Quantum Me

hani s in
Rigged Hilbert Spa e Language
by

Rafael de la Madrid Modino

DISSERTATION
Submitted in partial fulllment of the
requirements for the degree of
Do tor of Philosophy
Physi s

DEPARTAMENTO DE FISICA TEORICA

FACULTAD DE CIENCIAS. UNIVERSIDAD DE VALLADOLID
May 2001
 2002 Rafael de la Madrid Modino

Last updated: Mar h 28, 2002

To my parents, to my siblings,
and to those who always believed
in me
If you are lu ky enough to have lived in
Austin as a young man, then wherever
you go for the rest of your life, it stays
with you, for Austin is a movable feast
A knowledgments
To write a dissertation is an experien e that is at on e hallenging and rewarding. Although
one has total ontrol over his work, this dissertation would not have been possible without
the help of many people. Here they are:
My deepest appre iation to my supervisor Professor Manuel Gadella for his support, for
the insight into the RHS that I gained from him, and for letting me do my own thing.
For the past three years, I have been fortunate to work at the University of Texas at
Austin. It has been the fulllment of all my dreams. I have been in Austin be ause of one
single reason: Professor Arno Bohm. Thanks from the bottom of my heart.
Love and ae tion to my family for the enormous sa ri e they have made and for their
extraordinary sense of world wonder.
Many friends have helped me keep my mental balan e. Espe ially Al for opening a whole
new world to me, Arturo for putting me up so many times, Toro for being always so sweet,
and Casey for her love.
Spe ial thanks to Ramon Tre et for his radio show \Dialogos 3," that is denitely one of
the greatest in uen es on my life. Also thanks to the the rest of the radio shows in Radio 3.
So sorry I an't listen to you that often any more.
My gratitude to my oworkers for the ex hange of ideas; espe ially to Manish and Nathan
for their help with the English, to Luismi and Jose Manuel for their help with the appli ation
forms, to Mariano del Olmo for his advise on dissertation style and to the students of Phy
381N and Steve Newberry for proofreading Chapters 2 and 3 and for making invaluable
suggestions. Thanks also to Prof. Galindo for invaluable suggestions, whi h prompted this
revised version of the dissertation.
Finally, thanks so mu h to the ity of Austin for being a movable feast.

This dissertation was typeset with LATEX1 by the author.

1
Latex is a do ument preparation system developed by Leslie Lamport as a spe ial version of Donald
Knuth's TEXProgram.
Contents

1 Introdu tion 1
1.1 A Brief History of the Rigged Hilbert Spa e . . . . . . . . . . . . . . . . . . 3
1.2 Harmoni Os illator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 A Rigged Hilbert Spa e of the Square Barrier Potential . . . . . . . . . . . . 9
1.4 S attering o the Square Barrier Potential . . . . . . . . . . . . . . . . . . . 12
1.5 The Gamow Ve tors of the Square Barrier Potential Resonan es . . . . . . . 15
1.6 Time Reversal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.7 Synopsis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2 Mathemati al Framework of Quantum Me hani s 23
2.1 Linear Spa es . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.1.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.1.2 Linear Spa es and S alar Produ t . . . . . . . . . . . . . . . . . . . . 25
2.1.3 Linear Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.1.4 Antilinear Fun tionals . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2 Topologi al Spa es . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.2.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.2.2 Open Sets and Neighborhoods . . . . . . . . . . . . . . . . . . . . . . 34
2.2.3 Separation Axioms . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.2.4 Continuity and Homeomorphi Spa es . . . . . . . . . . . . . . . . . 38
2.3 Linear Topologi al Spa es . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.3.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.3.2 Cau hy Sequen es . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.3.3 Normed, S alar Produ t and Metri Spa es . . . . . . . . . . . . . . . 43
2.3.4 Continuous Linear Operators and Continuous Antilinear Fun tionals 45
2.4 Countably Hilbert Spa es . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.4.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.4.2 Dual Spa e of a Countably Hilbert Spa e . . . . . . . . . . . . . . . . 53
2.4.3 Countably Hilbert Spa es in Quantum Me hani s . . . . . . . . . . . 54
2.5 Linear Operators on Hilbert Spa es . . . . . . . . . . . . . . . . . . . . . . . 56
2.5.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.5.2 Bounded Operators on a Hilbert Spa e . . . . . . . . . . . . . . . . . 57
2.5.3 Unbounded Operators on a Hilbert Spa e . . . . . . . . . . . . . . . . 62

xi
xii Contents

2.6 Nu lear Rigged Hilbert Spa es . . . . . . . . . . . . . . . . . . . . . . . . . . 65

2.6.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.6.2 Nu lear Rigged Hilbert Spa es . . . . . . . . . . . . . . . . . . . . . . 66
3 The Rigged Hilbert Spa e of the Harmoni Os illator 69
3.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.2 Algebrai Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.3 Constru tion of the Topologies . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.3.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.3.2 Hilbert Spa e Topology . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.3.3 Nu lear Topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.3.4 Physi al Interpretation of ,  and H . . . . . . . . . . . . . . . . . 85
3.3.5 Extension of the Algebra of Operators . . . . . . . . . . . . . . . . . 85
3.4 The RHS of the Harmoni Os illator . . . . . . . . . . . . . . . . . . . . . . 89
3.4.1 The Conjugate Spa e . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
3.4.2 Constru tion of the Rigged Hilbert Spa e . . . . . . . . . . . . . . . . 91
3.4.3 Continuous Linear Operators on the Rigged Hilbert Spa e . . . . . . 93
3.5 Basis Systems, Eigenve tor De omposition and the Gelfand-Maurin Theorem 94
3.5.1 Basis Systems and Eigenve tor De omposition|a Heuristi Introdu -
tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.5.2 Gelfand-Maurin Theorem . . . . . . . . . . . . . . . . . . . . . . . . 104
3.6 Gelfand-Maurin Theorem Applied to the Harmoni Os illator . . . . . . . . 109
3.6.1 Spe tral Theorem Applied to the Energy Operator . . . . . . . . . . 109
3.6.2 Spe tral Theorem Applied to the Position and Momentum Operators 110
3.6.3 Realizations of the RHS of the Harmoni Os illator by Spa es of Fun -
tions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3.6.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
3.7 A Remark Con erning Generalizations . . . . . . . . . . . . . . . . . . . . . 128
3.7.1 Realization of the Abstra t RHS by Spa es of Fun tions . . . . . . . 128
3.7.2 General Statement of the Gelfand-Maurin Theorem . . . . . . . . . . 133
3.7.3 Generalization of the Algebra of Operators . . . . . . . . . . . . . . . 134
3.7.4 Appendix: Continuity of the Algebra of the Harmoni Os illator . . . 135
4 A Rigged Hilbert Spa e of the Square Barrier Potential 137
4.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
4.2 Sturm-Liouville Theory Applied to the Square Barrier Potential . . . . . . . 141
4.2.1 S hrodinger Equation in the Position Representation . . . . . . . . . 141
4.2.2 Self-Adjoint Extension . . . . . . . . . . . . . . . . . . . . . . . . . . 142
4.2.3 Resolvent and Green Fun tions . . . . . . . . . . . . . . . . . . . . . 143
4.2.4 Diagonalization of H and Eigenfun tion Expansion . . . . . . . . . . 146
4.2.5 The Need of the RHS . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
4.2.6 Constru tion of the Rigged Hilbert Spa e . . . . . . . . . . . . . . . . 154
4.2.7 Dira Basis Ve tor Expansion . . . . . . . . . . . . . . . . . . . . . . 155
Contents xiii

4.2.8 Energy Representation of the RHS . . . . . . . . . . . . . . . . . . . 156

4.2.9 Meaning of the Æ -normalization of the Eigenfun tions . . . . . . . . . 157
4.3 Con lusion to Chapter 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
4.4 Appendi es to Chapter 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
4.4.1 Appendix 1: Self-Adjoint Extension . . . . . . . . . . . . . . . . . . . 160
4.4.2 Appendix 2: Resolvent and Green Fun tion . . . . . . . . . . . . . . 161
4.4.3 Appendix 3: Diagonalization and Eigenfun tion Expansion . . . . . . 163
4.4.4 Appendix 4: Constru tion of the RHS . . . . . . . . . . . . . . . . . 165
4.4.5 Appendix 5: Dira Basis Ve tor Expansion . . . . . . . . . . . . . . . 167
4.4.6 Appendix 6: Energy Representation of the RHS . . . . . . . . . . . . 168
5 S attering o the Square Barrier Potential 171
5.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
5.2 Lippmann-S hwinger Equation . . . . . . . . . . . . . . . . . . . . . . . . . . 174
5.2.1 Lippmann-S hwinger Kets . . . . . . . . . . . . . . . . . . . . . . . . 174
5.2.2 Radial Representation of the Lippmann-S hwinger Equation . . . . . 175
5.2.3 Solution of the Radial Lippmann-S hwinger Equation . . . . . . . . . 177
5.2.4 Dire t Integral De omposition Asso iated to the In-States . . . . . . 178
5.2.5 Dire t Integral De omposition Asso iated to the Observables . . . . . 183
5.3 Constru tion of the Lippmann-S hwinger Kets and Dira Basis Ve tor Ex-
pansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
5.4 S-matrix and Mller Operators . . . . . . . . . . . . . . . . . . . . . . . . . 188
5.5 Appendi es to Chapter 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
5.5.1 Appendix 7: Free Hamiltonian . . . . . . . . . . . . . . . . . . . . . . 190
5.5.2 Appendix 8: Spa es of Hardy Fun tions . . . . . . . . . . . . . . . . 198
6 The Gamow Ve tors of the Square Barrier Potential Resonan es 203
6.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
6.2 S-matrix Resonan es . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
6.3 The Gamow Ve tors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
6.3.1 Lippmann-S hwinger Equation of the Gamow Ve tors . . . . . . . . . 208
6.3.2 The Gamow Ve tors in Position Representation . . . . . . . . . . . . 209
6.3.3 The Gamow Ve tors in Energy Representation . . . . . . . . . . . . . 212
6.4 Complex Basis Ve tor Expansion . . . . . . . . . . . . . . . . . . . . . . . . 216
6.5 Semigroup Time Evolution of the Gamow Ve tors . . . . . . . . . . . . . . . 217
6.6 Time Asymmetry of the Purely Outgoing Boundary Condition . . . . . . . . 219
6.6.1 Outgoing Boundary Condition in Phase . . . . . . . . . . . . . . . . . 219
6.6.2 Outgoing Boundary Condition in Probability Density . . . . . . . . . 220
6.7 Exponential De ay Law of the Gamow Ve tors . . . . . . . . . . . . . . . . . 221
6.8 Con lusion to Chapter 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
6.9 Appendix 9: Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
xiv Contents

7 The Time Reversal Operator in the Rigged Hilbert Spa e 227

7.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
7.2 The Standard Time Reversal Operator (T = I = 1) . . . . . . . . . . . . . 230
7.3 The Time Reversal Doubling (T = I = 1) . . . . . . . . . . . . . . . . . . 234
7.4 Appendix 10: Time Reversal . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
8 Con lusions 245
Chapter 1
Introdu tion
In this hapter, we sket h the ontents of this dissertation. These ontents will be mostly
on erned with the properties of Dira kets, Lippmann-S hwinger kets, and Gamow ve tors.

Jim looked at the trash, and then looked at me, and ba k at the
trash again. He had got the dream xed so strong in his head that he
ouldn't seem to shake it loose and get the fa ts ba k into pla e
again, right away. But when he did get the things straightened around,
he looked at me steady, without ever smiling, and says:
\What do dey stan' for? I's gwyne to tell you. When I got all wore
out wid work, en wid de allin' for you, en went to sleep, my heart
wuz mos' broke bekase you wuz los', en I didn' k'yer no mo' what
be ome er me en de raf'. En when I wake up en ne you ba k agin,
all safe en soun', de tears ome en I ould a got down on my knees
en kiss' yo' foot I's so thankful. En all you wuz thinkin 'bout wuz
how you ould make a foul uv ole Jim wid a lie. Dat tru k dah is
trash; en trash is what people is dat puts dirt on de head er dey fren's
en makes 'em ashamed."
Then he got up slow, and walked to the wigwam, and went in
there, without saying anything but that. But that was enough. It
made me feel so mean I ould kissed his foot to get him to
take it ba k.
It was fteen minutes before I ould work myself up to go and
humble myself to a nigger{but I done it, and I warn't ever sorry for
it afterwards, neither. I didn't do him no more mean tri ks, and I
wouldn't done that one if I'd a knowed it would make him feel that
way.

Mark Twain, The adventures of Hu kleberry Finn

1
1.1 A Brief History of the Rigged Hilbert Spa e 3

This dissertation is about the des ription of Dira kets, Lippmann-S hwinger kets and
Gamow ve tors in Rigged Hilbert Spa e language. The Dira kets are the state ve tors
asso iated to any element in the spe trum of an observable. The Lippmann-S hwinger kets
are the eigenkets of the Hamiltonian that are relevant in s attering theory. They orrespond
to the monoenergeti \in" and \out" s attering states. The Gamow ve tors are the kets
that represent the state ve tor of a resonan e. Our main goal is to show that the Rigged
Hilbert Spa e is the most suitable formalism to des ribe these kets. Rather than working in
an abstra t fashion, examples shall be used to illustrate this des ription. The two examples
we shall mainly use are the harmoni os illator and the square barrier potential.
In this dissertation, no experimental data is dis ussed. We shall rather fo us on the
methods, the ideas and prin iples in terms of whi h su h data an be interpreted and un-
derstood. We shall use the S hrodinger equation subje t to dierent boundary onditions as
a model for the des ription of the data. Dierent boundary onditions upon the S hrodinger
equation will yield Dira kets, Lippmann-S hwinger kets or Gamow ve tors. Although su h
a model involves an idealization, this is probably the best way to understand what these
state ve tors are.
We should note that the RHS is not an interpretation of Quantum Me hani s, but rather
the most natural, on ise and logi language to formulate su h heuristi physi al on epts
as Dira kets, Lippmann-S hwinger kets or Gamow ve tors.

1.1 A Brief History of the Rigged Hilbert Spa e

In the late 1920's, Dira introdu ed a new mathemati al model of Quantum Me hani s
based upon a uniquely smooth and elegant abstra t algebra of linear operators dened on
an innite dimensional omplex ve tor spa e equipped with an inner produ t norm [1℄.
Dira 's abstra t algebrai model of bras and kets (from the bra ket notation for the inner
produ t) proved to be of great heuristi value in the ensuing years, espe ially in dealing
with Hamiltonians whose spe trum is ontinuous. However, there were serious diÆ ulties in
nding a version of linear algebra whi h ould be employed for making the a tual numeri al
al ulations.
The Hilbert spa e (HS) was the rst mathemati al idealization proposed for Quantum
Me hani s [2℄. However, as von Neumann explains in the introdu tion to his book [2℄, the
HS theory and Dira 's formalism are two dierent things. Although there were attempts to
realize the Dira model in Hilbert spa e, there was a number of serious problems resulting
from the fa t that this formalism annot allo ate su h things as bras, kets or the Dira
delta fun tion or give a mathemati al meaning to the Dira basis ve tor expansion, whi h
are essential in any physi al formulation of Quantum Me hani s that deals with ontinuous
spe trum. Indeed in his textual presentation [1℄ Dira himself states that \the bra and ket
ve tors that we now use form a more general spa e than a Hilbert spa e" (see [1℄, page 40).
In the late 1940's, L. S hwartz gave a pre ise meaning to the Dira delta fun tion as a
fun tional over a spa e of test fun tions [3℄. This led to the development of a new bran h
of fun tional analysis, the theory of distributions [3℄.
4 1 Introdu tion

About the same time, von Neumann published the theory of dire t integral de omposi-
tions of a Hilbert spa e indu ed by a self-adjoint operator [4℄ (also valid for more general
ases). This spe tral theory was loser to lassi al Fourier analysis, and represented an
improvement over former von Neumann's spe tral theory [2℄.
I. Gelfand always thought that von Neumann's spe tral theory was not the whole story
of the theory of linear operators dened on innite dimensional ve tor spa es. Prompted by
the theory of distributions, he and his s hool introdu ed the Rigged Hilbert Spa e (RHS).
Starting out with this RHS and von Neumann's dire t integral de omposition, they were
able to prove the so- alled Nu lear Spe tral Theorem [5℄ (also known as the Gelfand-Maurin
Theorem). This theorem provides a more thorough information on the spe tral properties
of an operator and treats the ontinuous and the dis rete spe trum on the same footing.
One of the aspe ts of Dira 's formalism, the ontinuity of the elements of the algebra
of observables, was dis ussed in the early 1960's in Refs. [6, 7℄. If two operators of the
algebra of observables satisfy the anoni al (Heisenberg) ommutation relation, at least one
of them annot be ontinuous (i.e., bounded) with respe t to the Hilbert spa e topology. In
Refs. [6, 7℄, it is shown that there are subdomains of the Hilbert spa e that an be endowed
with topologies that make those operators ontinuous; the largest of these subdomains is
the S hwartz spa e.
In the 1960's, some physi ists [8, 9, 10℄ independently realized that the RHS provides a
rigorous mathemati al rephrasing of all of the aspe ts of Dira 's formalism. In parti ular,
the Nu lear Spe tral Theorem restates Dira basis ve tor expansion along with the Dira
bras and kets within a mathemati al theory. Later on, other authors ame to the same
on lusion [11℄. Nowadays the RHS is textbook material [12, 13, 14, 15, 16, 17, 18℄.
During the past few years, the RHS has emerged as the natural mathemati al language
in the theory of s attering and de ay ( f. Refs. [19, 20, 21, 22℄ and referen es therein). The
RHS has also proved to be very useful in other areas of theoreti al physi s su h as in the
onstru tion of generalized spe tral de ompositions of haoti maps [23, 24℄. In fa t, it
seems that the RHS is the best known language to deal with s attering and de ay in a
onsistent way. This is the very reason why we are using it here.
The S hrodinger equation is the dynami al equation that governs the behavior of a
quantum system. Thus any attempt to show that the RHS ontains the mathemati al
methods needed by Quantum Me hani s should show that the natural framework for the
solutions of the S hrodinger equation is the RHS. We re all that none of Refs. [19, 20, 21, 22℄
took the S hrodinger equation as the dynami al equation. The obje tive of this dissertation
is to obtain the Dira , Lippmann-S hwinger, and Gamow kets as solutions of the S hrodinger
equation subje t to dierent boundary onditions, and to show that these solutions fall in
the RHS rather than just in the HS [25, 26, 27℄.
In the end, the results of this dissertation will allow us to draw a very important on-
lusion: the RHS is the natural language to deal with s attering and de ay.
1.2 Harmoni Os illator 5

1.2 Harmoni Os illator

If the spe trum of an observable is dis rete, the mathemati al methods of the Hilbert spa e
are suÆ ient for the purposes of Quantum Me hani s. However, if the spe trum of an
observable has a ontinuous part, the mathemati al methods of the Hilbert spa e are not
suÆ ient, and an extension of these methods is needed.
Physi ists use Dira 's bra-ket formalism in order to handle ontinuous spe tra. Four of
the most important features of this formalism are:
1. To ea h element  of the spe trum of an observable A, there orresponds a ket ji
that is an eigenve tor of A with eigenvalue ,
Aji = ji : (1.2.1)

2. A wave fun tion ' an be expanded by these eigenkets,1

Z
'= d jihj'i : (1.2.2)
Spe trum(A)

3. The eigenkets are normalized a ording to the following rule:

hj0i = Æ( 0) ; (1.2.3)
where Æ ( 0 ) is the Dira delta fun tion.
4. All algebrai operations su h as the ommutator of two observables A and B are
always well dened,
[A; B ℄ = AB BA : (1.2.4)
In Quantum Me hani s, observables are assumed to be represented by self-adjoint, linear
operators dened on a Hilbert spa e H. If the operator A asso iated to an observable is
unbounded (whi h is the most ommon ase in Quantum Me hani s), then A is only dened
on a subdomain D(A) on whi h A is self-adjoint. In this ase, the Hilbert spa e methods are
not suÆ ient to make sense of (1.2.1)-(1.2.4). The RHS formalism provides the mathemati s
that are needed to make sense of them.
On the other hand, one of the key assumptions of Quantum Me hani s is that the
quantity
('; A') (1.2.5)
represents the expe tation value of the measurement of the observable A in the state ', and
that p

' A = ('; A2 ') ('; A')2 (1.2.6)
represents the un ertainty of the measurement of the observable A in the state ' (we assume
the wave fun tion ' to be normalized to 1). The expe tation value (1.2.5) annot be
1
Eq. (1.2.2) is referred to as the Dira basis ve tor expansion.
6 1 Introdu tion

omputed for every element of the Hilbert spa e H, but only for those ' 2 H that also belong
to D(A). Similarly, the un ertainty (1.2.6) annot be omputed for every element of H, but
just for those ' 2 D(jAj) [28℄. If we take as physi al states those normalizable fun tions for
whi h physi al quantities su h as the expe tation value (1.2.5) and the un ertainty (1.2.6)
an be omputed, then it is lear that not every square normalizable fun tion (i.e., every
element of H) an represent a physi al state. As we shall see, the natural spa e of physi al
wave fun tions is a subspa e  of H, be ause all physi al quantities an be omputed for
its elements. Further,  has all the ni eties of Dira 's formalism.
For example, let us onsider the harmoni os illator. The algebra of the harmoni os il-
lator ontains the observables position Q and momentum P . These observables are dened
as linear operators over the Hilbert spa e H, and they fulll the Heisenberg ommutation
relation:
[P; Q℄ = P Q QP = i~I : (1.2.7)
It is well known that Eq. (1.2.7) implies that either P or Q is an unbounded operator. This
implies that either P or Q annot be dened on the whole Hilbert spa e|they are, in fa t,
dened on ertain dense subdomains D(P ) and D(Q) on whi h P and Q are self-adjoint.
Therefore, the expression P Q QP is not dened on the whole Hilbert spa e. Moreover,
sin e D(P ) and D(Q) do not remain stable under the a tion of P and Q, the expression
P Q QP is only dened on those ' 2 H su h that ' 2 D(Q), ' 2 D(P ), P ' 2 D(Q) and
Q' 2 D(P ). Therefore, the Heisenberg ommutation relation (1.2.7) is not dened on the
whole of H, but only on a subspa e of it. We re all that Eq. (1.2.7) leads to the Heisenberg
un ertainty relation:
~

' P
' Q  : (1.2.8)
2
Now, if we want the expe tation values of H , P and Q,
('; A') ; A = H; P; Q ; (1.2.9)
the un ertainties of H , P and Q,

' A ; A = H; P; Q ; (1.2.10)
and the Heisenberg un ertainty relation (1.2.8) to be well dened, then the square normal-
izable wave fun tion ' must be not only in H, but also in D(P ), D(Q), D(H ), D(jP j),
D(jQj), D(jH j).
Hen e, a subdomain  of H where all of the physi al quantities (1.2.7)-(1.2.10) an be
omputed is needed. Clearly,  should be stable under the a tion of P , Q and H . It seems
that the best andidate for  is given by the interse tion of the domains of all the powers
of P , Q and H ,
1
\
= D(An) : (1.2.11)
n=0
A=P;Q;H
The spa e in Eq. (1.2.11) is the maximal invariant subspa e of the algebra of the harmoni
os illator. On , all physi al quantities su h as expe tation values and un ertainties an be
1.2 Harmoni Os illator 7

omputed. Algebrai relations su h as the Heisenberg ommutation relation are well dened
on . In parti ular, the Heisenberg un ertainty prin iple is well dened on .
The spe trum of the Hamiltonian of the harmoni os illator is dis rete, and its eigen-
ve tors are square normalizable (a tually, they are elements of ). This means that, as far
as the eigenve tors of H are on erned, there is no need to go beyond the Hilbert spa e
H. However, the spe trum of the position and momentum observables is ontinuous, and
oin ides with the set of real numbers. Following the pres ription (1.2.1), we asso iate an
eigenve tor jpi to ea h of the elements p of the ( ontinuous) spe trum of P ,
P j pi = pj pi ; 1 < p < +1 : (1.2.12)
A ording to (1.2.2), a wave fun tion an be expanded by these eigenkets,
Z +1
'= dp jpihpj'i : (1.2.13)
1
Obviously, the kets jpi are not in the Hilbert spa e|a larger linear spa e is needed to
a ommodate them. It happens that those jpi a quire meaning as antilinear fun tionals
over the spa e . That is, jpi 2  , where  represents the set of antilinear fun tionals
over the spa e . Similar onsiderations hold for the position operator Q,
Qjxi = xjxi ; jxi 2  ; 1 < x < +1 : (1.2.14)
Z +1
'= dx jxihxj'i ; ' 2  : (1.2.15)
1
In this way, the Gelfand triplet
  H   (1.2.16)
of the harmoni os illator arises in a natural way. The Hilbert spa e H omes from the
requirement that the wave fun tions must be square normalizable. The subspa e  is the
set of physi al wave fun tions, i.e., the wave fun tions on whi h any expe tation value,
any un ertainty and any ommutator an be omputed. The dual spa e  ontains the
eigenkets asso iated to the ontinuous spe trum of the observables of the algebra. These
eigenkets are dened as fun tionals over the spa e , and they an be used to expand any
' 2  as in Eq. (1.2.13) or Eq. (1.2.15).

These ideas will be elaborated in Chapter 3, where the Rigged Hilbert Spa e of the
harmoni os illator is onstru ted.2 The harmoni os illator will be studied from a dierent
point of view to that used in textbooks on Quantum Me hani s. The standard approa h
to the harmoni os illator is to start out with the (position) S hrodinger realization of the
algebra of operators, i.e., one takes for granted the well-known dierential expressions for
Q, P and H . From these expressions one derives, for instan e, the Heisenberg ommutation
2
Chapter 3 is a substantial improvement of and an extension to Ref. [29℄.
8 1 Introdu tion

relation. One an also derive that the Hamiltonian has a ountable number of eigenval-
ues whose orresponding eigenve tors are given by the Hermite polynomials. The above
pres riptions of Dira 's formalism are also assumed, although it is not mentioned that the
Hilbert spa e mathemati s annot in orporate them. In this dissertation, we shall not take
for granted the position realization of the algebra of the harmoni os illator, but rather
derive this realization from algebrai assumptions. We shall just assume some algebrai
relations to be fullled by the operators P , Q and H , namely the Heisenberg ommutation
relation
[P; Q℄ = P Q QP = i~I ; (1.2.17)
and the expression of H in terms of P and Q,
1 2 ! 2 2
H= P + Q : (1.2.18)
2 2
We shall make an additional essential assumption: the existen e of an eigenve tor 0 of the
energy operator,
H0 = 1=2 ~! 0 : (1.2.19)
From this algebrai starting point, we shall derive rst that H possesses a ountable number
of eigenvalues ~w(n + 1=2), n = 0; 1; 2; : : :, orresponding to some eigenve tors n . The
linear spa e spanned by the n will be alled . This linear spa e will be equipped with
two dierent topologies: the usual Hilbert spa e topology, whi h generates the Hilbert spa e
H from , and a stronger, nu lear topology, whi h generates the spa e  from . This
nu lear topology will make the elements of the algebra ontinuous operators. On e  is
onstru ted, we shall onstru t  and therewith the Rigged Hilbert Spa e of the harmoni
os illator:
  H   : (1.2.20)
The eigenkets jpi and jxi will be ontinuous antilinear fun tionals over , i.e., they will be
elements of  . The eigenket equations Qjxi = xjxi, P jpi = pjpi will nd their mathemat-
i al setting as fun tional equations over . The statement of the Gelfand-Maurin Theorem
will be given, whi h will guarantee the existen e of a omplete set of generalized eigenve -
tors of the position and momentum operators. It will be shown that this theorem is the
mathemati al statement that justies the heuristi Dira basis ve tor expansions (1.2.13)
and (1.2.15). We shall derive the S hrodinger representation of the harmoni os illator. In
this representation, the standard expressions for P , Q and H in terms of dierential oper-
ators will be obtained. The position realization of the RHS (1.2.20) by spa es of fun tions
and distributions will be also obtained. The spa e  will be realized by the S hwartz spa e
S (R ), and  will be realized by the spa e of tempered distributions S (R ) . Thus the
position realization of the RHS (1.2.20) will read
S (R )  L 2 (R )  S (R ) : (1.2.21)
The eigenve tors n of H will be realized by the Hermite polynomials.
1.3 A Rigged Hilbert Spa e of the Square Barrier Potential 9

Therefore, we shall give a proper mathemati al framework for the operations that the
physi s of the harmoni os illator seems to need, and we will throw light onto the problem
of how the S hrodinger realization of the algebra of operators of the harmoni os illator an
be singled out. The important point is that this realization, whi h is introdu ed ad ho in
the literature, an be derived from proper algebrai assumptions within the RHS formalism.

1.3 A Rigged Hilbert Spa e of the Square Barrier Po-

tential
The fundamental equation of Quantum Me hani s is the S hrodinger equation. Thus, show-
ing that the RHS ontains the mathemati al methods needed by Quantum Me hani s is tan-
tamount to showing that the natural framework for the solutions of the S hrodinger equation
is the RHS. To show this, we shall use the example of the square barrier potential [25, 26℄.
The time dependent S hrodinger equation reads as

i~ '(t) = H'(t) ; (1.3.1)
t
where H denotes the Hamiltonian, and '(t) denotes the value of the wave fun tion ' at
time t. Dira 's formalism solves this equation formally as follows: for ea h energy E in the
spe trum Sp(H ) of the Hamiltonian, there exists a ket jE i that is an eigenve tor of H ,
H jE i = E jE i ; E 2 Sp(H ) : (1.3.2)
These eigenkets form a omplete basis system that expands any wave fun tion ' as
Z Z
'= dE jE ihE j'i  dE '(E )jE i : (1.3.3)

The time dependent solution of Eq. (1.3.1) is obtained by Fourier-transforming the time
independent solution of Eq. (1.3.3),
Z
'(t) = dE e iEt=~ '(E ) : (1.3.4)

If the spe trum of the Hamiltonian has a ontinuous part, and if the energy E belongs
to this ontinuous part of the spe trum, then the orresponding eigenket jE i that solves
Eq. (1.3.2) is not square integrable, i.e., jE i is not an element of the Hilbert spa e. As in
the ase of the harmoni os illator, the Hilbert spa e annot handle these non-normalizable
kets, whereas the RHS formalism an.
The main short oming of the RHS formalism is that it does not provide a pres ription
to onstru t the spa es ,  , or the eigenkets jE i. The general statement of the Nu lear
Spe tral Theorem [5℄ just assures the existen e of the eigenkets jE i, and assumes the spa es
,  to be given beforehand. Therefore, a systemati pro edure to onstru t the RHS
10 1 Introdu tion

of S hrodinger Hamiltonians is needed. The fourth hapter of this dissertation provides

this systemati pro edure [25, 26℄. In order to make things lear, we shall illustrate this
pro edure through the square barrier potential, although the same method an be applied
to a large lass of potentials.

The pro edure to onstru t the RHS of the square barrier potential is as follows. First, we
write down the time independent S hrodinger equation in the radial position representation:
hrjH jE i  hhrjE i = E hrjE i ; (1.3.5)
where h is the following S hrodinger dierential operator:
~2 d2
h + V (r ) ; (1.3.6)
2m dr2
and 8
< 0 0<r<a
V (r ) = V0 a < r < b (1.3.7)
:
0 b<r<1
is the square barrier potential. By applying the Sturm-Liouville theory (Weyl theory) [30℄
to the time independent S hrodinger equation (1.3.5), we obtain a domain D(H ) on whi h
the dierential operator h is self-adjoint. This domain indu es the self-adjoint Hamiltonian
H . The next step is to ompute the Green fun tions (i.e., the resolvent) of H , the spe trum
of H (whi h in our example is [0; 1)), and the unitary operator U that diagonalizes H . The
operator U allows us to obtain the energy representation of the Hilbert spa e and the dire t
integral de omposition indu ed by the Hamiltonian. The dire t integral de omposition is
not enough for the purposes of Quantum Me hani s. The reasons why the dire t integral
de omposition (i.e., the Hilbert spa e methods) is not enough for the purposes of Quantum
Me hani s are the same as in the ase of the harmoni os illator:
(i) The expe tation values and the un ertainties of the Hamiltonian in any physi al wave
fun tion should be well dened.
(ii) Algebrai operations should be well dened. Sin e D(H ) is not stable under the
a tion of H , the powers of H are not well dened on all of the elements of H. Hen e, a
subdomain  in luded in D(H ) that remains stable under the a tion of H and all of its
powers is needed,
H n :  7 !  ; n = 0; 1; 2; : : : (1.3.8)
(Obviously, if Eq. (1.3.8) holds, then the expe tation values and the un ertainties of H in
any ' of  are well dened.)
(iii) For ea h E 2 Sp(H ), there is a Dira ket jE i su h that the eigenequation (1.3.2)
and the Dira basis ve tor expansion (1.3.3) hold. The kets jE i are dened in terms of the
1.3 A Rigged Hilbert Spa e of the Square Barrier Potential 11

eigenfun tions hrjE i of (1.3.5) as

jE i :  7! C Z 1
' 7 ! h'jE i := '(r)hrjE idr : (1.3.9)
0

After realizing that the Hilbert spa e is not suÆ ient to a ount for (i)-(iii), we onstru t
the RHS
  H   (1.3.10)
of the square barrier potential. This RHS a ounts for (i)-(iii), be ause of the following
reasons:
(1) The spa e  is stable under the a tion of H (this will give (1.3.8)). On the spa e
, all algebrai operations involving the Hamiltonian H are well dened. In parti ular, the
expe tation values of the Hamiltonian in any element of  are well dened. The elements
of  are represented by well-behaved fun tions, in ontrast to the elements of the Hilbert
spa e, whi h are represented by sets of equivalent fun tions that an vary arbitrarily on any
set of zero Lebesgue measure. As in the example of the harmoni os illator, we on lude that
not every element of the Hilbert spa e an be a physi ally a eptable wave fun tion|only
the elements of  fulll all the onditions to be a wave fun tion.
(2) The ket jE i, as dened by (1.3.9), is a well-dened antilinear fun tional on , i.e.,
jE i 2 . In the energy representation, jE i a ts as the antilinear S hwartz delta fun tional.
Moreover, jE i is an eigenve tor of H as in Eq. (1.3.2). To see this, we have to re all that in
RHS language, Eq. (1.3.2) means that
hH'jE i = E h'jE i ; 8' 2  : (1.3.11)
The a tion of H an be extended to the kets jE i in  as follows:
h'jH jE i = hH'jE i ; 8' 2  : (1.3.12)
Be ause H is ontinuous on , the operator H  is a uniquely dened extension of H . Using
the denition (1.3.12), we rewrite Eq. (1.3.11) as
h'jH jE i = E h'jE i ; 8' 2  : (1.3.13)
Omitting the arbitrary ' in this equation leads to
H  jE i = E jE i ; (1.3.14)
whi h is the same as Eq. (1.3.2). (Note that in Eq. (1.3.14) we have denoted the a tion of
the Hamiltonian on the ket jE i by H  and not just by H . We shall use this notation in
order to stress that the Hamiltonian is a ting on ve tors that lie outside the Hilbert spa e.)
(3) Any element of  an be expanded in terms of the eigenkets jE i as in Eq. (1.3.3).
From (1)-(3) it follows that, when ontinuous spe trum is present, the natural framework
for the solutions of the S hrodinger equation is the Rigged Hilbert Spa e rather than just
the Hilbert spa e.
12 1 Introdu tion

1.4 S attering o the Square Barrier Potential

The above pro edure to onstru t RHSs of S hrodinger Hamiltonians also shows that the
RHS an in orporate boundary onditions imposed upon the S hrodinger equation:

S hrodinger equation
+ !   H  .
boundary onditions

The Hilbert spa e H is needed to in orporate the requirement that the wave fun tions are
square integrable. Moreover, H singles out the s alar produ t used to ompute probability
amplitudes. The spa e  is needed to in orporate the Dira kets asso iated with the
eigenfun tions of the time independent S hrodinger equation subje t to boundary onditions.
The spa e  is needed to in orporate the wave fun tions on whi h the Dira kets a t as
ontinuous antilinear fun tionals and for whi h all the algebrai operations and all the
expe tation values are well dened.
We are now going to see that the RHS formalism is also able to in orporate the boundary
onditions of a s attering system. In essen e, the RHS an a ommodate the Lippmann-
S hwinger equation.3 To illustrate this, we shall use the example of s attering o the square
barrier potential.
Loosely speaking, we send a beam of prepared initial in-states 'in towards the square
barrier potential. After the ollision takes pla e, 'in be omes 'out . We then measure the
probability to nd a nal out-state out . The amplitude of this probability is given by
( out ; 'out ) =( out ; S'in ) ; (1.4.1)
where S is the S -matrix. The anoni al understanding is that the initial in-state 'in and
the nal out-state out are asymptoti forms of the so- alled in-state '+ and out-state
in the remote past and in the distant future, respe tively. In terms of these, the probability
amplitude (1.4.1) reads
( ; '+ ) : (1.4.2)
The asymptoti states 'in and out are related to the \exa t" states '+ and by the
Mller operators,

+ 'in = '+ ; (1.4.3a)

out = : (1.4.3b)
It is ustomary to split up the (total) Hamiltonian H into the free Hamiltonian H0 and the
potential V ,
H = H0 + V : (1.4.4)
For a mathemati al approa h to the Lippmann-S hwinger equation in terms of RHSs of Hardy fun tions
3

see Ref. [31℄.

1.4 S attering o the Square Barrier Potential 13

The potential V is interpreted as the intera tion between the omponents of the initial
prepared states, for instan e, the intera tion between the in-going beam and the target.
The initial in-state 'in and the nal out-state out evolve under the in uen e of the free
Hamiltonian H0 , whereas the in-state '+ and the out-state evolve under the in uen e
of the (total) Hamiltonian H .
Therefore, the dynami s of a s attering system is governed by the S hrodinger equation
subje t to ertain boundary onditions. These boundary onditions spe ify what is \in"
and what is \out." The Lippmann-S hwinger equation for the in- and out-kets jE  i has
those \in" and \out" boundary onditions built into it,4

jE i = jE i + E H1  i V jE i : (1.4.5)
0

Eq. (1.4.5) is an integral equation, and is equivalent to the S hrodinger equation

H  jE  i = E jE  i (1.4.6)
subje t to ertain boundary onditions. The most important of these boundary onditions
is built into the \innitesimal imaginary parts" i, whi h hara terize what is \in" (+i)
and what is \out" ( i). We then say that the jE  i are eigenve tors of the Hamiltonian
whose orresponding eigenvalues have an \innitesimal imaginary part."
Needless to say, the Lippmann-S hwinger kets jE  i are, mathemati ally speaking, de-
ned as antilinear fun tionals. The in-ket jE + i a ts on the in-states '+ , while the out-ket
jE i a ts on the out-states . Sin e the eigenvalues of the kets jE i have an \innitesimal
imaginary part," the wave fun tions h'+ jE + i and h jE i should have meaning not only
for real energies, but also for energies with an \innitesimal imaginary part." Mathemat-
i ally this means that the wave fun tions h'+jE + i and h jE i should be the boundary
values of analyti fun tions of the ( omplex) variable E . The analyti al properties satised
by the in-ket jE +i (or, equivalently, by the wave fun tion h'+ jE +i) are dierent to those
satised by the out-ket jE i (or, equivalently, by the wave fun tion h jE i). In in orpo-
rating these two dierent types of boundary onditions into the RHS framework, we will
end up onstru ting two dierent RHSs. One RHS orresponds to the in-states '+ ,
  H   ; (1.4.7)
while the other RHS orresponds to the out-states ,
+  H  + : (1.4.8)
The Lippmann-S hwinger kets belong to the dual spa es of these RHSs,
jE i 2  : (1.4.9)
4
In Eq. (1.4.5), the symbol jE i denotes an eigenket of the free Hamiltonian H0 , not the eigenket of the
total Hamiltonian of Eq. (1.3.9).
14 1 Introdu tion

The wave fun tions '+ are usually alled in-states, whereas the wave fun tions
are alled out-states. O asionally, we shall all the observables (or out-observables),
be ause they are determined by the registration apparatus. In order to grasp the meaning
of this terminology, let us onsider the matrix element ( ; '+ ). This s alar produ t is the
amplitude of the probability to observe the out-state in the in-state '+ ,
P'+! = j( ; '+ ) j 2 : (1.4.10)
Sin e is determined by the property that we want to measure, it stands to reason that
we all it observable and denote it by a spe i symbol. In order to stress the distin tion
between states and observables, the probability (1.4.10) may be written as
P'+! = Tr(P W'+ ) ; (1.4.11)
where Tr stands for tra e and
W'+  j'+ih'+ j ; (1.4.12)
P  j ih j : (1.4.13)

The Lippmann-S hwinger equation will be studied in Chapter 5 within the example of the
square barrier potential. We shall rst write Eq. (1.4.5) in the radial position representation,

hrjE i = hrjE i + hrj E H1  i V jE i : (1.4.14)

0

Next, we shall obtain the Lippmann-S hwinger eigenfun tions hrjE i. The ontinuation of
these eigenfun tions to omplex values of the energy, that we denote by hrj(E  i) i, will
be used to dene the a tion of the Lippmann-S hwinger kets:
Z 1
h j i := lim
'+ E+ !0
dr h'+ jrihrj(E + i)+ i ; '+ 2  ; (1.4.15a)
Z0 1
h jE i := lim
!0
dr h jrihrj(E i) i ; 2 + : (1.4.15b)
0
This denition needs a omment. The a tion of the Lippmann-S hwinger kets is dened
as the limits in Eq. (1.4.15) in order to keep tra k of the i boundary onditions. The
i boundary onditions just mean that we are approa hing the ut (i.e., the spe trum of
H ) either from above (+i) or from below ( i). Therefore, the a tion of the Lippmann-
S hwinger kets jE i should be viewed as the limit of the integrals in Eq. (1.4.15) when 
tends to 0.
The onditions under whi h the ket (1.4.15a) is well dened are in general dierent to
those under whi h (1.4.15b) is well dened. Sin e these onditions determine the spa e of
wave fun tions on whi h the kets a t, the spa e  on whi h the in-ket jE +i a ts is dierent
1.5 The Gamow Ve tors of the Square Barrier Potential Resonan es 15

from the spa e + on whi h the out-ket jE i a ts. Although the pre ise form of the spa es
 will not be given, we shall provide a list of ne essary onditions that must be satised
by the elements of  . For the sake of deniteness, we shall assume sometimes that those
spa es are, in the energy representation, subspa es of spa es of Hardy lass (see also [31℄).
On e the Lippmann-S hwinger kets are onstru ted, the omplex basis ve tor expansions
of the states '+ and of the observables follow:
Z 1
'+ = dE jE + ih+ E j'+i ; (1.4.16a)
Z0 1
= dE jE ih E j i : (1.4.16b)
0

We will also onstru t the Mller operators and the S-matrix, and express the matrix element
(1.4.2) in terms of the in- and out-Lippmann-S hwinger kets,
Z 1
( ; '+ ) = dE h jE iS (E )h+E j'+i : (1.4.17)
0

This expression will be used later to derive the omplex basis ve tor expansion generated
by the Gamow ve tors.
We remark that the RHS (1.3.10) was alled a RHS of the square barrier potential and
not the RHS of the square barrier potential, be ause dierent boundary onditions upon
the S hrodinger equation yield dierent RHSs for the same potential. The spa e  of
Eq. (1.3.10) is neither + nor  , be ause  in orporates neither the \in" nor the \out"
boundary onditions of the s attering o the square barrier potential [32℄.

1.5 The Gamow Ve tors of the Square Barrier Poten-

tial Resonan es
The Gamow ve tors are the state ve tors of resonan es. They are dened as eigenve tors
of the Hamiltonian with a omplex eigenvalue. The des ription of the Gamow ve tors, im-
possible in the Hilbert spa e, an be a omplished within the RHS formulation of Quantum
Me hani s.
Experimentally, resonan es often appear as peaks in the ross se tion whose shape resem-
ble the well-known Breit-Wigner distribution. The Breit-Wigner distribution has two har-
a teristi parameters: the energy ER at whi h the distribution rea hes its maximum, and
its width R at half-maximum. The inverse of R is the lifetime of the de aying state [33℄.
The peak of the ross se tion with Breit-Wigner shape is related to a rst-order pole of the
S -matrix in the energy representation S (E ) at the omplex number zR = ER i R =2. The
theoreti al expression of the ross se tion in terms of S (E ) ts the shape of the experimental
ross se tion in the neighborhood of ER . This is why the rst-order pole of the S -matrix is
often taken as the theoreti al denition of a resonan e.
16 1 Introdu tion

Although a resonan e has a nite lifetime, it is otherwise assigned all the properties
that are also attributed to stable parti les, like angular momentum, harge, spin, parity and
other parti le labels. For example, onsider [34℄ the bombardment of stable Pb206 nu lei by
a beam of  parti les whose energy is peaked around 5.4 MeV. The ross se tion for +Pb206
s attering has an in redibly sharp resonan e whose width is of the order of 10 18 eV. For
times (after the +Pb206 s attering has taken pla e) mu h less than 138 days, there will
be nu lei in the target that have all the hemi al and physi al properties asso iated with
the atomi numbers Z = 84, A = 210, and we all these nu lei Po210 . The probability
to nd Po210 is not stationary, however, but de reases exponentially with a hara teristi
de ay time of 138 days. For times short ompared to 138 days, Po210 is to all intends an
atomi nu leus. In fa t, we in lude it (and the rest of unstable nu lei) in the periodi table
of elements along with the stable nu lei.
In parti le physi s the situation is the same ( f. for instan e [35℄). Unstable parti les are
listed along with the stable ones in the Parti le Data Table [36℄ and attributed values for the
mass, the spin and the width (or lifetime). Thus, stable parti les dier from the unstable
ones by the value of their width, whi h is zero in the ase of stable parti les and dierent from
zero in the ase of unstable ones. Hen e, phenomenologi ally, unstable parti les are not less
fundamental than the stable ones, whi h are, a ording to urrent experimental eviden e,
only the proton, the ele tron, the photon, the neutrinos and possibly the graviton.
Theoreti ally, stable and unstable parti les are usually treated on a dierent footing.
The reason is that an unstable parti le, unlike a stable one, annot be des ribed within the
Hilbert spa e formalism. However, there are some theoreti al models that treat stable and
unstable parti les on the same footing. For instan e, in the eightfold way of Gell-Mann and
Ne'eman [37℄ many multiplets ontain both stable and unstable parti les|no fundamental
distin tion between stable and unstable parti les is made.
Be ause resonan es are parti les with a nite lifetime|not just peaks in the ross
se tion|a state ve tor des ription for resonan es is needed. The Gamow ve tors are the
natural state ve tors of resonan es [27℄. The des ription of resonan es by Gamow ve tors
allows us to interpret them as autonomous experimentally de aying physi al systems.
The energy eigenfun tion with omplex eigenvalue was originally introdu ed by Gamow
in his paper on -de ay of atomi nu lei [38℄, and used thereafter by a number of authors
(see for example, Refs. [39, 40, 41, 42, 43℄ and referen es therein). The real part of the
omplex eigenvalue is asso iated with the energy of the resonan e, and the imaginary part is
asso iated with the inverse of the lifetime. The Gamow eigenfun tions have an exponentially
de aying time evolution, in a ordan e with the exponential law observed in  de ay of
radioa tive nu lei [44, 45, 46, 47℄. The Gamow eigenfun tions are obtained as solutions of
the S hrodinger equation subje t to the purely outgoing boundary ondition. This ondition
was introdu ed by Siegert [48℄.
Gamow's treatment is merely heuristi though, and it annot be made rigorous in the
Hilbert spa e theory, be ause self-adjoint operators on a Hilbert spa e an only have real
eigenvalues. Re all however that Dira 's bra-ket formulation of Quantum Me hani s was
also heuristi and without mathemati al justi ation until the RHS formulation of Quantum
1.5 The Gamow Ve tors of the Square Barrier Potential Resonan es 17

Me hani s was suggested [8, 9, 10℄. During the past few years, it has be ome lear that
the RHS mathemati s also asserts the legitima y of Gamow's proposition ( f. Refs. [19, 20,
21, 22℄ and referen es therein). In RHS language, the Gamow ve tors are eigenve tors of
the dual extension of the self-adjoint Hamiltonian. This extension an surely have omplex
eigenvalues.5
A omplementary approa h to resonan es started with Breit and Wigner, who des ribed
a resonan e by means of the Breit-Wigner distribution [50℄. (Curiously enough, this distri-
bution had been independently introdu ed ve years earlier by Fo k [51℄.) Now, if a ve tor
is to obey the exponential de ay law and also to orrespond to the Breit-Wigner distribu-
tion, then this distribution must be nonzero over the full energy real line (see Ref. [52℄ and
referen es therein). Be ause the spe trum of the Hamiltonian is bounded from below, say
[0; 1), the Breit-Wigner distribution then has to be dened also at energies that do not
belong to the physi al spe trum [52℄. This seems to imply that the exponential de ay law
is in ompatible with the Breit-Wigner distribution, be ause the Breit-Wigner distribution
leads to the exponential law only when is dened over the full energy real line ( 1; 1)
rather than just over the physi al spe trum [0; 1). However, it has been shown that even
though the spe trum of the Hamiltonian is [0; 1), the Breit-Wigner distribution an be
dened on the full energy real line by means of RHSs of Hardy fun tions [20℄, and hen e the
Breit-Wigner distribution yields the exponential law. The essential ingredient to do so is the
so- alled van Winter's theorem [53℄. This theorem allows us to pie e together the physi al
spe trum, whi h oin ides with [0; 1), and the support of the Breit-Wigner distribution,
whi h oin ides with ( 1; 1).
Thus, there are two ways of des ribing a resonan e: the Gamow ve tors, whi h are
eigensolutions of the S hrodinger equation subje t to a purely outgoing boundary ondition,
and the Breit-Wigner distribution, whi h arises from the resonan e pole of the S -matrix. It
is the major goal of this dissertation to show that the energy representation of the Gamow
ve tors is given by the Breit-Wigner distribution. The square barrier potential will be used
to illustrate this point.

The Gamow ve tors of the square barrier potential will be onstru ted in Chapter 6.
The Gamow eigenkets will be dened as the solutions of a homogeneous integral equation
of the Lippmann-S hwinger type. If we denote the Gamow ket asso iated to the omplex
energy zR = ER i R =2 by jzR i, then jzR i satises the following integral equation:
jzR i = z H1 + i0 V jzR i : (1.5.1)
R 0
This equation was introdu ed (in a dierent language) by A. Mondragon6 et al. in Ref. [40℄.
It is well known that the poles of a s attering system ome in pairs, i.e., if zR = ER i R =2
5
Eigenve tors of the dual extension of self-adjoint operators with omplex eigenvalues in the RHS were
systemati ally studied for the rst time in the redu tion of SO(2,1) with respe t to its non ompa t sub-
group [49℄.
6
I am indebted to Prof. Alfonso Mondragon for his areful and patient explanations on Eq. (1.5.1).
18 1 Introdu tion

is a pole of the S -matrix, then zR = ER + i R =2 is also a pole of the S -matrix. The Gamow
ve tor asso iated to the pole zR is denoted by jzR+ i, and satises the following integral
equation:
jzR+i = z H1 i0 V jzR+i : (1.5.2)
R 0
In Chapter 6, we will solve the integral equations (1.5.1) and (1.5.2) in the radial position
representation. In this representation, these integral equations are equivalent to the time
independent S hrodinger equation subje t to a purely outgoing boundary ondition. The
resonan e spe trum is then singled out by this purely outgoing boundary ondition. As
we shall see, this is the same resonan e spe trum as that dened by the poles of the S -
matrix [27℄. The Gamow kets will be shown to be generalized eigenve tors of the Hamiltonian
with omplex eigenvalues:
H jzR i = zR jzR i ; jzR i 2 + ; (1.5.3a)
H  jzR+ i = zR jzR+ i ; jzR+ i 2  : (1.5.3b)
Next, we shall ompute the energy representation of these Gamow ve tors. We shall onsider
two energy representations. One energy representation will be asso iated to the physi al
spe trum, whi h is [0; 1) in our example. The other energy representation will be asso iated
to the support of the Breit-Wigner distribution, whi h is ( 1; 1). We will show that the
[0; 1)-energy representation of the Gamow ve tors is the omplex delta fun tion, and that
its ( 1; 1)-energy representation is given by the Breit-Wigner distribution.
On e the Gamow kets are onstru ted, we shall see that their time evolution is governed
by a semigroup [20℄. More pre isely, we shall see that the time evolution of jzR i an be
dened only for positive values of time, whereas the time evolution of jzR + i an be dened
only for negative values of time:
e iH  t=~j zR i = e izR t=~ jzR i = e iERt=~ e Rt=(2~) jzR i ; for t > 0 only ; (1.5.4a)
e iH  t=~ jzR + i = e izR t=~ jz  + i = e iER t=~ e R t=(2~) jz  + i ;
R R for t < 0 only : (1.5.4b)
Therefore, the Gamow ve tors that we shall onstru t have all the properties that are
demanded from a resonan e state:
1. They are eigenve tors of the (dual extension of the self-adjoint) Hamiltonian with
omplex eigenvalues. These eigenvalues are also poles of the S -matrix.
2. They orrespond to the Breit-Wigner amplitude in the ( 1; +1)-energy representa-
tion.
3. Their time evolution is governed by a semigroup, and obeys the exponential de ay
law.
The Gamow ve tors will be used also as basis ve tors. The expansion generated by
the Gamow ve tors will be alled the omplex basis ve tor expansion. We shall see that
1.6 Time Reversal 19

the Gamow ve tors do not form a omplete basis system. An additional set of Dira kets
orresponding to the energies that lie in the negative real axis of the se ond sheet of the
Riemann surfa e will be added to omplete them. As we shall see, the expansion of an
in-state '+ 2  reads
Z 1 1
X
'+ = jE iS (E )h+E j'+idE 2i rn jzn ih+zn j'+i ; (1.5.5)
0 n=0
where zn = En i n =2 represents the n-th resonan e energy of the square barrier potential,
and rn represents the residue of the S -matrix S (E ) at zn . In Eq. (1.5.5), the innite sum
ontains the resonan es ontribution, whereas the integral is asso iated to the ba kground.
As we said above, the Gamow ve tors have a semigroup time evolution. This semigroup
time evolution expresses the time asymmetry built into a de aying pro ess. Some authors
su h as Fonda et al. [52℄, Cohen-Tannoudji et al. [54℄, or Goldberger and Watson [55℄ have
alled this time asymmetry the irreversibility of a de aying pro ess. In re ent years, many
authors using various languages have laimed that time asymmetry is a feature of the time
evolution of any losed quantum systems (not just of a resonan e pro ess). For instan e,
Gell-Mann and Hartle have introdu ed the time asymmetry of losed quantum systems in
terms of de oheren e histories [56℄. Haag uses the on ept of event [57℄. Bohm, Antoniou,
and Kielanowski use the preparation-registration arrow of time [58℄. Although we shall not
dis uss time asymmetry in this dissertation, we would like to mention that, for this author,
the time asymmetry of a losed quantum system is built into the propagators (for more on
this see Ref. [32℄, where the arrow of time of Quantum Ele trodynami s is dis ussed).

1.6 Time Reversal

We shall also study how the time asymmetry of the resonan es behaves under the a tion of
the time reversal operator [59℄. In order to do it, we shall apply the time reversal operator
to the Gamow ve tors. Essentially, we will show in Chapter 7 that the so- alled \growing"
Gamow ve tor is really the time reversed of the so- alled \de aying" Gamow ve tor [60℄.
We shall also study more exoti possibilities, whi h are based on the work by Wigner.
When onstru ting proje tive representations of the Poin are group extended by time re-
versal and parity, Wigner [61, 62℄ found that there are four possibilities. Three of these possi-
bilities imply a doubling of the spa e supporting the representation. Later on, J. F. Cari~nena
and M. Santander7 studied the proje tive representations of the Galilei group extended by
time inversion and parity [63℄. They also found four possibilities for the ase with mass.
As in the relativisti ase, the standard ase does not yield a doubling of the spa e that
supports the representation, whereas the other three possibilities do yield a doubling.
Based on the work by Wigner [61, 62℄, Bohm has tried to nd a meaning to the doubling
of spa es [64℄. In Chapter 7, we shall onstru t this doubling expli itly for one of the
non-standard time reversal operators in the nonrelativisti domain.
7
I thank Professor M. Santander for making me aware of his paper with Professor J. F. Cari~nena and
for his explanations on it.
20 1 Introdu tion

1.7 Synopsis
The organization of this dissertation is as follows:
In Chapter 2, we review the mathemati al methods of the Rigged Hilbert Spa e. The
algebrai stru tures (linear spa es), the topologi al stru tures (topologi al spa es), and their
ombinations (linear topologi al spa es) are introdu ed in a pedestrian way. The ountably
Hilbert spa es, whi h are the lass of linear topologi al spa es almost ex lusively used in
Quantum Me hani s, are studied in more detail. At the end of Chapter 2, the Hilbert spa e
mathemati al methods used in this dissertation are presented.
In Chapter 3, we onstru t the RHS of the harmoni os illator.8 This system is stud-
ied from a dierent point of view to that used in Quantum Me hani s textbooks. Instead
of assuming that the position and momentum operators are given by the multipli ation
and derivative operators, we shall make three simple algebrai assumptions: the Heisen-
berg ommutation relation, the expression of the Hamiltonian in terms of the position
and momentum operators, and the existen e of an eigenve tor of the Hamiltonian. From
these algebrai assumptions, we shall onstru t the RHS of the harmoni os illator and the
S hrodinger representation of the algebra of the harmoni os illator.
In Chapter 4, we onstru t a RHS of the square barrier Hamiltonian by means of the
Sturm-Liouville theory. This theory provides the dire t integral de omposition of the Hilbert
spa e. From this dire t integral de omposition, we shall onstru t the RHS.
In Chapter 5, we turn to the des ription of the Lippmann-S hwinger equation within the
RHS formalism. First, the Lippmann-S hwinger eigenfun tions will be omputed. We shall
dene the Lippmann-S hwinger eigenkets in terms of the Lippmann-S hwinger eigenfun -
tions and see that they are dened on dierent spa es of wave fun tions. The Lippmann-
S hwinger kets will be used as basis ve tors to expand the wave fun tions. As well, the
Mller operators and the S -matrix are expli itly onstru ted.
In Chapter 6, we onstru t the Gamow ve tors of the square barrier resonan es. First,
we ompute the resonan e energies as poles of the S -matrix. The integral equation of
A. Mondragon et al. for the Gamow ve tors will be translated into the RHS language. The
Gamow eigenfun tions in the position representation are obtained as the solutions of the
time independent S hrodinger equation subje t to the purely outgoing boundary ondition.
These eigensolutions will be asso iated to ertain eigenfun tionals (Gamow kets). The
[0; 1)-energy representation of the Gamow eigenfun tion will be related to the omplex
delta fun tion, and the ( 1; 1)-energy representation of the Gamow eigenfun tion will
be related to the Breit-Wigner amplitude. The semigroup time evolution of the Gamow
ve tors will also be omputed. The Gamow ve tors will be used as basis ve tors. We
shall see that the Gamow ve tors do not form a omplete basis|an additional set of kets
needs to be added in order to obtain a omplete basis. The time asymmetry of the purely
outgoing boundary ondition will be dis losed. To nish the hapter, we shall elaborate on
the exponential de ay law of the Gamow ve tors.
In Chapter 7, we study the behavior of resonan es under the time reversal operation. We
8
This hapter is a substantial improvement of and an extension to Ref. [29℄.
1.7 Synopsis 21

shall study the standard time reversal operator and also a non-standard one, whi h yields
a doubling of the RHS.
Chapter 2
Mathemati al Framework of
Quantum Me hani s
In this hapter, we review the mathemati al methods of the Rigged Hilbert Spa e. The
algebrai stru tures (linear spa es), the topologi al stru tures (topologi al spa es), and their
ombinations (linear topologi al spa es) are introdu ed in a pedestrian way. The ountably
Hilbert spa es, whi h are the lass of linear topologi al spa es almost ex lusively used in
Quantum Me hani s, are studied in more detail. At the end of this hapter, the Hilbert
spa e mathemati al methods used in this dissertation are presented.

They rushed down the street together, digging every-

thing in the early way they had, whi h later be ame so mu h
sadder and per eptive and blank. But then they dan ed down
the streets like dingledodies, and I shambled after as I've been
doing all my life after people who interest me, be ause the
only people for me are the mad ones, the ones who are mad to
live, mad to talk, mad to be saved, desirous of everything at
the same time, the ones who never yawn or say a ommonpla e
thing, but burn, burn, burn like fabulous yellow roman an-
dles exploding like spiders a ross the stars and in the middle
you see the blue enterlight pop and everybody goes
\Awww!".

Ja k Keroua , On the road

23
2.1 Linear Spa es 25

2.1 Linear Spa es

2.1.1 Introdu tion
There are some major prin iples in Quantum Me hani s that seem to ome from experi-
mental data. Among them, there are the linear superposition prin iple and the probabilisti
nature of Quantum Me hani s. These two prin iples suggest that the mathemati al ideal-
ization of Quantum Me hani s should in lude a linear spa e with a s alar produ t (
;
)
dened on it. Then ( ; (
;
)) will be our primary mathemati al obje t.

2.1.2 Linear Spa es and S alar Produ t

A linear spa e  is a set of elements '; ; ; : : : whi h is assigned an algebrai stru ture that
is a generalization of ertain aspe ts of the three-dimensional real spa e R 3 . The elements,
also alled ve tors, are dened to obey rules whi h are well-known properties of ve tors in
R 3 . The ve tor spa es whi h we use are in general not three-dimensional, but an have any
dimension N , often innite, and are dened, in general, over the omplex numbers C rather
than over the real numbers R . There are two algebrai operations, the addition of ve tors
and the multipli ation of a ve tor by a s alar. The rules for these operations that dene the
ve tor spa e are similar to those in R 3 .
Denition A linear spa e (also alled ve tor spa e)  over the omplex numbers C is a
set of elements ', , ; : : : for whi h the sum ' + of any two elements '; and the
multipli ation by a omplex number  2 C ,  , are dened and are elements of , and
have the following properties
(VS1) ' + = + ' ; 8'; 2  ; (2.1.1)
(VS2) ( + ) + ' =  + ( + ') ; 8; ; ' 2  ; (2.1.2)
(VS3) There exists a 0 2  su h that 0 + ' = ' ; 8' 2  ; (2.1.3)
(VS4) 8 ' 2  there exists 2  su h that ' + = 0 (we write  ') ; (2.1.4)
(VS5) ()' = (') ; 8;  2 C ; 8' 2  ; (2.1.5)
(VS6) ( + )' = ' + ' ; 8;  2 C ; 8' 2  ; (2.1.6)
(VS7) (' + ) = ' +  ; 8 2 C ; 8'; 2  ; (2.1.7)
(VS8) 1' = ' ; 8' 2  : (2.1.8)
From these, it follows that the zero element is unique and that, for ea h ' in , the
element ' is unique; moreover, 0' = 0 and ( 1)' = ' for all ' in  and 0 = 0 for all
 in C . A linear spa e over the eld of real numbers an be des ribed in exa tly the same
way with the word \real" substituted for the word \ omplex." The spa es that we shall use
in Quantum Me hani s will have additional properties besides (2.1.1)-(2.1.8).
A subset S in a linear spa e  is alled a subspa e of  if S is a linear spa e under
the same denitions of the operations of addition and multipli ation by a number inherited
from , i.e., if it follows from '; 2 S and  2 C that ' 2 S and ' + 2 S .
26 2 Mathemati al Framework of Quantum Me hani s

An expression of the form 1 '1 + 2 '2 +


+ n 'n, where the 's are in C and the ''s
in , is alled a linear ombination of the ve tors '1 ; '2 ; : : : ; 'n. The ve tors '1 ; '2 ; : : : ; 'n
are said to be linearly dependent if there exist numbers 1 ; 2 ; : : : ; n, not all zero, for whi h
1 '1 + 2 '2 +


+ n 'n = 0. If the equation 1 '1 + 2 '2 +


+ n 'n = 0 holds only for
1 = 2 =


= n = 0, then the ve tors '1 ; '2 ; : : : ; 'n are alled linearly independent. A
spa e  is said to be nite dimensional or, more pre isely, n-dimensional if there are n and
not more than n linearly independent ve tors in . If the number of linearly independent
ve tors in  is arbitrarily large, then  is said to be innite-dimensional. Every system of
n linearly independent ve tors in an n-dimensional spa e  is alled a basis for .
If '1 ; '2 ; : : : ; 'n is a basis for an n-dimensional spa e  and ' is an arbitrary ve tor in
, then '; '1; '2; : : : ; 'n are linearly dependent, so that
' + 1 '1 + 2 '2 +


+ n 'n = 0 ; (2.1.9)
for some ; 1 ; 2 ; : : : ; n not all zero. Then  6= 0, for otherwise we should have
1 '1 + 2 '2 +


+ n 'n = 0 ; (2.1.10)
where 1 ; 2 ; : : : ; n are not all zero, whi h ontradi ts the supposition that the ve tors
'1 ; : : : ; 'n are linearly independent. But, if  6= 0, it follows from (2.1.9) that
' = 1 '1 + 2 '2 +


+ n 'n ; (2.1.11)
where i = i =. This representation of the element ' is unique. Thus, every ve tor ' in an
n-dimensional spa e  an be uniquely represented in the form (2.1.11), where '1 ; : : : ; 'n is
a basis for . The numbers 1 ; : : : ; n are alled the oordinates of the ve tor ' relative to the
basis '1 ; : : : ; 'n. Noti e that when the ve tors are added, their orresponding oordinates
relative to a xed basis are added and, when a ve tor is multiplied by any number, all the
oordinates are multiplied by that number.
Clearly the ve tors ~a; ~b; : : : in the three-dimensional spa e R 3 fulll the relations (2.1.1)-
(2.1.8). The set of omplex innitely dierentiable ontinuous fun tions whi h vanish rapidly
at innity ( alled the S hwartz spa e) also fullls these relations. One often says that the
abstra t ve tor spa e stru ture dened by the above rules is realized by other mathemati al
obje ts, if these obje ts appear to us more \real" than the \abstra t" ve tors. Thus if
one feels more familiar with fun tions one may prefer the \realization" of  by a spa e of
fun tions over the spa e  itself.
In physi s, the abstra t mathemati al obje ts are realized by obje ts with a physi al
interpretation. Thus, a physi ist's realization of a linear spa e is not by other more familiar
or more interesting mathemati al obje ts, but by physi al obje ts. In parti ular, in quantum
physi s, the elements of the spa e  will be the mathemati al images of pure physi al states
whi h will be alled state ve tors. Thus, a ve tor stru ture is \realized" by a on rete spa e
whose elements are interpreted as the physi al states of a quantum system.
For the purposes of Quantum Me hani s, a linear spa e is a set with very little math-
emati al stru ture. We will equip it with another stru ture by dening a s alar produ t.
This notion is again a generalization of the dot produ t in R 3 .
2.1 Linear Spa es 27

Denition A linear spa e is alled a s alar produ t spa e (or Eu lidean spa e or pre-Hilbert
spa e) if for ea h pair of ve tors '; 2  we an dene a omplex number ('; ) satisfying
the following properties:
(SP1) ('; ) = ( ; ') 8' ; 2  (the overline denotes omplex onjugation) ; (2.1.12)
(SP2) (';  1 +  2 ) = ('; 1 ) +  ('; 2 ) ; 8 '; 1 ; 2 2  ; 8 ;  2 C ; (2.1.13)
(SP3) ('; ')  0 ; and ('; ') = 0 i ' = 0 : (2.1.14)

This fun tion is alled a s alar produ t and ('; ) is alled the s alar produ t of the
elements ' and .
The usual s alar produ t in R 3 ; (~a; ~b) = ~a
~b learly fullls the onditions (2.1.12)-(2.1.14)
with all numbers being real instead of omplex.
As in R 3 , one alls two ve tors ' and orthogonal if
('; ) = 0 : (2.1.15)
With the s alar produ t dened by (2.1.12)-(2.1.14) one denes the norm k'k of a ve tor
' by p
k'k = + ('; ') : (2.1.16)
The norm of a ve tor is an extension of the notion of length of a ve tor in R 3 . For any
ve tor dierent from the zero ve tor one an always dene a ve tor ^ = =k k, whi h
has the property k ^k = 1 and whi h is alled a normalized ve tor.
Sometimes one needs in a ve tor spa e a more general notion than the s alar produ t,
the bilinear Hermitian form.
Denition A omplex-valued fun tion h('; ) of two ve tor arguments is a Hermitian form
if it satises
h('; ) = h( ; ') ; (2.1.17)
h(';  ) = h('; ) ; (2.1.18)
h('1 + '2 ; ) = h('1 ; ) + h('2 ; ) : (2.1.19)
If in addition h satises
h('; ')  0 (2.1.20)
for every ve tor ', then h is said to be a positive Hermitian form . A positive Hermitian form
is alled positive denite if from h('; ') = 0 it follows that ' = 0. Thus a Hermitian form
fullls (2.1.12) and (2.1.13), but not the ondition (2.1.14) for a s alar produ t. However, a
positive denite Hermitian form is a s alar produ t.
Positive Hermitian forms, whi h are not ne essarily s alar produ ts, satisfy the Cau hy-
S hwartz-Bunyakovski inequality:
jh('; )j2  h('; ')h( ; ) : (2.1.21)
28 2 Mathemati al Framework of Quantum Me hani s

If h is positive denite, equality holds i ' =  for some  2 C .

Sometimes we have dierent realizations of the same algebrai stru ture. In these ases,
the spa es are, from an algebrai point of view, the same.
Denition An isomorphism between two algebrai stru tures A and B is a one-to-one
orresponden e between the sets A and B (i.e., to every a 2 A there orresponds exa tly
one b 2 B and vi e versa: a $ b), whi h preserves the algebrai operations.
For example, two linear s alar produ t spa es  and are isomorphi if there exists a
mapping f :  ! whi h is one-to-one and onto and that fullls
f (' +  ) = f (') + f ( ) ; 8 ;  2 C ; 8' ; 2  ; (2.1.22)
('; ) = (f ('); f ( )) ; 8;'; 2 ; (2.1.23)
i.e., f preserves the sum, the multipli ation and the s alar produ t. Isomorphi s alar
produ t spa es (and in parti ular Hilbert spa es) are also alled isometri . It often happens
that two s alar produ t spa es are isomorphi as ve tor spa es, i.e., there is a one-to-one
orresponden e whi h fullls (2.1.22), but are not isomorphi as s alar produ t spa es, i.e.,
the orresponden e does not fulll (2.1.23).

2.1.3 Linear Operators

Ve tors in R 3 an be transformed into ea h other. One example is the rotation R of a ve tor
~a into a ve tor ~b = R~a. In analogy to this, one denes transformations or linear operators
on a ve tor spa e . A fun tion A; A :  ! , that maps ea h ve tor ' in a ve tor spa e
 into a ve tor 2 , A' = , is alled a linear operator if for all '; 2  and  2 C it
fullls the onditions
A(' + ) = A' + A ; (2.1.24)
A(') = A' : (2.1.25)
An operator is alled antilinear if it fullls
A(') = A' (2.1.26)
instead of (2.1.25), where  is the omplex onjugate of .
For two operators dened on the whole spa e , the operations of addition A + B ,
multipli ation by a omplex number A, and multipli ation AB , are dened in the following
way:
(A + B )' := A' + B' ; (A)' := (A') ; (AB )' := A(B') ; (2.1.27)
for all ' 2 . It is easily veried that A + B , A and AB are linear operators dened on
the whole spa e  if A and B are linear operators dened on the whole spa e . In nite
2.1 Linear Spa es 29

dimensional spa es with a topology (linear topologi al spa es dened in Se tion 2.3) there is
a large lass of operators that an be dened on the whole spa e, the ontinuous operators.
In general this is not the ase and the denition of A + B and AB is more ompli ated and
involves questions on the domains and on the ranges of the operators.
For every linear operator A dened on the whole spa e , one an dene an operator
Ay on the elements in  for whi h
(Ay ; ') := ( ; A') ; 8' 2 : (2.1.28)
The operator Ay is alled the adjoint operator of A. An operator for whi h Ay = A is alled
self-adjoint or Hermitian.1
In the general ase, an operator A need not to be dened on the whole spa e  but only
on ertain subset D(A) of .
Denition Let ; be two linear spa es and let D(A) be a subspa e of . A mapping
A : D(A)   ! is alled a linear operator if
A(' +  ) = A' + A ; 8 ;  2 C and 8 '; 2 D(A) ; (2.1.29)
and is alled an antilinear operator if
A(' +  ) = A' + A ; 8 ;  2 C and 8 '; 2 D(A) : (2.1.30)
D(A) is the domain of A and R(A) = fA' j ' 2 D(A)g  is the range of A.
Let Ai :   D(Ai ) ! (i = 1; 2) be two linear operators with domains D(Ai ). Then
A1 + A2 is a linear operator with domain D(A1 ) \ D(A2 ) dened as
(A1 + A2 )(') := A1 ' + A2 ' (2.1.31)
for every ' in D(A1 ) \ D(A2 ). In the same way, Ai is the operator dened on D(Ai) as
(Ai)(') := Ai ' (2.1.32)
for ea h ' 2 D(Ai ). The produ t of A1 and A2 is dened as
(A1 A2 )(') = A1 (A2 ') (2.1.33)
for the ve tors ' in  su h that ' is in D(A2 ) and A2 ' is in D(A1 ). With these operations
of addition and multipli ation by s alars, the set of all linear operators mapping  into
form a ve tor spa e.
1
We will usually use the term Hermitian if we do not want to distinguish between the mathemati ally
pre isely dened notions self-adjoint, essentially self-adjoint, and symmetri . We will present all these
on epts in Se tion 2.5 along with the pre ise denition of the adjoint operator.
30 2 Mathemati al Framework of Quantum Me hani s

Of spe ial interest are the zero operator, denoted 0, and the unit operator or identity
operator, denoted I , whi h are dened by
0' = 0 ; I' = ' ; (2.1.34)
for every ' 2 . Note that 0 on the left side of the rst equation is the zero operator, while
0 on the right is the zero ve tor in (2.1.3).
The denition of linear operators was inspired by the properties of transformations on
the three-dimensional spa e. Linear operators on a ve tor spa e  may be thought of
as analogous to transformations on the three-dimensional Eu lidean spa e, but they an
also have other physi al interpretations. In parti ular, in quantum physi s they represent
physi al observables.
Very important notion for quantum physi s is that of an eigenvalue and an eigenve tor
of an operator in a ve tor spa e.
Denition A nonzero ve tor 2  is alled an eigenve tor of the linear operator A if
A = with  2 C: (2.1.35)
 is alled the eigenvalue of A orresponding to the eigenve tor .
For a given operator A, there may be many (perhaps innitely many) dierent eigen-
ve tors with dierent eigenvalues. There may also be n (nite or innite) many dierent
eigenve tors with the same eigenvalue . In this ase,  is alled n-fold degenerate. In a
nite dimensional spa e every linear operator (matri es) has at least one eigenve tor. In
an innite dimensional spa e this is in general not fullled. For instan e, the operator
dierentiation i dxd dened on the Hilbert spa e L2 (R ) has no eigenve tor belonging to
L2 (R ).
If A is a Hermitian operator dened on a s alar produ t spa e, then eigenve tors and
eigenvalues have the following properties:
1. All eigenvalues are real.
2. If '1 and '2 are eigenve tors of A with eigenvalues 1 and 2 , respe tively, and if
1 6= 2 , then '1 and '2 are orthogonal to ea h other, i.e., ('1 ; '2 ) = 0.
In quantum physi s, an operator represents an observable of a physi al system. Its eigen-
values then represent the numbers whi h are obtained in a measurement of this observable.
In the nite dimensional ase (and in some spe ial innite dimensional ases), the eigen-
ve tors of a Hermitian operator an be used to expand any state (wave fun tion) in terms of
them. In the innite dimensional ase, this expansion will need the on ept of a generalized
eigenve tor and a generalized eigenvalue (see Se tion 3.5).
Denition An operator B is alled the inverse of an operator A if BA = AB = I . The
operator B is denoted by A 1.
2.1 Linear Spa es 31

A linear operator U is alled a unitary operator if U y U = UU y = I .

Be ause of the denition of the inverse operator, one an dene a unitary operator also
by the ondition U y = U 1 . It is worthwhile noting that not every operator has an inverse.
Another important notion is that of the ommutator of two operators.
Denition Let A and B be two operators dened on . The ommutator of A and B is
dened by
[A; B ℄  AB BA or [A; B ℄' = AB' AB' ; 8' 2 : (2.1.36)
A and B are said to ommute if
[A; B ℄  AB BA = 0 or AB' AB' = 0 ; 8' 2 : (2.1.37)

The olle tion of linear operators dened on the whole linear spa e forms a new algebrai
stru ture, where the algebrai operations of sum of two operators, produ t of a number with
an operator and produ t of two operators are dened by (2.1.27). This algebrai stru ture is
alled an asso iative algebra. An asso iative algebra an also be dened abstra tly without
any referen e to linear operators by the following denition:
Denition A set A is an (asso iative) algebra with unit element i
(A1) A is a ve tor spa e.
(A2) For every pair A; B 2 A, a produ t AB 2 A is dened su h that
(AB )C = A(BC ) ; (2.1.38)
A(B + C ) = AB + AC ; (2.1.39)
(A + B )C = AC + BC ; (2.1.40)
(A)B = A(B ) = AB : (2.1.41)
(A3) There exists an element I 2 A su h that
IA = AI = A ; 8A 2 A : (2.1.42)

A subset A1 of an algebra A is alled a subalgebra of A if A1 is an algebra with the same

denitions of the operations of addition, multipli ation by a number, and multipli ation as
inherited from A. That is, if from A; B 2 A1 and  2 C , it follows that A + B 2 A1 ,
A 2 A1 , and AB 2 A1 .
(A4) An algebra A is alled a -algebra if we have on the algebra a y-operation (involu-
tion), A ! Ay , that has the following dening properties:
(A + B )y = Ay + B y ; (2.1.43)
32 2 Mathemati al Framework of Quantum Me hani s

(AB )y = B y Ay ; (2.1.44)
(Ay )y = A ; (2.1.45)
Iy = I ; (2.1.46)
where A; B 2 A and ;  2 C .
From the denition (2.1.27) of the sum and the produ t of two operators and the produ t
of an operator with a number, and from the denition (2.1.28) of the adjoint operator, one
an see that the set of linear operators fullls all the axioms (A1)-(A4) of a -algebra.
Thus the set of linear operators dened on the whole ve tor spa e  forms a -algebra.
A subalgebra of this algebra is alled an operator -algebra. It an be shown that in a
ertain sense every -algebra an be realized as an operator -algebra in a s alar-produ t
spa e (generalization of the Gelfand-Naimark-Segal re onstru tion theorem). In Quantum
Me hani s, physi al systems are assumed to be des ribed by operator algebras.
A set X1 ; X2 ; : : : ; Xn of elements of A is alled a set of generators, and A is said to be
generated by the Xi (i = 1; 2; : : : ; n) i ea h element of A an be written as
n
X n
X
A = I + i X i + ij Xi Xj + : : : ; (2.1.47)
i=1 i;j =1

where ; i ; ij ; : : : 2 C .
Dening algebrai relations are relations among the generators
P (Xi) = 0 ; (2.1.48)
where P (Xi) is a polynomial with omplex oeÆ ients of the n variables Xi . An element
B 2 A, X X
B = bI + bi Xi + bij Xi Xj + : : : ; (2.1.49)
where b; bi ; : : : 2 C , is equal to the element A in (2.1.47) i (2.1.49) an be brought into
the form (2.1.47) with the same oeÆ ients ; i ; ij ; : : : by the use of the dening relations
(2.1.48).

2.1.4 Antilinear Fun tionals

In the previous se tion, we have introdu ed the on ept of an eigenve tor of an operator in
a ve tor spa e. In Quantum Me hani s, some of the eigenve tors that we need are antilinear
mappings from a spa e of states into the omplex numbers. In this se tion, we dene them
and explain some of their basi properties.
Denition Let  be a omplex linear spa e. A fun tional (or a fun tion) on  is a mapping
F from the spa e  into the omplex numbers C ; F :  ! C . (If  is a real spa e then the
mapping is into the real numbers R .)
2.1 Linear Spa es 33

If F satises
F (' +  ) = F (') + F ( ) ; 8'; 2  ; 8;  2 C ; (2.1.50)
then F is alled an antilinear fun tional. If F satises
F (' +  ) = F (') + F ( ) ; 8'; 2  ; 8;  2 C ; (2.1.51)
then F is alled a linear fun tional. (If  is a real spa e there is no distin tion between
linear and antilinear fun tionals.) A linear or antilinear fun tional is thus a spe ial ase of
a linear or antilinear operator between two linear spa es (see (2.1.29) and (2.1.30)) if the
spa e is the spa e of omplex numbers C .
A fun tional is also the analog of a omplex-valued fun tion F (x) of a real variable x
varying on R , F : R ! C , only now the variable is not a real number x 2 R but a ve tor
' 2 . We will onsider here antilinear fun tionals rather than linear fun tionals (in the
mathemati al literature one usually onsiders linear fun tionals).
An example of an antilinear fun tional on a s alar produ t spa e is given by
F : ! C
' ! F (') = ('; ) ; (2.1.52)
where is a xed element in  and ('; ) is the s alar produ t of with ', where ' varies
over . Be ause of this example and be ause in the general ase we want to onsider a
fun tional to be a generalization of the s alar produ t, one uses for the antilinear fun tional
F (') the Dira 's bra-ket symbol (see referen e [1℄)
F ( ')  h'j F i : (2.1.53)
We shall use the two notations on urrently. Dira kets will be given a mathemati al
meaning as antilinear fun tionals (whi h in addition are ontinuous, notion that will be
dened in Se tion 2.2).
Any two antilinear fun tionals F1 and F2 on a linear spa e  may be added and multiplied
by numbers a ording to
(F1 + F2 )(') = F1 (') + F2 (') ; ;  2 C ; (2.1.54)
or, using the notation (2.1.53),
h'jF1 + F2i = h'jF1i +  h'jF2i : (2.1.55)
The fun tional F1 + F2 dened by (2.1.54) is again an antilinear fun tional over . Thus,
the set of antilinear fun tionals on a ve tor spa e  is a linear spa e itself. This spa e
is alled the onjugate spa e or dual spa e (more pre isely, the algebrai dual or algebrai
onjugate spa e) of the spa e  and is denoted by alg .
34 2 Mathemati al Framework of Quantum Me hani s

2.2 Topologi al Spa es

2.2.1 Introdu tion
When we dened the ve tor spa e we took a set  of elements (whi h we alled '; ; : : :)
and endowed this set with an algebrai stru ture by dening two operations, addition of two
elements and multipli ation of ' 2  by an  2 C . We demanded that these operations
fullled ertain rules (see Se tion 2.1.2). The resulting stru tured set was alled a linear
spa e. (We thereafter dened another algebrai operation on , the s alar produ t|see
Se tion 2.1.2|and alled  a s alar produ t spa e). Now we take a set, whi h we all again
 (but whi h is not yet a ve tor spa e), and endow it with another stru ture, a topologi al
stru ture. The resulting stru tured set will be alled a topologi al spa e.
The topology on a spa e provides us with a way of phrasing su h on epts of nearness,
ontinuity, onvergen e, ompletion, et that we are familiar with for the spa e of real
numbers. We shall start with the on ept of \open set," whi h is a generalization of the
notion of open set in R . However, there are several equivalent ways of dening a topology
(via open sets, or losed sets, or neighborhoods,...) and for more restri ted ases one an
dene the topology in more spe i ways, e.g. by onvergen e of sequen es. We want to start
in the most general setting and then to spe ialize the on epts without mu h dis ussions in
order to arrive rapidly at the parti ular spa es that we need in Quantum Me hani s.

2.2.2 Open Sets and Neighborhoods

Denition Let  be a set and let P () = fS j S  g be the olle tion of all subsets of
. A subset  of P () is alled a topology of  if the following onditions are fullled:
(O1) ; 2  and  2  (; is the empty set) : (2.2.1)
(O2) The union of arbitrarily many elements of  is an element of  : (2.2.2)
(O3) The interse tion of a nite number of elements of  is in  : (2.2.3)
The pair (;  ) is alled a topologi al spa e and the elements of  are alled open sets.

With the given denition of topology we an dene the onvergen e of sequen es of

n=0 of the set , whi h is a generalization of the
elements (points) '1 ; '2 ; : : : ; 'n; : : :  f'n g1
notion of onvergen e for real numbers.

Denition A sequen e of points '1; '2; : : : ; 'n; : : : 2  is said to onverge to ' 2  if for
every open set O with ' 2 O there exists a positive integer N = N (O) su h that 'n 2 O
for all n > N (O).

This denition means that beginning from a large enough N the elements of the sequen e
are as lose to ' as we desire.
2.2 Topologi al Spa es 35

Example Let  be the set of real numbers R . The meaning of the onvergen e of the
sequen e fyn g, written as limn!1 yn = x (or yn ! x), is the following: the open sets O
ontaining x in the previous denition are given by
U (x)  fy 2 R j jx y j <  ;  > 0g : (2.2.4)
Then, a ording to the denition of onvergen e, for every U (x) there exists an N su h that
for all n > N , yn 2 U (x), i.e., jx ynj < . This is the well-known denition of onvergen e
of a sequen e of real numbers. The open sets U (x) in (2.2.4) are alled neighborhoods of x.
The generalization of the on ept of a neighborhood to a topologi al spa e is the following:
Denition If  is a topologi al spa e and ' 2 , a neighborhood (hereafter abbreviated
nhood) of ' is a set U whi h ontains an open set O ontaining ' (that is, ' 2 O  U ).
The olle tion U' of all nhoods of ' is the nhood system at '. Nhoods need not be open
but we shall only use systems of open nhoods, i.e., U 2 U' whi h also are in  .
One an easily see that a sequen e f'ng1 n=0 onverges to an element ' i ea h nhood
of ' ontains every point of the sequen e whose index is larger than some positive integer
depending on the given nhood. Thus, it is a generalization of the notion of onvergen e for
real numbers.
Denition A subset S   is said to be a topologi al subspa e of  if S is given the
topology
S = fS \ O j O 2  g : (2.2.5)
To des ribe a given topology, we do not need to know the whole olle tion of open sets:
it is enough to know a proper sub olle tion.
Denition A base B of a topology  on  is a sub olle tion of  su h that every open
set O is a union of some open sets in B, i.e., ea h O 2  an be given as
[
O = B ; B 2 B : (2.2.6)


Thus, given a base B we generate all the open sets (and therefore we des ribe the
topology ompletely) taking all possible unions of sets in B. In mu h the same way that a
base des ribes the whole olle tion of open sets, a nhood system an be ompletely des ribed
by a nhood base.
Denition A nhood base (or a system of basi nhoods) at ' in the topologi al spa e  is a
sub olle tion B' taken from the nhood system U' , having the property that ea h U 2 U'
ontains some V 2 B' . On e a nhood base at ' has been hosen (there are many to hoose
from, all produ ing the same nhood system at ') its elements are alled basi nhoods.
36 2 Mathemati al Framework of Quantum Me hani s

Obviously, the nhood system at ' is itself a nhood base at '. In the topologi al spa e
of the real numbers R , the open set (2.2.4) is a nhood of x and fU (x);  > 0g is a system
of basi nhoods at x. But also
Bx = fU1=m (x) j m = 1; 2; : : :g (2.2.7)
is a system of basi nhoods at the point x 2 R whi h onsists of a ountable number of
nhoods. For the real numbers we know that a sequen e fyn g is already onvergent to a
point x, yn ! x, i for every m 2 N , jx yn j < m1 for all positive integers n greater than a
ertain natural number N = N (m) depending on m. Thus, the ountable system of nhoods
(2.2.7) denes already the onvergen e in R and denes the topology ompletely.
In general, a topologi al spa e does not need to have a ountable system of (basi )
nhoods at ea h of its points. But there are many spa es whi h have this property.
Denition A spa e  is said to satisfy the rst axiom of ountability if it has a ountable
system of basi nhoods at ea h of its points. We also alled these spa es rst ountable.
Most of the spa es in whi h we are interested satisfy the rst axiom of ountability.
The most important feature of this kind of topologies is that we an des ribe them om-
pletely using onvergen e on sequen es (that is, spe ifying whi h sequen es onverge to
whi h points).
Using the above denitions one an prove that in a topologi al spa e  a system of
nhoods U' at a point ' has the properties:
(N1) If U 2 U' ; then ' 2 U ; (2.2.8)
(N2) If U; V 2 U' ; then U \ V 2 U' ; (2.2.9)
(N3) If U 2 U' ; then there is a V 2 U' su h that U 2 U for ea h 2 V ; (2.2.10)
(N4) If U 2 U' ; and U  V then V 2 U' ; (2.2.11)
and furthermore;
(N5) O   is open i O ontains a nhood of ea h of its points : (2.2.12)
Conversely, if in a set  a olle tion U' of subsets of  is assigned to ea h ' 2  so as
to satisfy (2.2.8)-(2.2.11) and if we dene \open" using (2.2.12), the result is a topology on
 (i.e., a olle tion of subsets of  satisfying (2.2.1)-(2.2.3)) in whi h U' is a nhood system
at ', for ea h ' 2 . Therefore, whenever nhoods have been assigned to ea h point in a set,
satisfying the properties (2.2.8)-(2.2.11), the topology is ompletely spe ied. This means
that we an equivalently des ribe a topology (that is, to des ribe the on epts of nearness,
ontinuity, onvergen e,...) using as starting point the open sets or the nhood systems at
ea h point. Obviously, one an also des ribe the topology ompletely assigning a system of
basi nhoods to ea h point.
A given set  an be equipped with various topologies. Dierent topologies on the
same set lead to dierent meanings of nearness, ontinuity, onvergen e,... If  is equipped
with two dierent topologies, say with 1 and 2 , and if 1  2 , then 1 is alled oarser
2.2 Topologi al Spa es 37

than 2 and the onvergen e with respe t to 1 weaker than the onvergen e with respe t
to 2 . Correspondingly, 2 is alled ner than 1 and the onvergen e with respe t to 2 is
alled stronger than with respe t to 1 . Sin e every U 2 1 is also in 2 , it follows from the
denition of onvergen e that every strongly onvergent sequen e is also weakly onvergent.
One an arrive at the same topology in a spa e (i.e., the same system of open sets)
starting from two dierent systems of nhoods. For example, in dening the natural topology
on the real line we an, on the one hand, take as nhoods the open intervals (2.2.4) with
real 's and, on the other hand, take the nhoods (2.2.7) with rational 's. As mentioned
above, both systems of nhoods des ribe the same topology. In general, we will all two
dierent systems of nhoods equivalent, if they lead to the same topology. The following
simple ondition is both ne essary and suÆ ient for the equivalen e of two given nhood
systems fU g and fV g: every nhood U ontains a nhood V , and every nhood V ontains a
nhood U .
A topology an also be des ribed in terms of losed sets. To introdu e this notion, we
rst need the following denition:
Denition Let (; ) be a topologi al spa e and let S  . ' 2  is alled an adheren e
point of S if for every U 2 U' , then U \ S =
6 ;.
In parti ular, every point of the set S is an adheren e point. There are two possibilities
for the adheren e points of a set S :
1. There exists a nhood of ' (the adheren e point) whi h ontains only a nite number
of points of S . We are not interested in this ase.
2. Every nhood of the adheren e point ' ontains an innite number of distin t points
of S . Then ' is alled a limit point of S .
A limit point ' of S may or may not belong to S . A set S is said to be losed if it ontains
all of its adheren e points. If a set S is not losed one obtains the losure S of S by adjoining
to S those of its adheren e points whi h do not already belong to it. Thus the losure S of
S is the olle tion of all adheren e points of S . The losure of any set S is losed, and S is
losed i S = S . The on epts of open and losed set are dual to ea h other. In fa t, a set
M   is losed (i.e., M = M ) i its omplement  M is open.
Denition A set D in a topologi al spa e  is alled dense in  if D = . A topologi al
spa e  is separable i  has a ountable dense subset.
The real line is separable, sin e the rational numbers are dense in R , and most of the
spa es used in Quantum Me hani s are separable.

2.2.3 Separation Axioms

The above denition of a topology is still too general. The topologies that are of importan e
in physi s satisfy more requirements. These topologies all fulll strong onditions for the
38 2 Mathemati al Framework of Quantum Me hani s

meaning of separation of two points ' and in . These onditions will allow us to
\distinguish" between two dierent points of the spa e using only the topology.
Denition A topologi al spa e  is a T0 -spa e (or, the topology on  is T0 ) if whenever
' and are distin t points in , there is an open set ontaining one and not the other.
A topologi al spa e  is a T1 -spa e if whenever ' and are distin t points in , there
is a nhood of ea h not ontaining the other.
 is said to be a T2 -spa e (also alled Hausdor) if whenever ' and are distin t points
of , there are disjoint open sets U and V in  with ' 2 U and 2 V .
Every T2 -spa e is T1 , and every T1 -spa e is T0 . In a T1 -spa e, every nite set is losed.
In a T2 -spa e, every onvergent sequen e has exa tly one (unique) limit point. For this
reason, the minimum that we will require of our topologies is that they be Hausdor spa es.
But often we will make even stronger separation demands on our spa es.
Denition A topologi al spa e  is said to be regular if whenever S is losed and ' is not
in S , then there are disjoint open sets U and V with ' 2 U and S  V .
We dene a T3 -spa e to be a regular T1 -spa e.
A topologi al spa e  is normal if whenever S and P are disjoint losed sets in  , there
are disjoint open sets U and V with S  U and P  V .
A normal T1 -spa e will be alled T4 .
Roughly speaking, in T3 - and T4 -spa es we an \distinguish" (or \separate") points
from sets and sets from sets, respe tively. Every T4 -spa e is T3 , and every T3 -spa e is T2 .
Most spa es we shall onsider will be T4 . The lass of T4 -spa es in lude all metrizable
and therefore all lo ally onvex spa es whose topology is given by a ountable number
of seminorms. These in lude ountable normed spa es, ountable Hilbert spa es, and, in
parti ular, normed and s alar produ t spa es. The denition of these kinds of spa es will be
given is Se tion 2.4.1. All the spa es that we shall use in Quantum Me hani s for the spa e
 of a rigged Hilbert spa e   H   will be ountable Hilbert spa es and therefore T4 .
2.2.4 Continuity and Homeomorphi Spa es
An important notion that depends upon the topology is the notion of a ontinuous mapping.
Intuitively, a map f is ontinuous at a given point ' if the images of the points lose to
' are lose to f ('). Thus the on ept of ontinuity is derived from that of nearness, and
therefore is given by the topology.
Denition Let (; ) and ( ;  ) be two topologi al spa es and f :  ! . Then f is
ontinuous at ' 2  i for ea h nhood V of f (') in  , there is a nhood U of ' in  su h
that f (U )  V . We say f is ontinuous on  i f is ontinuous at ea h ' 2 .
One an use the open sets to des ribe ontinuous maps on the whole spa e. A map
2.3 Linear Topologi al Spa es 39

f :  ! is ontinuous on  i the inverse image of every open set of is an open set of

, i.e., O 2  ) f 1(O) 2 .
In the ase of spa es that satisfy the rst axiom of ountability, we an use the sequen es
(that determine ompletely the topology) to des ribe the ontinuity of a fun tion at a given
point. A map f :  ! , where  and satisfy the rst axiom of ountability, is
ontinuous at a point ' i whenever 'n ! ' with respe t to  , then f ('n) ! f (') with
respe t to  .
A map is alled a topologi al map (or homeomorphism) if it is one-to-one, onto and
ontinuous and f 1 is also ontinuous. In this ase, we say that  and are homeomorphi .
Homeomorphi spa es are, from a topologi al point of view, the same. That is, there is no
topologi al property that allows us to distinguish them. Moreover, a topologi al property is
anything that is onserved under homeomorphisms.
In order to onvey the meaning of the ombination of an algebrai stru ture with a topo-
logi al stru ture, as is needed for the denition of a linear topologi al spa e in Se tion 2.3,
we have to dene the dire t produ t of two topologi al spa es.
Denition Let (1; 1 ) and (2; 2 ) be topologi al spa es. Then the topology on the dire t
produ t spa e 1  2 is generated by the base
B = fO1  O2 j O1 2 1 ; O2 2 2 g ; (2.2.13)
i.e., the (topologi al) produ t spa e is (1  2 ;  ), where  is the olle tion of arbitrary
unions of the sets that belong to B in (2.2.13).

2.3 Linear Topologi al Spa es

2.3.1 Introdu tion
A linear topologi al spa e (also alled a topologi al ve tor spa e) is a ombination of a linear
stru ture (see Se tion 2.1.2) and a topologi al stru ture (see Se tion 2.2), both introdu ed
on one and the same set . However, these stru tures are not independent of ea h other.
The linear operations, whi h are mappings on , are required to be ontinuous in order
that these two stru tures mat h ea h other. The general pro edure to onstru t topologi al
algebrai stru tures (topologi al algebras, topologi al groups, topologi al ve tor spa es) is:
1. One endows a given set  with an algebrai stru ture.
2. One endows the same set  with a topologi al stru ture.
3. One demands that the algebrai operations on  be ontinuous mappings.
The reason why one onstru ts these mathemati al stru tures is that there exist realiza-
tions of su h stru tures (with some additional properties) that are very useful in mathemati s
and in physi s. For instan e, linear topologi al spa es are realized in mathemati s by lasses
40 2 Mathemati al Framework of Quantum Me hani s

of fun tions (e.g., the S hwartz spa e). In physi s, these abstra t mathemati al entities are
used to des ribe some stru tures in nature. For example, topologi al groups (in parti ular
parameter or Lie groups) are the mathemati al image of symmetry transformations of the
registration apparatuses (dete tor) relative to the preparation apparatuses (a elerator).
Linear topologi al spa es and their algebras of linear operators provide the mathemati al
framework to des ribe the states and the observables of quantum physi s, respe tively.
For the ombination of a topologi al stru ture with the algebrai stru ture, the following
denition is an example of the general pro edure des ribed above.
Denition A set  is alled a linear topologi al spa e (l.t.s.) or a topologi al ve tor spa e
(t.v.s.) if
(LT1)  is a linear spa e : (2.3.1)
(LT2)  is a topologi al spa e : (2.3.2)
(LT3) The algebrai operations are ontinuous : (2.3.3)

Assumption (2.3.3) means that the mapping

:C  ! 
(; ') ! ' (2.3.4)
and the mapping
: ! 
('; ) ! '+ (2.3.5)
are ontinuous. The ontinuity of these operations gives a pre ise meaning to intuitive
notions su h as an \innite linear ombination" of ve tors or the limit of an innite sequen e
of ve tors. A l.t.s. is often denoted by (; ; +;
) in order to spe ify the linear and the
topologi al stru tures. We shall just denote a l.t.s. by  if no onfusion is possible.
From the ontinuity of the algebrai operations it follows that if U (0) is a nhood of the
zero element, then V = ' + U (0) (i.e., the set obtained by adding ' to all the elements of
U (0), also alled the translate of U (0) by ') is a nhood of '. In other words, the nhood
system at ' is just the family of translates by ' of members of the nhood system at 0.
Therefore, the topology of a l.t.s. an be ompletely spe ied by the system of nhoods at
the zero element.
TIf U0 is a base of nhoods at the zero element in the l.t.s. , then  is T2 if and only
if U 2U0 U = f0g, i.e., i the interse tion of the nhoods of zero is pre isely zero. We shall
always assume that the topology of a l.t.s. is T2 . Moreover, we shall onsider mostly T4
spa es. In parti ular, for all the spa es  in the Rigged Hilbert Spa e   H   we
shall hoose ex lusively T4 spa es.
The simplest example of a l.t.s. is the real line R when endowed with the usual addition
and multipli ation (whi h provide the linear algebrai stru ture) and with the topology of
2.3 Linear Topologi al Spa es 41

the absolute value dened in Se tion 2.2.2. In a similar manner, the omplex numbers C
an be also onsidered as a l.t.s.
A useful on ept in R is that of a bounded set. A set S  R is bounded if there exists an
M > 0 su h that jxj < M for all x 2 S . The generalization of this notion to an arbitrary
l.t.s. is:

Denition A subset B of a l.t.s.  is said to be bounded if for every nhood of zero U (0)
there exists a  > 0 su h that B  U (0). U (0) = f' j ' 2 U (0)g, the set obtained by
multiplying ea h element of U (0) by , is alled a multiple of U (0).

Roughly speaking, a set is bounded if every nhood of zero has a multiple that swallows
it up. By using the nhoods of R in Se tion 2.2.2, one an show that this denition agrees
with the above denition of boundedness of S  R .
It is easy to see that if 1 and 2 are two topologies on a l.t.s.  and if 1  2 , then
every set B whi h is bounded with respe t to the ner topology 2 is also bounded with
respe t to the oarser topology 1 .

2.3.2 Cau hy Sequen es

In the topologi al ve tor spa e of real numbers R , a sequen e fyn g1 n=0 is alled Cau hy if
for every  > 0 there is a positive integer N = N () (depending only on ) su h that for
all n; m  N we have jyn ym j < . This means that a sequen e is Cau hy if beginning
from a large enough N the elements of the sequen e are more and more lose to ea h
other. We an reformulate this denition in terms of nhoods of the zero element in R : a
system of nhoods at x = 0 is the olle tion (see Se tion 2.2.2) U0 = fU (0) j  > 0g where
U (0) = fy 2 R j jy j < g. Then a sequen e fyn g1 n=0 of real numbers is Cau hy i for
every nhood U (0) of 0, there is a positive integer N (U ) su h that yn ym 2 U (0) for all
n; m > N . We shall generalize this on ept to an arbitrary l.t.s.

Denition A sequen e f'ng1n=0 of elements in a l.t.s.  is alled Cau hy if for every nhood
U of the zero element there exists a natural number N = N (U ), depending only on U , su h
that 'n 'm 2 U for all n; m > N .

Every onvergent sequen e is Cau hy, but the onverse is not always true, i.e., a Cau hy
sequen e need not onverge to a point in the spa e. In the l.t.s. of the real numbers R , a
sequen e is Cau hy i it is onvergent to some (unique) real number. In the l.t.s. of rational
numbers Q this is not the ase, sin e there are Cau hy sequen es of rational numbers whi h
do not onverge to any rational number (for example, any sequen e of rational numbers
onverging to  ).

Denition A l.t.s.  is alled omplete (more pre isely, sequentially omplete) if every
Cau hy sequen e has a limit in .
42 2 Mathemati al Framework of Quantum Me hani s

This denition means that in a sequentially omplete spa e we always get to a point in
the spa e whenever we follow a sequen e of elements that be ome more and more lose to
ea h other. When the same set is endowed with two dierent topologies we usually say that
the spa e is  - omplete, if we want to emphasize whi h topology we are onsidering.
In general, we seek spa es that are omplete. This is why if a l.t.s. is not omplete
we omplete it by adjoining all the limit elements of Cau hy sequen es to it. Then, the
in omplete spa e an be viewed as a dense subspa e of its ompletion.

Denition A omplete l.t.s.  is said to be the ompletion of an in omplete l.t.s.  if there

is a map i :  !  whi h is one-to-one, linear and ontinuous with ontinuous inverse i 1
su h that i() is dense in  .

Note that the fun tion i (usually alled an embedding) is not onto (if so,  would already
be a omplete spa e). The ompletion of a spa e  is unique up to a linear homeomorphism
whi h leaves  pointwise xed. As an example, R is a ompletion of Q .
Completeness is a very important requirement in mathemati s. Without it one an-
not prove existen e theorems nor dene derivatives or integrals. In physi s, ompleteness
annot be established dire tly from physi al observation be ause ompleteness involves an
innite number of entities (Cau hy sequen es) and all physi al observations involve only a
nite number of states. Thus, it annot be \dedu ed" dire tly from experiments and only
the overall su ess of a mathemati al theory an show whi h ompletion|more pre isely,
ompletion with respe t to whi h topology|is preferable for quantum physi s.
We have given above only the denition of sequential ompleteness, whi h is suÆ ient
when the topology is fully des ribed in terms of the onvergen e of sequen es (that is, when
the topology satises the rst axiom of ountability). If the spa e is not rst ountable,
its ompletion annot be dened in terms of Cau hy sequen es. It has to be dened in
terms of nets, whi h we do not want to introdu e here. With this more general denition of
ompletion, every l.t.s. an be ompleted in the sense of the above denition, the ompletion
is unique (up to a linear homeomorphism) and the spa e an be onsidered as a dense
subspa e of its ompletion. The spa e  of the Rigged Hilbert Spa e   H   will
always be hosen to satisfy the rst axiom of ountability. Therefore, it an be ompleted
using Cau hy sequen es. The spa e  will in general not be rst ountable, and its
ompletion must be onstru ted using the general denition.
Sin e any metrizable spa e (a lass that in ludes s alar produ t spa es, normed spa es
and ountably normed spa es) is rst ountable (see Se tion 2.3.3), we an omplete it by
using Cau hy sequen es. Vaguely speaking, the ompletion is a omplished in the follow-
ing way: two Cau hy sequen es ('1 ; '2 ; '3 ; : : :) and ( 1 ; 2 ; 3 ; : : :) in  are onsidered
equivalent if beginning from a large enough term their elements are more and more lose
to ea h other. More pre isely, ('1 ; '2 ; '3 ; : : :)  ( 1 ; 2 ; 3 ; : : :) i for every U 2 U0 there
is a positive integer N (U ) su h that 'n m 2 U for all n; m > N . We denote by
[('1 ; '2 ; : : : ; 'n ; : : :)℄ the set of sequen es whi h are equivalent to ('1 ; '2 ; : : : ; 'n; : : :). On
2.3 Linear Topologi al Spa es 43

this spa e of equivalen e sequen es, one denes an algebrai addition

[('1 ; '2 ; : : : ; 'n; : : :)℄ + [( 1 ; 2 ; : : : ; n ; : : :)℄ := [('1 + 1 ; '2 + 2 ; : : : ; 'n + n ; : : :)℄ (2.3.6)
and a multipli ation by s alars
[('1 ; '2 ; : : : ; 'n; : : :)℄ := [('1 ; '2 ; : : : ; 'n; : : :)℄ : (2.3.7)
We dene the metri on this spa e as
d ([f'ng1
n=0 ℄; [f g1
n n=0 ℄) := nlim
!1
d('n ; n) ; (2.3.8)

where d is the metri on the in omplete spa e and d will be the metri on the omplete
spa e. Finally, we dene the map i :  !  as i(') = [('; '; : : : ; '; : : :)℄, i.e., the elements
' of the in omplete l.t.s.  are represented in  by the innite rows ('; '; '; : : :). In
the end, i is a 1:1, linear, ontinuous mapping and i() is dense in  . Moreover, we an
extend other algebrai operations on  to  in the same way we extended the sum and
the produ ts by s alars. For example, we an extend a s alar produ t on  to  via the
denition
([f'n g1
n=0 ℄; [f n gn=0 ℄) := nlim
1 (' ; )
!1 n n
(2.3.9)
where (
;
) is the s alar produ t on  and (
;
) is the s alar produ t on .

2.3.3 Normed, S alar Produ t and Metri Spa es

In a linear spa e , we an introdu e algebrai operations that have in prin iple nothing
to do with a topology, but that an be used to dene one. For instan e, in Se tion 2.1.2
we introdu ed a s alar produ t on a ve tor spa e. The resulting stru ture, alled s alar
produ t spa e, has signi an e even without any topologi al onsiderations. However, this
s alar produ t an be used to dene several topologies on the s alar produ t spa e (a Hilbert
spa e topology, a nu lear topology,...). Another example of an algebrai operation that an
give rise to a topology is the norm.
Denition Let  be a linear spa e. A norm k
k on  is a fun tion whi h asso iates to
ea h ' 2  a nite real number k'k fullling
(N1) k' + k  k'k + k k ; 8 '; 2  : (2.3.10)
(N2) k'k = jj k'k ; 8 ' 2 ; 8 2 C : (2.3.11)
(N3) k'k  0 ; and k'k = 0 only if ' = 0 : (2.3.12)

A linear spa e  equipped with a norm k
k is usually denoted by (; k
k), and it is
alled a normed spa e. From (2.3.10) and (2.3.11), it follows that k0k = 0.
44 2 Mathemati al Framework of Quantum Me hani s

With the use of the norm we an spe ify a system of nhoods at 0 to dene a topology,
alled the norm topology. We dene the nhoods of the zero element by
U(0) = f' j k'k < g ;  > 0 : (2.3.13)
Then the nhoods of any are dened by
U ( ) = + U (0) = f' j k' k < g : (2.3.14)
The system of nhoods at zero
U0 = fU(0) j  > 0g (2.3.15)
provides a topology for the normed spa e (; k
k). Equipped with this topology, (; k
k)
is a l.t.s. In pla e of (2.3.15), one an hoose a ountable system of nhoods at zero
W0 = fU1=m (0) j m = 1; 2; :::g : (2.3.16)
One an show that the systems of nhoods (2.3.15) and (2.3.16) are equivalent, i.e., they
yield the same topology. In parti ular, this means that every normed spa e satises the rst
axiom of ountability.
In the l.t.s. (; k
k) , we an give a meaning to the topologi al notions we have dis ussed
above ( onvergen e of sequen es, ompleteness,...). For example, a ording to the general
denition given in Se tion 2.3.2, a sequen e f'n g1 n=0 is Cau hy if for every U (0) in (2.3.13)
there exists an N = N (U ) su h that for all n; m > N , 'n 'm 2 U (0). This means
that for every  > 0 there exists a natural number N = N () su h that for all n; m > N ,
k'n 'mk < . This is the denition of Cau hy sequen es that one usually nds in tra ts
on normed spa es. As in the general ase, a normed spa e  is alled omplete if every
Cau hy sequen e onverges to an element in . If a normed spa e (; k
k) is not omplete
then it an be ompleted. A omplete normed spa e is alled a Bana h spa e.
Now, given a s alar produ t (
;
) on a linear spa e  we an dene the norm provided
by the s alar produ t as p
k'k := + ('; ') ; 8 ' 2  : (2.3.17)
It is easy to see that (2.3.17) is a well dened norm that satises the requirements (2.3.10)-
(2.3.12) if the s alar produ t satises (2.1.12)-(2.1.14). Therefore, we an make a s alar
produ t spa e  a l.t.s. by using the system of nhoods at zero (2.3.15) or (2.3.16) with k
k
dened by (2.3.17). Although a s alar produ t always des ribes a norm (through (2.3.17)),
the onverse is not always true. Therefore a s alar produ t spa e is always a normed spa e
but the onverse does not ne essarily hold.
Denition A s alar produ t spa e is alled a Hilbert spa e if it is omplete with respe t to
the topology generated by the norm given by the s alar produ t as in (2.3.17). We shall
usually denote a Hilbert spa e by H.
Thus a Hilbert spa e is the ompletion of the s alar produ t spa e of Se tion 2.1.2 with
respe t to the topology given by the system of nhoods (2.3.15) or (2.3.16) . Sin e the Hilbert
2.3 Linear Topologi al Spa es 45

spa e has been so important in mathemati s and physi s, a s alar produ t spa e (in whi h
one does not introdu e any topology) is often alled a pre-Hilbert spa e. Every pre-Hilbert
spa e be omes a Hilbert spa e when we omplete it with respe t to the topology given by
(2.3.15) with (2.3.17). It is worthwhile noting that the Hilbert spa e topology is not the
only topology for whi h one an omplete a s alar produ t spa e . In Se tion 2.4 we will
dis uss other dierent topologies that an be introdu ed on .
Denition A real-valued fun tion d, dened for ea h pair of elements ', of a set , is
alled a metri if it satises
(M1) d(' + )  d('; ) + d(; ) ; 8 '; ; 2  : (2.3.18)
(M2) d('; ) = d( ; ') ; 8 '; 2  : (2.3.19)
(M3) d('; )  0 ; d('; ') = 0 ; and d('; ) > 0 if ' 6= : (2.3.20)
A set  provided with a metri is alled a metri spa e and d('; ) is alled the distan e
between ' and .
Let V ('; ) be the set of all elements 2  su h that d('; ) < . Then, the olle tion
U' = fV ('; ) j  > 0g (2.3.21)
is a system of basi nhoods at ' that generate a topology on . Endowed with this topology,
a metri spa e is a l.t.s. A topologi al spa e is alled metrizable if its topology an be dened
by a metri d. Every metrizable spa e is rst ountable, sin e the system of nhoods
fV ('; 1=n) j n = 1; 2; : : :g (2.3.22)
is equivalent to (2.3.21). A metrizable spa e is also T4 .
The real numbers and the omplex numbers are both metrizable spa es, the metri being
given by
d(x; y ) := jx y j ; x; y 2 R (2 C ) : (2.3.23)
If we are given a norm k
k dened on a linear spa e, we an dene a metri asso iated to
it by d('; ) = k' k. Therefore, normed and s alar produ t spa es are metrizable, and
their topology as metrizable spa es oin ides with the topology dened by the norm or by
the s alar produ t.

2.3.4 Continuous Linear Operators and Continuous Antilinear

Fun tionals
Linear operators and antilinear fun tionals were dened in Se tions 2.1.3 and 2.1.4. In
quantum physi s, the operators (representing quantum me hani al observables and quantum
me hani al states) are linear and the fun tionals (representing kets or generalized states)
are antilinear. Therefore, we shall use here linear operators and antilinear fun tionals;
orresponding mathemati al statements hold for antilinear operators and linear fun tionals.
46 2 Mathemati al Framework of Quantum Me hani s

Denition Let  and be two l.t.s. A map A : D(A)   ! is alled a ontinuous

linear mapping or just a ontinuous operator i
1. A is linear ( f. Se tion 2.1.3),
2. A is ontinuous ( f. Se tion 2.2.4).

The notion of ontinuity of an operator on a l.t.s. an be lo alized at zero, in the same

way that the topology an be lo alized at zero. More pre isely, a linear mapping A from a
l.t.s.  into another l.t.s. is ontinuous on the whole spa e  i it is ontinuous at the
zero element. Therefore, a linear mapping A :  ! is ontinuous i for every nhood U of
0 2 there exists a nhood V of 0 2  su h that A(V )  U . When the l.t.s.  and are
rst ountable (that is, the topology an be des ribed in terms of onvergen e of sequen es),
then the sequential riterion for ontinuity (see Se tion 2.2.4) an also be lo alized at zero:
an operator A :  ! is ontinuous i whenever 'n ! 0 in , then A('n ) ! 0 in .
The notion of boundedness is related to the ontinuity of an operator:
Denition A linear operator A :  ! is alled bounded i it transforms every  -
bounded set B   into a  -bounded set A(B )  ( f. Se tion 2.3.1 for the denition of
a bounded set).
One an show that every ontinuous operator dened on a l.t.s. is bounded. Moreover,
if the l.t.s.  and satisfy the rst axiom of ountability, an operator A :  ! is
ontinuous if and only if it is bounded. Therefore, in all normed, ountably normed and
metrizable spa es (whi h are rst ountable) one an use the words ontinuous operator and
bounded operators inter hangeably.
Continuous operators as ompared to non- ontinuous operators have ni er properties
and are easier to handle be ause they an always be dened on the whole spa e . Even if
initially they are only dened on a dense subspa e D(A), their denition an be extended
to the whole spa e in a ontinuous manner. As an example of this extension, onsider two
rst ountable l.t.s.  and su h that is omplete. Let A : D(A)   ! be a densely
dened ontinuous operator. Then A an be uniquely extended to the whole spa e  in a
ontinuous way as follows: if ' 2  but ' is not in D(A) we an always nd a sequen e
f'ng in D(A) su h that 'n ! ' with respe t to  . Sin e f'ng is -Cau hy, fA'ng is
 -Cau hy, and fA'n g has a well dened limit in . We an dene the a tion of A on
' to be this limit
A' := nlim A' ; 8 ' 2  :
!1 n
(2.3.24)
The operator dened in (2.3.24) is well dened on the whole spa e , extends the a tion of A
on D(A) and is ontinuous. For the spa es  of the RHS   H   and for the operators
dened on them this extension will always be possible. In fa t, we shall always assume that
every ontinuous linear (as well as antilinear) operator has already been extended to the
whole spa e .
2.3 Linear Topologi al Spa es 47

If A :  ! is a ontinuous linear operator then (A),  2 C , is also a ontinuous linear

operator. If A; B are ontinuous linear operators then A + B and A
B are also ontinuous
linear operators. Thus if the generators Xi of an algebra A are given by ontinuous operators
then the whole algebra is an algebra of ontinuous operators and every A 2 A given by
(2.1.47) with a nite numbers of terms are dened on the whole spa e . (The question of
the onvergen e of innite sums of the form (2.1.47) an only be addressed after a topology
has been introdu ed on A). If one wants to do al ulations it is of great importan e to
have an algebra of ontinuous operators whi h are dened on the whole spa e, be ause then
one does not have to worry about domain questions, i.e., one does not have to answer the
question whether B' is in the domain D(A) of an operator A to al ulate AB'. Also one will
not have to deal with the absurd situation that the expe tation value ('; A'), representing
average value of a quantum me hani al observable A measured in quantum me hani al state
', is \innite" when ' is not in D(A). It would, therefore, be desirable that all quantum
me hani al observables were given by ontinuous operators on a suitable l.t.s. This means
that the mathemati al image of all quantum me hani al observables should be a set (perhaps
an algebra) of ontinuous operators on some l.t.s. Only ve tors ' of su h l.t.s. an represent
physi al states. The non- ontinuous operators should be forbidden be ause they may lead
to nonphysi al innite predi tions.
Mu h of the trouble of the Hilbert Spa e formalism omes from domain questions. Al-
ready the simplest operators of Quantum Me hani s, the operators momentum P and po-
sition Q, whi h fulll the algebrai relation P Q QP = i1 (Heisenberg ommutation
relation), annot be represented by ontinuous operators in the Hilbert spa e H. Thus the
Hilbert spa e ontains some \non-physi al states" in whi h these operators are not dened.
This is one of the reasons why we have to introdu e a ountably norm topology  in ad-
dition to the algebrai stru ture of a s alar produ t spa e. The ompletion with respe t
to this topology generates a spa e . This spa e, whi h is as subspa e of H, allows for a
representation by  - ontinuous operators that satisfy the Heisenberg ommutation relation
or similar algebrai relations (e.g., the ommutation relations of non- ompa t groups of im-
portan e in physi s). Our task is thus to nd a topology  su h that the phenomenologi al
ommutation relations of Quantum Me hani s are represented by ontinuous operators on
some spa e . For the Heisenberg ommutation relations and many other algebrai re-
lations (in luding the ommutation relations of all Lie groups) a ountably Hilbert spa e
( f. Se tion 2.4) will do the job.
The on ept and properties of a ontinuous antilinear fun tional F :  ! C follows
from the ase of a ontinuous linear mapping just hanging linearity for antilinearity and
onsidering the spa e C as a l.t.s.

Denition A map F :  ! C is alled a ontinuous antilinear fun tional if

1. F is antilinear ( f. Se tion 2.1.3),

2. F is ontinuous ( f. Se tion 2.2.4).

48 2 Mathemati al Framework of Quantum Me hani s

Dira kets, Lippmann-S hwinger kets and Gamow ve tors will be represented by ontin-
uous antilinear fun tionals.
The olle tion  of ontinuous antilinear fun tionals on a l.t.s. , i.e., the set
 = fF :  ! C j F is antilinear and  ontinuousg ; (2.3.25)
is alled the onjugate of , the topologi al dual of , or the adjoint of .
The onjugate
spa e depends, as all topologi al notions do, on the parti ular topology that has been
hosen. The onjugate spa e is a linear spa e under the usual sum of two fun tionals
and multipli ation of a fun tional by a omplex number. Sin e the elements of  must
be ontinuous, the topologi al dual is always a subspa e of the algebrai dual dened in
Se tion 2.1.4. A topology an also be assigned to  to make it a l.t.s.
Example If (; k
k) is a Bana h spa e, an antilinear fun tional F on  is ontinuous i
the quantity
kF k := sup jFk'('k)j (2.3.26)
'2; '6=0
is a nite real number. kF k is alled the norm of the antilinear fun tional F . One an
prove that (2.3.26) is indeed a well dened norm that satises (2.3.10)-(2.3.12). Moreover,
the adjoint spa e  of a Bana h spa e  is a Bana h spa e itself when we dene the norm
of a fun tional by (2.3.26). An antilinear fun tional F over a Bana h spa e is bounded i
there exists a positive onstant K su h that
jF (')j  K k'k ; 8 ' 2  : (2.3.27)
Sin e a Bana h spa e is rst ountable, F is ontinuous i F is bounded. In fa t, kF k in
(2.3.26) is the minimum of the real numbers K that satisfy (2.3.27).
The adjoint H of a Hilbert spa e H, that in parti ular is a Bana h spa e, an be
onstru ted in a similar fashion and an be endowed with the norm topology generated by
(2.3.26). On e this is done, the following important theorem holds:
Theorem (Riesz-Fre het) For every H- ontinuous antilinear fun tional F on a Hilbert
spa e H there exists a unique ve tor fF 2 H su h that
F (g ) = (g; fF ) ; 8g 2 H; (2.3.28)
and su h that kfF kH = kF kH .
The Riesz-Fre het theorem provides a one-to-one ontinuous linear mapping of H onto
H,
: H ! H
F ! fF ; (2.3.29)
2.4 Countably Hilbert Spa es 49

that preserves the norms of the spa es. Therefore, a Hilbert spa e and its adjoint are
isometri spa es ( f. Se tion 2.1.2). This is usually abbreviated as
H ' H : (2.3.30)
In general, two l.t.s.  and are alled isomorphi if there exists a one-to-one mapping
h of  onto whi h is linear and ontinuous and su h that its inverse is ontinuous. The
mapping h is alled an isomorphism.2 Thus an isomorphism is a mapping that preserves
the linear topologi al stru ture of the spa es. Two isomorphi spa es are, from a linear
topologi al point of view, the same, and are usually identied,
 ' : (2.3.31)
When  and are normed spa es, a linear mapping h of  onto is an isomorphism i
there are positive onstants K1 and K2 with
K1 k'k  kh(')k  K2 k'k : (2.3.32)
Two isomorphi metri spa es are usually alled isometri .
A ontinuous linear operator A dened on the whole of a l.t.s. ,
A :  ! ; (2.3.33)
an be extended into  by
h'jAF i := hA'jF i ; ' 2  ; F 2  : (2.3.34)
The dual extension A dened by (2.3.34) is a well dened linear operator on 
A :  !  : (2.3.35)

2.4 Countably Hilbert Spa es

2.4.1 Introdu tion
In Se tion 2.3 we studied how to ombine a linear and a topologi al stru ture. The resulting
l.t.s. stru ture is still too general for the purposes of Quantum Me hani s. We now distin-
guish a lass of l.t.s. that is of servi e in Quantum Me hani s: ountably Hilbert spa es. A
ountably Hilbert spa e is a linear spa e on whi h a ountable number of s alar produ ts is
dened, i.e., for every '; 2  there exist
('; )1 ; ('; )2 ; : : : ; ('; )p; : : : ; (2.4.1)
2
Sin e algebrai isomorphisms were also designated as isomorphisms in Se tion 2.1.2, the terms topologi al
isomorphism and topologi ally isomorphi may be used to avoid misunderstanding.
50 2 Mathemati al Framework of Quantum Me hani s

whi h fulll the dening onditions (2.1.12)-(2.1.14) of the s alar produ t. From these s alar
produ ts one an dene the norms
q
k'kp := ('; ')p ; p = 1; 2; : : : (2.4.2)
One an also dene a ountably set of arbitrary norms k'kp not ne essarily given by s alar
produ ts. In this ase, the spa e is alled ountably normed. The s alar produ ts (norms)
in a ountably Hilbert (normed) spa e must be related to ea h other. This relation makes
these norms mat h ea h other in the sense given by the following denitions:
Denition Let k
k1 and k
k2 be two norms dened on the same linear spa e . These
two norms are alled omparable if for every ' 2  there exists a onstant C > 0 su h that
k'k1  C k'k2 ; 8 ' 2  : (2.4.3)
The norm k
k1 is alled weaker than the norm k
k2 and k
k2 is alled stronger than
k
k1. Two norms are equivalent if there exist two onstants C and D su h that
k'k1  C k'k2 ; k'k2  Dk'k1 ; (2.4.4)
for every ' 2 .
Every sequen e that is Cau hy with respe t to the stronger norm is also Cau hy with
respe t to the weaker norm. If two norms are equivalent, a sequen e is Cau hy with respe t
to one of the norms i it is Cau hy with respe t to the other norm.
Denition Two n=1   whi h is
norms are alled ompatible i every sequen e f'n g1
Cau hy with respe t to both norms and whi h onverges to 0 with respe t to one of them,
also onverges to 0 with respe t to the other norm.
Let k
k1 and k
k2 be two omparable and ompatible norms on a linear spa e  su h
that k
k1 is weaker than k
k2 . We an omplete  with respe t to the norm k
k1 to
obtain a omplete normed spa e 1 . Similarly, we an omplete  with respe t to the norm
k
k2 to obtain 2. We then have
1  2   : (2.4.5)
If k
k1 and k
k2 are equivalent, then both ompletions yield the same spa e,
1 = 2   : (2.4.6)

Denition A spa e  is a ountably Hilbert spa e (or a ountably s alar produ t spa e) if
an in reasing denumerable number of s alar produ ts
('; ')1  ('; ')2 


 ('; ')p 


(2.4.7)
2.4 Countably Hilbert Spa es 51

are dened on  su h that the norms

q
k'kp := ('; ')p ; p = 0; 1; 2; : : : (2.4.8)
are omparable and ompatible. The nhoods of zero that generate the topology are given
by
Up;(0) = f' j k'kp < g ;  > 0; p = 1; 2; : : : (2.4.9)
The topology generated by (2.4.9) make  a linear topologi al spa e, i.e., the algebrai
operations are ontinuous.
Instead of (2.4.9), one an also hoose the ountable number of nhoods
1
Up;m (0) = f' j k'kp < g ; p; m = 1; 2; : : : (2.4.10)
m
It is not hard to see that the systems of nhoods (2.4.9) and (2.4.10) are equivalent. Therefore,
in a ountably s alar produ t spa e the rst axiom of ountability holds, and its topology 
is ompletely spe ied by the denition of onvergen e of sequen es. As it is easily seen, a
sequen e f'n g1 n=1 of elements in a ountably Hilbert spa e  onverges to zero with respe t
to this topology i it onverges to zero with respe t to every norm k
kp, p = 0; 1; 2; : : : In
symbols,
'n n!1 ! 0 i k'nkp n!1

! 0 ; for every p = 1; 2; 3; : : :
C
(2.4.11)
and 
'n n!1! ' i k'n 'kp n!1 ! 0 ; for every p = 1; 2; 3; : : :
C
(2.4.12)
Sin e a ountably Hilbert spa e is rst ountable, the ontinuity of the linear ombina-
tions ' +  an be equivalently stated in terms of sequen es as:
! ' then also 'n n!1

1. if 'n n!1 ! ' for every  2 C ;


C
2. if n n!1 ! ' for every ' 2  ;
!  then also n' n!1



3. if 'n n!1! ' and n n!1! then 'n + n n!1
 
!' + :
If a given system of ountable s alar produ ts (
;
)p does not fulll the inequalities
(2.4.7), it an be repla ed by a new equivalent system of s alar produ ts that has this
property. We just need to dene a new in reasing sequen e of s alar produ ts as
X p
('; ')0p := ('; ')i ; p = 1; 2; 3; : : : (2.4.13)
i=1

The systems of s alar produ ts (
;
)0p and (
;
)p yield the same topology. Therefore, the
ondition (2.4.7) does not restri t the lass of spa es onsidered.
When the sequen e of norms in (2.4.8) annot be dened in terms of s alar produ ts, we
all the spa e ountably normed. A ountably Hilbert spa e is always ountably normed,
52 2 Mathemati al Framework of Quantum Me hani s

but not vi e versa. At rst glan e, it may appear that the lass of ountably Hilbertp spa es
onstitutes a narrow lass of ountably normed spa es, be ause the norms k'kp = ('; ')p
are only spe ial ases of general ountable olle tions of norms. However, due to the fa t that
we are onsidering denumerable olle tions of norms, the dieren e is mu h less pronoun ed
than for the ase of one norm (Bana h spa e) and one s alar produ t (Hilbert spa e). Under
very mild assumptions any initial system of norms k'k0p on p
a given ountably normed spa e
an be repla ed by another system of norms k'kp = ('; ')p dened by some s alar
produ ts without altering the topology on the spa e. We will always onsider that this is
the ase.
Example An important example of ountably Hilbert spa e is the S hwartz spa e|also
alled the spa e of test fun tions. We onsider the set S (R ) of fun tions '(
) : R ! C whi h
are innitely dierentiable and the derivatives  k '(x)=xk of whi h tend to 0 as x ! 1
faster than any power of 1=jxj, for k = 0; 1; 2; : : : The norms that dene the topology are
 
 k  q '(x) 
k'k p = sup x
xq 
 ; p = 0; 1; 2; : : : (2.4.14)
k;qp

As mentioned before, we usually an nd a sequen e of s alar produ ts that generate the

same topology as the sequen e of norms do. In the ase of S (R ), these s alar produ ts are
dened by
Z 1 X  q '(x)  q (x)
('; )p = (1 + x2 )2p q dx ; p = 1; 2; : : : : (2.4.15)
1 0qp
x q x

The norms (2.4.14) and the s alar produ ts (2.4.15) lead to equivalent topologies on S (R ).
Therefore, S (R ) is a ountably Hilbert spa e.
Example The linear spa e K (a) of all innitely dierentiable fun tions '(x) that vanish
whenever jxj > a an be made a ountably normed spa e by dening the norms
 k 
 d '(x) 
k'kp := sup 
 dxk 
 ; p = 0; 1; 2; : : : (2.4.16)
k=0;1;:::;p

A ountably Hilbert spa e is always metrizable, i.e., we an dene a metri on it that

yields the original topology. In terms of the norms (2.4.8), this metri is given by
1
X 1 k' kn :
d('; ) = (2.4.17)
n=1
2n 1 + k' kn
The fun tion dened in (2.4.17) meets the onditions (2.3.18)-(2.3.20) for a metri . Thus
one an apply all the results for the well studied metri spa es to the ountably Hilbert
spa es.
2.4 Countably Hilbert Spa es 53

A ountably Hilbert (normed) spa e is alled Fre het or an F-spa e if it is omplete with
respe t to the topology generated by the sequen e of s alar produ ts (norms). To nd a
p  to be omplete, we denote
ne essary and suÆ ient ondition for a ountably Hilbert spa e
by n the ompletion of  relative to the norm k'kn = ('; ')n . Then n is a Hilbert
spa e. Sin e
k'k1  k'k2 


 k'kn 


; (2.4.18)
we have ( f. Eq. (2.4.5))
1  2  : : :  n  : : :  : (2.4.19)
One an prove that  is omplete with respe t to the topology given by the nhoods (2.4.9)
i 1
\
 = n : (2.4.20)
n=1
We shall always assume that our ountably Hilbert spa e is Fre het, i.e., it fullls (2.4.20).
In Se tion 2.2.2 we saw that dierent systems of nhoods in a topologi al spa e an lead
to equivalent topologies. The question arises whether the topology in a ountably normed
spa e is really not equivalent to the topology given by one single norm. On the one hand,
every normed spa e (; k
k) is a ountably normed spa e: one has just to hoose a ountable
system of norms k
kp, p = 1; 2; : : :, su h that k
kp is equivalent to k
k for every p. On
the other hand, given a ountably normed spa e , whose topology is given by the innite
sequen e of norms
k
k1  k
k2 


 k
kp 


; (2.4.21)
its topology is equivalent to the topology given by a single norm k
k i there is only a nite
number of non-equivalent norms in the sequen e (2.4.21). Therefore, the essential dieren e
between a normed spa e and a ountably normed spa e is that in the latter the topology is
given by an innite number of non-equivalent norms.

2.4.2 Dual Spa e of a Countably Hilbert Spa e

The dual spa e  ( f. Se tion 2.3.4) of a ountably Hilbert spa e  is the olle tion of
antilinear fun tionals on  that are ontinuous with respe t to the topology generated by
the norms (2.4.8). If we denote the adjoint of the Hilbert spa es n in (2.4.19) by n , then
these spa es form an in reasing hain
1  2  : : :  n  : : :   ; (2.4.22)
as opposed to the de reasing hain (2.4.19). Sin e a ountably Hilbert spa e is rst ount-
able, a linear fun tional F on  is ontinuous i it is bounded (see Se tion 2.3.4). One an
also see that F is bounded i there exist a positive onstant K and a norm k
kq in the
sequen e (2.4.8) su h that
F (')  K k'kq (2.4.23)
54 2 Mathemati al Framework of Quantum Me hani s

holds for every ' 2 . This means that an antilinear fun tional is ontinuous (bounded)
with respe t to the sequen e of norms (2.4.8) i it is ontinuous (bounded) with respe t to
one norm in this sequen e. Therefore, the dual spa e of a ountably Hilbert spa e an be
written as ( ompare to Eq. (2.4.20))
1
[
 = n :
 (2.4.24)
n=1

One an introdu e a topology in the linear spa e  in various ways. For instan e, one
an take as the nhoods of zero in  the sets
UW ('1 ; '2 ; : : : ; 'n ; ) = fF 2  j jF ('k )j   ; 1  k  ng : (2.4.25)
Here '1 ; '2 ; : : : ; 'n are elements of , and  is an arbitrary positive number. The topology
generated by these nhoods is alled the weak topology on the spa e  and is denoted by
W . Along with the weak topology one an onstru t the strong topology, whose nhoods of
zero are dened by  
US (B ; ) = F 2  j sup jF (')j <  ;
 (2.4.26)
'2B
where B is any bounded set in  ( f. Se tion 2.3.1), and  > 0. We denote the strong
topology by S . As the names indi ate, the strong topology is a tually stronger than the
weak topology, i.e., W  S .
We onsider, nally, the adjoint spa e  of  . In this spa e also, one an dene
dierent topologies. We shall only onsider a topology built from the strongly bounded sets
in  (that is, bounded with respe t to S ). With ea h S -bounded set B and ea h number
 > 0 we asso iate the set
 
U (B; ) = 'e 2  j sup j'e(F )j <  : (2.4.27)
F 2B

We take the olle tion of all sets U (B ; ) for a system of nhoods at zero in  . With this
topology the se ond adjoint  is isomorphi to the original ountably normed spa e ,
i.e.,  '  . A l.t.s.  for whi h  '  is alled re exive. Thus any ountably Hilbert
spa e is re exive. In parti ular, every Hilbert spa e is also re exive ( f. Eq. (2.3.30)).

2.4.3 Countably Hilbert Spa es in Quantum Me hani s

The primary stru ture that physi ists work with is a linear spa e with a (primary) s alar
produ t ('; ) dened on it and an algebra of linear operators A. The (primary) s alar
produ t onstitutes one of the most fundamental entities: j('; )j2 represents the probability
to nd the property in the state ', whi h are the quantities that are to be ompared with
the experimental data. The linear operators A 2 A represent the observables measured
in the quantum system upon onsideration. This algebrai stru ture has, in prin iple, no
topology atta hed to it. But in Quantum Me hani s we need a topologi al stru ture so
2.4 Countably Hilbert Spa es 55

that the elements of the algebra of observables are ontinuous operators and all algebrai
operations are allowed. This is the reason why we need a sequen e of s alar produ ts.
Therefore, we onsider the ase of a linear spa e in whi h, in addition to a sequen e
of s alar produ ts (2.4.7), there is also another s alar produ t
(
;
) :  7! C
' 7! ( ; ') (2.4.28)
dened on it. In prin iple the s alar produ t (2.4.28) is unrelated to the other s alar
produ ts (2.4.7) that generate a ountably Hilbert spa e topology  . To make the s alar
produ t (2.4.28) and the sequen e (2.4.7) mat h ea h other, we assume that (2.4.28) is a
 - ontinuous mapping. Thus, in addition to (2.1.12)-(2.1.14), we demand that
'n n!1! ' implies ('n; ) n!1

! ('; ) ; 8 2 :
C
(2.4.29)
We are now going to show that whenever a primary s alar produ t (2.4.28) is ontinuous
with respe t to the topology generated by a denumerable sequen e of s alar produ ts, we
an in lude this s alar produ t as the rst element of that sequen e without altering the
topology:
From (2.4.29) we an see that the linear fun tional
F : 7 ! C
' 7 ! F (') = ( ; ') (2.4.30)
and the antilinear fun tional
F : 7 ! C
' 7 ! F (') = ('; ) (2.4.31)
are - ontinuous. Sin e any ontinuous fun tional on a ountably Hilbert spa e is
bounded, there is a norm k
kq and a onstant C > 0 su h that
jF (')j  C k'kq ; jF (')j  C k'kq ; j( ; ')j  C k kq k'kq : (2.4.32)
We now dene the sequen e of s alar produ ts
('; )00 := ('; ) ; (2.4.33)
0
('; )p := C ('; )p+q 1 ; p = 1; 2; : : : (2.4.34)
The new sequen e of s alar produ ts ontains the original s alar produ t as the zeroth
element, satises
('; ')  ('; ')00  ('; ')01 


 ('; ')0p 


; (2.4.35)
and generates the same topology as the original s alar produ ts ('; ')p do.
56 2 Mathemati al Framework of Quantum Me hani s

Therefore, we an always assume that the s alar produ t (2.4.28) is already the zeroth
s alar produ t of the sequen e
('; ')0  ('; ')1 


 ('; ')p 


(2.4.36)
of s alar produ ts that will dene the topologies on . We are mostly interested in two
topologies indu ed by the s alar produ ts (2.4.36) on . The rst topology is generated by
the nhoods of zero given by
1
Um (0) = f' j k'k0 < g ; m = 1; 2; : : : (2.4.37)
m
This is the Hilbert spa e topology ( f. Se tion 2.3.3) and is denoted by H . The se ond
topology is the ountably Hilbert spa e topology  , whose nhoods of zero are given by
1
Up;m (0) = f' j k'kp < g ; p = 0; 1; 2; : : : ; m = 1; 2; : : : (2.4.38)
m
If we omplete the linear spa e with respe t to these two topologies, we obtain the hain
of spa es
   H: (2.4.39)
H is obtained by adjoining to the limit elements of H-Cau hy sequen es whereas  is
obtained by adjoining to the limit elements of  -Cau hy sequen es. The algebrai spa e
is  -dense in  and H -dense in H, and the omplete ountably Hilbert spa e  is
H -dense in H. The se ond in lusion in (2.4.39) omes from the fa t that every  -Cau hy
sequen e is also H -Cau hy be ause fUm (0)g  fUp;m (0)g (and then H   ), but not vi e
versa.
In appli ations to physi s, the s alar produ ts (2.4.36) are introdu ed in order to obtain
a topology so that all the elements of the algebra of observables are ontinuous. They
are dened in terms of the (primary) s alar produ t and the algebra of observables. For
example, the ountable number of s alar produ ts an be dened as
('; )p  ('; Ap ) ; p = 0; 1; 2; : : : ; A 2 A ; (2.4.40)
where (
;
) is the (primary) s alar produ t that des ribes the probabilities. The quantities
j('; )pj = j('; Ap )j = j('; )j have also an interpretation, namely the probability to
nd the property represented by ' in the transformed state  = Ap . Therefore, the
s alar produ ts ('; )p, and therewith the topology  and the spa e , depend upon the
parti ular system under study.

2.5 Linear Operators on Hilbert Spa es

2.5.1 Introdu tion
In Quantum Me hani s, the observables are represented by a linear operators dened on
some linear s alar produ t spa e ( ; (
;
)). The ompletion of this spa e with respe t to
2.5 Linear Operators on Hilbert Spa es 57

the Hilbert spa e topology leads to the Hilbert spa e H (see Se tion 2.4.3). Therefore, any
observable an be viewed as an operator dened on the domain of the Hilbert spa e H.
This will allow us to apply the Hilbert spa e methods to these operators. Some of these
methods will be very useful in the Rigged Hilbert Spa e theory.

2.5.2 Bounded Operators on a Hilbert Spa e

Certain lasses of bounded operators play an essential role in Quantum Me hani s: nu lear
operators are needed in the onstru tion of the nu lear Rigged Hilbert Spa e (see Se tion 2.6)
and operators with nite tra e (whi h are dened below) are to represent mixed states.
Before introdu ing the on ept of bounded operator, we need some preliminary denitions.
Denition Let H be a Hilbert spa e and M be a losed subspa e of H. The orthogonal
omplement M? of M is the set of elements in H whi h are orthogonal to every element of
M,
M? := ff 2 H j (f; g) = 0 ; 8g 2 Mg : (2.5.1)
If M is a losed subspa e of a Hilbert spa e H, then every f 2 H an be uniquely written
as f = g + g?, where g 2 M and g? 2 M?. We usually say that H is the dire t sum of the
spa es M and M?, and denote
H = M  M? : (2.5.2)
Denition A set feng1n=1  H is an orthonormal basis for H if:
1. The elements of the basis are orthonormal to ea h other,
(en ; em ) = Ænm ; n; m = 1; 2; : : : ; (2.5.3)
where Ænm is the Krone ker delta.
2. Every f 2 H an be expanded in terms of this basis as a series of the form
1
X
f= (en ; f )en ; (2.5.4)
n=1
whi h onverges in the sense of the norm of H.
In a general Hilbert spa e, an orthonormal basis need not be ountable. It an be proven
though, that a Hilbert spa e is separable i it has a ountable orthonormal basis. We shall
only onsider separable Hilbert spa es.
Example Dene l2 to be the set of sequen es fxn g1n=1 of omplex numbers whi h satisfy
P1
n=1jxnj2 < 1 with the s alar produ t
1
X
1 1
(fxn gn=1 ; fyngn=1 ) := x n yn : (2.5.5)
n=1
58 2 Mathemati al Framework of Quantum Me hani s

l2 is a separable Hilbert spa e and the set

f(1; 0; 0; : : : ; 0; : : :); (0; 1; 0; : : : ; 0; : : :); (0; 0; 1; : : : ; 0; : : :); : : : ; (0; 0; 0; : : : ; 1; : : :); : : :g
(2.5.6)
is an orthonormal basis for l2 .
Example
R1 The spa e L2 (R ; dx) is the set of omplex-valued fun tions on R whi h satisfy
1 jf (x)j dx < 1. L (R ; dx) is a Hilbert spa e under the s alar produ t
2 2
Z 1
(f; g ) = f (x) g (x) dx : (2.5.7)
1
The Hermite polynomials form an orthonormal basis for L2 (R ; dx).
We now list some denitions and results on erning bounded linear operators dened on
a Hilbert spa e H. Corresponding statements hold for operators of a Hilbert spa e H1 into
another Hilbert spa e H2 .
Denition A linear operator A dened on a Hilbert spa e H is alled bounded if there exists
a positive number K su h that
kAf k  K kf k (2.5.8)
holds for every f 2 H. This denition of bounded operator is equivalent to the denition
given in Se tion 2.3.4.
The olle tion of all bounded operators on H is denoted by L(H). The spa e L(H) is a
linear spa e under the usual sum of two operators and multipli ation of an operator by a
number. The norm of a bounded operator is dened by

kAk := sup
kAf k : (2.5.9)
f 2H;f 6=0 kf k

One an prove that (2.5.9) is a well dened norm that satises the onditions (2.3.10)-
(2.3.12). The spa e L(H) be omes a Bana h spa e when the norm of its elements is dened
by (2.5.9).
In Se tion 2.1.2 we gave a preliminary denition of the adjoint of an operator. Now we
give a more thorough denition.
Denition Let A be a bounded operator on a Hilbert spa e H. The adjoint operator Ay
of A is dened on the elements g for whi h there exists a z 2 H fullling
(Af; g ) = (f; z ) (2.5.10)
for every f 2 H. The adjoint is then dened by Ayg = z. Thus (2.5.10) an be restated as
(Af; g ) = (f; Ayg ) ; 8f 2 H ; 8g 2 D(Ay ) : (2.5.11)
2.5 Linear Operators on Hilbert Spa es 59

A bounded operator A is alled Hermitian or self-adjoint if A = Ay , i.e., if

(Af; g ) = (f; Ag ) ; 8f; g 2 H : (2.5.12)

An important lass of operators on Hilbert spa es is that of proje tions.

Denition If P 2 L(H) and P 2 = P , then P is alled a proje tion. If in addition P = P y,
then P is alled an orthogonal proje tion.
The range R(P ) of a proje tion P is always a losed subspa e on whi h P a ts like the
identity. If in addition P is orthogonal, then P a ts like the zero operator on (R (P ))?.
Conversely, given a losed subspa e M of H, we an dene a proje tion operator PM onto
M as follows: sin e M indu es on H a de omposition of the form (2.5.2), any f 2 H an
be written as f = g + g?, where g 2 M and g? 2 M?. We dene PM f = PM (g + g?) := g .
The operator PM is a well dened orthogonal proje tion. Therefore, there is a one to one
orresponden e between orthogonal proje tions and losed subspa es.
Denition An operator U on H is alled unitary if kUf k = kf k for every f 2 H. A unitary
operator satises the relations
U y U = UU y = I : (2.5.13)
Given a losed subspa e M  H, an operator U 2 L(H) is alled a partial isometry on M
if U is unitary when restri ted to M, i.e., if
U y U = PM ; (2.5.14)
where PM is the proje tion onto M.
Evidently, any unitary operator is a partial isometry on the whole of H. A unitary
operator U is always bounded and kU k = 1.
If A is a matrix on C n , then the eigenvalues of A are the omplex numbers  su h that
the determinant of I A is equal to zero. The set of su h  is alled the spe trum of A.
It an onsist of at most n points sin e det(I A) is a polynomial of degree n. If  is not
an eigenvalue, then (I A) has an inverse sin e det(I A) 6= 0. In this ase,  is in the
resolvent set of A. These notions an be extended to the ase of a linear transformation on
a Hilbert spa e.
Denition Let A 2 L(H). A omplex number  is said to be in the resolvent set, Re(A),
of A if I A is a bije tion with a bounded inverse. If  2= Re(A), then  is said to be in
the spe trum, Sp(A), of A. We distinguish two subsets of the spe trum:
1. An f 2 H whi h satises Af = f for some  2 C is alled an eigenve tor of A;  is
alled the orresponding eigenvalue. If  is an eigenvalue, then  is in the spe trum
of A. The set of all eigenvalues is alled the dis rete spe trum of A.
60 2 Mathemati al Framework of Quantum Me hani s

2. If  is not an eigenvalue and if  is not in Re(A), then  is said to be in the ontinuous

spe trum of A.

A very important lass of bounded operators is that of ompa t operators.

Denition An operator A 2 L(H) is alled ompa t (or ompletely ontinuous) if for every
bounded sequen e ffn g  H, fAfn g has a subsequen e onvergent in H.
Example The simplest example of ompa t operator is an operator A of the form
Af := (e; f )h ; (2.5.15)
where e and h are xed ve tors of unit length, and  is a xed number. This operator maps
all of H onto the one-dimensional subspa e spanned by h, and is alled a one-rank operator.
We an also dene a linear operator A as
N
X
Af := i (ei ; f ) hi (2.5.16)
i=1

for some xed olle tions of ve tors fei gNi=1 and fhi gNi=1 in H. The range of this operator is
the nite dimensional subspa e spanned by the ve tors fhi gNi=1 . The operator A in (2.5.16)
is alled a nite rank operator. Every nite rank operator is ompa t.
The spe trum Sp(A) of a ompa t operator A is a dis rete set having no limit points
ex epts perhaps  = 0. Further, any nonzero  2 Sp(A) is an eigenvalue of nite multipli ity
(i.e., the orresponding spa e of eigenve tors is nite dimensional).
A self adjoint ompa t operator, i.e., a ompa t operator A su h that (Af; g ) = (f; Ag )
for every f; g 2 H, has a parti ularly simple stru ture. If A is a ompa t self adjoint operator,
then one an hoose an orthonormal basis e1 ; e2 ; : : : in H whi h onsists of eigenve tors of
A, Aen = n en . The eigenvalues 1 ; 2 ; : : : orresponding to the eigenve tors e1 ; e2 ; : : : are
real and onverge to zero as n ! 1, i.e., limn!1 n = 0. Conversely, every operator A
whi h is dened in terms of some orthonormal basis e1 ; e2 ; : : : ; by Aen = n en , where the
n are real numbers and limn!1 n = 0, is self adjoint and ompa t.
An operator A is positive-denite if (Af; f )  0 for every ve tor f 2 H. The eigenvalues
of a positive-denite operator are either positive or equal to zero. A ompa t operator diers
from a positive-denite operator only by an isometri fa tor, i.e., the following theorem
holds:
Theorem Let A be a ompa t operator on a Hilbert spa e H. Then A has the form
A = U jAj ; (2.5.17)
where jAj is a positive-denite ompa t operator, and U is a partial isometry on the range
of jAj.
2.5 Linear Operators on Hilbert Spa es 61

Any ompa t operator an be approximated by a sum of one-rank operators (2.5.15).

Spe i ally, a ompa t operator A an be represented as the sum of a series
1
X
Af = n (en ; f ) hn : (2.5.18)
n=1

The en are the eigenve tors of the operator jAj in the de omposition (2.5.17) orresponding
to the eigenvalues n, i.e., jAjen = n en . The hn are given by hn = Uen . (In parti ular, the
en and the hn are the elements of two orthonormal basis in H, and 1 ; 2 ; : : : are positive
numbers that tend to zero as n ! 1). Conversely, every series of the form (2.5.18), in
whi h en , hn , n have the aforementioned properties, denes a ompa t operator.
The requirement that the eigenvalues n (of the operator jAj appearing in the de om-
position A = U jAj of a ompa t operator A) tend to zero is too weak. We now onsider
operators that satisfy more stringent onditions.
Denition A ompa t operator A = U jAj is alled Hilbert-S hmidt if P1n=1 2n < 1, where
the n are the eigenvalues of the operator jAj.
Therefore, an operator is P
of Hilbert-S hmidt type i admits a de omposition of the form
(2.5.18) su h that the series 1 2
n=1 n onverges. One an also see that in orderPan operator
A be of Hilbert-S hmidt type, it is ne essary and suÆ ient that the series 1 n=1 kAen k
2
onverge for at least one orthonormal basis e1 ; e2 ; : : : in H.
An even more restri tive requirement that the operator A be Hilbert-S hmidt is that it
be a nu lear operator.
Denition A ompa t operator is alled nu lear (or tra e lass) if P1n=1 n < 1, where
the n are the eigenvalues of the operator
P1 jAj appearing in the de omposition
P1 A = U jAj.
Sin e the onvergen e of the series n=1 2n follows from the onvergen e of n=1 n , every
nu lear operator is of Hilbert-S hmidt type.
It is lear that anPoperator A is nu lear i it admits a de omposition of the form (2.5.18)
su h that the series 1 n=1 n onverges. The nu lear operators will serve in the denition of
nu lear spa es ( f. Se tion 2.6), whi h are the most important lass of l.t.s. used in Quantum
Me hani s.
The tra e of an operator is a generalization of the usual notion of the sum of the diagonal
elements of a matrix. For any positive operator A 2 L(H) we dene
1
X
Tr(A) := (en ; Aen ) ; (2.5.19)
n=1

where fen g is an orthonormal basis of H. The number Tr(A) is alled the tra e of A and
is independent of the orthonormal basis hosen. When Tr(A) is nite, then A is alled an
operator with nite tra e. If A is a positive-denite ompa t operator, then A is nu lear i
62 2 Mathemati al Framework of Quantum Me hani s

A has a nite tra e. In this ase,

1
X 1
X
Tr(A) = (en ; Aen ) = n : (2.5.20)
n=1 n=1

In Quantum Me hani s, a general (mixed) state is assumed to be des ribed by a positive

operator W with nite tra e. W is usually hosen su h that Tr(W ) = 1 (if Tr(W ) 6= 1,
we just dene the equivalent normalized state W 0  W=Tr(W )). If A is a linear operator
representing a physi al observable, then the quantity Tr(AW ) is to represent the probability
to observe A in the state W .

2.5.3 Unbounded Operators on a Hilbert Spa e

Most important observables that o ur in Quantum Me hani s are represented by linear
operators that are unbounded with respe t to the Hilbert spa e topology. In this se tion
we will introdu e some of the basi denitions and theorems ne essary for dealing with this
type of operators.
An operator A is unbounded if the quantity (2.5.9) is not nite. Unbounded operators
are usually dened on some subdomain of the Hilbert spa e. We will always suppose that
this domain is dense.
In order to ompare operators that are not dened on the whole of H, we introdu e the
following denition:

Denition Let A and B be two operators dened on H. Let D(A) be the domain of A and
D(B ) the domain of B . A is said to be an extension of B if D(B )  D(A) and Af = Bf
for every f 2 D(B ). In this ase we shall write B  A. One may also all B the restri tion
of A to D(B ).
For some operators A there is a natural way of dening an extension A. One takes a
Cau hy sequen e ffn g in D(A). If the sequen e fAfn g is also Cau hy, and if one denotes
by f and g the limits of ffn g and fAfn g respe tively, it is natural to dene Af = g . Sin e
f is not ne essarily in D(A), one may dene an extension A of A by applying the above
pro edure to all Cau hy sequen es ffn g in D(A) whi h are su h that fAfn g is also Cau hy.
However, this onstru tion makes sense only if the element g is independent of the hoi e of
a parti ular Cau hy sequen e ffn g onverging to f , i.e., if whenever ffng and ffn0 g are two
Cau hy sequen es in D(A) onverging to the same limit f and fAfn g and fAfn0 g are also
Cau hy, then limn!1 Afn = limn!1 Afn0 . An operator A verifying this ondition is said
to be losable, and the extension A is alled the losure of A. An operator A is said to be
losed if A = A.
Closedness is a weaker ondition than ontinuity sin e, if an operator A on H is ontin-
uous, then
lim fn = f ; fn 2 D(A) ;
n!1
(2.5.21)
2.5 Linear Operators on Hilbert Spa es 63

implies that the sequen e fAfn g onverges, while if it is only losed, then the onvergen e
of the sequen e ffn g  D(A) does not imply the onvergen e of the sequen e fAfn g.
The spe tral notions for a bounded operator an be generalized to the unbounded ase
when the operator is losed.
Denition Let A be a losed operator on a Hilbert spa e H. A omplex number  is in the
resolvent set, Re(A), of A if I A is a bije tion from D(A) onto H with a bounded inverse.
The denitions of spe trum, dis rete spe trum and ontinuous spe trum are the same for
unbounded operators as they are for bounded operators. We will sometimes refer to the
spe trum of non losed, but losable operators. In this ase we always mean the spe trum
of the losure.
The adjoint of an unbounded operator A an be dened in a similar way to the bounded
ase whenever the domain of A is dense in H.
Denition Let A : H ! H be a linear operator (not ne essarily bounded) on a Hilbert
spa e H whose domain D(A) is a dense linear subspa e of H. The domain D(Ay) of the
adjoint operator Ay is the set of all ve tors f 2 H for whi h there exists a z 2 H fullling
(f; Ag ) = (z; g ) (2.5.22)
for every g 2 D(A). Then, by denition, Ayf = z . Sin e D(A) is dense, the ve tor z is
uniquely determined and Ay is well dened. We then write (2.5.22) as
(Ayf; g ) = (f; Ag ) ; 8g 2 D(A) ; 8f 2 D(Ay) : (2.5.23)

The adjoint operator is always losed. The relation between an unbounded operator A
and its adjoint Ay an be more ompli ated than for the bounded ase:
Denition An operator A on H is alled symmetri if D(A) is dense in H and (Af; g) =
(f; Ag ) for every f , g 2 D(A). This means that a densely dened operator is symmetri i
A  Ay . A is alled self-adjoint if D(A) is dense in H and A = Ay. A is alled essentially
self adjoint (e.s.a.) if A is self adjoint.
If A is a symmetri operator, then A is losable and A = Ayy. An e.s.a. operator has
a unique self adjoint extension that oin ides with its adjoint. Physi al observables are
assumed to be represented by e.s.a. operators.
Evidently, any self adjoint operator is e.s.a., and any e.s.a. operator is symmetri . In
fa t, an operator A (not ne essarily bounded) is
symmetri i A  A = Ayy  Ay ; (2.5.24)
e:s:a: i A  A = Ayy = Ay ; (2.5.25)
self adjoint i A = A = Ayy = Ay : (2.5.26)
64 2 Mathemati al Framework of Quantum Me hani s

The spe trum of a self-adjoint operator is always a losed subset of the real axis.
In Quantum Me hani s, the elements A of the algebra of observables A are dened on
some linear s alar produ t spa e ( ; (
;
)), and are required to fulll (A'; ) = ('; A )
for every ', 2 (i.e., they are required to be symmetri ). These operators are usually
unbounded. When this is the ase, they annot be extended to the whole Hilbert spa e
H (whi h is the ompletion of with respe t to the Hilbert spa e topology) due to the
following theorem:

Theorem (Hellinger-Toeplitz) Let A be an everywhere dened linear operator on a Hilbert

spa e H with (f; Ag ) = (Af; g ) for all f and g in H. Then A is bounded.

The Hellinger-Toeplitz theorem tells us that symmetri unbounded operators annot be

dened on the whole of H. Thus su h operators an be only extended at most into ertain
dense subspa es of H. Given two unbounded densely dened operators A and B in A, the
denition of A + B or AB may be diÆ ult: A + B is a priori only dened on D(A) \ D(B ),
and AB is only dened on the elements ' 2 D(B ) su h that B' 2 D(A). However, if there
exists a ommon invariant subdomain  for the algebra of operators A, i.e., a subspa e 
su h that   D(A) and A :  !  for every A 2 A, then all algebrai operations are
allowed and domain questions do not arise. The need for this domain, that is not provided
by the Hilbert spa e theory, is one of the reasons why we need to go beyond the Hilbert
spa e to the Rigged Hilbert Spa e.
As an example, let L2 (R ; dx) be the Hilbert spa e of square integrable fun tions on the
real line. Then the multipli ation (position) operator

Q : f (x) ! xf (x) (2.5.27)

and the dierentiation (momentum) operator

1 df (x)
P : f (x) ! (2.5.28)
i dx
are not bounded on L2 (R ; dx). Therefore, the ommutation relation

[Q; P ℄ = QP P Q = iI (2.5.29)

is not dened for every element in the Hilbert spa e. However, the a tions of P and Q
an be restri ted to the S hwartz spa e S (R ), that is in luded in the domains of P and Q.
On this subdomain both P and Q are bounded ( ontinuous) with respe t to the topology
generated by the s alar produ ts (2.4.15). On S (R ), the ommutation relation (2.5.29) is
well dened and all algebrai operations are allowed. This will serve as a motivation for
a physi ist to onsider using ountably Hilbert spa es su h as S (R ) rather than just the
Hilbert spa e L2 (R ; dx).
2.6 Nu lear Rigged Hilbert Spa es 65

2.6 Nu lear Rigged Hilbert Spa es

2.6.1 Introdu tion
The lass of ountably Hilbert spa es that is of servi e in Quantum Me hani s is that of
nu lear spa es. Nu lear spa es will appear in onne tion with the spe tral analysis of self
adjoint operators. This spe tral analysis will be provided by the Gelfand-Maurin theorem
(see Se tion 3.5).
In order to introdu e the on ept of nu learity, we onsider a ountably Hilbert spa e 
on whi h an in reasing sequen e of s alar produ ts
('; ')1  ('; ')2 


 ('; ')n 


(2.6.1)

p n whi h are obtained by ompleting the spa e

is dened. We onsider the Hilbert spa es
 with respe t to the norms k'kn = ('; ')n. These ompletions lead to the hain of
spa es
1  2 


 n 


  : (2.6.2)
By onstru tion,  is dense in ea h spa e n . We denote by '[n℄ and '[m℄ the same element
' 2 , onsidered as an element of n and m , respe tively. If m  n, then it follows from
(2.6.1) that the identity mapping
:   n !   m
'[n℄ ! '[m℄ (2.6.3)
is a ontinuous mapping from a dense set in n onto a dense set in m . We an extend this
mapping to a ontinuous linear transformation T nm whi h maps the spa e n onto a dense
subset of m ( f. Se tion 2.3.4).
A ountably Hilbert spa e  is alled nu lear if for any m there is an n su h that the
mapping T nm of the spa e n into the spa e m is nu lear, i.e., it has the form
1
X
T nm ' = k (ek ; ')n hk ; (2.6.4)
k=1

whereP' 2 n, fek g and fhk g are orthonormal systems in n and m , respe tively, k > 0
and 1 k=1 k < 1.
We an extend the on ept of nu learity to a ountably normed spa e. However, this
generalization does not lead to an extension of the lass of spa es onsidered: in any nu lear
ountably normed spa e it is possible to dene a sequen e of s alar produ ts in su h a way
that the spa e be omes a nu lear ountably Hilbert spa e without altering its topology.
Nu lear spa es posses ertain properties that make them suitable for the purposes of
Quantum Me hani s. Here we list the most relevant:
1. Any losed subspa e of a nu lear spa e is nu lear.
66 2 Mathemati al Framework of Quantum Me hani s

2. If  is a nu lear ountably Hilbert spa e, then the strong and the weak topology3 on
 agree.
3. If  is nu lear, then the strong and weak topologies on  (whi h were dened in
Se tion 2.4.2) oin ide.
4. A nu lear spa e is separable (i.e., it ontains a dense ountable subset).
5. A nu lear spa e is omplete with respe t to the weak onvergen e.
6. A Hilbert (or a Bana h) spa e is nu lear only if it is nite dimensional.

There is a number of ountably Hilbert spa es that are nu lear. For example, the
S hwartz spa e S (R ) (see Se tion 2.4.1) is nu lear with respe t to the topology generated
by the s alar produ ts (2.4.15). The spa e K (a) of Se tion 2.4.1 is also nu lear.

2.6.2 Nu lear Rigged Hilbert Spa es

By the use of the on epts dis ussed so far, it is now easy to introdu e the basi notion of
(nu lear) Rigged Hilbert Spa e.
Let  be a nu lear ountably Hilbert spa e. We introdu e a s alar produ t (
;
) into
 satisfying (2.1.12)-(2.1.14). This s alar produ t is also required to be ontinuous with
respe t to the ountably
p Hilbert spa e topology on . The ompletion of  with respe t to
the norm k'k = ('; ') yields the Hilbert spa e H. Therefore, the mapping T that brings
any element of  into the ompletion H is ontinuous. Usually, we identify the spa e 
with the spa e T () and write
  H: (2.6.5)
By onstru tion, the topology of  is stronger (ner) than the topology indu ed by H on
. Along with the spa es  and H we onsider the adjoint spa e  of  and the adjoint
spa e H of H. The adjoint T  of T is an operator mapping H into  . T  is dened
by the equation
h'jT h0i = hT 'jh0i (2.6.6)
for every h0 2 H and ' 2 . Sin e every antilinear fun tional h0 on the Hilbert spa e H
an be written in the form (see Fre het-Riesz Theorem in Se tion 2.3.4)
h0 (f ) = (f; h1 ) ; (2.6.7)
where h1 is some element of H, then T  an be onsidered as a mapping of H into  .
A sequen e f'k g of elements in a ountably Hilbert spa e  is said to be weakly onvergent to ' if
3

limn!1 F ('k ) = F (') for every fun tional F on . By strong onvergen e we mean the onvergen e with
respe t to the ountably Hilbert topology generated by the s alar produ ts (2.6.1).
2.6 Nu lear Rigged Hilbert Spa es 67

A Rigged Hilbert Spa e4 (abbreviated RHS) or a Gelfand triplet is a triplet of spa es ,

H, , having the properties stated above:  is a nu lear ountably Hilbert spa e on whi h
a s alar produ t is dened, H is the ompletion of  with respe t to this s alar produ t,
and  is the adjoint spa e of . For any RHS there exists a ontinuous linear operator T
whi h maps  one-to-one onto a dense subset of H, and its adjoint T  maps H one-to-one
onto a dense subset in  . Therefore, we will denote a RHS by
  H   : (2.6.8)
Sin e T is ontinuous, then there exists a norm k'km and an M > 0 su h that
p
kT 'k =
('; ')  M k'km : (2.6.9)
Thus T an be extended onto the entire spa e n , n  m. We denote the orresponding
operator by Tn . It an be proven that there is a value of n for whi h Tn , mapping the
Hilbert spa e n into H, is a nu lear operator. The operator Tn, mapping H into n , is
also nu lear.
The nu learity of Tn will allow us to write T in a simple form. Sin e Tn is nu lear, there
exist orthonormal basis fhk g and fek g in H and n su h that for every element ' 2 n
one has 1
X
Tn ' = k (ek ; ')hk ; (2.6.10)
k=1
P1
where k  0 and the series onverges. Sin e Tn ' = T ' if ' belongs to , then
k=1 k
for elements ' 2  formula (2.6.10) takes the form
1
X
T' = k (ek ; ')hk : (2.6.11)
k=1
One an asso iate with a RHS a two-sided innite de reasing hain of spa es
 


  n 


 0 


 n 


  ; (2.6.12)
su h that for any integer n there exists a nu lear mapping T nn+1 of the spa e n+1 onto a
dense subset of n , and su h that
1
\ 1
[
= n ;  =  n: (2.6.13)
n=1 n=1
In order to onstru t the hain (2.6.12), we take into a ount the fa t that a nu lear spa e
 is the interse tion of a de reasing hain of Hilbert spa es (see Se tion 2.4.1)
1
\
= n ; 1  2 


 n 


  ; (2.6.14)
n=1
4
The word \rigged" in Rigged Hilbert Spa e has a nauti al onnotation, su h as the phrase \fully rigged
ship"; it has nothing to do with any unsavory pra ti e su h as \xing" or predetermining a result. The
phrase \rigged Hilbert spa e" is a dire t translation of the phrase \osnash hyonnoe Hilbertovo prostranstvo"
from the original Russian.
68 2 Mathemati al Framework of Quantum Me hani s

and for every n the mapping T nn+1 is nu lear. Now the spa e  is the union of an in reasing
hain of Hilbert spa es (see Se tion 2.4.2)
1
[
 = 

n;  1   2 


  n 


  ; (2.6.15)
n=1

where  n  n . We denote by T nn+1 , for n < 1, the operator adjoint to T nn 1 . This
operator is also nu lear. In order to onne t the hains (2.6.14) and (2.6.15), we note that
there is a value n for whi h the operator Tn , mapping n into H, is nu lear. Then the
mapping T n of H into  n is also nu lear. Without loss of generality we may suppose that
n = 1. We now denote H by 0 , and the mappings T1 and T 1 by T 10 and T 0 1 , respe tively.
We thereby obtain the sequen e of spa es (2.6.12).
Example We dene the s alar produ t on S (R ) by
Z +1
('; ) = '(x) (x)dx : (2.6.16)
1
Completion of S (R ) with respe t to this s alar produ t yields the Hilbert spa e L2 (R ; dx).
Sin e S (R ) is nu lear and the s alar produ t (2.6.16) is ontinuous with respe t to the
topology on S (R ), then the triplet
S (R )  L2 (R ; dx)  S (R ) (2.6.17)
is a (nu lear) Rigged Hilbert Spa e. The spa e S (R ) is alled the spa e of tempered distri-
butions. The \plane waves" eix may be onsidered as elements of S (R ) . The fun tional
asso iated to ea h plane wave eix is dened by
Z 1
h'jFi  h'j i :=
eix '(x) eix dx : (2.6.18)
1
It is easy to see that jF i is a well dened ontinuous antilinear fun tional on S (R ).
Chapter 3
The Rigged Hilbert Spa e of the
Harmoni Os illator
In this hapter, we onstru t the RHS of the harmoni os illator. This system is studied
from a dierent point of view to that taken in Quantum Me hani s textbooks. Instead
of assuming that the position and momentum operators are given by the multipli ation
and derivative operators, we shall make three simple algebrai assumptions: the Heisen-
berg ommutation relation, the expression of the Hamiltonian in terms of the position
and momentum operators, and the existen e of an eigenve tor of the Hamiltonian. From
these algebrai assumptions, we shall onstru t the RHS of the harmoni os illator and the
S hrodinger representation of the algebra of the harmoni os illator.

As I sat there, brooding on the old unknown world,

I thought of Gatsby's wonder when he rst pi ked out the
green light at the end of Daisy's do k. He had ome a long
way to this blue lawn and his dream must have seemed so
lose that he ould hardly fail to grasp it. He did not know
that it was already behind him, somewhere ba k in the vast
obs urity beyond the ity, where the dark elds of the re-
publi rolled on under the night.
Gatsby believed in the green light, the orgiasti future that
year by year re edes before us. It eluded us then, but that's
no matter|tomorrow we will run faster, stret h out our
arms farther.... And one ne morning||
So we beat on, boats against the urrent, borne ba k
easelessly into the past.

F. S ott Fitzgerald, The Great Gatsby

69
3.1 Introdu tion 71

3.1 Introdu tion

We shall treat in detail one of the simplest physi al models, the one-dimensional harmoni
os illator, using the framework of the pre eding se tions. Our formulation will easily gen-
eralize to more ompli ated physi al models, and we will list the algebras for whi h these
generalizations are already known.
The standard approa h to the harmoni os illator is to start out with the (position)
S hrodinger realization of the algebra of operators, i.e., one takes for granted the well
known dierential expressions for the operators position Q, momentum P and energy H
of the harmoni os illator. From these expressions one derives, for instan e, the Heisen-
berg ommutation relation. These operators are impli itly assumed to be dened on the
same domain, whi h is assumed to remain stable under their a tion, and so all algebrai
operations su h as the multipli ation of two operators are allowed. The operators Q and
P are assumed to have eigenkets jxi and jpi satisfying Qjxi = xjxi and P jpi = pjpi for
every real x and p, although a satisfa tory mathemati al meaning within a Hilbert spa e
formulation is not possible. Dira basis expansion is also used, although the Hilbert spa e
spe tral de omposition does not orrespond to it.
Here we shall obtain this realization but starting from a dierent point. We shall just
assume some algebrai relations to be fullled by the operators P , Q and H , namely the
Heisenberg ommutation relation and the expression of H in terms of P and Q. We shall
make an additional essential assumption: the existen e of an eigenve tor of the energy
operator. The operators will be dened on a ommon linear spa e that remains stable
under their a tions.
From this algebrai starting point, we shall derive rst that H possesses a ountable
number of eigenvalues ~w(n + 1=2), n = 0; 1; 2; : : :, orresponding to some eigenve tors
n , as it appears in the literature. The linear spa e spanned by the n will be alled .
In Se tion 3.3 this linear spa e is equipped with two dierent topologies: the usual Hilbert
spa e topology, whi h generates the Hilbert spa e H from , and a stronger nu lear topology,
whi h generates the spa e  from . This nu lear topology will make the elements of the
algebra ontinuous operators.
In Se tion 3.4, the spa e of antilinear fun tionals is dened, and the Rigged Hilbert
Spa e
  H   (3.1.1)
for the harmoni os illator is onstru ted. Se tion 3.5 gives the denition of generalized
eigenve tors. This denition will provide the proper mathemati al setting for the eigenket
equations Qjxi = xjxi and P jpi = pjpi. The eigenkets jpi and jxi will be ontinuous
antilinear fun tionals over , i.e., they will be elements of  . A statement of the Gelfand-
Maurin Theorem will be given, whi h will guarantee the existen e of a omplete set of
generalized eigenve tors of the position and momentum operators, as it is usually assumed.
It will be shown that this theorem is the mathemati al statement that justies the heuristi
Dira basis ve tor expansion. In Se tion 3.6, we derive the S hrodinger representation of
the harmoni os illator. In this representation the standard expressions for P , Q and H in
72 3 The Rigged Hilbert Spa e of the Harmoni Os illator

terms of dierential operators will be obtained. The realization of the RHS (3.1.1) by spa es
of fun tions and distributions is also des ribed in Se tion 3.6. The spa e  will be realized
by the S hwartz spa e S (R ), and  by the spa e of tempered distributions S (R ) . Thus
the RHS (3.1.1) will be realized in the position representation by
S (R )  L2 (R ; dx)  S (R ) : (3.1.2)
Therefore, we shall provide a proper mathemati al framework for the operations that
are needed in physi s, and we will throw light onto the problem of how the S hrodinger
realization of the algebra of operators an be singled out. The important point is that
this realization, whi h is introdu ed ad ho in the literature, an be derived from proper
algebrai assumptions within the RHS formalism.

3.2 Algebrai Operations

The algebra ( f. Se tion 2.1.3) A of observables for the one dimensional harmoni os illator
is generated by the operators H (representing the observable energy), P (representing the
observable momentum) and Q (representing the observable position). The dening algebrai
relations are:
1 ! 2 2
H = P2 + Q ; [P; Q℄ = i~I ; (3.2.1)
2 2
where ~ is a universal onstant (Plan k's onstant), ; ! are hara teristi onstants of the
system (mass and frequen y, respe tively) and [P; Q℄  P Q QP is the ommutator of P
and Q. The elements of A are assumed to be linear operators dened on a linear spa e .
There is a s alar produ t (
;
) dened on that provides the probability amplitudes (but
is not a Hilbert spa e). Further, P , Q and H are supposed to be symmetri operators of
into . That is,
A : ! ; (3.2.2)
and
(A'; ) = ('; A ); 8'; 2 ; (3.2.3)
where the operator A an be P , Q or H . The assumptions about the algebra of observ-
ables stated so far do not spe ify the mathemati al stru ture ompletely. There are many
realizations of the ve tor spa e on whi h A is an algebra of operators. We have to make
one further assumption in order to fully spe ify the realization of A, i.e., the realization of
(3.2.1). This additional requirement an be formulated in the following way:
there exists at least one non degenerate eigenvalue of
H whose orresponding eigenve tor is an element of : (3.2.4)
In short, our starting point is to assume that the physi s of the harmoni os illator is
des ribed by an algebra of observables that satisfy the (algebrai ) assumptions (3.2.1)-
(3.2.4).
3.2 Algebrai Operations 73

In order to onstru t the spa e , we make the elements of A a t on the eigenve tor of
(3.2.4). The representation of A by linear operators on obtained in this way is alled the
ladder representation. The pro edure to nd the ladder representation is well known and
will be sket hed only very brie y. One denes
r 
1 ! i
a := p Q+ p P ; (3.2.5)
2 r~ ! ~
 
1 ! i
a := p
y Q p P ; (3.2.6)
2 ~ ! ~
1 1
N := ay a = H I: (3.2.7)
!~ 2
These operators learly fulll, as a onsequen e of (3.2.3),
('; a ) = (ay '; ) ; 8 '; 2 ; (3.2.8)
('; N ) = (N'; ) ; 8 '; 2 : (3.2.9)
Eq. (3.2.1) implies that a and ay fulll
[a; ay ℄ = a ay ay a = I : (3.2.10)
Assumption (3.2.4) implies that there exists a ' 6= 0 in , su h that
N' = ' : (3.2.11)
From (3.2.9) and (3.2.11) it follows that
(' ; ' ) = (' ; N' ) = (N' ; ' ) = (' ; ' ) : (3.2.12)
Therefore,  = , i.e.,  is real. From the ommutation relation (3.2.10), it then follows
that
N (a' ) = ay a a' = (a ay I )a' = a(ay a I )'
= a(N I )' = a( 1)' = ( 1)a' : (3.2.13)
This implies that either a' is an eigenve tor of N with eigenvalue ( 1) or a' = 0.
Further, from (3.2.8) and from the ommutation relation (3.2.10) it follows that
kay'k2 = ('; aya' ) + ('; I') = ka'k2 + k'k2 6= 0 ; (3.2.14)
6 0. In addition, equation
sin e ' is dierent from the zero ve tor. This means that ay ' =
(3.2.10) implies that
N (ay ' ) = ( + 1)ay' ; (3.2.15)
i.e., ay' is an eigenve tor of N with eigenvalue ( + 1).
74 3 The Rigged Hilbert Spa e of the Harmoni Os illator

We now start with the eigenve tor ' , whi h was assumed to exist, and su essively
dene the ve tors
' m = am ' m = 0; 1; 2; 3; : : : (3.2.16)
Sin e a ording to (3.2.13) ea h appli ation of a lowers the eigenvalue by 1, we have
N' m = ( m)' m ; m = 0; 1; 2; : : : (3.2.17)
This means that ' m = am ' , m = 0; 1; 2; : : :, are eigenve tors of N with eigenvalue
( m) unless ' m is the zero ve tor. We shall now show that after a nite number of
steps m0 , the ve tor ' m0 is the zero ve tor
' m0 = am0 ' = 0 : (3.2.18)
To prove this statement, we al ulate the s alar produ t of (3.2.16) and (3.2.17)
(' m ; N ' m ) = ( m) (' m ; ' m ) = (' m ; aya' m ) = ka' m k2 : (3.2.19)
If ' m 6= 0, then (3.2.19) leads to
( m) = kka'  mk
2
' m k 2 : (3.2.20)
Sin e the norm of a non-zero ve tor is always positive, equation (3.2.20) implies that
( m)  0 whenever ' m 6= 0. Now, if ' m were dierent from zero for every
m = 1; 2; : : : , then  m  0 ould not be fullled, sin e  is a xed real number.
Therefore, there must exist an m0 2 N su h that
' m 6= 0 for m < m0 ; (3.2.21)
and
' m0 = a' (m0 1) = 0 : (3.2.22)
This proves (3.2.18).
After normalization, we denote the last non-zero ve tor by
'
0   (m0 1) : (3.2.23)
k' (m0 1)k
From 0 (for whi h a0 = 0) one denes the sequen e of ve tors
0
1 = p1 ay0
1!
1 y2
2 = p (a ) 0
2!
:::: :: :::::
1
n = p (ay )n 0
n!
:::: :: ::::: (3.2.24)
These ve tors have the following properties:
3.2 Algebrai Operations 75

1. They are eigenve tors of the number operator N and of the Hamiltonian H
Nn = nn ; n = 0; 1; 2; : : : (3.2.25)
Hn = ~! (n + 1=2)n n = 0; 1; 2; : : : (3.2.26)
2. For every n, there exists a n+1 that is dierent from the zero ve tor.
3. The a tions of ay and a on the sequen e (3.2.24) are given by
p p
ay n = n + 1n+1 ; an = nn 1 : (3.2.27)

Equation (3.2.26) is usually interpreted by saying that the energy of the harmoni os illator
is quantized and annot take any arbitrary value. Equation (3.2.27) means that if we start
with an eigenstate n of H orresponding to the eigenvalue En = ~! (n +1=2), appli ation of
the operator ay yields an eigenve tor asso iated with the eigenvalue En+1 = ~! (n+1=2)+~! ,
and appli ation of a yields, in the same way, the energy En 1 = ~! (n + 1=2) ~! . For this
reason, ay is said to be a reation operator and a an annihilation operator: their a tion on
an eigenve tor of H makes an energy quantum ~! appear or disappear.
The spa e of the ladder representation of A is the linear spa e spanned by the eigen-
ve tors 0 ; 1 ; : : : ; n ; : : : of (3.2.24), i.e., is the set of all (nite) linear ombinations
M
X
=  n n ; (3.2.28)
n=0
where n 2 C and M is a natural number whi h is arbitrarily large but nite. In we have
the s alar produ t (
;
) for whi h (3.2.3), (3.2.8) and (3.2.9) holds. With respe t to this
s alar produ t, the ve tors n are orthogonal and normalized,
(n ; m ) = Ænm : (3.2.29)
We all the set of ve tors fng an algebrai orthonormal basis for the spa e .
The set an be also viewed as the linear spa e of all sequen es of the form
M
X
 (0 ; 1; : : : ; M ; 0; 0; : : :)   n n ; (3.2.30)
n=0
where n 2 C . The algebrai operations of this linear spa e are dened omponentwise: let
M1
X
= (0 ; 1 ; : : : ; M1 ; 0; 0; : : :)   n n (3.2.31)
n=0
and
M2
X
' = (0 ; 1 ; : : : ; M2 ; 0; 0; : : :)  n n (3.2.32)
n=0
76 3 The Rigged Hilbert Spa e of the Harmoni Os illator

be two elements of (M2 > M1 ). Then

+ ' := (0 + 0 ; 1 + 1 ; : : : ; M1 + M1 ; : : : ; M2 ; 0; 0; : : :)
M2
X
= (n + n )n ; (3.2.33)
n=0
and
M1
X
 := (0 ; 1 ; : : : ; M1 ; 0; 0; : : :) = (n )n : (3.2.34)
n=0
The s alar produ t of and ' is
M2
X M2
X
( ; ') = (n n ; n n) = n n : (3.2.35)
n=0 n=0
Then the norm of is given by
M1
X M1
X
k k 2 = (n n; n n ) = jnj2 : (3.2.36)
n=0 n=1
Sin e the sums (3.2.30)-(3.2.36) only go up to a nite number, the question of the onver-
gen e of these sums does not arise. The algebrai operations (3.2.33), (3.2.34) and (3.2.35)
show is a linear s alar produ t spa e ( f. Se tion 2.1.2).
We now introdu e the spa e spanned by ea h n . This is the subspa e
Rn = fn j  2 C g : (3.2.37)
Rn is a one dimensional subspa e alled the energy eigenspa e asso iated to the n-th eigen-
value of H , be ause all of its elements are eigenve tors of the operator H with eigenvalue
En = ~! (n + 1=2). Although we an just as well work with the ve tors n , by using the
spa es Rn we obtain a formulation whi h immediately generalizes to the ase where Rn is
not one dimensional. The spa e , that is given by
M
X
= f = (0 ; 1 ; : : : ; M ; 0; 0; : : :)   n n ;  n 2 C ; M 2 Ng ; (3.2.38)
n=0
is usually rewritten as the algebrai dire t sum of the spa es Rn :
X
=  Rn : (3.2.39)
algebrai

The right hand side of the equation (3.2.39) means that every 2 an be uniquely written
as a nite linear ombination
 r0 + r1 +


+ rM (3.2.40)
of elements rn = n n 2 Rn that are orthogonal to ea h other (i.e., (rn ; rm ) = 0 if n 6= m).
3.3 Constru tion of the Topologies 77

3.3 Constru tion of the Topologies

3.3.1 Introdu tion
The spa e has so far an algebrai stru ture, namely it is a linear spa e provided with a
s alar produ t. In order to be able to use ertain tools (su h as the Gelfand-Maurin theorem
of Se tion 3.5) we need to equip with a topologi al stru ture, i.e., we need to onstru t a
linear topologi al spa e ( f. Se tion 2.3).
We will dene a topology in terms of the notion of sequen e onvergen e. All the rest of
the topologi al notions su h as ontinuity, denseness, boundedness, losure, ompleteness,
et . will be derived from the notion of sequen e onvergen e. When this notion fully de-
s ribes the topology, the topologi al spa e is said to be rst ountable ( f. Se tion 2.2.2).
The topologies introdu ed in this se tion are all rst ountable. Only Se tion 3.4 will deal
with topologies that are not rst ountable.
Intuitively, a sequen e f'ng onverges to a point ' if whenever we follow the terms of
that sequen e we get loser and loser to the limit point ' with respe t to a ertain sense of
loseness. This denition of onvergen e must be su h that the linear and the topologi al
stru tures an be pie ed together. This is a omplished by requiring that the algebrai
operations be ontinuous with respe t to the topology under onsideration. Therefore, a
linear topologi al spa e has
1. A linear stru ture.
2. A topology  that provides a notion of onvergen e of sequen es. If a sequen e f'n g
onverges to a point ', then we denote

!' :
'n n!1 (3.3.1)

3. The algebrai operations are ontinuous. That is,

(3a) If 'n ! ', then 'n ! ' for every  2 C .
(3b) If C 3 n !  2 C , then n ' ! ' for every '.
(3 ) If 'n ! ' and n ! , then 'n + n ! ' + .
This is not the most general denition of a linear topologi al spa e ( f. Se tion 2.3) but it
is suÆ ient when the topology is rst ountable.
A sequen e f'n g will be said to be Cau hy if the terms of the sequen e get more and
more lose to ea h other as n ! 1. We then write
'n ! 0:
'm n;m!1 (3.3.2)
One may expe t that whenever we follow the elements of a Cau hy sequen e, we always end
up in an element of the spa e. However, this is the ase only in ertain kind of spa es, that
are alled omplete. Therefore, a spa e is omplete with respe t to a given topology if every
Cau hy sequen e has a limit that belongs to the spa e. When a spa e is not omplete, it
78 3 The Rigged Hilbert Spa e of the Harmoni Os illator

an be ompleted by adding to it all the limit points of Cau hy sequen es. In this ase, the
in omplete spa e an be viewed as a dense set of the omplete spa e ( f. Se tion 2.3.2).
We shall introdu e two dierent topologies on , the Hilbert spa e topology and a
nu lear topology. Completion with respe t to ea h topology will lead to the spa e H and ,
respe tively. These two topologies will be fully des ribed by the denition of onvergen e
of sequen es (i.e., they are rst ountable).

3.3.2 Hilbert Spa e Topology

First, we introdu e into the well known Hilbert spa e topology, whi h we shall denote by
H . For ea h 2 , we dene the norm of by
p
k k := ( ; ); (3.3.3)
where ( ; ) is the s alar produ t of . The norm (3.3.3) an be used to introdu e a
meaning of onvergen e, i.e., to spe ify in whi h sense the terms of a sequen e get loser
and loser to a limit point. We will say that a sequen e f n g in onverges to 2 with
respe t to the Hilbert spa e topology H if
k n k!0 as n ! 1 : (3.3.4)
This means that f n g tends to if for every  > 0 there is a positive integer N () su h that
k n k <  for every n > N . Symboli ally, this ondition is written as
H
n n!1 ! i k n k n!1
!0: (3.3.5)
The on ept of Cau hy sequen e an be stated as: a sequen e f n g in is Cau hy with
respe t to H if for every  > 0 there exists a positive integer N su h that k n mk < 
for every m; n > N . In other words, the sequen e f n g is Cau hy if
k n m k!0 as n; m ! 1 : (3.3.6)
The spa e is not omplete with respe t to the Hilbert spa e topology. That is to say,
there exist Cau hy sequen es of elements n in (i.e., sequen es fullling (3.3.6)) that do
not have a H -limit element in (i.e., there is no 2 su h that (3.3.4) holds). As an
example, let us onsider the following innite sequen e:
 
0 = ( 0 ; 0; 0; : : :)  0
1 1
0 1  
1 = ( ; ; 0; : : :)  0 + 1
1 2 1 2
:::: ::: :::::::::
n
0 1 n X i
n = ( ; ;:::; ; 0; : : :) 
1 2 n+1 i=0
i+1
:::: ::: ::::::::: ; (3.3.7)
3.3 Constru tion of the Topologies 79

where the n are the eigenve tors (3.2.24) of N . The sequen e (3.3.7) is H -Cau hy, be ause
for any positive integers n; m (m > n without loss of generality)

k
m
X kik2 m
X 1
mk = ! 0:
2 = (3.3.8)
n
i=n+1
(i + 1)2 i=n+1
(i + 1)2 n;m!1

But (3.3.7) tends, with respe t to H , to

1
0 1 n n+1 X i
z = ( ; ;:::; ; ; : : :)  : (3.3.9)
1 2 n+1 n+2 i=0
i + 1
whi h is an innite sequen e and therefore is not in . This shows that is not omplete.
The spa e an be ompleted with respe t to the topology H by adding to all the
limit points of H -Cau hy sequen es of elements of . The resulting spa e, denoted by H,
is the Hilbert spa e. , that is a s alar produ t spa e whi h is not omplete with respe t
to H , is usually alled a pre-Hilbert spa e. The spa e H is the set of innite sequen es
1
X
h = (r0 ; r1 ; : : : ; rn ; : : :)  rn ; (3.3.10)
n=0
where rn = n n 2 Rn and n 2 C , su h that
1
X 1
X
krnk 2 = jnj2 < 1 ; (3.3.11)
n=0 n=0
or in a more ompa t notation
1
X
H = fh  (r0; r1; : : : ; rn; : : :) j krnk2 < 1 ; rn 2 Rng : (3.3.12)
n=0

Its topology is given by the pres ription (3.3.5) for sequen e onvergen e:
hn n!1! h , khn hk n!1
H C
!0: (3.3.13)
It an be proven that the spa e H dened by (3.3.12) is omplete with respe t to the
topology dened by (3.3.13).
A ve tor h 2 H given by (3.3.10) is uniquely determined by the sequen e of omplex
numbers
h  (0 ; 1 ; : : : ; n ; : : :) ; (3.3.14)
where rn = n n . This sequen e is not arbitrary, but it must fulll (3.3.11). In fa t, it an
be shown that the norm (3.3.3) of any h 2 H is given by
1
X
khk 2 = jnj2 : (3.3.15)
n=0
80 3 The Rigged Hilbert Spa e of the Harmoni Os illator

An innite sequen e (0 ; 1 ; : : : ; n ; : : :) that satises (3.3.11) is said to be square summable.

The set of square summable sequen es is denoted by l2 (C ). It is lear that h 2 H i its
orresponding sequen e (3.3.14) is square summable. Therefore, (3.3.14) and (3.3.15) set up
a one-to-one orresponden e of the Hilbert spa e H onto the spa e l2 (C ). The spa es H and
l2 (C ) are, from a linear topologi al point of view, the same. We then say that l2 (C ) is the
realization of H by the spa e of square summable sequen es. In terms of these sequen es,
the spa e H is given by
1
X 1
X
H = fh =  n n j  n 2 C ; jnj2 < 1g : (3.3.16)
n=0 n=0
The n of (3.3.14) an be obtained as the s alar produ t of h with n
n = (n; h) ; n = 0; 1; 2; : : : (3.3.17)
Thus we an write any element h 2 H as
1
X 1
X
h n (n ; g )  jn)(n; g) ; (3.3.18)
n=0 n=0
in analogy to the three-dimensional ase
3
X 3
X
~x = ~ei (~ei
~x) = ~ei xi : (3.3.19)
i=1 i=1
Equation (3.3.18) is usually interpreted by saying that the fn g form an orthonormal basis
for H and that the n are the omponents along the basis ve tors n .
We are now going to show that H is a tually a H - ompletion of . First, the algebrai
operations on an be readily extended to the Hilbert spa e H. The sum of two elements
1
X
h  (0 ; 1 ; : : : ; n ; : : :)  n n (3.3.20)
n=0
and 1
X
g  (0 ; 1 ; : : : ; n; : : :)  n n (3.3.21)
n=0
of H is dened omponentwise as
1
X
h + g := (0 + 0 ; 1 + 1 ; : : : ; n + n ; : : :)  (n + n )n : (3.3.22)
n=0
It an be proven that if h and g are elements of H, i.e., the sequen es (3.3.20) and (3.3.21)
satisfy (3.3.11), then h + g is also an element of H, i.e., it obeys
1
X
jn + nj2 < 1 : (3.3.23)
n=0
3.3 Constru tion of the Topologies 81

The multipli ation of an element h of H by a omplex number  is dened by

1
X
h := (0 ; 1 ; : : : ; n ; : : :)  (n)n : (3.3.24)
n=0

The s alar produ t on H is dened by

1
X 1
X
(h; g ) = (n n ; n n) = n n : (3.3.25)
n=0 n=0

If h and g are in H, then it an be shown that the series (3.3.25) onverges and that this
s alar produ t is well dened. Therefore, H is a linear s alar produ t spa e that is omplete
with respe t to the topology H generated by this s alar produ t, i.e., H is a Hilbert spa e.
Sin e every element 2 is given by a nite sequen e of the form
= (r0 ; r1 ; : : : ; rM ; 0; 0; : : :) (3.3.26)
that obviously satises (3.3.11), the spa e is a subset of H. The algebrai operations
(3.3.22), (3.3.24) and (3.3.25) on H learly extend the operations (3.2.33), (3.2.34) and
(3.2.35) on . We an see that is dense in H with respe t to this topology. In fa t, every
element 1 X
h  (0 ; 1 ; : : : ; n; : : :)  n n (3.3.27)
n=0
of H is the H -limit of a sequen e of elements of with the form
n
X
n  (0; 1 ; : : : ; n; 0; 0; : : :)   i i : (3.3.28)
i=0

p H is the H - ompletion of with respe t to the topology dened by the

In other words,
norm k k = ( ; ).
In terms of the spa es Rn , the spa e H is usually written as
X
H=  Rn : (3.3.29)
Hilbert

The right hand side of (3.3.29) is usually alled the Hilbertian dire t sum or orthogonal
dire t sum of the Rn be ause the spa es Rn are orthogonal to ea h other, sin e (rn ; rm ) = 0
for n 6= m, where rn 2 Rn and rm 2 Rm .
An example of an element of H is the sequen e
1
1 1 1 X n
z  (1; ; ; : : : ; ; : : :)  (3.3.30)
2 3 n n=0
n+1
82 3 The Rigged Hilbert Spa e of the Harmoni Os illator

of (3.3.9). For this sequen e, it holds that

1 1
kzk 2
X knk2 X 1 2
=
(n + 1)2
=
n2 = 6 : (3.3.31)
n=0 n=1

z is the H -limit element of the Cau hy sequen e (3.3.7), be ause for every n in (3.3.7)
1 1
k zk =
2
X kik2 =
X 1
!0: (3.3.32)
n
i=n+1
(i + 1)2 i=n+1
( i + 1)2 n!1

Summarizing, we started with the linear s alar produ t spa e of nite linear ombi-
nations of the eigenve tors n . In this spa e, we introdu ed the topology H through the
meaning of sequen e onvergen e (3.3.5). The spa e was not omplete with respe t to
H , i.e., there were Cau hy sequen es of elements of that did not have a H -limit element
in . was ompleted to a spa e H by adjoining to it all limit elements of H -Cau hy
sequen es. Thus, the Hilbertian dire t sum was obtained by ompleting the algebrai dire t
sum with respe t to H :
X X
=  Rn 7 ! H ompletion 7 ! H=  Rn : (3.3.33)
algebrai Hilbert

The operators of the algebra of observables A an be onsidered now as linear operators

dened on a subdomain of the Hilbert spa e. These operators an be extended to larger
subdomains of H (see Se tion 3.3.5). But these extensions are not ontinuous with respe t
to H . Moreover, their domains do not remain stable under the a tions of the operators.
Therefore, algebrai operations su h as the sum or multipli ation of two operators are not
always allowed. Sin e in physi s these kind of operations are always assumed to be well
dened, it is reasonable to sear h for a spa e  that

1. remains stable under the a tion of the elements of the algebra A,

A :  ! ; (3.3.34)

where A is any element of A.

2. Every A 2 A is ontinuous with respe t to ertain topology on . This ontinuity

will allow us to extend the operators from to their extension on  in a unique way.

The domain  is hara teristi of the parti ular physi al system (i.e., of the parti ular alge-
bra of observables) upon onsideration. The onstru tion of this domain for the harmoni
os illator is the subje t of the next se tion.
3.3 Constru tion of the Topologies 83

3.3.3 Nu lear Topology

We now onstru t the spa e , whi h will be the ompletion of with respe t to a nu lear
topology  . We take the original s alar produ t (
;
) and dene the quantities
('; )p := ('; (N + I )p ) ; 8'; 2 ; p = 0; 1; 2; : : : (3.3.35)
and q
k kp := ( ; )p ; 8 2 : (3.3.36)
From the properties of the linear operator N , it is easy to see that ('; )p of (3.3.35) fullls
the onditions (2.1.12)-(2.1.14) of a s alar produ t and that the quantities of (3.3.36) form
an in reasing sequen e of norms
k k0  k k1  k k2 


(3.3.37)
Further, these norms are ompatible, i.e., if a sequen e onverges to zero with respe t to
one norm and is a Cau hy sequen e with respe t to another, then it also onverges to zero
with respe t to this other norm ( f. Se tion 2.4.1).
We now dene the notion of onvergen e that will determine the nu lear topology  .
A sequen e f n g of elements in onverges to an element with respe t to the topology
 if f n g onverges to with respe t to every norm in (3.3.36):
! ,k

n n!1 n kp ! 0 ; for every p = 0; 1; 2; : : : (3.3.38)
The reason why we introdu e the s alar produ ts (3.3.35) and the topology (3.3.38) is
that the operator N will be ontinuous with respe t to  ( f. Se tion 3.3.5). Moreover, due
to the spe ial stru ture of the algebra of the harmoni os illator, the ontinuity of N will
imply the ontinuity of the rest of the observables of this algebra ( f. Se tion 3.3.5).
From n ! it follows that n H! , but not vi e versa. Therefore  is stronger
(ner) than H , and H is weaker ( oarser) than  .
A sequen e f n g in is  -Cau hy if for every p and for every  > 0 there exists a
positive integer N = N (; p) su h that
k n m p k < for every n; m > N ; (3.3.39)
i.e., f n g is  -Cau hy if it is Cau hy with respe t to every norm of (3.3.36).
We now omplete with respe t to  by adding to the limit points of all  -Cau hy
sequen es. The omplete linear topologi al spa e obtained in this way is denoted by .
 is alled a ountably Hilbert spa e ( f. Se tion 2.4). Sin e there are more H -Cau hy
sequen es than  -Cau hy sequen es (be ause a H -Cau hy sequen e must fulll (3.3.39)
only for p = 0), this implies
   H: (3.3.40)
Thus, is  -dense in , and sin e is H -dense in H,  is H -dense in H.
84 3 The Rigged Hilbert Spa e of the Harmoni Os illator

In order to onstru t  expli itly, let us see whi h of the innite sequen es that are
elements of H, i.e., that fulll (3.3.11), are also elements of . A sequen e
' = (r0 ; r1 ; : : : ; rn ; : : :) ; rn 2 Rn ; (3.3.41)
is an element of  i it is the limit point of the sequen e
n = (r0 ; r1 ; : : : ; rn ; 0; 0; : : :) (3.3.42)
of elements of with respe t to  , i.e., i
k n 'kp ! 0 ; for every p = 0; 1; 2; : : : (3.3.43)
Equation (3.3.43) is equivalent to
(( n '); (N + I )p( n ')) ! 0 ; for every p = 0; 1; 2; : : : ; (3.3.44)
whi h is equivalent to
1
X 1
X
(ri ; (N + I )pr i) = (i + 1)pkri k2 n!1
! 0 ; for every p = 0; 1; 2; : : : (3.3.45)
i=n+1 i=n+1
Therefore, ' is the  -limit point of a  -Cau hy sequen e of elements of i it fullls
(3.3.45). Then the spa e  is given by
1
X 1
X
 = f' = n n j n 2 C ; (n + 1)pjn j2 < 1 for every p = 0; 1; 2; : : :g : (3.3.46)
n=0 n=0
Obviously,  is a linear spa e under the algebrai operations inherited from H.
If we denote the ompletion of with respe t to ea h norm k kp of (3.3.36) by p, for
every p = 0; 1; 2; : : :, then it an be shown that
1
\
= p : (3.3.47)
p=0
 is also alled the -dire t sum of the spa es Rn, that is denoted by
X
=  Rn : (3.3.48)
nu lear
This  -dire t sum has been onstru ted by ompleting the algebrai dire t sum with respe t
to the  topology
X X
=  Rn 7 !  ompletion 7!  Rn : (3.3.49)
algebrai nu lear
The operators in the algebra of observables an be extended to  ( f. Se tion 3.3.5).
These extensions will be proven to be ontinuous with respe t to  and the spa e  will
be proven to remain stable under the a tion of the operators of the algebra. Therefore, all
algebrai operations will be allowed. This is, in fa t, the very reason why we have introdu ed
the spa e .
3.3 Constru tion of the Topologies 85

3.3.4 Physi al Interpretation of ,  and H

To see what these various spa es might mean for physi s, we re all that Rn is the energy
eigenspa e orresponding to the energy eigenvalue
 1
E n = ~! n + : (3.3.50)
2
P1
If h = (0 ; 1 ; : : : ; n ; : : :) = n=0 n n , where n = (n ; h), then
1
X
h2H , jnj2 < 1 ; (3.3.51)
n=0
X1
h2 , (n + 1)pjn j2 < 1 for p = 0; 1; 2; : : : ; (3.3.52)
n=0
h2 , all n but a nite number are equal to 0 : (3.3.53)
Clearly, dire t experimental data an only tell us something about . In fa t for most
real physi al systems whose idealization is the harmoni os illator, e.g. diatomi mole ules,
only the very lowest energy levels are relevant; for higher energy the diatomi mole ule
is no longer a harmoni os illator and nally not even an os illator.  and H are both
idealizations, though  appears \ loser" to reality.
The reason why we prefer the mathemati al idealization provided by  over the one pro-
vided by H an be vaguely summarized by saying that  admits Dira 's bra-ket formalism.
Two aspe ts of this formalism are:
1. All algebrai operations involving the observables are allowed and no questions re-
garding the domain of denition arise.
2. For every observable there exists a omplete system of eigenve tors su h that every
wave fun tion an be expanded in terms of these eigenve tors.
The rst aspe t follows from the fa t that all the elements of the algebra A leave invariant
, are ontinuous operators with respe t to  and therefore uniquely dened on the whole
spa e . This will be dis ussed in the next se tion. The se ond aspe t will need the on epts
of dual spa e and generalized eigenfun tion, and will be dis ussed in Se tion 3.5.

3.3.5 Extension of the Algebra of Operators

The operators of the algebra of observables were assumed to be dened on the spa e . Sin e
 H, these observables an be onsidered as linear operators dened on the subdomain
of the Hilbert spa e H. They an be extended to larger subdomains of H by using the
notion of losure (see Se tion 2.5.3). If n 2 and n H! f , but f 62 , then A n 2
for every n but A is not dened on f . If A n H!g , then we dene the losure A of A by
Af = g . We an do this only for those f 2 H whi h are H -limit points of some sequen es
86 3 The Rigged Hilbert Spa e of the Harmoni Os illator

n 2 and for whi h A n H- onverges. The domain D(A) of A is the set of elements for
whi h the above pro edure an be applied. Evidently, the losure of an operator extends
the operator itself. Thus, in orresponden e to the relation
H (3.3.54)
between the spa es, we have the relation
AA (3.3.55)
between the operators. However, in general, D(A) 6= H. This means that there are elements
h in H on whi h A is not dened, i.e., kAhk = 1.
As an example, we onsider the operator N . This operator an be extended from to
larger subdomains of H. But N annot be extended to the whole of H. For instan e, the
a tion of N annot be extended to the Hilbert spa e element
1 1
z  (1; ; : : : ; ; : : :) (3.3.56)
2 n+1
be ause
n  (1; 12 ; : : : ; n +1 1 ; 0; : : :) n!1
! z;
H
(3.3.57)
but
N n  (0; 1; 12 ; 32 ; : : : ; n +n 1 ; 0; : : :) n!1
! (0; 1; 12 ; 23 ; : : : ; n +n 1 ; : : :) ;
H
(3.3.58)

whi h is not an element of H. In fa t, the largest subdomain of H to whi h N an be

extended is given by
1
X 1
X
fh = n n j n2 jnj2 < 1g : (3.3.59)
n=0 n=0
The extension
P
of any other operator A of the algebra A an be onstru ted in a similar way.
If h = 1 n=0 n n is an element of H, then the a tion of the extension of A (that is also

denoted by A) on h is given by
1 ! 1
X X
Ah = A n n := n (An) : (3.3.60)
n=0 n=0

This extension, however, is not dened for every element of the Hilbert spa e, but only for
those h 2 H for whi h kAhk < 1. Therefore, the operators of A annot be extended to the
whole of H and their extensions are not H - ontinuous. The domains of the extensions of
the operators of A are, in general, dierent for dierent operators, and do not remain stable
under the a tion of the extensions. In order to avoid domain questions, we need the spa e
. This spa e is the largest subspa e of H on whi h all the extensions of the operators are
well dened and that remains stable under the a tion of these extensions. From now on, we
3.3 Constru tion of the Topologies 87

will always onsider the domain of the operators to be . We will denote the (Hilbert spa e)
extension of P , Q, H : : : to  also by P , Q, H : : : These extensions are also symmetri ,
(A'; ) = ('; A ) ; 8'; 2  ; (3.3.61)
where A an be P , Q or H .
It an be shown that the operator H is not only symmetri but also essentially self-adjoint
(e.s.a.) ( f. Se tion 2.5.3), i.e., it an be shown that
H = Hy : (3.3.62)
In fa t, it an be proven (see referen e [65℄) that the requirement that H has at least one
eigenve tor (see (3.2.4)) is equivalent to the requirement that H is e.s.a.:
There exists a 'E su h that H'E = E'E i H = H y : (3.3.63)
We remark that either of these requirements leads to a representation of A whi h integrates
to a representation of the group generated by P , Q and I (Weyl group): the requirement
that H is e.s.a. be ause of the Nelson theorem (see [65℄) and the requirement that H has
one eigenve tor be ause it leads to the ladder representation, and ladder representations are
always integrable.
Sin e N = 1=(~! )H 1=2I , N is also e.s.a. That is,
Ny = N : (3.3.64)
As a onsequen e of the fa t that P and Q are elements of the Lie algebra of the Weyl group,
it follows that P and Q are also e.s.a. by a theorem of Nelson and Stinespring (see [66℄).
That is,
Py = P ; Qy = Q : (3.3.65)
H + I is, ex ept for some onstant fa tors, the Nelson operator, and is also e.s.a.
In our example, one an easily see that N is e.s.a. without invoking the Nelson theorem.
It an be easily proved by using one of the riteria of essentially self-adjointness:
Lemma: An operator A is e.s.a. if (A + I ) 1 is ontinuous and has a dense domain in
H.
The spe trum ( f. Se tion 2.5) of (N + 1) 1 is 1=(n + 1), n = 0; 1; 2; : : : Consequently,
it is a ontinuous operator. Its domain is dense in H. Therefore, N is e.s.a. As a
onsequen e, N + I is e.s.a. Further, (N + I )p is e.s.a.1 for every p = 1; 2; : : :
The extensionPof any operator A of A from toP is onstru ted as follows: given
any element ' = 1 n=0 n n of , the sequen e n = i=0 i i of elements of onverge
n

1
That (N + I )p is e.s.a. an be proved in many ways. It also follows from the fa t that (N + I )p is an
ellipti element in the enveloping algebra of a group representation (see referen e [66℄).
88 3 The Rigged Hilbert Spa e of the Harmoni Os illator

to ' with respe t to  , i.e., n !'.


The a tion of A on ' is dened by A' := 
limn!1 A n , or
1
X
A' := n An : (3.3.66)
n=0
These extensions are unique if the operators are  - ontinuous on ( f. Se tion 2.3.4). Sin e
the produ t and the sum of two ontinuous operators are ontinuous operators, the elements
of our algebra A are ontinuous if a and ay are ontinuous operators. In parti ular, P , Q and
H are ontinuous if a and ay are ontinuous. In a spa e on whi h the topology an be dened
by the onvergen e of sequen es, i.e., a spa e in whi h the rst axiom of ountability holds,
a linear operator A is ontinuous i for all sequen es f'n g with 'n ! 0 it follows that
A'n ! 0 ( f. Se tion 2.3.4). Sin e our topology  satises the rst axiom of ountability,
the operator a is ontinuous if from 'n !0 it follows that ay 'n !0:
y 

To prove that ay is - ontinuous we use a lemma (see Appendix 3.7.4) that states
that for every norm k kp of (3.3.36) there is a real onstant  < 1 su h that
( ; a(N + I )pay )  ( ; (N + I )p+1 ) ; 8 2 : (3.3.67)
Let n ! 0 as n ! 1. Then k nkp ! 0 for every p = 0; 1; 2; : : :,
( n ; (N + I )p n) ! 0 for every p = 0; 1; 2; : : : (3.3.68)
To show that ay n !0, we have to show that
kay n kq ! 0 (3.3.69)
for every q = 0; 1; 2; : : :, i.e., that
(ay n ; (N + I )q ay n ) = ( n ; a(N + I )q ay n) ! 0 (3.3.70)
for every q = 0; 1; 2; : : : By (3.3.67)
( n; a(N + I )q ay n )  ( n ; (N + I )q+1 n) : (3.3.71)
By (3.3.68) the right hand side of (3.3.71) tends to zero for every q = 0; 1; 2; : : :, and
onsequently also the left hand side, whi h proves (3.3.70). Then ay is ontinuous.
The proof of the - ontinuity of a is analogous.
We remark that the onvergen e of kay 'n kq ! 0 as n ! 1 for a xed q follows from
the onvergen e of k'n kq+1 ! 0. Therefore it is important to have a ountably innite
rather than a nite number of norms; in the ase of a nite number of norms, ay is not a
ontinuous operator, sin e the topology generated by a nite number of norms is equivalent
to the topology generated by one norm. In parti ular, this implies that ay annot be a
ontinuous operator with respe t to the Hilbert spa e topology.
We have shown that a and ay , and therewith the elements of the algebra of observables,
are  - ontinuous operators on the linear topologi al spa e . Their  - ontinuous exten-
sions to  are then unique. We denote the  -extensions of the operators a, ay , P , Q,
3.4 The RHS of the Harmoni Os illator 89

H; : : : to  also by a, ay, P , Q, H; : : : Domain questions do not arise when we perform the

algebrai operations, be ause the operators are dened on the whole spa e , and be ause
 remains invariant under their a tion. It is worthwhile noting that the -extension to 
of any operator A is the same as the Hilbert spa e extension of A to .
The se ond aspe t of the Dira formalism, the existen e of a omplete set of eigenve tors,
follows from the nu lear spe tral theorem (see Se tion 3.5 below). In order to be able to
apply this theorem to our example, we need to prove that the topology  is nu lear.
Before introdu ing the notion of nu lear topology, we need the following denition: a
bounded self-adjoint operator B dened on a Hilbert spa e H is Hilbert-S hmidt if B an
be written as
1
X
B= k Pk ; (3.3.72)
k=1
wherePthe Pk are proje tion operators ( f. Se tion 2.5.1) onto nite dimensional spa es Hk
and 1 k=1 (jk j dim Hk ) < 1. Instead of giving the original denition of nu lear spa e
2
( f. Se tion 2.6.1), we shall use a theorem by Roberts (see [10℄) whi h gives a ne essary and
suÆ ient ondition for a spa e to be nu lear: a linear topologi al spa e  is nu lear if there
exists an e.s.a.  - ontinuous operator A 2 A, whose inverse is Hilbert-S hmidt.
It is now very easy to see that our  is nu lear be ause N , and therefore N + I , is
e.s.a.,
P1 the spe trum of (N + I ) 1 is (n + 1) 1 ; n = 0; 1; 2; : : :, Rn is one dimensional and
n=0 1=(n + 1) < 1. Thus N + I is the operator that fullls the above denition.
2
Having established that  is nu lear, we an now addu e the nu lear spe tral theorem
to show that the essentially self-adjoint operators P and Q ea h has a omplete set of
eigenve tors in the sense of Dira . Unlike the ompa t operators, for whi h su h a set
of ve tors an be found from among the elements of the orresponding Hilbert spa e, the
eigenve tors given by the nu lear spe tral theorem neither reside in the spa e  nor in H.
Instead, these ve tors a quire mathemati al sense as elements of the topologi al dual of ,
and therewith the more pre ise terminology generalized eigenve tors. Before presenting the
nu lear spe tral theorem in Se tion 3.5 we shall dis uss the dual spa e of , the subje t of
the next se tion.

3.4 The RHS of the Harmoni Os illator

3.4.1 The Conjugate Spa e
As mentioned above, the generalized eigenve tors of the Dira basis ve tor expansion will
be des ribed by ontinuous antilinear fun tionals over the spa e . This is the notion that
we are about to present (see also Se tion 2.3.4).
An antilinear fun tional F on the linear spa e  is a fun tion F (') from  into the
omplex plane C whi h satises
F (' +  ) = F (') + F ( ) ; 8'; 2  ; 8; ; 2 C : (3.4.1)
90 3 The Rigged Hilbert Spa e of the Harmoni Os illator

In the bra-ket notation F (') = h'jF i, Eq. (3.4.1) reads

h' +  jF i = h'jF i +  h jF i : (3.4.2)
In the spa e , on whi h the topology  is dened, we an use the notion of ontinuity
with respe t to this topology to introdu e the on ept of ontinuous mapping: a fun tion
F is ontinuous if the image of every onvergent sequen e is also onvergent,
'n n!1! ' =) F ('n) n!1

! F ( ') ;
C
(3.4.3)
where ! indi ates onvergen e with respe t to the topology  and C! means onver-
gen e in the sense of omplex numbers.
A ontinuous antilinear fun tional F on  is a fun tion from  into C that is antilinear
and ontinuous. One an prove that a fun tional F on the ountably Hilbert spa e  is
 - ontinuous i there exist a positive onstant K and a norm k
kq among the olle tion
of norms (3.3.36) that obey
jF (')j < K k'kq ; 8 ' 2  : (3.4.4)
We now onsider the olle tion of all ontinuous antilinear fun tionals over , whi h is
denoted by  . The sum of two fun tionals and the multipli ation of a fun tional by a
number are dened by:
(F1 + F2 )(') = F1 (') + F2 (') ; 8' 2  ; 8;  2 C ; (3.4.5)
or in bra-ket notation
h'jF1 + F2i = h'jF1i +  h'jF2i : (3.4.6)
One an prove that if F1 ; F2 2  , i.e., if F1 and F2 satisfy (3.4.1) and (3.4.3), then F1 +F2
is also in  . Thus,  is a linear spa e under the operations dened by (3.4.5).
The dual spa e of the Hilbert spa e H an be onstru ted in a similar way. A fun tion
F from H into C is a H - ontinuous antilinear fun tional if
1. F is antilinear,
F (f + g ) = F (f ) + F (g ) ; 8f; g 2 H ; 8;  2 C : (3.4.7)

2. F is H - ontinuous,
fn ! f =) F (fn) C! F (f ) :
H
(3.4.8)
The adjoint spa e H of H is the olle tion of all H - ontinuous antilinear fun tionals over
H. The spa e H an be endowed with a linear stru ture if the sum of two fun tionals and
the multipli ation of a fun tional by a number are dened as in (3.4.5).
From the relation   H, it an be shown that
H   : (3.4.9)
3.4 The RHS of the Harmoni Os illator 91

Let F 2 H. If ffng is any sequen e of elements of H that onverges to f 2 H with

respe t to H, then F (fn) onverges to F (f ),
H H
fn ! f =) F (fn ) ! F (f ) : (3.4.10)
Let 'n be a sequen e of elements of  that onverges to ' 2  with respe t to .
Sin e 'n !' implies that 'n H!', if follows from (3.4.10) that F ('n ) H!F (').
Therefore, 
'n ! ' =) F ('n ) H! F (') ; (3.4.11)
whi h proves that F 2  .

3.4.2 Constru tion of the Rigged Hilbert Spa e

We are now in a position to onstru t the Rigged Hilbert Spa e for the Harmoni Os illator.
We begin this onstru tion onsidering the antilinear fun tional on H dened by
Ff (g ) := (g; f ) ; 8g 2 H ; (3.4.12)
where f is a xed element of H. It is easy to see that Ff , whi h is determined by the ve tor
f 2 H, fullls the ondition (3.4.7) if the fun tion (
;
) fullls the onditions for a s alar
produ t. Further, Ff 2 H , i.e., it also fullls (3.4.8). In order to prove this, we take a
sequen e gn of elements of H that onverges to g 2 H with respe t to H . In a Hilbert
spa e, the H - onvergen e implies that
(gn ; h) ! (g; h)
C
(3.4.13)
for ea h h 2 H. This s alar produ t onvergen e for ea h h 2 H is alled the weak
onvergen e of gn to g , in ontrast to the norm onvergen e gn H! g ( alled also strong
onvergen e in H). Applying Eq. (3.4.13) to f of (3.4.12), it follows that Ff (gn ) ! Ff (g ).
Thus, Eq. (3.4.12) denes a H - ontinuous antilinear fun tional for every f 2 H. Further-
more, the onverse is also true (see Fre het-Riesz Theorem in Se tion 2.3.4), i.e., for every
antilinear H - ontinuous fun tional F (H) 2 H there exists a unique ve tor fF 2 H su h
that
hgjF (H)i = F (H)(g) = (g; fF ) ; for every g 2 H : (3.4.14)
Therefore we an identify the Hilbert spa e H and its onjugate spa e H by equating the
fun tional F (H) 2 H with the ve tor fF 2 H given by (3.4.14)
H 3 F (H)  fF 2 H : (3.4.15)
Then we have that
   H ' H : (3.4.16)
For H - ontinuous fun tionals F (H) the symbols h
j
i and (
;
) are equivalent after the
identi ation F (H)  fF . This identi ation is possible be ause the a tion of the fun tional
F (H) at any ve tor g 2 H is equal to the s alar produ t of g with fF :
hgjF (H)i = hgjfF i = (g; fF ) : (3.4.17)
92 3 The Rigged Hilbert Spa e of the Harmoni Os illator

In the bra-ket notation, the identi ation F (H)  fF is written as jF (H) i  jfF ) or even
as jF i  jf ). However, for the lass of  - ontinuous fun tionals F the symbol h'jF i (the
a tion of the fun tional F at the point ') is equal to ('; F ) (the s alar produ t of ' and
F ) only if F 2 H . That is,
h'jF i = h'jF (H) i = ('; F ) only if F  F (H) 2 H : (3.4.18)
From Eqs. (3.4.9) and (3.4.16), it follows that
  H   : (3.4.19)
This triplet of spa es is the Rigged Hilbert spa e (RHS) or the Gelfand Triplet for the
Harmoni os illator.
In the spa e  , one an introdu e various topologies and therewith various meanings of
sequen e onvergen e ( f. Se tion 2.4.2). An example is the weak topology, that is denoted
by   or by W ( f. Se tion 2.4.2). This topology leads to a meaning of sequen e onvergen e
that is analogous to the weak onvergen e in H: a sequen e of fun tionals fF g1 =1  

onverges (weakly) to a fun tional F with respe t to   if
h'jF i ! h'jF i ; for every ' 2  : (3.4.20)
However, this notion of sequen e onvergen e does not spe ify the weak topology ompletely,
i.e.,   is not rst ountable.
On e  is equipped with the topology   , we an onstru t its dual spa e, that is
denoted by  . The elements of  are the   - ontinuous antilinear fun tionals 'e on
, i.e., the mappings 'e from  into C that satisfy
1. 'e is linear,
'e(F1 + F2 ) = 'e(F1 ) +  'e(F2 ) ; 8F1 ; F2 2  and 8;  2 C ; (3.4.21)
or in bra-ket notation
hF1 + F2j'ei = hF1j'ei +  hF2 j'ei : (3.4.22)
2. 'e is ontinuous with respe t to   .
The spa e  is also a linear topologi al spa e if addition and multipli ation are dened
by
hF j'e1 +  'e2i = hF j'e1i +  hF j'e2i (3.4.23)
and the (weak) onvergen e is dened by
 
! 'e , hF j'e i ! hF j'ei for every F 2  :
'e (3.4.24)
One an prove that to ea h element ' 2  there orresponds an antilinear ontinuous
fun tional 'e in  dened by
'e(F ) = F (') ; (3.4.25)
3.4 The RHS of the Harmoni Os illator 93

or in the Dira 's notation

hF j'ei = h'jF i : (3.4.26)
Thus, we an identify every element 'e 2  with an element ' 2  through (3.4.26) and

vi e versa. Further, it an be shown that the onvergen e dened by (3.4.24) is the same
as the onvergen e with respe t to  . Therefore, the spa es  and  are, from a linear
topologi al point of view, the same and an be identied
 '  : (3.4.27)
The Hilbert spa e H ertainly satises (3.4.27) be ause it already satises H ' H. Sin e
the fun tionals over H are given by (3.4.12), the Hilbert spa e relation that orresponds to
the relation (3.4.26) is
(f; h) = (h; f ) ; (3.4.28)
whi h is a property of the s alar produ t.

3.4.3 Continuous Linear Operators on the Rigged Hilbert Spa e

The operators of the algebra of observables were assumed to be dened on the linear s alar
produ t spa e . Later on, they were extended to the spa e  in a unique way. We now
extend their a tions to the dual spa e  . For every operator A on , one an dene the
onjugate operator (also alled the dual operator) A on  by
(A F )(') := F (A') ; 8' 2  ; 8F 2  ; (3.4.29)
or in bra-ket notation
h'jAjF i = hA'jF i : (3.4.30)
If A is a  - ontinuous operator on , then A is a   - ontinuous operator on  . In
parti ular, this implies that
 
A F  ! A F whenever F  ! F : (3.4.31)
We have dened the notion of ontinuous operator through the notion of sequen e
onvergen e. This is possible only in spa es where the rst axiom of ountability is
satised. The Hilbert spa e and the spa e  are su h spa es. In these ases, every
ontinuous operator is bounded and every bounded operator is ontinuous. If the
spa es are not rst ountable, the notion of ontinuity of an operator annot be fully
spe ied by the notion of sequen e onvergen e. This is why in the spa e |that
is not rst ountable|operators A that fulll (3.4.31) are not ne essarily bounded.
If a  - ontinuous operator A is also symmetri , then we have in orresponden e to the
relation (3.4.19) between the spa es the relation
A  A  Ay  A (3.4.32)
between the operators. When the operator A is also e.s.a., Eq. (3.4.32) be omes
A  A = Ay  A : (3.4.33)
In parti ular, the adjoint operators P  , Q , H  of the operators P , Q, H are   - ontinuous
operators on  and satisfy (3.4.33).
94 3 The Rigged Hilbert Spa e of the Harmoni Os illator

3.5 Basis Systems, Eigenve tor De omposition and the

Gelfand-Maurin Theorem
The next se tion is presented a heuristi motivation for Dira basis ve tor expansion. The
mathemati al details are overed in Se tion 3.5.2.

3.5.1 Basis Systems and Eigenve tor De omposition|a Heuristi

Introdu tion
The simplest example of basis system and eigenve tor de omposition is given by the three-
dimensional spa e R 3 . In R 3 it is ustomary to hoose a system of three orthonormal ve tors
e1 ; e2; e3 satisfying
ei
ej = Æij ; i; j = 1; 2; 3 ; (3.5.1)
where ei
ej is the s alar produ t of ei and ej and Æij is the Krone ker delta dened by

1 if i = j
Æij = 0 if i 6= j : (3.5.2)

The basis system fullling (3.5.1) an be hosen somewhat arbitrarily. But it is onvenient
to hoose it in su h a way that the parti ular physi al problem under onsideration takes its
simplest mathemati al form. For example, if one des ribes a three-dimensional rigid body
with moment of inertia tensor I, then it is useful to hoose the basis system fei g su h that
ei
I
ej = I(j) Æij ; (3.5.3)
i.e., to hoose the fei g to be eigenve tors of the (rank 2) tensor I. Every ve tor v 2 R 3 an
be expanded with respe t to this basis system of eigenve tors of I,
3
X
v= ei vi ; (3.5.4)
i=1

where the vi = ei
v are the oordinates (or omponents) of the ve tor v with respe t to the
basis fei g of eigenve tors of the symmetri tensor I.
In analogy to the three-dimensional spa e R 3 , one introdu es a system of basis ve tors
in a general linear spa e . If  is an N -dimensional linear spa e, then there are N linearly
independent ve tors fei gNi=1 that form an orthonormal basis system for . We denote these
basis ve tors ei also by jei ). The s alar produ ts of the elements of the basis system are
written in one of the following ways:
ei
ej  (ei ; ej )  (ei jej ) = Æij ; i; j = 1; 2; : : : ; N : (3.5.5)
As the basis system for the spa e , it is often extremely onvenient to hoose the eigen-
ve tors of an operator A whi h represents an important observable (most frequently one
3.5 Basis Systems, Eigenve tor De omposition and the Gelfand-Maurin Theorem 95

hooses the energy operator H , the position operator Q or the momentum operator P ).
Therefore, one seeks a set of basis ve tors ei 2  whi h also fulll
Aei = ai ei (3.5.6)
for some ai 2 C. These eigenve tors are often labeled by their eigenvalues ai and denoted
by
ei  jai ) : (3.5.7)
It an be proven that if A is a Hermitian operator on an N -dimensional spa e , then there
exists an orthonormal basis system of eigenve tors of A,
Ajai ) = ai jai ) ; i = 1; 2; : : : ; N ; (3.5.8)
(ai jaj ) = Æij ; i; j = 1; 2; : : : ; N ; (3.5.9)
su h that every ve tor ' 2  an be written as
N
X
'= jai)(aij') : (3.5.10)
i=1
This result is the spe tral theorem for a Hermitian operator A dened on a nite dimensional
s alar produ t spa e . Eq. (3.5.10) is alled the spe tral de omposition of the ve tor ' or
the eigenve tor expansion of ' with respe t to the basis system fjai )g. The omplex numbers
'i  (ai j') (3.5.11)
are the omponents of the ve tor ' with respe t to the basis fjai )g. The set of ai 's (whi h
are real if A is Hermitian) is alled the spe trum of A.
In general, the dimension of the linear spa e  is not nite. In this ase, the above result
annot be applied, and a proper generalization to the innite dimensional ase is needed.
In this innite dimensional ase, there are two possibilities depending on the spe trum of
the observable upon onsideration. When the possible measurements of an observable are
elements of a dis rete set of numbers, then only a dis rete set of eigenvalues is ne essary.
In this ase, only an innite dimensional generalization of (3.5.8)-(3.5.10) is needed. How-
ever, there seem to be observables in physi s whose possible measurements are elements of
a ontinuous set of numbers (e.g., the momentum and position an, in many ases, take
any real value). In this se ond ase, we need not only the innite dimensional general-
ization of (3.5.8)-(3.5.10) but also the ontinuous innite dimensional generalization. This
generalization is the Dira basis ve tor expansion ( f. [1℄) or, in mathemati al terms, the
Gelfand-Maurin Theorem ( f. [5℄). This theorem is valid under ertain onditions on the
spa e . Sin e the eigenve tor de omposition is essential in quantum physi s, we will only
onsider spa es for whi h this theorem holds.
To explain the Gelfand-Maurin Theorem in detail requires mu h more mathemati s.
These mathemati s are provided in Se tion 3.5.2. In this se tion, we just give an intuitive
statement, whi h an be a epted in analogy to (3.5.10).
96 3 The Rigged Hilbert Spa e of the Harmoni Os illator

We onsider the ases of an innite dis rete number of eigenvalues and a ontinuous set
of eigenvalues in parallel. The self-adjoint operator with dis rete spe trum will be alled
H . Its spe trum will onsist of the innitely many real eigenvalues En , n = 0; 1; 2; : : : The
self-adjoint operator with ontinuous spe trum will be denoted by Q.2 The spe trum of
Q will be the ontinuous interval of real numbers [m; M ℄. Then the (heuristi ) spe tral
theorem asserts:
There exists a system of eigenve tors, jEn) in the dis rete ase and jxi in the ontinuous
ase,
H jEn) = En jEn ); n = 0; 1; 2; : : : ; (3.5.12)
Qjxi = xjxi; 1 < m  x  M < +1; (3.5.13)
su h that every ' 2  an be expanded in terms of these eigenve tors,
1
X
' = jEn)(Enj'); (3.5.14)
n=0
Z M
' = dxjxihxj'i ; (3.5.15)
m
and ' = 0 if and only if all its omponents are zero, i.e., (En j') = 0 for all En or hxj'i = 0
for all x.3 A system of eigenve tors jEn ) or jxi with these properties is alled omplete or a
basis system. The jxi are alled kets, the hxj are alled bras and the hxj'i are alled bra-kets
(see [1℄). The bra-ket hxj'i is a generalization of the usual s alar produ t.
Thus the spe tral theorem asserts the existen e of a omplete system of eigenve tors of
a self-adjoint operator. (En j') or hxj'i are alled the oordinates or omponents of ' with
respe t to the basis system fjEn)g or fjxig. They an be thought of, in analogy to the
N -dimensional ase stated in Eq. (3.5.11), as the s alar produ ts of the eigenve tors with
the ve tor '
(En j') = (jEn ); ') ; (3.5.16)
hxj'i = (jxi; ') : (3.5.17)
Thus the (En j') are the dis rete innite dimensional generalization of the 'i in (3.5.11),
and the hxj'i are the ontinuous innite dimensional generalizations of the 'i .
Whereas the jEn ) are proper eigenve tors, i.e., they are normalizable, the jxi are not.
This is why the jxi are alled generalized eigenve tors or eigenkets and denoted by orner-
kets j
i, in ontrast to the normalizable eigenve tors jEn), that are denoted by round-
kets j
). Though we an manipulate the orner-kets as if they were proper eigenve tors,
For instan e, H an be the Hamiltonian for the harmoni os illator and Q the position operator for the
2

harmoni os illator.
3
The simple nondegenerate form (3.5.14), (3.5.15) is valid if the operator A (H or Q) is y li , i.e., if
there exists an f 2  su h that fAn f  f(n) g spans the entire spa e . This means that any ' 2  an be
P
written as ' = n f(n) (n) , where (n) are omplex numbers. Degenerate spe tra, whi h o ur when more
than one quantum number is needed, will be dis ussed in Se tion 3.7.2.
3.5 Basis Systems, Eigenve tor De omposition and the Gelfand-Maurin Theorem 97

mathemati ally there is an important dieren e between the dis rete basis ve tors jEn ) and
the ontinuous basis ve tors jxi: the jEn ) are in  while the jxi are in , the spa e of
ontinuous antilinear fun tionals over . Further, (En j') is indeed the s alar produ t of
the normalized ve tor jEn) with ', whereas hxj'i is the a tion of the fun tional jxi at the
ve tor '.
If an operator H has dis rete spe trum En , n = 0; 1; 2 : : :, then all the orresponding
eigenve tors jEn ) enter into the dis rete basis ve tor expansion (3.5.14) and there are no
other eigenve tors that enter into this basis ve tor expansion. If an operator Q has a
ontinuous spe trum, then in general|and this depends upon the properties of the spa e
|there are more generalized eigenve tors of Q (i.e., more kets whi h fulll (3.5.13)) than
appear in the eigenve tor expansion (3.5.15). Whereas the dis rete eigenvalues of a self-
adjoint operator are always real, the generalized eigenvalues of a self-adjoint operator need
not be real. They an be real or omplex, and even if they are real, they do not ne essarily
belong to the spe trum, i.e., appear in the integral (3.5.15). However, for a self-adjoint
operator there is always a real subset of the set of generalized eigenvalues su h that the set
of orresponding eigenve tors is omplete.
The most general form of the spe tral theorem for an operator A representing a physi al
observable is a ombination of (3.5.14) and (3.5.15)
X Z
'= jai )(aij') + dajaihaj'i ; (3.5.18)
i

where the sum runs over the dis rete spe trum of A and the integral runs over the ontinuous
spe trum of A.4 This is the ase for the Hamiltonian of the hydrogen atom. It is possible
that some or all of the values ai appearing in the sum also appear in the integral. In that
ase, they are alled dis rete eigenvalues embedded in the ontinuous spe trum. If this
happens for ak , then jak ) is still orthogonal to all the jai in luding jak i
(ak jai = 0 ; (ak jak i = 0 : (3.5.19)

To see that the oordinates (En j') are indeed what their notation indi ates, namely the
s alar produ t of the ve tor ' with the basis ve tor jEn), we al ulate the s alar produ t of
(3.5.14) with the eigenve tor jEm ):
1
X
(jEm ); ') = (jEm ); jEn))(En j') : (3.5.20)
n=0
4
The nu lear spe tral theorem for an arbitrary self-adjoint operator a tually does not assert (3.5.18)
but rather (3.5.43) with a general measure d(x), and it does not say anything about the spe tral measure
d(x) in addition to the assertion of its existen e. However, all operators used in this dissertation are of the
P
spe ial kind that either d(x) = (x)dx with (x) a positive smooth measurable fun tion (su h operators
are said to have an absolutely ontinuous spe trum) or d(x) = i Æ (x xi )dx (these are the operators
with dis rete spe trum), or they have both an absolutely ontinuous and a dis rete spe trum. So (3.5.18),
after a normalization hange (3.5.45), is the most general form used in this dissertation.
98 3 The Rigged Hilbert Spa e of the Harmoni Os illator

Sin e jEm ) and jEn) are eigenve tors of the same Hermitian operator H ,
(jEm ); jEn)) = 0 if En 6= Em : (3.5.21)
For En = Em we normalize them
(jEn); jEn )) = k jEn) k2 = 1 : (3.5.22)
We ombine (3.5.21) and (3.5.22) and obtain
(jEm ); jEn)) = (Em jEn ) = ÆEn Em = Ænm ; n; m = 0; 1; 2; : : : : (3.5.23)
Thus, the eigenve tors of the self-adjoint operator H have the property (3.5.5) as required
for orthonormal basis ve tors. Inserting (3.5.23) into (3.5.20) one obtains
1
X
(jEm ); ') = Æmn (En j') = (Em j') : (3.5.24)
n=0

This is the expe ted identity (3.5.16).

The spe tral theorem (3.5.14) leads to other spe tral de ompositions: one an omit the
arbitrary ve tor ' 2  on both sides of (3.5.14) and obtain the spe tral resolution of the
identity operator I
1
X
I= jEn)(Enj ; (3.5.25)
n=0
where the quantities jEn )(En j are alled proje tion operators. One an also apply the oper-
ator H to both sides of (3.5.14)
1
X 1
X
H' = H jEn )(En j') = En jEn )(En j') ; (3.5.26)
n=0 n=0

and then omit the arbitrary ve tor ' on both sides

1
X
H= En jEn )(En j : (3.5.27)
n=0

This identity between the operator H and the weighted sum of proje tion operators jEn )(En j
is alled the spe tral resolution of the self-adjoint operator H with a dis rete spe trum.
The s alar produ t of any two elements '; 2  an be written in terms of the ompo-
nents of these ve tors along the basis ve tors jEn ) as
1
X 1
X
( ; ') = ( jEn)(En j') = (En j ) (En j') : (3.5.28)
n=0 n=0
3.5 Basis Systems, Eigenve tor De omposition and the Gelfand-Maurin Theorem 99

In parti ular, if one hooses = ', then one obtains

1
X 1
X
k'k 2 = ('; ') = ('jEn )(En j') = j(Enj')j2: (3.5.29)
n=0 n=0
Equation (3.5.28) is the analog to the formula
3
X 3
X
v
x= vi ei
ej xj = vi xi (3.5.30)
i;j =1 i=1

for the ordinary s alar produ t in R 3 .

In the three-dimensional spa e R 3 , a ve tor v is spe ied by its omponents (v1 ; v2 ; v3 )
with respe t to the basis system fe1 ; e2 ; e3 g. Moreover, any sequen e of three real numbers
(v1 ; v2 ; v3 ) determines a ve tor v. In the dis rete innite dimensional ase, any ve tor ' is
also ompletely spe ied by its omponents (En j') with respe t to a given basis jEn). But,
unlike the three-dimensional ase, the sequen e of omponents
((E0 j'); (E1 j'); (E2 j'); : : : ; (En j'); : : :) (3.5.31)
is not arbitrary but has to fulll
1
X
j(Enj')j2 < 1 ; (3.5.32)
n=0
i.e., it must be square summable ( f. Eq. (3.5.29)). The sequen e (3.5.31) is alled the
realization of ' by the sequen e of its omponents. In the ase of a Hilbert spa e H, ea h
' 2 H is asso iated to a sequen e of omponents (3.5.31) that fullls (3.5.32), and H is said
to be realized by the spa e of innite square summable sequen es.
The elements of the spa e  are usually required to satisfy further onditions besides
(3.5.32). The ' 2  must be su h that every operator A representing a physi al observable
is dened on the whole spa e . This implies that A' must be well dened for ea h ' 2 ,
i.e., (A'; A') must be nite. If the observable under onsideration is represented by the
operator H , then H and all of its powers must be well dened on , i.e., (H p'; H p') must
be nite for ea h p = 0; 1; 2; : : : This leads to
1
X
(H p'; H p') = ('jH pjEn )(En jH pj')
n=0
X1
= En2p j(En j')j2 < 1 for any p = 0; 1; 2; : : : (3.5.33)
n=0
Thus, not only f(Enj')g1n=0 but also fE (En j')gn=0 , p = 1; 2; : : :, have to be square
p 1
summable for any ' 2 . When ea h element of a spa e  is asso iated to a sequen e
' $ ( (E0 j'); (E1 j'); : : : ; (En j'); : : :) (3.5.34)
100 3 The Rigged Hilbert Spa e of the Harmoni Os illator

whi h fullls (3.5.33), we say that  is realized by the spa e of rapidly de reasing innite
sequen es.
We now turn to the ontinuous spe trum ase and repeat the above onsiderations for
this ase. We al ulate the s alar produ t of ' with the generalized eigenve tor jxi 5 using
Eq. (3.5.15) Z +1
(jxi; ') = dy (jxi; jy i) hy j'i ; (3.5.35)
1
where we have assumed that the ontinuous spe trum runs over the whole real line (whi h
is the ase for e.g. the position operator Q of the harmoni os illator). Using the denition
hxjyi  (jxi; jyi) ; (3.5.36)
Eq. (3.5.35) an be rewritten as
Z +1
(jxi; ') = dy hxjy ihy j'i : (3.5.37)
1
In analogy to the N -dimensional ase (3.5.10), hxj'i is the omponent of the ve tor ' along
the dire tion of the basis ve tor jxi, while (jxi; ') is the \s alar produ t" of ' with the basis
ve tor jxi. This analogy also suggests that the omponent of ' along jxi should be given
by the \s alar produ t" of ' with jxi, i.e., it should be of the form hxj'i  (jxi; '). Then,
one would have Z +1
hxj'i = dy hxjy i hy j'i : (3.5.38)
1
The omponents hy j'i are fun tions of the ontinuous variable y in the same way as the s alar
produ ts (En j') are fun tions of the dis rete variable En . Equation (3.5.38) therefore says
that the mathemati al quantity hxjy i has the property that it maps the fun tion '(y ) =
hyj'i by integration into '(x) = hxj'i, its value at the xed point x. There is no well-
behaved (or even a lo ally integrable) fun tion hxjy i whi h has the property (3.5.38).
Quantities like hxjy i, whi h are dened by integration, are alled distributions or gener-
alized fun tions (see, for example, [67℄). The distribution hxjy i dened by (3.5.38) is alled
the Dira delta fun tion (though it is not a fun tion). It is denoted in analogy to (3.5.23)
by
hxjyi = Æ(x y) : (3.5.39)
Æ (x y ) is the ontinuous analog to ÆEn Em : the Krone ker delta is usually dened by (3.5.2),
but it ould as well have been dened as the fun tion that maps any sequen e f(En j')g by
summation into (Em j'), the m-th omponent of the sequen e,
1
X
(Em j') = ÆEm En (En j') : (3.5.40)
n=0
5
More pre isely, we should say that we al ulate the value (jxi; ') of the fun tional jxi at the ve tor
' 2 .
3.5 Basis Systems, Eigenve tor De omposition and the Gelfand-Maurin Theorem 101

Dening the Krone ker delta by (3.5.40) gives a learer analogypto (3.5.38).
The jxi are not dimensionless. They have the dimension 1= dim(dx). For example, if
dx has the dimension m, then hx0 jxi has the dimension m 1 , and jxi has the dimension
m 1=2 .
The eigenve tors jEn) are normalized a ording to (3.5.23). The normalization (3.5.39)
for the generalized eigenve tors jxi is alled Æ -fun tion normalization.
Instead of the generalized eigenve tors with Æ -fun tion normalization (3.5.39), one an
also hoose generalized eigenve tors of Q with a dierent normalization. One an dene a
new set of generalized eigenve tors by

jxg := jxi p 1 ; (3.5.41)

(x)
where (x) is a real nonnegative and integrable fun tion. These new kets are still eigenve -
tors of the operator Q,
Qjxg = xjxg : (3.5.42)
Now instead of (3.5.15), the eigenfun tion expansion in terms of the jxg kets reads
Z
'= d(x) jxg fxj'i ; (3.5.43)

where d(x) = (x) dx. In order for the new omponents of ' (the fxj'i) to be the s alar
produ t of ' with the new eigenve tors jxg , i.e., in order that
Z
fxj'i = d(y) fxjyg fyj'i ;
 (3.5.44)

one has to demand

d(y ) fxjy g = dy hxjy i = dy Æ (x y ) : (3.5.45)
So, the normalization of the new generalized eigenve tors is
  1
fxjyg = ddy(y) Æ(x y) =  1(y)Æ(x y):
 (3.5.46)

Thus, when the integration ontains the weight fun tion (x), the generalized eigenve tor
normalization will ontain the fa tor  1 (x).
The most appropriate hoi e for (x) depends upon the property of the operator Q and
its relation to the other operators of the problem. The hoi e (x) = 1 is not always the
most onvenient. For instan e, it may be onvenient to hoose (x) in su h a way that the
measure d(x) in (3.5.43) is invariant under some important symmetry transformation of
the physi al system.
The s alar produ t of ' 2  with 2  is given in the notation of (3.5.38) by
Z +1
( ; ') = h j 'i = dx h jxi hxj'i : (3.5.47)
1
102 3 The Rigged Hilbert Spa e of the Harmoni Os illator

In the same way Z +1

('; ) = dx h'jxi hxj i : (3.5.48)
1
Taking omplex onjugates on both sides of (3.5.48), we get
Z +1
('; ) = dx h'jxi hxj i : (3.5.49)
1
But the s alar produ t fullls
( ; ') = ('; ) : (3.5.50)
Thus, omparing (3.5.47) and (3.5.49) we see that the following relation is natural:
hxj i = h jxi ; 2  ; jxi 2  : (3.5.51)
Using the notation
hxj'i  '(x) (3.5.52)
and (3.5.51), one an write (3.5.47) in the form
Z +1
( ; ') = dx (x) '(x) : (3.5.53)
1
In parti ular, if one hooses = ', one obtains
Z +1 Z +1
k'k 2 = ('; ') = dx '(x) '(x) = dx j'(x)j2 : (3.5.54)
1 1
This implies that the omponents of a ve tor ' 2  with respe t to the ontinuous basis
system jxi annot be given by any arbitrary fun tion '(x). Rather, they an be given only
by those fun tions for whi h the integral on the right-hand side of Eq. (3.5.54) exists, i.e.,
the fun tions '(x) must be at least square integrable.
The asso iation of ' with its omponents
 3 ' $ hxj'i = '(x) (3.5.55)
is alled the realization of the spa e  by the spa e of fun tions '(x). The fun tion '(x) =
hxj'i is alled the wave fun tion of the ve tor '; in parti ular, if Q is the position operator,
then '(x) is alled the position wave fun tion.
If the integrals in (3.5.53) and (3.5.54) are ordinary Riemann integrals, then the linear
spa e of fun tions '(x) is not omplete with respe t to the norm dened by (3.5.54). This
means that there will be sequen es of fun tions n (x) that are Cau hy sequen es,
Z +1
k n m k2 = dx j n (x) m (x) j2 ! 0 as n; m ! 1 ; (3.5.56)
1 Riemann
3.5 Basis Systems, Eigenve tor De omposition and the Gelfand-Maurin Theorem 103

but for whi h there exists no fun tion (x) whi h is the limit element of that sequen e,
Z +1
k n k 2 = dx j n(x) (x)j2 ! 0 as n ! 1 : (3.5.57)
1 Riemann
In order to obtain a omplete spa e of fun tions with respe t to the s alar produ t
(3.5.53), one has to repla e the Riemann integrals in (3.5.56) and in (3.5.57) by Lebesgue
integrals. Then, for any sequen e of (Lebesgue) square integrable fun tions n (x) satisfying
Z +1
k n m k2 = dx j n (x) m (x) j2 ! 0 as n; m ! 1 ; (3.5.58)
1 Lebesgue
there is always a (Lebesgue) square integrable fun tion (x) satisfying
Z +1
k n k2 = dx j n(x) (x)j2 ! 0 as n ! 1 : (3.5.59)
1 Lebesgue
The ( omplete) spa e of Lebesgue square integrable fun tions is denoted by L2 (R ). Just
like the spa e of innite square summable sequen es, the spa e L2 (R ) of square integrable
fun tions is a realization of the Hilbert spa e.
The elements of the spa e  of physi al states ' are required, in addition to be square
normalizable, to be su h that the operator Q and all of its powers be well dened on every
' 2 . Then one must have
Z +1
k k
Qp ' 2 = (Qp '; Qp ') = dx x2p j'(x)j2 < 1 ; p = 0; 1; 2; : : : (3.5.60)
1
Thus, the fun tions '(x) that belong to the realization that fullls (3.5.60) must de rease
faster than any power of 1=x. If other operators are also to be dened everywhere on ,
further onditions will have to be imposed on the omponents hxj'i of ' 2 . Consequently,
the realization of  by a spa e of fun tions must be a subset of L2 (R ).
For instan e, we an dene a linear operator P on the spa e  by giving a pres ription
for its a tion on ea h fun tion '(x) whi h realizes a ve tor ' 2 . Let this operator be
dened by
d
hxj'i ! hxjP 'i := 1i dx hxj'i (3.5.61)
for every ' 2 . We also dene, a ording to (3.5.13), the operator Q on  by
hxj'i ! hxjQ'i := x hxj'i (3.5.62)
for ea h ' 2 . If we demand that Qp and P p, p = 0; 1; 2; : : :, are well dened at every
' 2 , then the fun tions '(x) must be innitely dierentiable, rapidly de reasing fun tions,
i.e., in addition to (3.5.60) the '(x) must be in C 1 (R ) and must satisfy
Z +1  p 
 d '(x) 2
k k
P p' 2 = (P p'; P p') = dx 
 dxp 
 < 1 ; p = 0; 1; 2; : : : (3.5.63)
1
104 3 The Rigged Hilbert Spa e of the Harmoni Os illator

The requirement that Qq P p', q; p = 0; 1; 2; : : :, is well dened leads to the onditions

Z +1  
 q dp '(x) 2
k k
Qq P p' 2 = dx x
 dxp 
 < 1 ; p; q = 0; 1; 2; : : : : (3.5.64)
1
This means that the realization of the spa e  is the linear spa e of innitely dieren-
tiable omplex-valued fun tions whi h together with their derivatives vanish at innite more
rapidly than any power of 1=x. This spa e is the S hwartz spa e S (R ), and we all these
fun tions well behaved. Conditions (3.5.64) are equivalent to
m
lim x n d '(x) = 0 ; n; m = 0; 1; 2; : : : (3.5.65)
x!1 dxm
The spa e S (R ) is not omplete with respe t to the norm onvergen e dened through
(3.5.54). In fa t, its ompletion with respe t to this norm is the spa e L2 (R ). However,
S (R ) is a omplete ountably Hilbert spa e with respe t to the topology generated by the
ountable number of s alar produ ts
Z 1  
2 2 1 p d2 2 1 p
( ; ')p := ( ; (P + Q + I ) ') = dx (x) +x + '(x) ; p = 0; 1; 2; : : :
2 1 dx2 2
(3.5.66)
A sequen e 'n 2 S (R ) onverges to ' 2 S (R ) with respe t to this topology if
k'n 'kp ! 0 ; p = 0; 1; 2; : : : ;
n!1
(3.5.67)
p
where k'kp = ('; ')p.
In Se tion 3.6.3, we shall show that the S hwartz spa e is the realization of the spa e 
for the harmoni os illator.
The Lebesgue integral, though mathemati ally well dened, is not easy to handle in
pra ti al omputations, while Riemann integrations are easy to al ulate. In physi s, the
Riemann integral is the one that is ex lusively used to perform omputations. The Hilbert
spa e uses Lebesgue integration, whereas the spa e  uses Riemann integration. This makes
 a mu h simpler spa e to work with and mu h more suitable for representing the physi al
wave fun tions than the Hilbert spa e. As an example, the integrals (3.5.54) in L2 (R )
are Lebesgue integrals, whereas the integrals (3.5.66) in S (R ) are Riemann integrals. This
makes S (R ) a mu h easier spa e to handle than L2 (R ).

3.5.2 Gelfand-Maurin Theorem

We are now in a position to address the Gelfand-Maurin Theorem, whi h provides the
mathemati al justi ation for the heuristi Dira basis ve tor expansion.
Before stating the Gelfand-Maurin Theorem, let us review the situation in nite dimen-
sional spa es and in the Hilbert spa e H.
A nonzero ve tor h in or in H is alled an eigenve tor of an operator A dened on
or on H if there exists a omplex number , alled the eigenvalue, su h that
Ah = h : (3.5.68)
3.5 Basis Systems, Eigenve tor De omposition and the Gelfand-Maurin Theorem 105

Theorem Every self-adjoint operator A dened on an N -dimensional s alar produ t spa e

has a omplete system of orthonormal eigenve tors hi = ji) 2 , i = 1; 2; : : : ; N ,
Aji ) = i ji) ; (3.5.69)
(i jj ) = Æij ; (3.5.70)
su h that every h 2 an be expanded as
N
X
h= ji)(ijh) : (3.5.71)
i=1
The set Sp(A) = f1 ; 2 ; : : : ; N g of the eigenvalues of A is alled the spe trum of A.
For an innite dimensional Hilbert spa e H this statement is no longer true. For example,
it is well known that the dierential operator
d'
P '(x) = i (x) (3.5.72)
dx
and the multipli ation operator
Q'(x) = x'(x) (3.5.73)
have no eigenve tors in L2 (R ). However, there is a lass of operators in the Hilbert spa e
alled ompa t operators for whi h the generalization of (3.5.69)-(3.5.71) holds.
A bounded operator A on a Hilbert spa e H is said to be ompa t if for every bounded
sequen e fhn g  H, fAhn g has a subsequen e onvergent in H. A self-adjoint ompa t
operator A on an innite dimensional Hilbert spa e H has only a dis rete spe trum whi h
oin ides with the set of its eigenvalues, Sp(A) = f1 ; 2 ; : : : ; n ; : : :g. In this ase, the
statements (3.5.69)-(3.5.71) for operators on nite dimensional spa es arry over to the
innite dimensional ase.
Theorem For any ompa t self-adjoint operator A dened on a Hilbert spa e H, there
exists an orthonormal set of eigenve tors hi  ji ), i 2 Sp(A) = f1 ; 2 ; : : : ; n ; : : :g,
Aji ) = i ji) ; (3.5.74)
(i jj ) = Æij ; (3.5.75)
su h that every h 2 H an be expanded as
1
X
h= ji)(ijh) ; (3.5.76)
i=1
where ea h eigenvalue i is repeated a ording to its (nite) multipli ity.
The observables that usually appear in Quantum Me hani s are des ribed by unbounded
operators dened on some dense subdomains of the Hilbert spa e. In this ase, the (Hilbert
106 3 The Rigged Hilbert Spa e of the Harmoni Os illator

spa e) spe trum of the operator is not dis rete in general, but has a ontinuous part ( f. Se -
tion 2.5.3). For example, P and Q in (3.5.72) and (3.5.73) ea h has a ontinuous spe trum
whi h oin ides with the real line. When the spe trum of an operator has a ontinuous
part, the spe tral de ompositions (3.5.71) and (3.5.76) are no longer valid be ause there are
no ve tors in the Hilbert spa e that are eigenve tors orresponding to eigenvalues in the
ontinuous part of the spe trum. In order to extend (3.5.71) and (3.5.76) to the ase where
the spe trum has a ontinuous part, we need the Gelfand-Maurin Theorem whi h is stated
below.
Dira formalism is the way physi ists handle ontinuous spe trum. For instan e, the
operators P and Q, as dened in (3.5.72) and (3.5.73), have a ontinuous spe trum that
overs the whole real line but do not have any eigenve tor in H. Nevertheless, physi ists,
following Dira , always asso iate an eigenket to ea h element of the ontinuous spe trum of
P and of Q,
P j pi = pj pi ; p 2 R ; (3.5.77)
Qjq i = q jq i ; q 2 R ; (3.5.78)
and use the assumption that these eigenve tors form a \ omplete" system in the sense that
every ' an be written as
Z +1
' = dp jpihpj'i ; (3.5.79)
1
Z +1
' = dq jq ihq j'i : (3.5.80)
1
When we omit ' in (3.5.79) and in (3.5.80), we obtain the resolution of the identity
Z +1
I = dp jpihpj ; (3.5.81)
1
Z +1
I = dq jq ihq j : (3.5.82)
1
However, the kets jpi and jq i are not in the Hilbert spa e, and the eigenvalues p and q are
not proper eigenvalues. The eigenkets orresponding to elements in the ontinuous spe trum
of an operator a quire mathemati al meaning as generalized eigenve tors orresponding to
generalized eigenvalues in the sense of the following denition:
Denition Let A be a  - ontinuous operator on  and A its dual extension to . A
generalized eigenve tor of the operator A orresponding to the generalized eigenvalue  is
an antilinear fun tional F 2  su h that
F (A') = F (') ; 8' 2  ; (3.5.83)
or in bra-ket notation
hA'jF i = h'jAF i = h'jF i : (3.5.84)
3.5 Basis Systems, Eigenve tor De omposition and the Gelfand-Maurin Theorem 107

Equations (3.5.83) and (3.5.84) are often written as

A F = F or A jF i = jF i ; (3.5.85)
respe tively. Following Dira , the ket in (3.5.85) is often labeled by its eigenvalue, jF i  ji.
Then we write
A ji = ji : (3.5.86)
If A is essentially self-adjoint, we may also write
Aji = ji : (3.5.87)

Let us assume that A has a generalized eigenve tor in the Hilbert spa e, i.e., F  f in
equation (3.5.84) is an element of H. Then (3.5.84) reads
(A'; f ) = ('; Ayf ) = ('; f ) (3.5.88)
for every ' 2 . Sin e  is H -dense in H, (3.5.88) implies that
A F  Ayf = f : (3.5.89)
Thus, a generalized eigenve tor whi h is also an element of the Hilbert spa e is an ordinary
eigenve tor of the Hilbert spa e adjoint operator orresponding to the same eigenvalue.
To avoid ompli ations whi h are inessential for our main purposes and inappli able for
the parti ular problem of the one-dimensional harmoni os illator, we restri t ourselves here
to y li operators.
Denition An operator A dened on a subdomain D(A) of a Hilbert spa e H is y li if
there exists a ve tor f 2 D(A) su h that fAk f g1
k=0 spans the entire Hilbert spa e.

For instan e, the operators P and Q in (3.5.72) and (3.5.73) are y li be ause the sets
fQk 0 ; k = 0; 1; 2; : : :g (3.5.90)
and
fP k 0 ; k = 0; 1; 2; : : :g ; (3.5.91)
where 0 is the zero-th Hermite polynomial, both span H.
If A has a ontinuous spe trum, the spe tral de omposition (3.5.76) is not possible in the
Hilbert spa e. However, a generalization of (3.5.76), alled the Gelfand-Maurin Theorem or
the Nu lear Spe tral Theorem, is possible in the Rigged Hilbert Spa e.
Theorem (Gelfand-Maurin Theorem or Nu lear Spe tral Theorem) Let   H   be
a Rigged Hilbert Spa e and A a y li , e.s.a.,  - ontinuous operator. Then, for ea h  in
the spe trum of A, there exists a generalized eigenve tor F  ji,
A ji = ji ;  2 Sp(A) ; (3.5.92)
108 3 The Rigged Hilbert Spa e of the Harmoni Os illator

i.e.,
hA'ji = h'jAji = h'ji ; 8' 2  : (3.5.93)
Furthermore, there is some uniquely dened positive measure d() on Sp(A) su h that for
every '; 2  Z
( ; ') = d()h jihj'i ; (3.5.94)
Sp(A)
where
hj'i = h'ji : (3.5.95)
Furthermore, if f () is a well-behaved fun tion on Sp(A), then
Z
( ; f ( A ) ') = d()h jihj'if () (3.5.96)
Sp(A)

If we set = ' in (3.5.94), we see that if all the omponents hj'i of the spe tral
de omposition of ' with respe t to the operator A vanish, then k'k = 0, i.e., ' = 0.
Be ause of this property, the set of generalized eigenve tors ji o urring in (3.5.92) is
alled omplete in analogy to the ompleteness of the system of ordinary eigenve tors in a
Hilbert spa e.
In general, an e.s.a. operator has more generalized eigenve tors than those that appear in
the spe tral de omposition (3.5.94). In parti ular, a generalized eigenvalue may be omplex.
The spe tral de omposition provided by the Gelfand-Maurin theorem needs not be
unique and it is also valid in many ases when  is not a nu lear spa e.
The statement of the Gelfand-Maurin Theorem is still too general for the purposes of
Quantum Me hani s. For the ases of physi al interest, the measure d() that appears in
this theorem has a dis rete and an absolutely ontinuous part, i.e., it an be written as
X
d() = (i )Æ ( i ) + ()d : (3.5.97)
dis rete
spe trum

After a delta-normalization given by (3.5.41), equations (3.5.94) and (3.5.97) yield

X Z
( ; ') = ( ji)(i j') + ontinuous d h jihj'i ; '; 2  : (3.5.98)
dis rete spe trum
spe trum

Eq. (3.5.98) is the form in whi h the Gelfand-Maurin theorem is used in physi s be ause
it is pre isely the Dira basis ve tor expansion. This expansion treats the elements of the
dis rete spe trum and the elements of the ontinuous spe trum of the operator A on the
same footing: there is always an eigenve tor orresponding to ea h element of the spe trum
of the operator. If this element belongs to the dis rete spe trum, then the orresponding
eigenket is an ordinary eigenve tor (i.e., it is square normalizable). If the element is in the
ontinuous part of the spe trum, then the orresponding eigenket is a generalized eigenve tor
3.6 Gelfand-Maurin Theorem Applied to the Harmoni Os illator 109

(i.e., it is a fun tional). It is worthwhile noting that the spe tral de omposition (3.5.98) is
only valid for elements ; ' in the spa e , but not for every element h in the Hilbert spa e
H.
The Gelfand-Maurin Theorem provides a mathemati al rephrasing of several formal
expressions used in Quantum Me hani s. For instan e, omission of in (3.5.98) leads to
the spe tral de omposition of any wave fun tion ' 2  ,
X Z
'= ji)(ij') + djihj'i : (3.5.99)
i

In the same way, we an obtain the spe tral resolution of the identity operator
X Z
I= ji)(ij + d jihj (3.5.100)
i

and of the operator itself

X Z
A= i ji )(i j + d  jihj : (3.5.101)
i

As an example, Eq. (3.5.79) is a parti ular ase of (3.5.94) with Sp(P ) = R ,  = p 2 R

and d() = dp, the Lebesgue measure on R , and similarly for the position operator.

3.6 Gelfand-Maurin Theorem Applied to the Harmoni

Os illator
In this se tion, we will apply the Gelfand-Maurin Theorem to the operators of the algebra of
observables of the harmoni os illator. We will show that the dening algebrai assumptions
(3.2.1)-(3.2.4) and the  - ontinuity of the algebra of observables lead to the S hrodinger
representation in the S hwartz spa e S (R ). The operators P , Q and H will be realized by
the standard dierential operators and the spa e  will be realized by the S hwartz spa e.

3.6.1 Spe tral Theorem Applied to the Energy Operator

We now re all the spe tral properties of H . These spe tral properties were derived in the
onstru tion of the RHS for the harmoni os illator.
The spe trum of H is the dis rete set
1
Sp(H ) = fEn = ~! (n + ) ; n = 0; 1; 2; : : :g : (3.6.1)
2
Corresponding to ea h eigenvalue En , there is an eigenve tor n  jn) of H :
H jn) = En jn) : (3.6.2)
110 3 The Rigged Hilbert Spa e of the Harmoni Os illator

The eigenve tors jn) are proper eigenve tors, i.e., jn) 2 , rather than generalized eigenve -
tors, i.e., elements of  . These eigenve tors form a basis system for , i.e., every ' 2 
an be written as 1
X
'= jn)(nj') ; (3.6.3)
n=0
where the omponents (nj') 2 C satisfy
1
X
(n + 1)p j(nj')j2 < 1 (3.6.4)
n=0
for every p = 0; 1; 2; : : :

3.6.2 Spe tral Theorem Applied to the Position and Momentum

Operators
We now want to al ulate the spe tra of the operators Q and P and the generalized eigen-
ve tors that orrespond to the elements of these spe tra. These generalized eigenve tors
will be ontinuous antilinear fun tionals over the spa e  onstru ted in Se tion 3.3.
We rst determine whether the Gelfand-Maurin Theorem of Se tion 3.5.2 an be applied
to the operators Q and P . The spa e , whose topology is dened by the ountable number
of s alar produ ts
('; )p = ('; (N + I )p ) ; (3.6.5)
was proved to be a nu lear spa e ( f. Se tion 3.3.5). The operators position Q and momen-
tum P were proved to be  - ontinuous ( f. Se tion 3.3.5). Q and P are y li operators,
sin e the sets
fQn0 j n = 0; 1; 2; : : :g (3.6.6)
and
fP n0 j n = 0; 1; 2; : : :g ; (3.6.7)
where 0 is the zeroth eigenve tor of the Hamiltonian operator, both span the whole of H.
Therefore, we an apply the Gelfand-Maurin to these operators. This theorem assures the
existen e of a omplete set of generalized eigenve tors of the operators Q and P ,
Q jxi = xjxi ; x 2 Sp(Q) ; (3.6.8)
P  jpi = pjpi ; p 2 Sp(P ) ; (3.6.9)
and either of these two sets an be used for the spe tral de omposition of any ve tor ' in
: Z
'= d(x) jxihxj'i ; (3.6.10)
Sp(Q)
or Z
'= d(p) jpihpj'i ; (3.6.11)
Sp(P )
3.6 Gelfand-Maurin Theorem Applied to the Harmoni Os illator 111

where d(x) and d(p) are measures on Sp(Q) and Sp(P ), respe tively.
It is well known that the spe trum of both P and Q is the real line. However, it is not
so widely known that the derivation of this is far from being trivial (see [68℄ and referen es
therein). The approa h of a physi ist is usually the reverse of the one des ribed here, namely
a physi ist nds the dening assumptions (3.2.1)-(3.2.4) of the harmoni os illator from the
spe tra of Q and P , whi h are onje tured from experimental data to be the real line. We
shall derive these spe tra in the present se tion. We will see that d(x) = dx, Sp(Q) = R ,
d(p) = dp and Sp(P ) = R . We shall see that the set of generalized eigenvalues of Q (of
P ) agrees with the spe trum of Q (of P ) when we hoose for the spa e  in the RHS
the ountably Hilbert spa e dened by the ountable number of s alar produ ts (3.6.5).
However, sin e the set of generalized eigenvalues depends upon the hoi e of the spa e ,
we ould hoose a dierent ountably Hilbert spa e and obtain also omplex generalized
eigenvalues of the essentially self-adjoint operator Q (or P ).
We begin by examining for whi h values x 2 C the equation
Q jxi = xjxi (3.6.12)
an be fullled, i.e., for whi h omplex numbers x the equation
hQ'jxi = h'jQjxi = xh'jxi (3.6.13)
holds for every ' 2 . Sin e every ' 2  an be expanded in terms of the basis of
eigenve tors n = jn) of H as in (3.6.3), it will be suÆ ient to know for whi h x the
equation
(njQ jxi = x(njxi (3.6.14)
holds for every jn).
From (3.2.5), (3.2.6) and (3.2.27) it follows that
s
~
Qjn) = (a + ay)jn)
2!
s
~  pn jn p 
= 1) + n + 1jn + 1) : (3.6.15)
2!
Taking the \s alar produ t" of this equation with jxi we obtain
s
~ p p
(njQ jxi = ( n (n 1jxi + n + 1(n + 1jxi) : (3.6.16)
2!
Equation (3.6.16) is mathemati ally well dened even though we used the term \s alar
produ t" of Qjn) with jxi, whi h is not well dened sin e jxi 2 . The pre ise
meaning of (3.6.16) is as follows: sin e jn) 2 , so is '  Qjn), be ause Q leaves 
invariant. Therefore, we an onsider the value of the fun tional jxi  Fx 2  at
112 3 The Rigged Hilbert Spa e of the Harmoni Os illator

the point ' 2 , Fx(') = h'jFx i. A ording to (3.4.25), this is related to the value
e
of the fun tional ' 2    at the point Fx 2  by
e
Fx (') = '(Fx ) ; (3.6.17)
or in bra-ket notation
e
h'jFx i = hFx j'i  hFx j'i ; (3.6.18)
e
where the identi ation  3 '  ' 2  was used. Returning to '  Qjn) = Qn,
(3.6.18) is written as
hxjQjn) = hxjQn i = hQnjxi = (njQjxi : (3.6.19)
This is the quantity that appears on the left hand side of (3.6.16). The quantities
(n 1jxi and (n + 1jxi are similarly dened by hoosing ' = jn 1) and ' = jn + 1)
respe tively.
On the other hand, taking the s alar produ t of (3.6.12) with jn), (more pre isely, the
a tion of the fun tional Q jxi at jn)) we obtain
(njQ jxi = x (njxi : (3.6.20)
Comparing (3.6.16) with (3.6.20) yields
s
~  p p 
x (njxi = n(n 1jxi + n + 1(n + 1jxi ; (3.6.21)
2!
or with n + 1 = m,
r
pm(mjxi = 2! p
x(m 1jxi m 1(m 2jxi : (3.6.22)
~
Sin e Eq. (3.6.15) is valid for n = 1; 2; : : : ; Eq. (3.6.22) is valid for m = 2; 3; : : : For n = 0
(m = 1), we obtain instead of (3.6.15)
s s
~ p ~
Qj0) = 0 + 1 j0 + 1) = j1) ; (3.6.23)
2! 2!
and instead of (3.6.22) r
p 2!
1 (1jxi = x (0jxi : (3.6.24)
~
Thus we see that (3.6.22) is a re urren e relation for (mjxi: if (0jxi is known, we an
determine (1jxi by (3.6.24) and then determine (2jxi by (3.6.22). With (1jxi and (2jxi we
an determine (3jxi by (3.6.22), and so on.
To nd out what the transition oeÆ ients (mjxi are, we introdu e
r
!
y x (3.6.25)
~
3.6 Gelfand-Maurin Theorem Applied to the Harmoni Os illator 113

and
p
(njxi
fn (y )  2n n! ; (3.6.26)
(0jxi
whi h is dened for all x su h that (0jxi 6= 0 (if (0jxi = 0, then by (3.6.22) and (3.6.24)
(njxi = 0 for all n). Then from (3.6.22) it follows that
r s s
m 2 m 1
f (y ) = y fm 1 (y ) f (y ) ; (3.6.27)
2m m! m 2m 1 (m 1)! 2m 2 (m 2)! m 2
or
fm (y ) = 2yfm 1 (y ) 2(m 1)fm 2 (y ) (3.6.28)
From (3.6.24) we have
f1 (y ) = 2yf0 (y ) ; (3.6.29)
and from (3.6.26)
f0 (y ) = 1 : (3.6.30)
Equations (3.6.28)-(3.6.30) are the re urren e relations for the Hermite fun tions and have
solutions for any omplex number y . Thus for any omplex value x there is a solution
(njxi of the re urren e relation (3.6.22). Sin e Q is an e.s.a. operator, the (Hilbert spa e)
spe trum of Q must be real ( f. Se tion 2.5.3). Therefore the generalized eigenvalues that
appear in the integral de omposition (3.6.10) must be real, and we need to onsider only
the solutions fm (y ) for y 2 R .
For real values of y , the solutions fm (y ) of (3.6.28)-(3.6.30) are the Hermite polynomials:
2
2 dn (e y )
fn (y ) = Hn (y ) = ( 1)n ey : (3.6.31)
dy n
Thus from (3.6.26) we an obtain the transition oeÆ ient (njxi for every real value of x for
whi h (0jxi is dened. We restri t ourselves to those solutions of (3.6.22) for whi h (0jxi is
nite, be ause j(0jxij2 , the probability for obtaining the value x in a measurement of Q in
the ground state 0 = j0 )(0 j, is assumed to be nite.
Combining (3.6.25), (3.6.26) and (3.6.31), we have
 !  r
1
(njxi = p n (0jxiHn x (3.6.32)
2 n! ~
for 1 < x < +1.
Sin e every ' 2  an be expanded as
1
X
'= jn)(nj') ; (3.6.33)
n=0
the jxi an be dened at ea h ' by
1
X
hxj'i = hxjn)(nj') : (3.6.34)
n=0
114 3 The Rigged Hilbert Spa e of the Harmoni Os illator

The quantities hxj'i  '(x) are alled the position wave fun tions or the wave fun tions in
the position representation. The quantities hxjn)  n(x) are alled the energy eigenfun -
tions, sin e they fulll
hxjH jn) = En hxjn) : (3.6.35)
Be ause of Eq. (3.6.34), the energy eigenfun tions hxjn) an be viewed also as \transition
elements" between the x- and the n-representation.
If we onsider the expansion
Z
n = d(x) jxihxjn) (3.6.36)
Sp(Q)

of the energy eigenve tors n in terms of the eigenkets of Q as a fun tional a ting on the
generalized eigenve tor Fx0 = jx0 i 2  , x0 2 Sp(Q), then a ording to (3.6.18) we obtain
from (3.6.36) Z
n (Fx0 ) = hx jni =
e 0 d(x) hx0 jxihxjn i : (3.6.37)
Sp(Q)
Thus d(x) hx0 jxi must be the Dira measure, i.e., the distribution dened by (3.6.37) must
have the property of the Dira delta-fun tion
d(x) hx0jxi = dx Æ (x0 x) : (3.6.38)
We now al ulate the s alar produ t of n = jn) and m = jm) using (3.6.36)
Z
Æmn = (m ; n ) = (mjn) = d(x) (mjxi hxjn) : (3.6.39)
Sp(Q)

We shall make use of

hxjn) = (njxi : (3.6.40)
We insert (3.6.32) and (3.6.40) into (3.6.39) and obtain
r Z r r
1 ! !
d(x) j(0jxij Hm (
2
x)Hn ( x) = Æmn : (3.6.41)
2 2 n!m!
n m
Sp(Q) ~ ~

Comparing (3.6.41) with the orthogonality relations for the Hermite polynomials,
Z +1
1 y 2 H (y )H (y )
n!2n
p dy e m n = Ænm ; (3.6.42)
1
and taking into a ount that the Hermite polynomials are only orthogonal polynomials if
asso iated with the interval 1 < y < +1 and the weight e y2 (one an dene Hn(y ) by
(3.6.42) and derive (3.6.28)-(3.6.30) for real y ) we on lude
r
! (!=~)x2
d(x) j(0jxij = dx
2 e (3.6.43)
~
3.6 Gelfand-Maurin Theorem Applied to the Harmoni Os illator 115

and
Sp(Q) = fx j 1 < x < +1g : (3.6.44)
If we agree to normalize the generalized eigenve tors su h that
hx0jxi = Æ(x0 x) ; (3.6.45)
then a ording to (3.6.38)
d(x) = dx : (3.6.46)
From (3.6.43) and (3.6.46) we on lude
r
! (!=~)x2
j(0jxij 2
=
~
e : (3.6.47)

Thus, up to an arbitrary phase fa tor (whi h we hoose to be unity),

 ! 1=4
(0jxi = : e (!=2~)x2
(3.6.48)
~
With this and (3.6.32) we obtain the transition oeÆ ients (njxi between the x- and n-
basis, i.e., the harmoni -os illator energy wave eigenfun tions n(x):
 ! 1=4   r
(njxi =
~
p 1n Hn ! x e (!=2~)x2 = hxjn) = n (x) : (3.6.49)
2 n! ~

We now repeat for the operator P the pro edure that we have gone through for the
operator Q. The generalized eigenve tors of P will be denoted by jpi:
P  j pi = pj pi : (3.6.50)
The a tion of P on jn) is, using (3.2.5), (3.2.6) and (3.2.27),
r
~!
P jn) = i (a ay )jn)
r
2
~! p p 
= i njn 1) n + 1jn + 1) : (3.6.51)
2
If we apply the fun tional jpi 2  on the ve tor P jn) 2  and use (3.6.50), we obtain
r
~!  p p 
phpj n ) = i nhpjn 1) n + 1hpjn + 1) ; (3.6.52)
2
or r
~!  p n (n p 
p( n j pi = i 1j pi n + 1(n + 1jpi : (3.6.53)
2
116 3 The Rigged Hilbert Spa e of the Harmoni Os illator

If we introdu e the new quantities (njpg dened by

(njpg := i n (njpi ; (3.6.54)
then
i(n 1jpi = in (n 1jpg ; (3.6.55)
i(n + 1jpi = in (n + 1jpg ; (3.6.56)
so (3.6.53) may be written as
r
~!  pn(n p 
p( n j pg = 1jpg + n + 1(n + 1jpg : (3.6.57)
2
p
We seepthat this is exa tly the same re urren e relation as in (3.6.21), with x !=~ repla ed
by p= ! ~. Thus by the same argument as for (njxi, we nd (using (3.6.54)) that
 1 1=4 1  1 
( n j pi = in p Hn p p e p2 =2!~ : (3.6.58)
! ~ 2n n ! ~!

Therefore, the eigenve tors jn) of the energy operator H for the harmoni os illator have
the very parti ular property that the transition oeÆ ients (3.6.49) between these ve tors
and the x-basis have the same fun tional form as the transition oeÆ ients (3.6.58) between
these ve tors and the p-basis ex ept for a phase fa tor.
By the same argument as above for the operator Q, we on lude that the spe trum of
P is ontinuous,
Sp(P ) = fp j 1 < p < 1g ; (3.6.59)
and that if we normalize the generalized eigenve tors jpi a ording to
hp0jpi = Æ(p0 p) ; (3.6.60)
then the measure d(p) is the Lebesgue measure on the real line,
d(p) = dp : (3.6.61)
The transition oeÆ ients hpjn) in
Z +1
jn) = dp jpihpjn) (3.6.62)
1
are alled the energy wave eigenfun tions in the momentum representation and are denoted
by
bn (p)  hpjn) : (3.6.63)
Also, for any arbitrary ve tor ' the transition oeÆ ient
'b(p) = hpj'i (3.6.64)
3.6 Gelfand-Maurin Theorem Applied to the Harmoni Os illator 117

in Z +1
'= dp jpihpj'i (3.6.65)
1
is alled the momentum wave fun tion or the wave fun tion in the momentum representation
of '. We have used the notation 'b(p) = hpj'i instead of the notation '(p) = hpj'i be ause
the fun tion 'b(p) of p is in general dierent to the fun tion '(x) = hxj'i of x. Therefore, to
avoid onfusion, we label the two dierent fun tions 'b(p) and '(x) by two dierent symbols.

3.6.3 Realizations of the RHS of the Harmoni Os illator by Spa es

of Fun tions
Thus far, we have obtained the matrix elements of Q in the x-representation,
hxjQjn) = xhxjn) ; hxjQj'i = xhxj'i ; (3.6.66)
and the matrix elements of P in the p-representation,
hpj P j n ) = phpj n ) ; hpj P j 'i = phpj 'i : (3.6.67)
We now want to al ulate hxjP j'i, the matrix elements of P in the x-representation,
and hpjQj'i, the matrix elements of Q in the p-representation. We do this in two steps:
1. We introdu e the new mathemati al obje ts hxjpi and hpjxi. They are generalizations
of the s alar produ t, but are the \s alar produ ts" between the eigenve tors jxi 2 
and jpi 2  . Thus, they are something like the hxjy i in (3.5.39) of Se tion 3.5, i.e.,
distributions that are dened by integration. Like the (njxi and the (njpi in (3.6.36)
and (3.6.62), the hxjpi (and hpjxi) are transition oeÆ ients between basis systems.
But whereas hxjn) are transition oeÆ ients between the ontinuous basis system fjxig
and the dis rete basis system fjn)g, the hxjpi are the transition oeÆ ients between
the ontinuous basis system fjxig and the ontinuous basis system fjpig.
2. We ompute hxjP j'i and hpjQj'i using the expressions for hxjpi and hpjxi.
The mathemati al obje t hpjxi appears when we take the \s alar produ t" of
Z +1
n = dx jxihxjn) (3.6.68)
1
with jpi (or, more pre isely, we onsider n as a fun tional at the generalized eigenve tor
jpi 2 ; p 2 Sp(P ), and use (3.6.68)):
Z +1
hpj n ) = dx hpjxihxjn) : (3.6.69)
1
118 3 The Rigged Hilbert Spa e of the Harmoni Os illator

On the other hand, the mathemati al obje t hxjpi appears when we take the s alar produ t
of Z +1
n = dp jpihpjn) (3.6.70)
1
with hxj (or, more pre isely, we onsider n as a fun tional a ting on the generalized eigen-
ve tor jxi 2  , x 2 Sp(Q), and use (3.6.70)):
Z +1
hxjn) = dp hxjpihpjn) : (3.6.71)
1
In (3.6.69) and (3.6.71), hxjn) and hpjn) are given by (3.6.49) and (3.6.58), respe tively.
The Hermite polynomials have the property
Z +1
2 ei  2 =2 H ( ) ;
in e  =2 Hn ( ) = d p e n (3.6.72)
1 2
p w
where  = p~pw and  = ~ x. Inserting (3.6.49) and (3.6.58) into this relation, it follows
that Z +1
eixp=~
( n j pi = dx p (njxi ; (3.6.73)
1 2 ~
or taking the omplex onjugate
Z +1 ixp=~
e
hpj n ) = dx p hxjn) : (3.6.74)
1 2 ~
Comparing (3.6.74) with (3.6.69), we nd that the hpjxi are given by

hpjxi = p 1 e ixp=~ : (3.6.75)

2 ~
In the same way one obtains from (3.6.71) and (3.6.72)

hxjpi = p 1 eixp=~ : (3.6.76)

2 ~
Eqs. (3.6.75) and (3.6.76) together give
hxjpi = hpjxi: (3.6.77)
It is now simple to al ulate the matrix element of P in the basis of generalized eigenve tors
of Q using (3.6.76):
Z +1 Z +1
eixp=~
hxjP j'i = dp p hxjpihpj'i = dp p p hpj 'i
1 1 2 ~
Z +1 Z +1
~  ~ d
= dp
i x
hxjpihpj'i = i dx dp hxjpihpj'i : (3.6.78)
1 1
3.6 Gelfand-Maurin Theorem Applied to the Harmoni Os illator 119

Thus
d
hxjP j'i = ~i dx hxj'i : (3.6.79)
In the same way using (3.6.75) one obtains

hpjQj'i = ~i dpd hpj'i : (3.6.80)

Therewith, we have shown that the operators Q and P are \realized" in the spa e of position
wave fun tions hxj'i = '(x) by the multipli ation operator
Q'(x) = x'(x) (3.6.81)
and by the dierentiation operator
~d
P '(x) = '(x) ; (3.6.82)
i dx
respe tively. These are the standard expressions that are usually assumed to represent
the position and momentum operators. We have derived them here 2
from the Heisenberg
ommutation relation [P; Q℄ = i~I , the relation H = 21 P 2 + !2 Q2 , and the additional
assumption of the existen e of an eigenve tor of H .
We shall now derive the position representation of the energy operator H , i.e., we shall
al ulate the matrix element hxjH jn). In this position representation, the energy eigenvalue
equation
Hn = En n (3.6.83)
is alled the time-independent S hrodinger equation. The Hamiltonian for the harmoni
os illator is given by
1 ! 2 2
H = P2 + Q : (3.6.84)
2 2
Let us take the matrix element of H between hxj and jn) (or, more pre isely, the a tion of
of the fun tional hxj at the point H jn)):
2
hxjH jn) = 21 hxjP 2jn) + !2 hxjQ2jn) : (3.6.85)

From (3.6.79), it follows that

hxjP 2jn) = hxjP 2jni
~ d
=
i dx
hxjP jni
 ~ 2
d2
=
i dx2
hxjni
 ~ 2 d 2
=
i dx2
hxjn) : (3.6.86)
120 3 The Rigged Hilbert Spa e of the Harmoni Os illator

From (3.6.66),
hxjQ2 jn) = x2 hxjn) : (3.6.87)
Inserting (3.6.86) and (3.6.87) into (3.6.85), we have for the matrix element of the energy
operator
~2
d2 ! 2 2
hxjH jn) =
2 dx2
h x j n ) +
2
x hxjn)
 ~2 d2 ! 2 2 
= + x hxjn) : (3.6.88)
2 dx2 2
Therefore, the S hrodinger representation of the eigenvalue equation (3.6.83) reads
 d2 ! 2 2 
~2
+ x n(x) = En n (x) ; (3.6.89)
2 dx2 2
where hxjn) = n (x) is given in terms of the n-th Hermite polynomial as in (3.6.49).
Eq. (3.6.89) is the time-independent S hrodinger equation. This equation, that is usu-
ally taking as the starting point in the study of the harmoni os illator, has been derived
here from the algebrai assumptions (3.2.1)-(3.2.4).
So far we have dis ussed the a tion of the operators Q, P and H on the wave fun tions
'(x) = hxj'i without spe ifying the parti ular properties of these fun tions. We shall now
show that as a onsequen e of the properties of the spa e , it follows that the position
realization of  is the S hwartz spa e S (R ) ( f. Se tion 2.4.1):
Sin e the spa e  remains stable under the a tion of the algebra of observables, 
remains stable under the a tion of any power of P and Q. This means that the quantities
Z +1
('; Qn ') = dx xn j'(x)j2 < 1 ; n = 0; 1; 2; : : : ; (3.6.90)
1
Z +1
dm
('; P m ') = ( i~)m dx '(x) '(x) < 1 ; m = 0; 1; 2; : : : ; (3.6.91)
1 dxm
Z +1
dm
n m
('; Q P ') = ( i~) m dx '(x) xn m '(x) < 1 ; n; m = 0; 1; 2; : : : (3.6.92)
1 dx
must be well dened for every ' 2 . This implies that the fun tions '(x) = hxj'i in the
realization spa e must be innitely dierentiable and that the fun tions and their derivatives
must de ay at innity faster than any power of x. Therefore, '(x) 2 S (R ). Moreover, the
topology on  is equivalent to the topology on S (R ). To show that equivalen e, we re all
that the topology on  is des ribed by the following pres ription for sequen e onvergen e:
a sequen e 'k 2  onverges to ' 2  if
k'k 'kp ! 0 ; p = 0; 1; 2; : : : ; (3.6.93)
p
where k'kp = ('; (N + I )p). In the realization of , the onditions (3.6.93) are equivalent
to the following: a sequen e 'k (x) = hxj'k i, whi h is the realization of the  - onvergent
3.6 Gelfand-Maurin Theorem Applied to the Harmoni Os illator 121

sequen e 'k ! ', onverges mto '(x) = hxj'i if the xn dxd m 'k (x) onverge uniformly on
m

every bounded region to xn dxd m '(x) for ea h n; m = 0; 1; 2; : : : Therefore the topology of

the realization of  is equivalent to the topology of the S hwartz spa e.
Sin e the spa e  is realized by the S hwartz spa e S (R ), the dual spa e  is realized
by the spa e of tempered distributions S (R ) . The realization of the generalized eigenve tor
jxi of Q is the Dira delta fun tion (see Eq. (3.6.45)), whi h is a distribution that belongs
to S (R ) ,
 3 jxi ! hx0jxi = Æ(x0 x) 2 S (R ) : (3.6.94)
The realization of the generalized eigenve tor jpi of P is the exponential fun tion eixp, whi h
is also a distribution,6

 3 jpi ! hxjpi = p 1 eixp=~ 2 S (R ) : (3.6.95)

2 ~
To say that Æ (x0 x) and p21~ eixp=~ are distributions in S (R ) means that they only make
sense as kernels of integrals that involve fun tions '(x) 2 S (R ),
Z +1
dx0 Æ (x0 x)'(x0 ) = '(x) ; '(x) 2 S (R ) ; (3.6.96)
1
Z +1
1
dp p eixp=~ 'b(p) = '(x) ; '(x) 2 S (R ) ; (3.6.97)
1 2 ~
or in bra-ket notation Z +1
dx0 hxjx0 ihx0 j'i = hxj'i ; (3.6.98)
1
Z +1
dp hxjpihpj'i = hxj'i : (3.6.99)
1
Finally, it is lear that the Hilbert spa e H is realized by the spa e of Lebesgue square
integrable fun tions L2 (R ; dx).
Summarizing, ea h ve tor ' in the ve tor spa e  an be fully hara terized by its
omponents with respe t to the ontinuous basis system of eigenve tors jxi of Q,
Z +1
'= dx jxihxj'i : (3.6.100)
1
Thus to the ve tor ' orresponds the fun tion hxj'i = '(x) and to the ve tor P ' orre-
sponds the fun tion hxjP j'i  P '(x). Equation (3.6.79) then states that in the realization
of the spa e of ve tors ' by the spa e of wave fun tions hxj'i = '(x), the momentum
operator is realized by the dierential operator times ~=i,
d
P ! i~ dx ; (3.6.101)
6
The exponential fun tion is not square integrable.
122 3 The Rigged Hilbert Spa e of the Harmoni Os illator

Eq. (3.6.66) states that the position operator is realized by the operator of multipli ation
by x
Q ! x; (3.6.102)
and Eq. (3.6.88) states that the energy operator is realized by the dierential operator
 d2 ! 2 2 
~2
H ! + x : (3.6.103)
2 dx2 2
The realization of the spa e  for the harmoni os illator given by the asso iation
 3 ' ! '(x) = hxj'i 2 S (R ) (3.6.104)
between the ve tor ' and its \ ontinuous omponents" hxj'i = '(x) establishes an equiva-
len e between two Rigged Hilbert Spa es, the RHS of the harmoni os illator
  H   (3.6.105)
and the RHS of S hwartz spa e fun tions
S (R )  L2 (R ; dx)  S  (R ) : (3.6.106)
This realization is alled the S hrodinger (position) representation or the x-representation.
We stress that the S hrodinger representation ould not be derived from the Heisenberg
ommutation relation
[P; Q℄ = i~I (3.6.107)
and the relation
1 2 ! 2 2
H= P + Q (3.6.108)
2 2
alone, but required an additional assumption: the operator H has at least one proper
eigenve tor in the spa e of states.
It is usually very useful to show the realizations of the abstra t mathemati al obje ts
(RHS, operators, fun tions, eigenve tors) through a diagram. For instan e, the position
representation of the operator Q is visualized by the following diagram:
x-representation diagram for the operator Q

Q; ' 2   H   3 jxi
l l l l l l
x; '(x) = hxj'i 2 S (R )  L2 (R ; dx)  S (R ) 3 hx0 jxi = Æ(x x0)
(3.6.109)
On the top line of the diagram (3.6.109), we have the abstra t obje ts. On the bottom line,
we have the x-realizations of all these abstra t obje ts: Q is realized by the multipli ation
3.6 Gelfand-Maurin Theorem Applied to the Harmoni Os illator 123

operator, ' is realized by the position wave fun tion '(x),  by the S hwartz spa e S (R ), H
by the Hilbert spa e of square integrable fun tions L2 (R ; dx),  by the spa e of tempered
distributions S (R ) and jxi by the Dira delta fun tion Æ (x x0 ).
In a similar way, we an onstru t the position representation diagram for the momentum
operator P ,
x-representation diagram for the operator P

P; ' 2   H   3 j pi
l l l l l l
i~ dxd ; '(x) = hxj'i 2 S (R )  L2 (R ; dx)  S (R ) 3 hxjpi = epixp=~
2~
(3.6.110)
As in the diagram (3.6.109), the top line of (3.6.110) ontains the abstra t mathemati al
obje ts and the bottom line ontains their x-realizations.
The x-diagram for the energy operator reads
x-representation diagram for the operator H

H; ' 2   H   3 jni
l l l l l l (3.6.111)
~2 d2 2
2 dx2 + !2 x2 ; '(x) 2 S (R )  L2(R ; dx)  S (R ) 3 n(x)
The momentum representation leads to similar onsiderations. The operator P is realized
by the multipli ation operator (see Eq. (3.6.67))
P 'b(p) = p'b(p) ; (3.6.112)
and the operator Q by the dierentiation operator (see Eq. (3.6.80))
~ d
Q'b(p) = 'b(p) : (3.6.113)
i dp
The spa e  is realized by the S hwartz spa e of fun tions 'b(p). In the p-representation,
the generalized eigenve tor jxi is realized by the exponential fun tion
 3 jxi ! hpjxi = p 1 e ipx=~ ; (3.6.114)
2 ~
and the eigenve tor jpi by the Dira delta fun tion
 3 jpi ! hp0jpi = Æ(p0 p) : (3.6.115)
124 3 The Rigged Hilbert Spa e of the Harmoni Os illator

As a onsequen e of (3.6.114),
Z +1 Z +1
1
'b(p) = hpj'i = dx hpjxihxj'i = dx p
e ixp=~ '(x) ; (3.6.116)
1 1 2 ~
Therefore, we an transform from the position representation '(x) into the momentum
representation 'b(p) using (3.6.95) and (3.6.116). Note that Eqs. (3.6.95) and (3.6.116) show
that the x- and the p-representation are related by the Fourier transform.
We an also onstru t diagrams similar to those onstru ted for the position represen-
tation. For instan e, the p-representation of the position operator leads to
p-representation diagram for the operator Q

Q; ' 2   H   3 jxi
l l l l l l
~d
i dp ; 'b(p) = hpj'i 2 S (R )  L2(R ; dp)  S (R ) 3 hpjxi = p21~ e ipx=~
(3.6.117)
For the momentum operator we have:
p-representation diagram for the operator P

P; ' 2   H   3 j pi
l l l l l l
p; 'b(p) = hpj'i 2 S (R )  L2 (R ; dp)  S (R ) 3 hp0jpi = Æ(p0 p)
(3.6.118)
Finally, the p-representation diagram for the energy operator reads
p-representation diagram for the operator H

H; ' 2   H   3 jn)
l l l l l l
~2 !2 d2 + 21 p2 ; 'b(p) 2 S (R )  L2 (R ; dp)  S (R ) 3 hpjn) = bn(p)
2 dp2
(3.6.119)
For the sake of ompleteness, we re all the energy representation '(n) = (nj'i. In this
representation, the variable n  En is dis rete, and the realization of the ve tors ' are given
3.6 Gelfand-Maurin Theorem Applied to the Harmoni Os illator 125

by sequen es of omplex numbers '(n) = (nj'i rather than by fun tions. The Hilbert spa e
H is realized by the sequen e of square integrable fun tions (3.3.51) and the spa e  by the
spa e of rapidly de reasing sequen es (3.3.52). The matrix elements of the operators H , Q
and P are
(njH jm) = En Ænm ; (3.6.120)
s
~  p p 
(mjQjn) = n Æm;n 1 + n + 1 Æm;n+1 ; (3.6.121)
2!
and r
~!  pn Æ p 
(mjP jn) = i m;n 1 n + 1 Æm;n+1 ; (3.6.122)
2
respe tively. In this representation, these operators are visualized as innite matri es whose
entries are given by the equations (3.6.120)-(3.6.122).

3.6.4 Summary
In the pre eding se tions, we have onstru ted the Rigged Hilbert Spa e for the one-
dimensional harmoni os illator. We started out with the algebra of observables A generated
by P , Q and H . These operators fulll the algebrai relations
1 2 w2 2
[P; Q℄ = i~I ; H= P + Q : (3.6.123)
2 2
The elements of A were assumed to be symmetri operators dened on a linear spa e ,
(A'; ) = ('; A ) ; '; 2 ; A 2 A; (3.6.124)
where (
;
) is the s alar produ t that provides the probabilities. This spa e was assumed
to remain stable under the a tion of the elements of A. We made the ru ial additional
assumption that the operator H has at least one eigenve tor7 0 in the spa e ,
1
H0 = ~!0 : (3.6.125)
2
From this one eigenve tor we dened the ve tors

jn) = p1 (ay)n0 ; n = 0; 1; 2; : : : ; (3.6.126)

n!
whi h fulll
H jn) = ~! (n + 1=2)jn) : (3.6.127)
7
This is equivalent to the assumption that H is essentially self-adjoint on the invariant dense subspa e
of the algebra A.
126 3 The Rigged Hilbert Spa e of the Harmoni Os illator

These eigenve tors span the linear spa e . With respe t to the s alar produ t on , the
eigenve tors jn) are orthonormal to ea h other,
(njm) = Ænm : (3.6.128)
The spa e was ompleted to the Hilbert spa e H using the topology generated by the
norm p
k'k = ('; ') : (3.6.129)
The ( omplete) Hilbert spa e H is
1
X 1
X
H = f' = jn)(nj') j j(nj')j2 < 1g : (3.6.130)
n=0 n=0

The operators in the algebra of observables were extended from into larger subdomains of
H. However, these extensions are not ontinuous with respe t to the Hilbert spa e topology,
and the domains do not remain stable under the a tion of the operators. In order to nd
a ommon invariant subdomain  for the algebra A that is endowed with a topology that
makes these observables ontinuous operators, we introdu ed the sequen e of s alar produ ts
('; )p = ('; (N + I )p ) ; p = 0; 1; 2; : : : (3.6.131)
The ompletion of with respe t to the topology generated by these s alar produ ts is the
spa e
1
X 1
X
 = f' = jn)(nj') j (n + 1)p j(nj')j2 < 1 ; p = 0; 1; 2; : : :g : (3.6.132)
n=0 n=0

The elements of the algebra A were extended ontinuously into . The spa e  remains
stable under the a tion of these extensions and all these extensions are ontinuous with
respe t to the topology on . Therefore, all the algebrai al ulations needed in physi s
involving the elements of the algebra of observables are allowed.
The operators P and Q are essentially self-adjoint8 and have a ontinuous spe trum
that oin ides with the real line. In order to asso iate an eigenve tor to ea h element of the
spe trum of these operators, we introdu ed the adjoint spa e  and onstru ted the RHS
  H   : (3.6.133)
In this RHS, the Gelfand-Maurin Theorem holds. This theorem assured the existen e of a
omplete system of generalized eigenve tors of Q
Q jxi = xjxi ; jxi 2  ; (3.6.134)
8
As a onsequen e of H being essentially self-adjoint.
3.6 Gelfand-Maurin Theorem Applied to the Harmoni Os illator 127

su h that any ' 2  was expanded in terms of these generalized eigenve tors of Q,
Z +1
'= dx jxihxj'i : (3.6.135)
1
Eq. (3.6.135) is the mathemati al rephrasing of the heuristi Dira basis ve tor expansion.
Using the spe tral de omposition (3.6.135), we derived the x-realization of the abstra t
' by fun tions '(x) in the S hwartz spa e,
 3 ' ! '(x) = hxj'i 2 S (R ) : (3.6.136)
In parti ular, to every eigenve tor jn) of H there orresponds a fun tion hxjn) = n (x),
jn) ! hxjn) = n(x) ; (3.6.137)
given in terms of the Hermite polynomials. The x-representation of the operator Q is the
multipli ation operator
Q' ! x'(x) ; (3.6.138)
the x-representation of the operator P is given by the dierentiation operator
d
P' ! ~i dx '(x) ; (3.6.139)
and the x-representation of the Hamiltonian is
 ~2 d2 ! 2 2 
H' ! 2 dx2
+
2
x '(x) : (3.6.140)

The realization of  given by the asso iation between the ve tor ' and its \ ontinuous
omponents" '(x) yields the realization of the abstra t RHS
  H   (3.6.141)
by the RHS of S hwartz spa e fun tions
S (R )  L2 (R ; dx)  S  (R) : (3.6.142)
The generalized eigenve tor jxi of Q is realized by the Dira delta fun tion
 3 jxi ! hx0 jxi = Æ(x0 x) 2 S (R ) ; (3.6.143)
and the generalized eigenve tor jpi of P by the exponential fun tion

 3 jpi ! hxjpi = p 1 eixp=~ 2 S (R ) : (3.6.144)

2 ~
Therewith, we have derived the S hrodinger representation of the harmoni os illator
from the algebrai assumptions (3.6.123)-(3.6.125). We remark again that the S hrodinger
128 3 The Rigged Hilbert Spa e of the Harmoni Os illator

representation is only one of the many possible representations of (3.6.123), and that it is
the additional assumption (3.6.125) whi h singles out the S hrodinger representation among
all the possible representations.9
From a mathemati al point of view, the RHS formulation extends the Hilbert spa e
formulation and justies the mathemati ally undened operations that physi ists are a us-
tomed to in their al ulations. In parti ular, using the Rigged Hilbert Spa e formalism we
are able to reprodu e the main features of the Dira formalism.

3.7 A Remark Con erning Generalizations

3.7.1 Realization of the Abstra t RHS by Spa es of Fun tions
The realization of the RHS of the harmoni os illator by the RHS of S hwartz fun tions
suggests that any RHS asso iated to the spe tral de omposition of an operator an be
realized by spa es of fun tions.
Let A be an operator dened on the RHS   H   and
Z
'= d() jihj'i (3.7.1)
Sp(A)

be the spe tral de omposition of ' 2  provided by the Gelfand-Maurin Theorem. The
quantity hj'i whi h appears in this spe tral de omposition may be regarded as a omplex
fun tion of the real variable  2 Sp(A), i.e., we an dene
'(
) : Sp(A) 7! C
 7 ! '() := hj'i : (3.7.2)
(In this se tion, we shall write '(
) when we want to speak about a fun tion and '() when
we want to speak about the value of this fun tion at a parti ular point ). If we write
Z Z
('; ') = d() h'ji hj'i = d() jhj'ij2 ; (3.7.3)
Sp(A) Sp(A)

we immediately realize that '(
) in (3.7.2) must be a square integrable fun tion with respe t
to d(). We shall denote the spa e of fun tions fullling (3.7.3) as L2 (Sp(A); d()). The
s alar produ t on L2 (Sp(A); d()) is dened as
Z
('(
); (
))L2 := d() '() () : (3.7.4)
Sp(A)
9
There are several equivalent forms of the assumption (3.6.125):
(a) The assumption that H is essentially self adjoint.
(b) The assumption that P , Q, and I are the generators of a group, the Weyl group (subgroup of the
symmetry group of non-relativisti spa e-time, the Galilei group).
3.7 A Remark Con erning Generalizations 129

If ', 2 , we have
Z
('(
); (
))L2 = d() h'ji hj i = ('; ) : (3.7.5)
Sp(A)

Thus the mapping ' $ '(
) that takes ' 2 H into the fun tion '(
) 2 L2(Sp(A); d())
preserves the s alar produ ts,
('; )H = ('(
); (
))L2 : (3.7.6)
Further, this mapping is obviously linear, be ause hj' +  i = hj'i +  hj i for
;  2 C . Thus, this mapping an be represented by means of a linear operator U y su h
that
Uy :  7 ! L2 (Sp(A); d())
' 7 ! U y ' = '(
) : (3.7.7)
Sin e U y preserves s alar produ ts, U y is an isometry ( f. Se tion 2.5.2). Moreover, it an
be proved that the image of  by U y is dense in L2 (Sp(A); d()). Therefore, U y has a
unique extension to H. This extension (whi h we also denote by U y ) is a unitary operator
from H onto L2 (Sp(A); d()),
Uy : H 7 ! L2 (Sp(A); d())
f 7 ! U yf : (3.7.8)
It is important to remark that the equation f () = hjf i, i.e., the statement that the value
of the fun tion f (
) at the point  equals the a tion of the fun tional hj at f , holds only
when f is an element of . For a general f in H, the orresponding U y f  f (
) has no
meaning as a fun tion, but only as a lass of equivalen e of fun tions whi h dier on a set
of zero Lebesgue measure.
We an endow U y  with a topology  by transporting the topology of  into U y  via
U . Sin e  is assumed to satisfy the rst axiom of ountability, so does U y . Therefore, we
y
an transport the topology  on  into U y  by using the notion of sequen e onvergen e.
Then, we say that 'n (
) ! '(
) with respe t to  i 'n ! ' with respe t to  . With this
denition, all topologi al properties are transferred from  into U y  by means of U y . In
parti ular, U y  is a  - omplete nu lear spa e, and it is  -dense in L2 (Sp(A); d()), i.e.,
for any f (
) 2 L2 (Sp(A); d()) there exists a sequen e of fun tions f'n(
)g1 n=1 in U  su h
y
that Z
d() j'n() f ()j2 n!1 !0: (3.7.9)
Sp(A)
Our next step is to extend the operator U y on  to an operator U  on  . This operator
is dened by
hU y'jU F i := h'jFi ; ' 2  ; F 2  : (3.7.10)
130 3 The Rigged Hilbert Spa e of the Harmoni Os illator

U  is a well dened linear operator from  into (U y ) . To show this, we just need to
prove that U  F is a ontinuous antilinear fun tional on U y . The antilinearity follows
from the denition (3.7.10), and the ontinuity from the fa t that if U y 'n ! U y ', then

hU y'njU F i = h'nji ! h'ji = hU y'jU F i : (3.7.11)

Therefore, U  F is an element of (U y ) .

The a tion of U  extends the a tion of U y . This means that if f 2 H, then U  f = U y f .
In fa t,
hU y'jU f i = h'jf i  ('; f ) = (U y'; U yf )  hU y'jU yf i (3.7.12)
for any ' 2  or, equivalently, for any U y ' 2 U y . Therefore, the fun tionals U  f and
U y f an be identied.
One an also prove that U  :  ! (U y ) is ontinuous when  is endowed with
the   topology and (U ) with the  topology. Hen e

U   = (U y ) (3.7.13)

and both spa es have the same linear topologi al stru ture.
In summary, we have onstru ted a realization of the RHS

  H   (3.7.14)

by the RHS of spa es of fun tions

U y   L2 (Sp(A); d())  (U y ) (3.7.15)

using the unitary operator U y provided by the Gelfand-Maurin Theorem.

At this point, it is important to stress that the unitary operator U y whi h yields the
realization of   H   depends drasti ally on A (the operator providing the generalized
eigenve tors). The dependen e of the spa e L2 (Sp(A); d()) on A is twofold: both Sp(A)
and d() depend on A. However, it is also possible that two dierent observables A and
B (with two dierent U y 's) lead to the same spa e L2 (Sp(A); d()). For instan e, in
the ase of the harmoni os illator we have seen that A = Q and B = P both lead to
L2 (Sp(A); d()) = L2 (R ). However, the Gelfand-Maurin theorem for Q gives a unitary
operator U y from H onto L2 (R ), and for P gives the operator V y = F U y , where F is the
Fourier transform operator.
Along with the realization of the ve tors of the triplet   H   , we an also onsider
a realization of observables. If A is an operator on , then U y AU is the orresponding
operator on U y . We all U y AU the realization of A on U y .
As we did in the ase of the harmoni os illator, we show the realization of an abstra t
RHS through the following diagram:
3.7 A Remark Con erning Generalizations 131

A; ' 2   H   3 ji
# # Uy # Uy # U # (3.7.16)
U y AU; U y ' = '(
) 2 U y  L2 (Sp(A); d())   3 U  ji

The top line of the diagram (3.7.16) ontains the abstra t mathemati al obje ts, and
the bottom line ontains their realizations.
An abstra t RHS and its realization are equivalent Rigged Hilbert Spa es|there is no
linear topologi al property that distinguishes one from the other. This suggests the following
denition:

Denition Two Rigged Hilbert Spa es   H   and  G   are equivalent i

there exists a unitary operator U y from H onto G su h that
1. U y H = G , U y  = .
2. U y and U are ontinuous with respe t to the topologies on  and .
As a onsequen e, U y an be extended to an operator U  :  !  dened by
hU y 'j U  F i = h'j F i : (3.7.17)
This extension U  is a bi ontinuous bije tive10 linear mapping from  onto  and hen e
U   =  .
The RHS spe tral de omposition
Z
'= d() jihj'i (3.7.18)
Sp(A)

of ' 2  is a \ ontinuous innite dimensional" generalization of the Hilbert spa e spe tral
de omposition of a ompa t self-adjoint operator. However, there are some dieren es
between these two spe tral de ompositions. The ontent of (3.7.18) is that any ve tor in 
an be written in terms of the generalized eigenve tors of A. Therefore, the eigenve tors of
A form a \system of generators" for the spa e . This system of generators is not a basis
for  in the usual Hilbert spa e sense for the following reasons:
1. The ve tors ji do not, in general, belong to  or to H.
10
One-to-one, onto, with ontinuous inverse.
132 3 The Rigged Hilbert Spa e of the Harmoni Os illator

2. The nite or ountably innite sum

1
X 1
X
'= n en = (en ; ')en (3.7.19)
n=0 n=0
in the Hilbert spa e spe tral de omposition is repla ed by an integral
Z
'= d() ji hj'i : (3.7.20)
Sp(A)

In the Hilbert spa e spe tral de omposition (3.7.19), n is the dis rete oeÆ ient (or weight)
for en . The n an be viewed as the n-th omponent of ' with respe t to the basis en . In the
RHS spe tral de omposition (3.7.20), '() = hj'i is the ontinuous oeÆ ient (or weight)
for ji. For a xed  2 Sp(A), '() = hj'i an be viewed as the \-th" omponent of '
with respe t to the system of generators ji.
The de omposition (3.7.19) leads to a realization of the P Hilbert spa e H by the spa e
of sequen es f n g of omplex numbers having the property 1 n=0 j n j < 1. Analogously,
2
(3.7.20) leads to a realization of the spa e  by the spa e of fun tions '().
The Gelfand-Maurin Theorem allows us to spe trally de ompose the s alar produ t of
any two ve tors ', 2  as
Z
( ; ') = d() h'jihj i : (3.7.21)
Sp(A)
In analogy to (3.7.21), we also spe trally de ompose the a tion of any fun tional F at a
ve tor ', Z
h'j F i = d() h'jihjF i : (3.7.22)
Sp(A)
Hen e, quantities of the type hjF i are distributions that are well dened only as a kernel
of integration whenever we write the a tion of a fun tional F as an integral operator. As
an example, the fun tional jxi that asso iates to any ' 2 S (R ) the value of the fun tion at
the point x,
jxi : S (R ) 7 ! C
' 7 ! h'jxi := '(x) ; (3.7.23)
an be written as an integral operator:
Z +1
h'jxi = dx h'jx0 ihx0 jxi : (3.7.24)
1
The fun tional jxi in (3.7.23) is the S hwartz delta fun tional, whereas the distribution
hx0 jxi in (3.7.24) is the Dira delta fun tion,
hx0jxi = Æ(x0 x) : (3.7.25)
Therefore, the Dira delta fun tion appears when we spe trally de ompose the a tion of the
S hwartz delta fun tion as an integral operator.
3.7 A Remark Con erning Generalizations 133

3.7.2 General Statement of the Gelfand-Maurin Theorem

The version of the Gelfand-Maurin Theorem stated in Se tion 3.5.2 is only suitable for
y li operators. In general, the linear operators in the algebra of observables are not
y li . Then more than one quantum number is needed to hara terize pure states, and one
needs a omplete system of ommuting operators to obtain a omplete set of generalized
eigenve tors.
Denition The olle tion of operators fAk gNk=1 is a system of ommuting operators if
1. [Ai ; Ak ℄ = 0 for all i; k = 1; : : : ; N ,
PN
2. k=1 A2k is essentially self-adjoint.
Let A be the algebra generated by a olle tion of operators fAk gNk=1. Then the olle tion
fAk gNk=1 is said to be a omplete system if there exists a ve tor ' 2  su h that the spa e
fA' j A 2 Ag (3.7.26)
spans the Hilbert spa e H.
The notions of generalized eigenve tor and generalized eigenvalue of an operator an be
extended to the ase of a system fAk gNk=1 . An antilinear fun tional F on  is a generalized
eigenve tor for a system fAk gNk=1 if for every k = 1; : : : ; N there exists a omplex number
(k) su h that
Ak F = (k) F : (3.7.27)
The numbers   ((1) ; (2) ; : : : ; (N ) ) are alled generalized eigenvalues orresponding to
the generalized eigenve tor F  j(1) ; (2) ; : : : ; (N ) i.
Theorem (Gelfand-Maurin Theorem) Let fAk gNk=1 be a omplete system of ommuting,
e.s.a.,  - ontinuous operators on the Rigged Hilbert Spa e   H   . Then, there
exists a set of generalized eigenve tors,
j(1) ; (2) ; : : : ; (N ) i 2  ; (3.7.28)

Ak j ;  ; : : : ;  i =  j ;  ; : : : ;  i ;
(1) (2) (N ) (k ) (1) (2) (N ) (3.7.29)
 2  = Sp(Ak ) ;
(k ) (k ) (3.7.30)
and a uniquely dened measure d() on  = (1)  (2) 


 (N ) , (where  denotes
the Cartesian produ t), su h that for every ; ' 2 
Z
( ; ') = d() h j(1); (2) ; : : : ; (N ) ih(1) ; (2) ; : : : ; (N ) j'i ; (3.7.31)

or omitting ,
Z
'= d() j(1); (2) ; : : : ; (N ) ih(1) ; (2) ; : : : (N ) j'i : (3.7.32)

134 3 The Rigged Hilbert Spa e of the Harmoni Os illator

This theorem gives the mathemati al formulation of the famous Dira onje ture if the
starting point is a pre isely dened algebra of observables.
The mathemati al task that has to be a omplished if one starts out with a well de-
ned algebra is to nd a omplete ommuting system and its spe trum. The problem of
determining when a system is omplete is far from trivial. Already for the simplest ases of
enveloping algebras of group representations the number of ommuting observables is not
independent of the parti ular ommuting system.
The problem of the physi ist is usually the reverse. From the experimental data one
nds out how many quantum numbers are required, and what their possible values are.
This gives a minimum number of operators for the omplete ommuting system be ause
the \matrix elements" of the Ak 's al ulated from the properties of this algebra must agree
with the experimental values of the orresponding observables.

3.7.3 Generalization of the Algebra of Operators

The onstru tion of the nu lear spa e  arried out for the harmoni os illator an be
immediately generalized to more general algebras of operators. The analog of the lemma
(3.3.67) is
('; X (
+ I )p X')  ('; (
+ I )p+1 ') ; (3.7.33)
P 2
where X is one of the generators Xi and
= Xi is the Nelson operator (Lapla ian).
Eq. (3.7.33) holds for all enveloping algebras (lemma by Nelson). Therewith the ontinuity
of the algebra in a linear topologi al spa e in whi h the topology is dened by the ountable
number of s alar produ ts
('; )p = ('; (
+ I )p ) (3.7.34)
follows immediately. Further, if
is e.s.a., then all symmetri generators are also e.s.a. (the-
orems by Nelson and Stinespring).
Eq. (3.7.33) is mu h stronger than what is required for the proof of the ontinuity of the
generators. The ontinuity of the generators (and therewith of the whole algebra) an be
proved if instead of p + 1 on the right hand side of (3.7.33) one has p + n, where n is any
positive integer. Therefore, it appears that the ontinuity of the generators an already be
proved for any nitely generated asso iative algebra.
The nu learity is a mu h harder property to establish. It has been proven for the ases
that the algebra is the enveloping algebra E (G) of the following groups G:
1. G is nilpotent (be ause then E (G) is isomorphi to the enveloping algebra generated
by P , Q ,  = 1; 2; : : : m, with [P ; Q ℄ = Æ I for some m (a theorem by Kirillov
in [69℄) and we have just an m-dimensional generalization of the harmoni os illator).
2. G is semi-simple (Bohm in [70℄).
3. G = AK , where  stands for semidire t produ t, with A Abelian and K ompa t
(B. Nagel in [71℄).
4. G is the Poin are group, for some of the representations (see [71℄).
3.7 A Remark Con erning Generalizations 135

3.7.4 Appendix: Continuity of the Algebra of the Harmoni Os-

illator
In this appendix, we provide a proof for Eq. (3.3.67)
( ; a(N + I )pay )  ( ; (N + I )p+1 ) ; 8 2 : (3.7.35)
Before pro eeding with the proof, we need some preliminary results. From the om-
mutation relation (3.2.10), it follows that
(N + I )ay = ay(N + 2I ) ; (3.7.36)
a(N + I ) = (N + 2I )a : (3.7.37)
It also holds that
( ; (N + I ) )  ( ; (N + I )2 ) ; 8 2 : (3.7.38)
Eq. (3.7.38) omes from the fa t that N is a positive operator, i.e., ( ; N )  0 for
ea h in and then
( ; (N + I )2 ) ( ; (N + I ) ) = (N ; N ) + ( ; N )
= kN k2 + ( ; N )  0 : (3.7.39)
From the positive deniteness of N , it also follows that if m and n are positive integers
and m  n, then
( ; (N + I )m )  ( ; (N + I )n ) ; 8 2 : (3.7.40)
We are now going to prove Eq. (3.7.35) by indu tion. Eq. (3.7.35) is true for p = 1,
be ause
( ; a(N + I )ay ) = ( ; aay (N + 2I ) )
= ( ; (N + I )(N + 2I ) )
= ( ; (N + I )2 ) + ( ; (N + I ) )
 2( ; (N + I )2 ) ; (3.7.41)
where in the last step we have made use of Eq. (3.7.38).
We now assume that (3.7.35) is true for p  q , i.e.,
( ; a(N + I )p ay )  ( ; (N + I )p+1 ) ; 8 2 ; p = 1; 2; : : : ; q : (3.7.42)
We have to prove that (3.7.35) is also true for p = q + 1 using (3:7:42). So we al ulate
( ; a(N + I )q+1 ay ) = ( ; a(N + I )(N + I )q 1 (N + I )ay )
= ( ; (N + 2I )a(N + I )q 1 ay (N + 2I ) ) (3.7.43)
= ((N + 2I ) ; a(N + I )q 1 ay (N + 2I ) )
136 3 The Rigged Hilbert Spa e of the Harmoni Os illator

 ((N + 2I ) ; (N + I )q (N + 2I ) ) (3.7.44)
= [((N + I ) ; (N + I )q (N q
+ I ) ) + ( ; (N + I ) )
+((N + I ) ; (N + I )q ) + ( ; (N + I )q (N + I ) )℄
= [( ; (N + I )q+2 ) + ( ; (N + I )q )
+( ; (N + I )q+1 ) + ( ; (N + I )q+1 )℄
 4( ; (N + I )q+2 ) ; (3.7.45)
where we have used Eqs. (3.7.36)-(3.7.37) in step (3.7.43), Eq. (3.7.42) in step (3.7.44) and
Eq. (3.7.40) in the last step.
Consequently, (3.7.35) has been shown to be fullled also for p = q + 1 and, therefore,
it is true for any integer p.
Chapter 4
A Rigged Hilbert Spa e of the Square
Barrier Potential
In this hapter, we onstru t a RHS of the square barrier Hamiltonian. In order to do it, we
shall use the Sturm-Liouville theory. This theory provides the dire t integral de omposition
of the Hilbert spa e. From this dire t integral de omposition, we shall onstru t the RHS.

CHARLIE [Stopping HAPPY's movement and reply. To BIFF℄

Nobody dast blame this man. You don't understand: Willy
was a salesman. And for a salesman, there is no ro k bottom
to the life. He don't put a bolt to a nut, he don't tell you
the law or give you medi ine. He's a man way out there in
the blue, riding on a smile and a shoeshine. And when they
start not smiling ba k|that's an earthquake. And then you
get yourself a ouple of spots on your hat, and you're n-
ished. Nobody dast blame this man. A salesman is got to
dream, boy. It omes with the territory.
BIFF: Charley, the man didn't know who he was.
HAPPY[infuriated℄: Don't say that!
BIFF: Why don't you ome with me, Happy?
HAPPY: I'm not li ked that easily. I'm staying right in this
ity, and I'm gonna beat this ra ket! [He looks at BIFF, his hin
set.℄ The Loman Brothers!

BIFF: I know who I am, kid.

Arthur Miller, Death of a Salesman

137
4.1 Introdu tion 139

4.1 Introdu tion

In the previous hapter, we have onstru ted the RHS of the harmoni os illator. This sys-
tem has a Hamiltonian whose spe trum is dis rete, i.e., the solutions of the time independent
S hrodinger equation orresponding to the harmoni os illator are square normalizable. Be-
ause the spe trum of its Hamiltonian has no ontinuous part, the harmoni os illator annot
have s attering states.
We now turn to study systems whose time independent S hrodinger equation has non-
square normalizable solutions. That is, systems whose Hamiltonian has a ontinuous spe -
trum. We shall fo us on the square barrier potential, be ause its S hrodinger equation an
be solved expli itly. The square barrier potential will give us the long-sought example of the
Rigged Hilbert Spa e generated by a S hrodinger Hamiltonian with ontinuous spe trum.
First, we review the gist of the Dira formalism for the ase of a Hamiltonian with
ontinuous spe trum. The dynami al equation that governs the behavior of a quantum
system at any time is the time dependent S hrodinger equation:

i~ '(t) = H'(t) ; (4.1.1)
t
where H denotes the Hamiltonian of the system and '(t) denotes the value of the wave
fun tion ' at time t. The Dira formalism solves this equation formally as follows: for
ea h energy E in the spe trum Sp(H ) of the Hamiltonian, there exists a ket jE i that is an
eigenve tor of H ,
H jE i = E jE i ; E 2 Sp(H ) : (4.1.2)
These eigenkets form a omplete basis system that expands any wave fun tion ' as
Z Z
'= dE jE ihE j'i  dE '(E )jE i : (4.1.3)

The time dependent solution of Eq. (4.1.1) is obtained by Fourier-transforming the time
independent solution (4.1.3),
Z
'(t) = dE e iEt=~ '(E ) : (4.1.4)

If the spe trum of the Hamiltonian has a ontinuous part, and if the energy E belongs
to this ontinuous part of the spe trum, then the orresponding eigenket jE i that solves
Eq. (4.1.2) is not square integrable, i.e., jE i is not an element of the Hilbert spa e.
It is the purpose of this hapter to show that the Rigged Hilbert Spa e is the mathe-
mati al framework that supports the above formal manipulations. We will show that the
expansion (4.1.3) is not valid for every element of the Hilbert spa e H, but only for those '
that belong to the spa e of wave fun tions   H. We will also show that the kets jE i an
be understood mathemati ally as ontinuous antilinear fun tionals over the spa e of wave
fun tions , i.e., jE i 2  .
140 4 A Rigged Hilbert Spa e of the Square Barrier Potential

A ording to the RHS mathemati s, equation (4.1.2) means that

hH'jE i = E h'jE i ; 8' 2  : (4.1.5)

The a tion of H an be extended to the kets jE i in  as follows:

h'jH jE i = hH'jE i ; 8' 2  : (4.1.6)

Be ause H is ontinuous on , the operator H  is a uniquely dened extension of H . Using

the denition (4.1.6), we rewrite Eq. (4.1.5) as

h'jH jE i = E h'jE i ; 8' 2  : (4.1.7)

Omitting the arbitrary ' in this equation leads to

H  jE i = E jE i ; (4.1.8)

whi h is the same as Eq. (4.1.2). (Note that in Eq. (4.1.8) we have denoted the a tion of
the Hamiltonian on the ket jE i by H  and not just by H . We shall use this notation in
order to stress that the Hamiltonian is a ting on ve tors that lie outside the Hilbert spa e.)
The statement of the Nu lear Spe tral Theorem [5℄ only assures the existen e of the
generalized eigenve tors jE i, but it does not say how to onstru t them or how to onstru t
the spa e , whi h is assumed to be given beforehand. In this hapter, we provide an
example of a Hamiltonian with ontinuous spe trum where all the quantities are expli itly
onstru ted. As mentioned above, this example is the three-dimensional square barrier
potential. We shall use the Sturm-Liouville theory (Weyl theory) [30℄ to nd the RHS of
this potential.
By applying the Sturm-Liouville theory to the S hrodinger equation of the square barrier
potential, we will obtain a domain D(H ) on whi h the Hamiltonian is self-adjoint. The Green
fun tions, the spe trum, and the unitary transformation that diagonalizes our Hamiltonian
will be also omputed. The diagonalization of the Hamiltonian will allow us to obtain the
energy (spe tral) representation and the dire t integral de omposition of the Hilbert spa e
indu ed by our Hamiltonian. We will see why this dire t integral de omposition is not
enough for the purposes of Quantum Me hani s and why the RHS is ne essary. Next, we
will onstru t the spa e . The RHS

  H   (4.1.9)

of the square barrier potential will follow. Dira kets will be a ommodated as elements of
, and the S hwartz delta fun tional will appear in the energy (spe tral) representation of
the triplet (4.1.9). The Nu lear Spe tral Theorem will be proved, and it will be shown that
this theorem is just a restatement of the (heuristi ) Dira basis ve tor expansion (4.1.3).
4.2 Sturm-Liouville Theory Applied to the Square Barrier Potential 141

4.2 Sturm-Liouville Theory Applied to the Square Bar-

rier Potential
4.2.1 S hrodinger Equation in the Position Representation
In order to al ulate the set of real generalized eigenvalues of the square barrier Hamil-
tonian (the physi al spe trum) and their orresponding generalized eigenve tors, we solve
equation (4.1.8) in the position representation,
h~xjH jE i = E h~xjE i : (4.2.1)
The expression of the Hamiltonian in the position representation is
 
~2
h~xjH jE i = 2m

+ V (~x) h~xjE i ; (4.2.2)

where
is the three-dimensional Lapla ian and

8
< 0 0<r<a
V (~x) = V (r) = V0 a < r < b (4.2.3)
:
0 b<r<1
is the square barrier potential. Writing Eqs. (4.2.1) and (4.2.2) in spheri al oordinates
and restri ting ourselves to the ase of zero angular momentum, we obtain the radial time-
independent S hrodinger equation,
 
~2d2
+ V (r) (r; E ) = E(r; E ) : (4.2.4)
2m dr2
Thus our Hamiltonian in the radial representation is given by the dierential operator
d2
~2
h + V (r ) : (4.2.5)
2m dr2
Throughout this hapter, the symbol h will be used to denote the formal dierential operator
(4.2.5).
The Sturm-Liouville theory studies the dierential operator
d d
(p(x) ) + q (x) ; (4.2.6)
dx dx
where p(x) and q (x) are fun tions of the real variable x, x running over an interval of
the real axis. In our example, x will be the radial oordinate r running over the interval
[0; 1), p(x) the onstant ~2 =2m and q (x) the square barrier potential (4.2.3). In this ase,
the Sturm-Liouville dierential operator (4.2.6) oin ides with the S hrodinger dierential
operator (4.2.5) and therefore we are allowed to apply the Sturm-Liouville theory to our
problem.
142 4 A Rigged Hilbert Spa e of the Square Barrier Potential

Mathemati ally, all the information about the dierential operator h provided by the
Sturm-Liouville theory (resolvent, spe trum, spe tral representation,...) is obtained from
the generalized eigenvalue equation
 
~2 d2
h(r; E ) = + V (r) (r; E ) = E(r; E ) ; E 2 C ; (4.2.7)
2m dr2
with various boundary onditions. As mentioned in the introdu tion, the \monoenergeti "
eigensolutions of (4.2.7) are not in general square integrable, i.e., they are not in the Hilbert
spa e. Those \monoenergeti " eigensolutions will be asso iated to antilinear fun tionals
FE 2  by Z 1
FE (')  dr '(r)(r; E ) : (4.2.8)
0
These fun tionals are generalized eigenve tors of the Hamiltonian H ,
H  FE = EFE ; (4.2.9)
or more pre isely,
h'jH jFE i = hH'jFE i = E h'jFE i ; 8' 2  : (4.2.10)
From a physi al point of view, Eq. (4.2.7) is the time-independent S hrodinger equation.
Dierent boundary onditions imposed upon it yield either Dira kets, Lippmann-S hwinger
kets or Gamow kets.

4.2.2 Self-Adjoint Extension

Our rst obje tive will be to dene a linear operator on a Hilbert spa e orresponding to
the formal dierential operator h and investigate its self-adjoint extensions. Among all
the possibilities, we shall hoose the self-adjoint extension that ts spheri ally symmetri
potentials. Later se tions will deal with the spe tral properties of this self-adjoint extension
and with the RHS indu ed by it.
The Hilbert spa e that is in the RHS of the square barrier potential is realized by the
spa e L2 ([0; 1); dr) of square integrable fun tions f (r) dened on the interval [0; 1). In this
se tion, we nd a subdomain D(H ) of this Hilbert spa e on whi h the dierential operator
h is self-adjoint. This domain must be a proper dense linear subspa e of L2 ([0; 1); dr). The
a tion of h must be well-dened on D(H ), and this a tion must remain in L2 ([0; 1); dr).
We need also a boundary ondition that assures the self-adjointness of the Hamiltonian.
Among all the possible boundary onditions that provide a self-adjoint extension (see Ap-
pendix 4.4.1), we hoose f (0) = 0. These requirements an be written as
f (r) 2 L2 ([0; 1); dr) ; (4.2.11a)
(hf )(r) 2 L2 ([0; 1); dr) ; (4.2.11b)
f (r) 2 AC 2 [0; 1) ; (4.2.11 )
f (0) = 0 ; (4.2.11d)
4.2 Sturm-Liouville Theory Applied to the Square Barrier Potential 143

where AC 2 [0; 1) denotes the spa e of fun tions whose derivative is absolutely ontinuous
( f. Appendix 4.4.1). Condition (4.2.11a) just means that the wave fun tions are square nor-
malizable. Condition (4.2.11b) assures that the a tion of h on any f (r) is square integrable.
Condition (4.2.11 ) is the weakest ondition suÆ ient for the se ond derivative of f (r) to
be well-dened. In our example, this ondition implies that f (r) and f 0 (r) are ontinuous
at r = a and at r = b. Equation (4.2.11d) sele ts, among all the possible possible boundary
onditions that provide a domain on whi h the dierential operator h is self-adjoint (see
Appendix 4.4.1), the self-adjoint extension needed in physi s.
The reason why we hoose (4.2.11d) is the following: in physi s [1, 54, 72, 73℄, the set
of boundary onditions imposed on the S hrodinger equation (4.2.7) always in ludes
(0; E ) = 0 ; (4.2.12a)
(r; E ); and 0 (r; E ) are ontinuous at r = a and at r = b : (4.2.12b)
Condition (4.2.12b) is implied by (4.2.11 ), so we just need to re over (4.2.12a). This is why
we impose (4.2.11d).
The set of onditions (4.2.11) leads to the domain
D(H ) = ff (r) j f (r); hf (r) 2 L2 ([0; 1); dr); f (r) 2 AC 2 [0; 1); f (0) = 0g : (4.2.13)
In hoosing (4.2.13) as the domain of our formal dierential operator h, we dene a linear
operator H by
 
d2
~2
(Hf )(r) := hf (r) = + V (r ) f (r ) ; f (r ) 2 D (H ) : (4.2.14)
2m dr2

4.2.3 Resolvent and Green Fun tions

The Green fun tion is the kernel of integration needed to write the resolvent of H as an
integral operator, Z 1
1
(E H ) f (r ) = G(r; s; E )f (s) ds : (4.2.15)
0
In Dira notation this equation reads
Z 1
hr j ( E H ) j f i =
1 hrj(E H ) 1jsihsjf i ds ; (4.2.16)
0
and therefore
G(r; s; E ) = hrj(E H ) 1 jsi : (4.2.17)
The so- alled outgoing and in oming Green fun tions are dened by
G (r; s; E ) = lim+ G(r; s; E  i) : (4.2.18)
!0

The pro edure to ompute the Green fun tion of our operator (4.2.14) is explained in [30℄
(see also [74℄). For the sake of ompleteness, we in lude in Appendix 4.4.2 the statement
144 4 A Rigged Hilbert Spa e of the Square Barrier Potential

of the theorem that is used to al ulate G(r; s; E ). The expression of the Green fun tion
will be given in terms of eigenfun tions of the dierential operator h subje t to dierent
boundary onditions ( f. Theorem 1 in Appendix 4.4.2).
We shall onsider three regions of the omplex plane and ompute the Green fun tion
for ea h region separately. In all our al ulations, we will use the following bran h of the
square root fun tion:
p
: fE 2 C j  < arg(E )   g 7 ! fE 2 C j =2 < arg(E )  =2g : (4.2.19)

Region <(E ) < 0, =(E ) 6= 0

For <(E ) < 0, =(E ) 6= 0, the Green fun tion (see Theorem 1 in Appendix 4.4.2) is given
by
8
> p 2m=~2 e s;E )
e(r;E ) (
r<s
< 2
2m=~ E 2 J
e 3 E)
(
G(r; s; E ) = <(E ) < 0 ; =(E ) 6= 0 : (4.2.20)
>
: p 22m=m=~~2 2 E e(s2;EJe)3(
e r;E )
(E )
r>s

The eigenfun tion e(r; E ) satises the S hrodinger equation (4.2.7) and the boundary on-
ditions
e(0; E ) = 0 ; (4.2.21a)
e(r; E ) 2 AC 2 ([0; 1)) ; (4.2.21b)
e(r; E ) is square integrable at 0 : (4.2.21 )
The boundary onditions (4.2.21) an be written as
e(0; E ) = 0 ; (4.2.22a)
e(a 0; E ) = e(a + 0; E ) ; (4.2.22b)
e0 (a 0; E ) = e0 (a + 0; E ) ; (4.2.22 )
e(b 0; E ) = e(b + 0; E ) ; (4.2.22d)
e0 (b 0; E ) = e0 (b + 0; E ) ; (4.2.22e)
e(r; E ) is square integrable at 0 ; (4.2.22f)
and lead to
8 q 2m q
2m Er
2 Er ~2
< e e 0<r<a
>
> ~
q q
2m (E V0 )r 2m (E V0 )r
e(r; E ) =
>
>
J
e1 (E )e
q
~ 2 + 2 (Eq)e
e J ~2 a<r<b (4.2.23)
: e 2m2 Er 2m2 Er
J
3 (E )e ~ J
+ e4 (E )e ~ b < r < 1:
The fun tions Je1 Je4 are su h that e(r; E ) satises the boundary onditions (4.2.22), and
their expressions are given in Eq. (4.4.12) of Appendix 4.4.2.
4.2 Sturm-Liouville Theory Applied to the Square Barrier Potential 145
e r ; E ) satises the S hr
The eigenfun tion ( odinger equation (4.2.7) and the boundary
onditions
e r ; E ) 2 AC 2 ([0; 1)) ;
( (4.2.24a)
e r ; E ) is square integrable at 1 :
( (4.2.24b)
The boundary onditions (4.2.24) an be written as
e a 0; E ) = (
( e a + 0; E ) ; (4.2.25a)
e 0 (a 0; E ) = 
 e 0 (a + 0; E ) ; (4.2.25b)
e b 0; E ) = (
( e b + 0; E ) ; (4.2.25 )
e 0 (b 0; E ) = 
 e 0 (b + 0; E ) ; (4.2.25d)
e r ; E ) is square integrable at 1 ;
( (4.2.25e)
and lead to
8 q
2m Er q
2m Er
>
>
< A
e1 (E )e
q
~2 A
+ e2 (E )e ~2
q
0<r<a
2m2 (E V0 )r 2m2 (E V0 )r
e r; E ) =
(
>
> q
A
e3 (E )e ~ A
+ e4 (E )e ~ a<r<b (4.2.26)
: 2m Er
e ~2 b < r < 1:

The fun tions Ae1 Ae4 are su h that (e r ; E ) satises the boundary onditions (4.2.25), and
their expressions are given in Eq. (4.4.13) of Appendix 4.4.2.
Region <(E ) > 0, =(E ) > 0
When <(E ) > 0, =(E ) > 0, the expression of the Green fun tion is
8
> p22m=~2 (r;E ) + (s;E ) r < s
< m=~2 E 2iJ4 (E )
G(r; s; E ) = ~2 (s;E ) + (r;E ) r > s
<(E ) > 0; =(E ) > 0 : (4.2.27)
>
: p22m=
m=~2 E 2iJ4 (E )

The eigenfun tion (r; E ) satises the S hrodinger equation (4.2.7) and the boundary on-
ditions (4.2.21),
8 q
>
> sin( 2m Er ) 0<r<a
>
< ~2q q
2m 2m
(r; E ) = J1(E )eiq ~2 (E V0)r + J2(Eq)e i ~2 (E V0 )r
a<r<b (4.2.28)
>
>
> 2m 2m
: J3(E )ei ~2 Er + J4(E )e i ~2 Er b < r < 1:
The fun tions J1 J4 are determined by the boundary onditions (4.2.22), and their ex-
pressions are listed in Eq. (4.4.16) of Appendix 4.4.2.
146 4 A Rigged Hilbert Spa e of the Square Barrier Potential

The eigenfun tion + (r; E ) satises the S hrodinger equation (4.2.7) and the boundary
onditions (4.2.24),
8 q q
>
> + (E )ei 2~m2 Er + + (E )e i 2~m2 Er
A A 0<r<a
< 1 q 2 q
2 i 2~m2 (E V0 )r
A A
m (E V )r
 + (r ; E ) = i
+ (E )e ~2
3
0 +
+ 4 (E )e a<r<b (4.2.29)
>
> q
: i 2m2 Er
e ~ b < r < 1:

The fun tions A+1 A+4 are determined by the boundary onditions (4.2.25), and their
expressions are listed in Eq. (4.4.17) of Appendix 4.4.2.
Region <(E ) > 0, =(E ) < 0
In the region <(E ) > 0, =(E ) < 0, the Green fun tion reads
8
> p22m=~2 (r;E )  (s;E ) r < s
< m=~2 E 2iJ3 (E )
G(r; s; E ) = ~2 (s;E )  (r;E ) r > s
<(E ) > 0; =(E ) < 0 : (4.2.30)
>
: p22m=
m=~2 E 2iJ3 (E )

The eigenfun tion (r; E ) is given by (4.2.28). The eigenfun tion  (r; E ) satises the
S hrodinger equation (4.2.7) and the boundary onditions (4.2.24),
8 q q
i 2~m2 Er 2m
A>
>
< 1 (E )e q A
+ 2 (E )e i ~2 Er q
0<r<a
(r ; E ) = A i 2~m2 (E V0 )r i 2~m2 (E V0 )r

>
>
3q( E ) e A
+ 4 ( E ) e a<r<b (4.2.31)
: i 2~m2 Er
e b < r < 1:

The fun tions A1 A4 are su h that  (r; E ) and its derivative are ontinuous at r = a
and at r = b. Their expressions are listed in Eq. (4.4.19) of Appendix 4.4.2.

4.2.4 Diagonalization of H and Eigenfun tion Expansion

In the present se tion, we diagonalize our Hamiltonian H and onstru t the expansion of
the wave fun tions in terms of the eigenfun tions of the dierential operator h. In order
to do so, we will ompute the spe trum of H and then onstru t a unitary operator U
that transforms from the position representation into the energy representation. We will
see that the spe trum of H is the positive real line [0; 1). In the energy representation, H
will a t as the multipli ation operator, the Hilbert spa e will be realized by L2 ([0; 1); dE )
and the domain of the Hamiltonian will be realized by the maximal domain on whi h the
multipli ation operator is well-dened. On our way, we shall take advantage of some results
of the Sturm-Liouville theory that are proved in [30℄. For the sake of ompleteness, we
in lude in Appendix 4.4.3 the statements of the theorems that are used in this se tion.
4.2 Sturm-Liouville Theory Applied to the Square Barrier Potential 147

Spe trum of H
We rst ompute the spe trum Sp(H ) of the operator H by applying Theorem 4 of Ap-
pendix 4.4.3 (see also [30℄). Sin e H is self-adjoint, its spe trum is real. This spe trum
is the subset of the real line on whi h the Green fun tion fails to be analyti . This non-
analyti ity of G(r; s; E ) will be built into the fun tions ij (E ) that appear in Theorem 4 of
Appendix 4.4.3.
From the expression of the Green fun tion omputed above, it is lear that the subsets
( 1; 0) and (0; 1) should be studied separately. We will denote either of these subsets by
.
Subset =( 1; 0)
We rst take  from Theorem 4 of Appendix 4.4.3 to be ( 1; 0). We hoose a basis
for the spa e of solutions of the equation h = E that is ontinuous on (0; 1)   and
analyti ally dependent on E as
8 q
2m Er q
2m Er
>
>
<
e1 B
( E ) eq
~2 + e2 ( B
E ) e ~2
q
0<r<a
1 (r; E ) = e3 (E )e 2~m2 (E V0 )r + e4 (E )e
B B
2m (E V0 )r
~2 a < r < b (4.2.32a)
>
> q
: 2m
e ~2 Er b < r < 1;

2 (r; E ) = (
e r; E ) : (4.2.32b)
The fun tions Be1 Be4 are su h that 1 (r; E ) and its derivative are ontinuous at r = a and
at r = b. Their expressions are listed in Eq. (4.4.29) of Appendix 4.4.3.
Obviously,
e(r; E ) = Je3 (E )1 (r; E ) + Je4 (E )2 (r; E ) ; (4.2.33)
whi h along with Eq. (4.2.20) leads to
" #
G(r; s; E ) = p
2m=~2 1
1 (r; E ) +
J e4 (E )
 (r; E ) 2 (s; E ) ;
2m=~2 E 2 J e3 (E ) 2
r < s ; <(E ) < 0 ; =(E ) 6= 0 : (4.2.34)
Sin e
2 (s; E ) = 2 (s; E ) ; (4.2.35)
we an write Eq. (4.2.34) as
" #
G(r; s; E ) = p
2m=~2 1
 ( r ; E )  ( s; E ) +
Je4(E )  (r; E ) (s; E ) ;
1 2
2m=~2 E 2 Je3(E ) 2 2

r < s ; <(E ) < 0 ; =(E ) 6= 0 : (4.2.36)

148 4 A Rigged Hilbert Spa e of the Square Barrier Potential

On the other hand, by Theorem 4 in Appendix 4.4.3 we have

2
X
G(r; s; E ) = ij (E )i (r; E )j (s; E ) r < s: (4.2.37)
i;j =1

By omparing Eqs. (4.2.36) and (4.2.37) we see that

0 ~2 1 1
0 p 2m=
2m=~2 E 2
ij (E ) = 
p 2m= ~2 1 Je4 (E )
A ; <(E ) < 0 ; =(E ) 6= 0 : (4.2.38)
0 2m=~2 E 2 Je3 (E )

The fun tions ij (E ) are analyti in a neighborhood of  = ( 1; 0). Therefore, the interval
( 1; 0) is in the resolvent set Re(H ) of the operator H .
Subset  = (0; 1)
Now we study the ase  = (0; 1). In order to be able to apply Theorem 4 of Ap-
pendix 4.4.3, we hoose the following basis for the spa e of solutions of h = E that is
ontinuous on (0; 1)   and analyti ally dependent on E :
1 (r; E ) = (r; E ) ; (4.2.39a)
8 q
>
> os( 2m Er ) 0<r<a
>
< 2
~q q
2 (r; E ) = i 2~m2 (E V0 )r i 2~m2 (E V0 )r
>
>
C1 (E )e q + 2 (Eq)e C a < r < b (4.2.39b)
> i 2~m2 Er i 2~m2 Er
: C3(E )e + C4 (E )e b < r < 1:
The fun tions C1 C4 , whose expressions are given in Eq. (4.4.30) of Appendix 4.4.3, are
su h that 2 and its derivative are ontinuous at r = a and at r = b.
Eqs. (4.2.29), (4.2.31) and (4.2.39) lead to

+ (r; E ) =
C4 (E )  (r; E ) + J4(E )  (r; E ) (4.2.40)
1 2
W (E ) W (E )
and to
 (r ; E ) =
C 3 (E )  (r ; E ) J3 (E )  (r ; E ) ; (4.2.41)
1 2
W (E ) W (E )
where
W (E ) = J4 (E )C3 (E ) J3 (E )C4 (E ) : (4.2.42)
By substituting Eq. (4.2.40) into Eq. (4.2.27) we get to

G(r; s; E ) = p
2m=~2 1 C4(E )  (r; E ) + J4(E )  (r; E )  (s; E ) ;
2m=~2 E 2iJ4 (E )
1
W (E ) W (E ) 2 1

<(E ) > 0; =(E ) > 0 ; r > s : (4.2.43)

4.2 Sturm-Liouville Theory Applied to the Square Barrier Potential 149

By substituting Eq. (4.2.41) into Eq. (4.2.30) we get to

 
G(r; s; E ) = p
2m=~2 1 C3 (E )  (r; E ) J3(E )  (r; E )  (s; E ) ;
2m=~2 E 2iJ3 (E ) W (E )
1
W (E ) 2 1

<(E ) > 0; =(E ) < 0 ; r > s ; (4.2.44)

Sin e
1 (s; E ) = 1 (s; E ) ; (4.2.45)
Eq. (4.2.43) leads to

G(r; s; E ) = p
2m=~2 1 C4 (E )  (r; E ) (s; E ) + J4(E )  (r; E ) (s; E )
2m=~ E 2iJ4 (E )
1 1
2 W (E ) W (E ) 2 1

<(E ) > 0; =(E ) > 0 ; r > s ; (4.2.46)

and Eq. (4.2.44) leads to
 
G(r; s; E ) = p
2m=~2 1 C3 (E )  (r; E ) (s; E ) J3(E )  (r; E ) (s; E )
2m=~2 E 2iJ3 (E ) W (E )
1 1
W (E ) 2 1

<(E ) > 0; =(E ) < 0 ; r > s ; (4.2.47)

The expression of the resolvent in terms of the basis 1 ; 2 an be written as (see Theorem 4
in Appendix 4.4.3)
2
X
G(r; s; E ) = ij+ (E )i (r; E )j (s; E ) r > s: (4.2.48)
i;j =1

By omparing (4.2.48) to (4.2.46) we get to

0 ~2 1 1
p22m= C4 (E )
m=~2 E 2i J4 (E )W (E )
0
ij+ (E ) = 
p22m=~2 1 1
A ; <(E ) > 0 ; =(E ) > 0 ; (4.2.49)
m=~2 E 2i W (E )
0

By omparing (4.2.48) to (4.2.47) we get to

0 ~2 1 C3 (E ) 1
p22m=
m=~2 E 2i J3 (E )W (E )
0
ij+ (E ) = 
p22m=~2 1 1
A ; <(E ) > 0 ; =(E ) < 0 ; (4.2.50)
m=~2 E 2i W (E )
0

From Eqs. (4.2.49) and (4.2.50) we an see that the measures 12 , 21 and 22 in Theorem 4
of Appendix 4.4.3 are zero and that the measure 11 is given by
Z
1 E2 Æ  + + (E + i) dE
11 ((E1 ; E2 )) = lim lim  11 ( E i)  11
Æ!0 !0+ 2i E1 +Æ
Z E2
1 2m=~2 1
= p dE ; (4.2.51)
E1 4 2m=~ E J3 (E )J4 (E )
2
150 4 A Rigged Hilbert Spa e of the Square Barrier Potential

whi h leads to
1 2m=~2 1
(E )  11 (E ) =
4
p
2m=~2 E jJ4(E )j2 ; E 2 (0; 1) : (4.2.52)

+ (E ) has a bran h ut along (0; 1), and therefore (0; 1) is in luded in

The fun tion 11
Sp(H ). Sin e Sp(H ) is a losed set, Sp(H ) = [0; 1). Thus the resolvent set of H is
Re(H ) = C [0; 1).

Diagonalization and Eigenfun tion Expansion

We are now in a position to diagonalize the Hamiltonian. By Theorem 2 of Appendix 4.4.3,
there is a unitary map Ue dened by
Ue : L2 ([0; 1); dr) 7 ! L2((0; 1); (E )dE ) Z
1
f (r) 7 ! fe(E ) = (Uf
e )(E ) = drf (r)(r; E ) ; (4.2.53)
0

that brings D(H ) onto the spa e

Z 1
D (Ee ) =f fe(E ) 2 L2 ((0; 1); (E )dE ) j dE E 2 jfe(E )j2 (E ) < 1g : (4.2.54)
0

Eqs. (4.2.53) and (4.2.54) provide a -diagonalization of H . If we seek a Æ -diagonalization,

i.e., if we seek eigenfun tions that are Æ -normalized, then the measure (E ) must be absorbed
by the eigenfun tions and by the wave fun tions.1 This is why we dene
p
 (r; E ) := (E ) (r; E ) ; (4.2.55)
whi h is the eigensolution of the dierential operator h that is Æ -normalized, and
p
fb(E ) := (E )fe(E ) ; fe(E ) 2 L2 ((0; 1); (E )dE ) ; (4.2.56)
and onstru t the unitary operator
Ub : L2 ((0; 1)); (E )dE ) 7 ! L2 ((0; 1); dE ) p
fe 7 ! fb(E ) = Ub fe(E ) := (E )fe(E ) : (4.2.57)
The operator that Æ -diagonalizes our Hamiltonian is U := Ub Ue ,
U : L2 ([0; 1)); dr) 7 ! L2 ((0; 1); dE )
f 7 ! Uf := fb: (4.2.58)
1
The meaning of the Æ -normalization of the eigenfun tions will be explained in Se tion 4.2.9.
4.2 Sturm-Liouville Theory Applied to the Square Barrier Potential 151

The a tion of U an be written as an integral operator,

Z 1
fb(E ) = (Uf )(E ) = drf (r) (r; E ) ; f (r) 2 L2 ([0; 1); dr) : (4.2.59)
0
The image of D(H ) under the a tion of U is
Z 1
D (Eb) := U D(H ) = f fb(E ) 2 L2 ((0; 1); dE ) j E 2 jfb(E )j2 dE < 1g : (4.2.60)
0
Therefore, we have onstru ted a unitary operator
U : D(H )  L2 ([0; 1); dr) 7 ! D(Eb)  L2 ((0; 1); dE )
f 7 ! fb = Uf (4.2.61)
that transforms from the position representation into the energy representation. The op-
erator U diagonalizes our Hamiltonian in the sense that Eb  UHU 1 is the multipli ation
operator,
Eb : D(Eb)  L2 ((0; 1); dE ) 7 ! L2((0; 1); dE )
fb 7 ! (Ebfb)(E ) := E fb(E ) : (4.2.62)
The inverse operator of U is given by (see Theorem 3 of Appendix 4.4.3)
Z 1
f (r ) = (U 1 fb)(r ) = dE fb(E ) (r; E ) ; fb(E ) 2 L2 ((0; 1); dE ) : (4.2.63)
0
The operator U 1 transforms from the energy representation into the position representa-
tion.
The expressions (4.2.59) and (4.2.63) provide the eigenfun tion expansion of any square
integrable fun tion in terms of the eigensolutions  (r; E ) of h.
The unitary operator U an be looked at as a sort of generalized Fourier transform: the
Fourier transform onne ts the position and the momentum representations. U onne ts the
position and the energy representations. The role played by the plane waves e ipx (whi h are
generalized eigenfun tions of the operator id=dx) is here played by the  (r; E ) (whi h are
generalized eigenfun tions of the dierential operator h). Therefore  (r; E )  hrjE i, whi h
is the Æ -normalized eigensolutions of the S hrodinger equation, an be viewed as \transition
elements" between the r- and the E -representations.
The label f of the fun tions in the position representation is dierent from the label fb
of the fun tions in the energy representation be ause they have dierent fun tional depen-
den es. The same applies to the Hamiltonian H , the domains, et . This is not the standard
pra ti e in the physi s literature, where dierent representations are usually identied and
labeled by the same symbol (see, for instan e, [72, 54, 73, 15℄).
We remark that the operator U that diagonalizes H is not unique. In fa t, dierent
eigenkets, i.e., dierent boundary onditions imposed upon (4.2.7), lead to dierent opera-
tors U .
152 4 A Rigged Hilbert Spa e of the Square Barrier Potential

4.2.5 The Need of the RHS

The Sturm-Liouville theory only provides a domain D(H ) on whi h the Hamiltonian H is
self-adjoint and a unitary operator U that diagonalizes H . This unitary operator indu es a
dire t integral de omposition of the Hilbert spa e (see [4, 5℄),
Z
H 7 ! UH  H =  b H(E )dE
Sp(H )
f 7 ! Uf  ffb(E )g; 2 H (E ) ; fb(E ) (4.2.64)
where H is realized by L2 ([0; 1); dr), and Hb is realized by L2 ([0; 1); dE ). The Hilbert
spa e H(E ) asso iated to ea h energy eigenvalue of Sp(H ) is realized by the Hilbert spa e
of omplex numbers C . On Hb , the operator H a ts as the multipli ation operator,
Hf 7 ! UHf  fE fb(E )g ; f 2 D(H ) : (4.2.65)
The s alar produ t on Hb an be written as
  Z  
f;
bb g = fb(E ); bg(E ) dE ; (4.2.66)
H
b Sp(H ) E

where the s alar produ t (
;
)E on H(E ) is the usual s alar produ t on C ,

 
fb(E ); bg(E ) = fb(E ) bg(E ) : (4.2.67)
E
As we shall explain below, the dire t integral de omposition does not shelter some of the
basi requirements needed in Quantum Me hani s. These requirements an be sheltered by
the RHS.
One of the most important prin iples of Quantum Me hani s is that the quantity ('; H')
should t the experimental expe tation value of the observable H in the state '. However,
('; H') is not dened for every element in H, but only for those square normalizable wave
fun tions that are also in D(H ). Therefore, not every square normalizable fun tion an
represent a \physi al wave fun tion," but only those that are (at least) in D(H ). Another
fundamental assumption of quantum physi s is that the quantity
disp' H = ('; H 2') ('; H')2 (4.2.68)
represents the dispersion of the observable H in the state ', and that
q

' H  disp' H (4.2.69)

represents the un ertainty of the observable H in the state '. Sin e (4.2.68) and (4.2.69)
are only dened when ' is an element of D(H 2 )  D(H ), not every element of D(H ) an
be assigned to a \physi al wave fun tion," but only those fun tions that are (at least) in
4.2 Sturm-Liouville Theory Applied to the Square Barrier Potential 153

D(H 2). Therefore, we would like to nd a subdomain  in luded in D(H ) on whi h the
expe tation values
('; H n') ; n = 0; 1; 2; : : : ; ' 2  (4.2.70)
are well-dened.
Another important requirement of Quantum Me hani s is that algebrai operations su h
as the sum and multipli ation of two operators are well-dened. In the HS formalism,
these algebrai operations are not always well-dened be ause the domains on whi h these
operators are self-adjoint do not remain stable under their a tions in general. In fa t,
mu h of the trouble of the HS formalism omes from domain questions. In our ase, the
domain D(H ) in (4.2.13) does not remain stable under H . We therefore would like to nd
a subdomain  in luded in D(H ) that remains stable under the a tion of H and all of its
powers,
H n :  7 !  ; n = 0; 1; 2; : : : (4.2.71)
One an see that if Eq. (4.2.71) holds, then the expe tation values (4.2.70) are well-dened
for ea h ' in , i.e., if the domain  remains stable under the a tion of H , then the
expe tation values of H in any state ' 2  are well-dened.
In Quantum Me hani s, it is always assumed that for ea h E 2 Sp(H ) there is a Dira
ket jE i su h that
H  jE i = E jE i (4.2.72)
and su h that the Dira basis ve tor expansion (4.1.3) holds. Equation (4.2.72) has no
solution in the Hilbert spa e when E belongs to the ontinuous part of the spe trum of the
Hamiltonian. In fa t, Eq. (4.2.72) has to be related to the equation
h~xjH jE i = E h~xjE i ; (4.2.73)
whi h in the radial representation reads
h (r; E ) = E (r; E ) ; (4.2.74)
where h is the dierential operator (4.2.5) and  (r; E ) is the delta-normalized eigenfun tion
(4.2.55). Sin e  (r; E )  hrjE i lies outside L2 ([0; 1); dr), i.e.,
Z 1
dr j (r; E )j2 = 1 ; (4.2.75)
0

the orresponding eigenket jE i, whi h is dened by

jE i :  7 ! C Z 1
' 7 ! h'jE i := '(r) (r; E )dr ; (4.2.76)
0

should also lie outside the Hilbert spa e. A tually, jE i is an element of  .

In summary, what our mathemati al framework should provide us with is:
154 4 A Rigged Hilbert Spa e of the Square Barrier Potential

1. a dense invariant domain on whi h all the powers of H and all the expe tation values
(4.2.70) are well-dened,
2. smooth enough wave fun tions so that Eq. (4.2.72) holds,
3. Dira basis ve tor expansion must follow.
In the dire t integral de omposition formalism, there is not enough room for either of these
three requirements. This is why we introdu e the RHS.

4.2.6 Constru tion of the Rigged Hilbert Spa e

The rst step is to make all the powers of the Hamiltonian well-dened. In order to do so,
we onstru t the maximal invariant subspa e D of the operator H ,
1
\
D := D (H n ) : (4.2.77)
n=0

The spa e D is the largest subspa e of D(H ) that remains stable under the a tion of the
Hamiltonian H and all of its powers. It is easy to he k that
D = f' 2 L2 ([0; 1); dr) j hn'(r) 2 L2 ([0; 1); dr); hn'(0) = 0; '(n)(a) = '(n)(b) = 0;
n = 0; 1; 2; : : : ; '(r) 2 C 1 ([0; 1))g : (4.2.78)
The onditions '(n) (a) = '(n) (b) = 0 in (4.2.78) ome from taking the dis ontinuities of the
potential V (r) at r = a and at r = b into onsideration (see [10℄).
The se ond step is to nd a subspa e  on whi h the eigenkets jE i of H are well-dened
as antilinear fun tionals. For ea h E 2 Sp(H ), we asso iate a ket jE i to the generalized
eigenfun tion  (r; E ) through
jE i :  7 ! C Z 1
' 7 ! h'jE i := '(r) (r; E )dr = (U')(E ) : (4.2.79)
0

As a tual omputations show, the ket jE i in (4.2.79) is a generalized eigenfun tional of H

if  is in luded in the maximal invariant subspa e of H ,
  D: (4.2.80)
Due to the non-square integrability of the eigenfun tion  (r; E ), we need to impose further
restri tions on the elements of D in order to make the eigenfun tional jE i in Eq. (4.2.79)
ontinuous, Z 1
dr j(r + 1)n(h + 1)m '(r)j2 < 1; n; m = 0; 1; 2; : : : (4.2.81)
0
4.2 Sturm-Liouville Theory Applied to the Square Barrier Potential 155

The imposition of onditions (4.2.81) upon the spa e D leads to the spa e of test fun tions
of the square barrier potential,
Z 1
 = f' 2 D j dr j(r + 1)n (h + 1)m '(r)j2 < 1; n; m = 0; 1; 2; : : :g : (4.2.82)
0
On , we dene the family of norms
sZ
1
k'kn;m := dr j(r + 1)n(h + 1)m '(r)j2 ; n; m = 0; 1; 2; : : : (4.2.83)
0

The quantities (4.2.83) fulll the onditions to be a norm ( f. Proposition 1 of Appendix 4.4.4)
and an be used to dene a ountably normed topology  on  (see [5℄),
! ' i k' 'kn;m !1

' !1 ! 0 ; n; m = 0; 1; 2; : : : (4.2.84)
One an see that the spa e  is stable under the a tion of H and that H is  - ontinuous
( f. Proposition 2 of Appendix 4.4.4).
On e we have onstru ted the spa e , we an onstru t its topologi al dual  as the
spa e of  - ontinuous antilinear fun tionals on  (see [5℄) and therewith the RHS of the
square barrier potential (for l = 0)
  L2 ([0; 1); dr)   : (4.2.85)
The ket jE i in Eq. (4.2.79) is a well-dened antilinear fun tional on , i.e., jE i belongs
to  ( f. Proposition 3 of Appendix 4.4.4). The ket jE i is a generalized eigenve tor of the
Hamiltonian H ( f. Proposition 3 of Appendix 4.4.4),
H  jE i = E jE i ; (4.2.86)
i.e.,
h'jH jE i = hH'jE i = E h'jE i ; 8' 2  : (4.2.87)
On the spa e , all the expe tation values of the Hamiltonian and all the algebrai
operations involving H are well-dened, and the generalized eigenvalue equation (4.2.87)
holds. As we shall see in the next se tion, the fun tions ' of  an be expanded by a Dira
basis ve tor expansion.

4.2.7 Dira Basis Ve tor Expansion

We are now in a position to derive the Dira basis ve tor expansion. This derivation onsists
of the restri tion of the Weyl-Kodaira expansions (4.2.59) and (4.2.63) to the spa e . If
we denote hrj'i  '(r) and hE jri   (r; E ), and if we dene the a tion of the left ket hE j
on ' 2  as hE j'i := 'b(E ), then Eq. (4.2.59) be omes
Z 1
hE j 'i = dr hE jri hrj'i ; ' 2  : (4.2.88)
0
156 4 A Rigged Hilbert Spa e of the Square Barrier Potential

If we denote hrjE i   (r; E ), then Eq. (4.2.63) be omes

Z 1
hr j 'i = dE hrjE i hE j'i ; ' 2  : (4.2.89)
0

This equation is the Dira basis ve tor expansion of the square barrier potential. In fa t,
when we formally write (4.1.3) in the position representation, we get to (4.2.89).
In Eq. (4.2.89), the wave fun tion hrj'i is spanned in a \Fourier-type" expansion by the
eigenfun tions hrjE i. In this expansion, ea h eigenfun tion hrjE i is weighted by hE j'i =
'b(E ), whi h is the value of the wave fun tion in the energy representation at the point
E . Thus any fun tion '(r) = hrj'i of  an be written as a linear superposition of the
monoenergeti eigenfun tions  (r; E ) = hrjE i.
Although the Weyl-Kodaira expansions (4.2.59) and (4.2.63) are valid for every element
of the Hilbert spa e, the Dira basis ve tor expansions (4.2.88) and (4.2.89) are only valid
for fun tions ' 2  be ause only those fun tions fulll both
'b(E ) = h'jE i (4.2.90)
and
h'jH jE i = hH'jE i = E h'jE i : (4.2.91)
Another way to rephrase the Dira basis ve tor expansion is the Nu lear Spe tral
(Gelfand-Maurin) Theorem. Instead of using the general statement of [5℄, we prove this theo-
rem using the ma hinery of the Sturm-Liouville theory (see Proposition 4 of Appendix 4.4.5).
The Nu lear Spe tral Theorem allows us to write the s alar produ t of any two fun tions
'; of  in terms of the a tion of the kets jE i on '; :
Z 1
('; ) = dE h'jE ihE j i ; 8'; 2  : (4.2.92)
0

It also allows us to write the matrix elements of the Hamiltonian and all of its powers
between two elements '; of  in terms of the a tion of the kets jE i on '; :
Z 1
('; H n )= dE E n h'jE ihE j i ; 8'; 2  ; n = 1; 2; : : : (4.2.93)
0

4.2.8 Energy Representation of the RHS

In this se tion, we onstru t the energy representation of the RHS. Sin e the unitary operator
U transforms from the position representation into the energy representation, the a tion of
U on the RHS provides the energy representation of the RHS.
We have already shown that in the energy representation the Hamiltonian H a ts as the
multipli ation operator Eb. The energy representation of the spa e  is dened as
b := U  : (4.2.94)
4.2 Sturm-Liouville Theory Applied to the Square Barrier Potential 157

It is very easy to see that 

b is a linear subspa e of L2 ([0; 1); dE ). In oder to endow 
b with
a topology b , we arry the topology on  into  b,

b := U : (4.2.95)

With this topology, the spa e b is a linear topologi al spa e. If we denote the dual spa e
of 
b by 
b  , then we have
U   = (U ) = 
b : (4.2.96)
If we denote jEb i  U  jE i, then we an prove that jEb i is the antilinear S hwartz delta
fun tional, i.e., jEb i is the antilinear fun tional that asso iates to ea h fun tion 'b the omplex
onjugate of its value at the point E (see Proposition 5 of Appendix 4.4.6),
jEbi : b 7 ! C
'b 7 ! h'bjEb i := 'b(E ) : (4.2.97)
Therefore, the S hwartz delta fun tional appears in the (spe tral) energy representation
of the RHS asso iated to the Hamiltonian. If we write the a tion of the S hwartz delta
fun tional as an integral operator, then the Dira Æ -fun tion appears in the integrand of
that integral operator.
It is very helpful to show the dierent realizations of the RHS through the following
diagram:
H ; '( r )   L2 ([0; 1); dr)   jE i position repr:
#U #U # U (4.2.98)
Eb ; 'b(E ) b  L2 ([0; 1); dE )  
b jEbi energy repr:
On the top line of the diagram (4.2.98), we have the RHS, the Hamiltonian, the wave
fun tions and the Dira kets in the position representation. On the bottom line, we have
their energy representation ounterparts.

4.2.9 Meaning of the Æ-normalization of the Eigenfun tions

In this se tion, we show that the Æ -normalization of the eigenfun tions is related to the
measure d(E ) that is used to ompute the s alar produ t of the wave fun tions in the
energy representation, Z 1
('; ) = '(E ) (E )d(E ) : (4.2.99)
0
We will see that if the measure in (4.2.99) is the Lebesgue measure dE , then the eigen-
fun tions are Æ -normalized, and that if the measure is (E )dE , then the eigenfun tions are
-normalized.
158 4 A Rigged Hilbert Spa e of the Square Barrier Potential

For the sake of simpli ity, in this se tion we label the wave fun tions in the position and
in the energy representation with the same symbol. With this notation, Eq. (4.2.89) leads
to Z 1
'( r ) = dE '(E ) (r; E ) ; (4.2.100a)
Z0 1
(r ) = dE (E ) (r; E ) : (4.2.100b)
0
Sin e '(r); (r) 2 L2 ([0; 1); dr), their s alar produ t is well-dened,
Z 1
('; ) = dr '(r) (r) : (4.2.101)
0
Plugging (4.2.100) into (4.2.101), we obtain
Z 1 Z 1 Z 1
('; ) = dE dE 0 '(E ) (E 0 ) dr  (r; E ) (r; E 0 ) : (4.2.102)
0 0 0
If we use the Lebesgue measure dE , then the s alar produ t (4.2.99) an be written as
Z 1
('; ) = dE '(E ) (E ) : (4.2.103)
0
Comparison of (4.2.102) and (4.2.103) leads to
Z 1
dr  (r; E ) (r; E 0) = Æ (E E 0) ; (4.2.104)
0
i.e., the eigenfun tions  (r; E ) are Æ -normalized.
We now onsider the ase in whi h the eigenfun tions are -normalized. If we use the
measure d(E ) = (E )dE , then the s alar produ t of ' and is given by
Z 1
('; ) = ' (E )  (E )(E ) dE ; (4.2.105)
0
p p
where 'p  (E ) := '(E )= (E ) and  (E ) := (E )= (E ). If we dene  (r; E ) :=
 (r; E )= (E ), then Eq. (4.2.100) leads to
Z 1
'( r ) = ' (E ) (r; E )(E ) dE ; (4.2.106a)
Z0 1
(r ) =  (E ) (r ; E )(E ) dE : (4.2.106b)
0
Plugging Eq. (4.2.106) into (4.2.101), we obtain
Z 1 Z 1 Z 1
('; ) = dE dE 0 ' (E )  (E 0 )(E )(E 0 ) dr  (r; E ) (r; E 0 ) : (4.2.107)
0 0 0
Comparison of (4.2.107) and (4.2.105) leads to
Z 1
1
dr  (r; E ) (r; E 0 ) = Æ (E E 0) ; (4.2.108)
0 (E )
i.e., the eigenfun tions  (r; E ) are -normalized.
4.3 Con lusion to Chapter 4 159

4.3 Con lusion to Chapter 4

In this hapter, we have onstru ted the Rigged Hilbert Spa e of the square barrier Hamil-
tonian
  L2 ([0; 1); dr)   (4.3.1)
and its energy representation

b  L2 ([0; 1); dE )  b  : (4.3.2)

The spe trum of the Hamiltonian H is the positive real semiaxis. For ea h value E of the
spe trum of H , we have onstru ted a Dira ket jE i that is a generalized eigenfun tional of
H whose orresponding generalized eigenvalue is E . In the energy representation, jE i a ts as
the antilinear S hwartz delta fun tional. On the spa e , all algebrai operations involving
the Hamiltonian H are well-dened. In parti ular, the expe tation values of the Hamiltonian
in any element of  are well-dened. Any element of  an be expanded in terms of the
eigenkets jE i by a Dira basis ve tor expansion. The elements of  are represented by well-
behaved fun tions in ontrast to the elements of the Hilbert spa e whi h are represented by
sets of equivalent fun tions that an vary arbitrarily on any set of zero Lebesgue measure.
Therefore, it seems natural to on lude that a physi ally a eptable wave fun tion is not
any element of the Hilbert spa e, but rather an element of the subspa e .
In our quest for the RHS of the square barrier potential, we have found a systemati
method to onstru t the RHS of a large lass of spheri ally symmetri potentials:

1. Expression of the formal dierential operator.

2. Hilbert spa e H of square integrable fun tions on whi h the formal dierential operator
a ts.

3. A domain D(H ) of the Hilbert spa e on whi h the formal dierential operator is
self-adjoint.

4. Green fun tions (resolvent) of this self-adjoint operator.

5. Diagonalization of the self-adjoint operator, eigenfun tion expansion of the elements of

H in terms of the eigensolutions of the formal dierential operator, and dire t integral
de omposition of H indu ed by the self-adjoint operator.

6. Subspa e  of D(H ) on whi h all the expe tation values of H are well-dened and on
whi h the Dira kets a t as antilinear fun tionals.

7. Rigged Hilbert spa e   H   .

160 4 A Rigged Hilbert Spa e of the Square Barrier Potential

4.4 Appendi es to Chapter 4

4.4.1 Appendix 1: Self-Adjoint Extension
In this appendix, we list the possible self-adjoint extensions asso iated to the dierential
operator h. We rst need rst some denitions ( f. [30℄).
Denition 1 By AC 2 ([0; 1)) we denote the spa e of all fun tions f whi h have a
ontinuous derivative in [0; 1), and for whi h f 0 is not only ontinuous but also absolutely
ontinuous over ea h ompa t subinterval of [0; 1). Thus f (2) exists almost everywhere,
and is integrable over any ompa t subinterval of [0; 1). At 0 f 0 is ontinuous from the
right.
The spa e AC 2 ([0; 1)) is the largest spa e of fun tions on whi h the dierential operator
h an be dened. In the ase of the square barrier potential (4.2.3), if f (r) belongs to
AC 2 ([0; 1)), then f (r) and f 0 (r) are ontinuous at r = a and at r = b.
Denition 2 We dene the spa es
Hh2 ([0; 1)) := ff 2 AC 2([0; 1)) j f; hf 2 L2([0; 1); dr)g (4.4.1)
H2 ([0; 1)) := ff 2 AC 2 ([0; 1) j f; f (2) 2 L2 ([0; 1); dr)g (4.4.2)
H02 ([0; 1)) := ff 2 H2 ([0; 1)) j f vanishes outside some
ompa t subset of (0; 1)g : (4.4.3)
Using these spa es, we an dene the ne essary operators to al ulate the self-adjoint
extensions asso iated to h.

Denition 3 If h is the formal dierential operator (4.2.5), we dene the operators

H0 and H1 on L2 ([0; 1); dr) by the formulas
D(H0) = H02 ([0; 1)); H0f := hf; f 2 D(H0) : (4.4.4)
D(H1) = Hh2 ([0; 1)); H1 f := hf; f 2 D(H1) : (4.4.5)

The operators H0 and H1 are sometimes alled the minimal and the maximal operators
asso iated to the dierential operator h, respe tively. The domain D(H1 ) is the largest
domain of the Hilbert spa e L2 ([0; 1); dr) on whi h the a tion of the dierential operator
h an be dened and remains inside L2 ([0; 1); dr). Further, H0y = H1 .
The self-adjoint extensions of H0 are given by the restri tions of the operator H1 to
domains determined by the onditions (see [30℄, page 1306)
f (0) +  f 0 (0) = 0 ; 1 <   1: (4.4.6)
4.4 Appendi es to Chapter 4 161

These boundary onditions lead to the domains

D(H ) = ff 2 D(H1) j f (0) +  f 0(0) = 0g ; 1 <   1: (4.4.7)
On these domains, the formal dierential operator h is self-adjoint. The boundary ondition
that ts spheri ally symmetri potentials is f (0) = 0, i.e.,  = 0. This ondition sele ts our
domain (4.2.13),
D(H ) = D=0 (H ) = ff 2 D(H1) j f (0) = 0g : (4.4.8)

4.4.2 Appendix 2: Resolvent and Green Fun tion

The following theorem provides the pro edure to ompute the Green fun tion of the Hamil-
tonian H ( f. Theorem XIII.3.16 of Ref. [30℄):
Theorem 1 Let H be the self-adjoint operator (4.2.14) derived from the real formal
dierential operator (4.2.5) by the imposition of the boundary ondition (4.2.11d). Let
=E 6= 0. Then there is exa tly one solution (r; E ) of (h E ) = 0 square-integrable
at 0 and satisfying the boundary ondition (4.2.11d), and exa tly one solution (r; E ) of
(h E ) = 0 square-integrable at innity. The resolvent (E H ) 1 is an integral operator
whose kernel G(r; s; E ) is given by
8
< 2~m (r;E ) (s;E )
2 W (;) r<s
G(r; s; E ) = (4.4.9)
: 2m (s;E ) (r;E ) r > s;
~2 W (;)

where W (; ) is the Wronskian of  and 

W (; ) = 0 0  : (4.4.10)

If we dene
r
2m
k
e := E; (4.4.11a)
~2
r
2m
Qe := (E V0 ) ; (4.4.11b)
~2

then the fun tions Je(E ) of Eq. (4.2.23) are given by

" ! ! #
k k
Je1(E ) = 12 e e e e e e
Qa 1+ eka + 1+ e ka ; (4.4.12a)
Qe Qe
" ! ! #
k eka k
Je2(E ) = 12 eQa
e e e e
1 e + 1 e ka ; (4.4.12b)
Qe Qe
162 4 A Rigged Hilbert Spa e of the Square Barrier Potential
" ! ! #
Qe Qb Qe
Je3(E ) = 21 e e
kb 1+ e J
e e
1 (E ) + 1
e e
e Qb J2(E ) ; (4.4.12 )
k
e k
e
" ! ! #
Qe Qb Q
Je4(E ) = 21 eekb
e
1 e e e Je1 (E ) + 1 + e e Qb J2(E ) ;
e e
(4.4.12d)
k k
and the fun tions Ae(E ) of Eq. (4.2.26) by
!
k
Ae3(E ) = 12 e e e e
Qb 1 ; e kb (4.4.13a)
Qe
!
1 k
e
Ae4(E ) = 2 eQb 1 + e e ekb;
e
(4.4.13b)
Q
" ! ! #
1 Q
e Qe
Ae1(E ) = 2 e ka 1 + e eQa Ae3(E ) + 1 e e Qa Ae4(E ) ;
e e e
(4.4.13 )
k k
1
"
Qe
!
Q
e
!
pe #
Ae2(E ) = 2 eka 1 e eQaAe3(E ) + 1 + e e a Ae4(E ) :
e e
(4.4.13d)
k k
The expression for the Wronskian of e and 
e is

W (e;  kJe3 (E ) :
e ) = 2e (4.4.14)
If we dene
r
2m
k := E; (4.4.15a)
~2
r
2m
Q := (E V0 ) ; (4.4.15b)
~2
then the fun tions J (E ) of Eq. (4.2.28) are given by
 
1 iQa k
J1 (E ) = 2
e sin(ka) + os(ka) ;
iQ
(4.4.16a)
 
1 iQa k
J2 (E ) = 2
e sin(ka)
iQ
os(ka) ; (4.4.16b)
    
1 ikb Q iQb Q
J3 (E ) = 2
e 1+
k
e J1 (E ) + 1
k
e J2 (E ) ;
iQb (4.4.16 )
    
1 ikb Q iQb Q
J4 (E ) = 2
e 1
k
e J1 (E ) + 1 +
k
e J2 (E ) ;
iQb (4.4.16d)

and the fun tions A+ (E ) of Eq. (4.2.29) by

 
1 iQb k ikb
A3 (E ) = 2 e
+ 1+
Q
e ; (4.4.17a)
4.4 Appendi es to Chapter 4 163
 
1 k ikb
A 4 = eiQb 1
+ (E )
2 Q
e ; (4.4.17b)
    
1 ika Q iQa + Q
A1 (E ) = 2 e
+ 1+
k
e A3 (E ) + 1
k
e A4 (E ) ; (4.4.17 )
iQa +
    
1 ika Q iQa + Q
A2 (E ) = 2 e
+ 1
k
e A3 (E ) + 1 +
k
e A4 (E ) : (4.4.17d)
iQa +

The Wronskian of  and + is

W (; +) = 2ikJ4 (E ) : (4.4.18)
The fun tions A (E ) of Eq. (4.2.31) are given by
 
1 iQb k
A3 (E ) = 2
e 1
Q
e ikb ; (4.4.19a)
 
1 iQb k
A4 (E ) = 2
e 1+
Q
e ikb ; (4.4.19b)
    
1 ika Q iQa Q
A1 (E ) = 2
e 1+
k
e A3 (E ) + 1
k
e A4 (E ) ;
iQa (4.4.19 )
    
1 ika Q iQa Q
A2 (E ) = 2
e 1
k
e A3 (E ) + 1 +
k
e A4 (E ) :
iQa (4.4.19d)

The Wronskian of  and  is

W (;  ) = 2ikJ3 (E ) : (4.4.20)

4.4.3 Appendix 3: Diagonalization and Eigenfun tion Expansion

The theorem that provides the operator U that diagonalizes H is ( f. Theorem XIII.5.13 of
Ref. [30℄)
Theorem 2 (Weyl-Kodaira) Let h be the formally self-adjoint dierential operator
(4.2.5) dened on the interval [0; 1). Let H be the self-adjoint operator (4.2.14). Let  be
an open interval of the real axis, and suppose that there is given a set 1 , 2 of fun tions,
dened and ontinuous on (0; 1)  , su h that for ea h xed E in , 1 (
; E ), 2 (
; E )
forms a basis for the spa e of solutions of h = E . Then there exists a positive 2  2
matrix measure fij g dened on , su h that
1. the limit Z d 
[(Uf )i (E ) = lim
!0
lim f (r)i (r; E )dr (4.4.21)
d!1
exists in the topology of L2 (; fij g) for ea h f in L2 ([0; 1); dr) and denes an iso-
metri isomorphism U of E ()L2 ([0; 1); dr) onto L2 (; fij g);
164 4 A Rigged Hilbert Spa e of the Square Barrier Potential

2. for ea h Borel fun tion G dened on the real line and vanishing outside ,
U D(G(H )) = f[fi ℄ 2 L2 (; fij g) j [Gfi ℄ 2 L2 (; fij g)g (4.4.22)
and
(UG(H )f )i (E ) = G(E )(Uf )i (E ); i = 1; 2; E 2 ; f 2 D(G(H )) : (4.4.23)

The theorem that provides the inverse of the operator U is ( f. Theorem XIII.5.14 of
Ref. [30℄)
Theorem 3 (Weyl-Kodaira) Let H , , fij g, et ., be as in Theorem 2. Let E0 and E1
be the end points of . Then
1. the inverse of the isometri isomorphism U of E ()L2 ([0; 1); dr) onto L2 (; fij g) is
given by the formula
Z 1 !
2
X
(U 1 F )(r) = lim lim Fi (E )j (r; E )ij (dE ) (4.4.24)
0 !E0 1 !E1 0
i;j =1

where F = [F1 ; F2 ℄ 2 L2 (; fij g), the limit existing in the topology of L2 ([0; 1); dr);
2. if G is a bounded Borel fun tion vanishing outside a Borel set e whose losure is
ompa t and ontained in , then G(H ) has the representation
Z 1
(G(H )f )(r) = f (s)K (H; r; s)ds ; (4.4.25)
0
where 2 Z
X
K (H; r; s) = G(E )i (s; E )j (r; E )ij (dE ) : (4.4.26)
i;j =1 e

The spe tral measures are provided by the following theorem ( f. Theorem XIII.5.18 of
Ref. [30℄):
Theorem 4 (Tit hmarsh-Kodaira) Let  be an open interval of the real axis and O
be an open set in the omplex plane ontaining . Let 1 , 2 be a set of fun tions whi h
form a basis for the solutions of the equation h = E , E 2 O, and whi h are ontinuous on
(0; 1)  O and analyti ally dependent on E for E in O. Suppose that the kernel G(r; s; E )
for the resolvent (E H ) 1 has a representation
8 P
2
i;j =1 ij (E )i (r ; E )j (s; E ) r<s
>
<
G(r; s; E ) = (4.4.27)
>
: P2 +
i;j =1 ij (E )i (r ; E )j (s; E ) r > s;
4.4 Appendi es to Chapter 4 165

for all E in Re(H ) \ O, and that fij g is a positive matrix measure on  asso iated with H
as in Theorem 2. Then the fun tions ij are analyti in Re(H ) \ O, and given any bounded
open interval (E1 ; E2 )  , we have for 1  i; j  2,
R E2 Æ
ij ((E1 ; E2 )) = limÆ!0 lim!0+ 21i E1 +Æ [ij (E i) ij (E + i)℄dE
R
(4.4.28)
= limÆ!0 lim!0+ 21i EE12+ÆÆ [ij+ (E i) ij+ (E + i)℄dE :

The fun tions Be(E ) of Eq. (4.2.32a) are given by

!
k
Be3(E ) = 12 e
e
e
Qb 1+ eekb ; (4.4.29a)
Qe
!
1 k
e
Be4(E ) = 2 eQb 1 e eekb;
e
(4.4.29b)
Q
" ! ! #
1 Q
e Qe
Be1(E ) = 2 e ka 1 + e eQa Be3 (E ) + 1 e e QaBe4 (E ) ;
e e e
(4.4.29 )
k k
" ! ! #
1 Qe Q
e
Be2(E ) = 2 eka 1 e eQa Be3 (E ) + 1 + e e QaBe4 (E ) :
e e e
(4.4.29d)
k k
The fun tions C (E ) of Eq. (4.2.39b) are given by
 
1 iQa k
C1(E ) = 2
e os(ka)
iQ
sin(ka) ; (4.4.30a)
 
1 iQa k
C2(E ) = 2
e os(ka) + sin(ka) ;
iQ
(4.4.30b)
    
1 ikb Q iQb Q
C3(E ) = 2
e 1+
k
e C1 (E ) + 1
k
e C 2 (E ) ;
iQb (4.4.30 )
    
1 ikb Q iQb Q
C4(E ) = 2
e 1
k
e C 1 (E ) + 1 +
k
e C 2 (E ) :
iQb (4.4.30d)

4.4.4 Appendix 4: Constru tion of the RHS

Proposition 1 The quantities
sZ
1
k'kn;m := dr j(r + 1)n (h + 1)m '(r)j2 ; n; m = 0; 1; 2; : : : ; ' 2  ; (4.4.31)
0

are norms.
166 4 A Rigged Hilbert Spa e of the Square Barrier Potential

Proof It is very easy to show that the quantities (4.4.31) fulll the onditions to be a
norm,
k' + kn;m  k'kn;m + k kn;m ; (4.4.32a)
k'kn;m = jj k'kn;m ; (4.4.32b)
k'kn;m  0 ; (4.4.32 )
If k'kn;m = 0; then ' = 0 : (4.4.32d)
The only ondition that is somewhat diÆ ult to prove is (4.4.32d): if k'kn;m = 0, then
(1 + r)n (h + 1)m '(r) = 0 ; (4.4.33)
whi h yields
(h + 1)m '(r) = 0 : (4.4.34)
If m = 0, then Eq. (4.4.34) implies '(r) = 0. If m = 1, then Eq. (4.4.34) implies that 1 is
an eigenvalue of H whose orresponding eigenve tor is '. Sin e 1 is not an eigenvalue of
H , ' must be the zero ve tor. If m > 1, the proof is similar.
Proposition 2 The spa e  is stable under the a tion of H , and H is  - ontinuous.
Proof In order to see that H is  - ontinuous, we just have to realize that
kH'kn;m = k(H + I )' 'kn;m
 k(H + I )'kn;m + k'kn;m
= k'kn;m+1 + k'kn;m : (4.4.35)
We now prove that  is stable under the a tion of H . Let ' 2 . To say that ' 2  is
equivalent to say that ' 2 D and that the norms k'kn;m are nite for every n; m = 0; 1; 2; : : :
Sin e H' is also in D, and sin e the norms kH'kn;m are also nite (see Eq. (4.4.35)), the
ve tor H' is also in .
Proposition 3 The fun tion
jE i :  7 ! C Z 1
' 7 ! h'jE i := '(r) (r; E )dr = (U')(E ) : (4.4.36)
0

is an antilinear fun tional on  that is a generalized eigenve tor of (the restri tion to  of)
H.
Proof From the denition (4.4.36), it is pretty easy to see that jE i is an antilinear
fun tional. In order to show that jE i is ontinuous, we dene
M(E ) := sup j (r; E )j : (4.4.37)
r2[0;1)
4.4 Appendi es to Chapter 4 167

Sin e
jh'jE ij = jU'(E )j
Z 
 1 
= 
 dr '(r) (r; E )
Z 01
 dr j'(r)jj (r; E )j
0 Z 1
 M(E ) dr j'(r)j
Z0 1
1
= M(E ) 1+r
(1 + r)j'(r)j
dr
0
Z 1 1=2 Z 1 1=2
1
 M(E ) dr
(1 + r)2
dr j(1 + r)'(r)j2
0 0
Z 1 1=2
1
= M( E ) dr k'k1;0
0 (1 + r)2
= M(E )k'k1;0 ; (4.4.38)
the fun tional jE i is ontinuous when  is endowed with the  topology.
In order to prove that jE i is a generalized eigenve tor of H , we make use of the onditions
(4.2.78) and (4.2.81) satised the elements of ,
h'j H  j E i = hH'jE i
Z 1 
d2
~2
= dr + V ( r ) '( r )  ( r ; E )
0 2m dr2
" #1  
~2 d'(r ) ~2 d (r; E ) 1
=  (r ; E ) + '( r )
2m dr 0
2 m dr 0
Z 1  2 2 
~ d
+ dr '(r) + V (r )  (r ; E )
0 2m dr2
= E h'j E i : (4.4.39)
Similarly, one an also prove that
h'j ( H  ) n j E i = E n h'j E i : (4.4.40)

4.4.5 Appendix 5: Dira Basis Ve tor Expansion

Proposition 4 (Nu lear Spe tral Theorem) Let
  L2 ([0; 1); dr)   (4.4.41)
be the RHS of the square barrier Hamiltonian H su h that  remains invariant under H
and H is a  - ontinuous operator on . Then, for ea h E in the spe trum of H there is a
168 4 A Rigged Hilbert Spa e of the Square Barrier Potential

generalized eigenve tor jE i su h that

H  jE i = E jE i (4.4.42)
and su h that Z
('; ) = dE h'jE ihE j i ; 8'; 2  ; (4.4.43)
Sp(H )
and Z
('; H n )= dE E n h'jE ihE j i ; 8'; 2  ; n = 1; 2; : : : (4.4.44)
Sp(H )

Proof Let ' and be in . Sin e U is unitary,

('; ) = (U'; U ) = (';
b b) : (4.4.45)
The wave fun tions 'b and b are in parti ular elements of L2 ([0; 1); dE ). Therefore their
s alar produ t is well-dened,
Z
(';
b b) = dE 'b(E ) b(E ) : (4.4.46)
Sp(H )

Sin e ' and belong to , the a tion of ea h eigenket jE i on them is well-dened,
h'jE i = 'b(E ) ; (4.4.47a)
hE j i = b(E ) : (4.4.47b)
Plugging Eq. (4.4.47) into Eq. (4.4.46) and Eq. (4.4.46) into Eq. (4.4.45), we get to Eq. (4.4.43).
The proof of (4.4.44) is similar:
('; H n ) = (U'; UH n U 1 U )
= (Z;
b Eb n b)

= dE 'b(E )(Eb n b)(E )

Sp(H )
Z
= dE E n 'b(E ) b(E )
Sp(H )
Z
= dE E n h'jE ihE j i : (4.4.48)
Sp(H )

4.4.6 Appendix 6: Energy Representation of the RHS

Proposition 5 The energy representation jEbi of the eigenket jE i is the antilinear S hwartz
delta fun tional.
4.4 Appendi es to Chapter 4 169

Proof Sin e
h'bjU jE i = hU 1'bjE i
= hZ'jE i
1
= '(r) (r; E )dr
0
= 'b(E ) ; (4.4.49)
the fun tional U  jE i = jEb i is the antilinear S hwartz delta fun tional.
Chapter 5
S attering o the Square Barrier
Potential
In this hapter, we turn to the des ription of the Lippmann-S hwinger equation within the
RHS formalism. The Lippmann-S hwinger eigenfun tions will be omputed rst. We shall
dene the Lippmann-S hwinger eigenkets in terms of these eigenfun tions and see that they
a t on dierent spa es of wave fun tions. The Lippmann-S hwinger kets will be used as
basis ve tors to expand the wave fun tions. The Mller operators and the S -matrix will be
expli itly onstru ted.

It is so hard to be good!

Thales of Miletus

171
5.1 Introdu tion 173

5.1 Introdu tion

In the previous hapter, we onstru ted a RHS of the square barrier potential. The RHS was
meant to in orporate ertain boundary onditions imposed upon the S hrodinger equation:

S hrodinger equation
+ !   H  .
boundary onditions

The Hilbert spa e H was needed to in orporate the requirement that the wave fun tions be
square integrable. The spa e  was needed to in orporate the Dira kets asso iated to the
eigenfun tions of the time independent S hrodinger equation subje t to ertain boundary
onditions. The spa e  was needed to in orporate the wave fun tions on whi h the Dira
kets a t as ontinuous antilinear fun tionals. The spa e  was identied with the spa e of
physi ally preparable wave fun tions, be ause in  all of the algebrai operations and all of
the expe tation values are well dened.
In this hapter, we onsider the problem of s attering o the square barrier potential.
Loosely speaking, we send a beam of prepared initial in-states 'in towards the square barrier
potential. After the ollision takes pla e, the in-state 'in be omes 'out . We then measure
the probability to nd a nal out-state out . The amplitude of this probability is given by
the following s alar produ t:
( out ; 'out ) =( out ; S'in ) ; (5.1.1)
where S is the S -matrix. The anoni al understanding is that the initial in-states 'in and
the nal out-states out are asymptoti forms of the so- alled in-state ve tor '+ and out-
state ve tor in the remote past and in the distant future, respe tively. In terms of these,
the probability amplitude (5.1.1) an be written as
( ; '+ ) : (5.1.2)
The asymptoti states 'in and out are related to the \exa t" states '+ and by the
so- alled Mller operators:

+ 'in = '+ ; (5.1.3a)

out = : (5.1.3b)
Along with the (total) Hamiltonian H , it is ustomary to onsider another \free" Hamil-
tonian H0 , whi h is assumed to dier from H by the (square barrier) potential V ,
H = H0 + V : (5.1.4)
The potential V represents the intera tion between the omponents of the initial prepared
states, for instan e, the intera tion between the in-going beam and the target. The anoni al
174 5 S attering o the Square Barrier Potential

understanding is that the initial in-state 'in and the nal out-state out evolve under the
in uen e of the free Hamiltonian H0 , whereas the in-state '+ and the out-state evolve
under the in uen e of the (total) Hamiltonian H .
The dynami s of a s attering system is therefore governed by the S hrodinger equation
subje t to ertain boundary onditions. These boundary onditions spe ify what is \in"
and what is \out."
The Lippmann-S hwinger equation for the in- and out-kets jE  i has the s attering
boundary onditions built into it. As we shall see, the Lippmann-S hwinger equation tells
us what is \in" and what is \out" by spe ifying ertain analyti al properties of its solutions.
The analyti al properties satised by the in-ket jE + i (or, equivalently, by the wave fun tion
h'+jE +i) are dierent to those satised by the out-ket jE i (or, equivalently, by the wave
fun tion h jE i). In in orporating these two dierent types of boundary onditions into
a RHS framework, we will end up onstru ting two dierent RHSs:
  H   : (5.1.5)
On our way, we will also onstru t the Mller operators and the S-matrix, and we will
express the matrix element (5.1.2) in terms of the in- and out-Lippmann-S hwinger kets:
Z 1
( ; '+) = dE h jE iS (E )h+E j'+i : (5.1.6)
0
This expression will be used to derive the omplex basis ve tor expansion in Chapter 6.

5.2 Lippmann-S hwinger Equation

5.2.1 Lippmann-S hwinger Kets
One of the fundamental equations of s attering theory is the Lippmann-S hwinger equation,1
jE i = jE i + E H1  i V jE i : (5.2.1)
0
This equation is also written as
jE i = jE i + E H1  i V jE i : (5.2.2)
In Eqs. (5.2.1) and (5.2.2), the kets jE  i represent generalized eigenve tor of the total
Hamiltonian H ,
H  jE  i = E jE  i ; (5.2.3)
whereas jE i represents a generalized eigenve tor of the free Hamiltonian H0 ,
H0 jE i = E jE i : (5.2.4)
1
In this hapter, the symbol jE i will denote the generalized eigenket of the free Hamiltonian that appears
in the Lippmann-S hwinger equation (5.2.1), and not an eigenket of the total Hamiltonian as in Chapter 4.
5.2 Lippmann-S hwinger Equation 175

5.2.2 Radial Representation of the Lippmann-S hwinger Equa-

tion
Sin e our square barrier potential is spheri ally symmetri , we shall work in the radial
representation. In this representation and for l = 0, H0 a ts as the formal dierential
operator h0 ,
~2 d2
H0 f (r) = h0 f (r) = f (r ) ; (5.2.5)
2m dr2
V a ts as multipli ation by the square barrier potential V (r),
8
< 0 0<r<a
V (r ) = V0 a < r < b (5.2.6)
:
0 b < r < 1;
and H a ts as the formal dierential operator h,
 
~2 d2
Hf (r) = hf (r) = + V (r ) f (r ) : (5.2.7)
2m dr2
In the radial representation, Eqs. (5.2.1) and (5.2.2) be ome

hrjE i = hrjE i + hrj E H1  i V jE i ; (5.2.8a)

0
1
hrjE i = hrjE i + hrj E H  i V jE i : (5.2.8b)

In Eq. (5.2.8), the quantities hrjE i are eigenfun tions of the formal dierential operator h0 ,
~2 d2
h0 hrjE i = hr j E i = E hr j E i ; (5.2.9)
2m dr2
whereas the quantities hrjE i are eigenfun tions of the formal dierential operator h satis-
fying proper boundary onditions (that we will spe ify later),
 
~2 d2
hhrjE i = + V (r ) hr j E  i = E hr j E  i : (5.2.10)
2m dr2
In the absen e of potential, the Lippmann-S hwinger eigenfun tions tend to the free Hamil-
tonian eigenfun tions,
lim hrjE  i = hrjE i : (5.2.11)
V0 !0
The generalized eigenve tors jE i of H0 and the eigenfun tions hrjE i of h0 are related by
Z 1
h'j E i = dr h'jrihrjE i : (5.2.12)
0
176 5 S attering o the Square Barrier Potential

The generalized eigenve tors jE  i of H and the eigenfun tions hrjE i of h are related by
Z 1
h j i = lim
'+ E+!0
dr h'+ jrihrj(E + i)+ i ; '+ 2  ; (5.2.13a)
Z0 1
h jE i = lim
!0
dr h jrihrj(E i) i ; 2 + : (5.2.13b)
0
The wave fun tions '+ are usually alled in-states, whereas the the wave fun tions are
alled out-states. However, we shall all the observables (or out-observables), be ause
they are determined by the registration apparatus. In order to grasp the meaning of this ter-
minology, let us onsider the matrix element ( ; '+). This s alar produ t is the amplitude
of the probability to observe the out-state in the in-state '+ . Sin e is determined by
the property we want to measure, it stands to reason that we all it observable and denote
it by a spe i symbol.
The a tion of the Lippmann-S hwinger kets is dened as the limits in (5.2.13). We are
now going to elaborate on that denition.
The dieren e between the in-states and the observables is re e ted not only in the
notation, but also in the fa t that they belong to dierent subspa es of the Hilbert spa e.
The reason for this is the following:2 the boundary onditions built into the Lippmann-
S hwinger equation for the in-ket jE +i (or, equivalently, for the eigenfun tion hrjE +i) are
dierent to the boundary onditions for the ket jE i (or, equivalently, for the eigenfun tion
hrjE i). Sin e the boundary onditions determine the spa e of test fun tions on whi h the
kets a t, the in-ket jE +i a ts on a spa e  , and the out-ket jE i a ts on a spa e + ,
whi h is dierent to  .
The dieren e in the boundary onditions for the in- an out-kets is built into the i of
Eq. (5.2.1). The meaning of the i is that we are approa hing the ut (i.e., the spe trum
of H ) either from above (+i) or from below ( i). Therefore, the a tion of the Lippmann-
S hwinger kets jE  i should be viewed as the limit of the a tion of ertain kets j(E  i) i
that have meaning when  6= 0,
j(E  i)i !!
0
jE  i : (5.2.14)
If we want this limit pro ess to be well-dened, the following integrals should be well-
dened: Z 1
h'+j(E + i)+ i = dr h'+jrihrj(E + i)+ i ; '+ 2  ; (5.2.15a)
Z0 1
h j(E i) i = dr h jrihrj(E i) i ; 2 + : (5.2.15b)
0
Thus the a tion of the Lippmann-S hwinger kets, that has to be viewed as the limit of the
a tion of the kets (5.2.15) when  tends to zero, would be dened by
h'+jE +i = lim
!0
h'+j(E + i)+ i ; '+ 2  ; (5.2.16a)
h jE i = lim
!0
h j(E i) i ; 2 + : (5.2.16b)
2
For a mathemati al approa h to this question in terms of RHSs of Hardy fun tions see Ref. [31℄.
5.2 Lippmann-S hwinger Equation 177

In addition, we would like our spa es  to be stable under the a tion of the Hamiltonian.
This is why we should also impose the following onditions:
Z 1
(E + i)n h j(E
'+ + i)+ i= dr hH n'+ jrihrj(E + i)+ i ; '+ 2  ; (5.2.17a)
Z0 1
(E i)n h j(E i) i = dr hH n jrihrj(E i) i ; 2 + ; (5.2.17b)
0
for every n = 0; 1; 2; : : : In the position representation, the onditions (5.2.17) should lead
to restri tions in the asymptoti behavior of the wave fun tions '+ (r) and (r). In the
energy representation, the onditions (5.2.17) should lead to analyti al properties of the
wave fun tions 'b+ (E ) and b (E ).
The pre ise onne tion between the onditions satised by '+ (r) ( (r)) and those
satised by 'b+ (E ) ( b (E )) is still an open problem. Our guess is that 'b+ (E ) will be the
limit value of a Hardy lass fun tion from below, and that b (E ) will be the limit value of a
Hardy lass fun tion from above ( f. Appendix 5.5.2 for the denition and general properties
of Hardy lass fun tions).

5.2.3 Solution of the Radial Lippmann-S hwinger Equation

Instead of solving the integral equations (5.2.8), we solve the equivalent dierential equations
 
~2 d2
hhrjE  i = + V (r ) hr j E  i = E hr j E  i (5.2.18)
2m dr2
subje t to the boundary onditions that are built into the integral equations (5.2.8). These
boundary onditions are
h0j E  i = 0 ; (5.2.19a)
hrjE i is ontinuous at r = a and at r = b ; (5.2.19b)
d
dr
hrjE i is ontinuous at r = a and at r = b ; (5.2.19 )
lim hrjE i = hrjE i ; (5.2.19d)
V0 !0
hrjE +i  e ikr S (E )eikr as r ! 1 ; (5.2.19e)
hrjE i  eikr S  (E )e ikr as r ! 1 ; (5.2.19f)
where r
2m
k= E (5.2.20)
~2
and S (E ) is the S -matrix in the energy representation.
It is well-known ( f. [72, 73℄) that the in- and out-eigenfun tions are given by
(r; E )
 (r; E ) =
J (E ) ; (5.2.21)
178 5 S attering o the Square Barrier Potential

where (r; E ) is the eigenfun tion (4.2.28),

8 q
>
> sin( 2m Er ) 0<r<a
>
< ~2q q
(r; E ) = i 2~m2 (E V0 )r i 2~m2 (E V0 )r
>
>
J1 ( E ) e q
+ (J
2 qE ) e a<r<b (5.2.22)
> i 2~m2 Er i 2~m2 Er
: J3 (E )e + J4 (E )e b < r < 1;
and J(E ) are the Jost fun tions,
J+(E ) = 2iJ4(E ) ; (5.2.23a)
J (E ) = 2iJ3(E ) : (5.2.23b)
In terms of the Jost fun tions, the S -matrix is given by

S (E ) =
J (E ) : (5.2.24)
J+ (E )
From Eq. (5.2.21) it follows that the in- and out-eigenfun tions are proportional to ea h
other,
+ (r; E ) = S (E ) (r; E ) : (5.2.25)
It is worthwhile noting that the boundary ondition that singles out the in- and out-
Lippmann-S hwinger eigenfun tions is their asymptoti behavior at innity as fun tions
of r (see Eqs. (5.2.19e) and (5.2.19f)). That asymptoti behavior is transfered into the
energy representation as a ertain analyti al property of the eigenfun tion  (r; E ) as a
fun tion of E . Sin e the boundary onditions satised by the elements of the spa e of
test fun tions are related to the boundary onditions satised by the eigenfun tions of the
S hrodinger equation, it seems reasonable to expe t that the boundary onditions that
single out the '+ (r) and the (r) are related to their asymptoti behavior, and that
this asymptoti behavior is transfered into the energy representation as a ondition on the
analyti al properties of the wave fun tions 'b+ (E ) and b (E ).

5.2.4 Dire t Integral De omposition Asso iated to the In-States

On e we have obtained the in-Lippmann-S hwinger eigenfun tion, we an onstru t the
dire t integral de omposition indu ed by it. In order to do so, we follow the pro edure of
Se tion 4.2.4.
The domain D(H ) on whi h the formal dierential operator h is self-adjoint was obtained
in Se tion 4.2.2,
D(H ) = ff (r) j f (r); hf (r) 2 L2 ([0; 1); dr); f (r) 2 AC 2[0; 1); f (0) = 0g : (5.2.26)
As we saw in Se tion 4.2.2, this domain indu es a self-adjoint operator H ,
 
~2d2
(Hf )(r) := hf (r) = + V (r ) f (r ) ; f (r ) 2 D (H ) : (5.2.27)
2m dr2
5.2 Lippmann-S hwinger Equation 179

The spe trum of H is [0; 1) ( f. Se tion 4.2.4).

It is worthwhile noting that the spa e  asso iated to the in-Lippmann-S hwinger
eigenfun tion + (r; E ) will be a subspa e of D(H ). In fa t, it will be a subspa e of the
maximal invariant subspa e of H . However, the spa e  will be dierent to the spa e 
of Chapter 4, be ause the boundary onditions fullled by the elements of  are dierent
to the boundary onditions fullled by the elements of .
The Green fun tion of H was omputed in Se tion 4.2.3 for dierent regions of the
omplex plane. Sin e we want to expand the wave fun tions in terms of the eigenfun tion
+ (r; E ), we write the Green fun tion in terms of + (r; E ). From Eqs.(4.2.27) and (5.2.21),
we an see that
8
> p22m=~2 + (r ; E )  (s; E ) r < s
< m=~2 E +
G(r; s; E ) = ~2 + (s; E )  (r ; E ) r > s
<(E ) > 0; =(E ) > 0 : (5.2.28)
>
: p22m=
m=~2 E +

From Eqs. (4.2.30) and (5.2.21), we an see that

8
> p22m= ~2  (r ; E )  (s; E ) r < s
< m=~2 E
G(r; s; E ) = ~2  (s; E )  (r ; E ) r > s
<(E ) > 0; =(E ) < 0 : (5.2.29)
>
: p22m=
m=~2 E

We are now in a position to ompute the generalized Fourier transform U+ indu ed by

the Lippmann-S hwinger eigenfun tion + (r; E ). In order to be able to apply Theorem 4
of Se tion 4.4.3, we hoose the following basis for the spa e of solutions of h = E that is
ontinuous on (0; 1)   and analyti ally dependent on E :
1 (r; E ) = + (r; E ) ; (5.2.30a)
8 q
>
> os( 2m Er ) 0<r<a
>
< 2
~q q
2m 2m
2 (r; E ) = C1 (E )eiq ~2 (E V0 )r + C2 (Eq)e i ~2 (E V0 )r
a < r < b (5.2.30b)
>
>
> 2m 2m
: C3 (E )ei ~2 Er + C4 (E )e i ~2 Er b < r < 1:
The fun tions C1 C4 are given by Eq. (4.4.30) of Appendix 4.4.3.
Eqs. (4.2.29), (4.2.31) and (5.2.30) lead to
2iJ4 (E )C4 (E ) J (E )
 + (r ; E ) = 1 (r; E ) + 4 2 (r; E ) (5.2.31)
W (E ) W (E )
and to
2iJ4 (E )C3 (E ) J3(E )  (r; E ) ;
 (r ; E ) = 1 (r; E ) 2 (5.2.32)
W (E ) W (E )
where
W (E ) = J4 (E )C3 (E ) J3(E )C4 (E ) : (5.2.33)
180 5 S attering o the Square Barrier Potential

By substituting Eq. (5.2.31) into Eq. (5.2.28) we get to

 
2m=~2 2iJ4 (E )C4 (E ) J (E )
G(r; s; E ) = p 1 (r; E ) + 4 2 (r; E ) 1 (s; E ) ;
2m=~ E
2 W (E ) W (E )
<(E ) > 0; =(E ) > 0 ; r > s : (5.2.34)
By substituting Eq. (5.2.32) into Eq. (5.2.29) we get to

G(r; s; E ) = p
2m=~2 J4(E )  2iJ4(E )C3 (E )  (r; E ) J3(E )  (r; E )  (s; E ) ;
2m=~2 E J3 (E ) W (E ) 1
W (E ) 2 1

<(E ) > 0; =(E ) < 0 ; r > s ; (5.2.35)

where we have used the fa t that

 (r ; E ) =
J4(E ) +(r; E ) : (5.2.36)
J3(E )
Sin e
1 (s; E ) =
J4(E )  (s; E ) ; (5.2.37)
J3(E ) 1
Eq. (5.2.34) leads to
2m=~2 1 h i
G(r; s; E ) = p 2 i J3 ( E ) C4 ( E ) 1 ( r ; E )1 ( s; E ) + J3 ( E ) 2 ( r ; E )1 ( s; E )
2m=~2 E W (E )
<(E ) > 0; =(E ) > 0 ; r > s ; (5.2.38)
and Eq. (5.2.35) leads to
2m=~2 1 h i
G(r; s; E ) = p 2iJ4 (E )C3 (E )1 (r; E )1 (s; E ) + J3 (E )2 (r; E )1 (s; E )
2m=~2 E W (E )
<(E ) > 0; =(E ) < 0 ; r > s : (5.2.39)
The expression of the resolvent in terms of the basis 1 ; 2 an be written as (see Theorem 4
in Appendix 4.4.3)
2
X
G(r; s; E ) = ij+ (E )i (r; E )j (s; E ) ; r > s: (5.2.40)
i;j =1

By omparing (5.2.40) to (5.2.38) we get to

0 ~2 2iJ3 (E )C4 (E ) 1
p22m=
m=~2 E W (E ) 0
ij+ (E ) =  2 m=~ 2 J3 (E )
p2m=~2 E W (E)
A ; <(E ) > 0 ; =(E ) > 0 : (5.2.41)
0
5.2 Lippmann-S hwinger Equation 181

By omparing (5.2.40) to (5.2.39) we get to

0 ~2 2iJ4 (E )C3 (E ) 1
p22m=
m=~2 E W (E ) 0
ij+ (E ) =  2 m=~ 2 J3 (E )
p2m=~2 E W (E)
A ; <(E ) > 0 ; =(E ) < 0 : (5.2.42)
0
From Eqs. (5.2.41) and (5.2.42) we an see that the measures 12 , 21 and 22 in Theorem 4
of Appendix 4.4.3 are zero and that the measure 11 is given by
Z
1 E2 Æ  + + (E + i) dE
11 ((E1 ; E2 )) = lim lim  11 ( E i)  11
Æ!0 !0+ 2i E1 +Æ
Z E2
1 2m=~2
= p dE ; (5.2.43)
E1  2m=~2 E
whi h leads to
1 2m=~2
+ (E )  11 (E ) = p ; E 2 (0; 1) : (5.2.44)
 2m=~2 E
By Theorem 2 of Appendix 4.4.3, there is a unitary map Ue+ dened by
Ue+ : L2 ([0; 1); dr) 7 ! L2 ((0; 1); +(E )dE )Z
1
f (r) 7 ! fe(E ) = (Ue+ f )(E ) = drf (r)+(r; E ) ; (5.2.45)
0
that brings D(H ) onto the spa e
Z 1
D (Ee) =f fe(E ) 2 L2 ((0; 1 ); +(E )dE ) j dE E 2 jfe(E )j2 + (E ) < 1g : (5.2.46)
0
Eqs. (5.2.45) and (5.2.46) provide a + -diagonalization of H . If we seek a Æ -diagonalization,
then the measure + (E ) must be absorbed by the eigenfun tions
p
 + (r; E ) := + (E ) + (r; E ) ; (5.2.47)
and by the wave fun tions
p
fb(E ) := + (E )fe(E ) ; fe(E ) 2 L2 ((0; 1); +(E )dE ) : (5.2.48)
The fun tion  + (r; E ) is the eigensolution of the Lippmann-S hwinger equation (5.2.8) that
is Æ -normalized. Using Eq. (5.2.48) we an onstru t the unitary operator
Ub+ : L2 ((0; 1)); +(E )dE ) 7 ! L2((0; 1); dE ) p
fe 7 ! fb(E ) = (Ub+fe)(E ) := + (E )fe(E ) : (5.2.49)
The operator that Æ -diagonalizes our Hamiltonian is U+ := Ub+ Ue+ ,
U+ : L2 ([0; 1)); dr) 7 ! L2 ((0; 1); dE )
f 7 ! U+ f := fb: (5.2.50)
182 5 S attering o the Square Barrier Potential

The a tion of U+ an be written as an integral operator,

Z 1
fb(E ) = (U+ f )(E ) = drf (r) +(r; E ) ; f (r) 2 L2 ([0; 1); dr) : (5.2.51)
0

The image of D(H ) under the a tion of U+ is

Z 1
D(Eb) := U+D(H ) = ffb(E ) 2 L2 ((0; 1); dE ) j E 2 jfb(E )j2 dE < 1g : (5.2.52)
0

Therefore, we have onstru ted a unitary operator

U+ : D(H )  L2 ([0; 1); dr) 7 ! D(Eb)  L2 ((0; 1); dE )
f 7 ! fb = U+ f (5.2.53)
that transforms from the position representation into the energy representation. The oper-
ator U+ diagonalizes H , i.e., Eb  U+ HU+ 1 is the multipli ation operator,

Eb : D(Eb )  L2 ((0; 1); dE ) 7 ! L2 ((0; 1); dE )

fb 7 ! Ebfb(E ) := E fb(E ) : (5.2.54)
The inverse operator of U+ is given by (see Theorem 3 of Appendix 4.4.3)
Z 1
f (r) = U+ 1 fb(r ) = dE fb(E ) + (r; E ) ; fb(E ) 2 L2 ((0; 1); dE ) : (5.2.55)
0

The operator U+ 1 transforms from the energy representation into the position representa-
tion.
The expressions (5.2.51) and (5.2.55) provide the eigenfun tion expansion of any wave
fun tion in terms of the Æ -normalized eigensolutions  + (r; E ) of the Lippmann-S hwinger
equation.
The unitary operator U+ an be also looked at as a sort of generalized Fourier transform
that onne ts the position and the energy representations. The eigenfun tions  + (r; E ) an
be viewed as \transition elements" between the r- and the E -representations.
Thus we have onstru ted the dire t integral de omposition of the Hilbert spa e asso i-
ated to the eigenfun tion + (r; E ) of the Lippmann-S hwinger equation,
Z
H 7 ! U+ H  H =  b H(E )dE
Sp(H )
f 7 ! U+ f  ffb(E )g; f 2 D(H ) ; fb(E ) 2 H(E ) : (5.2.56)

In Eq. (5.2.56), the Hilbert spa es H, Hb and H(E ) are the same as in Eq. (4.2.64).
5.2 Lippmann-S hwinger Equation 183

5.2.5 Dire t Integral De omposition Asso iated to the Observ-

ables
In this se tion, we ompute the unitary operator U indu ed by the Lippmann-S hwinger
eigenfun tion  (r; E ). Sin e the omputations are very similar to the omputations per-
formed in Se tion 5.2.4 for the in-states, we will restri t the explanations to the minimum
and just write down the results.
The fun tions
1 (r; E ) =  (r; E ) ; (5.2.57a)
8 q
>
> os( 2m Er ) 0<r<a
>
< 2
~q q
2 (r; E ) = i 2~m2 (E V0 )r i 2~m2 (E V0 )r
> C
1 (E )e q C
+ 2 (Eq)e a < r < b (5.2.57b)
>
: (E )ei 2~m2 Er + (E )e i 2~m2 Er
>
C3 C4 b < r < 1:
form a basis for the spa e of solutions of h = E that is ontinuous on (0; 1)   and ana-
lyti ally dependent on E . Therefore, we are allowed to apply Theorem 4 of Appendix 4.4.3.
Eqs. (4.2.29), (4.2.31) and (5.2.57) lead to
2iJ3 (E )C4 (E ) J (E )
+(r; E ) = 1 (r; E ) + 4 2 (r; E ) (5.2.58)
W (E ) W (E )
and to
2iJ3 (E )C3 (E ) J3 (E )  (r ; E ) ;
 (r ; E ) = 1 (r; E ) 2 (5.2.59)
W (E ) W (E )
where
W (E ) = J4 (E )C3 (E ) J3 (E )C4 (E ) : (5.2.60)
By substituting Eq. (5.2.58) into Eq. (5.2.28) we get to

G(r; s; E ) = p
2m=~2 J3(E )  2iJ3(E )C4 (E )  (r; E ) + J4(E )  (r; E )  (s; E ) ;
2m=~2 E J4 (E ) W (E ) 1
W (E ) 2 1

<(E ) > 0; =(E ) > 0 ; r > s : (5.2.61)

By substituting Eq. (5.2.59) into Eq. (5.2.29) we get to
 
2m=~2 2iJ3 (E )C3 (E ) J3(E )  (r; E )  (s; E ) ;
G(r; s; E ) = p  1 ( r ; E )
2m=~2 E W (E ) W (E ) 2 1

<(E ) > 0; =(E ) < 0 ; r > s : (5.2.62)

Sin e
1 (s; E ) =
J3 (E )  (s; E ) ; (5.2.63)
J4 (E ) 1
184 5 S attering o the Square Barrier Potential

Eq. (5.2.61) leads to

2m=~2 1 h i
G(r; s; E ) = p 2 i J3 ( E ) C4 ( E ) 1 ( r ; E )1 ( s; E ) J4 ( E ) 2 ( r ; E )1 ( s; E )
2m=~2 E W (E )
<(E ) > 0; =(E ) > 0 ; r > s ; (5.2.64)
and Eq. (5.2.62) leads to
2m=~2 1 h i
G(r; s; E ) = p 2 i J4 ( E ) C3 ( E ) 1 ( r ; E )1 ( s; E ) J4 ( E ) 2 ( r ; E )1 ( s; E )
2m=~2 E W (E )
<(E ) > 0; =(E ) < 0 ; r > s : (5.2.65)
The expression of the resolvent in terms of the basis 1 ; 2 an be written as (see Theorem 4
in Appendix 4.4.3)
2
X
G(r; s; E ) = ij+ (E )i (r; E )j (s; E ) ; r > s: (5.2.66)
i;j =1

By omparing (5.2.66) to (5.2.64) we get to

0 ~2 2iJ3 (E )C4 (E ) 1
p22m=
m=~2 E W (E ) 0
ij+ (E ) =  2 m=~ 2 J4 (E )
p2m=~2 E W (E)
A ; <(E ) > 0 ; =(E ) > 0 : (5.2.67)
0

By omparing (5.2.66) to (5.2.65) we get to

0 ~2 2iJ4 (E )C3 (E ) 1
p22m=
m=~2 E W (E ) 0
ij+ (E ) = 
p22m= 2
~ J4 (E )
A ; <(E ) > 0 ; =(E ) < 0 : (5.2.68)
m=~2 E W (E )
0

From Eqs. (5.2.67) and (5.2.68) we an see that the measures 12 , 21 and 22 in Theorem 4
of Appendix 4.4.3 are zero and that the measure 11 is given by
Z
1 E2 Æ  + 
+ (E + i) dE
11 ((E1 ; E2 )) = Ælim lim  (E
!0 !0+ 2i E1 +Æ 11
i) 11
Z E2
1 2m=~2
= p dE ; (5.2.69)
E1  2m=~2 E
whi h leads to
1 2m=~2
 (E )  11 (E ) = p ; E 2 (0; 1) : (5.2.70)
 2m=~2 E
In order to Æ -normalize, we dene
p
 (r; E ) :=  (E )  (r ; E ) ; (5.2.71)
5.3 Constru tion of the Lippmann-S hwinger Kets and Dira Basis Ve tor Expansion 185

whi h is the eigensolution of the Lippmann-S hwinger equation (5.2.8) that is Æ -normalized,
and p
fb(E ) :=  (E )fe(E ) ; fe(E ) 2 L2 ((0; 1); (E )dE ) : (5.2.72)
The unitary operator that Æ -diagonalizes the Hamiltonian is,
Z 1
fb(E ) = (U f )(E ) = drf (r) (r; E ) ; f (r) 2 L2 ([0; 1); dr) : (5.2.73)
0
The inverse operator of U is given by (see Theorem 3 of Appendix 4.4.3)
Z 1
f (r ) = (U 1 fb)(r ) = dE fb(E ) (r; E ) ; fb(E ) 2 L2 ((0; 1); dE ) : (5.2.74)
0
Therefore, we have onstru ted a unitary operator
U : D(H )  L2 ([0; 1); dr) 7 ! D(Eb)  L2 ((0; 1); dE )
f 7 ! fb = U f (5.2.75)
that transforms from the position representation into the energy representation. The oper-
ator U diagonalizes H . The operator U 1 transforms from the energy representation into
the position representation.
The expressions (5.2.73) and (5.2.74) provide the eigenfun tion expansion of any wave
fun tion in terms of the Æ -normalized eigensolutions  (r; E ) and the dire t integral de om-
position of H asso iated to the observables.

5.3 Constru tion of the Lippmann-S hwinger Kets and

Dira Basis Ve tor Expansion
In this se tion, we sket h the onstru tion of the spa es  on whi h the Lippmann-
S hwinger kets a t as antilinear fun tionals.
In order to asso iate a ket jE + i to the eigenfun tion hrjE +i, we dene
jE + i :  7 ! C Z 1
' 7 ! h' jE i := lim
+ + +
!0
dr h'+ jrihrj(E + i)+ i : (5.3.1)
0
As mentioned above, the a tion of the Lippmann-S hwinger ket jE + i should be viewed as
the limit of the a tion of some ket j(E + i)+ i when  tends to zero. This is why we dene
the a tion of jE + i on '+ as the limit of the integral in Eq. (5.3.1), and not just as
Z 1
dr h'+ jrihrjE +i : (5.3.2)
0
The fun tions '+ 2  on whi h the a tion of the in-ket of Eq. (5.3.1) is well dened satisfy
(at least) the following onditions:
186 5 S attering o the Square Barrier Potential

i.) They belong to the maximal invariant subspa e of H .

ii.) They are su h that the limit in Eq. (5.3.1) makes sense.
iii.) They satisfy (5.2.17a) for every n = 0; 1; 2; : : :
By means of the unitary operator U+ , whi h was onstru ted in Se tion 5.2.4, we an
obtain the energy representation of the spa e  ,
U+  = 
b jR+ : (5.3.3)
We have denoted the energy representation of the spa e  by  b jR+ (rather than by  b ),
be ause its elements are boundary values on the positive real line of fun tions that have
meaning for omplex energies. As we said above, those fun tions 'b+ (E ) seem to be realized
by analyti fun tions from below. For the sake of deniteness, we shall assume that
b jR+ = S \ H2 jR+ ; (5.3.4)
where S is the S hwartz spa e and H2 is the spa e of Hardy fun tions from below ( f. Ap-
pendix 5.5.2).
Under the assumptions made on the elements of the spa e  , one an prove that
jE i is a well dened antilinear fun tional and that jE +i is a generalized eigenve tor of the
+
Hamiltonian H (the proof is almost identi al to the proof of Proposition 3 of Appendix 4.4.4),
H  jE + i = E jE + i ; (5.3.5)
i.e.,
h'+jH jE +i = hH'+jE +i = E h'+jE +i ; 8'+ 2  : (5.3.6)
The in-ket jE + i an be used to expand the in-states '+ 2  in a Dira basis ve tor
expansion. This expansion is the restri tion of the eigenfun tion expansions (5.2.51) and
(5.2.55) to the spa e  ,
Z 1
h j i=
+E '+ dr h+ E jrihrj'+i ; '+ 2  ; (5.3.7a)
Z 1 0
hr j i =
'+ dE hrjE +ih+ E j'+ i ; '+ 2  : (5.3.7b)
0

Similarly, the out-ket jE

i asso iated to the eigenfun tion hrjE i is dened by
jE i : + 7 ! C Z 1
7 ! h jE i := lim !0
dr h jrihrj(E i) i : (5.3.8)
0

The fun tions 2 + satisfy (at least) the following onditions:

i.) They belong to the maximal invariant subspa e of H .
5.3 Constru tion of the Lippmann-S hwinger Kets and Dira Basis Ve tor Expansion 187

ii.) They are su h that the limit in Eq. (5.3.8) makes sense.
iii.) They satisfy (5.2.17b) for every n = 0; 1; 2; : : :
By means of the unitary operator U , whi h was onstru ted in Se tion 5.2.5, we an
obtain the energy representation of the spa e + ,
U + = b +jR+ : (5.3.9)
For the sake of deniteness, we shall assume that
b +jR+ = S \ H+2 jR+ ; (5.3.10)
where H+2 is the spa e of Hardy fun tions from above ( f. Appendix 5.5.2).
The out-ket is also a generalized eigenve tor of H ,
h j H  j E i = hH j E i = E h j E i ; 8 2 + : (5.3.11)
The Dira basis ve tor expansion indu ed by the out-ket reads
Z 1
h Ej i = dr h E jrihrj i; 2 + ; (5.3.12a)
Z 1 0
hr j i = dE hrjE ih E j i ; 2 + : (5.3.12b)
0

The Mller operators

 an be expressed in terms of the operators U ( f. Se tions 5.2.4
and 5.2.5) and U0 ( f. Appendix 5.5.1) as ( f. [75℄)

 = Uy U0 : (5.3.13)
Obviously,
 are unitary operators in the Hilbert spa e L2 ([0; 1); dr). The Mller opera-
tors an be used to onstru t the spa e in of asymptoti in-states 'in and the spa e out
of asymptoti out-observables out ,
 in =
y  : (5.3.14)
out

A ve tor 'in belongs to in if

h+E j'+i = hE j'ini ; (5.3.15)
where '+ =
+ 'in . A ve tor out belongs to out if
h E j i = hE j out i; (5.3.16)
where =
out . From the last two equations it follows that

 jE i = jE  i : (5.3.17)
188 5 S attering o the Square Barrier Potential

The following diagram summarizes the results on erning the states:

H0 ; 'in (r) in  L2 ([0; 1); dr)  in jE i position repr:
#
+ #
+ #
+
H ; '+ ( r )   L2 ([0; 1); dr)   jE + i position repr:
# U+ # U+ # U+
Eb ; 'b+ (E ) b jR+  L2([0; 1); dE )  (b jR+ ) jEb+i
energy repr:
(5.3.18)
The results on erning the observables are summarized by the following diagram:
H0 ; out (r) out  L2 ([0; 1); dr)  out jE i position repr:
#
#
#

H; (r ) +  L2 ([0; 1); dr)  + jE i position repr:
#U #U # U
Eb; b (E ) b +jR+  L2 ([0; 1); dE )  (b +jR+ ) jEb i energy repr:
(5.3.19)

5.4 S-matrix and Mller Operators

Our next goal is to onstru t the S -matrix and to ompute the expression of the S -matrix
element in terms of the Lippmann-S hwinger kets (see Eq. (5.4.11) below).
As we said in Se tion 5.1, the S -matrix element
( out ; 'out ) = ( out ; S'in ) = ( ; '+ ) (5.4.1)
is to represent the probability to dete t the property des ribed by out (or, equivalently,
by ) in the prepared ingoing beam hara terized by 'in (or, equivalently, by '+ ). The
expression of the S -matrix operator in terms of the Mller operators read
S =
y
+ : (5.4.2)
The operator S is a unitary operator in the Hilbert spa e L2 ([0; 1); dr). In the energy
representation, the operator (5.4.2) a ts as the operator multipli ation by the fun tion
S (E ) = J (E )=J+(E ). To be more pre ise, if we dene the operator Sb as
Sb : L2 ([0; 1); dE ) 7 ! L2 ([0; 1); dE )
fb 7 ! (Sbfb)(E ) = S (E ) fb(E ) ; (5.4.3)
then it an proved that
Sb = U0 SU0 1 : (5.4.4)
5.4 S-matrix and Mller Operators 189

In order to prove Eq. (5.4.4), we rst prove that

(U g)(E ) = S (E )(U+ g)(E ) ; 8g 2 L2 ([0; 1); dr) : (5.4.5)
Sin e
 + (r ; E ) = S (E ) (r ; E ) ; (5.4.6)
and sin e 1 ; E > 0;
S (E ) = (5.4.7)
S (E )
we on lude that
 (r ; E ) = S (E ) + (r ; E ) : (5.4.8)
By substituting Eq. (5.4.8) into the integral expression (5.2.73) of the operator U we
get to Z
1
(U g)(E ) = dr g(r)S (E )+ (r; E ) : (5.4.9)
0
Comparison of (5.4.9) to the integral expression (5.2.51) of U+ leads to (5.4.5).
Now,
b
(U0 SU0 1)f = (U0
y
+U0 1)f b
= (U0 U0 1 U U+y U0U0 1 )f b
= (U U+y )f b
= Sf ; bb (5.4.10)
b
where we have applied Eq. (5.4.5) to g = U+y f in the next to the last step.
As we said in Se tion 5.1, the S -matrix element an be written in terms of the Lippmann-
S hwinger eigenkets as
Z 1
( ; '+ ) = dE h jE iS (E )h+E j'+i : (5.4.11)
0

The proof of Eq. (5.4.11) is as follows: Let 2 + and '+ 2  . Sin e and
'+ belong, in parti ular, to the Hilbert spa e L2 ([0; 1); dr ), we an let the unitary
operator U a t on both of them,
( ; '+ ) = (U ; U '+ ) : (5.4.12)
The ve tors U and U '+ belong to L2 ([0; 1); dE ). Therefore,
(U ; U '+ )=
Z 1
dE U ( )(E )(U )( )
'+ E : (5.4.13)
0
From Eq. (5.4.5) it follows that
(U '+)(E ) = S (E )(U+ '+ )(E ) : (5.4.14)
190 5 S attering o the Square Barrier Potential

Thus,
(U ; U '+ )=
Z 1
dE U ( )(E )S (E )(U+ '+)(E ) : (5.4.15)
0
Sin e 2 +, we are allowed to write
(U )(E ) = h jE i : (5.4.16)
Sin e '+ 2  , we are allowed to write
(U+'+ )(E ) = h+ E j'+ i : (5.4.17)
Substitution of (5.4.16) and (5.4.17) into (5.4.15) leads to (5.4.11).
A similar argument to that used to prove Eq. (5.4.11) an be used to prove that
Z 1
( out ; S'in ) = dE h out jE iS (E )hE j'ini : (5.4.18)
0

Many formal identities follow from Eqs. (5.4.11) and (5.4.18). For instan e,
hE j S j E 0 i = h E j E 0 + i = S ( E ) Æ ( E E 0 ) ; (5.4.19)
Z 1 Z 1
dr hE jrihrjE 0i = dr h E jrihrjE 0 i = Æ (E E0) ; (5.4.20)
0 0
and Z 1
dr h E jrihrjE 0+ i = S (E )Æ (E E 0) : (5.4.21)
0
The \proof" of these identities follows the pattern of Se tion 4.2.9.

5.5 Appendi es to Chapter 5

5.5.1 Appendix 7: Free Hamiltonian
In this appendix, we ompute the RHS asso iated to the free Hamiltonian. We will follow
the method used in Chapter 4 for the total Hamiltonian.

Self-Adjoint Extension
The rst step is to dene a linear operator on a Hilbert spa e orresponding to the formal
dierential operator
~2 d2
h0  : (5.5.1)
2m dr2
The Hilbert spa e that is in the RHS of the free Hamiltonian is realized by the spa e
L2 ([0; 1); dr) of square integrable fun tions f (r) dened on the interval [0; 1). The same
5.5 Appendi es to Chapter 5 191

pro edure that was used to nd the domain of the total Hamiltonian an be applied to the
free Hamiltonian,
D(H0) = ff (r) j f (r); h0f (r) 2 L2 ([0; 1); dr); f (r) 2 AC 2 [0; 1); f (0) = 0g : (5.5.2)
On D(H0 ) the formal dierential operator h0 is self-adjoint. In hoosing (5.5.2) as the
domain of our formal dierential operator h0 , we dene a linear operator H0 by
~2 d2
H0 f (r) := h0 f (r) = f (r) ; f (r) 2 D(H0 ) : (5.5.3)
2m dr2
Resolvent and Green Fun tions
The expression of the free Green fun tion G0 (r; s; E ) is be given in terms of eigenfun tions
of the dierential operator h0 subje t to ertain boundary onditions ( f. Theorem 1 in
Se tion 4.4.2).
Region <(E ) < 0, =(E ) 6= 0
For <(E ) < 0, =(E ) =
6 0, the free Green fun tion (see Theorem 1 in Se tion 4.4.2) is
given by
8
>
< p 22m=m=~~2 2 E e0(r;E)2e 0(s;E) r<s
G0 (r; s; E ) = <(E ) < 0 ; =(E ) 6= 0 : (5.5.4)
>
: p 22m=m=~~2 2 E e0(s;E)2e 0(r;E) r>s

The eigenfun tion e0 (r; E ) satises the equation

h0 e0 (r; E ) = E e0 (r; E ) (5.5.5)
and the boundary onditions (4.2.21),
q q
2m2 Er 2m2 Er
e0 (r; E ) =e ~ e ~ ; 0 < r < 1: (5.5.6)
e 0 (r ; E ) satises the equation (5.5.5) and the boundary onditions (4.2.24),
The eigenfun tion 
q
2m Er
e 0 (r ; E ) = e
 ~2 ; 0 < r < 1: (5.5.7)

Region <(E ) > 0, =(E ) > 0

When <(E ) > 0, =(E ) > 0, the expression of the free Green fun tion is
8
> p22m=~2 0 (r ; E ) 0 (s; E ) r < s
< m=~2 E +
G0 (r; s; E ) = 2 <(E ) > 0; =(E ) > 0 : (5.5.8)
>
: p22m= ~ 0 (s; E ) 0 (r ; E ) r > s
m=~2 E +
192 5 S attering o the Square Barrier Potential

The eigenfun tion 0 (r; E ) satises the S hrodinger equation (5.5.5) and the boundary on-
ditions (4.2.21), r
2m
0 (r; E ) = sin( 2 Er) ; 0 < r < 1 : (5.5.9)
~
The eigenfun tion 0+ (r; E ) satises the equation (5.5.5) subje t to the boundary onditions
(4.2.24), q
2m
0+ (r; E ) = ei ~2 Er ; 0 < r < 1 : (5.5.10)

Region <(E ) > 0, =(E ) < 0

In the region <(E ) > 0, =(E ) < 0, the free Green fun tion reads
8
> p22m=~2 0 (r ; E ) 0 (s; E ) r < s
< m=~2 E
G0 (r; s; E ) = ~2 0 (s; E ) 0 (r ; E ) r > s
<(E ) > 0; =(E ) < 0 : (5.5.11)
>
: p22m=
m=~2 E

The eigenfun tion 0 (r; E ) is given by (5.5.9). The eigenfun tion 0 (r; E ) satises the
equation (5.5.5) and the boundary onditions (4.2.24),
q
i 2~m2 Er
0 (r; E ) =e ; 0 < r < 1: (5.5.12)

Spe trum of H0
We ompute the spe trum Sp(H0 ) of the operator H0 by applying the method used in
Se tion 4.2.4 to ompute the spe trum of H .
Subset =( 1; 0)
We rst take  from Theorem 4 of Se tion 4.4.3 to be ( 1; 0). We hoose a basis for
the spa e of solutions of the equation h0  = E as
q
2m Er
e1 (r; E ) = e ~2 ; (5.5.13a)
e 0 (r ; E ) :
e2 (r; E ) =  (5.5.13b)
Obviously,
e0 (r; E ) = e1 (r; E ) e2 (r; E ) ; (5.5.14)
whi h along with Eq. (5.5.4) leads to
2m=~2 1
G0 (r; s; E ) = p [e1 (r; E ) e2 (r; E )℄ e2 (s; E ) ; r < s ; <(E ) < 0 ; =(E ) 6= 0 :
2m=~2 E 2
(5.5.15)
5.5 Appendi es to Chapter 5 193

Sin e
e2 (s; E ) = e2 (s; E ) ; (5.5.16)
we an write Eq. (5.5.15) as
2m=~2 1 h i
G0 (r; s; E ) = p e1 (r ; E )
 e2 (s; E ) 
e2 (r ; E )
e2 (s; E ) ;
2m=~2 E 2
r < s ; <(E ) < 0 ; =(E ) 6= 0 : (5.5.17)
On the other hand, by Theorem 4 in Se tion 4.4.3 we have
2
X
G0 (r; s; E ) = ij (E )ei (r; E )ej (s; E ) r < s: (5.5.18)
i;j =1

By omparing Eqs. (5.5.17) and (5.5.18) we see that

0 ~2 1 1
0 p 2m=
2m=~2 E 2
ij (E ) = 
p 2m=~2 1
A ; <(E ) < 0 ; =(E ) 6= 0 : (5.5.19)
0 2m=~2 E 2

The fun tions ij (E ) are analyti in a neighborhood of  = ( 1; 0). Therefore, the interval
( 1; 0) is in the resolvent set Re(H0 ) of the operator H0 .
Subset  = (0; 1)
In this ase, we hoose the following basis for the spa e of solutions of h0  = E :
1 (r; E ) = 0 (r; E ) ; (5.5.20a)
r
2m
2 (r; E ) = os( Er) : (5.5.20b)
~2

Eqs. (5.5.10), (5.5.12) and (5.5.20) lead to

0+ (r; E ) = i1 (r; E ) + 2 (r; E ) (5.5.21)
and to
0 (r; E ) = i1 (r; E ) + 2 (r; E ) : (5.5.22)
By substituting Eq. (5.5.21) into Eq. (5.5.8) we get to
2m=~2
G0 (r; s; E ) = p 1 (s; E ) [i1 (r; E ) + 2 (r; E )℄ ; r > s ; <(E ) > 0; =(E ) > 0 :
2m=~2 E
(5.5.23)
194 5 S attering o the Square Barrier Potential

By substituting Eq. (5.5.22) into Eq. (5.5.11) we get to

2m=~2
G0 (r; s; E ) = p 1 (s; E ) [ i1 (r; E ) + 2 (r; E )℄ ; r > s ; <(E ) > 0; =(E ) < 0 :
2m=~2 E
(5.5.24)
Sin e
1 (s; E ) = 1 (s; E ) ; (5.5.25)
Eq. (5.5.23) leads to
2m=~2 h i
G0 (r; s; E ) = p i 1 ( r ; E )  1 ( s; E ) +  2 ( r ; E ) 1 ( s; E ) ;
2m=~2 E
<(E ) > 0; =(E ) > 0 ; r > s ; (5.5.26)
and Eq. (5.5.24) leads to
2m=~2 h i
G0 (r; s; E ) = p i 1 ( r ; E ) 1 ( s; E ) +  2 ( r ; E ) 1 ( s; E )
2m=~2 E
<(E ) > 0; =(E ) < 0 ; r > s : (5.5.27)
The expression of the resolvent in terms of the basis 1 ; 2 an be written as (see Theorem 4
in Se tion 4.4.3)
2
X
G0 (r; s; E ) = ij+ (E )i (r; E )j (s; E ) ; r > s: (5.5.28)
i;j =1
By omparing (5.5.28) to (5.5.26) we get to
!
p 2m=~2 p 2m=~2
i
ij+ (E ) = 2m=~2 E 2m=~2 E ; <(E ) > 0 ; =(E ) > 0 : (5.5.29)
0 0
By omparing (5.5.28) to (5.5.27) we get to
!
p 2m=~2 p 2m=~2
i
ij+ (E ) = 2m=~2 E 2m=~2 E ; <(E ) > 0 ; =(E ) < 0 : (5.5.30)
0 0
From Eqs. (5.5.29) and (5.5.30) we an see that the measures 12 , 21 and 22 in Theorem 4
of Se tion 4.4.3 are zero and that the measure 11 is given by
Z
1 E2 Æ  + + (E + i) dE
11 ((E1 ; E2 )) = Ælim lim
!0 !0+ 2i E1 +Æ
 11 ( E i)  11
Z E2
1 2m=~2
= p dE ; (5.5.31)
E1  2m=~2 E
whi h leads to
1 2m=~2
0 (E )  11 (E ) = p ; E 2 (0; 1) : (5.5.32)
 2m=~2 E
The fun tion 11 + (E ) has a bran h ut along (0; 1), and therefore (0; 1) is in luded in
Sp(H0 ). Sin e Sp(H0 ) is a losed set, Sp(H0 ) = [0; 1).
5.5 Appendi es to Chapter 5 195

Diagonalization and Eigenfun tion Expansion

In the present se tion, we diagonalize our Hamiltonian H0 and onstru t the expansion of
the wave fun tions in terms of the eigenfun tions of the dierential operator h0 .
By Theorem 2 of Se tion 4.4.3, there is a unitary map Ue0 dened by
Ue0 : L2 ([0; 1); dr) 7 ! L2 ((0; 1); 0(E )dEZ)
1
f (r) 7 ! fe(E ) = Ue0 f (E ) = drf (r)0 (r; E ) ; (5.5.33)
0
that brings D(H0 ) onto the spa e
Z 1
D (Ee ) = ffe(E ) 2 L2 ((0; 1 ); 0(E )dE ) j dE E 2 jfe(E )j2 0 (E ) < 1g : (5.5.34)
0
In order to Æ -normalize, we dene
p
 0 (r; E ) := 0 (E ) 0 (r; E ) ; (5.5.35)
whi h is the eigensolution of the dierential operator h0 that is Æ -normalized, and
p
fb(E ) := 0 (E )fe(E ) ; fe(E ) 2 L2 ((0; 1); 0 (E )dE ) ; (5.5.36)
and onstru t the unitary operator
Ub0 : L2 ((0; 1)); 0(E )dE ) 7 ! L2((0; 1); dE ) p
fe 7 ! fb(E ) = Ub0 fe(E ) := 0 (E )fe(E ) : (5.5.37)
The operator that Æ -diagonalizes our Hamiltonian is U0 := Ub0 Ue0 ,
U0 : L2 ([0; 1)); dr) 7 ! L2 ((0; 1); dE )
f 7 ! U0f := fb: (5.5.38)
The a tion of U0 an be written as an integral operator,
Z 1
fb(E ) = U0 f (E ) = drf (r) 0(r; E ) ; f (r) 2 L2 ([0; 1); dr) : (5.5.39)
0
The image of D(H0 ) under the a tion of U0 is
Z 1
D (Eb) := U D(H0 ) = f fb(E ) 2 L2 ((0; 1); dE ) j E 2 jfb(E )j2 dE < 1g : (5.5.40)
0
Therefore, we have onstru ted a unitary operator
U0 : D(H )  L2 ([0; 1); dr) 7 ! D(Eb)  L2 ((0; 1); dE )
f 7 ! fb = U0 f (5.5.41)
196 5 S attering o the Square Barrier Potential

that transforms from the position representation into the energy representation. The opera-
tor U0 diagonalizes the free Hamiltonian in the sense that Eb  U0 H0 U0 1 is the multipli ation
operator. The inverse operator of U0 is given by (see Theorem 3 of Se tion 4.4.3)
Z 1
f (r) = U0 1 fb(r ) = dE fb(E ) 0 (r; E ) ; fb(E ) 2 L2 ((0; 1); dE ) : (5.5.42)
0
The operator U0 1 transforms from the energy representation into the position representa-
tion.
The expressions (5.5.39) and (5.5.42) provide the eigenfun tion expansion of any square
integrable fun tion in terms of the eigensolutions  0 (r; E ) of h0 . One an easily see that
lim U = U0 : (5.5.43)
V0 !0

Constru tion of the RHS of the Free Hamiltonian

The Sturm-Liouville theory only provides a domain D(H0 ) on whi h the Hamiltonian H0 is
self-adjoint and a unitary operator U0 that diagonalizes H0 . This unitary operator indu es
a dire t integral de omposition of the Hilbert spa e (see [4, 5℄),
Z
H 7 ! U0H  H =  b H(E )dE
Sp(H0 )
f 7 ! U0f  ffb(E )g; fb(E ) 2 H (E ) : (5.5.44)
As we saw in Chapter 4, the dire t integral de omposition does not provide us with a
dense invariant domain 0 on whi h all the powers of H0 and all the expe tation values of
H0 are well-dened, and on whi h the Dira kets a t as antilinear fun tionals. In order to
onstru t 0 , we rst onstru t the maximal invariant subspa e D0 of H0 ,
1
\
D0 := D(H0n) : (5.5.45)
n=0
It is easy to he k that
D0 = f' 2 L2 ([0; 1); dr) j hn0 '(r) 2 L2 ([0; 1); dr); hn0 '(0) = 0; n = 0; 1; 2; : : : ;
'(r) 2 C 1 ([0; 1))g : (5.5.46)
The se ond step is to nd a subspa e 0 on whi h the eigenkets jE i of H0 are well-dened
as antilinear fun tionals. That subspa e is given by
Z 1
0 = f' 2 D0 j dr j(r + 1)n(h0 + 1)m '(r)j2 < 1; n; m = 0; 1; 2; : : :g : (5.5.47)
0
On 0 , we dene the family of norms
sZ
1
k'k0n;m := dr j(r + 1)n(h0 + 1)m '(r)j2 ; n; m = 0; 1; 2; : : : (5.5.48)
0
5.5 Appendi es to Chapter 5 197

The quantities (5.5.48) fulll the onditions to be a norm (the proof is almost identi al to
the proof of Proposition 1 of Se tion 4.4.4) and an be used to dene a ountably normed
topology 0 on 0 (see [5℄),
0
' !1 ! ' i k' 'k0n;m !1 ! 0 ; n; m = 0; 1; 2; : : : (5.5.49)
One an see that the spa e 0 is stable under the a tion of H0 and that H0 is 0 - ontinuous
(the proof is almost identi al to the proof of Proposition 2 of Se tion 4.4.4).
On e we have onstru ted the spa e 0 , we an onstru t its topologi al dual 0 as the
spa e of 0 - ontinuous antilinear fun tionals on 0 (see [5℄) and therewith the RHS of the
free Hamiltonian
0  L2 ([0; 1); dr)  0 : (5.5.50)
For ea h E 2 Sp(H0 ), we asso iate a ket jE i to the generalized eigenfun tion  0 (r; E )
through
jE i : 0 7 ! C Z 1
' 7 ! h'jE i := '(r) 0 (r; E )dr = (U0 ')(E ) : (5.5.51)
0
The ket jE i in Eq. (5.5.51) is a well-dened antilinear fun tional on 0 , i.e., jE i belongs to
0 (the proof is almost identi al to the proof of Proposition 3 of Se tion 4.4.4). The ket
jE i is a generalized eigenve tor of the free Hamiltonian H0 (the proof is almost identi al to
the proof of Proposition 3 of Se tion 4.4.4),
H0jE i = E jE i ; (5.5.52)
i.e.,
h'jH0jE i = hH0'jE i = E h'jE i ; 8' 2 0 : (5.5.53)

Dira Basis Ve tor Expansion for H0

We are now in a position to derive the Dira basis ve tor expansion for the free Hamilto-
nian. This derivation onsists of the restri tion of the Weyl-Kodaira expansions (5.5.39)
and (5.5.42) to the spa e 0 . If we denote hrj'i  '(r) and hE jri   0 (r; E ), and if we
dene the a tion of the left ket hE j on ' 2 0 as hE j'i := 'b(E ), then Eq. (5.5.39) be omes
Z 1
hE j 'i = dr hE jrihrj'i ; ' 2 0 : (5.5.54)
0
If we denote hrjE i   0 (r; E ), then Eq. (5.5.42) be omes
Z 1
hr j 'i = dE hrjE ihE j'i ; ' 2 0 : (5.5.55)
0
This equation is the Dira basis ve tor expansion of the wave fun tion ' in terms of the free
eigenkets jE i. We an also prove the Nu lear Spe tral Theorem for the free Hamiltonian
(the proof is almost identi al to the proof of Proposition 4 of Se tion 4.4.5),
Z 1
('; H0n )= dE E n h'jE ihE j i ; 8'; 2 0 ; n = 1; 2; : : : (5.5.56)
0
198 5 S attering o the Square Barrier Potential

Energy Representation of the RHS of H0

We have already shown that in the energy representation the Hamiltonian H0 a ts as the
multipli ation operator Eb. The energy representation of the spa e 0 is dened as
b 0 := U00 : (5.5.57)
Obviously  b 0 is a linear subspa e of L2 ([0; 1); dE ). In oder to endow 
b 0 with a topology
b 0 , we arry the topology on 0 into  b 0,

b 0 := U0 0 : (5.5.58)

With this topology, the spa e  b 0 is a linear topologi al spa e. If we denote the dual spa e
of 
b 0 by  
b , then we have
0
U0 0 = (U0 0 ) = 
b :
0 (5.5.59)
If we denote jEb i  U0 jE i, then we an prove that jEbi is the antilinear S hwartz delta
fun tional, (the proof is almost identi al to the proof of Proposition 5 of Se tion 4.4.6),
jEbi : b 7 ! C
'b 7 ! h'bjEbi := 'b(E ) : (5.5.60)
It is very helpful to show the dierent realizations of the RHS through the following
diagram:
H0 ; '(r) 0  L2 ([0; 1); dr)  0 jE i position repr:
# U0 # U0 # U0 (5.5.61)
Eb; 'b(E ) b 0  L2 ([0; 1); dE )  
b
0 jEbi energy repr:
We should stress that the spa e 0 is neither in nor out |the boundary onditions
satised by the elements of the spa es in;out are dierent to the boundary onditions
satised by the elements of 0 .

5.5.2 Appendix 8: Spa es of Hardy Fun tions

In this Appendix, we list the denition and main properties of fun tions of Hardy lass and
review the Bohm-Gadella onstru tion of RHS of Hardy fun tions.

General Properties of Hardy Fun tions

A Hardy fun tion f (z ) on the upper half of the omplex plane C + is a fun tion satisfying
the following onditions [76, 77, 78, 79℄:
i.) It is an analyti fun tion on the open upper half plane, i.e., on the set of omplex
numbers with positive imaginary part.
5.5 Appendi es to Chapter 5 199

ii.) For any value of y > 0, the integral

Z 1
jf (x + iy)j2 dx (5.5.62)
1
onverges.
iii.) For all y > 0, these integrals are bounded by the same onstant K ,
Z +1
sup jf (x + iy)j2 dx < K : (5.5.63)
y>0 1
The set of Hardy fun tions on the upper half plane, often referred to as Hardy fun tions
from above, is a ve tor spa e that we denote by H+2 .
Similarly, Hardy fun tions on the lower half plane C are analyti on the open lower
half plane, and for these fun tions the onditions (ii) and (iii) hold with y < 0. We denote
the ve tor spa e of Hardy fun tions from below by H2 .
Boundary values for Hardy fun tions are dened at almost all (with respe t to the
Lebesgue measure) points of the real axis. For any Hardy fun tion, these boundary values
yield a square integrable fun tion f (x), whi h is uniquely dened on the real line, save for
a set of zero Lebesgue measure,
lim f (x  iy ) = f (x) ; f
y!0
2 H2 : (5.5.64)
The square norm of f (x) is also bounded by K . Thus, a fun tion in H2 uniquely determines
(almost everywhere) a square integrable fun tion on R .
An important theorem, due to Tit hmarsh [80℄, states that Hardy fun tions an be
re overed by their boundary values on the real line. If f (x) is the fun tion representing the
boundary values of a Hardy fun tion f (z ) on C  , then
Z
1 1 f (x)
f (z ) =  dx ; (5.5.65)
2i 1 x z
where the signs (+) and ( ) orrespond to Hardy fun tions on the upper and lower half
plane, respe tively.
Another important theorem on Hardy fun tions is that by Paley and Wiener [81, 76,
77, 78, 79℄, whi h determines whether a square integrable fun tion is also a Hardy fun tion.
The theorem asserts that the Fourier transform F is bije tion between H2 , the spa e of
Hardy fun tions from below, and L2 (R + ) the spa e of square integrable fun tions dened
on the positive real axis. The on lusion holds also for H+2 and L2 (R ). That is,
F L2(R ) = H+2 ; (5.5.66a)
F L 2 (R + ) = H 2 : (5.5.66b)
There is another version of the same result that an be summarized as follows:
FH2+ = L2 (R + ) ; (5.5.67a)
FH2 = L2 (R ) : (5.5.67b)
200 5 S attering o the Square Barrier Potential

Sin e the Fourier transform is an isometry on L2 (R ), we on lude that H+2 and H2 are
losed subspa es of L2 (R ), and hen e Hilbert spa es. Sin e L2 (R ) = L2 (R + )  L2 (R ),
where  stands for orthogonal dire t sum, we have
L2 (R ) = H+2  H2 : (5.5.68)
A theorem due to van Winter [53℄ establishes that a Hardy fun tion an be re overed
by its boundary values on the semi-axis R + . Whether the re overed fun tion is an element
of H+2 or H2 is to be determined by means of the Mellin transform. Thus, if we all H++ 2
the spa e of boundary values on R + of the fun tions in H+2 and H2 + the spa e of boundary
values on R + of the fun tions in H2 , we have the following bije tion:
H+2 = H++
2 ; (5.5.69a)
 H2 = H2 + ; (5.5.69b)
where the image of any f (x) 2 H2 by  is a fun tion whi h is equal to f (x) for x 2 R +
and is not dened for negative values of x.
The following are among the other interesting properties of Hardy fun tions [80℄:
i.) Let us dene the Hilbert transform for an L2 (R ) fun tion f as
Z 1
f (t)
Hf (x) = 1 P dt ; (5.5.70)
1t x
where P denotes the Cau hy prin ipal value. The Hilbert transform is linear and its image
also lies in L2 (R ). A square integrable omplex fun tion f (x), with real part u(x) and
imaginary part v (x), belongs to H2 if and only if
Hu = v and Hv = u : (5.5.71)
In parti ular, a Hardy fun tion annot be either real or purely imaginary on the whole real
line.
ii.) From i.), we immediately see that f (x) 2 H2 if and only if its omplex onjugate
f  (x) 2 H2 .
iii.) Hardy fun tions vanish at innity. More pre isely, they behave for large values of
jzj as 1=pz ( f. [79℄).
iv.) Some Hardy fun tions on C  admit analyti ontinuation beyond the real axis to
C . We may onsider the fun tions f (x) 2 L2 (R  ) su h that there is a positive number 

with the property that ejxjf (x) 2 L2 (R  ). Then, the Fourier transforms of these fun tions
are Hardy fun tions on C  , and they admit an analyti ontinuation beyond the real axis to
a strip of width . This means that if ejxj f (x) 2 L2 (R ), its Fourier transform is analyti
on fz 2 C ;  < Im z < 1g and if ejxj f (x) 2 L2 (R + ), its Fourier transform is analyti
on fz 2 C ; 1 < Im z < g ( f. [82℄).
v.) A fun tion whi h is simultaneously Hardy on both the upper and lower half planes
would be obviously entire, and, as a onsequen e of above ondition iii), is also bounded.
5.5 Appendi es to Chapter 5 201

Hen e, Liouville theorem asserts that su h a fun tion is onstant. A onstant fun tion
annot be square integrable unless it is zero almost everywhere. However, there exist entire
fun tions that are also Hardy either on the upper or the lower half plane.
vi.) It is now lear that the spa es H+2 and H2 have a trivial interse tion. However, the
spa es of fun tions whi h are restri tions of Hardy fun tions to the positive semiaxis R +
have a nontrivial interse tion. Moreover, the interse tion H++
2 \H2 is dense in L2 (R + ) [83℄.
+

Rigged Hilbert Spa es of Hardy Fun tions

We now summarize the Bohm-Gadella onstru tion to des ribe resonan es. This onstru -
tion is based on RHS of Hardy fun tions.
All fun tions fullling the following onditions yield two rigged Hilbert spa es [20℄:
i.) They belong to the S hwartz spa e S .
ii.) Their supports are in R  .
We all these spa es S  , respe tively. Take their Fourier transforms. Sin e the Fourier
transform of a S hwartz fun tion is again a S hwartz fun tion, the Fourier transforms of the
fun tions in S  have the following properties:
a.) They belong to the S hwartz spa e.
b.) They belong to H2 .
.) The spa e of all these Fourier transforms FS  oin ide with the interse tion of S
and H2 , i.e., FS  = S \ H2 .
d.) Sin e S  is dense in L2 (R  ), S \ H2 is dense in H2 with respe t to the Hilbert
spa e topology inherited from L2 (R ).
e.) Note however that the dire t sum of spa es S \ H+2  S \ H2 does not oin ide with
S , sin e the Fourier transform of any of its fun tions vanish at zero.
f.) Sin e S \ H2 are subspa es of S , they inherit the topology of S . They have good
enough properties so that
S \ H2  H2  (S \ H2 ) (5.5.72)
are well dened rigged Hilbert spa es.
We mentioned earlier in this Appendix that Hardy fun tions are determined by their
values on the positive semiaxis plus a spe i ation whi h says if they are Hardy on the upper
or the lower half planes. Thus, we have dened the spa es H+2 = H++ 2 and  H2 = H2 .
+
Now onsider:

S \ H+2 R+ = +(S \ H+2 ) ; (5.5.73a)
S \ H2 R+ =  (S \ H2 ) : (5.5.73b)

The spa es S \ H2 R+ are dense in L2 (R + ). Sin e  is a bije tion, we an transport the
topology from S \ H2 to S \ H2 R+ by means of . The transported topologies have the
same properties as the original ones and they are ner than the Hilbert topology on L2 (R + ).
In parti ular, S \ H2 R+ are metrizable topologi al ve tor spa es and
 
S \ H2 R+  L2 (R+ )  ( S \ H2 R+ ) (5.5.74)
202 5 S attering o the Square Barrier Potential
 
are RHS. The spa es S \ H+2 R+ and S \ H2 R+ have a nontrivial interse tion [84℄.
We an dene the dual of the mapping  using the following formula:
hf(x)j Fi = hf(x)jFi ; 8f(x) 2 S \ H2 ; 8 F 2 (S \ H2 ) : (5.5.75)
The mapping  is a bije tion. Moreover, the following property is fullled algebrai ally
and topologi ally:

 (S \ H2 )  ( S \ H2 R+ ) = ((S \ H2 )) ; (5.5.76)

i.e., the topology on ( S \ H2 R+ ) is transported from (S \ H2 ) by  .
It is important, however, that the mapping  does not extend . The ause lies in the
non unitarity of the latter.
Chapter 6
The Gamow Ve tors of the Square
Barrier Potential Resonan es
In this hapter, we study the resonan es of the square barrier potential. We rst ompute
the resonan e energies as poles of the S -matrix. The integral equation of A. Mondragon et
al. for the Gamow ve tors will be translated into the RHS language. Next, we ompute the
Gamow eigenfun tions in the position representation as the solutions of the time independent
S hrodinger equation subje t to the purely outgoing boundary ondition. The [0; 1)-energy
representation of the Gamow eigenfun tion will be related to the omplex delta fun tion,
and the ( 1; 1)-energy representation of the Gamow eigenfun tion will be related to the
Breit-Wigner amplitude. The semigroup time evolution of the Gamow ve tors will also be
omputed. The Gamow ve tors will be used as basis ve tors. We shall see that the Gamow
ve tors do not form a omplete basis|an additional set of kets needs to be added in order
to obtain a omplete basis. The time asymmetry of the purely outgoing boundary ondition
will be dis losed. To nish this hapter, we shall dis uss the exponential de ay law of the
Gamow ve tors.

203
It don't bring you
Well I know it ain't been roses lately
Baby it's just been thorns
And no matter what we do
Nothing seems to hange
Love has always been my shelter
For you it's been a storm
But for awhile I thought
We'd almost beat the rain
Now there's a hole here in my po ket
Where all my dreams have gone
Falling out like so many ni kels
and dimes
And last of all you
You'd always been my good lu k harm
I should've known that lu k
Is a waste of time
Cause it don't bring you love if you don't love
And it don't bring you time if you ain't got time
And it don't bring you strength baby if you ain't strong
And it don't bring you kindness if you ain't kind
Now there's a whole lot in life to be unsure of
But there's one thing I an safely say I know
That of all the things that nally desert us
Pride is always the last thing to go
But it won't bring you love if you don't love
And it won't bring you time if you ain't got time
And it won't bring you strength baby if you ain't strong
And it won't bring you kindness if you ain't kind
And now I wish you only the roses without the thorns
And I hope your dreams are always within rea h
And I wish you shelter baby from all your storms
They s ared you but they never seemed to tea h
That I an't bring you love if you don't love
And I an't bring you time if you ain't got time
And I an't bring you strength baby if you ain't strong
And I an't bring you kindness if you ain't kind
And I an't bring you kindness if you ain't kind

Mary Chapin Carpenter, State of the Heart

6.1 Introdu tion 205

6.1 Introdu tion

Most elementary parti les are only quasistable states de aying through various intera tions
and thus have nite lifetimes of various orders of magnitude [36℄. Several theoreti al s hemes
have been proposed to des ribe quasistable parti les. The S -matrix and the Gamow ve tors
are two of the most widely used s hemes.
Experimentally, resonan es often appear as peaks in the ross se tion that resemble the
well-known Breit-Wigner distribution [50℄. The Breit-Wigner distribution has two hara -
teristi parameters: the energy ER at whi h the peak rea hes its maximum, and its width R
at half-maximum. The inverse of R is the lifetime of the de aying state [33℄. The peak of
the Breit-Wigner is related to a rst-order pole of the S-matrix in the energy representation
S (E ) at the omplex number zR = ER i R =2. The shape of the theoreti al expression
of the ross se tion in terms of S (E ) ts the shape of the experimental ross se tion in the
neighborhood of ER . This is why the rst-order pole of the S -matrix is often taken as the
theoreti al denition of a resonan e.
Although a resonan e has a nite lifetime, it is otherwise assigned all the properties that
are also attributed to stable parti les, su h as angular momentum, harge, spin, parity and
other parti le labels. For example, a radioa tive nu leus has a nite lifetime, but otherwise
it possesses all the properties of stable nu lei. In fa t, radioa tive nu lei are in luded in the
periodi table of the elements together with the stable nu lei. Therefore, it seems natural
to seek a theoreti al des ription that provides \parti le status" to the quasistable states.
The Gamow ve tors provide this parti le status. The des ription of resonan es by Gamow
ve tors allows us to interpret resonan es as autonomous experimentally de aying physi al
systems. This des ription, impossible in the Hilbert spa e, an be a omplished within the
Rigged Hilbert Spa e.
The original energy eigenfun tion with omplex eigenvalue ER i R =2 was introdu ed
by Gamow [38℄. The quantities ER and R are interpreted as the resonan e energy and the
resonan e width of the de aying state, respe tively. However, Gamow's heuristi approa h
annot be made rigorous in the Hilbert spa e framework, be ause self-adjoint operators on
a Hilbert spa e an only have real eigenvalues. An extended framework is therefore needed.
As we shall see, the Rigged Hilbert Spa e is the most natural framework to des ribe Gamow
ve tors.
In this hapter, Gamow eigenkets will be obtained as solutions of a homogeneous integral
equation of the Lippmann-S hwinger type. In the radial position representation, this integral
equation is equivalent to the time-independent S hrodinger equation subje t to a purely
outgoing boundary ondition. The resonan e spe trum is therefore singled out by the purely
outgoing boundary ondition. As we shall see, this is the same resonan e spe trum as that
dened by the poles of the S -matrix. The Gamow eigenfun tions will be asso iated to ertain
eigenfun tionals, that we all Gamow kets. These Gamow kets are generalized eigenve tors
of the square barrier potential Hamiltonian with omplex eigenvalue ER i R =2.
The energy representation of the Gamow ve tors will be obtained. We shall see that
in the [0; 1)-energy representation (i.e., in the representation asso iated to the physi al
spe trum), the Gamow ve tor is represented by the omplex delta fun tion, whereas in the
206 6 Gamow Ve tors of the Square Barrier Potential Resonan es

( 1; 1)-energy representation (i.e., in the representation asso iated to the support of the
Breit-Wigner amplitude), the Gamow ve tor is represented by the Breit-Wigner amplitude.
We shall also obtain the time evolution of the Gamow ve tors, whi h is given by a semigroup.
Therefore, the Gamow ve tors have all of the properties that are demanded from a
resonan e state:
1. They are eigenve tors of the Hamiltonian with omplex eigenvalues.
2. They orrespond to the Breit-Wigner amplitude in the energy representation.
3. Their time evolution is given by a semigroup, and obeys the exponential de ay law.
The organization of this hapter is as follows. In Se tion 6.2, we ompute the resonan e
energies as poles of the S -matrix. In Se tion 6.3, we introdu e the integral equation that
is satised by the Gamow ve tors. Next, we ompute the Gamow eigenfun tions in the
position representation as the solutions of the time independent S hrodinger equation with
omplex eigenvalues subje t to a purely outgoing boundary ondition. These eigensolutions
will be asso iated to ertain eigenfun tionals (Gamow kets). The [0; 1)-energy represen-
tation of the Gamow eigenfun tion will be related to the omplex delta fun tion, and the
( 1; 1)-energy representation of the Gamow eigenfun tion will be related to the Breit-
Wigner amplitude. In Se tion 6.4, the Gamow ve tors will be used as basis ve tors. We
shall see that the Gamow ve tors do not form a omplete basis|an additional set of kets
needs to be added in order to obtain a omplete basis. The time evolution of the Gamow
ve tors is omputed in Se tion 6.5. Se tion 6.6 deals with the time asymmetry behind the
purely outgoing boundary ondition. Se tion 6.7 studies the exponential de ay law of the
Gamow ve tors.

6.2 S-matrix Resonan es

The S -matrix in the energy representation is given by (see Se tion 5.4)

S (E ) =
J (E ) ; E > 0 : (6.2.1)
J+ (E )
As it stands, this expression is valid only for positive energies. As we said in Se tion 6.1,
the S -matrix resonan es are asso iated to the poles of the analyti ontinuation of S (E )
into the whole omplex plane. Sin e S (E ) is not a single-valued fun tion, it is onvenient
to write the S -matrix as a fun tion of the momentum k before we perform the analyti
ontinuation,
S (k ) =
J (k ) ; k > 0 : (6.2.2)
J+(k)
Here the momentum k is given by r
2m
k= 2 E: (6.2.3)
~
6.2 S-matrix Resonan es 207

Eq. (6.2.3) provides a Riemann surfa e in a natural way. The analyti ontinuation of the
numerator and the denominator of S (k) yield two analyti fun tions J (k). Therefore, the
ontinuation of S (k) is analyti ex ept at its poles. These are pre isely the zeros of the
denominator of S (k) (see [85℄),
J+ (k ) = 0 ; (6.2.4)
where now k is omplex. From Eqs. (4.4.16d) and (5.2.23a) it follows that the equality
(6.2.4) is equivalent to the following:
   
Q iQ(b a) k Q iQ(b a) k
(1 )e sin(ka) + os(ka) + (1 + )e sin(ka) os(ka) = 0 :
k iQ k iQ
(6.2.5)
The solutions of (6.2.5) are the (S-matrix) resonan es of the square barrier potential. Equa-
tion (6.2.5) has a denumerable innite number of omplex resonan e energy solutions. These
solutions ome in pairs ER  i R =2 (see Figure 6.2 of Appendix 6.9). The pole ER i R =2 is
asso iated with the de aying part of the resonan e, and it is lo ated on the lower half-plane
of the se ond sheet of the two-sheeted Riemann surfa e orresponding to the square root
mapping (see Figure 6.2a of Appendix 6.9). The pole ER + i R =2 is asso iated with the
growing or formation part of the resonan e, and it is lo ated on the upper half-plane of the
se ond sheet of the Riemann surfa e (see Figure 6.2b of Appendix 6.9). In the momentum
plane, this pair of energy poles orresponds to a pair of poles Re(k) iIm(k) in the lower
half of the k-plane that are mirror images of one another with respe t to the imaginary axis
(see Figure 6.1 of Appendix 6.9).
The width of the resonan es in reases as the energy in reases, and therefore their lifetime
R = ~= R de reases. The resonan es whose energies are below the top of the barrier E = V0
are lose to the real axis. As E keeps in reasing the resonan es move away from the real
axis towards innity. The square barrier potential poles never orrespond to a bound or a
virtual state, i.e., they do not lie in the imaginary axis of the momentum plane. The square
barrier potential poles are always simple ( f. [86℄ for an example of a barrier with double
poles).
In order to distinguish ea h of the denumerable innite number of resonan e poles, the
de aying resonan e energies of the square barrier potential will be denoted by
n
zn = En i
; n = 1; 2; : : : ; (6.2.6)
2
whereas the growing resonan e energies will be denoted by
zn = En + i n
; n = 1; 2; : : : (6.2.7)
2
The orresponding momentum poles will be denoted respe tively by
p
kn = Re(kn ) i Im(kn ) = zn ; n = 1; 2; : : : ; (6.2.8)
and by p
kn = Re(kn ) i Im(kn ) = zn ; n = 1; 2; : : : (6.2.9)
208 6 Gamow Ve tors of the Square Barrier Potential Resonan es

6.3 The Gamow Ve tors

Gamow ve tors are usually dened as eigensolutions of the S hrodinger equation subje t to
a purely outgoing boundary ondition (see Se tion 6.3.2 below). Although we ould start
the study of Gamow ve tors with that denition, we will follow instead the pro edure used
for the Lippmann-S hwinger kets. We will dene a Gamow ve tor as the solution of an
integral equation. This integral equation has that purely outgoing boundary ondition built
into it. Needless to say, the solutions of that integral equation will be found by solving the
time independent S hrodinger equation subje t to the purely outgoing boundary ondition.

6.3.1 Lippmann-S hwinger Equation of the Gamow Ve tors

The Gamow ve tors are solutions of an integral equation of the Lippmann-S hwinger type.
If zR = ER i R =2 denotes the omplex energy asso iated to a resonan e of energy ER and
width R , then the de aying Gamow ve tor jzR i fullls
1
j zR i = z V jz i :
H 0 + i0 R
(6.3.1)
R

This equation was introdu ed (with a dierent notation) by A. Mondragon et al. in Ref. [40℄.
The +i0 in Eq. (6.3.1) means that we are working with the retarded free Green fun tion,
whi h has a purely outgoing boundary ondition built into it. The retarded free Green
fun tion is analyti ally ontinued a ross the ut into the lower half plane (of the se ond
sheet of the Riemann surfa e), where the omplex number zR is lo ated. Therefore, as
A. Mondragon has pointed out, Eq. (6.3.1) should be written as
1
jzR i = Elim
!zR E H0 + i0
V jE i : (6.3.2)

The notation in this equation expresses better the fa t that we rst have to ompute the
retarded free Green fun tion (E H0 + i0) 1 and then ontinue it a ross the ut into the
lower half plane.1 The minus sign in jzR i means that the de aying Gamow ve tor will be
dened as an antilinear fun tional over the 2 +.
As we said above, the integral equation (6.3.1) has a purely outgoing boundary ondition
built into it. To be more pre ise, Eq. (6.3.1) in the position representation is equivalent to
the time independent S hrodinger equation subje t to the ondition that far away from the
potential region the solution behave as a purely outgoing wave.
As we saw in Se tion 6.2, to ea h de aying pole zR = ER i R =2 of the S -matrix there
orresponds a growing pole zR = ER + i R =2. We now asso iate a growing Gamow ve tor
1
This also shows that a onsistent notation will always have ipping of signs like '+ 2  , 2 + ,
et . This ipping of signs omes from the fa t that we perform analyti ontinuations from the upper (lower)
rim of the ut, whi h is labeled by +i0 ( i0), into the lower (upper) half plane, whi h is labeled by C
(C + ).
6.3 The Gamow Ve tors 209

jzR +i to the pole of the S -matrix at zR . The integral equation satised by jzR + i should
read
jzR +i = z H1 i0 V jzR +i = Elim
1
!zR E H0 i0
V jE i : (6.3.3)
R 0
In ontrast to Eq. (6.3.1), Eq. (6.3.3) has a purely in oming boundary ondition built into
it. That is, Eq. (6.3.3) in the position representation is equivalent to the time independent
S hrodinger equation subje t to the ondition that far away from the potential region the
solution behave as a purely in oming wave.

6.3.2 The Gamow Ve tors in Position Representation

In this se tion, we obtain the solutions of the integral equations (6.3.1) and (6.3.3). In
oder to do so, we will work in the radial position representation. In this representation we
will solve the S hrodinger equation under purely outgoing boundary onditions. We will
see that there is a one-to-one orresponden e between the omplex poles of the analyti ally
ontinued S -matrix and the omplex generalized eigenvalues obtained under purely outgoing
boundary onditions.
In the radial representation, Eqs. (6.3.1) and (6.3.3) read
1 1
hrjzR i = hrj z H0 + i0
V jzR i = Elim h r j
!zR E H0 + i0
V jE i ; (6.3.4a)
R
1 1
hrjzR +i = hrj z H0 i0
V jzR + i = lim hrj
E !zR E H0 i0
V jE i : (6.3.4b)
R
Instead of solving these integral equations, we solve the equivalent S hrodinger dierential
equation  
~2 d2
+ V (r) hrjzR i = zR hrjzR i ; (6.3.5)
2m dr2
subje t to the boundary onditions built into them,
h0jzR i = 0 (6.3.6a)
hrjzR i is ontinuous at r = a and at r = b (6.3.6b)
d
hrjz i is ontinuous at r = a and at r = b
dr R
(6.3.6 )
hrjzR i  eikRr as r ! 1 ; (6.3.6d)
where r
2m
kR = zR : (6.3.7)
~2
In Eqs. (6.3.5) and (6.3.6), we have used the same symbol hrjzR i to denote both hrjzR i
and hrjzR + i. This will reate no onfusion, be ause whenever zR = ER i R =2, then
hrjzR i will mean hrjzR i, and whenever zR = ER + i R =2, then hrjzRi will mean hrjzR +i.
Condition (6.3.6d) is the purely outgoing boundary ondition. At rst glan e, it may look
210 6 Gamow Ve tors of the Square Barrier Potential Resonan es

like we have imposed also a purely outgoing boundary ondition upon the growing Gamow
eigenfun tions hrjzR + i. However, sin e kR is omplex, the fun tion eikR r is not always an
outgoing wave. In fa t, it is an outgoing wave only when Re(kR ) is positive, i.e., when we
are working with the de aying Gamow ve tor hrjzR i, and it is an in oming wave only when
Re(kR ) is negative, i.e., when we are working with the growing Gamow ve tor hrjzR + i. This
means that working with the momentum kR rather than with zR will allow us to obtain the
de aying and the growing Gamow ve tors at the same time.
The purely outgoing boundary ondition (6.3.6d) is often written as
du(r; zR )
lim ikR u(r; zR ) = 0 ; (6.3.8)
r!1 dr
where
u(r; zR ) = hrjzR i : (6.3.9)
One an easily he k that (6.3.8) is equivalent to (6.3.6d).
If we impose the onditions (6.3.6a)-(6.3.6 ) upon the general solution of Eq. (6.3.5), we
obtain that, up to a normalization fa tor, the solution has the form
8
< sin(kR r ) 0<r<a
(r; zR )  (r; kR ) = J1 (kR )e iQ R r + J2 (kR )e iQR r a<r<b (6.3.10)
:
J3(kR)eikRr + J4(kR )e ikRr b < r < 1 ;
where r r
2 2m 2m
QR = kR 2 V0 = 2 (zR V0 ) : (6.3.11)
~ ~
The eigensolution (6.3.10), whi h does not satisfy the purely outgoing boundary ondition
yet, is equal to the analyti ontinuation of the regular solution (r; E ) of Eq. (5.2.22). In
Eq. (6.3.10), there is no restri tion on the values that zR an take, i.e., before imposing
the purely outgoing boundary ondition zR an be any omplex number. If we now impose
(6.3.6d) upon the eigensolution (6.3.10), then the oeÆ ient J4 (kR ) = i=2 J+(kR ) must be
zero. Sin e this ondition is the same as the ondition (6.2.4) for the omplex poles of the
S -matrix, the set of generalized omplex eigenvalues zR must in lude the set of S -matrix
resonan e poles. We now show that these two sets of solutions are the same.
The boundary onditions (6.3.6) an be written in terms of the oeÆ ients of (6.3.10)
as
J1eiQRa + J2e iQRa = sin(kR a) (6.3.12a)
iQR (J1 e iQ R a J2e ) = kR os(kR a)
iQ R a (6.3.12b)
J3e = J1e + J2e
ik R b iQ R b iQ R b (6.3.12 )
ikR J3 e R = iQR (J1 e R J2 e R ) :
ik b iQ b iQ b (6.3.12d)
Writing this set of linear equations as a matrix equation we obtain
0 10 1 0 1
sin(kR a) 0 eiQR a e iQR a 1 0
C B J3 C = B 0 C :
B kR os(kR a) 0 iQR e iQ R a iQR e iQ R a C B C B C
B
 0 ik
e R b iQ
e R b e R A  J1 A  0 A
iQ b (6.3.13)
0 ikR eikR b iQR eiQR b iQR e iQR b J2 0
6.3 The Gamow Ve tors 211

This is a homogeneous system of four equations with four unknowns. The system has a
non-trivial solution i the determinant of the oeÆ ients is equal to zero,
 

 sin(kR a) 0 eiQR a e iQR a 

 kR os(kR a) 0 iQR eiQR a iQR e iQR a  = 0 : (6.3.14)

 0 e Rb
ik eiQR b e iQR b 
 0 ikR eikR b iQR e iQ R b iQR e iQR b 
Straightforward omputations show that the ondition (6.3.14) is the same as the ondition
(6.2.5). Thus the set of generalized eigenvalues of the time independent S hrodinger equation
subje t to purely outgoing boundary onditions is the same as the set of S -matrix poles.
As we mentioned earlier, the solutions of (6.2.5) ome in pairs of a growing and a de aying
pole. We have denoted those poles by zn and zn and their orresponding momenta by kn
and kn . The eigenfun tion asso iated to zn = En i n =2 is the de aying Gamow ve tor
in the position representation, whose radial part, up to a normalization fa tor, is
8
> 1 sin(k r ) 0<r<a
< J3 (kn )
> n
un(r; zn )  un(r; kn) = J31 (knn ) eiQn r + JJ32 ((kknn )) e iQn r
J (k )
a<r<b (6.3.15)
>
>
: ikn r e b < r < 1:
The eigenfun tion asso iated to zn = En + i n =2 is the growing Gamow ve tor in the position
representation, whose radial part, up to a normalization fa tor, is
8
> 1
< J3 ( kn )
> sin( kn r) 0<r<a
 
un (r; zn )  un(r; kn ) = JJ31 (( kknn )) e iQn r + JJ23 (( kknn )) eiQn r a<r<b (6.3.16)
>
>
: ik r
e n b < r < 1:
Form equations (6.3.10), (6.3.15) and (6.3.16) it follows that the Gamow eigenfun tions are
proportional to the analyti ontinuation of the regular solution,
2i
un(r; kn) =
J (kn) (r; kn) ; (6.3.17a)
2i
un(r; kn ) = 
J ( kn ) (r; kn) : (6.3.17b)

The Gamow ve tor (6.3.15) is dened up to a normalization fa tor. By normalization

we mean the fun tion Nn of zn by whi h we an multiply the Gamow eigenfun tion un (r; zn )
to obtain another eigenfun tion Nn un(r; zn ) with the same eigenvalue zn and satisfying the
same boundary onditions (6.3.6). If no onfusion arises, we denote the normalized Gamow
ve tor also by un (r; zn),
8
> Nn sin(k r ) 0<r<a
< J3 (kn )
> n
Nn J1 (kn ) iQn r
un (r; zn)  un (r; kn) =
> J 3 (kn ) e + NJn3J(2k(nk)n ) e iQn r a<r<b (6.3.18)
>
: N eikn r
n b < r < 1:
212 6 Gamow Ve tors of the Square Barrier Potential Resonan es

After this normalization Eq. (6.3.17a) be omes

2iNn
un(r; zn )  un (r; kn) =
J (kn) (r; kn) (6.3.19)

There is an extensive literature on the normalization of Gamow ve tors ( f. [40℄ and refer-
en es therein). We shall not treat this problem here, although we would like to mention
that the normalization proposed by A. Mondragon et al. [40℄ seems to be the most suitable.
The eigenfun tions un (r; zn) of the dierential operator h are obviously not square in-
tegrable, i.e., they do not belong to the Hilbert spa e L2 ([0; 1); dr). In order to onstru t
an eigenket jzn i of the Hamiltonian H asso iated to the eigenfun tion un (r; zn), we follow
the pattern of Se tion 5.3. The Gamow ket jzn i asso iated to the eigenfun tion un(r; zn ) is
dened by
jzn i : + 7 ! C Z 1
7 ! h jzn i := dr (r)u(r; zn) : (6.3.20)
0

Under the assumptions made upon the elements of + , the fun tion jzn i is a well-dened
antilinear fun tional. It is easy to show that the fun tion jzn i is a generalized eigenve tor
of H with omplex eigenvalue zn ,
H  j z n i = zn j z n i ; (6.3.21)
or more pre isely,
h jH jzn i = znh jzn i ; 8 2 + : (6.3.22)
In a similar vein, we an dene a ket jzn + i asso iated to the eigenfun tion un(r; zn ),
jzn +i :  7 ! C Z 1
' 7 ! h' jzn i :=
+ +  +
dr '+ (r)u(r; zn ) ; (6.3.23)
0

and prove that this is a well dened antilinear eigenfun tional of the Hamiltonian H with
omplex eigenvalue zn ,
h'+jH jzn +i = zn h'+jzn +i ; 8'+ 2  : (6.3.24)

6.3.3 The Gamow Ve tors in Energy Representation

On e we have onstru ted the Gamow ve tor jzn i in the position representation, it is
straightforward to ompute its energy representation. We shall obtain the Gamow ve tor
in two dierent energy representations. One of them is the standard energy representation
asso iated to the physi al spe trum, whi h is [0; 1) in our example. The other one is asso-
iated to the ( 1; 1) spe trum. These two representations are related by the fun tion  of
6.3 The Gamow Ve tors 213

Appendix 5.5.2. We shall show that in the [0; 1)-energy representation the Gamow ve tor
a ts as the antilinear omplex delta fun tional2 multiplied by a normalization fa tor. In the
( 1; 1)-energy representation, the Gamow ve tor an be asso iated to the Breit-Wigner
amplitude multiplied by a normalization fa tor.
First, we dene the antilinear omplex delta fun tional jzbn i,
jzbn i : b +jR+ 7 ! C
b 7 ! h b jzb i := e (zn ) ; (6.3.25)
n

where the fun tion e (z ) is dened by

e (z ) := b (z ) : (6.3.26)
The fun tion jzbn i is a well dened antilinear fun tional over the spa e b + jR+ , sin e the
elements b of  b + jR+ were taken to be in H2 .
+
It an be shown that the [0; 1)-energy representation of jzn i is the antilinear omplex
delta fun tional jzbn i save for a normalization fa tor An ,
U  jzn i = An jzbn i : (6.3.27)
The proof of Eq. (6.3.27) is as follows: from Eqs. (5.2.21) and (6.3.19) we on lude
that the Gamow eigenfun tion u(r; kn ) is proportional to the analyti ontinuation of
the Lippmann-S hwinger eigenfun tion  (r; k),
un (r ; kn ) = 2iNn  (r ; kn ) : (6.3.28)
Then the Gamow eigenfun tion is proportional to the analyti ontinuation of the
eigenfun tion  (r; E ),
un (r ; zn ) = An  (r ; zn ) : (6.3.29)
From Eq. (6.3.29) and from the (assumed) properties of the elements 2 + (see
Se tion 5.3) it follows that
b Z
h jU jzn i = h jzn i
1
= (r)un(r; zn)
=
0
An
Z1 () (; )
dr

dr r  r zn

= An zn b( )
0

= An zn ; h b jb i 8 b 2 b j
+ R+ ; (6.3.30)
where in the last step we have used the denition (6.3.25). This proves Eq. (6.3.27).
2
For a great deal of information about the antilinear omplex delta fun tional and its representations,
the reader is referred to [87℄.
214 6 Gamow Ve tors of the Square Barrier Potential Resonan es

If we write the a tion of jzbn i as an integral operator, the kernel of integration is the
omplex delta fun tion,
Z 1
h b jzbn i = dE An Æ (E zn ) b (E ) : (6.3.31)
0
We are now going to study the relation between the omplex delta fun tion and the
Breit-Wigner amplitude. We shall show that
1
 j i = Anjzbn i ; (6.3.32)
E zn
where  is the dual extension of the fun tion , and  is the fun tion that takes any fun tion
of  E zn i
b + into its restri tion to the positive real line ( f. Appendix 5.5.2). The fun tional j 1
of Eq. (6.3.32) is asso iated to the Breit-Wigner amplitude by
j E 1 z i : b + 7 ! C
n Z 1  
1 1 A
 b 7 ! h b j
1 1 i := dE 2i E z  1 b (E ) ;
n
E zn 1 n
(6.3.33)
where An is the normalization fa tor of Eq. (6.3.27). We shall all the fun tional dened
by Eq. (6.3.33) the Breit-Wigner fun tional. By Tit hmarsh theorem [80℄, the Breit-Wigner
fun tional is a well dened antilinear fun tional.
The key property that will be used to prove (6.3.32) is that the fun tions b of  b + jR+ are
boundary values of Hardy fun tions from above. In order to build the ground of that proof,
we rst show the relation between the [0; 1)-energy representation and the ( 1; 1)-energy
representation:
b ; b +jR+  L2 ([0; 1); dE )  (b +jR+ ) [0; 1) energy repr:
" " 
 1b ; b +  H+2  b + ( 1; 1) energy repr:
(6.3.34)
where H+ is the spa e of Hardy fun tions from above. It is worthwhile noting that although
2
we have denoted the fun tions  1 b and b by a dierent symbol, they are indeed the same
fun tion. More pre isely, they are dierent \pie es" of the same fun tion. In parti ular, the
value of their analyti ontinuation at a omplex number z is the same,
 1 b (z ) = b (z ) : (6.3.35)
Obviously, the fun tions e and  1eenjoy an analogous property,
 1 e (z ) = e (z ) : (6.3.36)
The reason why we use a dierent symbol for dierent \pie es" of the same fun tion is that
the the proof on the onne tion between the Breit-Wigner amplitude and the omplex delta
fun tion be omes more apparent:
6.3 The Gamow Ve tors 215

Let  1 b 2 b . Then e = b 2 H . By Tit hmarsh theorem [80℄,

 1  1 2

e ( ) = 21 Z 11 1 e ( )
+

 1 zn
i
dE (6.3.37)
E zn
 1 E :

Multiplying this equation by An we obtain

An  1 (zn ) =e 1 1 dE An  Z 1 e( ) E : (6.3.38)
2i 1 E zR
From Eqs. (6.3.33) and (6.3.38) it follows that
An  1 e( zn ) = h 1 bj E
1
zn
i: (6.3.39)
We now dene the a tion of  on j E 1zn i by
b
h jj E 1 z i := h
n
1 bj E
1
zn
i: (6.3.40)
Eqs. (6.3.39) and (6.3.40) lead to
b
h jj E 1 z i = An 
n
1 e( )
zn : (6.3.41)
Taking advantage of Eq. (6.3.36) we an write (6.3.41) as
b
h jj E 1 z i = An (zn ) ; 8 2 +jR+ :
n
e b b (6.3.42)

b
The right hand side of this equation equals the a tion of the omplex delta fun tion
at multiplied by An ,
b
h jj 1 i = A h jz i ; 8 2  j + ;
E zn
n bb n
b b
(6.3.43) +R

whi h proves (6.3.36).

Therefore, the Gamow eigenfun tion un(r; zn ), the omplex delta fun tion (multiplied
by a normalization fa tor) An Æ (E zn ) and the Breit-Wigner amplitude (multiplied by a
normalization fa tor) iAn =2 (E zn ) orrespond to the same obje t in dierent represen-
tations,
un (r; zn ) ! An Æ (E zn ) ; E 2 [0; 1) ! 1 An
2i E zn ; E2( 1; 1)
posit. repr. [0; 1)-energy repr. (1; 1)-energy repr.
(6.3.44)
216 6 Gamow Ve tors of the Square Barrier Potential Resonan es

We summarize the results of this se tion in the following diagram:

H; (r ) +  L2 ([0; 1); dr)  + hrjzn i  un(r; zn)
#U #U # U
Eb; b (E ) b +jR+  L2 ([0; 1); dE )  (b +jR+ ) h E jzn i  AnÆ(E zn )
" " 
Eb; b (E ) b +  H+2  b + h E jzn i  1 An
2i E zn
(6.3.45)
In this diagram, we have denoted both (E ) 2  j and 
b b + R+ (E ) 2  by the same
1b b+
symbol (E ), sin e no distin tion is ne essary any more. We stress that in the se ond line
b
of (6.3.45) the energy is allowed to run over the physi al spe trum [0; 1), whereas in the
bottom line the energy is allowed to run over ( 1; 1).

6.4 Complex Basis Ve tor Expansion

The Lippmann-S hwinger kets are basis ve tors that were used to expand a normalizable
smooth fun tion '+ as in (5.3.7). The Gamow ve tors are also basis ve tors. The expansion
generated by the Gamow ve tors is alled the omplex basis ve tor expansion. However, the
Gamow ve tors do not form a omplete basis system. The omplex basis ve tor expansion
needs an additional set of Dira kets orresponding to the energies that lie in the negative
real axis of the se ond sheet of the Riemann surfa e. This has been realized also by other
authors [40, 43, 42℄, who have used the Green fun tion to onstru t the omplex basis ve tor
expansion.
In this se tion, we expand a normalizable wave fun tion in terms of the Gamow ve tors
and a ontinuous set of Dira kets. The Gamow ve tors ontain the resonan e ontribution,
whereas the ontribution of the additional set of Dira kets is interpreted as a ba kground.
The omplex basis ve tor expansion is not valid for every normalizable wave fun tion, i.e.,
for every element of the Hilbert spa e, but only for fun tions '+ 2  . The te hni alities
of that expansion an be found in [88℄.
In a s attering experiment, we measure the transition probability from a state (in-state)
+
' into an observable (out-state) . In Se tion 5.4, we wrote the amplitude of this prob-
ability as Z 1


+
;' = h jE iS (E )hE +j'+idE : (6.4.1)
0
The fun tion S (E ) is the energy representation of the S-matrix. This fun tion an be
analyti ally ontinued into a two-sheeted Riemann surfa e, and the quantity S (E ) in (6.4.1)
represents the boundary value of this analyti fun tion on the upper lip of the ut in the
Riemann surfa e. We now extra t the resonan e ontribution out of (6.4.1). This resonan e
ontribution is arried by the Gamow ve tors. In order to do so, we deform the ontour of
6.5 Semigroup Time Evolution of the Gamow Ve tors 217

integration into the lower half-plane of the se ond sheet of the Riemann surfa e, where the
de aying resonan e poles are lo ated (see Figure 6.3a of Appendix 6.9). Using the results
that appear in [88℄, we an write (6.4.1) as


Z 1 1
X
; '+ = h jE iS (E )hE +j'+idE 2i rn h jzn ih+znj'+i ; (6.4.2)
0 n=0

where rn denotes the residue of the S -matrix at zn . The integral in Eq. (6.4.2) is done on
the negative real semiaxis of the se ond sheet of the Riemann surfa e. The series in (6.4.2)
an be shown to be onvergent [88℄. Omitting in (6.4.2), we get the omplex basis ve tor
expansion for the states,
Z 1 1
X
'+ = jE iS (E )h+E j'+idE 2i rn jzn ih+zn j'+i : (6.4.3)
0 n=0

In Eq. (6.4.3), the innite sum ontains the resonan es ontribution, while the integral
is interpreted as the ba kground ontribution. Needless to say, the Gamow ve tors in
Eq. (6.4.3) are dened up to a normalization fa tor.
Similarly, we obtain the omplex basis ve tor expansion for the observable [88℄, but
now we deform the ontour of integration into the upper half-plane of the se ond sheet of
the Riemann surfa e, where the growing resonan e poles are lo ated (see Figure 6.3b)
Z 1 1
X
= j iS (E )hE j idE + 2i
E+ rn jzn + ih zn j i: (6.4.4)
0 n=0

In this equation, rn denotes the residue of S -matrix at zn . The integration in (6.4.4) is
performed on the negative real semiaxis of the se ond sheet of the Riemann surfa e. The
series in (6.4.4) has been shown to be onvergent [88℄.

6.5 Semigroup Time Evolution of the Gamow Ve tors

Now, we want to study the time evolution of the Gamow ve tors. In order to do it, we
need to onsider the operator U (t) = e iHt=~ , whi h governs the time evolution of the ve tor
states in Hilbert spa e. The operator onjugate to U (t) is dened by
hjU (t)jF i = hU y(t)jF i ; (6.5.1)
where  denotes '+ 2  or 2 + and F belongs to  or +, respe tively. Eq. (6.5.1)
will also be denoted as
h j e iH  t=~ jF i = heiHt=~ jF i :
(6.5.2)
The onjugate operator U  (t) determines the time evolution of the elements in  , whenever
it an be dened.
218 6 Gamow Ve tors of the Square Barrier Potential Resonan es

We have assumed that  b  = S \ H2 . Under this assumption, the following statements


an be proved [20, 91℄:
i.) For any t  0, eiHt=~ +  + and eiHt=~ is ontinuous on + . For any t < 0, there
is a 2 + su h that eiHt=~ 2= +.
ii.) For any t  0, eiHt=~    and eiHt=~ is ontinuous on  . For any t > 0, there
is a '+ 2  su h that eiHt=~ '+ 2=  .
This leads to the following onsequen es:
i.) For any t  0, one an dene the onjugate of U (t) = e iHt=~ , whi h extends the
evolution operator U (t) = e iHt=~ to a ontinuous operator and denes the time evolution

on + . This annot be done for any t < 0. We denote this extension as U+ (t) = e+iH t=~ .
ii.) For any t  0, one an dene the onjugate of U (t) = e iHt=~ , whi h extends the
evolution operator U (t) = e iHt=~ to a ontinuous operator and denes the time evolution

on  . This annot be done for any t > 0. We denote this extension as U  (t) = e iH t=~ .
It seems natural to onsider these extensions as the operators that determine the time
evolution of the obje ts in the dual spa es. We observe that this time evolution is not
dened for all values of t. This has important onsequen es for the time evolution of the
Gamow ve tors.
Consider the de aying Gamow ve tor jzn i. From the above omments, it is not diÆ ult
to see [20℄ that the evolution operator U+ (t) a ts on jzn i if and only if t  0,
U+(t)jzn i = e iEn t=~ e n t=(2~)
j zn i ; only for t  0 ; (6.5.3)
or more pre isely,
h jU+(t)jzn i = e iEn t=~ e n t=(2~)
h jzn i ; 8 2 + ; t  0 only : (6.5.4)
We see that the Gamow ve tor de ays exponentially. Thus it fullls the properties that are
demanded from the de aying resonan e states. Mu h more signi ant is the property that
the time evolution of the de aying Gamow ve tors o urs for t > 0 only|the time evolution
of Gamow ve tors is time asymmetri .
Consider now the Gamow ve tor jzn+ i. The evolution operator U  (t) a ts on jzn+ i if
and only if t  0,
U (t)jzn+ i = e iEn t=~ e n t=(2~) jzn+ i ; only for t  0 ; (6.5.5)
or more pre isely,
h'+jU (t)jzn+i = e iEn t=~ e n t=(2~) h'+jzn+ i ; 8'+ 2  ; t  0 only : (6.5.6)
Therefore, the symmetri group evolution of the Hilbert spa e splits up into two semi-
groups, expressing time asymmetry on a mi ros opi level.3
3
For more on the des ription of time asymmetry in Quantum Me hani s in terms of the propagators,
the reader is referred to [32℄.
6.6 Time Asymmetry of the Purely Outgoing Boundary Condition 219

6.6 Time Asymmetry of the Purely Outgoing Bound-

ary Condition
The semigroup time evolution of the Gamow ve tors expresses the time asymmetry built
into them. We will show here that the purely outgoing boundary ondition that singles
out the resonan e energies has also a time asymmetry built into it. To be more pre ise,
we will show that the purely outgoing boundary ondition should read as purely outgoing
only for the de aying part of a resonan e and as purely in oming for the growing part of
the resonan e. Be ause the purely in oming ondition is the time reversed of the purely
outgoing one, the growing Gamow ve tor an be viewed as the time reversed of the de aying
Gamow ve tor [60℄ (see also next hapter).

6.6.1 Outgoing Boundary Condition in Phase

First, we study the meaning of the purely outgoing boundary ondition when it is imposed
on the de aying part of the resonan e. The omplex energy asso iated to the de aying
part of a resonan e is zn = En i n =2 (En ; n > 0) whi h lies in the fourth quarter of
the se ond sheet of the Riemann surfa e (see Figure 6.2a). Its orresponding momentum
kn = Re(kn ) iIm(kn ) (Re(kn ), Im(kn) > 0) lies in the fourth quadrant of the momentum
plane (see Figure 6.1). The de aying Gamow ve tor un (r; zn) of Eq. (6.3.15) was obtained
after imposing the purely outgoing boundary ondition (6.3.6d) upon (6.3.10). If we had not
imposed this ondition, we would had obtained a solution of the form (6.3.10), and every
omplex number would had been a generalized eigenvalue of the Hamiltonian. In the region
r > b, this solution would had been the sum of two linearly independent solutions
ude aying
in oming (r ; zn ; t) = J4e iknr e izn t=~


= J4e Im(kn )r nt=(2~) e iRe(kn )r iEn t=~ ; r > b; (6.6.1)
whi h we all in oming de aying Gamow ve tor, and
ude aying
outgoing (r ; zn ; t) = J3eikn r e iznt=~


= J3eIm(kn )r n t=(2~) eiRe(kn)r iEn t=~ ; r > b; (6.6.2)
whi h we all outgoing de aying Gamow ve tor. These names ome from the standard
interpretation (see, for instan e, [92℄) of plane waves with a omplex exponent: the expo-
nential with purely imaginary exponent|the term that arries the phase|is interpreted as
the term that governs the propagation of the wave, and the exponential with real exponent
is interpreted as the term that just hanges the amplitude of the wave on the surfa es of
equal phase [92℄. We are going to interpret (6.6.1) and (6.6.2) in the same fashion. The
terms between bra kets in (6.6.1) and (6.6.2) determine the amplitude of the waves. The
propagation of ude aying
outgoing is governed by e
iRe(kn )r e iEn t=~ , and therefore ude aying is an outgoing
outgoing
de aying
wave (in phase). Analogously, the propagation of uin oming is governed by e iRe(kn )r e iEn t=~ ,
and thus ude aying
in oming is an in oming wave (in phase). Imposing the purely outgoing boundary
220 6 Gamow Ve tors of the Square Barrier Potential Resonan es

ondition J4 = 0 is tantamount to forbidding ude aying

in oming . Thus for the de aying part of the
resonan e the purely outgoing boundary ondition allows only purely outgoing waves (in
phase).
The meaning of the purely outgoing boundary ondition applied to the growing part
of the resonan e is the opposite. The growing energy eigenvalue zn = En + i n =2 lies in
the rst quadrant of the se ond sheet of the Riemann surfa e (see Figure 6.2b), and its
momentum kn = Re(kn) iIm(kn ) lies in the third quadrant of the momentum plane
(see Figure 6.1). The growing Gamow ve tor un (r; zn ) of Eq. (6.3.16) was obtained after
imposing the ondition J4 = 0 on (6.3.10). If we had not imposed this ondition, in the
region r > b the solution would had been the sum of two linearly independent solutions
ugrowing 
in oming (r ; zn ; t) = J3e ikn r e izn t=~


= J3eIm(kn )r+ n t=(2~) e iRe(kn )r iEn t=~ ; r > b; (6.6.3)
that we all in oming growing Gamow ve tor, and
ugrowing 
outgoing (r ; zn ; t) = J4eikn r e izn t=~


= J4e Im(kn)r+ n t=(2~) eiRe(kn )r iEn t=~ ; r > b; (6.6.4)
that we all outgoing growing Gamow ve tor. The names ome also after the standard
interpretation [92℄ of plane waves with a omplex exponent. Therefore, the purely outgoing
boundary ondition J4 = 0, when applied to the growing part of a resonan e, bans ugrowing
outgoing
and allows only purely in oming waves (in phase).

6.6.2 Outgoing Boundary Condition in Probability Density

In the previous se tion, we showed how the time asymmetry built into the purely outgoing
boundary ondition ae ted the phase of the Gamow ve tors. In this se tion, we show the
same time asymmetry but now onsidering the probability density of the Gamow ve tors.
For the de aying part of the resonan e, the probability densities (before imposing the
purely outgoing boundary ondition) are obtained by taking the absolute value square of
(6.6.1)

in oming (r ; zn ; t) = juin oming (r ; zn ; t)j =

de aying jJ4j2e
de aying 2 2Im(kn )r n t=~

= jJ4j2e n =~(t+r=vn )
; r > b; (6.6.5)
that we all in oming de aying probability density, and of (6.6.2)

outgoing (r ; zn ; t) = juoutgoing (r ; zn ; t)j =

de aying jJ3j2e2Im(kn )r n t=~
de aying 2

= jJ3j2e n =~(t r=vn ) ; r > b ; (6.6.6)

that we all outgoing de aying probability density (vn = n =(2~Im(kn))). By imposing the
purely outgoing boundary ondition J4 = 0, we only allow (6.6.6) and forbid (6.6.5), whi h
6.7 Exponential De ay Law of the Gamow Ve tors 221

we interpret by saying that we have a purely outgoing probability density ondition for the
de aying part of the resonan e.
For the growing part of the resonan e, the probability densities (before imposing J4 = 0)
are the absolute value square of (6.6.3)
in oming (r ; zn ; t) = juin oming (r ; zn ; t)j = jJ3j2e2Im(kn)r+ n t=~
growing  growing  2
= jJ3j2e n =~(t+r=vn ) ; r > b ; (6.6.7)
that we all the in oming growing probability density, and of (6.6.4)
outgoing (r ; zn ; t) = juoutgoing (r ; zn ; t)j = jJ4j2e 2Im(kn )r+ nt=~
growing  growing  2
= jJ4j2e n =~(t r=vn ) ; r > b ; (6.6.8)
that we all the outgoing growing probability density. For this growing part, the ondition
J4 = 0 leads to the on lusion that in the growing stage of a resonan e only waves with
purely in oming probability density are allowed.
In short, the purely outgoing boundary ondition (6.3.6d) must be read as purely outgoing
(in phase or in probability density) only for the de aying part of the resonan e and as purely
in oming (in phase or in probability density) for the growing part of the resonan e.

6.7 Exponential De ay Law of the Gamow Ve tors

If we are dealing with a s attering system with resonan es, the omplex basis ve tor expan-
sion allows us to isolate the ontribution of ea h resonan e. To illustrate this, we are going
to see how the exponential de ay law holds if only the ontribution of a resonan e is taken
into a ount.
We want to determine the probability P
r0 (t) of dete ting the de aying state within
a shell of width
r0 outside the potential region (r > b). This is the probability that
is measured by the ounting rate of a dete tor pla ed, for example, outside a radioa tive
nu leus from whi h an  parti le is emitted. We assume that the dete tor surrounds the
nu leus ompletely and that is at a distan e r0 > b from the enter r = 0.
Theoreti ally, the probability P
r0 (t) to observe an in-state '+ at time t within the
interval
r0 around the surfa e r = r0 is given by
Z Z
P
r0 (t) = d
r2 drjhr; ; j'+(t)ij2 : (6.7.1)

r0

Experimentally, the probability of nding the de aying state parti le around r0 , that is,
the ounting rate of the dete tor, is not dened for all times t: a resonan e must be rst
prepared before the system an de ay. The time at whi h the preparation of the resonan e
is nished and at whi h the de ay starts an be hosen arbitrarily (we hoose it to be 0). For
example, the  parti le emitted by an -unstable nu leus travels at speed v = R =(2~Im(k))
and rea hes the point r0 at the time t(r0 ) = r0 =v . For times less than t(r0 ), the  parti le
222 6 Gamow Ve tors of the Square Barrier Potential Resonan es

is not there yet, and therefore the ounting rate measured by a dete tor pla ed at r0 is zero
for times t < r0 =v . Whatever would have been ounted by the dete tor before the instant
t(r0 ) at r0 annot be onne ted with the de aying state. Thus the theoreti al probability
to dete t a resonan e at r0 should be zero for t < r0 =v . This is an instan e of the time
asymmetry built into a de aying pro ess.
Experimentally as well, the de ay of unstable systems usually follows the exponential
law ( f. Refs. [44, 45, 46, 47℄).
The Hilbert spa e annot a ommodate either the time asymmetry of P
r0 (t) [89℄ or
the exponential de ay law [90℄. To a ount for these two features, we should use the Rigged
Hilbert Spa e. In the RHS formulation, the Gamow ve tors have an asymmetri time
evolution given by a semigroup e iH  t=~ ( f. Se tion 6.5 above), whi h a ounts for the time
asymmetry of a resonant pro ess. The behavior of the semigroup evolution is in ontrast to
the time-symmetri Hilbert spa e time evolution, whi h is given by a group.
We are going to show that the exponential de ay law holds if we onsider only the
resonan e (Gamow ve tor) ontribution to the probability (6.7.1). In Se tion 6.4, we used
the Gamow ve tors as basis ve tors to expand the normalized in-state '+ in terms of the
ba kground and the resonan e ontribution (see Eq. (6.4.3)). In order to ompute the n-th
resonan e ontribution to the probability (6.7.1), we approximate '+ by the Gamow ve tor
by negle ting the ba kground term and the ontribution of the rest of the resonan es in
(6.4.3),
u (r ; z )
'+ (r; ; ) ' nD (r; ; ) = n n Y0;0 (; ) : (6.7.2)
r
Thus the n-th resonan e ontribution to the probability is
Z Z
P
r0 (t) ' d
r2 drjhr; ; j D ij2
n (t) : (6.7.3)

r 0
The time evolution of the n-th Gamow ve tor is given by
D iH  t=~ D i(En t i n =2)t=~ D ;
n (t) =e n =e n (6.7.4)
and therefore
hr; ; j D i=e i(En i n =2)t=~ un (r ; zn ) Y (; ) :
n ( t) 0;0
r
(6.7.5)
Inserting (6.7.5) into (6.7.3) yields
Z
P
r0 (t) ' je n =(2~)t
j 2 drjun(r; zn)j2
Z
r0

= e n t=~
drj2iNn j2 jei(Re(kn ) iIm(kn ))r 2j

r0
Z r0 +
r0
= e n t=~
j2Nnj 2 dre2Im(kn )r
r0
2Im(kn )
r0
= e n t=~
j2Nnj2e2Im(kn)r0 e 2Im(kn )
1
6.8 Con lusion to Chapter 6 223

' e n t=~ j2Nnj2e2Im(kn )r0
r0 ;

= j2Nnj2
r0 e n =~ (t r0 =vn ) ; t > r0 =vn ; (6.7.6)

where we have used the approximation
r0 small in the next to the last step. Therefore,

P
r0 (t) ' j2Nnj2
r0 e n =~ (t r0 =vn ) ; t > r0 =vn : (6.7.7)

Equation (6.7.7) represents the n-th resonan e ontribution to the ounting rate measured
by a dete tor pla ed at r0 . This ounting rate rea hes its maximum at t = r0 =vn , and
de reases exponentially as time goes on. Therefore, the n-th Gamow ve tor (resonan e)
ontribution to the probability P
r0 (t) follows the exponential de ay law.

6.8 Con lusion to Chapter 6

In this hapter, we have onstru ted the Gamow ve tors of the square barrier potential
resonan es, and studied their properties. We have dened a de aying Gamow ve tor jzn i
as the solution of the integral equation (6.3.1), whereas a growing Gamow ve tor jzn+ i has
been dened as the solution of (6.3.3). We have seen that in the radial representation the
Gamow eigenfun tion hrjzn i satises the S hrodinger equation subje t to a purely outgoing
boundary ondition, whereas the growing Gamow eigenfun tion hrjzn+ i satises a purely
in oming boundary ondition. We have also seen that the purely outgoing boundary on-
dition produ es the same resonan e spe trum as the S -matrix spe trum of Figure 6.2. The
[0; 1)-energy representation of the eigenfun tion hrjzn i has been shown to be the omplex
delta fun tion multiplied by a normalization fa tor An Æ (E zn ). The ( 1; 1)-energy rep-
resentation of the eigenfun tion hrjzn i has been shown to be the Breit-Wigner amplitude
multiplied by a normalization fa tor 1=(2i) An=(E zn ).
The time evolution of the Gamow kets has been shown to be governed by a semigroup,
expressing the time asymmetry built into a resonan e.
The Gamow ve tors have been used as basis ve tors in the omplex basis ve tor ex-
pansions (6.4.3) and (6.4.4). However, they do not form a omplete basis, and therefore
a ontinuous set of Dira kets was added to omplete them. The expansions (6.4.3) and
(6.4.4) extra t the resonan e ontribution out of the in- and out-states, respe tively.
We have un overed the time asymmetry that arises from the purely outgoing boundary
ondition. We have seen that the purely outgoing boundary ondition should read as purely
outgoing only for the de aying part of the resonan e, and as purely in oming for the growing
part of the resonan e.
The exponential law has been shown to hold if the ba kground term of the omplex
basis ve tor expansion is negle ted|only the resonan e (Gamow ve tor) ontribution to
the probability is taken into onsideration.
224 6 Gamow Ve tors of the Square Barrier Potential Resonan es

6.9 Appendix 9: Figures

In this Appendix, we show the graphi s of the square barrier potential resonan e poles, both
in the momentum and energy omplex planes. We also show the ontours that where used
to obtain the omplex basis ve tor expansion.

Im k

Re k

Figure 6.1: The resonan e momenta of the square barrier potential.

6.9 Appendix 9: Figures 225

E (first sheet)

(a)

Sp(H)

E (second sheet)

E (second sheet)

(b)

Sp(H)

E (first sheet)

Figure 6.2: The resonan e energies of the square barrier potential.

226 6 Gamow Ve tors of the Square Barrier Potential Resonan es

E (first sheet)

(a)

Sp(H)

E (second sheet)

E (second sheet)

(b)

Sp(H)

E (first sheet)

Figure 6.3: Deformation of the path of integration into the se ond sheet of the energy
Riemann surfa e; (a) for the de aying states and (b) for the growing states.
Chapter 7
The Time Reversal Operator in the
Rigged Hilbert Spa e
In this hapter, we study the behavior of resonan es under the time reversal operation. We
shall study the standard time reversal operator and also a non-standard one, whi h yields
a doubling of the RHS.

VLADIMIR: We've nothing more to do here.

ESTRAGON: Nor anywhere else.
VLADIMIR: Ah Gogo, don't go like that.
To-morrow everything will be better.
ESTRAGON: How do you make that out?
VLADIMIR: Did you not hear what the hild said?
ESTRAGON: No.
VLADIMIR: He said that Godot was sure to ome
to-morrow. (Pause). What do you say to that?
ESTAGRON: Then all we have to do is to wait here.

Samuel Be kett, Waiting for Godot

227
7.1 Introdu tion 229

7.1 Introdu tion

In the previous hapter, we have seen that the time evolution of the Gamow ve tors is given
by a semigroup and therefore is time asymmetri . We now want to dis uss how this time
asymmetry behaves under the a tion of the time reversal operator in the nonrelativisti
domain [59℄. We shall forget about our beloved square barrier potential and work with a
\general" potential.
We have seen in the previous two hapters that a s attering pro ess should be des ribed
by two RHSs. One RHS orresponds to the states '+ ,
  Ha   ; (7.1.1)
whereas the other RHS orresponds to the observables ,
+  Ha  + : (7.1.2)
In both RHSs, the spa e Ha represents the Hilbert spa e of s attering states of the total
Hamiltonian H (\a " stands for absolutely ontinuous). On Ha , H has absolutely ontin-
uous spe trum only. We shall assume that this spe trum oin ides with the positive real
line.
Suppose the S -matrix has a pair of simple poles at zR = ER i R =2 and at zR =
ER + i R =2, and denote their orresponding Gamow ve tors as jzR i and jzR + i. These
Gamow ve tors have the following properties:
i.) The Gamow ve tors are fun tionals,
jzR i 2 + ; (7.1.3a)
jzR +i 2  : (7.1.3b)
ii.) They are generalized eigenve tors of the total Hamiltonian H ,
H  jzR i = zR jzR i ; (7.1.4a)
H  jzR + i = zR jzR + i : (7.1.4b)
iii.) The time evolution operator e iHt an be ontinuously extended to + for positive
values of time and to  for negative values of time. The ontinuity of the extensions refers
to the weak topology [93℄. In addition we have:
e iH  t
jzR i = e izR tjzR i = e iERte R t=2 jzR i ; for t > 0 ; (7.1.5a)
eiH  t jz  + i = e izR t jz  + i = e iER t e R t=2 jz  + i ; for t < 0 : (7.1.5b)
R R R

The a tion of e iH  t on jzR i for t < 0 and on jzR + i for t > 0 is, however, not dened. Thus
the Hilbert spa e group evolution splits into two semigroups. This splitting is a onsequen e
of the hoi e of  and the properties of Hardy fun tions. The hoi e of Hardy fun tions is
related to a ausality ondition, and therefore the splitting is also related to ausality. The
230 7 Time Reversal Operator in the Rigged Hilbert Spa e

splitting of the group of evolution into two semigroups shows the irreversible hara ter of a
resonan e [52, 54, 55℄.
In his study of the representations of the Poin are group extended by time reversal and
parity, Wigner found four dierent possibilities ( f. Table I of Appendix 7.4 and Refs. [61,
62℄). The rst possibility is the standard one, but the other three imply a doubling of the
spa e that supports the representation. J. F. Cari~nena and M. Santander onstru ted the
proje tive representations of the Galilei group extended by time inversion and parity [63℄.
They also found four possibilities for the ase with mass. As for the Poi are group, the
standard ase does not yield a doubling of the spa e that supports the representation,
while the other three do yield a doubling. Bohm [64℄ has studied the latter time reversal
(T = I = 1) in the relativisti ase, whi h yields a doubling. One of our goals is to
onstru t an analog to this doubling in the non-relativisti ase for s-waves (j = 0).
In the next se tion, we dis uss the ee t of the time reversal operator on Gamow ve -
tors in the standard ase. This standard ase is labeled by T = I = 1 (see Table I of
Appendix 7.4). In Se tion 7.3, we study the the ase T = I = 1 (see Table I of Ap-
pendix 7.4) and present the idea of time reversal doubling. In Appendix 7.4, we review some
general aspe ts of the time reversal operation.

7.2 The Standard Time Reversal Operator (T = I =

1)
In this se tion, we present the ee t of the standard time reversal operation on s attering
systems having (simple pole) resonan es. The notation we are using here does not dier
essentially from that in Ref. [59℄, although there is a ouple of dieren es:
i.) The restri tion to the positive real semiaxis of interse tions of Hardy spa es with the
S hwartz spa e are denoted here by

H2 \ S R+ : (7.2.1)
The plus sign stands for Hardy fun tions on the upper half plane and the minus sign for
Hardy fun tions on the lower half plane. S is the S hwartz spa e on the real line.
ii.) The extension of the evolution operator e iHt to the spa e + is denoted as U+ (t)
for t > 0, and the extension of e iHt to  is denoted as U  (t) for t < 0. These are the two
semigroups dis ussed in the introdu tion.
For simpli ity, we shall work with a spheri ally symmetri potential and onsider parti les
without spin or any other possible degrees of freedom. We restri t ourselves to zero values
of the angular momentum, and denote the orresponding Hilbert spa e by H0 . If the system
does not have bound states, the spa e H0 oin ide with Ha .
Let us re all that the unitary operators U diagonalize the total Hamiltonian H (or its
restri tion to its absolutely ontinuous spa e Ha , if H has bound states), in the sense that
these operators give a unitary equivalen e between H and the multipli ation operator on
L2 (R + ; dE ). They are a produ t of the inverses of the Mller operators times the operator
U0 that diagonalizes the free Hamiltonian, U = U0
y .
7.2 The Standard Time Reversal Operator (T = I = 1) 231

For s waves (j = 0), the standard hoi e of the time reversal operation is T = I = 1 (see
Table I of Appendix 7.4). Therefore, in the energy representation the time reversal operator
AT a ts as the omplex onjugation C . Sin e the mapping C transforms any fun tion of E
into its omplex onjugate, we have
 
C : H2 \ S R+ 7 ! H2 \ S R+ : (7.2.2)
Moreover, one an show that this map is ontinuous.
Our next goal is to dene time reversal operators AT  on  . These operators should
be equivalent to C and the equivalen e should be given by U . Their denition is:
AT  := Uy C U : (7.2.3)
This denition makes the following diagram:
 
H2 \ S R+ C
! H2 \ S R+
Uy # # Uy (7.2.4)

 ! 
AT 
These operators have the following properties:
1. AT  are ontinuous antilinear mappings from  onto  .
2. They an be extended to ( ontinuous) antiunitary mappings from Ha onto itself.
3. Their adjoints are given by
h iy
y y
AT  = U C U = Uy C U = AT  ; (7.2.5)
that is, they are adjoint to ea h other.
4. They are inverse to ea h other,
AT + AT = U+y CU U y CU+ = I ; on  ; (7.2.6)
AT AT + = U y CU+ U+y CU = I ; on + : (7.2.7)

Consider now a densely dened ontinuous antilinear operator A on H with the following
property: there are two RHSs   H   and  H   su h that Ay maps
ontinuously  into . Then, A an be extended by ontinuity to  using the following
formula:
h jAF i := hAy jF i = hF jAy i ; 8F 2  ; 8 2 : (7.2.8)
Thus A is a weak ontinuous antilinear mapping from  into  . The proof of this goes
exa tly as the proof for the linear ase [93℄.
232 7 Time Reversal Operator in the Rigged Hilbert Spa e

It is straightforward to apply this denition to the time reversal operator, after making
the identi ation  =  , =  , A = AT  , and Ay = AT  . Thus, we have the following
ontinuous antilinear extensions:
AT  :  7 !  : (7.2.9)
These extensions are one-to-one, onto mappings with ontinuous inverses, and they indeed
extend AT  as originally dened in (7.2.3).
We now obtain the images of Lippmann-S hwinger kets and of the Gamow ve tors by
AT  . To this end, let us onsider two arbitrary ve tors ' 2  .1 Their wave fun tions in

the energy representation are given by

'b (E ) = h E j' i = (U ' )(E ) 2 H2 \ S R+ : (7.2.10)
Using the denition of AT  we obtain
h'jAT jE i = hE jAyT 'i = C 'b(E ) = 'b(E ) = h'jE i ; (7.2.11)
that is,
AT  jE  i = jE i : (7.2.12)
Take now the Gamow ve tors jzR i and jzR + i. Then
h'+jAT +jzR i = h zR jAyT +'+i = C 'b+(zR ) = 'b+(zR ) = h'+jzR+i ; (7.2.13)
that is,
AT + jzR i = jzR + i : (7.2.14)
Analogously
AT jzR + i = jzR i : (7.2.15)
Next, we study the a tion of the standard time reversal operator on the time evolution
semigroups. We know that eiHt +  + if t > 0. Then,
AT + eiHt AT '+ = U+y CU eiHt U y CU+ '+
= U+y CeitE CU+ '+
= U+y e itE U+ '+
= ei( t)H '+ ; '+ 2  ; t > 0 : (7.2.16)
Analogously, if t < 0, we have that eiHt   . Then
AT eiHt AT + ' = ei( t)H ' ; ' 2 + ; t < 0 : (7.2.17)
Therefore,
AT + U+y (t)AT = U y ( t) ; t > 0 ; (7.2.18a)
AT U y (t)AT + = U+y ( t) ; t < 0 : (7.2.18b)
1
In this hapter, we shall denote the observables by ' in order not to repeat the formulas twi e.
7.2 The Standard Time Reversal Operator (T = I = 1) 233

We see that the operators AT  transform one semigroup into the other. The extension of
these formulas to the dual spa es yields

AT + U+ (t)AT = U  ( t) ; t > 0 ; (7.2.19a)

AT U  (t)AT + = U+ ( t) ; t < 0 : (7.2.19b)

One ould expe t that the operators AT  are the same operator restri ted to dierent
subdomains. That is true. As a matter of fa t, their extensions to Ha oin ide:
The proof of this statement is rather simple. Write
A2T + = U+y CU y
U+ CU : (7.2.20)
Sin e U = U0
y, one has
A2T + =
+U0yCU0
y
+U0yCU0
y : (7.2.21)
The S operator is given by
S =
y
+ ; (7.2.22)
and its adjoint is given by
Sy =
y+
: (7.2.23)
Sin e
U0 SU0 1 = U0SU0y = S (E ) (= S (E + i0) ; E> 0) ; (7.2.24)
we have that
y = (U
U0 S y U0
y)y = S (E ) : (7.2.25)
0 SU0

Plugging Eqs. (7.2.22)-(7.2.25) into (7.2.21) we obtain

A2T + =
+U0yCU0SU0yCU0
y
=
+U0yCS (E )CU0
y
=
+U0yS (E )U0
y
=
+S y
y
=
+
y+

y
= I; (7.2.26)
where I is the identity on Ha . The same is true for AT . Therefore AT + is an
invertible bounded operator su h that AT + = AT +1 on Ha . Sin e AT +1 and AT
oin ide on the dense subspa e  , they are equal on H0, and we have that AT =
1 =A .
AT + T+
234 7 Time Reversal Operator in the Rigged Hilbert Spa e

7.3 The Time Reversal Doubling (T = I = 1)

In this se tion, we present the onstru tion of the time reversal doubling. A ording to
Wigner [61, 62℄, there are three other possible representations of the Poin are group extended
by time reversal and parity besides the (standard) one of the previous se tion. All four
possibilities are listed in Table I of Appendix 7.4.
Let us onsider the following pair of RHSs:
 

H2 \ S R+
C 2  L2 (R + )
C 2  H2 \ S R+ 
C 2 ; (7.3.1)
where C 2 denotes the two-dimensional linear spa e of olumn ve tors whose entries are
omplex numbers. The elements of ea h spa e of the triplet (7.3.1) an be expressed as
two-dimensional ve tors whose entries belong to the spa e in the left hand side of the tensor
produ t. In the ase T = I = 1, the time reversal operator in the energy representation
is dened by ( f. Table I of Appendix 7.4)
 
C := 0 C ; (7.3.2)
C 0
where C denotes the omplex onjugation.
 This operator is antilinear and ontinuous from
H \ S R+
C onto H \ S R+
C .
By duality, it an be extended
2  2 2  2


to a ontinuous

antilinear mapping C from H \ S R+
C onto H \ S R+
C .
 2  2 2  2
Ea h spa e H2 \ S R+
C 2 has two distinguished subspa es,
 

0 ;

 = H \ S
+ 2  + (7.3.3a)
R
 
2

 0
 = H \ S R+
 ; (7.3.3b)
where  and  are arbitrary omplex numbers. We have, therefore, two new RHSs that an
be written in the following form:
  L2 (R + )  () ; (7.3.4)
where
 
L2+ (R + ) = L2 (R + )
 ; (7.3.5a)
0
 
L2 (R + ) = L2 (R + )
0 : (7.3.5b)

The dual of the spa es (7.3.3) an be written as
 
(+ ) = H \ S 2 

 +
 ; (7.3.6a)
 R 0
 
( ) = H \ S 2 

 +
0 : (7.3.6b)
R 
7.3 The Time Reversal Doubling (T = I = 1) 235

It is not diÆ ult to show that C is a ontinuous antilinear bije tion from  onto  ,
C  =  : (7.3.7)
Thus C an be ontinuously extended to the dual spa es,
C ( ) = () : (7.3.8)
As an operator on Ha
C 2 , the square of C is proportional to the identity,
C 2 = I = T I ; (7.3.9)
where I represents the identity on Ha
C 2 .
In order to larify the notation, we repla e the supers ript signs by r, with r = +; .
That is, we shall write r , and so on. This notation makes it lear that the signs above are
independent of the signs below.
Let us dene the operators
U := U
I ; (7.3.10)
where I is the identity on C 2 . We an write these operators as
 
U = U 0 ; (7.3.11)
0 U
and their adjoints as  
Uy = U1 = Uy 0 : (7.3.12)
0 Uy
It is lear that U maps Ha
C 2 onto L2 (R + )
C 2 . Using those operators, we an dene
the following spa es:
r := Uy r : (7.3.13)
Clearly, the spa es r are subspa es of Ha
C 2 . It is obvious that
 
r=+ = 
 ; (7.3.14a)
0
 
r= = 
0 : (7.3.14b)

The operators U and their respe tive inverses Uy an be ontinuously extended to the dual
spa es.
We are now in a position to introdu e the time reversal operators for our T = I = 1
hoi e. They an be dened as
AT  := Uy C U : (7.3.15)
These two operators have similar properties to those satised by AT  . We list here these
properties without proofs, sin e these proofs do not dier mu h from those for AT  :
236 7 Time Reversal Operator in the Rigged Hilbert Spa e

1. AT  are ontinuous antilinear mappings from r onto r , respe tively.They an
be ontinuously extended to antilinear mappings between the respe tive duals.
2. They are adjoint to ea h other,
AyT  = AT  : (7.3.16)
3. As operators on Ha
C 2 , they are antiunitary. In addition, they are inverse to ea h
other,
AT +AT = I ; (7.3.17a)
AT AT + = I : (7.3.17b)

4. On the Hilbert spa e Ha

C 2 , we have that
A2T  = I =) AT AT +AT + = AT =) AT + = AT : (7.3.18)

Formulas (7.3.16)-(7.3.18) are a onsequen e of the denition hosen in (7.3.15) for AT

(with minus sign), whi h has its origin in the fa t that C y = C . If we redened AT
without the minus sign in (7.3.15), we would have
AyT  = AT  ; (7.3.19a)
AT +AT = I; (7.3.19b)
AT AT + = I; (7.3.19 )
AT + = AT : (7.3.19d)
We an hoose either (7.3.16)-(7.3.18) or (7.3.19). The hoi e (7.3.19) has the advantage of
having a unique time reversal operator, and the distin tion between AT + and AT indi ates
the restri tion of a unique time reversal operator to dierent subspa es r . As we shall see
later, this hoi e has the disadvantage of leading to the appearan e of a minus sign in the
formulas of the a tion of the time reversal operator on the semigroups.
The importan e of the above onstru tion lies on the possibility of extending the time
reversal operator to the dual spa es, whi h ontain the Lippmann-S hwinger kets and the
Gamow ve tors.
We dene the Lippmann-S hwinger kets in the T = I = 1 ase as
 
jE ; r = +i = jE i 2 (r=+) ; (7.3.20a)
0 
 
jE  ; r = i= 0 2 (r= ) : (7.3.20b)
jE  i
The kets in Eqs. (7.3.20) are generalized eigenve tors of the operator H
I (H is the exa t
Hamiltonian, and I is the identity on C 2 ) with generalized eigenvalue E > 0. We now
7.3 The Time Reversal Doubling (T = I = 1) 237

determine the a tion of AT  on those eigenkets. We start with the following denition,
whi h has its origin in (7.2.8):
hjAT jE ; ri = hE ; rjAyT i ;  2 r : (7.3.21)
From (7.3.15) we obtain  
AT  =  A 0 AT
: (7.3.22)
T 0
Let us write  
 = ' ; (7.3.23)

where ' ;  2  . We shall study separately the ases r = . Take rst r = +. Then
 y  
 y
h E ; r = +jAT  i = (h E j; 0) ATy '
  A
T
 y
= h E jAT  i 
= h  jAT  jE  i
= h  jE  i
 
= (' ; ) jE  i
  0

= h jE  ; r = i : (7.3.24)
This and Eq. (7.3.21) yield
AT jE ; r = +i = jE  ; r = i : (7.3.25)
The  signs appear as the oeÆ ient of jE  ; r = i in Eq. (7.3.25) only if we make the
hoi e AT = Uy C U+ . The hoi e AT = Uy C U+ repla es the  signs in (7.3.25) by plus.
Now take r = . An analogous al ulation to (7.3.24) yields
AT jE ; r = i = jE ; r = +i ; (7.3.26)
where the  signs have the same origin as in the ase r = +. The hoi e AT = Uy C U+
repla es them by minus.
The next step is to dene the Gamow ve tors and to obtain their images under time
reversal. The Gamow ve tors jzR i and jzR+ i an be used to dene the following Gamow
ve tors for our T = I = 1 ase:
 
j
jzR ; r = +i := 0R ; z i (7.3.27a)
 
0
jzR ; r = i := jz i ; (7.3.27b)
R
 + 
jzR ; r = +i := jzR0 i ;
+ (7.3.27 )
 

jzR ; r = i := jz+i :
+ 0 (7.3.27d)
R
238 7 Time Reversal Operator in the Rigged Hilbert Spa e

The Gamow ve tors (7.3.27a) and (7.3.27b) are generalized eigenve tors of the operator
H
I with generalized eigenvalue zR , while the Gamow ve tors (7.3.27 ) and (7.3.27d) are
generalized eigenve tors of H
I with generalized eigenvalue zR . One an also show that
AT +jzR ; r = +i = jzR+ ; r = i ; (7.3.28a)
AT +jzR ; r = i = jzR+ ; r = +i ; (7.3.28b)
AT jzR+ ; r = +i = jzR ; r = i ; (7.3.28 )
AT jzR+; r = i = jzR ; r = +i : (7.3.28d)
The overall signs on the right hand side of Eqs. (7.3.28 ) and (7.3.28d) orrespond to the
hoi e
AT = Uy+ C U : (7.3.29)
For the hoi e
AT = Uy+ C U ; (7.3.30)
the overall signs on the right hand side of Eqs. (7.3.28 ) and (7.3.28d) are the opposite.
Now, we obtain the a tion of the time reversal operator on the time evolution semigroups.
The time evolution semigroups are dened on the dual spa es (r ) as
 
W(t) := U(t)
I = U(t) 0
: (7.3.31)
0 U (t)


The operator W+ (t) is well dened on (r+ ) for t > 0 only, while W  (t) is well dened on
(r ) for t < 0 only. From Eqs. (7.2.19) it follows that
AT +W+ (t)AT = W ( t) ; t > 0 ; (7.3.32a)
AT W  (t)AT + = W+ ( t) ; t < 0 : (7.3.32b)
This result has been obtained for the hoi e of AT as in (7.3.29). If we made the hoi e
(7.3.30), a minus sign would appear on the right hand side of Eqs. (7.3.32). As mentioned
above, we prefer the hoi e (7.3.29), be ause we want Eqs. (7.3.32) to not have that minus
sign.

7.4 Appendix 10: Time Reversal

Textbooks on Quantum Me hani s usually dene the time reversal operation in the position
representation as
C (~x; t) =  (~x; t) ; (7.4.1)
where the star denotes omplex onjugation. We are going to explain what this denition
means.
Following Wigner, time reversal is an operation su h that the following operations, when
performed sequentially, yield the identity:
7.4 Appendix 10: Time Reversal 239

time displa ement by t  time reversal  time displa ement by t  time reversal : (7.4.2)
If we denote the time reversal operator by C , a possible denition would be C (t) = ( t).
However, this kind of operation is obviously linear. The need for an antilinear time reversal
operation has been ni ely shown by Wigner in the following terms:
Consider a system whose Hamiltonian has a omplete set of eigenve tors 'n (for
instan e, the Harmoni os illator, the bound states of the Hydrogen atom, or any
system formed by the bound states of the Hamiltonian, if any). Then, any state
ve tor ' an be expanded by those eigenve tors,
' =
X an 'n ; (7.4.3)
n

where
H'n = E n 'n : (7.4.4)
The operations (7.4.2) yield the identity if and only if:
time displa ement by t  time reversal = time reversal  time displa ement by t :
(7.4.5)
Let us apply these operations to ' in (7.4.3). If we assume that the time reversal
operator C is linear, then
C' =
X
an C'n : (7.4.6)
n
Sin e [H; C ℄ = 0, C'n is also an eigenve tor of the Hamiltonian with the same eigen-
value En. Therefore, time displa ement by t on (7.4.6) gives
X an e iEn t C'n : (7.4.7)
n

A ording to the rule in (7.4.5), this should be equal to the result of performing rst
the time displa ement by t on '
X an eiEn t 'n ; (7.4.8)
n

and then the time reversal operation C , whi h (assuming that C is linear) leads to
X an eiEn t C'n : (7.4.9)
n

This result does not oin ide with the expression given by (7.4.7). However, they do
oin ide if C is dened as an antilinear operator.
240 7 Time Reversal Operator in the Rigged Hilbert Spa e

On e we have shown that the time reversal operator must be antilinear, we study its
a tion in the energy representation. In this representation, the Hamiltonian H a ts as the
multipli ation operator. If (E ) is a wave fun tion in the energy representation, then the
a tion of the time reversal operator on it is dened by
C (E ) :=  (E ) ; (7.4.10)
where we have hosen the omplex onjugation as the time reversal operator (as Wigner
does). Time displa ement by t on (7.4.10) gives
e iEt  (E ) : (7.4.11)
If we apply time reversal to (7.4.11), whi h is now equivalent to perform the omplex
onjugation operation, we obtain
eiEt (E ) : (7.4.12)
Finally, time displa ement by t on (7.4.12) gives
e iEt eiEt (E ) = (E ) : (7.4.13)
Hen e, the time reversal operator C dened by (7.4.10) fullls the above rule (7.4.2) and is
antilinear. Obviously,
C ( (E; t)) = C (e iEt (E )) = eiEt  (E ) : (7.4.14)
We an look at eiEt  (E ) as the result of applying time displa ement by t on  (E ).
Therefore, eiEt  (E ) is what should be identied with the  (E; t) (or  t (E )) that
appears in the literature. Note that  (E; t) = eiEt  (E ) = [e iEt (E )℄ = [ (E; t)℄ .
The same an be argued in the position representation, where time reversal is given by
Ce (~x) =  (~x) : (7.4.15)
However, in the momentum representation, the time reversal operator, whi h we denote
by C 0 , a ts as
C 0 '(p~) = ['( p~)℄ ; (7.4.16)
sin e the time reversal hanges ~p for p~.
In order to show it, let (~x) be a wave fun tion in the position representation. Then
the orresponding wave fun tion in the momentum representation is given by
b (p~)  F (p~) := 21
1 Z
e i~p~x (~
x) d~
x: (7.4.17)
1
where F and hat denote the Fourier transform. The Fourier transform takes the time
reversal operator in the position representation into the time reversal operator in the
momentum representation,
C0 = FCF 1 :e (7.4.18)
7.4 Appendix 10: Time Reversal 241

Hen e,
b
(C 0 )(p~) = e b
(F C F 1 )(p~)
= eZ
(F C )(p~)
= e
1 1 e i~p~x(C )(~x) d~x
2 1
= 21
1 Z
e i~p~x  (~
x) d~
x

b 1
= [ ( p~)℄ ; (7.4.19)
whi h proves Eq. (7.4.16).
Consider now an arbitrary representation supported by the Hilbert spa e H. Let U be
the unitary operator that transforms from the position into that arbitrary representation.
Analogously to (7.4.18), we dene the time reversal operator AT on H as
e 1   U C
AT  := U CU e (~
x) : (7.4.20)
If we denote T := AT , (t) := e itH , and all H 0 = U 1 HU the Hamiltonian in the
position representation, then
AT (t) = AT e iHt 
= e 1 e iHt 
U CU
= e 1 e iHt U(~
U CU x)
0
e iH t (~
= U Ce x)
iH 0t 
= Ue  (~x)
= eiHt U C
e (~x)
= iHt
e U CU e 1
= eiHt AT 
= (AT )( t)
= T ( t) : (7.4.21)
Thus AT (t) = T ( t), whi h generalizes the equation C (~x; t) =  (~x; t).
However, this is not the whole story. As mentioned above, Wigner [61, 62℄ realized that,
when onstru ting proje tive representations of the Poin are group extended by time inver-
sion and parity, new possibilities exist. These new possibilities are not independent of the
representation of the parity and imply a doubling of the spa e supporting the representa-
tion. We do not want to dis uss this onstru tion here. Instead, we present a table with
the four possibilities (see Table I below). The four possibilities are hara terized by two
parameters, whi h also appear among the parameters that hara terize the representations
of the extended Poin are group.
Consider the spa e, time, and total inversion operators on a Hilbert spa e H, whi h we
denote respe tively by UP , AT and AI . The operator UP is unitary, while AT and AI are
242 7 Time Reversal Operator in the Rigged Hilbert Spa e

antiunitary. From the nature of the orresponding physi al operations, it follows that the
operators UP2 , A2T , and A2I must be proportional to the unit operator. Sin e UP is unitary,
we an hoose its phase su h that
UP2 = I ; (7.4.22)
while su h a normalization is not possible for the antiunitary AT or AI . In fa t, the antiu-
nitarity and the asso iative law of the group multipli ation di tate that the squares of AT
and AI must ne essarily equal either +1 or 1:
A2T = T I ; T = 1 ; (7.4.23a)
A2I = I I ; I = 1 : (7.4.23b)
Moreover, the phase of AI an be hosen su h that
AI = UP AT : (7.4.24)
Corresponding to the values that T and I an take, there exist four extensions of the
ontinuous symmetry group. Barring the existen e of any additional onditions, all four are
possible, and Wigner [61℄ has derived these four lasses of proje tive representations of the
extended Poin are group. The results are summarized in the following table:2
Table I
T I UP AT

( 1)2j ( 1)2j 1 C
   
( 1)2j ( 1)2j 1 0 0 C
0 1 C 0
   
( 1)2j ( 1)2j 1 0 0 C
0 1 C 0
   
( 1)2j ( 1)2j 1 0 0 C
0 1 C 0

In this table, j refers to the spin of the parti le under onsideration, while C is the well
known (2j + 1)-dimensional matrix whose entries are ; = ( 1)j + Æ;  , j  ;   j .
In these representations, the ontinuous spa e-time transformations Ug , as well as any other
known observables su h as the internal symmetry generators B , have the following form:
   
Ug = Ug 0 ; B= B 0 : (7.4.25)
0 Ug 0 B
2
In the non-relativisti ase, J. F. Cari~nena and M. Santander have obtained a totally analogous result
for the Galilei group (in the ase with mass) [63℄.
7.4 Appendix 10: Time Reversal 243

The representation spa e of the extensions of the spa e-time symmetry group by P and
T is therefore redu ible under the restri ted symmetry transformations and observables.
From Table I we see that only the ase for whi h T = I = ( 1)2j leads to no doubling
of the spa e of the mi ros opi system under inversions. This is the only ase dis ussed in
relativisti quantum eld theory, and quantum elds have so far been onstru ted only for
this lass of the four lasses of proje tive representations of the extended Poin are group [94℄.
In the three other ases, the restri ted spa e-time symmetry transformation is doubled
after the time reversal operator is adjoined|the time reversal doubling. In these ases,
the two subspa es that are left invariant under Ug and B remain invariant also under UP ,
albeit they have opposite relative parity in the two ases for T I = 1. In these two
ases, the two subspa es an in fa t be distinguished by their parity eigenvalue, and to
the extent that we asso iate mi ros opi systems with irredu ible representations of the
symmetry group, the two subspa es would des ribe parti les with the same mass and spin
but opposite parity. This, however, is not the situation for the extended group hara terized
by T = I = ( 1)2j for whi h the relative parity of the two subspa es of states is +1. It is
this lass of proje tive representations whi h have been used in Ref. [64℄, be ause for these
representations a label r an be introdu ed, a two-valued parameter, whi h was used in
Ref. [64℄ to distinguish between the spa e of prepared states and the spa e of time reversed
registered observables. These two subspa es (have the same parity and) remain irredu ible
under Ug and B . From Table I it is lear that AT hanges the value r.
Chapter 8
Con lusions
In this last hapter, we present the on lusions of the dissertation.

There is never an ending to Paris and the memory of ea h

person who has lived in it diers from that of any other. We
always returned to it no matter who we were or how it was
hanged or with what diÆ ulties, or ease, it ould be rea hed.
Paris was always worth it and you re eived return for what-
ever you brought to it. But this is how Paris was in the early
days when we were very poor and very happy.

Ernest Hemingway, A movable Feast

245
Con lusions 247

The RHS language has been used to des ribe Dira kets, Lippmann-S hwinger kets and
Gamow ve tors in a onsistent way. We have seen that the mathemati al image of those
obje ts should be the following:

Physi al quantity Mathemati al image Symbol Name

Bound state of Normalizable eigenve tor of H jEn ) Bound state
energy En < 0 with eigenvalue En
S attering state of Generalized eigenve tor of H jE + i Lippmann-S hwinger
energy E > 0 with real eigenvalue E ket
Resonan e of energy Generalized eigenve tor of H jzR i Gamow ket
ER and width R with omplex eigenvalue zR

In terms of results, we would like to highlight the following:

 We have presented a systemati review of the mathemati al methods of the Rigged
Hilbert Spa e.
 We have reviewed and improved the onstru tion of the RHS of the harmoni os illator.
 We have shown that the natural framework for the solutions of the S hrodinger equation
is the RHS. We have illustrated this point by onstru ting a RHS of the square barrier
potential expli itly.
 We have shown that a onsistent des ription of the Lippmann-S hwinger equations
needs a pair of RHSs. We have also shown that the Lippmann-S hwinger kets a t as anti-
linear fun tionals over spa es of wave fun tions that are boundary values of fun tions that
an be ontinued analyti ally.
 We have translated A. Mondragon et al.'s integral equation for the Gamow ve tors
into the RHS language.
 We have onstru ted the Gamow ve tors of the square barrier potential Hamiltonian.
We have shown that the [0; 1)-energy representation of these ve tors is the omplex delta
fun tional, and that their ( 1; 1)-energy representation is the Breit-Wigner amplitude.
 We have dis losed the time asymmetry of the purely outgoing boundary ondition.
 We have studied the a tion of the time reversal operator on resonan es for the stan-
dard ase and for one of the ases that lead to a doubling of the spa e supporting the
representation. The doubling has been expli itly onstru ted within the RHS framework.
Bibliography
[1℄ P. A. M. Dira , The prin iples of Quantum Me hani s, 3rd ed., Clarendon Press, Oxford
(1947).
[2℄ J. von Neumann, Mathematis he Grundlagen der Quantentheorie, Springer, Berlin
(1931); English translation by R. T. Beyer, Prin eton University Press, Prin eton
(1955).
[3℄ L. S hwartz, Theory de Distributions, Hermann, Paris (1950).
[4℄ J. von Neumann, Ann. Math. (N. Y.) 50, 401 (1949).
[5℄ I. M. Gelfand, N. Y. Vilenkin, Generalized Fun tions, Vol. IV, A ademi Press, New
York (1964); K. Maurin, Generalized Eigenfun tion Expansions and Unitary Represen-
tations of Topologi al Groups, Polish S ienti Publishers, Warsaw (1968).
[6℄ L. C. Mejlbo, Math. S and. 13, 129 (1963).
[7℄ P. Kristensen, L. Mejlbo, E. Thue Poulsen, Commun. Math. Phys. 1 175 (1965);
Math. S and. 14 129 (1964).
[8℄ J.-P. Antoine, J. Math. Phys. 10, 53 (1969); J. Math. Phys. 10, 2276 (1969).
[9℄ A. Bohm, Rigged Hilbert Spa es, International Center for Theoreti al Physi s Le ture
Notes, Publi ation ICTP 64/9, Trieste (1964); Boulder Le tures in Theoreti al Physi s,
1966, Volume 9A, Gordon and Brea h, New York (1967).
[10℄ J. E. Roberts, J. Math. Phys. 7, 1097 (1966); Commun. Math. Phys. 3, 98 (1966).
[11℄ O. Melsheimer, J. Math. Phys. 15, 902; 917 (1974).
[12℄ D. Atkinson, P. W. Johnson, Quantum Field Theory, Rinton Press, Prin eton (2002).
[13℄ N. N. Bogolubov, A. A. Logunov, I. T. Todorov, Introdu tion to Axiomati Quantum
Field Theory, Benjamin, Reading, Massa husetts (1975).
[14℄ L. E. Ballentine, Quantum Me hani s, Prenti e-Hall International, In ., Englewood
Clis, New Jersey (1990).

249
250 Bibliography

[15℄ A. Bohm, Quantum Me hani s: Foundations and Appli ations, Springer-Verlag, New
York (1994).
[16℄ A. Z. Capri, Nonrelativisti Quantum Me hani s, Benjamin, Menlo Park, California
(1985).
[17℄ D. A. Dubin, M. A. Hennings, Quantum Me hani s, Algebras and Distributions, Long-
man, Harlow (1990).
[18℄ A. Galindo, P. Pas ual, Me ani a Cuanti a, Universidad-Manuales, Eudema (1989);
English translation by J. D. Gar a and L. Alvarez-Gaume, Springer-Verlag (1990).
[19℄ A. Bohm, J. Math. Phys. 22, 2813 (1981).
[20℄ A. Bohm, M. Gadella, Dira Kets, Gamow Ve tors and Gelfand Triplets, Springer
Le ture Notes in Physi s 348, Berlin (1989).
[21℄ C. G. Bollini, O. Civitarese, A. L. De Paoli, M. C. Ro a, J. Math. Phys. 37, 4235
(1996).
[22℄ J.-P. Antoine, Quantum Me hani s Beyond Hilbert Spa e, in Irreversibility and Causal-
ity, A. Bohm, H.-D. Doebner, P. Kielanowski [Eds.℄, Springer-Verlag (1998), page 3.
[23℄ I. Antoniou, S. Tasaki, Int. J. Quant. Chem. 44, 425 (1993).
[24℄ Z. Su hane ki, I. Antoniou, S. Tasaki, O. F. Brandtlow, J. Math. Phys. 37, 5837 (1996).
[25℄ R. de la Madrid, A. Bohm, M. Gadella, Forts hr. Phys. 50, 185 (2002); quant-
ph/0109154.

[26℄ R. de la Madrid, J. Phys. A: Math. Gen. 35, 319 (2002); quant-ph/0110165.

[27℄ R. de la Madrid, M. Gadella, Am. J. Phys. 70, 626 (2002); quant-ph/0201091.
[28℄ A. Galindo, Private Dis ussion.
[29℄ A. Bohm, The Rigged Hilbert Spa e and Quantum Me hani s, Le ture Notes in Physi s,
vol. 78, Springer, New York (1978).
[30℄ N. Dunford, J. S hwartz, Linear operators, vol. II., Inters ien e Publishers, New York
(1963).
[31℄ A. R. Bohm, R. de la Madrid, B. A. Tay, P. Kielanowski, Time Asymmetri Boundary
Conditions and the Denition of Mass and Width for Relativisti Resonan es, hep-
th/0101121.

[32℄ R. de la Madrid, The Importan e of Boundary Conditions in Quantum Me hani s, in

\Irreversible Quantum Dynami s", F. Benatti, R. Floreanini [Eds.℄, Springer Le ture
Notes in Physi s (to appear); quant-ph/0302184.
Bibliography 251

[33℄ V. E. Weisskopf, E. P. Wigner, Z. f. Phys. 63, 54 (1930); 65, 18 (1930).

[34℄ K. Gottfried, Quantum Me hani s, Volume I: Fundamentals, Benjamin, New York,
(1966).
[35℄ Hany Kamel Halim Kaldas, Relativisti Gamow Ve tors: State Ve tors for Unstable
Parti les, PhD Thesis, The University of Texas at Austin (2000).
[36℄ D. E. Groom et al., \Review of Parti le Physi s," The European Physi al Journal C15,
1 (2000).
[37℄ M. Gell-Mann, Y. Ne'eman, The Eightfold Way, W. A. Benjamin, In ., New York
(1964).
[38℄ G. Gamow, Z. Phys., 51, 204 (1928).
[39℄ R. E. Peierls, Pro . R. So . (London) A 253, 16 (1959).
[40℄ E. Hernandez and A. Mondragon, Phys. Rev. C 29, 722 (1984); A. Mondragon,
E. Hernandez, and J. M. Velazquez Ar os, Ann. Phys. (Leipzig) 48, 503 (1991).
[41℄ M. Baldo, L. S. Ferreira, and L. Streit, Phys. Rev. C 36, 1743 (1987). L. S. Ferreira,
E. Maglione, R. J. Liotta, Phys. Rev. Lett. 78, 1640 (1997).
[42℄ G. Gar a-Calderon and R. Peierls, Nu l. Phys. A 265, 443 (1976).
[43℄ T. Berggren, Nu l. Phys. A 109, 265 (1968).
[44℄ A. A. Rollesfon and R. M. Prior, Am. J. Phys. 46, 1007 (1978).
[45℄ F. W. Supon and J. J. Kraushaar, Am. J. Phys. 51, 761 (1983).
[46℄ D. Desmarais and J. L. Duggan, Am. J. Phys. 58, 1079 (1990).
[47℄ K. Ruddi k, Am. J. Phys. 63, 658 (1995).
[48℄ A. F. J. Siegert, Phys. Rev. 56, 750, (1939).
[49℄ G. Lindblad and B. Nagel, Ann. Inst. Heri Poin are, XIII, 27 (1970).
[50℄ G. Breit, E. P. Wigner, Phys. Rev. 49, 519 (1936).
[51℄ V. A. Fo k, Fundamentals of Quantum Me hani s, USSR Publishers, Mos ow (1931);
tranlated from Russian by E. Yankovsky.
[52℄ L. Fonda, G. C. Ghirardi, A. Rimini, Rep. Prog. Phys. 41, 587 (1978).
[53℄ C. van Winter, J. Math. Anal., 47, 633 (1974).
252 Bibliography

[54℄ C. Cohen-Tannoudji, B. Diu, F. Laloe, Quantum Me hani s, Wiley, New York (1977).
[55℄ M. L. Goldberger, K. M. Watson, Collision Theory, Wiley, New York (1964).
[56℄ M. Gell-Mann, J. B. Hartle, p. 311 in Physi al Origins of Time Asymmetry, J. J. Hal-
liwell, et al. [Eds.℄, Cambridge University (1994).
[57℄ R. Haag, Comm. Math. Phys. 132, 245 (1990), and Le tures at the Max Born Sympo-
sium on \Quantum Future," Przieka (1997).
[58℄ A. Bohm, I. Antoniou, P. Kielanowski, J. Math. Phys. 36, 2593 (1994).
[59℄ M. Gadella, R. de la Madrid, Int. Journ. Theor. Phys., 38, 93 (1999).
[60℄ Similar views are held by A. Mondragon, Private Dis ussions.
[61℄ E. P. Wigner, Group Theoreti al Con epts and Methods in Elementary Parti le Physi s,
F. Gursey [Ed.℄, Gordon and Brea h, New York, p. 37 (1964).
[62℄ E. P. Wigner, Symmetries and Re e tions, Indiana University Press, Bloomington,
p. 38 (1967).
[63℄ J. F. Cari~nena, M. Santander, J. Math. Phys. 22 (8), 1548 (1981).
[64℄ A. Bohm, Phys. Rev. A, 51, 1758 (1995); A. Bohm, S. Wi kramasekara, Found. Phys.,
27, 969 (1997).
[65℄ E. Nelson, Ann. Math. 70, 572 (1959).
[66℄ E. Nelson, W. F. Stinespring, Amer. Journ. Math. 81, 547 (1959).
[67℄ I. M. Gelfand, G. F. Shilov, Generalized Fun tions, Vol. I, A ademi Press, New York
(1967).
[68℄ J. Dixmier, Comp. Math. 13, 263 (1958).
[69℄ A. A. Kirillov, Dokl. Akad. Nauk. SSSR, 130, 996 (1960).
[70℄ A. Bohm, Journ. Math. Phys. 8, 1557 (1967), Appendix B.
[71℄ B. Nagel, Le ture Notes, College de Fran e (1970).
[72℄ R. G. Newton, S attering Theory of Waves and Parti les, M Graw-Hill, New York
(1966).
[73℄ J. R. Taylor, S attering theory, Jhon Wiley & Sons, In ., New York (1972).
[74℄ R. de la Madrid, Chaos, Fra tals & Solitons 12, 2689 (2001); quant-ph/0107096.
Bibliography 253

[75℄ W. Amrein, J. Jau h and K. Sinha, S attering Theory in Quantum Me hani s, Ben-
jamin, Reading, Massa husetts (1977).
[76℄ P. L. Duren, Theory of H p Spa es, A ademi Press, New York (1970).
[77℄ K. Homann, Bana h Spa es of Analyti Fun tions, Prenti e Hall, New Jersey (1962).
[78℄ P. Koosis, Introdu tion to H p Spa es, Cambridge, UK (1980).
[79℄ P. Koosis, The Logaritmi Integral, Cambridge, UK (1990).
[80℄ E. C. Tit hmarsh, Introdu tion to the Theory of Fourier Integrals, Clarendon Press,
Oxford, UK (1937).
[81℄ R. Paley and N. Wiener, Fourier Transforms in the Complex Domain, Ameri an Math-
emati al So iety, New York (1934).
[82℄ M. Reed and B. Simon, Methods of Modern Mathemati al Physi s: Fourier Analysis,
Self-Adjointness, A ademi Press, New York (1975).
[83℄ M. Gadella and A. Ordon~ez, Int. J. Theor. Phys. (1998).
[84℄ M. Gadella, J. Math. Phys. 24, 1462; 2142 (1983).
[85℄ J. E. Marsden, Basi Complex Analysis, Freeman, San Fran is o (1975).
[86℄ E. Hernandez, A. Jauregui, A. Mondragon, J. Phys. A: Math. Gen. 33, 4507 (2000).
[87℄ I. Antoniou, Z. Su hane ki, S. Tasaki, Series Representations of the Complex Delta
Fun tion, in Generalized fun tions, operator theory, and dynami al systems, I. Anto-
niou, G. Lumer [Eds.℄, Chapman & Hall/CRC, Bo a Raton (1999), p. 130.
[88℄ M. Gadella, Letters in Mathemati al Physi s 41, 279 (1997).
[89℄ A. Bohm, M. Gadella, G. B. Mainland, Am. J. Phys. 57, 1103 (1989).
[90℄ L. A. Khaln, Sov. Phys. JETP, vol. 6 (33), no. 6, 1053 (1958).
[91℄ A. Bohm, M. Gadella, S. Wi kramasekara, Some Little Things about Rigged Hilbert
Spa es and Quantum Me hani s and All That, in Genaralized Fun tions, Operator The-
ory, and Dynami al Systems, I. Antoniou, G. Lumer [Eds.℄, Chapman & Hall/CRC,
Bo a Raton (1999), p. 202.
[92℄ M. Born, E. Wolf, Prin iples of Opti s: ele tromagneti theory of propagation, inter-
feren e and dira tion of light, Pergamon Press (1980); J. Ja kson, Classi al Ele tro-
dynami s, se ond edition, John Wiley & sons, New York (1962).
[93℄ H. H. S haeer, Topologi al Ve tor Spa es, Springer Verlag, New York, Heidelberg,
Berlin (1970).
254 Bibliography

[94℄ S. Weinberg, The Quantum Theory of Fields, Cambridge University Press, Cambridge
(1995).