g) f
(g)G (dg) =
2
G
1
I 21 when|k k
| =
0 otherwise.
4
1
2
(5.195)

The last equation follows from (4.22) and the formula SU(2) 2k (g)f k
(g)SU(2) (dg) =
2
k k

, 2
1
2
k
k
SU(2) 2 (g)f k2 (g )SU(2) (dg) = k+1 I 2 , which is a special case of (3.77).
Hence, applying Schwarz inequality
with
respect to
the inner product A, B :=
Tr k,k
A B, to the case A := I k2 k
+ 1 k
, B := k + 1 k2 I k
, we obtain
2
Tr k,k
k + 1 2k k
+ 1 k
2

Tr k,k
(k + 1) k k2 Ik
Tr k,k
I (k
+ 1) k
k
.
2
Since k,k
is invariant with respect to the action of G,
(5.196)
200
Tr k,k
(k + 1) k k2 I k
2
2

= Tr k,k
(f (g) f k
(g))(k + 1)( k k2 I k
)
2
SU(2)
= Tr k,k
I k2 I k
=
2
(f 2k (g) f k
(g)) SU(2) (dg)
1
when|k k | =
2
0 otherwise,
1
2
(5.197)
where we used the condition Tr k 2k = 1.

2
Similarly, we have
Tr k,k
I k2 d k
k
k
=
2
1
when|k k
| =
0 otherwise.
2
1
2
(5.198)
Combining (5.196)(5.198), we have
Tr k,k
k + 1 k2 k
+ 1 k

2
1
when |k k
| =
0 otherwise.
2
1
2
Hence, combining the above relation with (5.194), we obtain (5.181). Due to
the equality condition for Schwarz inequality, the equality in (5.181) holds when
1
F[ ] =
k=0 k2 | k+1 I k2 .

k+1
When (g,1 ,2 ) sin 2 is written as
k=0 k2 sin 2 with k2 0, we have
cos | (g,1 ,2 )|2 sin2

2
2
2
2
SU(2)

2

2
d

k + 1 2 d

=
(5.199)
=
cos
k2 sin
cos |()|2 .
2 k=0
2
2
2
2
2
2
x0 (g)| (g)|2 SU(2) (dg) =
Chapter 6
Representation of General Lie Groups

and General Lie Algebras
Abstract Based on root systems, this chapter deals with representations of a Lie
group via those of the corresponding Lie algebra. This chapter discusses a noncompact Lie group as well as a compact Lie group. To discuss both, we focus on the
relation between a (real) Lie algebra and a complex Lie algebra. Since representations
of a non-compact group are very complicated, this chapter treats only a part of its
representations that are related to quantum theory. As such special representation
has analogies with representations of a compact Lie group, they can be more easily
understood than the general case. Since this chapter is composed of very advanced
topics and such sections are labeled with *, the reader can omit this chapter in the
first time.
6.1 Root System and Its Classification*

The previous section discusses specific types of Lie groups and Lie algebras. This
section deals with Lie groups and Lie algebras in more general forms. In Sect. 4.3,
we have discussed an irreducible skew-Hermitian representation of su(r ) based on
a root system. In the following, based on a root system, we discuss compact semi
simple Lie algebras in the general setting. For this purpose, we give the definition of
a root system in the general form, and discuss its properties.
Definition 6.1 A subset of a real linear space V with a positive definite inner
product (, ) is called a root system when the following conditions hold.
(R1) || < and the set does not contain 0 and spans V .
(R2) For an element , the element is contained in , but 2,
3, . . . are not contained in .
for two elements and of h0 .
(R3) We define the bilinear form , := 2(,)
(,)

Then, , is an integer for any elements , .
(R4) The map W () := , is closed with respect to .
Here, we should remark that , is not necessarily symmetric. For a root system
, the set W := {W } of maps forms a group, and is called the Weyl group.
The following theorem is known for a root system (Exercise 6.4).
DOI 10.1007/978-3-319-44906-7_6
201
202
6 Representation of General Lie Groups
Theorem 6.1 For a root system , there exists a subset = {1 , . . . , r } satisfying

the following conditions. The subset is called a fundamental system or a root basis.
(B1) forms a basis of the linear space V .
(B2) For any element , we have = rj=1 m j j , where all of m j are
positive integers or negative integers.
Definition 6.2 For a root system V and its fundamental system = {1 , . . . ,
r }, an element of is called a simple root. An element of is called a positive
root(negative root) when it is a linear sum of with positive (negative) coefficients.
Their set is denoted by + ()( ()). When there is no possibility for confusion,
we omit ().
The vector expression [, 1 , . . . , , r ] of V is called Dynkin index
(under the fundamental
system ). For two vectors and , we write when
the relation = rj=1 c j j holds with c1 , . . . , cr 0. The relation satisfies
the condition for partial order.
Then, we have the following proposition. (Exercise 6.4)
Proposition 6.1 For a given root system V , we choose an element V
satisfying (, )
= 0 for . Then, we can choose a fundamental system
satisfying
+ = { |(, ) > 0}.
(6.1)
Conversely, given a fundamental system , there exists an element V satisfying

(6.1).
For a fundamental system = { j }rj=1 of a root system , the matrix M :=
(l , l ) is called Cartan matrix. The Dynkin index of the j-th element j of the
fundamental system is given as j-th row vector in Cartan matrix. Then, Conditions
(R1-4) guarantees that the real number , takes one of 0, 1, 2, and 3 for
any two elements , (Exercise 6.1). In particular, when
= are elements
of the fundamental system, since the relation (, ) 0 is known (Exercise 6.5),
the real number , takes one of 0, 1, 2, 3. Hence, any diagonal entry of
Cartan matrix is 2, and its non-diagonal entry takes one of 0, 1, 2, 3.
In the following, we study Cartan matrix. When the real number l , l is 2
or 3 for l
= l , it is known that l , l is 1 (Exercise 6.1). For the classification
of Cartan matrix, we connect each point l based on the following rule. When
l , l = l , l = 1, we connect l and l with one line. When l , l =
2 and l , l = 1, we connect l and l with two lines, and write the inequality
> on the two lines to clarify that l is larger. This condition implies that (l , l ) >
(l , l ). When l , l = 3 and l , l = 1, we connect l and l with
three lines, and write the inequality > on the three lines to clarify that l is larger.
This diagram is called a Dynkin diagram.
In particular, when the all points are connected via other points, the Dynkin
diagram is called connected. When the Dynkin diagram corresponding to a root
system is connected, the root system is called connected.
6.1 Root System and Its Classification*
203
Ar ( r 1) :
Br ( r 2) :
C r ( r 3) :
Dr ( r 4) :
E6 :
E7 :
E8 :
F4 :
G2 :
Fig. 6.1 Dynkin diagram: One line between l and l means l , l = l , l = 1. Two lines
between l and l mean l , l = 2 and l , l = 1, in which, we write the inequality > on the
two lines because l is larger. Three lines between l and l mean l , l = 3 and l , l = 1,
in which, we write the inequality > on the three lines because l is larger
Proposition 6.2 There exists a root () in a connected root system such that
() for any root . Such a root () is called the maximum root
of .
For its proof, see Exercise 6.6. A connected Dynkin diagram is limited to one of
Ar , Br , Cr , Dr , E 6 , E 7 , E 8 , F4 , and G 2 , which are given in Fig. 6.1 [60, Theorem in
Sect. 11.4]. Especially, as shown in Fig. 6.1, we have A1 = B1 = C1 , B2 = C2 , and
A 3 = D3 .
Cartan matrix M = (l , l ) of Ar is the matrix given in (4.41).
rWhen l :=

have

,

=
,
which
implies
the
relations
=
M 1 l , we
l
l
l.l
l=1 l , l

and = rl=1 , l l . Especially, given two elements and in h, we write
when the relation l , l , holds for l = 1, . . . , r . Then, we define
two sets () and + () as follows.
() := { h |, l Z, l }
+ () := { ()|, l 0, l }.
(6.2)
204
Exercise 6.1 ([60, Table 1 of Sect. 9.4]) Show that the pair (, , , ) is limited
to (0, 0), (1, 1), (1, 2), (1, 3), and (2, 2) for , with double
sign in same order when . Show that the pair (2, 2) is limited to the
case when = .
Exercise 6.2 Assume that , satisfy
= . Show the following items.
(Use Exercise 6.1.)
a: When (, ) > 0, .
b: When (, ) < 0, + .
Exercise 6.3 Let r be the dimension of V . Let V + () := { V |(, ) > 0}. For
any r vectors 1 , . . . , r , show that rj=1 V + ( j ) is not empty.
Exercise 6.4 ([60, Sect. 10.1]) For a vector V , we define P := { V |
(, ) = 0} and + () := { |(, ) > 0}. For a vector V \ P , we
call + () decomposable when there exist 1 , 2 + () such that =
1 + 2 , indecomposable otherwise, and we denote the set of indecomposable roots
by (). Show the following items for a vector V \ P .
a: Any element + () is written as non-negative Z-linear combination of ().
(Use contradiction.)
b: When , (), (, ) 0 or = . (Use contradiction and Exercise 6.2.)
c: () is a liner independent set. (Use b and contradiction.)
d: () is a base of .
e: When is a base of , there exists a vector V \ P such that =
( ) and + = + ( ). (Use Exercise 6.3.)
Exercise 6.5 ([60, Sect. 10.1]) Assume that is a fundamental system of and

= . Show that (, ) 0 and is not a root. (Use contradiction and
Exercise 6.2.)
Exercise 6.6 ([60, Lemma A in Sect. 10.4]) Assume that is a connected root
system. Show Proposition 6.1 by showing the following items for , .
a: When there is no element such that , we have + .
b: When there is no element such that , we have (, ) 0 for
and there exists an element such that (, ) > 0. (Use contradiction and
Exercise 6.2.)
c: When
there is no element such that , the root is written as =

r with r > 0. (Use contradiction, a, b, (B2), and Exercise 6.5.)
d: When there is no element such that or , we have = .
(Use contradiction, b, c, and Exercise 6.2.)
6.2 Structure of Compact Lie Algebra*
205

In this section, we focus on a compact Lie group G and its compact Lie algebra G
without any condition for G and g. An irreducible unitary representation of the compact Lie group G can be given from a skew-Hermitian representation of its compact
Lie algebra G. For analysis of a real Lie algebra g, we often focus on its Lie subalgebra h satisfying [X, Y ] = 0 for any two elements X, Y h. Such a Lie subalgebra
h is called a Cartan subalgebra of g. Especially, a Cartan subalgebra h is called
a maximal Cartan subalgebra when a Cartan subalgebra including h is limited
only to h itself. There exist more than one maximal Cartan subalgebras, however, its
dimension does not depend on the choice of the maximal Cartan subalgebra, and is
called the rank of the Lie algebra g.
The structure of the Lie algebra mainly is decided by the maximal Cartan subalgebra h and its relation to the Lie algebra. Especially, when the Lie algebra g is compact,
all of maximal Cartan subalgebras are isomorphic to each other. As mentioned in
(3.47), the compact Lie algebra g can be decomposed to the direct sum of the semi
simple part g(1) := [g, g] and the commutative part c(g) = {X g |ad(X ) = 0}. Letting h be a maximal Cartan subalgebra of the semi simple part g(1) , we find that a
maximal Cartan subalgebra of g is given as h c(g). Then, the structure of the compact Lie algebra g can be characterized by that of the compact semi simple Lie algebra
g(1) as follows.
In the following, based on the root system, we investigate the structure of the
compact semi simple Lie algebra. Let g be a compact real semi simple Lie algebra
and h be its maximal Cartan subalgebra. Then, for an element of the dual space
h of h, we choose an element Z h such that
(X ) = (X, Z )g , X h .
(6.3)
Using the element Z , we define a positive definite inner product of h as

(, ) := (Z , Z )g = (Z ), , h .
(6.4)
Since all elements of h are commutative with each other, we can choose a basis of g
such that all elements X h simultaneously take the form given in (3.46) of Lemma
y
3.9. Hence, there exist elements {t } h and the basis {Ftx , Ft }t of the orthogonal
complement of h such that
y
ad(X )Ftx = t (X )Ft , ad(X )Ft = t (X )Ftx , X h,
(6.5)
y
ad(X )(Ftx + i Ft ) = it (X )(Ftx + i Ft ), X h .
(6.6)
206

y
Here, we remark that when Ft is replaced by Ft , instead of the element t , the

element t satisfies the above condition. In the following, we denote the set of t
and t by g .
Theorem 6.2 When g is a compact semi simple Lie algebra, the relation t
= t
holds for t
= t .
Theorem 6.3 The set g is a root system when g is a compact semi simple Lie
algebra. Conversely, for any root system , there exists a compact semi simple Lie
algebra g such that = g .
Theorems 6.2 and 6.3 follow from Theorems 6.6 and 6.7, which will be given
later.
Since g is a root system due to the above theorems, we choose a fundamental
y
y
system of the root system. When t g+ (), we replace Ft by Ft . Then,
+
g () equals the set {t }t . Hence, without loss of generality, we can assume that
y
g+ () = {t }t by choosing the proper sign of Ft . In the following, using the
y
y
x
x
element t of , we denote {Ft , Ft } by {Ft , Ft }. Then, the relations
[Fx , Fy ] = (Fx , Fx )g Z = (Fy , Fy )g Z
(6.7)
hold, and the subspace generated by {Fx , F , Z } is a Lie subalgebra isomorphic to

y
y
su(2). Since the relation (6.7) implies that (Fx , Fx )g = (F , F )g , we choose Fx ,
y
F such that
(Fx , Fx )g = (Fy , Fy )g =
1
1
=
.
(, )
(Z , Z )g
(6.8)
Hence, we have
[Fx , Fy ] =
1
Z.
(, )
(6.9)
y
x
g
g
Thus,
letting [] be the linear space generated by {F ,x F },y we have = h
( + g[]). Since the set {Z 1 , . . . , Z r } + {F , F } is a basis of the
linear space g, it is called a basis of the compact semi simple Lie algebra g. These
equations express the structure of the compact semi simple Lie algebra g. Further,
the relation (6.6) and Jacobi law imply
ad(X )[Fx + i Fy , Fx + i F ] = i( + )(X )[Fx + i Fy , Fx + i F ]

y
for any element X h. Describing the real coefficient parts and the pure imaginary
coefficient parts, separately, we obtain the following theorem.
Theorem 6.4 When + + , the elements [Fx , Fx ] [F , F ] and [Fx ,
y
y
y
x
F ] + [F , Fx ] are constant times of F+
and F+ under the above basis.
When +
/ + , both of the above elements are 0.
y
207
On the other hand, for a compact Lie algebra g, we choose a maximal Cartan
subalgebra h and
+ for a compact semi simple Lie algebra [g, g]. Then, we obtain
g = h c(g) ( + g[]). Hence, the analysis of a compact semi simple Lie
algebra can be reduced to the analysis of the root system. Given a fundamental
system , Weyl group W is generated by the subset {W } , and the image of
with respect to W is . Hence, all informations for the root system are contained in
the fundamental system. For the analysis of the fundamental system, it is sufficient
to analyze Cartan matrix. So, the analysis of the compact semi simple Lie algebra is
reduced to that of Cartan matrix. Since the classification of Cartan matrices can be
reduced to that of Dynkin diagrams, the classification of compact semi simple Lie
algebras can be reduced to that of Dynkin diagrams. Then, the following theorem
holds.
Theorem 6.5 A compact semi simple Lie algebra g is simple if and only if its Dynkin
diagram is connected.
The proof of this theorem will be given in the next section because it is easier to
show the same argument with its complexified Lie algebra. The connected Dynkin
diagrams Ar , Br , Cr , and Dr given in the previous section are called classical type,
they are the Dynkin diagrams of the root systems of su(r + 1), so(2r + 1, R), sp(r ),
and so(2r, R), respectively.
The remaining connected Dynkin diagrams E 6 , E 7 , E 8 , F4 , and G 2 are called
exceptional type, the corresponding Lie algebras are given as Lie subalgebras of
u(r ). The structure of the classical type of root systems are summarized in Table 6.1
[66, pp. 509510].
Table 6.1 Root systems of Ar , Br , Cr , Dr : indices j, j , and l of positive roots are chosen as j =
1, . . . , r 1, j = 1, . . . , r, l = j + 1, . . . , r . The number j of simple root j ( j = 1, . . . , r )
corresponds to the number in Fig. 6.1
Lie algebra
Simple root
Positive
Norm
Number Dimension
j, j =
root
of e j
of roots
1, . . . , r 1 r
Ar
su(r + 1)
e j e j+1 ,
er er +1
Br
so(2r + 1, R)
e j e j+1 ,
er
Cr
sp(r )
e j e j+1 ,
2e2r
Dr
so(2r, R)
e j e j+1 ,
er 1 + er
e j el
e j el ,
e j ,
e j + el
e j el ,
e j + el ,
2e2 j
e j el ,
e j + el
1
2(r +1)
1
2(2r 1)
1
2 r +1
1
2(2r 1)
r2 + r
r 2 + 2r
2r 2
2r 2 + r
2r 2
2r 2 + r
2r 2
2r 2 + r
208
In the following, we discuss the relation between the elements of the root system Ar and the elements of the Lie algebra su(r + 1). Since the root system is
, the set {c} also satisfies the
determined by the real numbers , = 2(,)
(,)
conditions for a root system. However, the root system generates a compact simple
Lie algebra, it needs to satisfy the condition (6.5) under the inner product given in
(6.4), in which, we adopt the Killing form as the inner product of the space h. Taking
account into this condition, we can uniquely determine the inner product of h and
the root system corresponding the compact simple Lie algebra. By using j,l h
x
defined in (4.38), the root system is given as { j,l , j,l }1 j<lr +1 . Since F j,l
,
y
z
F j,l , and F j,l given in Sect. 4.3 satisfy
y
x
x
(F j,l
, F j,l
)su(r +1) = (F j,l , F j,l )su(r +1) =
1 z
(F , F z )su(r +1) = (r + 1),
4 j,l j,l
y
z
1
x
satisfies (6.3). Since ( j,l , j,l ) = r +1
, Fx j,l := F j,l
and F j,l :=
Z j,l := 2(r1+1) F j,l
y
F j,l satisfy (6.8). Hence, when h is regarded as a subspace of the r + 1-dimensional

real linear space Vr with the CONS {e j }1 jr +1 , we obtain Table 6.1.

Exercise 6.8 Assume that a fundamental
system has more than two connected
components 1 , 2 , . . .. Define i+ := { i m } + . Show that i + j
(i
= j) is not contained in the root system for any element i i+ .
6.3 Structure of Complex Semi Simple Lie Algebra*

6.3.1 Complex Semi Simple Lie Algebra
The analysis of a complex semi simple Lie algebra g is easier than that of a compact
semi simple Lie algebra. In the case of complex Lie algebra, a Cartan subalgebra
is defined as a complex Lie subalgebra whose elements are commutative with each
other and are diagonalizable under the adjoint representation. Similarly, we define
its maximal Cartan subalgebra h and the rank of the complex semi simple Lie
algebra g. Given any element of the dual space h of h and any element X h,
we define E h such that (X ) = (X, E ), and define the inner product (, ) :=
(E , E ) = (E ) for two elements and of h .
Since {ad(X )} X h is simultaneously diagonalizable, we choose a basis {K t } of
the orthogonal complementary space to h among eigenvectors of {ad(X )} X h and
choose an element t of the dual space h such that ad(X )K t = t (X )K t for any
t. Such an element t is called a root, and we denote the set by g . The following
theorem holds.
Theorem 6.6 ([60, Proposition (a) of Sect. 8.4]) When g is a complex semi simple
Lie algebra, we have t
= t for t
= t .
209
The following theorem also holds for the set g .

Theorem 6.7 g is a root system for a complex semi simple Lie algebra g. Conversely, for an arbitrary root system , there exists a complex semi simple Lie algebra
g such that = g .1
Due to Theorem 6.6, for an element of g , there uniquely exists K such that
ad(X )K = (X )K , X h .
(6.10)
So, we can uniquely choose E such that

2
.
(E , E )g
(6.11)
2
E.
(E , E )g
(6.12)
(K , K )g =
Then,
[K , K ] =
Then, n+ :=< {K } + > and n :=< {K } + > are nilpotent Lie subalgebras. We obtain the Gauss decomposition g = n+ h n of the Lie algebra g.
It is known that a similar fact holds for the corresponding Lie group G. Since the
set {E } {K } is a basis of the the complex semi simple Lie algebra g, it
is called a basis of the complex semi simple Lie algebra g. Using
complex vectors
} + , we can define the Lie groups N+ := {exp( + K )}
{ } + and {
and N := {exp( K )} that correspond to the nilpotent Lie subalgebras,
where is a complex number for any + and { } + is a complex vector
with | + | entries. Now, we consider the right hand side action of N+ and the left
hand side action of N . Since both actions are commutative with each other, we can
define the quotient space for the combined action. Then, h gives
a set of representatives of the tangent space of the quotient space. So, {exp( rj=1 E j )} gives
a set of representatives of the quotient space. Hence, as a generalization of (4.2),
for an element g G except for measure 0 points, we can choose complex vectors
{ } + , { } + , and { j }rj=1 such that [87, p. 474]

g = exp

K

r

E j exp
K ,
exp
j=1
(6.13)
which is called the Gauss decomposition of the Lie group G.
1 The
former part is shown in [60, Theorem in Sect. 8.5], and the latter part can be checked by
considering the fact that the corresponding simple Lie algebras are given for all of simple root
systems in [60].
210
6.3.2 Relation to Compact Semi Simple Lie Algebra

From now, to show Theorem 6.5, we denote a compact semi simple Lie algebra by
g, and its complexification by gC , which is a complex semi simple Lie algebra. For
an element g+ , the elements K and E gC are given from the elements Fx ,
y
F , and Z of g as follows
E = i Z ,
K = i Fx Fy ,
K = i Fx + Fy .
(6.14)
The relations (6.5), (6.7) and (6.9) guarantee that the above given elements K and
E satisfy (6.10), (6.11) and (6.12).
Now, define the one-dimensional linear subspace g generated by K . (Remark
that g is different from g[].) Then, we have g = hC g . Theorem 6.4
yields
[g , g ] =
g+ +
/
{0} +
(6.15)
for
= . When = , we have
[g , g ] hC .
(6.16)
Conversely, based on the basis {E } {K } of the complex semi simple

Lie algebra, we define
Z := i E ,
Fx :=
i
(K + K ),
2
y
Fy :=
1
(K + K ),
2
the real linear space generated by Fx , F , Z forms a compact semi simple Lie
algebra. That is, there exists a one-to-one correspondence between complex semi
simple Lie algebras and compact semi simple Lie algebras. In the next two sections,
using this relation, we investigate the properties of a compact semi simple Lie algebra.
Here, we give the proof of Theorem 6.5. Firstly, we remark that given a compact
semi simple Lie algebra g, the compact semi simple Lie algebra g is simple when
its complexification gC is simple. Then, K is an eigenvector of the maximal Cartan
subalgebra h under the adjoint representation of the semi simple complex Lie algebra
gC . The action of K on the semi simple complex Lie algebra gC exchanges E
and K under the identification of the scalar times. Hence, for + , the subspace
spanned by {E , K , K } is included in the invariant subspace of the semi simple
complex Lie algebra gC with respect to the adjoint representation. Further, when
elements , + satisfy = + + , K and K are exchanged to each
other under the adjoint representation. Since, similarly, K and K are exchanged
to each other, these three elements are exchanged to each other.
We choose two elements , of the fundamental system such that ,
=
0. Since , is a non-negative integer, we have , + , which
211
implies that K and K are exchanged to each other. Hence, the complex Lie algebra corresponding to a connected Dynkin diagram does not have a non-trivial invariant subspace with respect to the adjoint representation. That is, the corresponding
complex Lie algebra is simple. Hence, the corresponding compact Lie algebra is
simple.
On the other hand, we assume that the corresponding fundamental system has
more than two components that are not connected to each other. Then, due to Exercise
6.8, the element i + j is not contained in for any element i i+ . Now, let
g C be a complex Lie subalgebra generated by one connected component of the
corresponding fundamental system. K maps any elements of g C to 0. Similarly,
E maps them to 0. Hence, the complex Lie subalgebra g C is an invariant subspace
of the complex Lie algebra gC with respect to the adjoint representation. Thus, the
corresponding compact Lie subalgebra g is also an invariant subspace of the compact
Lie algebra g with respect to the adjoint representation.
These facts indicate the equivalence relation among the following three arguments;
(1) The semi simple complex Lie algebra gC is simple. (2) The semi simple compact
Lie algebra gC is simple. (3) The corresponding Dynkin diagram is connected. Hence,
we obtain Theorem 6.5.
6.4 Skew-Hermitian Representation of Compact Semi

Simple Lie Algebra*
Next, to study a unitary representation of an arbitrary compact Lie group, we investigate a skew-Hermitian representation of the compact simple Lie algebra g. Similar to
representations of su(2) and su(d), for an irreducible skew-Hermitian representation
f of g on H, we define
E j := if(Z j ), j g , K := if(Fx ) f(Fy ), g .
Considering the complexification gC , we have E j = f(E j ) for j g and K =
f(F ) for g . Then, the relations (6.10) and (6.12) imply
[E j , E j ] = 0
[E j , K ] = (Z j )K
2
Ej.
[K j , K j ] =
( j , j )
(6.17)
(6.18)
Then, since the representation is skew-Hermitian, the relation (K ) = K holds

and E j is Hermitian. Hence, due to (6.17), H is decomposed to the direct sum space of
the simultaneous eigenspaces with respect to elements of h. The eigenvalues can be
written by using an element m of the dual space h of h. The simultaneous eigenspace
corresponding to the eigenvalues given by m is given as Umw := {v H|f(X )v =
212
i m(X )v, X h}. The condition given in the set Umw is equivalent to the condition
E j v = m(Z j )v. An element m h is called a weight for a representation f on a
representation space H when Umw
= {0}. Then, due to (6.18), K v is an element of
w
Um+
for v Umw . The action of an element e X G (X h) satisfies
f(e X )v = ei m(X ) v, v Umw .
(6.19)
When H is an irreducible skew-Hermitian representation space, we can uniquely

choose a weight such that
(1) Uw is a one-dimensional space.
(2) An element |; of Uw satisfies K |; = 0 for + .
Such a weight is called a highest weight. Especially, a vector |; satisfying
Condition (2) is called the highest weight
vector. The representation space with the

highest weight is generated by { + (K )l v } and is uniquely determined
by the highest weight . Hence, the highest weight is regarded as the label (i.e.,
an element of g ) to identify the irreducible skew-Hermitian representation. So, we
denote the representation space by U , and the representation on U by f .
There exists a skew-Hermitian representation with the highest weight if and
only if all of entries of Dynkin index of are non-negative integers , j , i.e.,
+ . Since the fundamental system is given as { j }, the definition of + ()
given in (6.2) guarantees that the latter condition is equivalent to the condition
+ (). That is, + () can be identified with the set g of irreducible skew-Hermitian
is also an irreducible representation space under
representations. In particular, g C
the adjoint representation, and its highest weight is the maximal root () given in
Proposition 6.2.
To characterize the set () := { h |Uw
= {0}}, we define the following map
as a generalization of the map W .

W,k () :=
k when k is an integer between 0 and , ,
otherwise.
y
= 2([Fx , F ]) and the Lie subalgebra generated by

Note that , = 2(,)
(,)
y
y
x
x
{F , F , [F , F ]} is isomorphic to su(2). Since the action of the Lie subalgebra
characterizes the action of W,k , we find that the set () is closed with respect to
W,k . Hence, the set () is {W1 ,k1 W2 ,k2 Wn ,kn ()} j ,k j Z . Further,
the dimension of U is characterized by the following Weyls dimension formula
[60, Corollary in Sect. 24.3];

dim U =
+ (
+ , )
=
(,
)
+
+
+ ,
,
,
(6.20)

where := 21 + . By using j defined in Sect. 6.1, can be written as =

r
j=1 j [60, Lemma A in Sect. 13.3].
6.4 Skew-Hermitian Representation of Compact Semi Simple Lie Algebra*
213
When g is a compact real semi simple Lie algebra, the root system can be decomposed to connected components 1 , . . . , l and each connected component j is a
root system. Let j be the fundamental system of j . Then, the set of weights of g
is given as + (1 ) + (l ). Further, due to Lemma 3.3, the set g of irreducible skew-Hermitian representations is given as + (1 ) + (l ). When
g is a compact Lie algebra, g can be decomposed to the direct sum of the compact
real semi simple Lie algebra [g, g] and the center c(g). Since any irreducible representation space of the center c(g) is one-dimensional, an irreducible representation
space of g is also an irreducible representation space of the compact real semi simple
Lie algebra [g, g]. Thus, an irreducible skew-Hermitian representation of g can be
determined by the action of [g, g] on the irreducible representation space U and the
scalar action of the center c(g) on the space U .
1
j (r j) j in the case of SU(r ).
Exercise 6.9 Show that = 21 rj=1
Exercise 6.10 Show that (6.20) equals (4.47) in the case of SU(r ).
6.5 Subalgebra and Root Subsystem*

When a homogeneous space has a symmetry with respect to the action of the Lie
group G, the homogeneous space is written as the quotient space G/H1 by using the
stabilizer H1 , which is a subgroup of the group G. Hence, to investigate the structure
of the homogeneous space, we need to study the relation between the group G and
the subgroup H1 . This discussion plays an important role for the generalization of
the concept of coherent states. In the following, we focus on the case when H1 is a
Lie subgroup. However, since it might be difficult to study the structure of the Lie
group directly, we study that of the corresponding Lie algebra. Hence, this section
deals with the compact real semi simple Lie algebra g and its Lie subalgebra h1 .
This section discusses only a compact real semi simple Lie algebra, however, this
discussion works as a preparation of our discussion of a non-compact Lie algebra.
Hence, the discussion in this section is essential for that for a non-compact Lie
algebra.
When a maximal Cartan subalgebra h2 of h1 is a Cartan subalgebra of g, h2 is
a subspace of a maximal Cartan subalgebra h of g. Hence, h2 can be regarded as
a subspace of h under the inner product of h . When a root system := (h1 )
of h1 is a subset of := g , h1 is called a regular subalgebra of g. When a Lie
subalgebra h1 contains a maximal Cartan subalgebra h of g, h is a maximal Cartan
subalgebra of h1 . Hence, for the common maximal Cartan subalgebra h, we can
define their root systems := g and := (h1 ). Then, is a subset of . In
this case, h1 is called a maximal regular subalgebra of g. In the following, we
discuss only the case when h1 is a maximal regular subalgebra of g.
Then, since h1 is a subalgebra, i.e.,
[h1 , h1 ] h1 ,
(6.21)
214
we have
( + ) ,
(6.22)
where we define S + S := { + | S, S } for two subsets S and S of a

linear space V . In this case, is called a root subsystem of .
Letting + be the set of positive roots of the root system , we give the set
of positive roots of
the root subsystem by + () := + () . Then, since
the set h1 := h ( + g[]) satisfies the condition (6.21) due to the condition
maximal Cartan subalgebra h of the
(6.22), h1 is a Lie subalgebra including the
compact semi simple Lie algebra g := h ( + g[]). Hence, the analysis of
the Lie subalgebra h1 including the maximal Cartan subalgebra can be reduced to
that of the root subsystem.
Now, let p be the orthogonal complement of h1 with respect to Killing form. Given
X, Y h1 and Z p, as [X, Y ] h1 , we have (X, [Y, Z ])g = ([X, Y ], Z )g = 0,
which implies that
[h1 , p] p .
(6.23)
Since the set of roots contained in p is \ , we obtain

( + (\ )) (\ ).
(6.24)
That is, (6.24) is the property derived from (6.22).

The following equivalence relation holds for the root subsystem.
Theorem 6.8 Let be a root subsystem of a root system whose fundamental
system is a subset of a real linear space V . The following relations are equivalent
to each other.
(1) + () is the set of positive roots of , and its fundamental system is included
in the fundamental system = {1 , . . . , r } of the root system . The relation
= + () also holds. Further, there exists an element + () such
that = { |, = 0}.
(2) There exists an element X h such that
+ ()\ = { |(X ) > 0}
= { |(X ) = 0}.
(6.25)
(3) The set + () satisfies (( + ()\ ) + ) ( + ()\ ).

(4) The set + ()\ is invariant for Weyl group W( ) (the group generated by
{W } ) with respect to .
Proof (2)(3) is trivial.
(3)(4): For + ()\ and , , is an integer. Hence, (3) guarantees that W () = , is an element of + ()\ .
215
(4)(2): The set cone( + ()\ ) := {X h |(X ) > 0, + ()\ } is

a convex cone. Especially, since cone( + ()\ ) cone( + ()\ ) = , the set
cone( + ()\ ) is not the empty set. Next, we define the action of an element
W of Weyl group W( ) for X h as (W (X )) := W ()(X ). Then, we have
W (X ) := X (X )Z . Since the set + ()\ is invariant with respect to the
action of Weyl group W( ), cone( + ()\ ) is also invariant with respect to the
action of Weyl group W( ). Hence, cone( + ()\ ) contains a non-zero fixed
point X with respect to the action of Weyl group W( ). Thus, for we have
(X ) = 0, which implies (2).
(2)(1): For , we choose elements i1 , . . . , im of and positive integers m i1 , . . . , m im such that = m i1 i1 + . . . m im im . Condition (2) yields that
i1 , . . . , im are elements of . The definition of the fundamental system guarantees that l = 1 and . That is, we have = + (). Since the set
= { j } is a basis of the linear space spanned by , we can choose j h such
that j , i = i, j .
Now, we assume that \ = { j1 , . . . , jl } and we define := j1 + . . . + jl .
Then, we have + (). So, we obtain the relation = { |, = 0}.

(1)(2) is trivial.
When a root subsystem of a root system with fundamental system satisfies
the above condition, ( + (), ) is called a hyperbolic system[79, Definition
VII.2.6.]. When ( + (), ) is a hyperbolic system, Condition (2) implies that
(( + ()\ ) + ( + ()\ )) ( + ()\ ).
(6.26)
Hence, the following theorem holds even though a Lie subalgebra h1 of g does not
include the maximal Cartan subalgebra.
Theorem 6.9 The following conditions are equivalent to each other for a compact
real semi simple Lie algebra g and its Lie subalgebra h1 .
(1) cg (c(h1 )) = h1 .
(2) The Lie subalgebra h1 contains a maximal Cartan subalgebra h of g.
Further, when the above condition holds and is the root system of h1 , there exists
a fundamental system of such that ( + (), ) is a hyperbolic system.
Proof (1)(2): Since an element of a maximal Cartan subalgebra h of the semi simple Lie algebra [h1 , h1 ] is commutative with any element of c(h1 ), h + c(h1 ) is a maximal Cartan subalgebra of h1 . The discussion in the beginning part of Sect. 6.2 guarantees that h1 is a maximal Cartan subalgebra. Let h be a maximal Cartan subalgebra
of g including h + c(h1 ). Since c(h1 ) h, we have h = cg (h) cg (c(h1 )) = h1 ,
which implies that h is a Cartan subalgebra of h1 . Since h + c(h1 ) is a maximal Cartan subalgebra of g, h + c(h1 ) is h. That is, h1 contains a maximal Cartan subalgebra
h of g.
(2)(1): Since the definition of the center implies that h1 cg (c(h1 )), it is sufficient
to show that p cg (c(h1 )) = {0}. Because c(h1 ) is included in a maximal Cartan
216
subalgebra, we have c(h1 ) = {X h |(X ) = 0, }. We choose an element

X h satisfying Condition (2) of Theorem 6.8. So, we have X c(h1 ). Also, (6.25)
guarantees that Y p satisfies [X, Y ]
= 0. Hence, we obtain Y
/ cg (c(h1 )), which
yields Condition (1).
Now, we assume Conditions (1) and (2), and let be the root system of h1 . Based
on the discussion of the proof of (1)(2), we choose two kinds of elements Fx and
y
F of h1 as elements of a basis of the orthogonal complement of h. Since Fx and
y
F belong to g, a root of h1 is also a root of g. That is, we have h1 g . Hence,
X c(h1 ) satisfies
(X ) = 0, (h1 ).
(6.27)
On the other hand, since cg (c(h1 )) = h1 , for a root g \h1 , there exists an
element X c(h1 ) such that (X )
= 0. Thus, there exists an element X c(h1 )
such that the relation (X )
= 0 holds for any root g \h1 . Hence, Proposition
6.1 guarantees that there exists a fundamental system of g such that the set of
positive roots g+ satisfies
(X ) > 0, g+ \h1 .
(6.28)
Due to (6.27) and (6.28), any element of h+1 := h1 g+ is given as a sum of

elements of (h1 ) := h1 , which is a fundamental system of h1 . Since =
{1 , . . . , r } forms a basis of a linear space, we can choose elements {1 , . . . , r }
such that i , j = i, j . Therefore, describing all of elements of \(h1 ) as
{i1 , . . . , il }, we have h1 = { g |i1 + + il , = 0}. Thus, we obtain

Condition (1) of Theorem 6.8, which yields (2).
Next, we consider the complexifications h1,C and pC of h1 and p. Then, the conditions (6.21) and (6.23) yield
[h1,C , h1,C ] h1,C , [h1,C , pC ] pC .
(6.29)
Now, we focus on the complex Lie algebras p+,C := + ()\ g , p,C :=

+ ()\ g , h1,+,C := + () g , and h1,,C := + () g generated
by {K } + ()\ , {K } + ()\ , {K } + () , and {K } + () , respectively. The relations (6.15) and (6.16) and the property of + () imply that
[h1,+,C , h1,+,C ] h1,+,C , [h1,,C , h1,,C ] h1,,C .
(6.30)
Similarly, the relations (6.15), (6.16) and (6.26) yield that

[p+,C , p+,C ] p+,C , [p,C , p,C ] p,C .
(6.31)
These facts show that p+,C , p,C , h1,+,C , and h1,,C are complex Lie subalgebras of
gC .
217
Here, let us consider the adjoint representation of the Lie subalgebra h1 including
a maximal Cartan subalgebra of g on g. Then, h1 g is an invariant subspace. Hence,
we can define a real representation of h1 on g / h1 . Since g / h1 is isomorphic to p,
this representation can be regarded as a real representation of h1 on p.
On the other hand, the root subsystem of the root system satisfies
((\ ) + (\ ))
(6.32)
if and only if the Lie subalgebra h1 including the maximal Cartan subalgebra corresponding to and the orthogonal complement p satisfy [p, p] h1 . In the following,
the root subsystem of is called irreducible when the condition holds. The condition is equivalent to the condition [pC , pC ] h1,C for the complexifications pC and
h1,C .
When ( + (), ) is a parabolic system for an irreducible root subsystem

of , ( + (), ) is called an irreducible parabolic system.
Theorem 6.10 The following conditions are equivalent to each other for a root
subsystem of a root system with a fundamental system in a real linear
space V .
(1) ( + (), ) is an irreducible parabolic system.
(2)
(( + ()\ ) + ( + ()\ )) =
+
(( ()\ ) + ) ( ()\ ).
(6.33)
(6.34)
Proof (1)(2): The relations (6.26) and (6.32) yield the relation (6.33). The relation
(6.34) is none other than the Condition (3) for a parabolic system.
(2)(1): Due to the relation (6.34), ( + (), ) is a parabolic system. The

relation (6.33) yields the relation (6.32). So, is irreducible.
When a Lie algebra g is simple, its root system is connected. An irreducible
parabolic system ( + (), ) is called connected irreducible parabolic system when is connected. Then, Condition (2) guarantees that there exists an element of \ . If there exist two elements 1 and 2 of \ , we can choose
elements 1 , . . . , l that connect the two elements 1 and 2 because the
fundamental system is connected, where the inner product between two distinct
roots i and j is assumed to be zero unless |i j| = 1. Further, we assume that
1 , 1 = l , 2 = 1. If these relations do not hold, they hold by exchanging 1 and 2 because of the classification given in Table 6.1. Hence, due to the
condition (6.24), 1 := Wl W1 ( 1 ) is contained in + \ . Thus, we have
W2 ( 1 ) = 1 + 2 , which contradicts (6.33). So, the set \ has the unique element . Thus, the set + ()\ is given as ( + span Z0 (\{})) . Due to
this fact, all of connected irreducible parabolic systems can be classified by use of
the Dynkin diagram describing and the unique simple root [79, Theorem A.V.1
(c)] [66, Appendix C]. The pair of the Dynkin diagram and the simple root is
218
An III :
( n 1, p + q = n + 1,1 p q )
Bn I :
(n 2)
Cn I :
( n 3)
Dn I :
( n 4)
Dn II :
( n 4)
E6 III :
E7 VII :
Fig. 6.2 Hermitian Dynkin diagram
classified as Fig. 6.2, in which, the black point corresponds to the simple root . The
figures given in Fig. 6.2 is called a Hermitian Dynkin diagram.
When ( + (), ) is a connected irreducible parabolic system, there exists
the maximum root () because the Lie algebra g corresponding to is simple,
but () does not belong to . This fact is useful for latter discussions. These
discussions yield the following Theorem.
Theorem 6.11 When a root subsystem of is irreducible and g is simple, we
obtain the following items. [p, h1 ] = p, |\ | = 1, and the real representation of
h1 on p is irreducible.
Proof Since p + h1 = [p + h1 , p + h1 ] = [p, h1 ] + [p, p] + [h1 , h1 ] and [p, p] + [h1 ,
h1 ] = h1 , we have [p, h1 ] = p. Further, the above discussion yields the relation
|\ | = 1, any element of + \ can be written as + m 1 1 + . . . + m l l .
Hence, the orbit of K with respect to h1,C contains K for any + \ . Thus,
the representation of h1,C on pC is irreducible.
When the representation of h1 on p has a non-trivial invariant subspace, the complexification of the invariant subspace is also an invariant subspace for complexifi-
219
cation of the representation, which contradicts the above fact. So, the representation

of h1 on p is also irreducible.
6.6 Structure of sp(r)*

Section 3.1.2 mentioned that the real symplectic group Sp(2r, R) inevitably appears
in the symmetry related to analytical mechanics. However, since Sp(2r, R) is a noncompact Lie group, there are several difficulties to treat it. Since it is impossible to
deal with it in this section,2 we discuss the Lie algebra sp(r ) of the compact Lie
group Sp(r ). As both complexifications of the Lie groups Sp(r ) and Sp(2r, R) are
the complex Lie group Sp(2r, C), both Lie groups have common structures. Since
the Lie algebra sp(r ) is an example of a compact semi simple Lie group belonging
to the Case Cr in Sect. 6.5, it is a good example other than su(r ) to understand the
general discussion in Sect. 6.5.
Any matrix on Cr can be converted to a matrix on R2r by replacing the multiplication of the pure imaginary number i by that of the matrix I J . An element
X = X 1 + i X 2 (X 1 and X 2 are real-valued matrices) of u(r ) can be regarded as an
element of sp(r ) as follows. Since the matrix (X
) := X 1 I2 + X 2 J satisfies
(X 2 I2 + X 2 J )T Jn = (X 1T I2 X 2T J )In J
= (X 1 I2 X 2 J )In J = (X 1 I2 X 2 J )In J
= In J (X 1 I2 + X 2 J ),
it is an element of sp(r ). As an extension of the map defined in Sect. 4.1, using W
defined in (4.9), we define
(Y ) := (Ir W )Y (Ir W )1
(6.35)
for a square matrix Y of order 2r . The map satisfies (X

) = (X X ). Then, the
relations (u(r
)) = sp(r ) sp(2r, R), (

U(r )) = Sp(r ) Sp(2r, R) hold.
Hence, we can regard u(r ) as a subalgebra of sp(r ). In the following, we
y
z
x
, F j,l , F j,l
, and F jz,u defined in (4.35), (4.36),
discuss sp(2r, R) by using F j,l
(4.37) and (4.44). Then, a maximal Cartan subalgebra h of sp(r ) is given by
h := {X 2 J |X 2 is a diagonal matrix of size r }, and is a subalgebra of u(r ). That
is, since u(r ) is a maximal regular subalgebra of sp(r ), a root of u(r ) is a root of
sp(r ). Hence, j,l su(r ) can be regarded as an element of sp(r ) as follows.
j,l (X 2 J ) = j|X 2 | j + l|X 2 |l, X 2 J h .
2 Sp(2r, R)
will be discussed in Sect. 7.8.
220
Then, the weight is defined so that the relation (X 2 J ) = 2r |X 2 |r holds

for an element X 2 J h.
We have
((X
), (Y
))sp(r ) = 4(r + 1) Tr X Y , X, Y u(r )
y
y
(|r r | F1z , |r r | F1z )sp(r ) = (|r r | F1 , |r r | F1 )sp(r ) = (r + 1).
+
The set u(r
) of positive roots and the fundamental system of sp(r ) are given by
+
u(r ) = { j,l }1 j<lr and = { j, j+1 } j=1,...,r 1 {}, respectively. Hence,
Z j,l :=
Z :=
i
1
z
z
J=
)
F j,l
(F
j,l
4(r + 1)
4(r + 1)
1
1
z,u
|r r | J =
(2F
r )
2(r + 1)
4(r + 1)
1
satisfy the condition (6.3). Since ( j,l , j,l ) = 2(r1+1) and (, ) = r +1
, we have
, r 1,r = 2. Thus, positive roots except for u(r ) are written as e j + el in
Table 6.1, and is given as
+
+
sp(r
) \u(r ) = { j,l := j,l + 2l,r + }1 jl<r ,
(6.36)
where we assume that j,l = 0 for j = l. On the other hand, the roots
{ j, j+1 } j=1,...,r 1 are given in the same way as su(r ). Here, we remark that the
inner product of h is different from that of su(r ) because the inner product is given
as the Killing form defined in the Lie algebra sp(r ), which is larger than su(r ). Due to
+
(6.36), (sp(r
) , u(r ) sp(r ) ) is a connected irreducible parabolic system. Further,
the matrix
Z j,l :=
1
1
(| j j| + |ll|) J =
(F
jz,u + Flz,u )
4(r + 1)
4(r + 1)
satisfies (6.3), and

x
x
Fx j,l := i F j,l
J = (F
j,l
),
1
Fx j,l := (| jl| + |l j|) F1z ,
2
satisfy (6.7) and (6.8).
Fy j,l := F j,l I = (F
j,l )
1
y
F j,l := (| jl| + |l j|) F1x
2
6.7 Complex Structure and Coherent State*
221

Section 6.5 dealt with a Lie subalgebra of g including its maximal Cartan subalgebra
h and its root subsystem. Using this discussion, we extend coherent states to a more
general setting because coherent states play an important role in a quantum system.
Remember that coherent states are given as an orbit of the ground state with respect
to a representation of a group of our interest. The ground state is a stable state and
invariant for a subgroup of our group. Hence, the discussions in Sect. 6.5 play an
important role. When the root systems of a compact semi simple Lie algebra g and
its Lie subalgebra h1 including its maximal Catran subalgebra, are characterized as
a parabolic system( + (), ), as explained later, the homogeneous space
G/H1 has the structure of a complex manifold, where G and H1 are the Lie groups
corresponding to g and h1 .
As an example, we consider a representation space U with an arbitrary highest weight + (). Assume that the stabilizer H1 is the invariant Lie subgroup {g G|f (g)|; ; |f (g) = |; ; |} of the highest weight vector
|; . Then, the corresponding Lie subalgebra h1 is the invariant Lie subalgebra
{X g |f (X )|; ; | |; ; |f (X ) = 0} g with respect to the highest weight vector |; . Letting be the root system of the Lie subalgebra h1 ,
we see that is { |, = 0}. Condition (1) of Theorem 6.8 guarantees
that ( + (), ) is a parabolic system. Hence, coherent states are given as the
orbit of the highest weight state |; ; | with respect to the representation of the
Lie group G. Then, the set of coherent states is characterized by the homogeneous
y
space G/H1 , where H1 is the Lie group corresponding to h1 . The set {Fx , F }+
generates the orthogonal complement p to h1 .
y
For
complex vector = { = x + i} + , we define g() :=
an arbitrary
y y
x y
exp( + 2i( F F )) and apply Gauss decomposition (6.13) to the complex Lie group corresponding to the complex Lie algebra gC . Then, there exist complex vectors = { } + , = { } + , and = { j }rj=1 such that

r

j E j exp
K .
(6.37)
f (g()) = exp
K exp
+
j=1
Since, as mentioned in Sect. 6.5, h1,,C and p,C are complex Lie subalgebras of
n = h1,,C p,C , there exists a complex vector { } + such that

exp

K
= exp

+ \
K exp
K .
(6.38)
+
As a generalization of U,s given in (4.31), we define U as the substitute of the

right hand side of (6.38) into the right hand side of (6.37). Note that this decomposition holds independently of the choice of the representation f . Applying them to
the vector | , , we have
222
f (g())| ; exp
K | ; .
(6.39)
+ \
The complex vector is given as a continuous function of the complex vector .

G and H1 are Lie groups corresponding to g and h1 , respectively, and H1 is the
stabilizer of the state | ; ; | with respect to the representation of the Lie
group G. Hence, due to (6.39), the homogeneous space G/H1 can be identified with
set of except for points with measure 0. Thus, the homogeneous space G/H1 is
isomorphic to the complex Lie group P,C generated by the complex Lie algebra
p,C . Hence, the homogeneous space G/H1 has the structure of a complex manifold.
This discussion holds independently of the choice of the highest weight h as
long as H1 is the stabilizer of | ; .
complex vector = { } + \ , we
Then, given a highest weight h and a
define the coherent vector | : := C exp( + \ K )|; , where C is
the normalizing constant. By choosing the complex vector satisfying (6.37) and
(6.38), the coherent state , := | : : | satisfies , =f (g())|; ;
|f (g()) . The closure of the set {, } has a one-to-one correspondence to the
orbit of |; ; | with respect to the representation of G, i.e., the homogeneous
space G/H1 . Due to (3.56), as a generalization of (4.34) to a compact semi simple
Lie group, the invariant measure on := G/H1 defined in the end of Sect. 3.7
satisfies

d
| : : | (d) = I,
(6.40)

which gives a POVM.

Conversely, when the root systems of a compact semi simple Lie algebra g and
a Lie subalgebra h1 including its maximal Catran subalgebra are characterized by
a parabolic system ( + (), ), due to Condition (1) of Theorem 6.8, there
exists a positive root + () such that = { |, = 0}. In this general setting, the homogeneous space G/H1 is characterized as the orbit of the state
|; ; | on the representation space of the highest weight . Hence, applying
the above discussion, we obtain the following theorem.
Theorem 6.12 Let g and h1 be the Lie algebras corresponding to a compact semi
simple Lie group G and its Lie subgroup H1 . When the Lie algebra h1 includes a
maximal Catran subalgebra of g and its root system is characterized by an irreducible
parabolic system ( + (), ), the homogeneous space G/H1 has the structure
of a complex manifold.
Next, let ( + (), ) be a connected irreducible parabolic system with
\ = {}. Also, let g and h1 be the compact semi simple Lie algebra and
its Lie subalgebra that correspond to the connected irreducible parabolic system.
When the irreducible representation space Un of g has the irreducible decomposi
j
tion nj=0 V[K, |n, n] with respect to the representation of h1 , Un is called
223
h1 -irreducible, where V[K, |n, n] is the irreducible representation space of the
j
Lie algebra h1 that contains K, |n, n. For example, in any Hermitian Dynkin
diagram, U is h1 -irreducible because g / h1 is irreducible with respect to the representation of h1 . Under An III Type of Hermitian Dynkin diagram, when p = 1, Um
is u(n 1)-irreducible for any integer m.
6.8 Structure of Non-compact Semi Simple Lie Algebra*

All of groups discussed in relation to analytical mechanics in Sect. 3.1.2 are noncompact groups. Hence, for an application to quantum system, we need to treat a
non-compact group. So, we investigate the structure of a non-compact semi simple
Lie group G. For this study, we discuss the corresponding Lie algebra g. Since it is
semi simple, the Killing form is non-degenerate. However, it is not negative definite
because the Lie algebra g is not compact. Thus, we focus on the complex Lie algebra
gC := g C, which is a complexification of g. Especially, the complexification of
a simple Lie algebra is a complex simple Lie algebra. Non-compact semi simple
Lie algebras whose complexification is a classical type of simple Lie algebra are
limited to sl(n, R), su( p, q), su (2m), so( p, q), so (2m), sp(2n, R), and sp( p, q)
given in Table 3.6 [67, p. 6]. In this section, we avoid to treat all of non-compact semi
simple Lie algebras, and discuss only a special class of non-compact semi simple
Lie algebras, whose analysis is not so difficult.
For this purpose, for a compact subalgebra h0 of g, we choose a real linear subspace
u of gC such that its dimension is g, its Killing form is negative definite, and h0 u.
Then, u is a compact semi simple Lie algebra. Letting k := g u and p := g i u, we
find that k is a compact Lie algebra and is called a maximal compact subalgebra.
Then, we obtain the direct sum decomposition g = k + p, which is called the Cartan
decomposition of g [67, p. 3]. Then, the Killing form is negative definite on k and
is positive definite on p. Since [u, u] u, [u, i u] i u, and [i u, i u] u, we have
[k , k ] k , [k , p ] p , [p , p ] k .
(6.41)
Since [g, g] = g, we also have [k, k] + [p, p] = k and [k, p] = p. Using the map :
X 1 + X 2 X 1 X 2 for X 1 k and X 2 p, we define the inner product B(X, Y ) :=
(X, (Y ))g , which is positive definite. For X 1 k and X 2 p, ad(X 1 ) and ad(X 2 )
are alternative and symmetric matrices under the CONS with respect to the inner
product, respectively. Define K := {exp(X )} X k and P := {exp(X )} X p . Then, K
is a maximal compact subgroup of G, and we obtain the Cartan decomposition
G = K P of the Lie group G.
However, since a non-compact simple Lie group G and its Lie algebra g have
too complicated structure, it is too difficult to discuss the general theory. To avoid
such a difficulty, we assume that there exists a compact maximal Cartan subalgebra
h. Since many physically important examples satisfy this condition, our analysis is
restricted to this case.
224
Similar to the case when g is compact, which is discussed in Sect. 6.2, we define
the inner product of h and define Z h for h . Then, there exists a basis of g
such that an element of h is written as an alternative matrix given in (3.46) of Lemma
3.9 under the basis. Hence, similar to the compact case, as subsets of h , we can
y
define the root system g of g and the root system k of k, and choose {Fx , F } for
g such that
[X, Fx ] = (X )Fy , [X, Fy ] = (X )Fx , X h .
Since g is semi simple, Theorem 6.6 guarantees that a different pair of basis Fx
y
and F does not correspond to the same root . Similar to (6.7), we can show
y
y
y
that [Fx , F ] = (Fx , Fx )g Z = (F , F )g Z , which implies that (Fx , Fx )g =
y
y
(F , F )g .
y
y
Since k is a compact Lie algebra, the real numbers (Fx , Fx )g and (F , F )g
y
x
are negative values for k . So, the subalgebra generated by {F , F , Z } is
y
isomorphic to su(2). Similar to (6.8), we choose the constant factors of Fx , F
y
y
1
x
x
x
x
such that (F , F )g = (F , F )g = (,) . On the other hand, since F and F
belong to p for p := g \ k , the real number (Fx , Fx )g is positive. So, the
y
subalgebra generated by {Fx , F , Z } is isomorphic to sl(2, R). Then, we choose
y
y
y
1
x
. So, the
the constant factors of F and F such that (Fx , Fx )g = (F , F )g = (,)
basis of g is uniquely determined from h.
Also, due to the final property [p, p] k of (6.41), the root system k of k is
irreducible as a root subsystem of the root system of g. In the discussions up to
the previous section, we have assumed that g is compact. However, the most crucial
assumption is the non-degeneracy of killing form, i.e., the semi-simplicity of g. In
this case, since g is not compact, but semi simple, a similar discussion is possible.
Then, by the same discussion as the previous section, we define the root system of
g, the fundamental system = {, . . . , r }, the set + () of positive roots, and
the root system k of k.
Then, since the Lie algebra k is assumed to contain the maximal Catran subalgebra
h of g, due to Theorem 6.9, ( + (), k ) is a parabolic system. Further, the
third relation (6.41) guarantees that ( + (), k ) is an irreducible parabolic
system.
Such a semi simple Lie algebra g is called Hermitian type and is classified
in Table 6.2 [79, Theorem A.V.1(c)] [66, Appendix C]. That is, a Hermitian type
Lie algebra g is described by an irreducible parabolic system. Conversely, for an
irreducible parabolic system ( + (), k ), we can construct a semi simple Lie
algebra g and its maximal compact Lie subalgebra k whose root systems are and k ,
by tracing the above discussion. Hence, the classification of Hermitian type simple
Lie algebras can be reduced to the classification of connected irreducible parabolic
systems.
6.9 Skew-Hermitian Representation of Non-compact Simple Lie Algebra*

Table 6.2 Hermitian type Lie algebra
g
Type
An III
p+q = n
Bn I
Cn I
Dn I
Dn III
E 6 III
E 7 VII
225
gC
su( p, q)
sl(n, C)
su( p) su(q) u(1)
so(2n 1, 2)
so(2n +1, C)
so(2n 1)
sp(2n, R)
sp(2n, C)
u(n)
so(2n 2, 2)
so(2n, C)
so(2n 2)
so (2n)
so(2n, C)
u(n)
e6
so(10) R
e7
e6 R
e14
6
e25
7
6.9 Skew-Hermitian Representation of Non-compact

Simple Lie Algebra*
Next, we address a skew-Hermitian representation of a non-compact simple Lie
algebra g. A skew-Hermitian representation of non-compact simple Lie algebra g
does not necessarily have the highest weight in general. However, a real quantum
system is usually characterized by a representation that has the highest weight. Such
a representation is called highest weight representation. It is often called a discrete
series representation [67].
When a non-compact simple Lie algebra g has a compact maximal Cartan subalgebra, it has a highest weight representation [67, Theorem 9.20]. In the following, for
simplicity, we address a Hermitian type simple Lie algebra g as a special case of simple Lie algebras. Also, we deal with a highest weight skew-Hermitian representation
of g.
Under a highest weight skew-Hermitian representation f of g, for a simple root
j and a positive root + (), we define
E j := if(Z j ), K := if(Fx ) f(Fy ).
(6.42)
Then, for an arbitrary root , we have

[E j , E j ] = 0
[E j , K ] = (Z j )K .
(6.43)
(6.44)
Since the representation f is skew-Hermitian, we find that (K ) = K and E j is

self-adjoint. In the following, we assume that the skew-Hermitian representation f is
irreducible, the relation K |; = 0 holds for any positive root + , and there
exists a normalized simultaneous eigenvector |; of E1 , . . . , Er (highest weight
vector). Then, the representation f is a highest weight representation.
In this case, an element m h of h is called a weight when
226
Umw := {v U|f(X )v = im(X )v, X h}

= {0}.
In particular, a weight h is called the highest weight when the relation
f(X )|; = i(X )|; holds for any element X h. The highest weight uniquely
determines the irreducible highest weight skew-Hermitian representation, and the
relation Uw = R|; holds. Then, we denote the representation space of the representation by U , and the representation by f . Then, the set of weights is written
as ().
Similar to the
discussion in the previous sections, the representation space U is
generated by { (K )m |; }. Hence, due to (6.43), U is given as U = m Umw
as the decomposition to simultaneous eigenspace of elements of h by using weights m.
y
y
For k+ (), the subspace generated by {Fx , F , [Fx , F ]} is isomorphic to
su(2). Hence, since the action of these elements are isomorphic to the skew-Hermitian
representation of su(2), the highest weight h satisfies
, = 2([Fx , Fy ]) Z0 .
(6.45)
On the other hand, for any element + () \ k+ (), the subspace generated
y
y
by {Fx , F , [Fx , F ]} is isomorphic to sl(2, R). Hence, since the action of these
elements is isomorphic to the skew-Hermitian representation of sl(2, R), the relation
m, 0 holds for an arbitrary element m (). Thus, unless = 0, the relation
, < 0
(6.46)
holds.
Conversely, when h satisfies the Harish-Chandra condition
+ , () < 0,
(6.47)
there exists an irreducible skew-Hermitian representation with the highest weight

h [79, p. 547] [67, p. 158] [44, Lemma 27]. Further, the formal dimension d
of the representation space U satisfies the formula by Harish-Chandra

d =
+ ()
| + , |
+ () ,
which can be regarded as a non-compact version of Weyls dimension formula [44,

Theorem 4] [79, p. 547].
Assume that the highest weight of the irreducible skew-Hermitian representation
satisfies
, = 0, k .
(6.48)
6.9 Skew-Hermitian Representation of Non-compact Simple Lie Algebra*
227
Then, similar to the compact case, the representation space U is called k-irreducible

when the irreducible decomposition of the representation of k is given as
j=0 V
j
[K |; ]. For example, when g is su(1, n 1), the highest weight representation
satisfying (6.48) is u(n 1)-irreducible.
Next, similar to Sect. 6.7, we define the coherent states for irreducible skewHermitian representation of g with non-trivial (non-zero) highest weight h . Let
h1 be the subalgebra that does not change the highest weight state |; ; |. Then,
h1 contains a compact maximal Cartan subalgebra h of g, and its root system h1 is
given as { |(, ) = 0}.
a complex vector = { }+ ,
We define + := { + |(, )
= 0}. For
we define a coherent vector | : := C exp( + K )|; , where the
normalizing constant C is chosen such that the norm of | : is 1. Then, similar

to the discussion in Sect. 6.7, applying Gauss decomposition (6.13) to the complex
Lie group corresponding to the complex Lie algebra gC , we can show that there
exist complex vectors , , satisfying (6.37) for a highest weight h and a
complex vector . Hence, the coherent state , := | : : | satisfies , =
f (g())|; ; |f (g()) . When Harish-Chandra condition (6.47) holds, due to
(3.56), as a generalization of (6.40) to a non-compact semi simple Lie group, we
have

d
| : : | (d) = I,
(6.49)

where is the invariant measure of := G/H1 defined in Sect. 3.7, and d is the
formal dimension of the representation f [44, Theorem 4]. This relation gives a
POVM by coherent states.
Further, we choose a highest weight representation such that the stabilizer of the
highest weight state is a maximal compact Lie subgroup K . Then, similar to Theorem
6.12, we can show the following theorem.
Theorem 6.13 Let G be a Hermitian type Lie group, and K be its maximal compact
Lie subgroup K . Then, the homogeneous space G/K has the structure of a complex
manifold.
6.10 General Non-compact Lie Algebra

and Its Skew-Hermitian Representation*
Up to the previous section, we imposed the semi simple condition for our Lie algebra
g. In the following, we deal with a real Lie algebra that is not necessarily semi
simple. However, since it is quite difficult to discuss a real Lie algebra g without
any additional condition, we assume that the Lie algebra g has a compact maximal
Cartan subalgebra h. Then, there exists a maximal compact subalgebra k including
h. Any element of the center c(g) = {X h |(X, X )g = 0} of h is commutative with
228
any element of g. We choose a subspace g0 of g that is isomorphic to the quotient

space g / c(g), and define h0 := h g0 . Similar to the semi simple case, we define the
element Z h0 for h0 and the positive definite inner product (,) on h0 . Since
all of elements of h0 are commutative with each other, we can choose the basis {F jx ,
y
F j } of the orthogonal complement of h0 in the space g0 such that all of elements of
h0 are written as the form (3.46) of Lemma 3.9. Then, the relation
y
ad(X )F jx = j (X )F j , ad(X )F j = j (X )F jx
(6.50)
holds for any element X h0 .

Let g+ be the set of elements j satisfying the above condition, and g be the set
y
of elements j and j . Then, we choose the signs of F j such that there exists an
element of X 1 h0 satisfying that g+ = { |(X 1 ) > 0}. Here, generates
h0 , but g is not necessarily semi simple. Hence, does not necessarily satisfy the
condition for a root system. So, it is called a root set. The root set k of a maximally
compact subalgebra k is a subset of . We define
p := \k .

Then, we consider the decomposition g0 = h0 + g[] p+ g[] of
k
g0 , where k+ := k + and p+ := p + . For the pair k , we can
show the equivalence among Conditions (2), (3) and (4) of Theorem 6.8. The
pair ( + , k ) is called a parabolic system when one of the above condig
tions holds. Further,
the adjoint representation of induces a representation of
g
g
p
+
on
/
[].
Additionally,
when
the induced representation of
:=
=
k
k
p
g
is
irreducible,
is
Hermitian-type.
Then,
the
highest
weight with respect to
k
the representation of k on g / k is the maximum element of p+ . So, we have
p+ = ( span Z0 k ) + , where k is a fundamental system of k .
When we define K for a representation f of g in the same way as (6.42), the relation [if(X ), K ] = (X )K holds for an arbitrary element X h in the same way.
In the following, we consider the highest weight representation characterized by the
highest weight vector. However, it is not so easy to write down the necessary and sufficient condition of the highest weight for the representation to be skew-Hermitian
in the general case. Especially, the representation space U is called k-irreducible
when the following three conditions hold; (1) The Lie algebra g is Hermitian-type.
(2) The highest weight is orthogonal tok . (3) The irreducible decomposition
j
of the representation space U is given as
j=0 V[K |; ] by using the highest
weight vector |; .
Next, we define the coherent state for the irreducible skew-Hermitian representation of g with the highest weight h0 in the same way as Sect. 6.7. Let
h1 be the subalgebra of elements of g that does not change the highest weight
state |; ; |. We assume that h1 is included in k. Then, h1 includes a compact maximal Cartan subalgebra h of g, and the root set h1 is given as {
+
+
|(Z ) = 0}. Using
:= { k |(Z )
= 0}, we define the coherent vector
| : := C exp( + K )|; for a complex vector = { }+ , where
C is chosen such that the norm of | : is 1.
6.10 General Non-compact Lie Algebra and Its Skew-Hermitian Representation*
229
Let H1 be the Lie group corresponding to the Lie algebra h1 . Then, the homogeneous space G/H1 is given as the closure of the set {| : : |} . Hence, when
the formal dimension d can be defined by (3.54), we regard as an element of the
homogeneous space G/H1 and focus on the invariant measure on the := G/H1
defined in the end of Sect. 3.7. As a generalization of(6.49) to the case when g is
not necessarily semi simple, we obtain the relation d | : : | (d) = I .
This relation shows that the resolution of the identity I by the Coherent state coherent
states , := | : : | forms a POVM.
In this way, the discussion given in the previous section can be generalized to the
case when g is not necessarily semi simple. The next chapter discusses an example
of such a Lie algebra g.
Chapter 7
Bosonic System and Quantum Optics
Abstract When more than one particles can simultaneously share the same state of
the quantum system, the particle is called a boson. A photon is a typical example
of a boson. In the bosonic system, a unitary representation of Heisenberg group is
given. In the photonic case, the coherent state of the representation is called coherent
light and is physically implementable. The operation corresponding to the representation of SU(1, 1) is called squeezing. When we apply the squeezing to the vacuum
state, which is the state with no photon, the resultant state is called a squeezed state
and plays an important role in the quantum optical system. In fact, the Heisenberg
group corresponds to the transformation of the position and the momentum in analytical mechanics. The canonical transformation is a typical transformation preserving
the product of the Heisenberg group. This chapter discusses the representation that
describes the transformation for the position and the momentum by Heisenberg group
as well as the canonical transformation.
7.1 One-Mode Bosonic System

7.1.1 One-Variable System
This section addresses the one-mode bosonic system. As the first step, we focus on
the quantum system of a particle with one-dimensional freedom, i.e., the one-variable
system, which is the quantum system with respect to the position of a particle on the
2
one-dimensional space R and is given as the linear space
L (R) 2of square integrable
1
2
functions from R to C with the norm f := 2 | f (x)| d x. In this system,
the physical quantity describing the position is given as the multiplication operator
Q defined as1
(Q f )(x) := x f (x).
1 For
(7.1)
the precise definition of the operators in this chapter, see Sect. 1.6.

DOI 10.1007/978-3-319-44906-7_7
231
232
7 Bosonic System and Quantum Optics
On the other hand, the momentum is described by the operator P := P by using the
Plank constant and the operator P defined by
(P f )(x) := i
d
f (x).
dx
(7.2)
Since the operators Q and P satisfy the conditions Q = Q and P = P, they are
self-adjoint operators[84].2 Then, these operators satisfy the following commutation
relation
[Q, P] = i I.
(7.3)
That is, two physical quantities satisfy the commutation relation

= iI.
[Q, P]
(7.4)
In the following, we write Plank constant as for the sake of the consistency
with the representation-theoretical notation. Hence, we define the skew-Hermitian
representation f of h(2, R) as
f(P) = iP, f(Q) = iQ, f(Z ) = iI.
(7.5)
This representation is isomorphic to the following representation via the unitary

transformation f (x) 1/4 f (1/2 x) on L 2 (R).
f(P) = i P, f(Q) = i Q, f(Z ) = iI.
(7.6)
When = 0, this representation is an irreducible representation. An arbitrary irreducible skew-Hermitian representation of h(2, R) is limited to the representation
isomorphic to the above representation, which is known as Stone-von Neumann
Theorem [65], [31, Sect. 1.5].
When the relation [X, [X, Y ]] = [Y, [X, Y ]] = 0 holds, we have the Hausdorff
formula
e X eY = e 2 [X,Y ] e X +Y .
1
(7.7)
When = 12 (q + i p) and = 12 (q + i p ), due to the above relation (7.7) and the

commutation relation (7.11), the above representation f of Heisenberg group H(2, R)
satisfies the relation
2 The
precise definitions of Q and P are given in Sect. 1.6.
233
W() := f(exp( p Q q P)) = exp(i( pQ qP)) = exp(a a)

= e
W()W( ) = e
||2
2
exp(a ) exp(a)
1

2 ( )
W( + ) = e
i

2 (q pq p )
(7.8)
W( + ).
(7.9)
Now, we rewrite
the operator W() as W( z ) by using the real vector z = (q, p)T .
i
is a unitary representation of Heisenberg group H(2, R) =
Then, W( z )e
{gh ( z , )} and is called Heisenberg representation.
On the other hand, the map R2

(q, p) W( ) is can be regarded as a proi

jective unitary representation of group R2 with factor system {e 2 (q pq p ) }(q, p),(q , p ) .
The representation is called Heisenberg representation. When we need to distinguish the representation from the Heisenberg representation of Heisenberg group
H(2, R) = {gh ( z , )}, it is called Heisenberg representation of R2 . Then, Stone-von
Neumann Theorem [65], [31, Sect. 1.5] guarantees that a projective irreducible unii

2 (q pq p ) }
tary representation of R2 with
(q, p),(q , p ) is limited to
factor system {e
Heisenberg representation W( z ).
Since this representation satisfies
W( )1 QW() = Q + q I, W( )1 PW() = P + p I,
(7.10)
it can be regarded as the representation correspondingto the displacement of position and momentum. Hence, the unitary operator W( ) is called displacement
transformation.
Exercise 7.1 Show that W( z )ei is a unitary representation of Heisenberg group

H(2, R) = {gh ( z , )}.
7.1.2 One-Mode Bosonic System and One-Variable System

Next, we discuss the bosonic system, in which, the number of possible states to
be taken by a particle is only one. Such a quantum system is called the one-mode
bosonic system. For example, when several photons proceed to the same direction
with the same frequency and are localized in the same place, we say that these photons
take the same mode. In the bosonic case, many particles can take the same mode
simultaneously. Hence, even though only one mode is available for each particle, the
quantum system is composed of more than one states because the states are assigned
according to the number n of the particles taking the unique mode. We denote the
state vector (number vector) describing n particles by |n. Here, we assume that
|n is a normalized vector and |n is orthogonal to |m for distinct integers n and m.
In particular, the density operator |00| expresses the vacuum state corresponding
situation
photon. Then, the

with no
one-mode bosonic system is given as the CONS
|xn |2 < generated by {|n} , where xn is a complex
2 := {xn }
n=0
n=0
n=0
number.
234
The annihilation operator a on 2 is defined by a|n = n|n 1. The adjoint

operator a is called the creation operator and satisfies a |n = n + 1|n + 1.
Then, based on the definition, we can show the following relation
[a, a ] = I,
(7.11)
where I is the identity operator. Since the operator N := a a satisfies
N|n = a a|n = a n|n 1 = n|n,
(7.12)
it is called the number operator.

However, when we treat the one-mode bosonic system only as the sequence space
2 , we have some inconveniences. Identifying the sequence space 2 with the functional space L 2 (R), we can employ several properties and concepts defined in L 2 (R)
so that we have much wider application. Now, we identify |n with the n-th Hermitian
n
x2
2
function n (x) := 1 n e 2 ddx n ex . Then, we have naturally the isomorphism
n!2
from the sequence space 2 to the space L 2 (R) of square integrable functions on R.
In the following, we identify the one-mode bosonic system 2 and the one-variable
system L 2 (R) via the isomorphism. Under the isomorphism, we have the relations
i
(a a ). Similarly, the relations a = 12 (Q + iP),
Q = 12 (a + a ) and P =
2
a = 12 (Q iP), and N = 21 (P2 + Q2 I ) hold.

As a typical example, we focus on the quantum system of photons with frequency
f = c (angular frequency = 2 f , c is velocity of light), i.e., the electromagnetic
field with wave length as Fig. 7.1. Here, the operator Q corresponds to the vector
potential A of electromagnetic field and the operator P does to the electric field E
of electromagnetic field. That is, the physical quantities A and E are given by using
the dielectric constant 0 and the Planck constant h as [105, Sect. 2.1]

A=

h
h
Q, E =
P.
20
20
(7.13)
By using a, the relations (7.10) can be unified to the following relation
E: electric field
=wave length
direction
A: vector potential
Fig. 7.1 Electromagnetic field with wave length
235
W()1 aW() = a + I.
(7.14)
In general, the Hamiltonian in the one-mode

system is often given as the
bosonic
Nn
constant times of N. Hence, the state N :=
n=0 (N +1)n |nn| is called the thermal
equilibrium state and plays an important role in the quantum optics [29]. Since
N satisfies the relation N = Tr N N, it is called the average number. We choose
tN
the parameter t dependently of the average number N such that N = Tre etN . Since
2

||N dqdp
p , we have N =
= q+i
. So, the thermal equilibrium state is the
C ||e
2 N
2
average state of the coherent states under the Gaussian integral.
As fundamental measurements on the quantum system L 2 (R), we can list the
measurements E Q and E P based on the spectral measures of the self-adjoint operators
P and Q and the measurement E N based
on the spectral measure of N. Due to (7.12),
the spectral decomposition of N is
n=0 n|nn|, and its eigenvalues are limited to
non-negative integers. Then, using the above relation, we define the coherent vector
under the Heisenberg representation of the Heisenberg group H(2, R) as
| := e
||2
2
exp(a )|0 = e
||2
2

n
|n.
n!
n=0
(7.15)
Since (7.8) guarantees the relation | = W()|0, the coherent vector can be written
by using the unitary representation. As the vector | satisfies a| = |, we have
p.
|a| = , i.e., |Q| = x and |P| = y because = q+i
2
Now, we discuss the equality condition of the uncertainty relation (5.140). The
relation (7.14) yields that
0 = 0|a a|0 = |W()a W() W()aW() |
1
=| ((Q q)2 + (P p)2 i[Q q I, P p I ])|
2
1
= (2|| P + 2|| Q 1),
2
(7.16)
which implies the equality of (5.140) when = ||. Especially, the equality condition of the first and second inequalities in (5.140) yield that
2|| P = 2|| Q =
1
.
2
(7.17)
That is, the dispersion for P and Q of the coherent state | does not depend on .
Here, we compare the coherent state defined here with the coherent states of
su(2), su(1, 1) defined in Sect. 4.2. Using (7.15), we have

dqdp
W()|00|W()
C
||
dqdp
= I,
2
(7.18)
236
which gives a measurement based on the coherent state ||. This measurement

is called the heterodyne measurement. The equation (7.18) corresponds to (4.34)
is the invariant measure on C,
for the coherent states of su(2), su(1, 1). Since dqdp
2
(7.18) implies that the formal dimension is 1, as discussed in the relation (3.56).
dqdp
is given as 2(1+s||
The invariant measure dqdp
2 )2 with substituting 0 into s. So, we
2
denote C by 0 .
Now, we introduce two types of functions by using the coherent states |. We call
Q X () := |X | the Q-function (Husimi function) of the operator X [61]. When
X is a density operator, i.e., a state, the Q-function gives the probability density function of the outcome of the heterodyne measurement. On the other
hand, for an oper,
ator X , there uniquely exists a function PX () such that X = C PX ()|| dqdp
2
and such a function PX () is called the P-function. When we perform the heterodyne
p
measurement to the system whose state is a coherent state |, the outcome = q+i
2
. So, we
obeys the Gaussian distribution e|| dadb
(Exercise 7.2), where = a+ib
2
2
have

dadb
dadb
| =
Q X () =|
PX ()||
PX ()|||2
2
2
C
C

2 dadb
.
(7.19)
=
PX ()e||
2
C
2
That is, the Q-function is given as the output distribution when the information is
generated by the P-function with the additive white Gaussian noisy operation.
For the operator X , Wigner function W X () is defined to be Tr X W(). When
:= 12 (a + ib), Wigner function W|| () of the coherent state || is calculated as
W|| () = e e
||2
2
= e
||2
2
+i(apqb)
(7.20)
That is, the absolute square of Wigner function |W|| ()|2 is the standard Gaussian
distribution for p and q. The reason why we consider the absolute square of Wigner
function is given in Sect. 7.7. Generally, Wigner function W X () can be characterized
by P-function as follows.

dadb
W()
PX ()||
W X () = Tr
2
C

||2 dadb
dadb
= e
.
=
PX ()|W()|
PX ()e 2
2
2
C
C
(7.21)
Q-function is also characterized by Wigner function as (Exercise 7.11)

Q X () =
W X ()e+ e
||2
2
dadb
.
2
(7.22)
When =
q+i
p,
2
237
Wigner function is written as

qp
q + ip
q
q + ip
ip
= Tr X W
= Tr X ei 2 W W
2
2
2
2

q
i p (a) i q p
i q2p
i
px
Xe
= Tr W
W
=e 2
X (x q, x)e d x
2
2

q
q i pt
q
e dt,
X (x q, x)ei p(x 2 ) d x =
X t ,t +
(7.23)
=
2
2
WX
which is the definition of Wigner function in many literatures [108, 55]. Here, (a)
follows from (3.65). Overall, the functions, Q-function, P-function, and Wigner function are used for visualization of the state in the phase space, which describes the
position space and the momentum space equally.
Exercise 7.2 Show that the distribution of the outcome =
measurement is
2
e|| dadb
2
q+i
p
2
of the heterodyne
when the state is the coherent state |.

Exercise 7.4 Calculate the distribution of the outcome =
measurement when the state is the number state |n.
q+i
p
2
of the heterodyne
Exercise 7.5 Calculate the distribution of the outcome n when we measure the
number operator N and the state is the coherent state |.
7.1.3 Extended Heisenberg Representation

In the above Heisenberg representation, the coherent vector cannot be characterized
by using a compact maximal Cartan subalgebra of the the Lie algebra corresponding
to Heisenberg group H(2, R). Hence, the characterization of coherent vectors given
in Chap. 3 cannot be applied. To realize such a characterization, we consider a larger
group than H(2, R).
Firstly, we define the action T of g U(1) on H(2, R) as
T (ei )gh (q, p, ) = gh (cos q sin p, sin q + cos p, ).
(7.24)
The semi direct product Lie group H(2, R) U(1) based on the action T is called
the extended Heisenberg group with degree 1. By letting F be the generator of
u(1), the Lie algebra h(2, R) u(1) of the extended Heisenberg group of degree 1
is generated by the elements F, Q, P, and Z satisfying the following commutation
relation.
[F, Q] = P, [F, P] = Q, [F, Z ] = 0.
(7.25)
238
Hence, due to the discussion in Sect. 7.1.1, the irreducible skew-Hermitian representation f of the Lie algebra h(2, R) u(1) is isomorphic to the following irreducible
skew-Hermitian representation given by Z , F R.
if(Q) =
Z Q, if(P) =
Z P, if(Z ) = Z I.
(7.26)
The operator E := if(F) satisfies the relation [iE, a] = ia due to the commutation relation (7.25). So, we obtain
E = if(F) = a a + F .
(7.27)
The subspace F, Z spanned by {F, Z } is a Cartan subalgebra of the Lie algebra

h(2, R) u(1). Hence, in the representation (7.26) and (7.27), the vectors {|n}
n=0
are the simultaneous eigenvectors (weight vector) of if(Z ) and E. The eigenvalues
(weight) of |n with respect to if(Z ) and E are Z and F n, respectively. In
this representation, the highest weight is ( Z , F ) and the highest weight vector is
the vacuum vector |0. Since any representation can be characterized by (7.26) and
(7.27), a skew-Hermitian representation of Lie algebra h(2, R) u(1) is determined
by the highest weight ( Z , F ) and the representation space is denoted by U Z , F .
When we need to identify the Lie algebra, the representation space is denoted by
U Z , F (h(2, R) u(1)). Since the case of Z = 1 can be reduced to the case of
Z = 1 via a suitable conversion with respect to (q, p), in the following, we discuss
only the case of Z = 1, in which, U Z , F is simplified to U F .
Next, we investigate what representation of the Lie algebra h(2, R) u(1) gives
a representation of the extended Heisenberg group of degree 1 H(2, R) U(1). For
this purpose, we need to define the map U : ei eiE . This map can be defined
only when F is an integer. Then, due to (7.27), U satisfies

U(ei )W( Z )U(ei )1 = W(ei Z ),
(7.28)
which is the condition (2.37) given in Sect. 2.3. Hence, based on the discussion
in Sect. 2.3, W is a unitary representation of the extended Heisenberg group with
degree 1 H(2, R) U(1), which is a semi direct product Lie group. This unitary
representation is called the extended Heisenberg representation. By letting =
q+i
p , the relation (7.28) is equivalent to the following relation
2

U(ei )W( Z x, Z y)U(ei )1

= W(cos Z x sin Z y, sin Z x + cos Z y).
Hence, when F R gives the irreducible skew-Hermitian representation f Z , F
of the Lie algebra h(2, R) u(1), the representation f Z , F can be extended to an
irreducible unitary representation of the Lie group H(2, R) U(1) if and only if F
is an integer.
239
be the double covering of U(1), (which is isomorphic to

Further, by letting U(1)
U(1)), the double covering group of the Lie group H(2, R) U(1) is the Lie group
An irreducible skew-Hermitian representation f F of the Lie algebra
H(2, R) U(1).
h(2, R) u(1) cannot be extended to an irreducible unitary representation of the Lie
group H(2, R) U(1), but can be extended to an irreducible unitary representation
of the double covering group if and only if F is a half integer, i.e., 2 F is an integer
and F is not an integer.
6.10 for the representation f Z , F is
Further, the coherent vector given in Sect.
xd y
is the normalized measure, the formal
| Z defined in (7.15). Hence, when Z d 2
dimension of the representation space U Z , F is Z , which can be checked by (7.18)
with a suitable change of variables from p and q to x and y.
Taking the derivative in (7.28) for the complex number , we have U(ei )a
U(ei )1 = ei a, which is equivalent to
U(ei )QU(ei )1 = cos Q + sin P, U(ei )PU(ei )1 = sin Q + cos P.
For example, since the Fourier transform F on R satisfies FQF 1 = P and
FPF 1 = Q, it equals U(ei/2 ). The tensor product space U Z , F UZ ,F of two
representation spaces U Z , F and UZ ,F has the following irreducible decomposition
U Z , F UZ ,F =
U Z +Z , F +F n .
(7.29)
n=0
This fact can be easily understood because the weight ( Z + Z , F + F n) has

n orthogonal eigenvectors. In general, in the representation space U Z +Z , F +F n ,
possible weights are limited to ( Z + Z , F + F m), where m is an integer satisfying m n. Each weight has only one eigenvector. These facts show the decomposition (7.29).
7.2 Multi-mode Bosonic System

7.2.1 Multi-variable System
This section addresses the multi-mode bosonic system, whose physical explanation
in Sect. 7.2.2. For this purpose, firstly we focus on the multi-variable system, which
is the same mathematical structure as the multi-mode bosonic system. The multivariable system is the quantum system describing the position of a particle in the
r -dimensional space Rr , and is mathematically given as the Hilbert space L 2 (Rr )
of C-valued square integrable functions on Rr , which is isomorphic to the space
L 2 (R)r . In this system, the physical quantity describing the position of the j-th
coordinate is given as the multiplication operator Q j
240
(Q j f )(x) := x j f (x).
(7.30)
The physical quantity describing the momentum of the j-th coordinate is given as
the operator P j
(P j f )(x) := i
f (x).
x j
(7.31)
Both Q j and P j are self-adjoint operators because they satisfy the conditions Qj =
Q j and Pj = P j [84]. They also satisfy the following commutation relations
[Q j , P j ] = i j, j I, [P j , P j ] = [Q j , Q j ] = 0.
(7.32)
Hence, a skew-Hermitian representation f of h(2r, R) can be defined as
f(P j ) = i P j , f(Q j ) = i Q j , f(Z ) = iI.
(7.33)
The representation is an irreducible representation, and it is known as Stone-von

Neumann Theorem [65] [31, Sect. 1.5] that any irreducible skew-Hermitian representation of h(2r, R) is isomorphic to the above representation.
Exercise 7.6 Show that [Q I + I Q, P I I P] = 0.
7.2.2 Relation Between Multi-mode Bosonic System

and Multi-variable System
Next, we address the bosonic system when the number of possible states of particles
is more than one. Such a quantum system is called the multi-mode bosonic system.
For example, in the quantum optical system, such a system is constructed by the
collection of several optical paths. That is, when we focus only on photons with one
frequency and we discuss several paths for photon, the quantum system is the multimode bosonic system. When we have r paths, the quantum system is the r -mode
bosonic system and the representation space is (2 )r as Fig. 7.2. Then, we denote
the number of particles in the j-th path (mode) by n j . The state vectors |n 1 , . . . , n r
form a CONS of (2 )r . The annihilation operator a j of the j-th mode is defined as
a j |n 1 , . . . , n j , . . . , n r :=
n j |n 1 , . . . , n j 1, . . . , n r ,
and its adjoint operator, the creation operator aj of the j-th mode satisfies
aj |n 1 , . . . , n j , . . . , n r :=

n j + 1|n 1 , . . . , n j + 1, . . . , n r .
7.2 Multi-mode Bosonic System
241
Fig. 7.2 Quantum system of 6 optical paths
Thus, the commutation relation

[a j , aj ] = j, j I
(7.34)
holds. Then, the sum N = N1 + + Nr of operators N j := aj a j gives the number

of particles of all modes.
In the following, similar to the case with one-mode, we identify the sequence
space (2 )r with L 2 (R)r and L 2 (Rr ) via Hermitian functions. Then, we obtain
the relations Q j = 12 (a j + aj ), P j = i 2 (a j aj ), a j = 12 (Q j + iP j ), and
aj = 12 (Q j iP j ). Next, we define an irreducible unitary representation of the
Lie group H(2r, R) from the irreducible skew-Hermitian representation of the
Lie algebra h(2r, R) defined in (7.33). Using a 2r -dimensional real-valued vector

z := (q1 , . . . , qr , p1 , . . . , pr )T and a 2r -dimensional
r operator-valued vector R :=
T
(Q1 , . . . , Qr , P1 , . . . , Pr ) , we obtain the relation j=1 p j Q j q j P j = ( z , Jr R).
Letting j = 12 (q j + i p j ) and using the Hausdorff formula (7.7) and the commuta
tion relation (7.34), we find that the operator Wr () := f(exp( rj=1 p j Q j q j P j ))
satisfies the relations
r

r
W () = exp
jaj jaj
j=1
= e
2
2
exp
r

j=1
j aj exp
r

jaj
(7.35)
j=1
r
1

Wr ( )Wr ( ) = e 2 ( j=1 j j + j j ) Wr ( ( + )).
is the
When the input variable of Wr () is z = (x1 , . . . , xr , y1 , . . . , yr ) and R
T
2r -dimensional vector (Q1 , . . . , Qr , P1 , . . . , Pr ) whose entries are the elements
Q1 , . . . , Qr , P1 , . . . , Pr of the Lie algebra, the operator Wr () equals the operator
r
r

exp(i(
z , Jr R)) and is denoted by W ( z ). Denoting exp(i( z , Jr R) + ) by W ( z , ),
r
W ( z , ) gives a unitary representation of H(2r, R).
242

2
When =1, we define the coherent vector | := |1 , . . ., r := e 2 exp( rj=1
j aj )|0, . . . , 0. Due to (7.35), we have | = Wr ()|0, . . . , 0. That is, the coherent
vector can be written by using unitary representation. Similar to (7.18), we have

Cr
||
d x1 d xr dy1 dyr
= I.
(2)r
(7.36)
7.3 Transformation by Unitary Group U(r)

7.3.1 Correlation by Beam Splitter
In quantum optics, a beam splitter is known as an important optical device in the multimode bosonic system of multiple optical paths. A beam splitter can be constructed
by glass plates and makes an interaction between two optical paths. As Fig. 7.2, the
unitary operator describing the time evolution by the beam splitter is given as the
representation U(g) of g U(r ) satisfying the condition
U(g)a j U(g)1 =

(g ) j, j a j .
(7.37)
j
When the matrix g U(r ) is decomposed into the real part and the imaginary part as
g = g1 + ig2 , where g1 and g2 are real-valued matrices, we can define (g)
:= g1
the action of g on a 2r -dimensional
I2 + g2 J Sp(2r, R). By using the map ,
with operator entries is given as
vector R
1

= (g)
1 R.
U(g)RU(g)
(7.38)
So, the unitary matrix g acts on a unitary operator Wr () as

U(g)Wr ()U(g)1 = Wr (g).
(7.39)
x +i y
When j = j2 j , the relation (7.39) is equivalent to the relation U(g)Wr ( z )U

(g)1 = Wr ((g)
z ).
In the following, we concretely give a skew-Hermitian representation of u(r )
satisfying (7.37) as follows. Taking the derivative at (7.37), for X u(r ), we have
[U(X ), a j ] =
r

X j,l al .
(7.40)
l=1
= (X
Hence, we define
Taking the derivative at (7.38), we have [U(X ), R]
)R.
y
x
E j := iU(i| j j|) and K j,l; := iU(F j,l ) su(F j,l ) according to the discussion
7.3 Transformation by Unitary Group U(r )
243
in Sect. 4.3. So, we have [E j , al ] = j,l al , [K j,l;+ , a j ] = j, j al , and [K j,l; , a j ] =

j ,l a j . Further, the properties of u(r ) yield that
[E j , E j ] = 0, [E j + E j +1 , K j,l; ] = j,l (F jz , j +1 )K j,l; ,
[K j,l;+ , K j,l; ] = E j El .
Thus, we have E j = aj a j + F I , K j,l;+ = aj al , and K j,l; = al a j . Therefore,
we obtain a skew-Hermitian representation of u(r ).
Here, the condition (7.40) for the semi direct product Lie algebra g = h(2r, R)
u(r ) corresponds to the condition (3.42) of the general case. Hence, combining
the above skew-Hermitian representation of u(r ) and the preceding skew-Hermitian
representation of h(2r, R), we can define the semi direct product representation f of
the Lie algebra g = h(2r, R) u(r ).
Then, discussing the representation of the subgroup U(r ), we obtain the irreducible decomposition of the representation space L 2 (R)r as
L 2 (R)r =
Us,n ,
(7.41)
n=0
where Us,n is the space generated by {|n 1 , . . . , n r }n 1 ++nr =n , and isomorphic to nfold symmetric tensor product space of Cr as representation spaces with respect
to SU(r ). Hence, the relation U(g)NU(g)1 = N holds for an arbitrary element
g U(r ).
When the Hamiltonian for the interaction is given as the sum of the product
a j aj of the annihilation and creation operators, the time evolution is written as
the action of u(r ). When the operators U(g) and U(g)1 are applied to the operators
I, Q1 , . . . , Qr , P1 , . . . , Pr , N from both sides, the resultant operators are given as the
linear sum of these operators. Since U(g) transforms these operators linearly, U(g)
is called a linear operation. Also, when the interaction is written as the products
a j aj of the annihilation and creation operators, it is called a linear interaction.
When g O(r ) U(r ) and we regard L 2 (R)r as the Hilbert space L 2 (Rr ) of
n-variable square integrable functions, we find that (U(g) f )(x) = f (gx) by considering the coordinate changes of Q 1 , . . . , Q r . For example, when we apply a
beam
to two optical paths and the interaction between two paths is given
splitter
ab
SU(2), the coherent vectors are changed as Fig. 7.3.
as
cd
Especially, the beam splitter described by the unitary matrix b :=
1
SU(2) is called the beam splitter of transmittance from

1

system 1 to system 2. In particular, U(b1/2 ) is simplified to U1,2 . Similarly, we can
define Uk, j . Using the beam splitter of transmittance , we define the quantum communication channel Att as
244
Fig. 7.3 Beam splitter
a 1 + b 2
Beam splitter
c 1 + d 2
Att () := Tr 2 U(b ) |00|U(b ) ,

which is called the attenuation channel of transmittance 1 . The noisy process
caused by thermal noise is given by the following trace-preserving completely positive map (TP-CP map) N [29]

N (||) =
||e
||2
N
d xd y
.
2 N
(7.42)
Since any state is determined by the P-function, (7.42) uniquely determines the TPCP map. Since N satisfies
N () = lim Tr 2 U(b ) N / U(b ) ,
+0
(7.43)
it can be regarded as a small interaction with a large system 2. In general, the large
system 2 is called the environment system.
Now, to implement the heterodyne measurement {||}, we set the states of the
systems 1 and 2 to be and the vacuum state |00|. Then, we insert them into the
beam splitter of transmittance 21 . Finally, we apply the measurements FQ and FP
of Q and P to the output systems of the systems 1 and 2, respectively. Using these
outcomes x and y, we define the final outcome := x + i y. So, the measurement
with the final outcome is the heterodyne measurement {||} for the system 1. This
fact can be shown by checking that both measurements have the same expectation
of n m = (x + i y)n (x i y)m for any integers n and m as follows.
In the case of the heterodyne measurement {||}, we have

1
n m ||dqdp = Tr
an ||(a )m dqdp
2 C
C

n 1
= Tr a
||dqdp(a )m = Tr an (a )m .
2 C
1
2
245
When the system 1 is input to the beam splitter of transmittance 21 , we have

Tr U(b 21 ) |00|U(b 21 ) (Q1 + iP2 )n (Q1 iP2 )m
= Tr |00|(U1,2 (Q1 + iP2 )U1,2 )n (U1,2 (Q1 iP2 )U1,2 )m

i
1
= Tr |00| (Q1 + Q2 + (P1 P2 ))n
2
2

m
1
i
(Q1 + Q2 ) (P1 P2 )
2
2
= Tr |00|(a1 + a2 )n (a1 + a2 )m = Tr an (a )m .
(7.44)
Therefore, we can conclude that the heterodyne measurement can be realized by

inputting the state to be measured and the vacuum state |00| into the beam splitter
of transmittance 21 .
Here, we need to an additional discussion to show the final equation of (7.44).
Since any density operator can be written as the sum of pure states, it is sufficient to
show the final equation in the pure states case. When = ||, it can be shown as
follows
(a1 + a2 )m | |0 = (a1 + a2 )m1 (a1 |) |0 = ((a1 )m |) |0.
Similarly, we have (| 0|)(a1 + a2 )n = (|a1n ) 0|, which implies the final
equation of (7.44).
7.3.2 Transformation Corresponding

to the Irreducible Decomposition
Using the above beam splitter, we can realize the transformation corresponding to
irreducible decomposition given in (7.29) of the Lie group H(2, R) U(1) of the onemode bosonic system. Firstly, we focus on two representation spaces U Z , F and U0,0
of the Lie group H(2, R) U(1) as one-mode bosonic systems. These two spaces are
isomorphic to L 2 (R). We denote their annihilation operators by a1 and a2 . The Lie
group H(2, R) does not act on the second space, and only the Lie algebra u(1) acts
on it so that the representation of the generator F is given as if(F) = a2 a2 . Then,
the tensor product space U Z , F U0,0 has the following irreducible decomposition
2
U Z , F U0,0 =
n=0 L (R) |n = n=0 U Z , F n .
(7.45)
When two representation spaces U Z ,1 , F,1 and U Z ,2 , F,2 of the Lie group H(2, R)
U(1) are given, the representation space can be reduced to a simple case by applying
246
a suitable beam splitter, as explained later. Then, the tensor product representation
on the two-mode bosonic system, i.e., the composite system U Z ,1 , F,1 U Z ,2 , F,2 is
given as
if(F) = (a1 a1 + a2 a2 ) + F,1 + F,2 ,

1
(if(Q) + f(P)) = Z ,1 a1 + Z ,2 a2 ,
2
if(Z ) = ( Z ,1 + Z ,2 )I.

When we choose g0 =
Z ,1
Z ,1 + Z ,2
Z ,2
Z ,1 + Z ,2

Z ,2
Z ,1+
Z ,2

, the representation U(g0 ) satis
Z ,1
fies U(g0 )a1 U(g0 ) = Z ,1+
a1 +
Z ,2

Z ,1
a1 Z ,1+
a2 . So, we have
Z ,2
Z ,1
Z ,1 + Z ,2
Z ,2
a and U(g0 )a2 U(g0 )
Z ,1 + Z ,2 2
Z ,2
Z ,1 + Z ,2
U(g0 )(if(F))U(g0 ) = a1 a1 a2 a2 + F,1 + F,2 ,

1
U(g0 ) (if(Q) + f(P))U(g0 ) = Z ,1 + Z ,2 a1
2
U(g0 )(if(Z ))U(g0 ) = ( Z ,1 + Z ,2 )I.
That is, the subspace L 2 (R) |nn| of U(g0 )U Z ,1 , F,1 U Z ,2 , F,2 U(g0 ) is an irreducible representation space with the highest weight F,1 + F,2 n. Hence, we
have
U(g0 )U Z ,1 , F,1 U Z ,2 , F,2 U(g0 ) =
U Z ,1 + Z ,2 , F,1 + F,2 n
n=0
= U Z ,1 + Z ,2 , F,1 + F,2 U0,0 .
7.3.3 Representation of Semi Direct Product Lie Group

H(2r, R) U(r) and Its Coherent State*
In the discussion of Sect. 7.2, the Lie algebra of the stabilizer of the Heisenberg group
H(2r, R) did not include any compact maximal Cartan subalgebra of the Lie algebra
of H(2r, R). Hence, we cannot characterize the coherent states by using a compact
maximal Cartan subalgebra as in the previous chapters. To characterize the coherent
states in this way, we need to consider a representation of a larger Lie group. For this
purpose, we focus on a unitary representation of the semi direct product Lie group
H(2r, R) U(r ). To address the unitary representation, we discuss skew-Hermitian
representations of the corresponding Lie algebra h(2r, R) u(r ). Then, we need to
247
fix the root system of the Lie algebra h(2r, R) u(r ). Since the center of the Lie
algebra is the one-dimensional subspace c := RZ generated by Z , we choose the root
system for h(2r, R) u(r )/ c. Then, we fix the root system for the Lie subalgebra
g0 :=< Q 1 , . . . , Q r , P1 , . . . , Pr > u(r ) of h(2r, R) u(r ) that is isomorphic to
h(2r, R)/ c. A maximal compact Cartan subalgebra h of h(2r, R) u(r ) is given as
c d(r ). Then, h0 := g0 (c d(r )) equals d(r ). So, we define j,l , j d(r ) as
elements of the dual space d(r ) of d(r ) as follows.
j,l (X ) := i( j|X | j l|X |l), j (X ) := i j|X | j, X d(r ).
y
(7.46)
x
When Fx j,l , F j,l , Fx j , F j are given as F j,l
, F j,l , 2j , 2j , we can see that the
condition (6.50) holds.
When a maximal compact Lie subalgebra k of g includes h, it is the Lie subalgebra
c u(r ). Hence, h0 is d(r ). The set k+ of positive roots of k is { j,l }1 j<r , and
p+ = + \ k+ is { j }rj=1 .
For the semi direct product representation W U given in Sect. 7.3.1, we have
y
K j = i(W U)(Fx j ) (W U)(F j ) = a j , K j = i(W U)(Fx j ) + (W U)

y
(F j ) = aj , K j,l = K j,l;+ = aj al , and K j,l = K j,l; = al a j . Letting := rj=1
j h0 , we obtain
r
r

c j F jz,u =
cj.
j=0
(7.47)
j=0
Since the relation K |0, . . . , 0 = 0 holds for an arbitrary element + , |0, . . . , 0

is the highest weight vector. Further, the relation (7.47) implies that
if
r

c j F jz,u |0, . . . , 0 =
j=0
r

c j (aj a j + F )|0, . . . , 0
j=0
r
r

z,u
= F
c j |0, . . . , 0 = F
c j F j |0, . . . , 0.
j=0
j=0
Then, the highest weight is (1, F ) (c d(r )) . We denote the representation

space by U1, F .
Since SU(r ) is a universal covering group, we have the unitary representation of
the subgroup G = H(2r, R) SU(r ) for any real number F . Although we have the
decomposition U(r ) = SU(r ) U(1), in order to define the action of U(1), F needs
to be an integer. Hence, the unitary representation on the representation space U1, F
can be extended to a unitary representation of the Lie group G = H(2r, R) U(r )
if and only if F is an integer. When F is a half integer, using the double covering
) of U(r ), we can extend the representation to a unitary representation of the Lie
U(r
).
group H(2r, R) U(r
248
The Lie algebra of the stabilizer H of the highest weight state |0, . . . , 00, . . . , 0|
is c u(r ). According to the discussion in Sect. 6.10, we construct the coherent
vector. We see that the coherent vector is | and that the parameter space G/H
r dp1 dpr
. Hence,
is Cr . The normalized invariant measure on Cr is C (d) = dq1 dq(2)
r
the formal dimension of the representation is 1. Especially, since the relation (6.19)
holds even for a representation of a non-compact group, denoting the representation
of g = e X U(r ) by U(g), we have
U(g)|0, . . . , 0 = ei F (X ) |0, . . . , 0
= ei F i Tr X |0, . . . , 0 = (det g) F |0, . . . , 0.
Hence, the element g U(r ) acts on the coherent vector | as follows
U(g)| = U(g)Wr ()|0 = Wr (g)U(g)|0
= Wr (g)(det g) F |0 = (det g) F |g.
7.4 Second Quantization and Boson

Based on the contents given until now, we explain the structure of the quantum system
more deeply. To discuss the quantum system, we need the two-step procedure for the
quantization. Firstly, we are given a quantum system H that describes the possible
states for one particle. For example, when we fix a specific polarization with a specific
frequency and we prepare r paths for a photon, the quantum system of the possible
states of one photon forms a r -dimensional quantum system H = Cr . Then, the state
corresponding each path is called a mode.
As mentioned in Sect. 4.4.2, when more than one particles can share the same
mode, the type of the particles is called boson, and when more than one particles
cannot share the same mode, the type of the particles is called fermion. Hence,
when the type of the particle is boson, it is possible that two particles take the same
mode. In this case, it is impossible to distinguish two particles. We can distinguish
only the number of particles in each mode. In the following, we focus only on
bosons. As mentioned in Sect. 4.4.2, when the total number of bosons is fixed to n,
the quantum system is given as the n-fold symmetric tensor product space Us,n (H)
of H. Hence, when the total number n of bosons is not fixed and may take an
arbitrary non-negative integer, the total quantum system is described as the space
2
L 2 (H) :=
n=0 Us,n (H). As explained in (7.41), the space L (H) is equivalent to
2
r
the space L (R) . Whereas deriving the quantum system H of states of one particle
is called the first quantization, deriving the total quantum system L 2 (H) is called
the second quantization. In this case, when we apply the coordinate transformation
g U(r ) to the original system H, the system L 2 (H) is naturally transformed by the
representation U(g). In the case of fermion, when the total number is fixed to n, the
7.4 Second Quantization and Boson
249
quantum system is given as the n-fold alternative tensor product space Ua,n (H) of
H. The total quantum system is described as the space rn=0 Us,n (H).
Assume that one boson may take a state in the system H A = {|u j } or H B =
{|v j }. In this case, the system of possible modes of one boson is the product space
H A H B . When more than two particles are possible, the total quantum system is
given as L 2 (H A H B ) =
n=0 Us,n (H A H B ). Since the space Us,2 (H A H B ) is
the symmetric tensor product space, the space is composed of symmetrized vectors
in (H A H B )2 . Here, we consider the case when the respective systems H A and
H B have one particle. In this case, the total number of particles is 2, and the state is
included in the symmetric tensor product space Us,2 (H A H B ). For example, when
the state of the particle in the system H A is |u j1 and the state of the particle in the
system H B is |v j2 , the state of the total system is given as 12 (|u j1 , v j2 + |v j2 , u j1 )
Us,2 (H A H B ). The subspace composed of such elements is isomorphic to the tensor
product space H A H B . That is, the space H A H B can be regarded as a subspace
of Us,2 (H A H B ). In other words, the tensor product space H A H B can be realized
by projection from L 2 (H A H B ) to the subspace H A H B .
In fact, to realize the composite system H A H B of the systems H A and H B
physically, we often employ this method. Even if the particle is a boson whose mode
is given as an element of the system H A H B , the realized system can be regarded as
not the space Us,2 (H A H B ) but the subspace H A H B by making several physical
constraints artificially, i.e., by assuming that the respective systems H A and H B have
only one particle. Indeed, when both systems H A and H B are isomorphic to H,
we denote H A H B by H2 . However, it does not mean that both systems are the
physically same system. When the two particles are in the same system H, in the
bosonic case, the real quantum system is described by the space Us,2 (H), which is
strictly smaller than the system H2 . Hence, in this case, the system H2 cannot be
realized physically.
Similarly, when there exists only one bosonic particle in n respective systems
H1 , . . . , Hn so that there exist totally n bonoic particles, the system H1 Hn is
realized as a subspace of the n-fold symmetric tensor product space Us,n (H1
Hn ). That is, even though n systems H1 , . . . , Hn are mathematically isomorphic to
each other, when they are physically distinguishable as in Sect. 4.4.1, the system
H1 Hn can be realized. The same discussion holds in the fermionic case by
replacing the symmetric subspace by the alternative subspace.
7.5 Squeezed State on One-Mode Bosonic System

In Sect. 3.1.2, the symplectic group Sp(2, R) is introduced as a Lie group preserving the product of Heisenberg group H(2, R). Since the symplectic group Sp(2, R)
corresponds to canonical transformation in analytical mechanics, we can expect that
there exists a projective unitary representation of the symplectic group Sp(2, R) that
corresponds to the canonical transformation in the quantum system. Such a projective
250
unitary representation f of Sp(2, R) is required to satisfy the condition

f(g)W(q, p, )f(g)1 = W(g(q, p, )).
(7.48)
Indeed, a construction of a projective unitary representation f satisfying the above

condition is equivalent to a construction of a projective unitary representation of the
semi direct product Lie group H(2, R) SU(1, 1) as an extension of W.
Taking the derivative with respect to p and q, we obtain the following condition

f(g)Qf(g)1
f(g)Pf(g)1

=g

Q
.
P
(7.49)
We should remark that the Lie group Sp(2, R) is the Lie group SL(2, R), and is
isomorphic to SU(1, 1) via the map given in (4.8). Then, defining a projective
unitary representation S of SU(1, 1) as S(g) := f((g)) for g SU(1, 1), we see
that the condition (7.49) is equivalent to the following condition

S(g)QS(g)1
S(g)PS(g)1
= (g)1

Q
.
P
(7.50)
This condition can be expressed as the action on the annihilation and creation operators a and a in the following way

S(g)aS(g)1
S(g)a S(g)1

=g
a
a
(7.51)
When we define the action g of g SU(1, 1) on a complex number by (4.10), the

condition (7.48) is equivalent to the following condition
S(g)W()S(g)1 = W(g).
(7.52)
Then, taking the derivative in (7.51), for an element X su(1, 1), we obtain

[S(X ), a]
[S(X ), a ]

= X
a
a
Similarly, taking the derivative (7.50), we also obtain

[S(X ), Q]
[S(X ), P]

= (X )

Q
.
P
(7.53)
Hence, under the condition (7.53), the operators E0,1 and K,1 defined in (4.12)
are limited to the operators satisfying the following conditions
251
1
1
a, [E0,1 , a ] = a
2
2
[K+,1 , a] = 0, [K+,1 , a ] = a
(7.54)
[K,1 , a] = a , [K,1 , a ] = 0.
(7.55)
[E0,1 , a] =
The conditions (7.54) and (7.55) imply K+,1 = 21 a2 and K,1 = 21 (a )2 . Further,
the second equation in (4.13) yields E0,1 = [K+,1 , K,1 ] = 21 a a 41 . Hence,
we obtain a skew-Hermitian representation of g = h(2, R) sl(2, R).
The property of the operator E0,1 guarantees that all of weight vectors are number
vectors |n. The highest weight vector is the vacuum state vector |0 and has weight
41 . However, the irreducible space containing the vacuum state vector |0 is the
space generated by number vectors {|2n} with even number, and is the even function space L 2even (R). Hence, there exists an irreducible space containing the highest
weight vector |1 with the highest weight 43 . This space is generated by number
vectors {|2n + 1} with odd numbers, and is the odd function space L 2odd (R). Since
these representations of sl(2, R) have the highest weights 41 and 43 , respectively,
there exist the corresponding unitary representations of the double covering group
R) of SL(2, R), as explained in Sect. 2.4. Hence, there exists a projective repSL(2,
resentation of SL(2, R) on L 2 (R) = L 2even (R) L 2odd (R). Thus, we can construct a
projective unitary representation of the Lie group H(2, R) SL(2, R) such that the
representations of its subgroups satisfy the condition (7.48).
Now, we observe this fact from another viewpoint. This projective representation
z
) gives the repcontains a representation of g = h(2, R) u(1). The operator S(F1
z
resentation of the generator of u(1). While E1 = 2iS(F1 ), the highest weight F
in the previous section is 21 .
Hence, we cannot determine the representation of the subgroup U(1), and can
Since this fact
determine the representation only of the double covering group U(1).
appears in the representation of the Lie group SL(2, R), we can determine the rep R). Here, we define a vector
resentation only of the double covering group SL(2,
R). For a complex numby squeezing the coherent vector by the action of SL(2,
ber D, we define the squeezed vector |2 , 1 := exp( 22 (a )2 22 a2 )|1 =
S(g2 ,1 )|1 . Especially, when 1 = 0, the vector is the coherent vector with respect
R), and is written as |2 , 0 = (1 |2 |2 ) 41 exp( 2 (a )2 )|0.
to SL(2,
2
Next, we investigate the dispersions of the physical quantities P and Q for the
squeezed vector. To discuss the weighted sum of the components of |2 ,1 [Q : P]
with an arbitrary weight, we remember the relation
Tr((g)1 )T (g)1 |2 ,1 [Q : P]
= 2 , 1 |(Q q I, P p I )((g)1 )T (g)1
Q qI
P pI
x+i
y . Then, we choose g = g
2 ,1
2
2
2 2
2
exp( 2 (a ) 2 a ), which implies that
for g SU(1, 1), where 1 =

we have S(g2 ,1 ) =

|2 , 1
defined in (4.7). So,
252

Q qI
|2 , 1
P pI

Q qI
= (g2 ,1 )1
W(g2 ,1 )S(g2 ,1 )|0
P pI

Q qI
qI
= W(g2 ,1 )(g2 ,1 )1
+ (g2 ,1 )
S(g2 ,1 )|0
P pI
pI
(g2 ,1 )1
= W(g2 ,1 )S(g2 ,1 )(g2 ,1 )1

Q qI
qI
(g2 ,1 )
+ (g2 ,1 )
|0
P pI
pI

Q
|0.
= W(g2 ,1 )S(g2 ,1 )
P
So, the relation (7.16) implies

Q qI
|2 , 1
2 , 1 |(Q q I, P p I )((g2 ,1 ) ) (g2 ,1 )
P pI

Q
= 0|(Q, P)S(g2 ,1 ) W(g2 ,1 ) W(g2 ,1 )S(g2 ,1 )
|0
P

Q
= 0|(Q, P)
|0 = 1.
P
1 T
We obtain Tr((g2 ,1 )1 )T (g2 ,1 )1 |2 ,1 [Q : P] = 1. Since (g2 ,1 )

SL(2, R), the relation (5.141) yields that
1
Tr((g2 ,1 )1 )T (g2 ,1 )1 [Q : P]
2

1
det((g2 ,1 )1 )T (g2 ,1 )1 [Q : P]
2
for an arbitrary state . When = |2 , 1 2 , 1 |, since the equality of the above
inequality holds, the equality condition guarantees the relation Tr((g2 ,1 )1 )T
(g2 ,1 )1 [Q : P] = 21 , i.e., the relation |2 ,1 [Q : P] = 21 (g2 ,1 )(g2 ,1 )T .
In general, a 2 2 strictly positive matrix can be written as a constant times
of ((g )1 )T (g )1 by using a matrix g SU(1, 1). Hence, we have the element
2 minimizing 21 Tr((g )1 )T (g )1 [Q : P] dependently of the weight matrix
((g )1 )T (g )1 .
Finally, we focus on the Wigner function W|2 ,1 2 ,1 | () of the squeezed state
|2 , 1 2 , 1 |. Using (7.52), we can calculate the Wigner function as
W|2 ,1 2 ,1 | () = Tr |2 , 1 2 , 1 |W()
)W()S(g1
)1
= Tr S(g2 ,1 )|1 1 |S(g2 ,1 ) W() = Tr |1 1 |S(g1
2 ,1
2 ,1
253
Fig. 7.4 Absolute square of

Wigner function: squeezed
state and coherent state
= Tr |1 1 |W(g1
)
2 ,1
=e
(g1,1 )1 1 (g1,1 )
2
|(g 1 |2
2 ,1
2
(7.56)
Hence, the absolute square of Wigner

function is a Gaussian distribution. Since

cosh + sinh cos
sinh sin
(g2 ,1 ) is the symmetric matrix
,
sinh sin
cosh sinh cos
where 2 = ei (see (4.8)), the covariance matrix of the Gaussian distribution for
p ) is
p and q ( = q+i
2
((g2 ,1 )2

2 ,2 sin
,1 + 2 ,2 cos
=
,1 2 cos ,2
2 ,2 sin
(7.57)
where ,1 := cosh2 + sinh2 and ,2 := cosh sinh . Hence, the squeezed vector is characterized as Fig. 7.4 based on the above covariance matrix.
7.6 Squeezed State on Two-Mode Bosonic System*

In this section, we discuss the squeezing operation
on

the two-mode bosonic system.
For this purpose, we define the action of g =

SU(1, 1) for = (1 , 2 )T
C2 as follows

1
1

.
(7.58)
2
2
254
T
2
That is, we
a vector = (1 , 2 ) C is transformed to another vector
assume that
1 + 2
.
g :=
1 + 2
Then, the representation S2 of SU(1, 1) corresponding to the squeezing operation on the two-mode bosonic system is given by S2 (g)W2 ()S2 (g)1 = W2 (g ).
Comparing the coefficients of complex numbers 1 and 2 , we obtain the following
relation for this representation.
S2 (g)a1 S2 (g)1
S2 (g)a2 S2 (g)1
= g 1
a1
a2
(7.59)
Here, by taking the derivative, an element X su(1, 1) satisfies

[S2 (X ), a1 ]
[S2 (g), a2 ]

= X
a1
a2
Hence, the elements E0,1 and K,1 given in (4.12) satisfy

1
1
a1 , [E0,1 , a2 ] = a2
2
2
[K+,1 , a1 ] = 0, [K+,1 , a2 ] = a1
[E0,1 , a1 ] =
[K,1 , a1 ] =
a2 ,
[K,1 , a2 ]
= 0.
(7.60)
(7.61)
= K+,1 , the relations (7.60) and (7.61) guarantee that K+,1 does not
Since K+,1
contain the operators a1 and a2 . So, we obtain K+,1 = a1 a2 and K,1 = a1 a2 .
Further, the second relation of (4.13) yields
1
1
(a a1 + a2 a2 + I ).
E0,1 = [K+,1 , K,1 ] =
2
2 1
Since a number vector |n, m satisfies

n m|n 1, m 1,
K,1 |n, m = n + 1 m + 1|n + 1, m + 1,
K+,1 |n, m =
the vectors |n, 0 and |0, n are highest weight vectors. So, the irreducible space
. In general, the coherent
containing |n, 0 or |0, n has the highest weight n1
2
state of the representation is called a two-mode squeezed state when the highest
weight vector is the vacuum vector |0, 0. Hence, the representation corresponding
, and can be extended to the
to a two-mode squeezed state has the highest weight 1
2
representation of the Lie groupSL(2, R).
7.7 Wigner Function and Fourier Transform of Heisenberg Representation*
255
7.7 Wigner Function and Fourier Transform

of Heisenberg Representation*
7.7.1 Fourier Transform of Heisenberg Representation
Next, we consider the Fourier transform based on Heisenberg representation. For
this purpose, we remember the discussion in Sect. 7.1.1, which guarantees that
the projective unitary representation of the Lie group R2 with the factor system
i

{e 2 (q pq p ) }(q, p),(q , p ) is limited to the unitary projective representation isomorphic to
the Heisenberg representation W. Hence, using the notation W appearing the Heisenberg representation W, we denote the Fourier transform of the two-dimensional group
R2 with the factor system by FW . Remember that in order to define the Fourier transform with respect to projective unitary representations, we need to fix the factor
system, as explained in Sects. 2.8 and 3.8. Since the irreducible projective unitary
representation with the factor system is unique and has the representation space
L 2 (R2 ), we consider the space L 2 (L 2 (R)) of square integrable space on L 2 (R) as
L 2 (L 2 (R)) := {X : linear operator on L 2 (R)|X 2 := Tr X X < },
(7.62)
which is called the space of HilbertSchmidt operators. Hence, the Fourier transform
is defined as a map FW from L 2 (R2 ) to L 2 (L 2 (R)). That is, letting := 12 (q + i p),
we obtain the Fourier transform of an element f L 2 (R2 ) by

FW [ f ] =
f ()W()
dqdp
.
2
(7.63)
1
For the Inverse Fourier transform, is defined as a map FW
from L 2 (L 2 (R)) to L 2 (R).
2
2
For an operator X L (L (R)), we have
1
[X ]() = Tr W()X,
FW
(7.64)
which equals Wigner function W X () of X . The discussion in Sect. 3.8.2 guarantees

1
1
that FW
FW and FW FW
are identify maps on L 2 (R2 ) and L 2 (L 2 (R)), respec1
are isometric. In this context, the operator
tively. That is, both maps FW and FW
X X corresponds to a density operator on L 2 (R). Thus, the operator X L 2 (L 2 (R))

can be written by use of Wigner function as

X = FW [W X ] =
W X ()W()
dadb
,
2
(7.65)
where := 12 (a + ib). So, when Tr X X = 1, |X is an entangled state with a

suitable reference system and the Wigner function W X contains all information for
the entangled state with respect to the symmetry for the Heisenberg Representation. In
256
1
this section, we employ the notation FW
[X ] instead of the notation W X because it is
better to avoid to denote X as the subscript. Further, when the operator X X describes
a density operator on L 2 (R), the vector |X on the system L 2 (R)2 describes an
entangled state with the reference system L 2 (R).
Here, we calculate Wigner functions, i.e., Inverse Fourier transforms of various
operators while the Wigner functions for the coherent state || and the squeezed
state |2 , 1 2 , 1 | are calculated in (7.20) and (7.56), respectively. First, we focus
on the operators a|| and a ||. Then,
1
FW
[a||]() = Tr W()|| = e
||2
2
+i(apqb)
1
FW
[a ||]() = Tr W()a || = Tr(a )W()||
= ( ) Tr W()|| = ( )e
||2
2
+i(apqb)
So, the relations

||2
a + a
q
e 2 +i(apqb)
|| () = i
p
2
2

q
FW
= i
[||]()
p
2

p ||2 +i(apqb)
a a
i || () = i
FW
e 2
q
2
2

p
FW
= i
[||]()
q
2
1
FW
(7.66)
(7.67)
hold (Exercise 7.9). Since any operator on the space L 2 (R) can be written as a linear
sum of the coherent states {||} , we have

q
a + a
FW
[X ]()
X () = i
p
2
2

a a
p
1
1
i X () = i
FW
FW
[X ]()
q
2
2
1
FW
p , F 1 [X ]() can be regarded as an

for any operator X on L 2 (R). Since = q+i
W
2
element of the function space L 2 (R2 ) with respect to two variables q and p. Then,
when q is the variable for the first mode and p is that for the second mode, we denote
the operations of multiplying q and p by the operators Q1 and Q2 , respectively.
Similarly, we denote the i multiplication of the derivatives withe respect to q and
p by the operators P1 and P2 , respectively. Hence, we have
257

1
1
1
FW
[QX ]() = P2 Q1 FW
[X ]()
2

1
1
1
[PX ]() = P1 Q2 FW
[X ]().
FW
2
Now, for an operator Y on L 2 (R), we define the operator L Y on the operator space on
L 2 (R) by L Y : X Y X . We notice that FW is a unitary operator from the function
space L 2 (R2 ) to the the operator space on L 2 (R). Then, we have
1
1
1
1
L Q FW = P2 Q1 , FW
L P FW = P1 Q2 .
FW
2
2
(7.68)
7.7.2 Uncertainty Relation with Wigner Function

In this subsection, we extend the uncertainty relation for wave packets discussed
in Sect. 5.6 to the case with Wigner function. For this purpose, we focus on the
uncertainty for Wigner function

R2
1
2||2 |FW
[X ]()|2 dqdp,
(7.69)
whose operational meaning is discussed in [50, Sect. 4.3] [48]. To consider a generalization of the uncertainty relation (5.139), we consider the energy constraint
X |Q2 + P2 |X = Tr(Q2 + P2 )X X E
(7.70)
for X L 2 (L 2 (R)). That is, we consider the following minimization problem;

min
X
R2

1
2||2 |FW
[X ]()|2 dqdp X |Q2 + P2 |X E .
(7.71)
In this problem, we address the trade-off between the dispersions of the operators
before and after the Inverse Fourier transform.
1
[X ] of L 2 (R2 ) via
When we transform the operator X to the element := FW
the Inverse Fourier transform, due to the condition (7.68), the condition (7.70) is
transformed to the condition |(P2 21 Q1 )2 + (P1 21 Q2 )2 | E. Then, the
value to be minimized is |Q21 + Q22 |. To discuss this minimization, we consider
the opposite minimization as
2
2

1
1
| P2 Q1 + P1 Q2 |.
2
2
:|Q21 +Q22 |C
min
(7.72)
258
For this minimization, we employ a method similar to the minimization (5.141).

For this purpose, we define the operators A j := Q j and A2+ j := P j . Then, we have
the 4 4 semi-definite matrix J := (Tr Ak A j )k, j . (For its semi-definiteness, see
Exercise 7.8.) For a function , we define as (q,

p) := ( p, q). Then, we
a+b

have |Ak A j | = (1) |Ak A j |, where k , j , a, b are given as follows.
When k = 1, 2, 3, 4, we decide k = 2, 1, 4, 3 and a = 1, 0, 1, 0, respectively. j
2 + Q2 |
= |Q2 + Q2 |
and b are decided in the same way from j. Since |Q
1
2
1
2
1
1
1
2
2
2
and |(P2 2 Q1 ) + (P1 2 Q2 ) | = |(P2 2 Q1 ) + (P1 21 Q2 )2 |,

the minimization problem can be converted to

2
2
1
1
P2 Q1 + P1 Q2
Tr
min
,
2
2
:Tr(Q21 +Q22 ) C
(7.73)
|
+ ||). Due to the symmetry, by using three real numbers
where := 21 (|
0, 0, and , the semi-definite matrix (Tr Ak A j )k, j can be written as
0
:=
i
2
J,,
0 2i
2i
.
0
2i 0
(7.74)
Since for any three real numbers 0, 0, and satisfying J,, 0, there
exists an element g Sp(4, R) such that g T J,, g = J , ,0 (Exercise 7.10). As
explained in the next section, there exists a unitary operator corresponding to the
element g. So, we can choose a state such that J = J,, . Thus, the above minimization problem can be converted to

,
2 ++
,,:2C,J,, 0
4
(7.75)
min
The condition J 0 is equivalent to

0
0
i
2
i
2
i
2
i
2
1
4
+ 2
i
1
4
i
+ 2
(7.76)
This condition is equivalent to || + 41 + 2 . Since it is enough to choose

= 1 (|| + 41 + 2 ) and = C2 , the minimization is converted to

min 2
2
C

C
1 C2
1
4
C
|| + + 2 + +
= min
2 + || + + +
.
C
4
8
2
4
16
(7.77)
259
When C > 2, the minimum equals 1 and is realized when = 21 ( C2 1). When
C 2, the minimum equals C1 + C4 and is realized when = 0. Since the inverse
function of C C1 + C4 is the function E 2(E E 2 1) and this minimization still holds by including mixed states, we have
min
Tr(Q21 + Q22 )
min
|Q21 + Q22 | = 2(E
:Tr((P2 21 Q1 )2 +(P1 21 Q2 )2 )E
:|(P2 21 Q1 )2 +(P1 21 Q2 )2 |E
E 2 1)
(7.78)
when E 1. Since the minimum eigenvalue of P2 + Q2 is 1, we have no minimum

for E < 1.
So, this discussion is summarized as follows.
Theorem 7.1 ([49])

1
min
2||2 |FW
[X ]()|2 dqdp X |Q2 + P2 |X E
X
R2

= 2(E E 2 1)
(7.79)
for E 1.
Exercise 7.8 Show that the Hermitian matrix J := (Tr Ak A j )k, j is positive semidefinite.
Exercise 7.10 Show that for any three real numbers 0, 0, and satisfying
J,, 0, there exists an element g Sp(4, R) such that g T J,, g = J , ,0 .
Exercise 7.11 Show (7.22) by using (7.65).
7.8 Squeezed State on Multi-mode Bosonic System*

Next, we address squeezing operations on a multi-mode bosonic system. The representation of the Lie group SU(1, 1) is not sufficient for describing a squeezing
operation over more than two modes. Alternatively, we employ the representation
Sr of the symplectic group Sp(2r, R) when the number of modes is r . In this case,
i.e., on the r -mode bosonic system, the representation Sr of an element g Sp(2r, R)
r (g)1 = g 1 R.
When we
is defined as a generalization of (7.51) so that Sr (g)RS
employ the map defined in (6.35) with a general form, we have

S(g)1
Sr (g)a
r
S(g)1
S (g)a
= 1 (g 1 )

a

a
(7.80)
260
In the following, to see what representation on the r -mode bosonic system is

derived from the above condition, we simply discuss the structure of the Lie algebra
sp(2r, R) corresponding to Sp(2r, R). The complexification of sp(2r, R) is sp(2r, C),
which is the complexification of sp(r ). Hence, the root system sp(2r,R) of sp(2r, R)
is that of sp(r ). Further, since a maximal compact subgroup of Sp(2r, R) is U(r ), the
root system sp(2r,R) is divided into the compact part u(r ) and the non-compact part
p . As u(r ) contains a maximal Cartan subalgebra of sp(r ), the sets of respective
+
+
positive roots are given as u(r
) = { j,l }1 j<lr and p = { j,l }1 jlr by using
the notations in Sect. 6.6. Especially, sp(2r, R) is Hermitian and r,r is . Since
y
Z j,l , Z j,l , Fx j,l , and F j,l are included in u(r ), they have the same form as in the
y
case of sp(r ). However, Fx j,l and F j,l have different forms from those in the case of
sp(r ), and are given as

Fx j,l
=

F j,l =
1 (| jl|
2
x
+ |l j|) F1
when j < l
x
| j j| F1
when j = l
1 (| jl|
2
+ |l j|) F1 when j < l

y
when j = l.
| j j| F1
Thus, the element of maximal Cartan subalgebra and the representations of Fx j,l and
y
F j,l on the r -mode bosonic system are given in Sect. 7.3. Although the discussion
in Sect. 7.3 gives no constraint for the parameter F corresponding to the highest
weight, it is calculated to be 21 on the r -mode bosonic system due to the following
discussion.
y
We need to fix the representations of Fx j,l and F j,l for discussing the representation Sr on the r -mode bosonic system, instead of their representations, we deal with
the operators K j,l and K j,l . When j = l, the commutation relations among K j,l ,
K j,l , and Z j,l are reduced to the commutation relations given in the case with
su(1, 1). Since only the operators a j and aj appear in the commutation relations,
the analysis of the action of these operators can be reduced to the one-mode bosonic
case discussed in Sect. 7.5. Hence, we have K j,l = 21 a2j and K j,l = 21 (aj )2 . Since
f(| j j|) = aj a j 21 I , the highest weight F given in Sect. 7.3 is decided to be
1
.
2
Similarly, when j < l, the commutation relations among K j,l , K j,l , and Z j,l
are reduced to the commutation relations of su(1, 1). So, only the operators a j , al ,
aj , and al appear in the commutation relations, the analysis of the action of these
operators can be reduced to the two-mode bosonic case discussed in Sect. 7.6. Hence,
we obtain K j,l = a j al and K j,l = aj al .
The actions of K j,l and K j,l preserve the number of total photons, that of
K j,l decreases the number by two, and that of K j,l increases the number by two.
Hence, the representation space L 2 (R)r is divided to the spaces L 2 (R)r :even and
L 2 (R)r :odd , where the space L 2 (R)r :even (L 2 (R)r :odd ) is composed of the states
whose total photon number is even (odd). The highest weight vector L 2 (R)r :even is
7.8 Squeezed State on Multi-mode Bosonic System*
261
|0, . . . , 0 and the highest weight is 21 . Hence, the representation of the subgroup
)
U(r ) of Sp(2r, R) cannot be given, and only that of the double covering group U(r
can be decided. Similarly, the representation of Sp(2r, R) cannot be decided on the
R) can be
r -mode bosonic system and that of the double covering group Sp(2r,
j
decided on the r -mode bosonic system. The space V[K ]|0, . . . , 0 is the space
of states whose total photon numbers is 2 j, and is irreducible with respect to the
representation of u(r ). Hence, the representation of L 2 (R)r :even is u(r )-irreducible.
j
The highest weight vector |0, . . . , 0 of the space V[K ]|0, . . . , 0 is written as
| 21 ; 21 by using the notation given in Sect. 6.9.
R) by
Now, we describe the representation of the double covering group Sp(2r,
+
R)
using Cartan decomposition. Since p = { j,l }1 jlr , any element of Sp(2r,
) and a complex symmetric matrix = {k,l }. So,
can be written by an element g of U(r

its representation is given as U(g) exp( 21 k,l k,l ak al C.C.), where C.C. denotes
the adjoint operator. Here,
there exists
an element g U(r ) such that the matrix
1
2
g g T is a diagonal matrix
.. , where j is areal number. Hence, letting
.
r
Srj ( j ) := exp( 21 j (aj )2 C.C.), we have

1

k,l ak al C.C.
U(g) exp
2 k,l

= U(g)U(g ) U(g ) exp(
1
k,l ak al C.C.)U(g ) U(g )
2 k,l
= U(g)U(g ) Sr1 (1 )Sr2 (2 ) Srr (r )U(g )

= U(g)U(g ) Sr1 (1 )U1,2 Sr1 (2 )U1,2 U1,r Sr1 (r )U1,r U(g ),
R) can be constructed from the comwhich implies that the representation of Sp(2r,
bination of the representations of the squeezing for the first mode and U(r ).
Next, we describe a squeezed vector of the r -mode bosonic system as the generalized coherent vector with respect to sp(2r, R) given in Sect. 6.9. Since a generalized
coherent vector is written by a complex-valued vector, the generalized coherent vector with respect to sp(2r, R) is written with a complex symmetric matrix = {k,l }
as

1
1
1
1

(7.81)
k,l ak al | ; ,
| : = C exp
2
2 k,l
2
2
where C is a normalized constant. Especially, the state | 21 : 21 : | is
transformed via the representation of the unitary matrix g U(r ) as
262
1
U(g)| :
2

1
1
1
= C exp
k,l U(g)ak U(g) U(g)al U(g) U(g)| ;
2 k,l
2
2

1
1
1
1
k,l gl ,l gk ,k ak al | ; = | : gg T .
= C exp
2 k,k ,l,l
2
2
2
When g is chosen as the element g diagonalizing , the above vector is C exp
( 21 (a1 )2 )|0 exp( 2r (ar )2 )|0. Hence, the normalized constant C is calculated
1
1
1
to (1 |1 |2 ) 4 (1 |r |2 ) 4 = (det(I )) 4 . That is, the above vector is the
tensor product |1 , 0 |r , 0 of the one-mode squeezed vector. The inner
product between the above vector and the highest weight vector is calculated to
1
be | 21 ; 21 | 21 : |2 = (det(I )) 2 . The highest weight vector of the
space L 2 (R)r :odd is |1, 0, . . . , 0, and the highest weight is 21 + 21 1,1 .
Chapter 8
Discretization of Bosonic System
Abstract Chapter 4 addressed the Lie groups U(r ) and SU(r ) composed of
unitary matrices on the space Cr . Dependently of the physical situation, all of unitary
matrices do not necessarily have the equal possibility to physically realize. When
we realize them artificially, it is natural that several specific transformations can be
easily realized. Due to such a situation, it is important to address the symmetry of
a specific subgroup as well as that of the Lie groups U(r ) and SU(r ). On the other
hand, the Heisenberg representation discussed in Chap. 7 is a representation on an
infinite-dimensional space, but the Lie group of interest has the finite dimension.
Besides, the representation is irreducible. Hence, we might construct a small subgroup of U(r ) or SU(r ) such that it plays a similar role to U(r ) or SU(r ) if we employ
a similar structure as the Heisenberg representation. For this purpose, we address the
discretizations of the Heisenberg representation and the representation of Sp(2r, R)
that corresponds to the squeezing operation.
8.1 Discrete Heisenberg Representation

8.1.1 Discrete Heisenberg Representation on Zd
In quantum information, when a d-dimensional system H is given, we usually fix one
CONS to describe our information processing. Such a CONS is called the computational basis, and is written as {|0, . . . , |d 1} with modulo d. Our information
processings, e.g., quantum error correction and quantum computation are designed
under the computational basis.
This section investigates the discretization of Heisenberg representation. Since
the Heisenberg representation is a projective unitary representation of the Lie group
R2 on an infinite-dimensional space, replacing R by the commutative algebra Zd :=
Z/dZ, we deal with a projective unitary representation of Z2d [107]. Since the group
Z2d is a commutative group, any irreducible unitary representation is one of the
following one-dimensional representations fs ,t (s , t = 0, . . . , d 1).
fs ,t (s, t)|v := e2i(ss +tt )/d |v, (s, t) Z2d ,

DOI 10.1007/978-3-319-44906-7_8
263
264
8 Discretization of Bosonic System
where |v is the basis of one of one-dimensional representation spaces. That is, there
are d 2 distinct one-dimensional irreducible representations. So, we can easily check
that the relation (2.56) holds.
However, there exists a d-dimensional irreducible projective unitary representation of Z2d . Firstly, we define the following matrices XZ and ZZ on the space H whose
CONS is the computational basis {|0, . . . , |d 1}.
XZ |x = |x + 1, ZZ |x = Zx |x,
(8.1)
Z := ei2/d . This definition is equivalent to the following conditions.

XZ =
|x + 1x|, ZZ =
Zx |xx|.
(8.2)
Z as
Then, we define W
Z (s, t) := Xs Zt =
W
Zxt |x + sx|, (s, t) Z2d .
(8.3)
Since the relation Z XZ = ZX holds, we have

Z (s , t ) = s t W
Z (s, t)W
Z (s + s , t + t ).
W
Z
(8.4)

Z is a projective unitary representation
As the factor system {Zs t } satisfies (2.5), W
2
of Zd . Since the dimension of the representation space is d and the order of the group

Z2d is d 2 , any irreducible projective unitary representation with factor system {Zs t }
Z (s, t) because of the projective representation
is isomorphic to the representation W
version of (2.56). Also, the commutation relation
Z (s , t ) = s tt s W
Z (s, t)
Z (s, t)W
Z (s , t )W
W
Z
(8.5)
(d+1)/2
holds. Here, we define Z to be Z

when d is an odd number, and to be ei/d
when d is an even number. Then, we define the projective unitary representation
WZ (s, t) as
Z (s, t) = st
WZ (s, t) := Zst W
Z
Zxt |x + sx|,
(8.6)
s and t are integers 0, . . . , d 1. Thus, we have

WZ (s, t)WZ (s , t )
s tt s
WZ (s + s , t + t )
when d is an odd number
= Z s tt

(1) Z s WZ (s + s , t + t ) when d is an even number,
(8.7)
265
where is chosen as an integer 0 or 1 dependently of s, t, s , and t . Hence, WZ (s, t)

is a projective unitary
representation of Z2d and can be regarded as a discrete version
of W((q + i p)/ 2). So, we call it the discrete Heisenberg representation of Z2d .
To distinguish W Z from WZ , we call W Z pre-discrete-Heisenberg representation.
Now, we define the dual computational basis for a given the computational basis
{|x}xZd as
1 kl
|eZ (l) :=
Z |k, l Zd .
d kZd
(8.8)
Then, the dual computational basis is composed of the eigenvectors of XZ . The matrix
ZZ cyclically changes the elements of dual computational basis.
Lemma 8.1 The relation
(WZ (s, t))d = I
(8.9)
holds. Hence, when a unitary matrix U , two integers s and t, and a constant c satisfy
U WZ (s, t)U = cWZ (s, t), the constant c is limited to be Zn .
Proof When d is an odd number, the relation (8.7) yields that (WZ (s, t))d =
WZ (ds, dt) = I , which implies the desired argument.
On the other hand, when d is an even number, the relation yields that
Z (s, t))d = st+2st++(d1)st W
(ds,dt) =
(W
Z
Z
Z
std(d1)/2
I = Zstd(d1) I.
Hence,
(WZ (s, t))d = Zdst Zstd(d1) I = Zd
st
I.
Since d is an even number, 2d is a divisor of d 2 . So, we obtain (8.9).
For any element (s, t) Zd and any integer n, matrices on the line
{WZ (ns, nt)}n=0,...,d1 are commutative with each other. When d is a prime p, the
number of non-identity matrices is p 2 1. There are p + 1 lines composed of commutative matrices and each line contains p 1 non-identity matrices, as Fig. 8.1.
That is, we can make p + 1 bases by diagonalizing these matrices.
Further, we can show that the central extension by Z2d based on the factor system (8.7) is isomorphic to discrete Heisenberg group H(2, Zd ) given in Example
2.4. Hence, we can define the map WZ,H as a unitary representation of the group
H(2, Zd ) by
WZ,H (s, t, dn ) := Zn WZ (s, t) when d is an odd number
n
) := Zn WZ (s, t) when d is an even number.
WZ,H (s, t, 2d
266
Fig. 8.1 p + 1 lines in Z p
with p = 5
s
Example 8.1 When d = 2, the discrete Heisenberg representation WZ is given as
follows.

01
,
WZ (0, 0) = I, WZ (1, 0) =
10

1 0
0 i
WZ (0, 1) =
, WZ (1, 1) =
.
0 1
i 0

Further, the set | 1d WZ (s, t)
(s,t)Z2d
forms a CONS on Cd Cd (see Exercise
8.2). So, the set gives a PVM on the composite system Cd Cd . This measurement
is called the generalized Bell measurement, and the states | 1d WZ (s, t) is called the
generalized Bell states.
Exercise 8.1 List up all elements of WZ (k, j) when d = 3.

Exercise 8.2 Show that | 1d WZ (s, t)
forms a CONS on Cd Cd .
2
(s,t)Zd
8.1.2 Discrete Heisenberg Representation on Fq

When d is not a prime number, there is an element s Z2d such that s has no inverse
element. Hence, Zd is not a field, which causes inconveniences in the latter discussion.
To resolve
these inconveniences, we consider another discrete version of W((x +
i y)/ 2) when d is a power q = p m of a prime number p. That is, we focus on

the Galois extension of the finite field F p := Z/ pZ, which is given as the finite
field Fq := F p [X ]/ f (X ) with an irreducible polynomial f with degree m. For this
267
purposes, we define the following matrix XF , ZF on the space H whose CONS is the
computational basis {|x}xFq ;
XF (s)|x = |x + s, ZF (t)|x = Ftr t x |x,
(8.10)
where F is defined to be ei2/ p and tr z is to be Tr Mz when Mz denotes the matrix

representation of the multiplication map x zx with identifying the finite filed Fq
with the vector space Fmp . This definition is equivalent to the following definition.
XF (s) =
|x + sx|, ZF (t) =
xFq
Ztr t x |xx|.
xFq
Since the relation XF (s)ZF (t)XF (s) = Ftr st ZF (t) holds, when we define
tr t x
2
F (s, t) := XF (s)ZF (t) =
W
xFq Z |x + sx| for any pair (s, t) Fq , we have

F (s , t ) = tr s t W
F (s, t)W
F (s + s , t + t ). Because the factor system { tr s t } satW
F
F
F is a projective unitary representation of F2 . Hence, we have the
isfies (2.5), W
q
commutation relation
F (s , t ) = tr (s tt s) W
F (s, t).
F (s , t )W
F (s, t)W
W
F
(8.11)
Since the square of the group Fq2 is the dimension of the representation space of
F (s, t), when the factor system of a projective unitary representation of F2 is equivW
q

alent to {Ftr s t }, the projective unitary representation is limited to a representation
F (s, t).
isomorphic to W
( p+1)/2
when the characteristic p of Fq is not 2, and
Now, we define F as F := Z
we define F as F := i when the characteristic p of Fq is 2. Then, we define the
projective representation WF (s, t) as
F (s, t) = tr ts
WF (s, t) := Ftr ts W
F
Ftr t x |x + sx|,
xFq
where we assume that tr ts takes values in integers 0, . . . , p 1. Then, the relation

WF (s, t)WF (s , t ) =
Ftr (s tt s) WF (s + s , t + t )
when p = 2

tr (s tt s)

(1) F
WF (s + s , t + t ) when p = 2
(8.12)

holds, where is chosen as an integer 0 or 1 dependently of s, t, s , and
t . Hence,
WF (s, t) also can be regarded as a discrete version of W((q + i p)/ 2). So, the
projective unitary representation of Fq2 given here is called the discrete Heisenberg
representation of Fq2 . When q is a prime number p, identifying F2p with Z2p , we
find that the discrete Heisenberg representation of F2p is the discrete Heisenberg
representation of Z2p . However, when q is a power p m of a prime number p, the
268
discrete Heisenberg representation of Fq2 is different from the discrete Heisenberg

representation of Zq2 . We need to remark that a representation based on the finite field
Fq is different from a representation based on the commutative finite algebra Zq .
Here, based on the computational basis {|x}xFq , we define the dual computational basis
1 tr kl
|eF (l) :=
|k, l Fq .
q kF F
(8.13)
Then, the dual computational basis forms the set of eigenvectors of XF . The operator
ZF maps the dual computational basis cyclically.
Further, similar to Lemma 8.1, we obtain the following Lemma.
Lemma 8.2 The equation (WF (s, t)) p = I holds. Hence, when a unitary matrix U ,
two integers s and t, and a constant c satisfy U WF (s, t)U = cWF (s, t), the constant
c is limited to be Fn .
Further, we find that the central extension by Fq2 based on the factor system (8.12)
is isomorphic to the discrete Heisenberg group H(2, Fq ) given in Example 2.6.
Then, we define the unitary representation WF,H of the group H(2, Fq ) as
WF,H (s, t, np ) := Fn WF (s, t) when p is a odd number
WF,H (s, t, 4n ) := Fn WF (s, t) when p = 2.
(8.14)
8.1.3 Multi-mode Discrete Heisenberg Representation

Next, based on the discrete Heisenberg representation of Z2d or Fq2 , we define
2r
a projective unitary representation of the group Z2r
d or the vector space Fq on
r
H = H1 Hr as follows. Here, Hi is a space isomorphic to H. Then, the
computational basis of the space Hr can be written as |s by using the element s of
Zrd or Frq . In the following, to treat the two cases with the finite field Fq and the finite
commutative algebra Zd unifiedly, we denote them by X. That is, we can substitute
Fq and Zd into X in this notation, and X has the structure of a finite commutative
algebra, at least. Then, WX,i (s, t) is defined to be the action WX (s, t) on the i-th
space Hi . So, we can define the projective unitary representation WrX as
WrX (s ) := WX,1 (s1 , t1 ) WX,r (sr , tr ),
r (s ) := W
X,1 (s1 , t1 ) W
X,r (sr , tr )
W
X
for any vector s = (s1 , . . . sr , t1 , . . . , tr ) X2r . Then, for s = (s1 , . . . sr ), t =

(t1 , . . . tr ) Xr , we define (s, t)Z := s1 t1 + + sr tr and (s, t)F := tr (s1 t1 + +
r (s, t)W
r (s , t ) = (s ,t)X W
r (s + s , t + t ). Also, for s =
sr tr ). So, we have W
X
X
X
Z

2r

(s, t), s = (s , t ) X , we define s , s X := (s , t)X (s, t )X . Thus,

WrX (s )WrX (s )
269
s ,s
X X WrX (s + s )
when d or p is an odd number
s ,s
(1) X X WrX (s + s ) when d or p is an even number,
(8.15)
where is an integer 0 or 1 dependently of s, t, s , and t .

r , and Wr are projective unitary representations of Z2r or F2r .
r , Wr , W
Hence, W
q
d
Z
Z
F
F
The projective unitary representation WrZ (WrF ) is called the discrete Heisenberg
2r
representation of Z2r
d (Fq ). Dependently of the situation, we use one of two types
of discrete Heisenberg representations in the following discussion. Also, the coms ,s
mutation relation WrX (s )WrX (s ) = X X WrX (s )WrX (s ) holds, and a similar com X.
mutation relation holds for another projective unitary representation W
We have the following lemma for the representation given here.
Lemma 8.3 The projective unitary representation WrX is irreducible. Further, when
an irreducible projective unitary representation of X2r has the same factor system
r.
as WrX , it is isomorphic to WrX . The same fact holds for W
X
Proof When s = 0, Tr WrX (s ) is 0, and Tr WrX (0) equals q r or d r , which is the
dimension of the representation space. Hence, Lemma 2.15 with the case of projective unitary representation guarantees that WrX is irreducible. Further, due to the
relation (2.56) with the case of projective unitary representation, a projective unitary
representation is isomorphic to WrX (s ) when its factor system equals that of WrX (s ).
r in the same way.

The same fact can be shown for W
X
Further, when d or p is an odd number and the positive integer k is invertible as
an element of X, a projective unitary representation whose factor system is
ks ,s X
ei(s ,s ) = X
(8.16)
satisfies the same fact as Lemma 8.3. That is, when we define the projective unitary
representation WrX:k by replacing X in the definition of WrX by Xk , WrX:k is an
irreducible projective unitary representation whose factor system is (8.16). Then, in
the same way as the proof of Lemma 8.3, we can show that any irreducible projective
unitary representation of Xr with the factor system (8.16) is isomorphic to WrX:k .
From Lemmas 8.1 and 8.2, we obtain the following lemma.
Lemma 8.4 When a unitary matrix U and a constant c satisfy U WrX (s )U =
cWrX (s ) for a given s, the constant c is limited to Xn .
Let WrX Wr X be a projective unitary representation on the space composed of
matrices on Hr . Then, all of its irreducible subspaces are one-dimensional spaces.
So, the irreducible decomposition is given as sX2r < WrX (s ) >. So, the representation is characterized as
s ,s X
WrX Wr X (s ) : WrX (s ) X
Hence, we obtain the following lemma.
WrX (s ).
(8.17)
270
Lemma 8.5 Assume that a unitary matrix U satisfies

WrX (s )U = ei(U,s ) U WrX (s ), X2r ,
(8.18)
where (U, s) is a constant depending on U and . Then, there exists an element

s ,s
s X2r such that the unitary U is a constant times of WrX (s ) and ei(U,s ) = X X .
Proof Modifying (8.18), we obtain WrX (s )U (WrX (s )) = ei(U,s ) U for an arbitrary

element s X2r . So, the above discussion yields this lemma.
Further, we can show that the central extension by Z2r
d based on the factor system
given in (8.15) is isomorphic to the discrete Heisenberg group H(2r, Zd ) defined in
Example 2.5. Now, we define the unitary representation WZ,H of the group H(2r, Zd )
as
WZ,H (s, t, dn ) := Zn WZ (s, t) when d is an odd number
n
) := Zn WZ (s, t) when d is an even number.
WZ,H (s, t, 2d
(8.19)
Similarly, the central extension by Fq2r based on the factor system given in (8.15)
is isomorphic to the discrete Heisenberg group H(2r, Fq ) defined in Example 2.7.
Now, we define the unitary representation WF,H of the group H(2r, Fq ) as
WF,H (s, t, np ) := Fn WZ (s, t) when p is an odd number
WF,H (s, t, 4n ) := Fn WZ (s, t) when p = 2.
(8.20)
In the above discussion, we have given the unitary representations of the discrete
Heisenberg groups H(2r, Zd ) and H(2r, Fq ) from the projective unitary representa2r
tions of Z2r
d and Fq . Conversely, it is possible to derive projective unitary represen2r
tations of Z2r
d and Fq from unitary representations of discrete Heisenberg groups
H(2r, Zd ) and H(2r, Fq ).
Related to this relation, we have the following lemma by using Lemma 8.3.
Lemma 8.6 The unitary representation WrX,H is irreducible. Further, we assume
that when an irreducible unitary representation of the discrete Heisenberg group
H(2r, X) is restricted in the center, the restricted representation is isomorphic to the
representation restricted in the center of WrX,H . Then, the original irreducible unitary
representation of H(2r, X) is isomorphic to WrX,H .
Proof Since the representation WrX,H is irreducible as a projective representation
of the subgroup X2r , the representation WrX,H is irreducible. Given an irreducible
unitary representation of H(2r, X), when it is regarded as a representation of the
center of H(2r, X), it gives the factor system of the projective unitary representation

of the subgroup X2r . Hence, the latter argument follows from Lemma 8.3.
Further, similar to WrX:k , we can define the representation WrX,H:k of the discrete
Heisenberg group H(2r, X). The unitary representation satisfies a lemma similar to
Lemma 8.6.
271
Exercise 8.3 Consider the algebraic extension F4 of F2 by using 2 = + 1. Then,

we define F2 morphism F from F22 to F4 as
(0, 0) 0, (1, 0) , (0, 1) + 1, (1, 1) 1.
Show the following relation;
WF4 (F(x), F(y))|F(z) = WF2 2 (x, y)|z
(8.21)
for x, y, z F22 .
8.1.4 Irreducible Representation of Subgroup

We call a subgroup N of X2r an X-subgroup when the subgroup N is closed with
respect to the multiplication of the finite commutative algebra X. We choose an Xsubgroup N such that s , s X = 0 for any elements s, s N . Such an X-subgroup
is called a self-orthogonal subgroup. In the following, we assume that N is a selforthogonal subgroup. Then, by defining N := {s X2r |s , s X = 0, s N },
an X-subgroup N is self-orthogonal if and only if N N . Especially, when N =
N , it is called a strictly self-orthogonal subgroup. We define the dual space of
N as N := X2r /N . For an element x N , using its representative sx , we define
x(s ) := s , sx X , which does not depend on the choice of the representative of x.
Under the projective unitary representation WrX , the representations of elements
of N are commutative with each other. So, WrX is a representation of the commutative group N . Hence, the irreducible components are one-dimensional. Thus, the
matrices {WX (s )}sN can be simultaneously diagonalized. The eigenvalue is given
by a homomorphism from the group N to the cyclic group {Xn }n , which is the
center of the discrete Heisenberg group H(2r, X). The homomorphism is given as
N s Xx(s ) by choosing a suitable element x N . Hence, we denote the simultaneous eigenspace of the matrices {WrX (s )}sN corresponding to the eigenvalues
Xx(s ) by Hx . Then, we let the projection to Hx be Px , i.e.,
WrX (s ) =

xN
Xx(s ) Px .
(8.22)
Since any element s X2r satisfies

WrX (s )WrX (s ) = WrX (s )WX (s )1 WrX (s )WrX (s )
s ,s
= WrX (s )X X WrX (s )

WrX (s )X[s ](s ) WrX (s ),
we have WrX (s )Hx = Hx+[s ] .

Thus, we have the following lemma.
(8.23)
s X ,
2r
(8.24)
272
Lemma 8.7 Assume that N is a self-orthogonal subgroup. Under the projective

unitary representation WrX , the representation of N on the space Hx is an irreducible
space and is not isomorphic to that on another space Hx . Further, the dimension of
Hx is |X|r /|N |. An irreducible projective unitary representations of N is given on
Hx with a suitable choice of x when its factor system equals that of WrX .
Especially, when X = Fq and N is an r -dimensional vector subspace, the dimension of Hx is q r k .
Proof For an element x N , we assume that the space Hx is not irreducible with
respect to the representation of N . Let Hx be an irreducible subspace of Hx . Since
the above discussion does not depend on the choice of the representative of an
element of Fq2r /N , an element of (N )c maps the space Hx to yN \{0} Hx+y . That
is, the subspace Hx ( yN \{0} Hx+y ) is closed with respect to the action of X2r ,
which contradicts the fact that H is irreducible. Hence, Hx is irreducible. Further,
the representation space Hx of the representation of N is not isomorphic to the
representation space Hx with another x . So, the same fact holds for the subgroup
N .
Due to the above discussion, since Hx and H y are mapped to each other by an
application of a unitary matrix, they have the same dimension. Since the dimension
of the space H is |X|r , the dimension of H y is |X|r /|N |. We also have the relation
|N | = |X|2r /|N |. This value equals (|X|r /|N |)2 |N |, which is the product of the
square of the dimension of the irreducible space and the number of irreducible spaces
in H. Hence, due to the relation (2.56) for the projective unitary representation, the
space H contains all of irreducible projective unitary representations with the same

factor system as WrX .
Next, we discuss the same discussion as the above when N is an X-subgroup of
the discrete Heisenberg group H(2r, X). In this case, we can define a self-orthogonal
subgroup in the same way. In the following, we assume that N is a self-orthogonal
subgroup of H(2r, X). A subset N Z of discrete Heisenberg group H(2r, X) is a
normal subgroup, where Z is the center of H(2r, X). Also, N is an X-subgroup,
and H(2r, X) equals (N Z ) N . Then, we have a lemma similar to Lemma 8.7
for a self-orthogonal subgroup N of H(2r, X). Hence, the inertia group of N given
in Sect. 2.5 is an X-subgroup composed of the identify element. Now, we denote
the restriction of WrX,H | N Z on H0 by WrX,H | N Z ,0 . Then, the irreducible unitary
representation WX,H of discrete Heisenberg group H(2r, X) equals WrX,H | N Z ,0 1

given in Construction 2.
Now, we apply Lemma 2.6 to N Z H(2r, X). Letting H,x be the character of the irreducible unitary representation WrX,H |Hx of N Z , we have the relation H,x (s ) = 0 for any element s N Z \ N Z because C(N Z ) = N Z .
Applying this fact to the character x of the projective unitary representation on Hx
of the X-subgroup N of X2r , we have
x (s ) = 0, s N \ N .
(8.25)
273
Exercise 8.4 Give a non-trivial example of a strictly self-orthogonal subgroup of

F42 .
8.2 Discrete Symplectic Group and Clifford Group

8.2.1 Definitions of Discrete Symplectic Group
and Clifford Group
In this subsection, to address the operation corresponding to squeezing operation
over the finite field Fq and the algebra Zd , we deal with the symplectic group
(discrete symplectic group) Sp(2r, Fq ) and Sp(2r, Zd ). Similar to the symplectic
group Sp(2r, R) on R, these groups are given as
Sp(2r, Fq ) := {g GL(2r, Fq )|g T Jr g = Jr }
Sp(2r, Zd ) := {g GL(2r, Zd )|g T Jr g = Jr },
where GL(2r, Fq ) and GL(2r, Zd ) are the groups composed of invertible square
matrices of oder 2r on Fq and Zd , respectively. Now, we denote the group composed
of square matrices of order 2 with determinant 1 (2-dimensional special linear group)
by SL(2, Fq ) and SL(2, Zd ), respectively. So, similar to the case of real numbers, we
have SL(2, Fq ) = Sp(2, Fq ) and SL(2, Zd ) = Sp(2, Zd ). The order | Sp(2r, Fq )| of
2
r
2k
the group Sp(2r, Fq ) is known to be q r
1) [2, p. 147]. When q = p 2 ,
k=1 (q
2
4
4
| Sp(2, F p2 )| = p ( p 1) while | Sp(4, F p )| = p ( p 2 1)( p 4 1). So, we find
that Sp(4, F p ) is much larger than Sp(2, F p2 ) (see Exercise 8.8).
On the other hand, in the case of Zd , when d is a product of powers of distinct
prime numbers p1 . . . , pl , we have | Sp(2, Zd )| = d 3 (1 p12 ) (1 pl2 ) [81,
Lemma 3.2], which can be checked as follows. Firstly, we consider | Sp(2, Z pk )|.
Assume that we denote the entries of an element of Sp(2, Z pk ) by ai, j . The number
of elements with invertible a1,1 is p k p k ( p k p k1 ). When a1,1 is not invertible, a2,1
must be invertible. The number of such elements is p k p k1 ( p k p k1 ). So, the total
number is
| Sp(2, Z pk )| = ( p k )2 ( p k p k1 + p k1 p k2 ) = ( p k )3 (1 p 2 ).
(8.26)
Next, we assume that the prime factorization of d is given as d = p1k1 plkl . Then,
Zd is isomorphic to Z pk1 Z pkl . So, Sp(2, Zd ) is isomorphic to Sp(2, Z pk1 )
1
1
l
Sp(2, Z pkl ). Hence, using (8.26), we have [81, Lemma 3.2]
l
| Sp(2, Z pk1 pkl )| = ( p1k1 plkl )3 (1 p12 ) (1 pl2 ).

1
(8.27)
274
In the following, X2r and Sp(2r, X) express Fq2r and Sp(2r, Fq ) for X = Fq and
do Z2r
d and Sp(2r, Zd ) for X = Zd .
Further, the symplectic group Sp(2r, X) naturally acts on the commutative
group X2r . Using this action, we can naturally define the semi direct product
X2r Sp(2r, X), which is called the Clifford group.1 An element of Clifford group
(s , g) X2r Sp(2r, X) naturally acts on s X2 as
(s , g) : s (gs ) + s.
Example 8.2 When q = 2 and r = 1, F22 is isometric to Klein four-group Z2 Z2 .
Also, the order of Sp(2, F2 ) is 6. The all of elements of the group are listed as

10
11
01
,
,
01
10
11

01
11
10
,
,
.
10
01
11
The orders of the second and third elements are 3. The orders of the 4th6th elements
are 2. Hence, this group is isomorphic to the permutation group of degree 3 (see
Exercise 2.1).
The action of Sp(2, F2 ) on F22 is the same as the action of S3 on Z2 Z2 given
in Example 2.2. Hence, the Clifford group F22 Sp(2, F2 ) is isomorphic to (Z2
Z2 ) S3 , i.e., S4 . In Example 8.3, we will see that the Clifford group F22 Sp(2, F2 )
is isomorphic to the regular hexahedron group, which is also isometric to S4 , in a
different way. For the regular hexahedron group, see Example 2.2 and Fig. 2.13.
Exercise 8.5 Show that Sp(2, Fq ) = SL(2, Fq ) and Sp(2, Zd ) = SL(2, Zd ).
Exercise 8.6 List up all elements of Sp(2, F3 ).
Exercise 8.7 Compare two groups GL(2, F2 ) and F
4 by using the F2 -morphism F
defined in Exercise 8.3. For this comparison, give an element of GL(2, F2 ) that does
not belong to F
4 via the F2 -morphism F.
Exercise 8.8 Compare two groups Sp(2, F4 ) and Sp(4, F2 ) by using the
F2 -morphism F defined in Exercise 8.3. For this comparison, give an element of
Sp(4, F2 ) that does not belong to Sp(2, F4 ) via the F2 -morphism F.
8.2.2 Generator
In the following, we investigate the generators of Sp(2, Fq ). When all of diagonal
entries of an element g of Sp(2, Fq ) are not zero, there exist elements a, a , b Fq
such that its Gaussian decomposition like (4.2) is available as
1 Appleby
[3] extended the Clifford group.

g=
1a
01

275
b1 0
0 b

1 0
.
a 1
(8.28)
That is, choosing

Mb :=
b1 0
0 b

, Q a :=

10
1a
, Pa :=
,
a1
01
(8.29)
we have g = Pa Mb Q a . Now, we denote the entries of g by ak,l . On the other hand,

when a1,1 is 0, all of diagonal entries of g J are invertible. Hence, any element
of Sp(2, Fq ) can be expressed by the products of Q a , Pa , Mb , J . Especially, since
Pa = J Q a J , the generators of Sp(2, Fq ) can be restricted to {Q a }a {Mb }b {J }.
Now, we assume that p = 2. We choose a specific element a0 Fq such that there
does not exist b Fq satisfying a0 = b2 . Then, any element of Fq except for 0 can be
written as b2 or b2 a0 . Hence, due to the relation Mb Q a Mb1 = Q b2 a , the generators of
Sp(2, Fq ) can be restricted to {Q 1 } {Q a0 } {Mb } {J }. Further, since Mb Mb =
Mb b , the generators of Sp(2, Fq ) can be restricted to {Q 1 } {Q a0 } {Mb }bG (Fq )
{J }, where G (Fq ) is the generators of Fq with respect to multiplication.
Now, we assume that p = 2. Since x x 2 is bijective on Fq , any element of Fq
except for 0 can be written as b2 . Hence, the generators of Sp(2, Fq ) can be restricted
to {Q 1 } {Mb }bG (Fq ) {J }.
Then, the generators satisfy the following relations
Q a Q a = Q a +a , Mb Q a = Q ba Mb , J Mb = Mb J, J 2 = I.
(8.30)
When all of diagonal entries of an element of Sp(2, Zd ) are invertible, the decomposition (8.28) is available. Hence, the above discussion can be applied to the case
of Sp(2, Zd ).
Next, we focus on the group Sp(2r, Fq ). We choose two r r matrices and
on Fq such that T = and det = 0. That is, is a symmetric matrix and is an
invertible matrix. We define three elements M , Q , and P of Sp(2r, Fq ) as

M :=
1 0
0 T

,
Q :=
Ir 0
Ir

,
P :=
Ir
0 Ir

.
Then, the matrices Jr , Q , P , and M satisfy the above relations. In the following,
we denote the basis of the vector space Frq on the finite field Fq by {|k)}rk=1 . Then,

for a Fq , we define a matrix k,l:a := a|k)(l| + rk =1 |k )(k | as an element of the
general linear group GL(Fq , r ) on the finite field Fq .
Consider an element of Sp(2r, Fq ) that has non-zero entries only on ( j, j), ( j, j +
r ), ( j + r, j), ( j + r, j + r ) entries. The set of such elements forms a subgroup of
Sp(2r, Fq ) that is isomorphic to Sp(2, Fq ). So, Sp(2r, Fq ) can be regarded as a sub-
276
group of Sp(2r, Fq ). Hence, the direct product group Sp(2, Fq ) Sp(2, Fq ) is

r
also a subgroup of Sp(2r, Fq ). Similarly, Fq Fq is a subgroup of GL(Fq , r ).

r
Then, we have the following theorem.

Theorem 8.1 When p = 2, the group Sp(2r, Fq ) is generated by the subgroup
Sp(2, Fq ) Sp(2, Fq ) and the set {M j,l:1 } j,l {M j,l:a0 } j,l , where a0 Fq is

r
defined above. When a0 is given as a finite sum of elements of {a 2 }aFq , the group
Sp(2r, Fq ) is generated by Sp(2, Fq ) Sp(2, Fq ) and {M j,l:1 } j,l .

r
When p = 2, the group Sp(2r, Fq )

Sp(2, Fq ) Sp(2, Fq ) and {M j,l:1 } j,l .

is
generated
by
its
subgroup
To show this Theorem, we prepare the following Lemma.

Lemma 8.8 (1) When p = 2, the group GL(Fq , r ) is generated by the subgroup
Fq Fq and { j,l:1 } j,l { j,l:a0 } j,l . (2) When a0 is given as a finite sum of

r
elements of {a 2 }aFq , it is generated by the subgroup Fq Fq and { j,l:1 } j,l .

r
(3) When p = 2, the group GL(Fq , r ) is generated by the subgroup its subgroup
Fq Fq and { j,l:1 } j,l .

r
Proof of Lemma 8.8 Elementary operations for matrices show that GL(Fq , r ) is
generated by its subgroup Fq Fq , {M j,l:a } j,l,aFq , and the set of permuta

r
tion matrices T j,l := | jl| + |l j|. Let D j (x) be the diagonal matrix whose j-th
diagonal entry is x and whose other diagonal entries are 1. Then, we have T j,l =
D j (1) j,l:1 l, j:1 j,l:1 . Hence, it is sufficient to show that j,l:1 can be generated
by the proposed subset. Now, we have D j (b)Dl (b1 )l, j:a D j (b1 )Dl (b) = l, j:b2 a
for any a Fq .
For (1), the above relation with a = 1 and the choice of a0 yield that j,l:1 can
be generated by the proposed subset. For (2), the relation Q a Q a = Q a +a guarantees
the desired argument. For (3), there is an element b such that b2 (1) = 1 when
p = 2. So, the above relation with a = 1 guarantees the desired argument.

Proof of Theorem 8.1 It is known that Sp(2r, Fq ) is generated by {Jr } {Q } {M }
2r
[80, Theorem 3.3].2 Let {|k)}2r
k=1 be the basis of the vector space Fq on the finite
field Fq . Then, we define the matrix Jr, j as
the paper [80] assumes the condition p = 2, it shows this fact in this case. However, the
proof is still valid even when p = 2.
2 Since
Jr, j = | j)( j + r | + | j + r )( j| +
277

(|k)(k| + |k + r )(k + r |),
(8.31)
k= j
which is contained in the subgroup Sp(2, Fq ) Sp(2, Fq ). Since the equa

r
tion Jr = Jr,1 Jr,r holds, Jr is also contained in the above subgroup. Further,
since M Q M 1 = Q T , any symmetric matrix can be diagonalized by applying the matrix M from both sides. Let j (x) be the matrix whose jth diagonal entry
is x and whose other diagonal entry is 0. Then, the matrix j (x) is contained in
the above subgroup. Hence, we obtain the desired argument from these facts and
Lemma 8.8.
8.2.3 Structure of Self-orthogonal Subgroup

In the following, we consider the expression of the self-orthogonal subgroup N on
the finite filed Fq2r based on an element of the symplectic group Sp(2r, Fq ). Let e j be
the element of Fq2r whose jth entry is 1 and whose other entries are 0. When the selforthogonal subgroup N is isomorphic to Fqk , i.e., the dimension of N is k, we have
k r . As a typical case, we define the Fq -subgroup Fq(k) generated by e1 , . . . , ek .
Hence, we can choose an element g Sp(2r, Fq ) such that any k-dimensional selforthogonal subgroup N is the image of Fq(k) by the map g. Especially, the image of Fq(k)
by the map g is the Fq -subgroup generated by the vectors e(g)1 , . . . , e(g)k , where
e(g) j is the jth column vector of g. Now, we call a 2r k matrix a generating matrix
of the self-orthogonal subgroup N when its image is the self-orthogonal subgroup
N . Hence, defining the 2r k matrix A(g) as the matrix composed of k vectors
e(g)1 , . . . , e(g)k , we find that A(g) is a generating matrix of the self-orthogonal
subgroup N . Also, for any element h GL(k, Fq ), A(g)h is a generating matrix of
the self-orthogonal subgroup N .
Especially, when N is a strictly self-orthogonal subgroup, we have k = r . So, it is
composed of two square matrices of order r , the upper half and the lower half. When
the upper half is invertible, by multiplying the inverse of the upper matrix from the
left side, the generating matrix A(g) is transformed to a matrix whose upper half is
the identity matrix. The lower half of the transformed matrix is an alternative matrix
because N is a strictly self-orthogonal subgroup. Hence, unless the upper half of
the generating matrix is non-invertible, any strictly self-orthogonal subgroup can be
described by an r r alternative matrix on the finite filed Frq . Note that when
p = 2, a matrix is alternative if and only if it is symmetric. So, an alternative matrix
may have non-zero diagonal element in this case.
Exercise 8.9 Give a generating matrix of a strictly self-orthogonal subgroup given
in Exercise 8.4.
278
8.3 Metaplectic Representation

8.3.1 Existence of Projective Unitary Representation
Next, we give a projective unitary representation of the symplectic group Sp(2r, X)
on Hr . Firstly, we consider the case when d or p is an odd number. Then, for two
elements g, g of Sp(2r, X), we have
s , s X = g(s ), g(s )X .
(8.32)
Due to (8.15), this fact shows that the projective representation s WrX g(s ) is
isomorphic to the projective representation s WrX (s ). Since Lemma 8.3 guarantees
that the projective representation s WrX (s ) is irreducible, there exists a unitary
matrix SrX (g) such that
SrX (g)WrX (s )SrX (g) = WrX (g(s )).
(8.33)
Further, since
SrX (g)SrX (g )WrX (s )SrX (g ) SrX (g) = WrX (gg (s )) = SrX (gg )WrX (s )SrX (gg ) ,
there exists a real number (g, g ) such that SrX (g)SrX (g ) = ei(g,g ) SrX (gg ). Hence,
g SrX (g) is a projective unitary representation. The projective unitary representation is called the Metaplectic representation of the symplectic group Sp(2r, X).
On the other hand, when d or p is an even number, the similar fact can be shown in
the same way basically. However, we need some additional ideas. For two elements
g, g Sp(2r, X), the relation (8.32) holds. However, the projective representation
s WrX (g(s )) is not necessarily isomorphic to the projective representation s
WrX (s ).
Here, we choose generators of X2r = Fq2r , Z2r
d as follows. When X = Zd , we
consider the vector e j whose jth entry is 1 and whose other entries are 0. Then,
2r
Z2r
d is a commutative group generated by {e j } j=1 . When X = Fq , we regard Fq as
a vector space on the finite field F p , and focus on the basis { f k } of Fq . Let f k, j be
the vector in Fq2r whose jth entry is f k and whose other entries are 0. Then, Fq2r is
a commutative group generated by { f k, j }. We denote the generators by {e j } with a
proper reordering.
The above given discrete Heisenberg representation can be given by the representations WrX (e j ) of the generators {e j } because the representation of another
element can be given from the products of those of the generators in the same
way as Sect. 8.1. Here, we define another projective representation s W rX (s ) by
W rX (e j ) = W rX (g(e j )). Here, the representation of another element is defined by
the above method. Then, since the commutation relation among the representations
of the generators for W rX is that of WrX , there exists an integer g,s for s such that
279
W rX (s ) = (1)g,s WrX (g(s )). Thus, because W rX (s ) is irreducible and is isomorphic

to WrX (s ), there exists a unitary matrix SrX (g) such that
SrX (g)WrX (s )SrX (g) = (1)g,s WrX (g(s )).
(8.34)
Further, since
(1)g ,s +g,g(s ) SrX (g)SrX (g )WrX (s )SrX (g ) SrX (g)
= WrX (gg (s )) = (1)gg ,s SrX (gg )WrX (s )SrX (gg ) ,
we have g ,s + g,g (s ) = gg ,s (mod) 2. So, there exists a real number (g, g ) such

that SrX (g)SrX (g ) = ei(g,g ) SrX (gg ). Hence, we find that g SrX (g) is a projective
unitary representation. Similar to the odd number case, this representation is called
the Metaplectic representation of the symplectic group Sp(2r, X). The properties
(8.33) and (8.34) guarantee that the representation is faithful.
8.3.2 Concrete Construction of Representation

In this subsection, for the case of X = Fq , we concretely construct the Metaplectic representation of Sp(2r, Fq ). It suffices for this purpose to give a matrix SrF (g)
for a generator g such that the relations (8.33) and (8.34) hold. When p = 2, the
representations of the generators are given as [80, Theorem 4.1]
SrF (Jr, j ) := Ir( j1)
SrF (M ) :=
1

tr x y
|xy|
Ir(r j)
F
q 1/2
(8.35)
x,yFq
| xx|
(8.36)
xFrq
SrF (Q ) :=
(x, x)F
|xx|.
(8.37)
xFrq
When r = 1, SrF (Jr ) is the discrete Fourier transform. Since {M } is the subgroup
GL(r, Fq ), the relation (8.36) can be regarded as the definition of a projective representation of the subgroup GL(r, Fq ), which is an invertible multiplication of on
the computational basis. Then, we can show that
SrF (M )|eF (l) = |eF ( 1 l).
280
Also, the above defined matrix SrF (g) satisfies

SrF (Jr, j )WrX (s, t)SrF (Jr, j ) = WrX (Jr, j (s, t))
SrF (M )WrX (s, t)SrF (M ) = WrX (s, 1 t)
SrF (Q )WrX (s, t)SrF (Q ) = WrX (s, t + s).
Hence, Theorem 8.1 guarantees that the relations (8.35), (8.36), and (8.37) give
a projective unitary representation satisfying (8.33). That is, due to Theorem 8.1,
we can implement the representation SrF physically by implementing the unitary
matrices corresponding to the projective representation of Sp(2, Fq ) on each system
H j and the unitary matrices {SrF (M j,l:1 )} j,l and {SrF (M j,l:a )} j,l across two quantum
systems H j and Hl . Further, it is known that, given a projective unitary representation SrF (g), there exist scalars c(g) such that c(g)SrF (g) is a unitary representation
[80, Theorem 4.3].
On the other hand, when p = 2, we can define SrF (Jr, j ) and SrF (M ) in the same
way. However, we need to define SrF (Q ) in a different way from the above. Since
the given generators contain only Q j (y) among the elements {Q }, we need to construct only unitary matrices SrF (Q j (y) ). For this purpose, we regard the finite field
Fq as a vector space Fm
2 on the finite field F2 , and choose the CONS {e1 , . . . , em }
such that tr e j ek = j,k . Then, we define the integer wgt(x) to be the number
of elements #{ j {1, . . . , m}|tr (xe j ) = 0}. So, for two elements x, y Fq , we
= 1}, where x j := tr (xe j )
have wgt(x + y) = wgt(x)
+ wgt(y) 2#{ j|x j = y j
F2 . Further, since x = j x j e j , we have tr (x y) = j x j y j = #{ j|x j = y j = 1
( mod 2)}. Hence,
wgt(x+y)
wgt(x)+wgt(y)
= F
tr (x y)
(8.38)
Because the map x x 2 is bijective on Fq , for an arbitrary element y Fq , there

exists its square root y Fq . Using this fact, we can define [46]
SrF (Q j (y) ) := I j1
wgt(y x)
|xx| I r j ,
(8.39)
xFq
which satisfies (8.34), as will be shown later. Hence, due to Theorem 8.1, the relations
(8.35), (8.36), and (8.39) give a projective unitary representation satisfying (8.33).
That is, due to Theorem 8.1, we can implement the representation SrF physically by
implementing the unitary matrices corresponding to the projective representation of
Sp(2, Fq ) on each system H j and the unitary matrix SrF (M j,l:1 ) across two systems
H j and Hl . Especially, in this case, the unitary matrix SrF (M j,l:1 ) is called the Control
NOT (C-NOT) operation with the control system Hl and the target system H j .
Indeed, any quantum circuit can be constructed from the combinations of C-NOT
and local operations as Fig. 8.2. Hence, it is an important topic in physics to implement
the C-NOT operation physically.

Fig. 8.2 Quantum circuit
281
Controlled not
X
Z
Proof of (8.34) In the following, we show that the unitary matrix defined in (8.39)
satisfies the condition (8.34) for the projective unitary representation SrF . For this
purpose, it is enough to discuss only the j-th Hilbert space. That is, the problem can
be reduced to the case when r = 1, i.e., it is enough to show the condition (8.34) for
the case of r = 1. Now, we abbreviate S1F to SF in this case. Due to (8.38), we can
check that
wgt(y s)tr (y s)2
SF (Q y )WF (s, t)SF (Q y ) = F
WF (s, t + ys).
Here, due to (8.34), there exists a unitary matrix U such that

U WF (s, t)U = (1) WF (s, t + ys).
Hence,
wgt(y s)tr (y s)2
(1) U WF (s, t + ys)U = F
SF (Q y ) WF (s, t + ys)SF (Q y ).
wgt(y s)tr (y s)2
Thus, Lemma 8.4 guarantees that F

is 1 or 1. That is, wgt(y s)
2
tr (y s) = 0 mod 2. Therefore, we have shown that the unitary matrix defined in
(8.39) satisfies the condition (8.34).

The above discussion can be applied to the group Sp(2r, Zd ). That is, the representations of Jr , M , and Q can be given as the above way. Further, since these
elements generate the group Sp(2, Zd ), its projective representation can be concretely
constructed from the above method.
8.3.3 Properties of Irreducible Decomposition

When X = Fq and p = 2, the representation space Hr of the Metaplectic representation is divided into two irreducible subspaces as [80, Theorem 4.2]
282
Hr :even :={| Hr |s Frq , s| = s|}

Hr :odd :={| Hr |s Frq , s| = s|}.
The two irreducible spaces Hr :even and Hr :odd correspond to the discrete versions
of the even function space L 2 (R)r :even and the odd function space L 2 (R)r :odd ,
respectively, which are the irreducible spaces of the projective unitary representation given in Sect. 7.8 of the symplectic group Sp(2r, R). The dimensions of these
subspaces are given as
dim Hr :even =
qr + 1
qr 1
, dim Hr :odd =
.
2
2
On the other hand, we have the following theorem in the case of p = 2.

Lemma 8.9 When p = 2, the Metaplectic representation SrF is irreducible.
Proof The symplectic group Sp(2r, Fq ) contains the direct product group
Sp(2, Fq ) Sp(2, Fq ) as a subgroup. Hence, due to Lemma 2.8, it is enough

r
to show only the case of r = 1. When we focus on the subgroup generated by

Q y , any irreducible subspaces of the representation of the subgroup are given as
1-dimensional spaces generated by |x. These 1-dimensional representation spaces
are not isometric to each other.
Next, we consider the action of element M . So, we find that |x and |x can
be mapped to each other for two non-zero elements x and x of Fq . Since the
element Jr maps the vector |0, we find that the Metaplectic representation S1F is
irreducible.

Further, since the discrete Heisenberg representation and the Metaplectic representation satisfy the condition (2.39), as mentioned in (8.33) and (8.34), we can define
the semi direct product representation VrX of the semi direct product X2r Sp(2r, X).
n(g,s ,s ) r
Hence, VrX (s , g)WrX (s )VrX (s , g) = X
WX (g(s ) + s), where n(g, s, s ) is an

integer depending on g, s, s . Thus, the semi direct product representation VrX is
faithful.
Next, we define the group
CX (Hr )
:= {U U(Hr )|s X2r , (s , g), s X2r , U WrX (s )U = ei(s ,U ) WrX (s )}.
Then, the group CX (Hr ) is a subgroup of U(Hr ). The subgroup naturally contains
the group U(1) as its subgroup of the scalar multiplication. Then, the following
theorem holds.
Theorem 8.2 The semi direct product representation VrZ of the discrete Heisenberg representation and the Metaplectic representation of the Clifford group Z2r
d
283
Sp(2r, Zd ) gives the isomorphism between the Clifford group Z2r

d Sp(2r, Zd ) and
the group CZ (Hr )/ U(1).
Proof Since the semi direct product representation is faithful, the semi direct product
r
representation is an injective map from the group Z2r
d Sp(2r, Zd ) to U(H ). On the
other hand, we can show that the semi direct product representation is a surjective
r
map from the group Z2r
d Sp(2r, Zd ) to C Z (H )/ U(1) by the following way.
For an element U C Z (Hr ), we choose a vector u(s ) such that g WrZ (s )g =
ei(s ,U ) WrZ (u(s )). Since the commutation relation between WrZ (s ) and WrZ (s ) is the
commutation relation between WrZ (u(s )) and WrZ (u(s )), we obtain
s , sZ = u(s ), u(s )Z .
(8.40)
Hence,
ei(s +s ,g) WrZ (u(s + s)) = g WrZ (s )WrZ (s )g = g WrZ (s )g g WrZ (s )g
= ei((s ,g)+(s ,g)) WrZ (u(s ))WrZ (u(s )).

This fact indicates that u(s + s) = u(s ) + u(s ). In particular, since u(0) = 0, we
have u(n s + ns ) = n (s ) + nu(s ) for any two integers n, n . Because this fact
shows the linearity of u, the map u can be written as a matrix. Further, the relation (8.40) guarantees that the map u is an element of Sp(2r, Zd ). Hence, we have

SrZ (u) U WrZ (s )U SrZ (u) = (1) ei (s ,U ) WrZ (s ). Due to Lemma 8.5, there exists an
r
s ) is a constant times of SrZ (u) U . Therefore, we
element s Z2r
d such that WZ (
obtain the desired argument.

Thus, we obtain the following lemma.
Lemma 8.10 The representation space of the tensor product representation between
the projective unitary representation VrF of the Clifford group Fq2r Sp(2r, Fq ) on the
space Hr and its complex conjugate representation VrF can be regarded as the space
composed of the linear maps form Hr to Hr . Then, the representation space is
irreducibly decomposed to the one-dimensional space spanned by the identity matrix
and its orthogonal complement.
Proof Firstly, we focus on the representation of the subgroup Fq2r . Then, each irreducible subspace is a one-dimensional subspace spanned by WrF (s ). For an arbitrary
non-zero element s, s Fq2r , there exists an element of Sp(2r, Fq ) that maps WrF (s )
to WrF (s ). Hence, < WrF (0) > and < {WrF (s )}s=0 > are invariant subspaces, which
shows the desired argument.

Example 8.3 When q=2 and r =1, we consider the Clifford group Fq2r Sp(2r, Fq )
and the representation VrF VrF . Now, we restrict this representation to the subspace
< WF (1, 0), WF (0, 1), WF (1, 1) >. Due to Example 8.1, this space is composed of
all of Hermitian traceless matrices. The space is isomorphic to R3 . The concrete form
284
of WF given in Example 8.1 shows that g,s in (8.34) is 0 for any g and s. Hence,
by regarding WF (1, 0), WF (0, 1), WF (1, 1) as the axis of the space R3 , the action of
the Clifford group F22 Sp(2, F2 ) on the space R3 is the natural action of the regular
hexahedron group on the space R3 .
When the Clifford group Fq2r Sp(2r, Fq ) is regarded as a subset of U(r ) via the
above representation based on Lemma 8.10, the Clifford group Fq2r Sp(2r, Fq ) is a
(1, 1)-design. Hence, Theorem 4.6 guarantees that it is also a (2, 0)-design. As shown
in Lemma 4.3, the tensor product representation (VrF )2 is irreducibly decomposed
to symmetric tensor product space Hs and alternative tensor product space Ha . In the
following, when q is an odd number, using Theorem 2.2, we investigate the projective
unitary representation (VrF )2 of the Clifford group Fq2r Sp(2r, Fq ) on H2r . When
we consider the representation only of Fq2r , the factor system of the representation
is the square of that of the irreducible projective unitary representation WrF . As
explained in the sentence before Lemma 8.4, when a projective unitary representation
has the same factor system, it is isometric to the projective unitary representation
WrF:2 . We denote the representation space by UFrq ,2 .
Next, in order to apply Theorem 2.2, we discuss H(2r, Fq ) instead of Fq2r . That is,
we focus on the unitary representation (WrF,H SrF )2 of H(2r, Fq ) Sp(2r, Fq ).
Here, we change the phase factor of SrF so that it is a unitary representation. Then, due
to the above discussion, the irreducible unitary representation component of H2r
with respect to the normal subgroup H(2r, Fq ) is isometric to WrF,H:2 . Hence,
Hs = UFrq ,2 C(q
+1)/2
(q 1)/2
Ha = UFrq ,2 C
(8.41)
.
(8.42)
Since the unitary representation WrF,H:2 can be constructed from the unitary representation WrF,H by replacing F by F2 , the associated representation of Sp(2r, Fq )
with respect to the unitary representation WrF,H:2 can be chosen as a unitary representation. This unitary representation is denoted by SrF:2 . Since the unitary representations of H(2r, Fq ) Sp(2r, Fq ) on Hs and Ha are irreducible, due to Theorem
r
2.2, there exist irreducible unitary representations of Sp(2r, Fq ) on C(q +1)/2 and
(q r 1)/2
given the right hand sides of (8.41) and (8.42). We denote these unitary
C
representations by SrF:s and SrF:a . So, (WrF,H SrF )2 can be described as
rF:2 SrF:s
(WrF,H SrF )2 = WrF,H:2 S
rF:2 SrF:a .
WrF,H:2 S
Next, we consider the case when X = Zd . The order of an element of Z2r

d is given
2r
Z
as a divisor of d. For a divisor m of d, let Nm be the set of elements of
d whose
2
order is m. Then, the number of elements of Nm is given as |Nm | = m m |m |Nm |.
Especially, when d = p k , we have |N pl | = p 2l p 2(l1) = p 2(l1) ( p 2 1).
Lemma 8.11 The representation space of the tensor product representation between
the projective unitary representation VrZ of the Clifford group Z2r
d Sp(2r, Zd ) on
285
the space Hr and its complex conjugate representation VrZ can be regarded as
the space of liner maps from Hr to Hr . Then, the irreducible subspace of this
representation is given as < {WrZ (s )}sNm >.
Proof Any irreducible subspace of the representation of the subgroup Z2r
d is given
as the 1-dimensional subspace spanned by WrZ (s ). For any two elements s and s
of order m, there exists an element g of Sp(2r, Zd ) such that g(s ) = s. That is,
WrZ (s ) = SrF (g)WrZ (s )SrF (g) . Conversely, when s has a different order from s,
there is no unitary matrix U such that WrZ (s ) = U WrZ (s )U . So, we have the desired
argument.

Exercise 8.10 Give a CONS of H1:even and H1:odd when d = 3.
8.4 MUB and SIC Vectors*

For a CONS B0 := {|0, . . . , |d 1}, we define another CONS B1 := {|e0 , . . . ,
|ed1 } by (2.51) via the discrete Fourier transform DFT. Then, the condition
1
|el | j| = , l, j Zd
d
(8.43)
holds. Since any base of the CONS B1 has a frequency of the same size based on the
other CONS B0 , it seems to have no bias on the other CONS B0 . Hence, two CONSs
B0 and B1 are called mutually unbiased to each other when they satisfy (8.43). So,
they are called mutually unbiased bases (MUB). For example, the CONS
1 2ilk/d
e
|k = DFT |l
|el :=
d kFq
d
is given by the discrete Fourier transform for the computational basis {|x}d1
x=0 on C .
Then, these CONSs are MUB. As will be explained later, when d + 1 bases mutually
unbiased to each other, we can construct a highly symmetric POVM, which enables
us to construct a useful protocol. MUB can also be applied to signal processing is
given in [1].
A MUB can be constructed as follows.
Theorem 8.3 ([110]) When the dimension d of the Hilbert space H is the power
d = p m of the prime number p, there exist d + 1 bases mutually unbiased to each
other.
When d = 2, three MUBs can be constructed in Riemann sphere as Fig. 8.3. This
construction of MUB is called the canonical MUB. An alternative construction was
given based on strictly self-orthogonal subgroup in the sense of Subsect. 8.1.4 under
the same condition [14, 70]. It is known that the number of MUB is at most d + 1 for
286
Fig. 8.3 MUBs in Riemann

sphere
= i
= e1
=1
=i
=0
any dimension d [14]. Also, in any prime power dimension, the canonical MUB is
uniquely determined by an extremal orbit of the (restricted) Clifford group except in
dimension 3 [119]. On the other hand, there are several MUBs that are not unitarily
equivalent to each other. Inequivalent MUBs are classified in [62]. A general review
for MUB is available in [28].
Proof When q = p m , we employ the discrete Heisenberg representation on the finite
field Fq so that we can construct d + 1 bases mutually unbiased to each other as
follows.
We can choose a pair (s, t) = (0, 0) of elements of Fq such that the set
{WrF (xs, xt)}xFq is composed of matrices commutative with each other. For p = 5,
see Fig. 8.1. Then, we can choose the CONS B(s,t) that diagonalizes these matrices
to the computational basis, and
simultaneously. Here, the CONS B(0,1) corresponds

1
Ftr xt |x. So, they are mutually
an element of the CONS B(1,0) is given as xFq q 1/2
unbiased with each other.
Generally, for two pairs (s, t), (s , t ) Fq2 \ (0, 0), we say that (s, t) is equivalent
to (s , t ), i.e., (s, t) (s , t ) when there exists a scalar x Fq such that (s, t) =
(xs , xt ). When two pairs (s, t), (s , t ) Fq2 are not equivalent to each other, there
exists an element g Sp(2, Fq ) such that g(s, t) = (0, 1), g(s , t ) = (1, 0).
Since the unitary matrix SrF (g) maps the bases B(s,t) and B(s ,t ) to the bases B(0,1)
and B(1,0) , the bases B(s,t) and B(s ,t ) are mutually unbiased to each other. Further,

because |(Fq2 \ (0, 0))/ | = q + 1, we obtain the desired argument.
MUB is a 2-design as follows.
Theorem 8.4 When d + 1 bases B1 , , Bd+1 are mutually unbiased to each other,
the set B1 Bd+1 is a (1, 1)-design. So, it is also a 2-design.
Further, it is known that the canonical MUB is the unique minimal 2-design
covariant with respect to the Clifford group except in dimension 3 [119].
287
Proof We denote the basis of B j by u 1, j , . . . , u d, j , and the vector | 1d I on

the composite system Cd Cd by . Then, the subspace B j =< u 1, j u 1, j
1 , . . . , u d1, j u d1, j 1 > of the composite system Cd Cd is a subd
d
space of the orthogonal complement to , and has the dimension d 1. Since
1
1
u i, j u i, j |u i , j u i , j = 0,
d
d
j = j ,

the subspaces B1 , . . . , Bd+1
are orthogonal to each other. Because the orthogonal

spans the
complement to has the dimension d 2 1, the subset B1 . . . Bd+1
orthogonal complement to
. Hence,
an
arbitrary
vector
on
the
composite
system
d
a
u
u
.
Cd Cd can be written as d+1
i, j
j=1
i=1 i, j i, j
Since
2
2

1
1

ai, j u i, j u i, j =
ai, j
|
ai, j |2 +
|ai, j |2

d j,i
d j

j,i
i
j,i
2

2

0

1

ai, j u i, j u i, j =
ai, j ,
A,B
d j,i

j,i

we have

j,i
ai, j u i, j

1
|u k,l u k,l u k,l u k,l |
u i, j
l,k d + 1

ai , j u i , j u i , j

j ,i
2

1
=
u k,l u k,l ||ai, j u i, j u i, j

d + 1 l,k j,i

2

1
1
1
=
ai, j
|
ai, j |2 +
|ai, j |2
d j,i
d(d
+
1)
d
+
1

j
i
j,i

d
1
|| +
I
=
ai, j u i, j u i, j
ai , j u i , j u i , j .
d + 1 j ,i
d +1
j,i

1
d
1
Thus, we obtain
l,k d+1 |u k,l u k,l u k,l u k,l | = d+1 || + d+1 I , which
implies that B1 Bd+1 is a (1, 1)-design. Therefore, Theorem 4.6 guarantees

that B1 Bd+1 is a (2, 0)-design.
288
Fig. 8.4 SIC vectors in

Riemann sphere
=0
1
1
Next, we focus on d 2 vectors {|s j }dj=0
satisfying the symmetry |sl |s j | = d+1
,
where j = l. Such vectors are called symmetric informationally complete (SIC)
vectors. When d = 2, SIC vectors
from
of tetrahedron

the vertexes

canbe constructed
on Riemann sphere by vectors |1, 23 |0 + 13 |1, 23 |0 + e2i/3 13 |1, 23 |0

+e4i/3 13 |1 (see Fig. 8.4).
SIC vectors satisfy the following properties.
2
1
Theorem 8.5 SIC vectors {|s j }dj=0
form a POVM
a (1, 1)-design, i.e., a (2, 0)-design.
2
1
d
d 2 1
j=0
|s j s j | = I , and form
The literatures [1, 90, 111, 112] are basic for SIC vectors. The existence of SIC
vectors has been analytically shown when the dimension d is one of integers from 2
to 16, or one of 19, 24, 28, 31, 35, 37, 43, and 48 [7, 97]. Also, when d 67, SIC
vectors has been approximately given by a numerical analysis [42, 90, 97]. Since,
in fact, SIC vectors are closely related to Lie algebra, Lie groups [5, 7, 115] and
Galois field [6] and has interesting symmetric properties [4, 113, 114, 120], they are
interesting mathematical objects. Due to the symmetry, SIC can be applied to state
estimation [95, 121, 115]. It was also applied to Bayesian statistics with quantum
setting [33].
Further, even when t is larger than 2, a construction of a (t, 0)-design has been
discussed based on SIC vectors [16, 95, 17, 42, 56, 97]. Also, SIC vectors can be
applied to equiangular lines, which are collections of vectors with only one angle
between them [25, 36, 41, 64, 71, 93, 103].
Proof Firstly, we show that the vectors s1 s1 , . . . , sd 2 sd 2 are linearly independent
on Cd Cd . For this purpose, we choose complex numbers a1 , . . . , ad 2 such
that i ai si si = 0, which implies that
0=
d|
289

i
0 = u 1 u 1 |
ai si si = a1 +

i

i=1
ai u i u i = a1 +
ai
1
ai .
d + 1 i=1
These two equations show that a1 = 0. Similarly, we have ai = 0. So, s1 s1 , . . . , sd 2

sd 2 are linearly independent. Since the composite
system Cd Cd has the dimen
2
d
d
sion d , any vector on C C is written as j ai u i u i .
Because
2

2

1
d

=

a
s
s
a
+
|ak |2

j
j
j
k

d +1 k
d +1 k

j
2

2

0

1

|

a j s j s j =
a j ,
A,B
d

j
j
we have

1

|sk sk sk sk |
ai si si
ajsj sj

d
j
i
k

2

d + 2
d
=
ak +
|ak |2

2
2

(d + 1) k
(d + 1) k

d
1

|| +
I
=
ai u i si
ajsj sj ,
d +1
d + 1 j
i

which implies that

1
1
d
|sk sk sk sk | =
|| +
I.
d
d +1
d +1
k
(8.44)
1
1
That is, {|s j }dj=0
is a (1, 1)-design. Further, Theorem 4.6 guarantees that {|s j }dj=0
1
is a (2, 0)-design. Taking the partial trace in (8.44), we obtain k d |sk sk | = I ,

which shows that { d1 |sk sk |} is a POVM on Cd .

Exercise 8.11 Show that vectors |1, 23 |0 + 13 |1, 23 |0 + e2i/3 13 |1,

2
|0 + e4i/3 31 |1 form SIC vectors.
3
2
Appendix A
Solutions of Exercises
Exercise 1.1
5
, PM (2) =
PM (1) = 16
5
,
16
and PM (3) = 38 .
Exercise 1.2
A = E 1 E 2 and B = F1 F2 , where E 1 =

11
1 1
1
1
F1 = 2
and F2 = 2
.
11
1 1
1
2

1 i
, E2 =
i 1
Exercise 1.3
The expectation can be calculated in two ways. (1) 1
Tr A = 0.
Exercise 1.4
The variance A can be calculated in two ways. (1) 12
Tr A2 = 1.
1
2
1
2
+ (1)

1 i
,
i 1
1
2
= 0. (2)
+ (1)2
1
2
Exercise 1.5
Since Tr B = 0, the value A B is Tr

i 0
= 0.
0 i

1
2
1
(AB
2
+ B A) = Tr
1
2
= 1. (2)
i 0
0 i

+
Exercise 1.6
The state |X A,B A,B X | is entangled if and only if the vector |X A,B is in a tensor
product form. This condition is equivalent to det X = 0.
Exercise 1.7
The orthogonality can be shown by using (1.16).

DOI 10.1007/978-3-319-44906-7
291
292
Appendix A: Solutions of Exercises
Exercise 1.8
n
is n
The
spectral decomposition of
i 1 ,i 2 ,...,i n :i 1 +i 2 ++i n =k E i 1 E i 2 E i n .
Exercise 1.9
Since lim T
lim
1
T
1
T
T
0
n
2k
k=0 3n
:=
Fk , where Fk
eit (h j h j ) dt = 0 for h j = h j , we have
eit H eit H dt = lim
1
T
eit (h j h j ) dt E j E j =
E j E j .
Exercise 1.10
The PVM {E i F j }i, j gives the simultaneous diagonalization of A I and I B,
where E i and F j are given in Exercise 1.2.
Exercise 1.11
For simplicity, we show the case with d = 1. Given two non-negative integers
, Z+ , we assume that 1. Then, we have

d
d 1
d
sup x x f (x) = sup x 1 f (x) + x +1 f (x)
dx
dx
dx
x
x

d 1
+1 d
sup x
f (x) + sup x
f (x) < .
d x 1
dx
x
x
When = 0,

d

sup x x f (x) = sup x +1 f (x) < .
dx
x
x
So, he function x f (x) belongs to S (Rd ).

For a given , Z+ , we assume that 1. Then, we have supx x ddx

+1
= supx x ddx +1 f (x) < . So, the function ddx f (x) belongs to S (Rd ).
d
dx

f (x)
Exercise 1.12
For simplicity, we show the case with d = 1. Given two non-negative integers
, Z+ , we assume that 1. Then, we have
k

d

d
d k
sup x g(x) f (x) = sup x
g(x)
f (x)
k
k
k dx
dx
dx
x
x
k=0

k=0
d k

dk
sup x k g(x) k f (x) < .
dx
dx
k x
So, the product f (x)g(x) belongs to S (Rd ).
293
Exercise 2.1
Let (a, b, c) be the cyclic permutation among a, b, c. Then, the permutation group
S3 is composed of (1, 2, 3), (3, 1, 2), (2, 3, 1), (1, 2), (2, 3), and (3, 1). The orders
of (1, 2, 3), (3, 1, 2), (2, 3, 1) are 3. The orders of (1, 2), (2, 3), (3, 1) are 2.
Exercise 2.2
This statement can be shown by the following relation. (1, 2)(3, 4)(1, 3)(2, 4) =
(1, 3)(2, 4)(1, 2)(3, 4) = (1, 4)(2, 3), (1, 3)(2, 4)(1, 4)(2, 3) = (1, 4)(2, 3)(1, 3)
(2, 4) = (1, 2)(3, 4), (1, 4)(2, 3)(1, 2)(3, 4) = (1, 2)(3, 4)(1, 4)(2, 3) = (1, 3)
(2, 4), ((1, 2)(3, 4))2 = ((1, 3)(2, 4))2 = ((1, 4)(2, 3))2 = e.
Exercise 2.3
The following map gives the isometric relation. e e, (1, 2)(3, 4) i, (1, 3)
(2, 4) j, (1, 4)(2, 3) k.
Exercise 2.4
It is enough to show that (a, b){e, (1, 2)(3, 4), (1, 3)(2, 4), (1, 4)(2, 3)}(a, b) =
{e, (1, 2)(3, 4), (1, 3)(2, 4), (1, 4)(2, 3)} because any permutation is given as the
products of transpositions. Consider the case of (a, b) = (1, 2) because other
cases can be shown in the same way. (1, 2)(1, 2)(3, 4)(1, 2) = (1, 2)(1, 2)(1, 2)
(3, 4) = (1, 2)(3, 4). (1, 2)(1, 3)(2, 4)(1, 2) = (2, 3)(1, 2)(1, 2)(1, 4) = (2, 3)
(1, 4) = (1, 4)(2, 3). (1, 2)(1, 4)(2, 3)(1, 2) = (2, 4)(1, 2)(1, 2)(1, 3) = (2, 4)
(1, 3) = (1, 3)(2, 4).
Exercise 2.5
Consider the subgroup S3 of S4 , which permutes the letters {1, 2, 3}. For any different
element g = e S3 , the set g(Z2 Z2 ) is different from Z2 Z2 . Hence, for any
different elements g1 = g2 S3 , the set g1 (Z2 Z2 ) is different from g2 (Z2 Z2 ).
So, each element of S3 gives different residue class. Since the order of S3 is 6 and the
order of S4 /Z2 Z2 is 24/4 = 6, all of elements of S4 /Z2 Z2 can be written by
S3 . Since they have the same products, the quotient group S4 /Z2 Z2 is isometric
to S3 .
Exercise 2.6
It is trivial that the subset {g G|T (g)0 = 0 } contains the identity element. When
g1 , g2 {g G|T (g)0 = 0 }, we have T (g1 g2 )0 = T (g1 )T (g2 )0 T (g1 )0 = 0 .
Hence, {g G|T (g)0 = 0 } is a subgroup.
Exercise 2.7
Firstly, we fix a point 0 in . For an element , we choose g G such that
g 0 = . Then, [g ] G/H is an isomorphism for homogeneous spaces.
294
Exercise 2.8
Let X be the set {g 1 h 1 gh|g, h G} of generators of the commutator subgroup [G, G]. For elements z G and x, y G, we have z(x yx 1 y 1 )z 1 =
(zx z 1 )(zyz 1 )(zx z 1 )1 (zyz 1 )1 X . That is, z X z 1 X . Since X generates
[G, G], we have z[G, G]z 1 [G, G]. Especially, considering the case when z is
an element of [G, G], we find that [G, G] is a normal subgroup.
Exercise 2.9
The desired statement follows from the calculation given in Exercise 2.4.
Exercise 2.10
Since
((h, k)(h , k ))(h , k ) = (hT (k)(h )(k, k ), kk )(h , k )
=(hT (k)(h )(k, k )T (kk )(h )(kk , k ), kk k ),
(h, k)((h , k )(h , k )) = (h, k)(h T (k )(h )(k , k ), k k )
=(hT (k)(h T (k )(h )(k , k ))(k, k k ), kk k ),
it is sufficient to show that
hT (k)(h )(k, k )T (kk )(h )(kk , k )
=hT (k)(h T (k )(h )(k , k ))(k, k k ).
(A.1)
This equation can be shown as follows.

T (k)(h T (k )(h )(k , k ))(k, k k )
=T (k)(h )T (k)(T (k )(h ))T (k)((k , k ))(k, k k )
(a)
=T (k)(h )(k, k )T (kk )(h )(k, k )1 (k, k )(kk , k )(k, k k )1

(k, k k )
=T (k)(h )(k, k )T (kk )(h )(kk , k ),
(A.2)
where (a) follows from (2.7) and (2.8).

Exercise 2.11
hk = (h, e)(e, k) = (h, k). khk 1 = (e, k)(h, e)(e, k 1 ) = (e, k)(h, k 1 ) =
(T (k)h, kk 1 ) = (T (k)h, e) = T (k)(h).
Exercise 2.12
The desired argument can be checked by the relations (h, k)(h , e)(h, k)1 =
(h, k)(h , e)(h 1 , k 1 ) = (h, k)(h h 1 , k 1 ) = (h, T (k)h h 1 , kk 1 ) = (h, T (k)h
h 1 , e) H .
295
Exercise 2.13
When T (k) is the identity map, the discussions in Sect. 2.2.2 give the central
extension.
Conversely, we assume that the given extension is a central extension. Since
any factor system can be replaced by a normalized factor system, we treat only
the case with a normalized factor system. We consider the case with (h , e) =
(e, k)(h , e)(e, k)1 = (e, k)(h , k 1 ) = (T (k)h (k, k ), kk 1 ) = (T (k)h
(k, k ), e) h = e. So, we obtain (k, k ) = e. Hence, we have T (k)h = h ,
which implies that T (k) is the identity map.
Exercise 2.14
The desired argument can be shown by focusing on the simultaneous diagonalization.
Exercise 2.15
When f is a self-conjugate representation, we have f (g) = f (g) = f (g). So, the
character takes real numbers.
Exercise 2.16
The representation is complexifiable if and only if it is commutative with the multiplication of the imaginary number i, which is equivalent to (2.44).
Exercise 2.17
It is enough to consider the case when j = 2 and i = 1 in (2.72). So, (2.72) implies
that
n + k
nk
n!
+
2
1
2
2
( n+k
+ 1)! nk
!
2
2

(n + 1)!
k+1 n+1
.
=
(k
+
1)
=
n + 1 n+k
+1
(n + 1)( n+k
+ 1)! nk
!
2
2
2
nk =
d Sn ,( n+k
2 , 2 )

cos sin
Any element g of SO(2, R) is written as
. So, any element con sin cos
nected to the connected components including I . Thus, SO(2, R) is connected.
Exercise 3.1
Exercise 3.2
Since gg = g g, there are d eigenvectors v1 , . . . , vd of g. Since g preserves the inner
product, the eigenvalue j corresponding to v j has the absolute value 1. Taking
the complex conjugate of the eigenvalue equation j v j = gv j , we have j v j =
gv j . This fact shows that the number of non-real eigenvectors is even, which is
denoted by 2r . We sort the eigenvectors vi such that initial 2r eigenvectors are
non-real, and v2 j is the complex conjugate of v2 j1 . Now, we define two vectors
w2 j1 := 12 (v2 j1 + v2 j ) and w2 j := i 2 (v2 j1 v2 j ) for j = 1, . . . , r . For
296
j 2r + 1, we define w j := v j . Since | j | = 1, j is written as cos + i sin . So,

we have gw2 j1 = cos w2 j1 + sin w2 j and gw2 j = sin w2 j1 + cos w2 j for
j = 1, . . . , r . Arranging the d vectors w1 , . . . , wd , we define orthogonal matrix g .
Then, g gg 1 satisfies the required condition.
Exercise 3.3
Due to Exercises 3.1 and 3.2, Any element g SO(d, R) is connected to the connected components including I . So, SO(d, R) is connected.
Exercise 3.4
We choose one-dimensional differentiable subsets gi (t) of G i . Then, g1 (t) and g2 (t)
are commutative with each other as elements of G 1 G 2 . This fact shows that
g(G 1 ) g(G 2 ) = g(G 1 G 2 ).
Exercise 3.5
The desired argument can be checked by expanding both sides of (3.20) and (3.21)
up to the order t 2 .
Exercise 3.6
It is sufficient to show that (3.41) satisfies Jacobi law, which can be shown as follows.
[(X 1 , X 2 ), [(Y1 , Y2 ), (Z 1 , Z 2 )]] + [(Y1 , Y2 ), [(Z 1 , Z 2 ), (X 1 , X 2 )]]
=[(X 1 , X 2 ), (T (Y2 )Z 1 T (Z 2 )Y1 + [Y1 , Z 1 ], [Y2 , Z 2 ])]
+ [(Y1 , Y2 ), (T (Z 2 )X 1 T (X 2 )Z 1 + [Z 1 , X 1 ], [Z 2 , X 2 ])]
=(T (X 2 )(T (Y2 )Z 1 T (Z 2 )Y1 + [Y1 , Z 1 ]) T ([Y2 , Z 2 ])X 1
+ [X 1 , (T (Y2 )Z 1 T (Z 2 )Y1 + [Y1 , Z 1 ]], [X 2 , [Y2 , Z 2 ]])
+ (T (Y2 )(T (Z 2 )X 1 T (X 2 )Z 1 + [Z 1 , X 1 ]) T ([Z 2 , X 2 ])Y1
+ [Y1 , (T (Z 2 )X 1 T (X 2 )Z 1 + [Z 1 , X 1 ]], [Y2 , [Z 2 , X 2 ]])
=(T (X 2 )(T (Y2 )Z 1 T (Z 2 )Y1 + [Y1 , Z 1 ]) T ([Y2 , Z 2 ])X 1
+ T (Y2 )(T (Z 2 )X 1 T (X 2 )Z 1 + [Z 1 , X 1 ]) T ([Z 2 , X 2 ])Y1
+ [X 1 , (T (Y2 )Z 1 T (Z 2 )Y1 + [Y1 , Z 1 ]]
+ [Y1 , (T (Z 2 )X 1 T (X 2 )Z 1 + [Z 1 , X 1 ]],
[X 2 , [Y2 , Z 2 ]] + [Y2 , [Z 2 , X 2 ]])
=([T (X 2 ), T (Y2 )]Z 1 + T (Z 2 )T (Y2 )X 1 T (Z 2 )T (X 2 )Y1
+ T (Z 2 )[X 1 , Y1 ] + [Z 1 , (T (X 2 )Y1 T (Y2 )X 1 ] + [Z 1 , [X 1 , Y1 ]],
[Z 2 , [X 2 , Y2 ]])
=[(Z 1 , Z 2 ), (T (X 2 )Y1 T (Y2 )X 1 + [X 1 , Y1 ], [X 2 , Y2 ])]
=[(Z 1 , Z 2 ), [(X 1 , X 2 ), (Y1 , Y2 )]].
297
Exercise 3.7
[f((X 1 , X 2 )), f((Y1 , Y2 ))]
=(f1 (X 1 ) + f2 (Y1 ))(f1 (X 2 ) + f2 (Y2 )) (f1 (X 2 ) + f2 (Y2 ))(f1 (X 1 ) + f2 (Y1 ))
=[f1 (X 1 ), f1 (Y1 )] + [f2 (X 2 ), f2 (Y2 )] + [f1 (X 1 ), f2 (Y2 )] + [f2 (X 2 ), f1 (Y1 )]
=f1 ([X 1 , Y1 ]) + f2 ([X 2 , Y2 ]) f1 (T (Y2 )X 1 ) + f1 (T (X 2 )Y1 )
=f([(X 1 , X 2 ), (Y2 , Y2 )]).
Exercise 3.8
Property (3.44) shows that the matrix ad(Y ) is represented as an alternative matrix.
Hence, exp(ad(Y )) is an orthogonal matrix. When we choose Y as g = exp(Y ), we
have (ad(g)X, ad(g)Z )g = (X, Z )g .
Exercise 3.9
Since
FU(1) FU1(1) [{an }](n ) =
1
2
an ein ein d =
nZ
an n,n = an ,
nZ
the map FU(1) FU1(1) is the identity map.

For any function (= 0) 2 (U(1)), FU(1) [] is not zero. So, the kernel of FU(1)
is {0}. Thus, the map FU1(1) FU(1) is also the identity map.
Exercise 3.10
FU1(1) [{an }]2 =
=

n,m
an am
1
2

0
1
2
2
0
an am ein eim d
n,m
ei(mn) d =
an am n,m = {an }2 .
n,m
Thus, FU1(1) is a unitary map. Since FU1(1) FU(1) is the identity map, FU(1) is a unitary
map.
298
Exercise 3.11

df
df
1
]() = i
(x)ei x d x
iFd [
d
dx j
(2) 2 Rd d x j

i
=
[(x)ei x ]
d
x j = d x 1 . . . d x j1 d x j+1 . . . d x d
(2) 2 Rd1

i
d f i x
+
(x)
e
dx
d
dx j
(2) 2 Rd1

1
(x) j ei x d x = j Fd [ f ]().
=
d
(2) 2 Rd1
Exercise 3.12
Fd1 [] ()
=
1
d
(2) 2
Rd

(x)e
(2) 2

i x
dx
Rd
(x) ei x d x = Fd [ ]().
Exercise 3.13
Fd []() =
=

1
d
(2) 2

1
d
(2) 2
(x)ei x d x
Rd
(x)ei x() d x
Rd
= Fd1 [].
Exercise 3.14
For simplicity, we show the case with d = 1. Using Exercise 3.11, we have

d
d 1
sup F[ f ]() = sup
d
d
2

1
d
= sup
(x)ei x d x
2 R d

1
= sup
(i x) (x)ei x d x
2 R

1
d

= sup
(i x) (x)ei x d x .
2 R d x

(x)ei x d x
R
299
Exercise 1.11 guarantees that ddx (i x) (x) belongs to S (R).

For a function f in S (Rd ), the values supx |x|2 | f (x)| and supx | f (x)| have finite
values C1 and C2 , we have |(x)| |xC21| and |(x)| C2 . We apply this fact to
d
(i x) (x). So, we have
dx

1
d

sup
(i x) (x)ei x d x
2 R d x

d
1
1
C1

dx +
(i x) (x) d x
C2 d x < .
2
2 R d x
2
|x|>1 |x |
|x|1
So, Fd [] belongs to S (Rd ).
Exercise 3.15

1 |Fd [2 ] =

1 (x)
=
Rd

=
Rd

=
Rd
1
d
(2) 2
1
d
(2) 2
1 (x)Fd [2 ](x)d x

1
i yx
(y)e
dy
dx
2
d
Rd (2) 2

i yx
e
1 (x)d x 2 (y)dy

Rd
Rd
Rd

ei yx 1 (x)d x 2 (y)dy = Fd1 [1 ]|2 .
Exercise 3.16
a Firstly, considering the complex integral in the path R + i y
R + i y

R R R + i y
. Then, we have
R
R+i y
R
2

(xi y
)2
e 2
dy =
e 2 x dy =
e 2 x dy.
2 R
2 R+i y
2 R
Taking the limit R , we have

2
2
(xi y
)2
e
dy =
e 2 x dy = 1.
2
2
Thus,

1
2
1
1

41
x 2 i x y
4
x2
F [1/
](y) =
ei x y
e
dy =
e 2
dy
2
2
2 2
1

2

1

14 2
(xi y
)2 x2
1 4 x2
=
e
dy =
e 2
=
(x).
2
2 2
300
b
1
2
41
1 1
) 4 (x )|F[1/
]
2
1
1

1 4
1 4
1
F [(x )]|1/
=
F[(x + )]|1/

=
2

1
1

1 4
1 4 1
1
=
F [(x + )]|1/
=
F [(x + )](y)1/
(y)dy
2

1
2
F[](y)ei x y e
y /2 dy.
=
2
(x y)
(y)dy = (
1 4
c Since ( 2
)
(y) is the Gaussian distribution with variance
, taking the limit
0, we have

1
F[](y)ei x y dy = F 1 [F[]].
(x) =
2
So, F 1 F is the identity map on S(R). Since

1
1
F 1 [F []] =
F [](y)ei x y dy =
F 1 [](z)ei x z dz = F [F 1 []],
2
2
F F 1 is the identity map on S(R).

d Since S(R) is a dense subset of L 2 (R), F 1 F and F F 1 are the identity map
on L 2 (R).
e Since Fd = F d and Fd1 = (F 1 )d , Fd1 Fd and Fd Fd1 are the identity
map on L 2 (Rd ).
Exercise 3.17
Substituting Fd [2 ] into 1 in Exercise 3.15, we have
Fd [2 ]|Fd [2 ] = Fd1 [Fd [2 ]]|2 . = 2 ]|2 .
So, Fd preserves the norm, i.e., Fd is a unitary map.
Exercise 3.18
Since S(Rd ) is a dense subset of L 2 (Rd ), Exercise 3.11 implies (3.70), which contains
(3.61) as a special case.
301
Exercise 3.19

1
eix ei px (x)d x
F[ei pQ ]() =
2

1
ei( p)x = F[]( p).
=
2
Exercise 3.20
When we replace by in (3.61), we have
F 1 [P]() = F 1 [](),
which implies that F 1 P = QF 1 , i.e., P = F 1 QF 1 . Substitute F 1 [] into
in (3.64), we have
1
ei pP []()ei pF QF []() = Fei pQ F 1 []()

=F[ei pQ F 1 []]() = F[F 1 []]( p) = ( p).
Exercise 3.21
Consider the inverse Fourier transform of the given function. Exercise 3.22 shows
that the product Fd1 [1 ] Fd1 [2 ] of the inverse Fourier transforms of 1 , 2 .
Exercise 3.14 shows that these inverse Fourier transforms Fd1 [1 ] and Fd1 [2 ]
belong to S (Rd ). So, Exercise 1.12 shows that their product Fd1 [1 ] Fd1 [2 ] also
belongs to S (Rd ). Since Exercise 3.14 shows that the Fourier transform Fd [Fd1 [1 ]
Fd1 [2 ]] belongs to S (Rd ), the given function belongs to S (Rd ).
Exercise 3.22
We can show this relation as follows.

1
Fd [1 ]( )Fd [2 ]( )d
d
(2) 2 Rd

1

=
Fd [1 ]( )2 (x)ei x( ) d xd
d
(2) Rd Rd

1

=
Fd [1 ]( )ei x d 2 (x)ei x d x
(2)d Rd Rd

1
=
Fd1 Fd [1 ](x)2 (x)ei x d x
d
(2) 2 Rd

1
1 (x)2 (x)ei x d x = Fd [1 2 ]().
=
d
(2) 2 Rd
302
Exercise 4.1
matrices
X satisfying
the condition
a The

set M

1,1 is the set of Hermitian
1
0
1
0
1 0
Tr X
= Tr X g
g = Tr X
= 0. Since Tr g X g
0 1
0 1
0 1

1 0
and the image of is limited to the connected component, we obtain the
0 1
desired statement.
b det g X g =
det g det X detg = det X .
z
x yi
c Since det
= z 2 x 2 y 2 , any element g SU(1, 1) gives
x + yi
z
a 3 3 matrix (g) SO(1, 2) based on the coordinate (x, y, z) R3 and the
above relation. The map is a homomorphism from SU(2) to SO(3, R), and is
not an isomorphism because the matrices I, I SU(1, 1) are mapped to the unit
matrix on R3 . In particular, we find that 1 (e) = {I, I }. So, we obtain the relation
SO(1, 2)0
= SU(1, 1)/Z2 .
Exercise 4.2
Using (4.13), we have
n
n
K+,s + [K+,s , K,s
])v0
K+,s K,s vn1 = (K,s
n
n1
=(K,s
K+,s + K,s
ns(2E0,s (n 1)I ))v0
n1
=(K,s
ns(2 (n 1)))v0 = ns(2 (n 1))vn1 .
Exercise 4.3
Since
n
]=
[E0,s , K,s
n1

n1 j
K,s [E0,s , K,s ]K,s
n
= nK,s
,
j=0
we obtain the first equation in (4.14). The second equation in (4.14) follows from
the relations
n ]=
[K+,s , K,s
n1

n1 j
K,s [K+,s , K,s ]K,s
j=0
n1

n1

j=0
n1 j
n1
(K,s
2sE0,s + K,s [2sE0,s , K,s
])
j=0
n1

n1 j
n1
(K,s
2sE0,s + K,s 2s(n 1 j)K,s
j=0
n1
=K,s
ns(2E0,s (n 1)I ).
n1 j
K,s 2sE0,s K,s
303
Exercise 4.4
Equation (4.25) is changed to

1
1
m( ) + 2 (1 + )2 m 2 = ( 1 )( + 1 + 1) +
2
4

+ 21 when = 1 + 21
=
21 when = 1 21 ,
2
which implies 3.28.

Exercise 4.5
Due to the invariance of the LHS of (4.32), it is sufficient to consider the case with
2
2 2
= (1 + (x 2 + y 2 )/2)2 , using
the state
|; . Since |; | : | = (1 + || )
r = x 2 + y 2 and t = r 2 /2, we have

d xd y
; | | : : |
|;
2(1 + ||2 )2
C

d xd y
= |; | : |2
2(1
+ ||2 )2
C
d xd y
= (1 + (x 2 + y 2 )/2)2
2(1 + (x 2 + y 2 )/2)2
C
2
dr dr
=
(1 + (x 2 + y 2 )/2)22
2
0 0
(1 + t)22 dt
=
0
1
1
[(1 + t)21 ]
.
0 =
2 1
2 + 1
Exercise 4.6
Due to the invariance of the LHS of (4.33), it is sufficient to consider the case with
2
2 2
the state
= (1 (x 2 + y 2 )/2)2 , using
|; . Since |; | : | = (1 || )
2
2
2
r = x + y and t = r /2, we have

d xd y
; | | : : |
|;
2(1
||2 )2
D

d xd y
= |; | : |2
2(1 ||2 )2
D
d xd y
= (1 (x 2 + y 2 )/2)2
2 + y 2 )/2)2
2(1
(x
D
2 2
dr dr
=
(1 (x 2 + y 2 )/2)22
2
0
0
304

=
(1 t)22 dt
1
1
[(1 t)21 ]10 =
.
2 1
2 + 1
Exercise 4.7
|; | : |2
d xd y
d xd y
.
= (1 + (x 2 + y 2 )/2)22
2
2
2(1 + || )
2
Exercise 4.8
|; | : |2
d xd y
d xd y
.
= (1 (x 2 + y 2 )/2)22
2(1 ||2 )2
2
Exercise 4.9
Since the basis of U[n,0,...,0] (U(r )) are identified with n-combination with repetition
1)!
.
of r things, which is (n+r
(r 1)!n!
Exercise 4.10
Substituting 1 into both of 1 and 2 in (4.48), we obtain the dimension
(r +1)r (r 1)
.
3
Exercise 4.11
Counting the dimension of both sides in (4.11), we have r 2 = 1 + dim U[1,0,...,0,1]
(U(r )).
Exercise 4.12
Taking the complex conjugate in (4.65), we obtain (4.68).
Exercise 5.1
We show only the case with t = 1, l = 2. Since
(x2 3 x3 2 )(x3 1 x1 3 )
=x2 1 + (x2 x3 3 1 x2 x1 3 3 x32 2 1 + x3 x1 2 3 ),
(x3 1 x1 3 )(x2 3 x3 2 )
=x1 2 + (x2 x3 3 1 x2 x1 3 3 x32 2 1 + x3 x1 2 3 ),
we have
[L 1 , L 2 ] = 2 [(x2 3 x3 2 ), (x3 1 x1 3 )] = 2 (x1 2 x2 1 ) = iL 3 .
305
Exercise 5.2
For k > l, we have
2
= Q 2k l2 + Q l2 k2 2Q k Q l k l Q k k Q l l .
Fk,l
Taking the sum for k = l, we have

2
Fk,l
=
k>l
Q 2k l2 + Q l2 k2 2Q k Q l k l Q k k Q l l
k>l
Q 2k l2
k=l
Q k Q l k l
k=l
Q 2k l2
k=l
Q k k
k=l
Q k Q l k l (d 1)
k=l
Q k k ,
which implies (5.40). Also, we have

2
Q k k
Q 2k k2 +
Q k k Q l l
k=l
Q 2k k2 +
Q k k +
Q k Q l k l ,
k=l
which implies (5.41).

Exercise 5.3
For k = 1, . . . , d 1, j > k, we have
1
2xk
r
wk
2
,
=
=
xk
r
d
2
x
k =1 k
which implies (5.50). Using the formula
d
dx
df
1
,
1 f (x)2 d x
cos1 f (x) =

d
d
d
2
2
2

x
x
k
k =k+1 x k

= d k =k k 2 dk =k+12 k = d
2
3/2
xk
k =k+1 x k ( k =k x k )
k =k x k

d
2
k =k+1 wk
=
,

r dk =k wk2
we have
306
which implies (5.51). Using the same formula, we have

d

x 2
xk x j
xk x j
k

= d k =k k 2 d
=
2 3/2
x j
d
d
2
2
x
(
x
)
k =k+1 k
k =k k
( k =k xk )
k =k+1 x k
wk w j

=
,
d
d
2
r ( k =k wk2 )
k =k+1 wk
Using (5.53), for j > l, we have
F j,l =w j

l1
k=1
k =k

j1
k=1

l1

wk2

wk2
+
w
d
l
2
l
r
k =l wk
wl
d
2 k
k =k+1 wk
d
k =l+1
wk wl

d

d
k =k
wk w j

wk2
d
2 k
k =k+1 wk

wk2
+
w
j
2
j
r
k = j wk
d
k = j+1
d
w j wl wk

d
2
2 k
k= j
k =k wk
k =k+1 wk

d
d
2
2
k = j+1 wk
k =l+1 wk
+ w j d
w
,

l
d
2
2
l
j
k =l wk
k = j wk

d
307
Using (5.53) and (5.54), we have

F j,d =w j

d2
k=1

k=1

d2

wk2
k =k
wd1
2
wd1 + wd2
j1
wd
wk wd

d
+ wd
d1
r

d
k =k
wk2 k
d
k =k+1
wk w j

wk2
d
2 k
k =k+1 wk

wk2
+
w
d
j
2
j
r
k = j wk
d
k = j+1
wk w j wd

d
2
2 k
k= j
k =k wk
k =k+1 wk

d
2
k = j+1 wk
wd1
+ wj 2
w
,
d
d
2
j
wd1 + wd2 d1
k = j wk

d

Exercise 5.4
We have

n+d 1
n+d 3
n
n2
(n + d 1) . . . (n + 1) (n + d 3) . . . (n 1)
=
(d 1)!
(d 1)!
[(n + d 1)(n + d 2) n(n 1)](n + d 3) . . . (n + 1)
=
(d 1)!
(d 1)(2n + d 2)(n + d 3) . . . (n + 1)
=
(d 1)!
(2n + d 2)((n + d 3) . . . (n + 1))
,
=
(d 2)!
308
Exercise 5.5
Since
(xk l xl k )(xk l xl k )
=xk l l,k xl l k,k xk k l,l + xl k k,l
+ (xk xk l l xk xl l k xl xk k l + xl xl k k ),
(xk l xl k )(xk l xl k )
=xk l l ,k xl l k ,k xk k l ,l + xl k k ,l
+ (xk xk l l xk xl l k xl xk k l + xl xl k k ),
we have
[L k,l , L k ,l ] = L k,l l,k L l,l k,k L k,k l,l + L l,k k,l ,
Exercise 5.6
Since
2 (x2 + i x1 )n
= n(n 1)(x2 + i x1 )n2
x12
2 (x2 + i x1 )n
= n(n 1)(x2 + i x1 )n2
x22
2 (x2 + i x1 )n
= 0,
x 2j
for j 3, we have Rd (x2 + i x1 )n = 0.
Exercise 5.7
n2
Rd
k=0

(1)k

n
x n2k x22k
2k 1
n2 1
k=0

n
(1)k (n 2k)(n 2k 1)
x n2k2 x22k
2k 1
n2
k=1
n2
k=1

n
(1)k 2k(2k 1)
x n2k x22k2
2k 1
x1n2k x22k2

(1)k1 (n 2k + 2)(n 2k + 1)
=0.

n
n
+ (1)k 2k(2k 1)
2k 2
2k
309
Exercise 5.8
Let x = sin 1 cos 2 . We have


sin2 1 sin 2
2(1 cos 1 )
0
0
0
2

=
d1
d2
sin2 1 sin 2
2(1 cos 1 )
0
0
2
sin 1
(x + i cos 1 )n
sin 1
=
d1
dx
2(1 cos 1 )
0
sin 1
2
sin 1
(x + i cos 1 )n
sin 1
=
d1
dx
2(1 cos 1 )
0
sin 1
2
1 (sin 1 + i cos 1 )n+1
sin 1
=
d1 (
n+1
2(1 cos 1 )
0
1 ( sin 1 + i cos 1 )n+1
sin 1 )
n+1
2(1 cos 1 )
2
1 (sin(n + 1)1 + i cos(n + 1)1 )
sin 1
=
d1 (
n+1
2(1 cos 1 )
0
1 ( sin(n + 1)1 + i cos(n + 1)1 )
sin 1 )
n+1
2(1 cos 1 )
2
sin 1
2i n
sin(n1 ).
=
d1
n+1 0
2(1 cos 1 )
2
d1
d2
d3
Exercise 5.9
c We have
2
sin sin 2
sin (1 + cos ) sin 2
d =
d
1 cos
1 cos2
0
0
2
2
(1 + cos ) sin 2
=
d =
2(1 + cos ) cos d
sin
0
0
2
2
=
2 cos2 d =
1 + cos 2d = 2.

d Since n 2, we have

=

=
2
sin sin(n + 1)
sin (1 + cos ) sin(n + 1)
d =
d
1 cos
1 cos2
0
(1 + cos ) sin(n + 1)
d
sin
(1 + cos )(sin n cos + cos n sin )
d
sin
310

=

=

=

=

=
2
(cos + cos2 ) sin n
d +
(1 + cos ) cos nd
sin
0
(cos + 1 sin2 ) sin n
d + 0
sin
2
(cos + 1) sin n
sin sin nd
d +
sin
0

2
( cos2 + 1) sin n
sin sin nd
d +
sin (1 cos )
0
2
sin2 sin n
sin sin n
d + 0 =
d.
sin (1 cos )
1 cos
0
Exercise 5.10

sin2 1 sin 2
2(1 cos 1 )
0
0
0
2

(1 + cos 1 ) sin 2
=
d1
d2
d3
2
0
0 0

sin 2
sin 2
= 2 2
= 2 2 .
= 2
d2
d3
d2
2
2
0
0
0
T ( f )(w) =
d1
d2
d3
Exercise 5.11
Using the commutation relation (5.2), we have
3

3

P j2 , Q k Pk = 2iPk2 ,
P j2 , Pk Q k = 2iPk2 .
j=1
j=1
So, we have
3

j=1
3
i
P j2 , A =
P 2.
k=1 k
Also,
1

x2 1
x2
, Q k Pk = i k3 ,
, Pk Q k = i k3 .
r
r
r
r
Since
3
xk2
k=1 r 3
r2
r3
= r1 , we have
3

j=1

Z e2
P j2 , A = i
.
4
0 r
311
Thus, we have the desired commutation relation

1 2
P + H.
2 j=1 j
3
i[H, A] =
Exercise 5.12
We show (5.100) only for k = 3.

4

w j 2
pk
j=1

2
=
j=1 (4 p0 p j
p3 )2 + (2 p0 )2 ( p 2 2 p32 + p02 )2 + (4 p02 p3 )2

( p 2 + p02 )4
4 p02 (4 p02 ( p 2 p32 ) + ( p 2 2 p32 + p02 )2 + 4 p02 p32 )

( p 2 + p02 )4
4 p02 ( p 2 + p02 )2
2 p0 2
=( 2
) .
2 4
2
( p + p0 )
p + p02
Next, we show (5.88) only for k = 2, l = 3.

4

w j w j
p2 p3
j=1
16 p02 p12 p2 p3 8 p02 ( p 2 2 p22 + p02 ) p2 p3 8 p02 ( p 2 2 p32 + p02 ) p2 p3 16 p04 p2 p3

=
( p 2 + p02 )4
=
=
8 p02 (2( p02 + p12 ) p2 p3 p2 p3 (2 p2 2 p22 2 p32 + 2 p02 )

( p 2 + p02 )4
8 p02 p2 p3 (2( p02 + p12 ) (2 p12 + 2 p02 )
( p 2 + p02 )4
Exercise 5.13
(5.107) can be shown as
= 0.
312

1 |4 F3 [2 ] =
1 (x) 4 F3 [2 ](x)d 3 x
R3

1
=
1 (x)
4 (y)F3 [2 ](x y)d 3 yd 3 x
3
(2) 2 R3
R3

1
1 (x) F3 [2 ](x y)4 (y)d 3 yd 3 x
=
3
(2) 2 R3 R3

1
=
1 (x) F31 [2 ](y x)4 (y)d 3 yd 3 x
3
(2) 2 R3 R3

1
(a)
1
3
=
(x)F
[
](y
x)
d
x
4 (y)d 3 y
1
2
3
3
R3 (2) 2
R3

(1 F31 [2 ] )(y) 4 (y)d 3 y = 1 F31 [2 ] |4 ,
=
R3
where (a) follows from Exercise 3.13.

Exercise 5.14
a We choose the spherical coordinate (r, 1 , 2 ) given in (5.7). We assume that p =
(0, 0, | p|). Then, choosing t := cos 1 , we have

ei(x p1 +yp2 +zp3 )

e|x|
d xd ydz
x 2 + y2 + z2
R3
2

ei| p|r cos 1 2
r sin 1
dr
d1
d2 er
=
r
0
0
0

ei| p|r cos 1 2
r sin 1
=2
dr
d1 er
r
0
0
1

r i| p|r t
r 4
=2
dr
dtr e e
=e
sin(| p|r ).
| p| 0
0
1
b
4
0 | p|
lim

4
er (ei| p|r ei| p|r )dr
0 i| p| 0
01
4
2 0 1
(
+
)=
= lim
.
0 i| p| i| p|
i| p| +
| p|2
er sin(| p|r )dr = lim
c The combination of a and b implies (5.109).

Exercise 5.15
Now, we consider the case when the spin is . Since the flavor system has the
permutation symmetry, it is enough to show the orthogonality the vectors in the spin
system when the vector in the flavor system is |udd. We have udd|n = | ,
udd|0 = | + | + | , and |(| + | + | ) = 0.
313
Exercise 5.16
Now, we consider the case when the spin is . It is enough to show the orthogonality
the vectors in the spin system when the vector in the flavor system is |uds. Since
uds|0 = | | , uds| 0 = | + | 2| ), and
uds| 0 = | + | + | , they are orthogonal to each other.

a
0 for two complex numbers a and b.
It is enough to show that (a , b )J
b
Exercise 5.17
(a, b)J

a
= Tr(aQ0 + aP0 ) (aQ0 + aP0 ) 0.
b
Exercise 6.1
Let is the angle between and . Then, cos = (, ). So, , = 2(,)
.
(,)
2
2
Thus, , , = 4 cos , which should be an integer. Hence, cos is limited
to 0, 41 , 21 , 34 , 1, i.e., , , is limited to 0, 1, 2, 3, 4. When , , = 4,
since , the pair (, , , ) is limited to (0, 0), (1, 1), (1, 2),
and (1, 3). When , , = 4, cos2 = 1, i.e., = . So, the pair
(, , , ) is (2, 2).
Exercise 6.2
a Since (, ) > 0, , and , are positive. So, Exercise 6.1 shows that one
of two is one, at least. When , = 1, = , = W () .
When = 1, . So, = ( ) .
b Substituting into in a, we obtain the desired statement.
Exercise 6.3
Consider liner transform that maps j to e j , where e j is the vector that has non-zero
component 1 only j-th entry. Since rj=1 V + (e j ) is not empty, rj=1 V + ( j ) is also
not empty.
Exercise 6.4
a Assume that there exists an element + () such that it is not written as
non-negative Z-linear combination of (). We choose := argmin (, ), where
the minimum is taken among elements + () satisfying the above condition.
Since is decomposable, there exists 1 , 2 + () such that = 1 + 2 .
We find (, ) = (1 , ) + (2 , ) and (1 , ), (2 , ) > 0. Due to the above
minimum choice, 1 and 2 are written as non-negative Z-linear combinations of
(). So, is also written as a non-negative Z-linear combination of (), which
is the contradiction of the assumption.
b Assume that (, ) > 0 and = . So, = . Exercise 6.2 guarantees
that . Thus, or belongs to + (). When + (),
314
= ()+, which contradicts the assumption (). When + (),

= ( ) + , which contradicts the assumption ().
c Assume that there exist real numbers r such that () r = 0. Define
subsets ()1 := {
0} and ()2 := { ()|r < 0}.
()|r >

r
=
So, we have
a
:=
()1
()2 r . Thus, b guarantees that

+

(a, a) = ()
1 , ()2 r (r )(, ) 0. So, a = 0. Thus, since (),
0 = (a, ) = ()1 r (, ) > 0, which contradicts the assumption.
d a and c guarantee the conditions (B2) and (B1), respectively.
e Due to Exercise 6.3, we can choose V + (). Due to condition (B2) and
a, we have + + ( ) and + ( ) = + ( ). So, + = + ( ).
Thus, ( ). Since the number of elements of both sets are dimension, we have
= ( ).
Exercise 6.5
Firstly, we assume that . However, it contradicts (B2). So, is not a root.
Next, we assume that (, ) > 0, which implies = . Since = , Exercise 6.2
guarantees that , which contradicts (B2). So, we have (, ) 0.
Exercise 6.6

a Assume that . = r with non-negative integer r . Choose
such that r = 0. So, + , which contradicts the assumption.
b Assume that there exists an element such that (, ) < 0, which implies
= . Since + , = . Exercise 6.2 guarantees that + , which
contradicts the assumption. So, we have (, ) 0 for . Since is not 0 and
spans the vector space V , there exists anelement such that (, ) > 0.
c Due to a and (B2), we is written as r with non-negative integer r .
Define two subsets 1 := { |r > 0} and 2 := { |r = 0}. Assume that
2 is not empty. Exercise 6.5 guarantees that (, ) = 1 r (, ) 0 for an
element 2 . Since is connected, there exist 1 and 2 such that
guarantees that (,
) < 0. Thus, (, ) < 0, which b.
(, ) = 0. Exercise 6.5
d Due to c, we have = r and = r with r , r > 0. (, ) =

, r r (, ), which is strictly positive due to b. When = , Exercise 6.2

guarantees that , which implies or , which contradicts the
assumption.
Exercise 6.7
X h satisfies
(X, [Fx , Fy ])g = ([X, Fx ], Fy )g = ((X )Fy , Fy )g = (X, Z )(Fy , Fy )g .
y
So, as Killing form is non-degenerate, we have [Fx , F ] = (F , F )g Z . Simiy

larly, we have [Fx , F ] = (Fx , Fx )g Z .
315
Exercise 6.8
The assumption yields that i , j = 0. Then, we have i , i + j
(i ,i )
i ,i )
= (i ,(
< 1. Since the above value is not a integer,
= (i +
j ,i + j )
i )+( j , j )
i + j does not belong to .
Exercise 6.9
The set + of positive roots is given as { j,l }1 j<lr , where j,l = j + + l1 .

1
j (r j) j .
Hence, = 21 1 j<lr ( j + + l1 ) = 21 rj=1
Exercise 6.10
It is sufficient to show that

l j+ l1
+,
t= j t
= , j,l j,l , for each root j,l + . Exerl j
cise 6.9 guarantees that , j = 21 (2 j (r j) ( j + 1)(r j 1) ( j 1)(r j +
1)) = 1 for j = 2, . . . , r 2, and , 1 = , r = 21 (2(r 1) 2(r 2)) = 1

for j = 1, r 1. So, , j,l = j l. Also, , j,l = l1
t= j t .
Exercise 7.1
The desired statement follows from (7.9) and (3.13).
Exercise 7.2
2
+i p
, the distribution of the outcome is || |2 dqdp
= e| | dqdp
=
When = q
2
2
2
e
( p p )2 +(qq )2
2
dqdp
.
2
Exercise 7.3
W|| () = |W()| = 0|W() W()W()|0
=0|W()W()W()|0 = e 2 () 0|W()W( + )|0
1
=e 2 () e 2 (++(+)) 0|W()|0 = e e
1
||2
2
Exercise 7.4
The distribution of the outcome is ||n|2 dqdp
= e||
2
dqdp
.
2
||2n dqdp
n! 2
= e
p 2 +q 2
2
(p
2 +q 2
2
n!
Exercise 7.5
2n
2
The distribution of the outcome n is ||n|2 = e| | |n!| .
Exercise 7.6
[Q I + I Q, P I I P] = [Q I, P I ] + [I Q, I P] = i I i I = 0.
)2n
316
Exercise 7.7
Assume that =
x+i
y
2
and =
x
+i y
2
= 1 + . Then, we have
Tr 2 U(b )|| N / U(b )

||2
d xd y
= | 1 + 1 + |e N
2 N
C

| 1|2 d x dy
N (||).
= ||e N
2 N
C
a1

It is enough to show that (a1 , . . . , a4 )J ... 0 for four complex numbers
Exercise 7.8
a4
a1 , . . . , a4 .
a1
4
4

(a1 , . . . , a4 )J ... = Tr(
ai Ai ) (
ai Ai ) 0.
a4
i=1
i=1
Exercise 7.9
The relation (7.66) follows as

p 2 +q 2
||2
q
q
e 2 +i(apqb) = i
e 4 +i(apqb)
p
2
p
2

p 2 +q 2
q + i p p2 +q 2 +i(apqb)
1
e 4
= a+
= ( + )e 4 +i(apqb)
2
2
a
+
a
1
= FW
[ ||]().
2
i
Similarly, the relation (7.67) follows as

p ||2 +i(apqb)
p p2 +q 2 +i(apqb)
2
i
e
e 4
= i
q
2
q
2

p 2 +q 2
q + i p p2 +q 2 +i(apqb)
i
e 4
= bi
= ( + )e 4 +i(apqb)
2
2
a
1
= FW
[i ||]().
2
317

cos sin
0
with
=
0
sin cos

cos sin
suitable two real numbers and . So, we have

sin cos

0
cos sin
=
.
sin cos
0
cos 0
0 sin
0 cos sin
0
Since the matrix g =

0 sin cos 0 satisfies
sin
0
0 cos
Exercise 7.10
Choose such that
cos sin
sin cos
0 2i 0
0
0 0 0 i
2
gT
i 0 0 0 g = i
2
2
0 2i 0 0
0
0 2i 0
0 0 2i
,
0 0 0
2i 0 0
the matrix g belongs to Sp(4, R). Thus, we have g T J,, g = J , ,0 .

Exercise 7.11
Using (7.65), we have

dadb
|
Q X () = |
W X ()W()
2
C

||2 dadb
dadb
=
.
= W X ()|W()|
W X ()e+ e 2
2
2
C
C
Exercise 8.1
001
WZ (0, 0) = I, WZ (1, 0) = 1 0 0 ,
010
010
1 0
0
WZ (2, 0) = 0 0 1 , WZ (0, 1) = 0 e2i/3 0 ,
100
0 0 e4i/3
2i/3
0 0 e4i/3
0e
0
0 , WZ (2, 1) = e2i/3 0 0 e4i/3 ,
WZ (1, 1) = ei/3 1 0
2i/3
0
1 0
0
0e
2i/3
1 0
0
0 0 e
0 ,
WZ (0, 2) = 0 e4i/3 0 , WZ (1, 2) = e2i/3 1 0
2i/3
4i/3
0
0 0 e
0e
4i/3
0e
0
WZ (2, 2) = e4i/3 0 0 e2i/3 .
1 0
0
318
Exercise 8.2
We have 1d WZ (s , t )| 1d WZ (s, t) =
1
d
Tr Zs tst WZ (s s , t t ) = s,s t,t .
So, {| 1d WZ (s, t)}(s,t)Z2d forms a CONS on Cd Cd .
Exercise 8.3
It is sufficient to show that (x, y) = (F(x), F(y))F4 for x, y F22 . This relation can
be shown as follows.
(1, 1)F4 = 0, (1, )F4 = 1, (, )F4 = 1, (, + 1)F4 = 0,
( + 1, + 1)F4 = 1, ( + 1, 1)F4 = 1.
Exercise 8.4
{(0, 0, 0, 0), (1, 1, 0, 0), (0, 0, 1, 1), (1, 1, 1, 1)}. It is easy to show that this subgroup
is a self-orthogonal subgroup. The dimension of this subgroup is 2. So, its orthogonal
group also has dimension 2. Thus, it is a strictly self-orthogonal group.
Exercise 8.5
Consider a matrix

ab
cd

ab
. Since
cd

x
ab
x
= (ad bc)(x y x y ),
,
y
y
cd
(A.3)
it belongs to Sp(2, Fq ) if and only if ad bc = 1.

Exercise 8.6

10
,
01

01
,
20

02
,
10

21
,
11

10
,
11

11
,
20

12
,
10

11
,
12

10
11
12
,
,
,
21
01
01

21
01
01
,
,
,
20
21
22

22
02
02
,
,
,
10
11
12

21
.
12
Exercise 8.7
Firstly, notice that the elements of GL(2, F2 ) corresponding to F4 via the F2 morphism F are limited to the following matrices.

1

11
01
10
.
, + 1
,
10
11
01
(A.4)

So, the matrix g0 :=
10
11
319

GL(2, F2 ) is not contained in F4 in this sense.
Exercise 8.8
g0 0
0 g0
but does not belong to Sp(2, F4 ) via the F2 -morphism F.
Use the element g0 given in Exercise 8.7. The matrix

belongs to Sp(4, F2 )
Exercise 8.9
The following is an example of a generating matrix of a strictly self-orthogonal
subgroup given in Exercise 8.4.
1
1
A(g) =
0
0
0
0
.
1
1
Exercise 8.10
The space H1:even is spanned by {|0, |1 + |2}. The space H1:odd is spanned by
{|1 |2}.
Exercise 8.11

2 1
1
2
1
2
1
2i/3 1
0| +
1|
|0 + e
|1 = + e2i/3 = + i.
3
3
3
3
3 3
2 2 3
So,

2
1
1
2

2i/3 1
0| +
1|
|0 + e
|1 =
.

3
3
3
3
3
Similarly, we have

2
1
2

4i/3 1
0| +
1|
|0 + e
|1

3
3
3
3

2
1
2

2i/3 1
4i/3 1
=
0| + e
1|
|0 + e
|1 =
.
3
3
3
3
3
320
Also, we have

2

2
1

2i/3 1
0| +
1| |1 =
0| + e
1| |1

3
3
3
3

2
1
1

4i/3
=
0| + e
1| |1 =
.
3
3
3
References
1. W. Alltop, Complex sequences with low periodic correlations (corresp.). IEEE Trans. Inf.
Theory 26(3), 350354 (1980)
2. A. Ambainis, J. Emerson, Quantum t-designs: t-wise independence in the quantum world,
in Twenty-Second Annual IEEE Conference on Computational Complexity (CCCf07) (2007),
pp. 129140
3. D.M. Appleby, SIC-POVMs and the extended Clifford Group. J. Math. Phys. 46, 052107
(2005)
4. D.M. Appleby, Symmetric informationally complete-positive operator valued measures and
the extended Clifford group. J. Math. Phys. 46, 052107 (2005)
5. D.M. Appleby, S.T. Flammia, C.A. Fuchs, The Lie algebraic significance of symmetric informationally complete measurements. J. Math. Phys. 52, 022202 (2011)
6. D.M. Appleby, H. Yadsan-Appleby, G. Zauner, Galois automorphisms of a symmetric measurement. Quantum Inf. Comput. 13(78) 06720720 (2013). http://arxiv.org/abs/1209.1813
7. D.M. Appleby, C.A. Fuchs, H. Zhu, Group theoretic, Lie algebraic and Jordan algebraic
formulations of the SIC existence problem. Quantum Inf. Comput. 15(12), 6194 (2015)
8. E. Artin, Geometric Algebra (Interscience Publishers, New York, 1957)
9. K. Asano, H. Nagao, Group Theory (Iwanami shoten, Tokyo, 1965). (In Japanese)
10. J.J. Aubert et al., Experimental Observation of a Heavy Particle. J. Phys. Rev. Lett. 33(23),
1404 (1974)
11. J.-E. Augustin et al., Discovery of a Narrow Resonance in e+ e Annihilation. Phys. Rev.
Lett. 33(23), 1406 (1974)
12. J. Baik, P. Deift, K. Johansson, On the distribution of the length of the longest increasing
subsequence of random permutations. J. Amer. Math. Soc. 12(4), 11191178 (1999); J. Baik,
P. Deift, K. Johansson, On the distribution of the length of the second row of a Young diagram
under Plancherel measure. Geom. Funct. Anal. 10(4), 702 (2000)
13. M. Bander, C. Itzykson, Group Theory and the Hydrogen Atom (I). Rev. Mod. Phys. 38, 330
(1966)
14. S. Bandyopadhyay, P.O. Boykin, V. Roychowdhury, F. Vatan, A new proof for the existence
of mutually unbiased bases. Algorithmica 34(4), 512528 (2002)
15. Eiichi Bannai, Etsuko Bannai, A survey on spherical designs and algebraic combinatorics on
spheres. Eur. J. Combinator. 30(6), 13921425 (2009)
16. E. Bannai, S.G. Hoggar, On tight t-designs in compact symmetric spaces of rank one. Proc.
Japan Acad. 61A, 78 (1985)
17. E. Bannai, S.G. Hoggar, Tight t-designs and squarefree integers. Eur. J. Comb. 10, 113 (1989)
DOI 10.1007/978-3-319-44906-7
321
322
References
18. B.J. Bjorken, S.L. Glashow, Elementary particles and SU(4). Physics Letters 11(3), 255257
(1964)
19. K. Chandrasekharan, Arithmetical Functions (Springer, Heidelberg, 1970)
20. H.-F. Chau, Unconditionally secure key distribution in higher dimensions by depolarization.
IEEE Trans. Inf. Theory 51(4), 14511468 (2005)
21. R. Cleve, D. Leung, L. Liu, C. Wang, Near-linear constructions of exact unitary 2-designs,
arXiv:1501.04592 (2015)
22. M. Combescure, D. Robert, Coherent States, and Applications in Mathematical Physics,
Theoretical and Mathematical Physics (Springer, Heidelberg, 2012)
23. C. Dankert, Efficient Simulation of Random Quantum States and Operators, Master thesis,
University of Waterloo (2005). http://arxiv.org/abs/quant-ph/0512217
24. C. Dankert, R. Cleve, J. Emerson, E. Livine, Exact and approximate unitary 2-designs and
their application to fidelity estimation. Phys. Rev. A 80, 012304 (2009)
25. P. Delsarte, J.M. Goethals, J.J. Seidel, Bounds for systems of lines, and Jacobi poynomials.
Philips Res. Rep. 30, 91 (1975)
26. P. Delsarte, J.M. Goethals, J.J. Seidel, Spherical codes and designs. Geom. Dedicata 6, 363
388 (1977)
27. D.P. DiVincenzo, D.W. Leung, B.M. Terhal, Quantum data hiding. IEEE Trans. Inf. Theory
48(3), 580598 (2002)
28. T. Durt, B.-G. Englert, I. Bengtsson, K. Zyczkowski,

On mutually unbiased bases. Int. J.
Quant. Inf. 8, 535 (2010)
29. J. Eisert, M.M. Wolf, Gaussian quantum channels, in Quantum Information with Continous
Variables of Atoms and Light N, ed. by J. Cerf, G. Leuchs, E.S. Polzik (Imperial College
Press, London, 2007), pp. 2342
30. V. Fock, Zur Theorie des Wasserstoffatoms. Z. Physik 98, 145 (1935)
31. G.B. Folland, Harmonic Analysis in Phase Space, Annals of Mathematics Studies 122 (Princeton University Press, Princeton, 1989)
32. G.B. Folland, A Course in Abstract Harmonic Analysis, Studies in Advanced Mathematics
(CRC Press, Boca Raton, 1995)
33. C.A. Fuchs, R. Schack, Quantum-Bayesian coherence: the no-nonsense version. Rev. Mod.
Phys. 85, 16931715 (2013)
34. H. Fhr, Abstract harmonic analysis of continuous wavelet transforms. Lecture Notes in
Mathematics, vol. 1863 (Springer, Berlin, 2005)
35. S.L. Glashow, J. Iliopoulos, L. Maiani, Weak Interactions with Lepton-Hadron Symmetry.
Physical Review D 2(7), 12851292 (1970)
36. C. Godsil, A. Roy, Equiangular lines, mutually unbiased bases, and spin models. Eur. J.
Combinator. 30, 246262 (2009)
37. R. Goodman, N.R. Wallach, Representations and Invariants of the Classical Groups, Encyclopedia of Mathematics and Its Applications, vol. 68 (Cambridge University Press, Cambridge,
1999)
38. M. Grassl, M. Rotteler, T. Beth, Efficient quantum circuits for non-qubit quantum errorcorrecting codes. International Journal of Foundations of Computer Science (IJFCS) 14(5),
757775 (2003)
39. D. Gross, K. Audenaert, J. Eisert, Evenly distributed unitaries: On the structure of unitary
designs. J. Math. Phys. 48(5), 052104 (2007)
40. D. Gross, F. Krahmer, R. Kueng, A partial derandomization of phaselift using spherical
designs. J. Fourier Anal. Appl. 21(2), 229266 (2015)
41. J. Haantjes, Equilateral point-sets in elliptic two- and three dimensional spaces. Nieuw Arch.
Wisk. 22, 355362 (1948)
42. R.H. Hardin, N.J.A. Sloane, McLarens improved snub cube and other new spherical designs
in three dimensions Discrete. Comput. Geom. 15, 429 (1996)
43. G.H. Hardy, S. Ramanujan, Asymptotic formulae in combinatory analysis. Proc. Lond. Math.
Soc. 17, 75115 (1918)
References
323
44. Harish-Chandra, Representation of Semisimple Lie group VI. Amer. Jour. Math., 78(3) 564
628 (1956)
45. A.W. Harrow, R.A. Low, Random quantum circuits are approximate 2-designs. Commun.
Math. Phys. 291(1), 257302 (2009)
46. M. Hayashi, Quantum information: an introduction (Springer, Heidelberg, 2006) (Originally
published in Japanese in 2004)
47. M. Hayashi, Discrete realization of group symmetric LOCC-detection of maximally entangled
state (2008). arXiv:0810.3381
48. M. Hayashi, Exponential decreasing rate of leaked information in universal random privacy
amplification. IEEE Trans. Inf. Theory 57, 39894001 (2011)
49. M. Hayashi, Fourier analytic approach to quantum estimation of group action. Commun.
Math. Phys. 347(1), 382 (2009)
50. M. Hayashi, A Group Theoretic Approach to Quantum Information (Springer, In preparation)
(Originally published in Japanese in, 2014)
51. M. Hayashi, F. Sakaguchi, Subnormal operators regarded as generalized observables and
compound-system-type normal extension related to su(1,1). J. Phys. A:Math. Gen. 33, 7793
7820 (2000)
52. A. Hayashi, T. Hashimoto, M. Horibe, Reexamination of optimal quantum state estimation
of pure states. Phys. Rev. A 72, 032325 (2005)
53. M. Hayashi, S. Ishizaka, A. Kawachi, G. Kimura, T. Ogawa, Introduction to Quantum Information Science, Graduate Texts in Physics. (Springer, Heidelberg, 2015) (Originally published
in Japanese in 2012)
54. M. Hillery, R.F. OfConnell, M.O. Scully, E.P. Wigner, Distribution functions in physics:
fundamentals. Phys. Rep. 106, 121167 (1984)
55. S.G. Hoggar, t-designs in projective spaces. Eur. J. Comb. 3, 233254 (1982)
56. S.G. Hoggar, Parameters of t-designs in F P d1 . Eur. J. Comb. 5, 29 (1984)
57. A.S. Holevo, Probabilistic and Statistical Aspects of Quantum Theory (North-Holland, Amsterdam, 1982) (Originally published in Russian 1980, 2nd Edition, Springer 2012.)
58. R. Howe, E.C. Tan, Non-Abelian Harmonic Analysis: Applications of SL (2, R) (Springer,
Heidelberg, 1992)
59. J.E. Humphreys, Introduction to Lie Algebras and Representation Theory, Graduate Texts in
Mathematics 9 (Springer, Heidelberg, 1972)
60. K. Husimi, Some formal properties of the density matrix. Proc. Phys. Math. Soc. Jpn. 22,
264314 (1940)
61. W.M. Kantor, MUBs inequivalence and affine planes. J. Math. Phys. 53(3), 032204 (2012)
62. E.H. Kennard, Zur Quantenmechanik einfacher Bewegungstypen. Zeitschrift fur Physik (in
German) 44(45), 326 (1927)
63. M. Khatirinejad, Regular Structures of Lines in Complex Spaces, Ph.D. thesis, Simon Fraser
University (2008)
64. A.A. Kirillov, Elements of the theory of representations (Springer, New York, 1976)
(Grundlehren der Mathematischen Wissenschaften 220, Berlin)
65. A.W. Knapp, Lie Groups Beyond an Introduction, vol. 140, 2nd edn., Progress in mathematics
(Birkhauser, 2002)
66. A.W. Knapp, Representation Theory of Semisimple Groups: An Overview Based on Examples,
vol. 36 (Princeton, Princeton Mathematics Series (Princeton University Press, 1986)
67. T. Komorowski, A criterion for essential self-adjointness of a symmetric operator defined by
some infinite hermitian matrix with unbounded entries. Integral Equ. Oper. Theory (2015).
doi:10.1007/s00020-015-2237-2
68. R. Kueng, D. Gross, Qubit stabilizer states are complex projective 3-designs (2015).
arXiv:1510.02767
Brukner, A. Zeilinger, Mutually unbiased binary observable sets on N qubits.
69. J. Lawrence, C.
Phys. Rev. A 65, 032320 (2002)
70. P.W.H. Lemmens, J.J. Seidel, Equiangular lines. J. Algebra 24, 494 (1973)
324
References
71. V. Levenshtein, Designs as maximum codes in polynomial metric spaces. Acta Appl. Math.
29, 1 (1992)
72. V. Levenshtein, On designs in compact metric spaces and a universal bound on their size.
Discrete Math. 192(13), 251271 (1998)
73. W. Lisiecki, Coherent state representations. A survey. Rep. Math. Phys. 35, 327358 (1995)
74. I.G. MacDonald, Symmetric Functions and Hall Polynomials (Clarendon Press, New York,
1979)
75. W. Matthews, S. Wehner, A. Winter, Distinguishability of quantum states under restricted
families of measurements with an application to quantum data hiding. Commun. Math. Phys.
291(3), 813843 (2009)
76. A. Messiah, Quantum Mechanics, vol. I (North Holland Publishing Company, Amsterdam,
1967). (Translation from the French by G.M. Temmer)
77. A. Messiah, Quantum Mechanics (Volume II) (North Holland Publishing, New York, 1981)
78. K.H. Neeb, Holomorphy and Convexity in Lie Theory, De Gruyter Expositions in Mathematics
28, Walter de Gruyter (1999)
79. M. Neuhauser, An explicit construction of the metaplectic representation over a finite field.
J. Lie Theory 12, 15 (2002)
80. M. Neuhauser, Explicit construction of the metaplectic representation for finite cyclic groups,
preprint: http://www.matha.rwth-aachen.de/~markus/
81. A. Perelomov, Generalized Coherent States and Their Applications, Texts and Monographs
in Physics (Springer, Heidelberg, 1986)
82. A.M. Perelomov, V.S. Popov, The Lorentz group as a dynamic symmetry group of the hydrogen atom. J. Exp. Theor. Phys. 23(1), 118 (1966). (Russian original - ZhETF, Vol. 50, No. 1,
p. 179, 1966)
83. D. Petz, An invitation to the algebra of canonical commutation relations. Leuven Notes Math.
Theor. Phys. 2, (1990)
84. A.V.M. Prasad, Padmavathamma, The Hardy-Ramanujan-Rademacher expansion for
Cm,m (n) using ford circles. Indian J. Pure Appl. Math. 22(11), 913925 (1991)
85. H. Rademacher, On the partition function p(n). Proc. Lond. Math. Soc. 43, 241254 (1938)
86. A.V. Razumov, M.V. Saveliev, Lie Algebras, Geometry, and Toda-type Systems (Cambridge
Lecture Notes in Physics, Cambridge, 1997)
87. M. Reed, B. Simon, Methods of Modern Mathematical Physics I: Functional Analysis (Academic Press, New York, 1980)
88. J.M. Renes, Frames, Designs, and Spherical Codes in Quantum Information Theory, Ph.D.
thesis, University of New Mexico (2004)
89. J.M. Renes, R. Blume-Kohout, A.J. Scott, C.M. Caves, Symmetric informationally complete
quantum measurements. J. Math. Phys. 45, 2171 (2004)
90. H.P. Robertson, The uncertainty principle. Phys. Rev. 34, 163164 (1929)
91. R.T. Rockafellar, Convex analysis (Princeton University Press, Princeton, 1997)
92. A. Roy, Complex lines with restricted angles, Ph.D. thesis, University of Waterloo (2006).
http://arxiv.org/abs/1306.0978
93. A. Roy, A.J. Scott, Unitary designs and codes. Des. Codes Cryptogr. 53(1), 1331 (2009)
94. A.J. Scott, Tight informationally complete quantum measurements. J. Phys. A: Math. Gen.
39, 13507 (2006)
95. A.J. Scott, Optimizing quantum process tomography with unitary 2-designs. J. Phys. A: Math.
Theor. 41(5), 055308 (2008)
96. A.J. Scott, M. Grassl, SIC-POVMs: a new computer study. J. Math. Phys. 51, 042203 (2010)
97. J.-P. Serre, Linear Representation of Finite Groups, 5th edn. (Springer, Heidelberg, 1996)
98. M. Suzuki, Group Theory I (Springer, Heidelberg, 1982) (Grundlehren der mathematischen
Wissenschaften)
99. G. Teschl, Mathematical Methods in Quantum Mechanics With Applications to Schrdinger
Operators Graduate Studies in Mathematics, vol. 99 (American Mathematical Society, Providence, 2014)
100. L.H. Thomas, The Motion of the Spinning Electron. Nature 117, 514 (1926)
References
325
101. C.A. Tracy, H. Widom, Comm. Math. Phys. 159, 151 (1994)
102. J.H. van Lint, J.J. Seidel, Equilateral point sets in elliptic geometry. Proc. Kon. Nederl. Akad.
Wet. Ser. A: Indag. Math. 69, 335348 (1966)
103. V.S. Varadarajan, Lie groups, Lie Algebras and Their Representations, Graduate Texts in
Mathematics 102 (Springer, Heidelberg, 1984)
104. D.F. Walls, G.J. Milburn, Quantum optics (Springer, Heidelberg, 2008)
105. Z. Webb, The Clifford group forms a unitary 3-design (2015). http://arxiv.org/abs/1510.02769
106. H. Weyl, Gruppentheorie und Quantenmechanik, Leipzig: Verlag von S. Hirzel in Leipzig
(1928) (English translation, The Theory of Groups and Quantum Mechanics, of the second
(1931) ed. was reprinted by Dover, 1950)
107. E.P. Wigner, On the quantum correction for thermodynamic equilibrium. Phys. Rev. 40, 749
759 (1932)
108. G. Wolf, Mathieu Functions and Hillfs Equation (2013). http://dlmf.nist.gov/28
109. W.K. Wootters, B.D. Fields, Optimal state determination by mutually unbiased measurements,
Ann. Phys. (N.Y.) 191, 363381 (1989)
110. G. Zauner, Quantendesigns - Grundzuge einer nicht kommutativen Designtheorie, Ph.D. thesis, University of Vienna (1999)
111. G. Zauner, Quantum designs: Foundations of a noncommutative design theory. Int. J. Quant.
Inf. 9, 445507 (2011)
112. H. Zhu, SIC POVMs and Clifford groups in prime dimensions. J. Phys. A: Math. Theor. 43,
305305 (2010)
113. H. Zhu, Quantum State Estimation and Symmetric Informationally Complete POMs. Ph.D.
thesis, National University of Singapore (2012). http://scholarbank.nus.edu.sg/bitstream/
handle/10635/35247/ZhuHJthesis.pdf
114. H. Zhu, Tomographic and Lie algebraic significance of generalized symmetric informationally
complete measurements. Phys. Rev. A 90, 032309 (2014)
115. H. Zhu, Quantum state estimation with informationally overcomplete measurements. Phys.
Rev. A 90, 012115 (2014)
116. H. Zhu, Multiqubit Clifford groups are unitary 3-designs (2015). http://arxiv.org/abs/1510.
02619
117. H. Zhu, Permutation symmetry determines the discrete Wigner function (2015). http://arxiv.
org/abs/1504.03773
118. H. Zhu, Mutually unbiased bases as minimal Clifford covariant 2-designs. Phys. Rev. A 91,
060301 (2015)
119. H. Zhu, Super-symmetric informationally complete measurements. Ann. Phys. 362, 311326
(2015)
120. H. Zhu, B.-G. Englert, Quantum state tomography with fully symmetric measurements and
product measurements. Phys. Rev. A 84, 022327 (2011)
Curriculum Vitae
Masahito Hayashi
Graduate School of Mathematics
Nagoya University
Furocho, Chikusaku
Nagoya, 464-8602, Japan, and
Centre for Quantum Technologies
National University of Singapore
3 Science Drive 2, Singapore 117542.
Phone: + 81(52)789-2549
Fax: + 81(52)789-5397
Email: masahito@math.nagoya-u.ac.jp
Homepage:
http://www.math.nagoya-u.ac.jp/
masahito/index_e.html
Personal
Born on December 28, 1971.
Born in Tokushima, Japan.
Japanese citizen.
Education
Ph.D. Depart of Mathematics, Graduate School of Science, Kyoto University,
1999.
M.A. Depart of Mathematics, Graduate School of Science, Kyoto University, 1996.
B.S. Physics & Mathematics, Faculty of Science, Kyoto University, 1994
Employment (Full time)
Research Fellow, the Japan Society for the Promotion of Science, 19982000
Researcher, Laboratory for Mathematical Neuroscience (Amaris Laboratory),
Brain Science Institute, RIKEN, 20002003
Research Manager, ERATO, Quantum Computation and Information Project,
Japan Science and Technology Agency (JST), 20032006
Group Leader, Quantum Information Theory Group, ERATO-SORST Quantum
Computation and Information Project, Japan Science and Technology Agency
(JST), 20062007
Associate Professor, Graduate School of Information Sciences, Tohoku University,
20072012
DOI 10.1007/978-3-319-44906-7
327
328
Curriculum Vitae
Full Professor, Graduate School of Mathematics, Nagoya University, 2012

current.
Employment (Part time)
Part-time Lecturer (Quantum Information), SOKENDAI (The Graduate University
for Advanced Studies), 2002
Adjunct Associate Professor, Superrobust Computation Project, Information Science and Technology
Strategic Core (21st Century COE by MEXT), Graduate School of Information
Science and Technology, The University of Tokyo, 20042007
Visiting Associate Professor, Graduate School of Information Sciences, Tohoku
University, 2007
Visiting Research Associate Professor, Centre for Quantum Technologies, National
University of Singapore, 20092012
Visiting Research Professor, Centre for Quantum Technologies, National University of Singapore, 2012current.
Visiting Professor, Faculty of Engineering and Information Technology, University
of Technology, Sydney, 2013
Honors, Awards, and Fellowships
1. 12th Japan Academy Medal: Information Theory and Quantum Information
Theory for Finite-Coding-Length This prize is awarded by The Japan Academy
to up to 6 prestigious Japanese researchers under 45 years old among all scienti_c
areas. Awardees are selected from among the annual recipients of the JSPS Prize.
2. FY2015 JSPS PRIZE: Information Theory and Quantum Information Theory
for Finite-Coding-Length The JSPS PRIZE is meant to recognize at an early
stage in their careers young researchers with fresh ideas who have the potential to
become world leaders in their _elds, while helping to enhance their opportunities
to advance their research and make breakthroughs. Twenty-_ve researchers under
45 years old were selected for this years PRIZE. Their _elds of research run the
spectrum from the humanities and social sciences to the natural sciences.
3. 2011 IEEE Information Theory Society Paper Award: Information spectrum
approach to second-order coding rate in channel coding IEEE Transactions on
Information Theory, Vol. 55, No. 11, 49474966 (2009). This prize is the most
distinguished paper award in the information theory community.
4. Senior Member of IEEE, 2013.
5. Japan IBM Prize in the Computer Science Section 2010: Universal protocol
in quantum information and its application to quantum key distribution This
prize is one of the most distinguished prizes in Japan in information science for
researchers under 45 across Japan.
6. Funai Foundation for Information Technology Award in the Computer Science
Category 2010: Universal quantum information protocol and its application to
quantum cryptography.
7. Research Fellowship for Young Scientists, the Japan Society for the Promotion of
Science, 1998.
Curriculum Vitae
329
8. 2001 SITA Encouragement Award (by The Society of Information Theory and its
Applications (SITA)): Variable length universal entanglement concentration by
local operations.
9. 16th TEPIA Video Award, 2006 (by Association for Technological Excellence
Promoting Innovative Advances (TEPIA)): This prize is awarded to Public Video
for ERATO Quantum computation and information project. I managed the production process as Research Manager.
Professional Activities
Editorial Board Member
1. Editorial Board Member, International Journal of Quantum Information (IJQI)
World Scientific, 2002Current.
2. Advisory Board Member, International Journal On Advances in Security, IARIA,
2009Current. (Editorial Board Member: 2008-2009)
Organization of International Meetings
1. Organizing Committee Chair: ERATO Conference on Quantum Information Science 2003 (EQIS 03), Doshiha Univ., Kyoto, Japan, September 46, 2003.
2. Organizing Committee and Program Committee: COE Symposium on Quantum
Information Theory, Doshiha Univ., Kyoto, Japan, September 23, 2003.
3. Program Committee Vice-Chair: ERATO Conference on Quantum Information
Science 2004 (EQIS 04), Hitotsubashi-memorial hall, Tokyo, Japan, September
15, 2004.
4. Program Committee Vice-Chair: ERATO Conference on Quantum Information
Science 2005 (EQIS 05), JST-Museum Hall, Tokyo, Japan, August 2630, 2005.
5. Organizing Committee Chair: COE-Kakenhi Workshop on Quantum Information
Theory and Quantum Statistical Inference, University of Tokyo, Hongo, Tokyo,
November 1718, (2005)
6. Local Committee: The 8th International Conference on Quantum Communication, Measurement, and Computing (QCMC2006), Tsukuba Epocal, Tsukuba,
Japan, November 28December 3, (2006).
7. Program Committee: Asian Conference on Quantum Information Science 2006
(AQIS06), Beijing, China, September 14, 2006.
8. Technical Program Committee, The First International Workshop on Quantum
Security (QSEC2007), Guadeloupe, French Caribbean, France, January 26,
2007. Organized by IEEE France and IARIA.
9. Organizing Committee Chair: Special seminar series on quantum information,
National Institute of Informatics (NII), Tokyo, February 1March 21, 2007.
10. Program Committee: The 3rd Workshop on Theory of Quantum Computation,
Communication, and Cryptography (TQC2008), The University of Tokyo, Tokyo
Japan, January 30February 1, 2008.
11. Technical Program Committee Co-Chair: ICQNM 2008, The Second International Conference on Quantum, Nano, and Micro Technologies, Sainte Luce,
330
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
Curriculum Vitae
Martinique, France, February 1015, 2008. Organized by IEEE France and

IARIA.
Main Organizer: GSIS & DEX-SMI Workshop on Quantum statistical inference
and entanglement Graduate School of Information Sciences, Tohoku University,
February, 1112, 2008
International Advisory Committee: International School and Conference on
Quantum Information (ISCQI - 2008), Institute of Physics (IOP), Bhubaneswar,
Orissa, India, March 412, 2008
Program Committee: Asian Conference on Quantum Information Science 2008
(AQIS08), Seoul, Korea, August 2526, 2008.
Organizing Committee: International Workshop on Statistical-Mechanical Informatics 2008 (IW-SMI2008), Sendai, Japan, September 1417, 2008.
Organizing Committee Chair: GSIS Workshop on Quantum Information Theory,
Sendai, Japan, November 57, 2008
Organizing Committee: Multicritical Behaviour of Spin Glasses and Quantum
Error Correcting Codes (MBQEC), Centennial Hall, Ookayama campus, Tokyo
Institute of Technology, Tokyo, Japan, November 1719 (2008)
Organizing committee chair: DEX-SMI workshop on quantum statistical inference, National Institute of Informatics (NII), Tokyo, Japan March, 24 (2009).
Program Committee: ICQNM 2009, The Third International Conference on
Quantum, Nano and Micro Technologies, Cancun, Mexico, February 16 (2009)
Program Committee: Workshop on Theory of Quantum Computation, Communication, and Cryptography (TQC2009), Waterloo, Canada, May 1113 (2009)
Program Committee: ICQNM 2010, The Fourth International Conference on
Quantum, Nano and Micro Technologies, St. Maarten, Netherlands Antilles,
February 1016 (2010).@
Program Committee (Special Area Chair): ICQNM 2011, The Fifth International
Conference on Quantum, Nano and Micro Technologies, Nice, France, August
2127 (2011).
Technical Program Committee: ICQNM 2012, The Sixth International Conference on Quantum, Nano and Micro Technologies, Rome, Italy, August 1924
(2012)
Program Committee: 12th Asian Quantum Information Science conference
(AQIS12), School of Physical Science and Technology, Soochow University,
Suzhou, China, August 2326 (2012).
Workshop Co-chair: Japan-Singapore Workshop on Multi-user Quantum Networks, Centre for Quantum Technologies, National University of Singapore,
Singapore, September 1720 (2012)
Special Area Chair: ICQNM 2013, The Seventh International Conference on
Quantum, Nano and Micro Technologies, Barcelona, Spain, August 2531
(2013)
Program Committee: 13th Asian Quantum Information Science conference
(AQIS 13), IMSC Chennai, India, August 2530 (2013)
Program Committee: The 7th International Conference on Information Theoretic
Security (ICITS), NTU, Singapore, November, 2830 (2013)
Curriculum Vitae
331
29. Program Committee: 2014 International Symposium on Information Theory

(ISIT), Honolulu, Hawaii, June 29July 4 (2014)
30. Program Committee: 14th Asian Quantum Information Science conference
(AQIS 14), Kyoto, Japan, August 2125 (2014)
31. Special Area Chair: ICQNM 2014, The Eighth International Conference on
Quantum, Nano and Micro Technologies, Lisbon, Portugal, November 1620
(2014)
32. Program Committee: The 9th Conference on the Theory of Quantum Computation, Communication and Cryptography (TQC2014), National University of
Singapore, Singapore, May, 21-23 (2014)
33. Organizer: Australia-Japan Workshop on Multi-user Quantum Networks, University of Technology, Sydney, October, 2224 (2014)
34. Program Committee: Quantum Information Processing (QIP 2015), Sydney January (2015)
35. Program Committee: 2015 International Symposium on Information Theory
(ISIT), Hong Kong, June 1419 (2015)
36. Special Area Chair: ICQNM 2015, The Seventh International Conference on
Quantum, Nano and Micro Technologies, Venice, Italy, August 2328, 2015
37. Program Committee: 2015 IEEE Information Theory Workshop (ITW), Jeju
Island, Koreag, October 1115, 2015
38. Special Area Chair: ICQNM 2016, The Tenth International Conference on Quantum, Nano and Micro Technologies, Nice, France, July 2428, 2016
39. Program Committee: 16th Asian Quantum Information Science conference
(AQIS16), Academia Sinica, Taipei, Taiwan, August 29-September 1, 2016
40. Conference Co-Chair: Beyond I.I.D. in Information Theory 2017, Institute for
Mathematical Sciences, National University of Singapore, Singapore, July 24
28, 2017
Review Committee for Ph.D. Thesis
1. 2005, Review Committee Member for Two Ph.D. Candidates, Graduate School
of Information Science and Technology, The University of Tokyo, Japan.
2. 2006, Review Committee Member for a Ph.D. Candidate, Graduate School of
Information Science and Technology, The University of Tokyo, Japan.
3. 2008, External examiner (referent), Faculty of Mathematics and Natural Sciences,
University of Leiden, Netherlands
4. 2008, External examiner, National Institute of Education, Nanyang Technological
University, Singapore.
5. 2008, External examiner, Department of Communication and Integrated Systems,
Graduate School of Science and Engineering, Tokyo Institute of Technology.
6. 2012, External examiner, School of Computer Science, McGill University,
Canada.
7. 2013, External examiner, Graduate School of Information Sciences, Tohoku University, Japan.
8. 2014, External examiner, Department of Electrical and Computer Engineering,
National University of Singapore.
332
Curriculum Vitae
Other Kinds of Review Committees

1. 2015, External Reviewer of Research Committee, the Czech Science Foundation.
2. 2014, External Reviewer of Research Committee, University of Macau (UM) for
the 2014 Multi-Year Research Grant (MYRG).
3. 2012, External Reviewer of TATA INSTITUTE OF FUNDAMENTAL
RESEARCH.
Other Professional Activities
1. I am a founder of the Asian Conference on Quantum Information Science (AQIS)
conference series, which is a major international conference series on quantum
information and computation.
2. As Research Manager, I oversaw the ERATO Quantum Computation and Information Project, which at the time was the largest research group for quantum
information and computation in Japan.
3. I have served as a referee for renowned international journals including Nature
Photonics, IEEE Transactions on Information Theory, Communications in Mathematical Physics, Physical Review Letters, Physical Review A, Journal of Physics
A, New Journal of Physics, Quantum Information and Compu- tation, Journal of
Mathematical Physics, and many prestigious journals.
Membership of Academic Societies
IEEE (The Institute of Electrical and Electronics Engineers): Senior Member.
The Mathematical Society of Japan.
JPS (The Physical Society of Japan).
IEICE (The Institute of Electronics, Information and Communication Engineers).
Index
A
Action, 24
Action of group, 24
Adjoint operator, 15
Adjoint representation, 87, 88, 131
Alternating group, 23
Alternative representation, 86
Alternative tensor product space, 249, 284
Alternative tensor representation, 88, 131
Angular momentum, 158
Annihilation operator, 234
Anti-blue, 179
Anti-down quark, 178
Anti-green, 179
Anti-quark, 178
Anti-red, 179
Anti-strange quark, 178
Anti-up quark, 178
Associated representation, 50
Associative law, 22, 91
Atomic number, 164
Attenuation channel, 244
Automorphic action, 27
Automorphism, 27
Average number, 235
Azimuthal quantum number, 155, 165
B
Baryon, 179, 181
Baryon decuplet, 181, 182
Baryon octet, 182
Basis
of compact semi simple Lie algebra, 206
of complex semi simple Lie algebra, 209
Bilinearity, 60, 76
Blue, 178
Boson, 14, 140, 231, 248

Bosonic system
multi-mode, 240, 259
one-mode, 231
r -mode, 240
two-mode, 253
Bottom quark, 185
Bounded, 15
C
Canonical MUB, 285
Cartan decomposition, 103, 223
Cartan matrix, 130, 202
Cartan subalgebra, 205
Casimir operator, 94, 154, 162
Center, 26, 78
Central extension, 30, 33
Centralizer, 78, 26
Character, 41, 52
Charm quark, 183
Classical type, 207
Clebsch-Gordan coefficient, 125, 176
Clifford group, 274
Closed extension, 16
Closed operator, 15
Coherent state, 222, 227, 229
Coherent vector, 127, 227
Color, 178
Color confinement, 180, 181
Control NOT(C-NOT), 280
Commutative, 77, 78
Commutative group, 23
Commutative law, 23
Commutator, 76
Commutator subgroup, 27
Compact

DOI 10.1007/978-3-319-44906-7
333
334
element of Lie algebra
element of Lie algebra, 92
Lie algebra, 93
Compact Lie group, 70
Complementary series representations, 120
Completely orthonormal system, 1
Completely reducible, 40, 85, 87
Completeness, 1, 15
Completion, 15
Complex conjugate matrix, 9
Complex conjugate representation, 53, 85,
87, 131, 179
Complex conjugate transposed matrix, 9
Complex Lie algebra, 76
Complex Lie group, 74
Complex Lie subalgebra, 77
Complexifiable, 55, 87
Complexification, 77
Computational basis, 263
Conjugacy class, 26
Conjugate isomorophism, 54
Conjugate isomorphic, 54
Conjugate linear, 54
Connected, 202
root system, 202
Connected irreducible parabolic system, 217
CONS, 1
Convex combination, 7
Creation operator, 234
C-representation, 87, 101
Cyclic group, 23, 26, 58
D
Decomposable, 40, 8587
Degenerate, 2
Density matrix, 4
Design
approximate f-, 149
approximate -, 149
f-, 144, 145
-, 145, 147
S-, 145
(t1 , t2 )-, 146, 148
Direct product (group), 28
Direct sum representation, 39, 86
Discrete Fourier transform, 58, 279
Discrete Heisenberg group, 35, 36, 265, 268,
270, 272
Discrete Heisenberg representation, 265,
267, 269
Discrete series representation, 225
Discrete symplectic group, 273
Index
Displacement transformation, 233
Double covering, 247
Double covering group, 97
Down quark, 177
Dual basis, 58
Dual computational basis, 265, 268
Dynkin diagram, 202
Dynkin index, 130, 132, 133, 147, 177, 178,
202, 212
E
Electric charge, 178
Elementary charge, 164
Entangled state, 8
Entanglement, 8
Environment system, 244
Equivalent
path, 95
Equivalent, 24, 46
Essentially self-adjoint, 16
Even permutation, 23
Exceptional type, 207
Expectation, 4
Extended Heisenberg group
with degree 1, 237
Extended Heisenberg representation, 238
Extension, 29
F
Factor system, 31, 33, 46
Faithful, 38, 8587
Fermion, 14, 140, 178, 180, 248
Finite group, 23, 59
Finitely generated, 23
First quantization, 248
Flavor, 177
Formal dimension, 103
Fourier transform, 63, 108, 255
Fundamental group, 95
Fundamental representation, 88, 131
Fundamental system, 130, 202
G
Gauss decomposition, 114, 209
General linear group, 37
Generate, 23
Generating matrix, 277
Generator, 23
Green, 178
Ground state, 122
Group, 22
Index
Group algebra, 59
Group of extension, 34
H
Haar measure, 101
Hamiltonian, 2, 11, 156, 175
Harish-Chandra condition, 226
Hausdorff formula, 232
Heisenberg representation, 233
Hermitian, 17
Hermitian Dynkin diagram, 218
Hermitian type, 224
Heterodyne measurement, 236
Highest weight, 121, 130, 132, 212, 226
Highest weight representation, 225
Highest weight vector, 212, 225
Hilbert space, 1
Homogeneous space, 24
Homomorphism
complex Lie algebra, 77
complex Lie group, 101
group, 26
linear algebra, 60
real Lie algebra, 77
Homomorphism theorem, 27
Horizontal permutation, 66
Horizontal permutation group, 66
Husimi function, 236
Hyperbolic system, 215
Hypercharge, 178
I
Ideal, 77
Identity element, 22
Indecomposable, 40, 8587
Index, 25
Index group, 26
Inertia group, 50
Invariance, 24, 39, 85
Invariant measure, 101, 103
Invariant subgroup, 25
Invariant subspace, 40, 85
Inverse element, 22
Inverse Fourier transform, 63, 108, 255
Irreducible, 40, 8587, 217
h1 , 223
k-, 227
Irreducible character, 41, 52
Irreducible decomposition, 41, 43, 45, 51,
52, 58, 134, 138, 139, 145, 146, 227,
228, 239, 243, 269, 283
Irreducible parabolic system, 217
335
Isomorphic
group, 26
Lie algebra, 77
projective unitary representation, 39
representation, 38
Isomorphism
group, 26
representation, 39
unitary representation, 39
Isospin, 178
J
Jacobi law, 76
K
Kernel, 27
Killing form, 90
Kinetic energy, 158
Klein four-group, 23
L
Laplacian, 158
Larmor interaction, 175
Left-invariant, 101
Legendre equation, 155
Legendre polynomial, 155
Lie group, 69
Linear algebra, 60
Linear interaction, 243
Linear operation, 243
LittlewoodRichardson coefficient, 134
LittlewoodRichardson rule, 134
M
Magnetic quantum number, 155, 165
Maximal Cartan subalgebra, 205
Maximal compact subalgebra, 223
Maximal compact subgroup, 103
Maximal regular subalgebra, 213
Maximum root, 203
Meson, 179
Meson octet, 180
Meson singlet, 180
Metaplectic representation, 278, 279
Mixed state, 4
Momentum operator, 152
Multi-mode bosonic system, 240, 259
Multiplicity, 52
Multiplicity coefficient, 56
Multi-variable system, 239, 240
Mutually unbiased, 285
336
Mutually unbiased bases(MUB), 285
N
Negative definite, 91
Negative root, 130, 202
Neutron, 177
Nilpotent, 78
Non-definite, 91
Non-degenerate, 2
Normal subgroup, 25
Normalized, 102, 103
Normalized factor system, 31
Normalizer, 26
Number operator, 234
Number vector, 233
O
One-mode bosonic system, 231, 233
One-variable system, 231, 233
Orbital angular momentum, 175
Orbital angular momentum operator, 152
Order, 23
Orthogonal representation, 53
P
Parabolic system, 228
Partial order, 202
Partial trace, 8
Partition function, 65
Path, 78
Permittivity, 164
Permutation group, 23, 64
P-function, 236
Plancherel formula, 61
Plancherel measure, 67, 110
Planck constant, 152, 232
Position operator, 152
Positive definite, 91
Positive root, 130, 202
Pre-discrete-Heisenberg representation, 265
Principal quantum number, 165
Principal series representations, 120
Projection, 25
Projection-valued measurement, 3
Projective representation, 37, 88
Projective skew-Hermitian representation,
88
Projective unitary representation, 37
Proper projective representation, 46, 89
Proper projective unitary representation, 46
Proton, 177, 182
Index
Pseudoscalar meson, 180
Pure state, 4
PVM, 3
Q
Q-function, 236
Quantum system, 1
Quark, 177
Quotient group, 26
Quotient space, 25
R
Radical, 78
Rank, 205, 208
Real Lie algebra, 76
Real Lie subalgebra, 77
Real representation
group, 52
Lie algebra, 86
Red, 178
Reduced density matrix, 8
Reduced mass, 164
Reducible, 39, 46, 85
Regular subalgebra, 213
Relativistic quantum mechanics, 178
Representation
adjoint, 87, 88, 131
alternative tensor, 88, 131
associated, 50
C-, 87, 101
complex conjugate, 53, 131
direct sum, 39
discrete Heisenberg, 265, 267, 269
discrete series, 225
for complex Lie algebras, 87
for complex Lie groups, 101
fundamental, 88, 131
group, 37
Heisenberg, 233
highest weight, 225
Lie algebra, 85
Metaplectic, 278
orthogonal, 53
pre-discrete-Heisenberg, 265
projective, 37, 88
projective skew-Hermitian, 88
proper projective, 46, 89
proper projective unitary, 46
real, 52
self-adjoint, 131
self-conjugate, 53
Index
semi direct product, 50
semi direct product unitary, 50
skew-Hermitian, 85
symmetric tensor, 88, 131
tensor product, 39, 137
unitary, 37
Representation for complex Lie algebras, 87
Representation for complex Lie groups, 101
Representation space, 37, 52
Residue class, 25
Residue group, 26
Riemann sphere, 115, 127
Right-invariant, 101
r -mode bosonic system, 240
Root, 130, 208
Root basis, 202
Root set, 228
Root subsystem, 214
Root system, 130, 201
S
Schrdinger equation, 11, 158, 164, 176
Schur duality, 138
Schur function, 137
Second quantization, 248
Self-adjoint, 16
Self-adjoint representation, 53, 85
Self-conjugate representation, 53
Self-orthogonal subgroup, 271
Self-adjoint representation, 131
Semi direct product, 28
Semi direct product Lie algebra, 90
Semi simple, 78
Semistandard Young tableau, 65
Separable state, 7
Simple, 27, 77
Simple root, 130, 202
Simply connected, 96
Size, 65
Skew-diagram, 134
Skew-Hermitian representation, 85
Skew-symmetry, 76
Skew-adjoint representation, 175
SL(2, C), 114
sl(2, C), 82, 114, 115, 117, 120
SL(2, R), 118, 251, 252, 254
sl(2, R), 82, 118, 119, 224, 226, 251
Solvable, 78
Sp(2r, R), 72
Special linear group, 37
Special orthogonal group, 162
Special unitary group, 37
337
Spherical coordinate, 152, 159, 174
Spherical Laplacian, 159
Spin, 175, 178
Spin-orbit interaction, 175
Squeezed vector, 251
Stabilizer, 25
Standard Young tableau, 65
Stereographic projection, 167
Stone-von Neumann Theorem, 232
Strange quark, 177
Strangeness, 178
Strictly self-orthogonal subgroup, 271
SU(1, 1), 116119, 127, 252254
su(1, 1), 82, 116120, 122, 124
SU(2), 79, 114116, 119, 122, 127, 178
su(2), 79, 115, 116, 119, 120, 122, 124, 129,
138, 175, 206
su(2, R), 251
SU(3), 177
su(3), 178
SU(6), 181
SU(r ), 129, 130, 132, 137139, 243, 247
su(r ), 87, 129132, 201, 208, 220
Subgroup, 23
Subrepresentation, 40
Symmetric, 16
Symmetric informationally complete (SIC)
vectors, 288
Symmetric tensor product space, 139, 248,
284
Symmetric tensor representation, 88, 131
Symplectic group, 259
System, 1
T
Tensor product representation, 39, 86, 137
Tensor product space, 6, 10
Tensor product state, 6, 10
Thermal equilibrium state, 235
Thomas interaction, 175
Top quark, 185
Total angular momentum, 175
Total orbital angular momentum, 152
Transitive, 24
Transitive law, 24
Transmittance, 243
Transposed matrix, 9
Transposition, 23, 66
Two-mode bosonic system, 253
Two-mode squeezed state, 254
Type I, 109
338
U
u(r ), 87, 131, 132, 207, 243, 261
U(r ), 131, 132, 134, 135, 242, 243, 247, 248,
260, 261
Unbounded, 15
Uncertainty relation, 186, 257
Underlying real form, 54, 87
Unimodular, 101, 109
Unit circle, 23, 104
Unit disk, 117, 127
Unitarily isomorphic, 39
projective representation, 39
Unitary group, 37
Unitary map, 37
Unitary representation, 37
Universal covering group, 97
Up quark, 177
V
Vacuum state, 233
Index
Variance, 4
Vectorial meson, 180
Vertical permutation, 66
W
Wave function, 4
Weight, 121, 130, 132, 212, 225
Weight vector, 121
Weyl group, 201
Weyls dimension formula, 133, 137, 212
Wigner function, 236
Y
Young diagram, 64, 132, 134
Young horizontal symmetrizer, 66
Young index, 132, 133
Young symmetrizer, 66
Young tableau, 65
Young vertical symmetrizer, 66

Teoria Grupos Quantum Mechanics

Uploaded by

Copyright:

Available Formats

Teoria Grupos Quantum Mechanics

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Teoria Grupos Quantum Mechanics

Uploaded by

Copyright:

Available Formats

What are the main topics covered in the book?

What are the main topics covered in the book?

What are some of the applications of representation theory discussed in the book?

What are some of the applications of representation theory discussed in the book?

MasahitoHayashi

Group Representation for Quantum Theory

Library of Congress Control Number: 2016950886

Preface to the Japanese Version

Group representational symmetry is one of most fundamental concepts in quantum

Preface to the Japanese Version

Preface to the Japanese Version

Preface to the Japanese Version

Preface to the Japanese Version

2 Group Representation Theory . . . . . . . . . . . . . . . . . . . . . .

1 Mathematical Foundation for Quantum System . . . .

Semi Direct Product and Its Representation . . . . . . . . . . . . . . . . .

3 Foundation of Representation Theory of Lie Group

Invariant Measures on Group and Homogeneous Space . . . .

4 Representations of Typical Lie Groups and Typical

5.3.3 Fourier Transform of Coulomb Potential* . . . . . . . .

6 Representation of General Lie Groups and General

8 Discretization of Bosonic System . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Transformation by Unitary Group Ur . . . . . . . . . . . . . . . .

Appendix A: Solutions of Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291

About the Author

Direct product group of groups H and K

Centralizer of subgroup M of group G

Extension of group G by U1 dened by 2 CG

Discrete symplectic group

Quotient space, quotient group (residue group) when H is

Direct sum of g1 and g2

Set of irreducible unitary representations of G

Set of nite-dimensional projective representations in

Set of irreducible skew-Hermitian presentations of Lie algebra g

Table 4 Symbols for Representation

Direct sum representation

CONS of U G weight vector for SU2

Table 6 Symbols for vector space

61, 104, 107

Set of square integrable functions of conjugated class

Set of real-valued functions in L2 G; inv

Table 7 Symbols for matrices and operators

19, 186, 234

Table 8 Other symbols

Complex conjugate matrix of X

Transposed complex conjugate matrix of X

Order (number of elements) of G

Product set of sets H and K

Element of HG : U1 corresponding to projective

Mathematical Foundation for Quantum

Abstract This book introduces group representation theory in terms of quantum

1.1 System, State, and Measurement

for a, b C. Then, the inner

case. see Sect. 1.6.

1 Mathematical Foundation for Quantum System

where ai is an eigenvalue of A and i is an eigenvector corresponding to the eigenvalue ai .

is defined as a self-adjoint operator in the infinite-dimensional case. The precise discussion is

1.1 System, State, and Measurement

Fig. 1.1 Measurement

possible to realize the measurement whose outcome corresponding to the base i is